Synthesis of CMP-deaminoneuraminic acid (CMP-KDN) using the CTP:CMP-3-deoxynonulosonate cytidylyltransferase from rainbow trout testis. Identification and characterization of a CMP-KDN synthetase.

The sugar nucleotide, cytidine 5'-(3-deoxy-D-glycero-D-galacto-2-nonulosonic phosphate) (CMP-KDN) is expected to serve as a donor of KDN residues in the synthesis of KDN-containing glycoconjugates. We report here the identification and characterization of CMP-KDN synthetase, a novel enzyme responsible for synthesis of CMP-KDN from KDN and CTP. The enzyme was partially purified from the testis of rainbow trout (Oncorhynchus mykiss), where KDN gangliosides were first discovered (Yu, S., Kitajima, K., Inoue, S., and Inoue, Y. (1991) J. Biol. Chem. 266, 21929-21935), and used to synthesize CMP-[14C]KDN, which was characterized by 1H NMR. Vmax/Km studies showed that KDN was a preferred nonulosonic acid substrate compared to N-acetylneuraminic acid (Neu5Ac) or N-glycolylneuraminic acid (Neu5Gc) (4.4 x 10(-3) min-1 for KDN versus 2.3 and 1.8 x 10(-3) min-1 for Neu5Ac and Neu5Gc, respectively). CMP-KDN synthetase activity was maximal at pH 9-10 and at 25 degrees C. The presence of either Mg2+ or Mn2+ was essential for CMP-KDN synthetase activity. 25 mM Mg2+ stimulated formation of CMP-KDN more than 10-fold, yet only stimulated formation of CMP-Neu5Ac and CMP-Neu5Gc 4-fold, relative to 1 mM Mg2+. A kinetic study using mixed substrates showed that both CMP-KDN and CMP-Neu5Ac synthetase activities in the partially purified enzyme were due to the same active site of a single enzyme. In contrast, Neu5Ac and Neu5Gc were the preferred nonulosonic acid substrates for the calf brain CMP-sialic acid synthetase. Thus, mammalian CMP-sialic acid synthetases recognizes similar, yet distinctively different, substrate specificity determinants. Thus, the trout testis enzyme was considered to synthesize activated sugar nucleotides required for synthesis of both (KDN)GM3 and (Neu5Ac)GM3. The expression of CMP-KDN synthetase was shown to be temporally correlated with development and to parallel the developmental expression of (KDN)GM3 in sperm.

temporally correlated with development and to parallel the developmental expression of (KDN)GM3 in sperm.
KDN' (3-deoxy-D-g~ycero-D-gu~ucto-2-nonulosonic acid), a unique deaminated analogue of sialic acid was first reported as a naturally occurring sialic acid in fish egg polysialoglycoproteins in 1986 (1). Subsequently, an increasing number of KDN-containing glycoconjugates have been reported (2)(3)(4)(5)(6)(7)(8)(9)(10). Possible physiological roles of these KDN glycoconjugates have also been reported recently (6,7,11). The occurrence and structures of KDN-containing glycoconjugates are summarized in Table I. A logical step in the investigation of KDN glycoconjugates is to determine how the KDN residues are incorporated into KDN-glycan chains. Presumably, the penultimate step in synthesis of KDN glycoconjugates would be synthesis of the activated KDN nucleotide, CMP-KDN, catalyzed by CTP:CMP-KDN cytidylyltransferase (CMP-KDN synthetase). A related enzyme, CTP:CMP-sialic acid cytidylyltransferase (CMP-sialic acid synthetase) has been identified in cells and tissues of animals (12)(13)(14)(15) and bacteria (16)(17)(18)(19). Since CMP-KDN synthetase has not been previously reported, we cannot rule out u priori the possibility that Neu5Ac or Neu5Gc residues are deacylated and deaminated at the level of the sugar nucleotide or after incorporation into polymeric products. To investigate these unresolved questions, studies were initiated to determine: 1) if CMP-KDN synthetase activity exists in rainbow trout testis, a tissue known to synthesize KDN-containing glycoconjugates; 2) to partially purify and characterize such an enzyme activity, if found; and 3) to use the enzyme to synthesize CMP- [14C] KDN, thereby providing a key substrate for future biosynthetic studies to determine how expression of KDN glycoconjugates are regulated.

Pleurodeles waltlii
Fucal+3(Gal~l+4)GlcNAc@1~3(KDN~2--+6)GalNAc (Le") Fish sperm KDN ganglioside Amphibian egg jelly coat: Fuc~l+2Gal@1-+4(Fuc~ul-+3)GlcNAcp1-t3(KDNn2"tG)GalNAc (LeY) thetase was used to synthesize CMP-KDN (12,20). Although CMP-Neu5Ac and CMP-Neu5Gc were efficiently synthesized by this enzyme, KDN was a very poor substrate and the amount of CMP-KDN formed was far less than CMP-sialic acid (20-22). In the present study, we have shown that rainbow trout testes contain CMP-KDN synthetase activity and have determined at what stage during spermatogenesis the synthetase is expressed. Because the trout testis enzyme prefers KDN, it should be referred to as a CMP-KDN synthetase, rather than CMP-sialic acid synthetase, to distinguish it from the more generic enzyme expressed in calf brain. CMP-KDN synthetase was also detected in rainbow trout oocytes which contained KDN in the egg cortical glycoproteins (23). Evidence in favor of a single enzyme rather than two different enzymes which are responsible for the synthesis of CMP-KDN and CMP-Neu5Ac was provided by the mixed substrate method in a kinetic study of their formation. Thus, the trout testis enzyme appears to have broad substrate specificity and can catalyze the synthesis of several CMP-nonulosonic acids.

Materials
Rainbow trout testes were obtained from 1-year-old fish collected monthly between May and November. The fish were provided through the courtesy of Gunma Prefectural Fisheries Experimental Station at Kawaba and stored at -80 "C until use. Mature sperm obtained in November were centrifuged at 5,000 rpm for 10 min to remove the seminal fluid and kept separately at -80 "C until use. Calf brain was obtained from a 4-month-old calf through Tokyo Shibaura Zouki. [U-14C]Neu5Ac (10 GBq/mmol), ~-[ u -'~C ] M a n (10 GBq/mmol), and sodium [I-"Clpyruvate (1.18 GBq/mmol) were purchased from Amersham (United Kingdom). Neu5Ac was obtained from Nacalai (Kyoto), and Neu5Gc was isolated and purified from rainbow trout polysialoglycoprotein, as previously described (24,25). CMP-Neu5Ac was purchased from Wako Chemicals (Osaka), and CMP-Neu5Gc was kindly provided from Mect.  was obtained from Seikagaku Kogyo, Co. (Tokyo). KDN was isolated from rainbow trout ovarian fluid KDN-rich glycoprotein (25) or enzymatically synthesized using the Escherichia coli acylneuraminate pyruvate lyase (Wako Chemicals, Osaka) according to the procedure of Aug6 and Gautheron (26). Calf intestine alkaline phosphatase was obtained from Boehringer Mannheim.
Synthesis and Purification of f'CINeu5Gc Neu5Gc was first converted by acylneuraminate pyruvate lyase to N-glycolylmannosamine and pyruvate, according to the method of Comb and Roseman (27). [14C]Neu5Gc was then synthesized from ManNGc and [14C]pyruvate by an aldol condensation, catalyzed by the same enzyme.
A 1-ml reaction mixture containing 35 pmol of NeuBGc, 1 unit of acylneuraminate pyruvate lyase, and 50 mM potassium phosphate buffer (pH 7.2) was incubated at 37 "C for 24 h. The reaction mixture was then applied to a DEAE-Sephadex A-25 column (HCO, form, 0.9 X 10 cm). The column was eluted with water, and the flowthrough fractions containing ManNGc were pooled. Part of the eluate was spotted on a TLC plate (Kieselgel 60, Merck) and developed in 95% ethanol/l M ammonium acetate (7:3, v/v) for 5 h. A single spot, identified as ManNGc appeared after spraying with 10% H,S04ethanol (28) and heating at 120 "C. The remaining eluate was lyophilized and stored at -20 "C.

CMP-Nonulosonate Synthetase Assay
Reaction mixtures contained 4 pl of the enzyme fraction in 20 pl of 100 mM Tris-acetic acid buffer (pH 9.0) with 0.18 mM [I4C]KDN (400 Bq) or ['4C]Neu5Ac or ["C]Neu5Gc, 9.6 mM 5'-CTP, and 25 mM Mg(0Ac)B. After incubation for 3 h at 25 "C, the reaction was stopped by the addition of 2 volumes (40 p l ) of cold ethanol. After standing for 30 min in an ice bath, the mixture was centrifuged at 5,000 rpm for 10 min. A 10-pl aliquot of the supernatant was applied on a cellulose sheet (20 x 20 cm) with fluorescent indicator (Kodak), which was developed in 95% ethanol/l M ammonium acetate (7:3, v/ v) for 2 h. After air drying, the CMP-[14C]KDN formed was detected by a Bio-imaging analyzer (Fujix BAS 2000). One unit of the enzyme was defined as a quantity of the enzyme required to synthesize 0.01 pmol of CMP-nonulosonate/l h at 25 "C.

Preparation of Calf Brain CMP-Neu5Ac Synthetase
The CMP-sialic acid synthetase from calf brain was prepared by the procedures described by van den Eijnden and van Dijk (12) and Higa and Paulson (20).

Preparation of Rainbow Trout Testis CMP-Nonulosonate Synthetase
The procedures of van den Eijnden and van Dijk (12) and Higa and Paulson (20) were followed in extraction of the CMP-nonulosonate synthetase from rainbow trout testis. Testis (44 g) from rainbow trout collected on August 19, 1991 (average 36 g/fish) were sliced after removal of connective tissue and blood-vessels and homogenized in 90 ml of 0.01 M sodium pyrophosphate (pH 10.2) containing 9 mg of soybean trypsin inhibitor. The homogenate was centrifuged at 15,000 rpm for 20 min and the resulting supernatant (CS, 94 ml) was ultracentrifuged at 100,000 g for 1 h, resulting in 81 ml of the supernatant (UCS). Saturated ammonium sulfate was gradually added to the supernatant UCS (up to 30%), and the mixture was gently stirred at 4 "C for 15 min and then left at 4 "C for 1 h. Precipitated proteins were removed by centrifugation at 12,000 rpm for 20 min, and 148 ml of the supernatant fraction were obtained and labeled 30% (NH4),S04-sup. Saturated ammonium sulfate was again added with stirring to a final concentration of 60%. The mixture was stirred for 15 min and left overnight at 4 "C. Precipitated proteins were collected after centrifugation at 12,000 rpm for 20 min, dissolved in 35 ml of 5 mM Tris-acetate buffer (pH 7.5) containing 0.05% 2mercaptoethanol, and dialyzed against 5 mM Tris-HC1 buffer (pH 7.6) containing 0.1% 2-mercaptoethanol and 1 mM MgCI, for 4-12 h. After dialysis, 40 ml of the 60% (NH1)2S04-ppt fraction were obtained. This enzyme fraction was used in experiments to characterize the CMP-nonulosonate synthetase activity (Figs. 2 and 3, and Tables 11, IV, V, and VI) and in preparation of CMP-nonulosonates.

Purification of Rainbow Trout Testis CMP-Nonulosonate Synthetase
The 60% (NH&S04-ppt fraction (1.5 ml) was dialyzed for 20 h against 100 mM Tris-acetate buffer (pH 7.5) containing 0.1% 2mercaptoethanol and 1 M glycerol. The dialysate was then applied to a Sephacryl S-300 column (1.8 X 66 cm) equilibrated and eluted with 100 mM Tris-acetate buffer (pH 7.5) containing 0.1% 2-mercaptoethanol and 1 M glycerol. Two-ml fractions were collected, and the elution pattern was monitored by (a) measurement of light absorption at 280 nm and (b) measurement of the enzyme activity to form CMP-KDN, CMP-NeuBAc, and CMP-Neu5Gc. The enzyme-containing fractions were pooled, concentrated to 5.7 ml in an Amicon ultrafilter using a YM-IO membrane, and were then rechromatographed on the same column.

Protein Determination
Protein was quantitated by use of a Bio-Rad Protein Assay Kit (Bio-Rad) with ovalbumin as the standard.

Molecular Weight Determination
The molecular weight of rainbow trout testis CMP-nonulosonate synthetase was estimated by Sephacryl S-300 gel chromatography. The column was first calibrated with the following molecular markers: ferritin (Mr, 440,000), bovine serum albumin (Mr, 67,000), and ovalbumin (Mr, 45,000). The molecular weight of CMP-nonulosonate synthetase was calculated by plotting the K., uersus the log of the molecular weight standards.
Enzymatic Synthesis of CMP-KDN and CMP-NeuSGc Preparation of CMP-KDN and CMP-Neu5Gc-Synthesis of CMPnonulosonates (KDN and Neu5Gc) from CTP and the corresponding nonulosonic acids was carried out at 25 "C in a 2.2-ml incubation mixture containing 100 mM Tris-acetate buffer, (pH 9.0), 4.5 mM KDN or Neu5Gc, 11 mM 5'-CTP, 25 mM Mg(OAc),, and 440 p1 of enzyme (60% saturated (NH4)*SO4 precipitate of rainbow trout testis homogenate). After 24 h of incubation, the reaction was stopped as described above by the addition of 4.4 ml of cold ethanol. After 30 min of incubation in an ice bath, the mixture was centrifuged at 5,000 rpm for 10 min, and the supernatant evaporated under reduced pressure to 200 pl.
Treatment of CMP-KDN and CMP-NeuSGc with Calf Intestine Alkaline Phosphatase-To 50 p1 of the concentrated solution containing CMP-KDN or CMP-NeuBGc, 50 pl of 50 mM ammonium bicarbonate containing 56 units of calf intestine alkaline phosphatase was added, and the resulting solution incubated at 25 "C for 10 h.
Isolation and Purification of CMP-KDN and CMP-NeuSGc-The alkaline phosphatase-treated reaction products, CMP-KDN or CMP-Neu5Gc, were centrifuged at 5,000 rpm for 5 min. The supernatant was applied to Whatman No. 3MM paper (46 X 57 cm), and the chromatogram was developed with 95% ethanol/l M ammonium acetate (7:3, v/v) for 8.5 h. After air-drying, CMP-KDN and CMP-Neu5Gc were located under UV light, and eluted from the chromatogram with 5% ethanol containing 1 mM NH,OH. The sample was then applied to a Sephadex G-25 column (1.25 X 72 cm) and eluted with 1 mM NH40H at 4 "C (29). The elution pattern was monitored by absorption at 271 nm and by the thiobarbituric acid method (30, 31) without prior hydrolysis for CMP-KDN, and with prior hydrolysis for CMP-Neu5Gc. The purified samples of CMP-KDN and CMP-Neu5Gc were lyophilized and stored at -20 "C.

Analysis of CMP-Nonulosonates by High Performance Liquid
Chromatography (HPLC) HPLC was carried out on an Irika HPLC system using a DC 613 cation-exchange column (6 X 150 mm; Showa Denko, Tokyo), which was eluted with 20 mM sodium phosphate buffer (pH 7.5)/acetonitrile (1:2.4, v/v) (25). The elution pattern was monitored by measurement of light absorption at 271 nm. CMP-KDN, synthesized by CMP-KDN synthetase as described above, and authentic samples of CMP-Neu5Ac and CMP-Neu5Gc were dissolved in water (3-15 nmol of CMP-nonulosonates/40 pl) and analyzed by HPLC.
Determination of the Thermal Stability of CMP-KDN by HPLC The temperature lability of CMP-KDN was determined by monitoring its decomposition by HPLC. An aqueous solution of CMP-KDN (2 nmol/20 pl, pH 7) was incubated at 43 and 80 "C for 10 min before HPLC analysis, as described above.

Determination of the p H Lability of CMP-KDN and CMP-Neu5Ac
Three to 6 nmol of CMP-KDN and CMP-Neu5Ac were incubated at pH 5.0 (in 90 pl of 67 mM sodium acetate buffer), pH 6.5 and 8.0 (in 180 p1 of 13 mM sodium phosphate buffer) at 37 "C. The time course of decomposition was followed by HPLC. At various times (0, 30,60, and 240 rnin), 2O-pl aliquots of samples were transferred to an ice bath and analyzed by HPLC.
Expression of CMP-Nonulosonate Synthetase Activity of Rainbow Trout Testis during Spermatogenesis Rainbow trout testes were obtained by dissecting 1-year-old fish monthly from May through November at the Gunma Prefectural Fisheries Experimental Station at Kawaba, Japan. The testes were to August were weighed and examined for the CMP-KDN and CMP-stored at -80 "C until use. Testes obtained during the period of May Neu5Ac activities as described above. Because there were mature spermatozoa (semen or sperm) in the sperm ducts of the testes obtained in September, October, and November, these were separately examined for the enzyme activity (mature sperm and immature testis had been separated immediately after dissection). Semen or sperm was centrifuged at 5,000 rpm for 5 min at 4 "C to separate plasma and spermatozoa. For mature spermatozoa the GMP-KDN and CMP-Neu5Ac synthetase activities were determined by the same procedures as described above for testis. Seminal plasma was also assayed for CMP-KDN synthetase activity.

Partial Purification of CMP-Nonulosonate Synthetase from Rainbow Trout Testes
As summarized in Table 11, CMP-nonulosonate synthetase was purified 116-fold with a 64% yield from the homogenate of trout testis. A nucleotide phosphatase activity present in the crude homogenate, and which hydrolyzed CTP, was mostly eliminated by precipitation with 30% saturated (NH4)$04. The apparent molecular weight of CMP-nonulosonate synthetase was approximated at 260,000, as judged by gel chromatography on Sephacryl S-300 (Fig. 1). SDS-polyacrylamide gel electrophoresis of the partially purified enzyme revealed several Coomassie Blue-staining bands (results not shown). Further attempts at purification of this enzyme by anion-exchange chromatography on DEAE-Sephadex resulted in nearly complete loss of enzyme activity irrespective of the presence or absence of glycerol.

Effect of Enzyme Concentration on the Formation of CMP-Nonulosonates
The effect of enzyme concentration on the rate of conversion of CTP and nonulosonic acids to CMP-nonulosonates was studied and formation of all three CMP-nonulosonates was found to increase linearly with the amount of CMP-nonulosonate synthetase in 20-pl incubation mixtures up to 0.032 units (data not shown).

Effects of Temperature and Incubation Time on the Formation of CMP-Nonulosonates
The time course for the CMP-nonulosonates at 15,25, and 37 "C was followed, and the data suggested that all three sialic acids were converted to their CMP derivatives (not shown). This supposition was confirmed by showing directly that not only KDN, but also Neu5Ac and Neu5Gc, were converted to CMP-KDN, CMP-Neu5Ac, and CMP-NeuSGc, respectively (Fig. 2). I t is also evident from the data that under identical conditions the formation of CMP-[14C]KDN was significantly greater than either CMP-[14C]Neu5Ac or CMP-[14C]Neu5Gc. In contrast, the calf brain CMP-Neu5Ac synthetase was highly active on Neu5Ac and Neu5Gc but was only slightly active on KDN (Table 111). Thus, the rainbow trout testis enzyme can be designated a CMP-KDN synthetase, while the calf brain enzyme should be referred to as a CMP-sialate synthetase. The optimum temperature for the rainbow trout testis CMP-KDN synthetase activity was approximately 25 "C. Rainbow trout is inhabitable at 15 "C or slightly below 15 "C, a temperature at which the CMP-nonulosonates were efficiently synthesized. This is in contrast to the Escherichia coli K1 CMP-sialic acid synthetase, which is a low temperature-sensitive enzyme and as such, shows little activity at 15 "C (32). At 37 "C the initial rates of formation of CMPnonulosonates were highest, but incubation at 37 "C for more than 3 h resulted in decomposition of CMP-KDN.

The Effects of Divalent Cations on Formation of CMP-Nonulosonates
The partially purified CMP-nonulosonate synthetase of rainbow trout testis required divalent cation for activity. Table IV summarizes the effect of divalent cations on the enzyme activity. The presence of Mg2+ or Mn2+ is essential for enzyme activity. When the enzyme activity a t 25 mM Mg2+ was compared with that at 1 mM Mg2+, the formation of CMP-KDN was stimulated more than 10-fold, while that of CMP-Neu5Ac and CMP-Neu5Gc was only 4-fold. Optimum

TABLE IV
Effect of divalent cations on the activity of rainbow trout testis CMP-nonulosonate synthetase using KDN, NeUSAc, NeUSGc, and CTP Incubation mixtures contained 100 mM Tris-acetate buffer (pH 8.0 or 9.0) and the concentration of divalent cation indicated. Incubation was at 25 "C for 3 h. Yields of CMP-nonulosonates formed are expressed in % relative to the amount of the respective nonulosonic acids used. The effect of pH upon formation of CMP-nonulosonates by the rainbow trout testis CMP-KDN synthetase was examined. With three different nonulosonates as substrates, the enzyme showed maximal activity between pH 9.0 and 10.0 in the presence of 25 mM Mg2+. The optimum pH was shifted to Since the CMP-nonulosonate synthetase activity appeared pH 8 in the presence of 2 or 5 mM Mn2+ (data not shown).
to show a relatively broad substrate specificity (Fig. 2), we

Kinetic Studies of the Rainbow Trout Testis CMP-Nonulosonate Synthetase
carried out a detailed kinetic analysis, comparing the relative activity of KDN, Neu5Ac, and Neu5Gc. The Michaelis constants of these nonulosonates at a fixed saturated concentration of CTP (9.6 mM) were determined. The apparent Michaelis constants were estimated by Lineweaver-Burk plots (data not shown). To compare the activities among the three sialic acid substrates, the V,,,/K, values were determined and are summarized in Table V. The value for KDN (4.4 x min") is nearly two times greater than that of Neu5Ac (2.3 X min") and Neu5Gc (1.8 x min"). This leads us to conclude that the partially purified CMP-KDN synthetase prefers KDN as its nonulosonate substrate and that this enzyme can effectively provide CMP-KDN for synthesis of KDN-containing glycoconjugates, such as (KDN)GM3 in trout testis. The Michaelis constants of CTP at a fixed concentration of KDN, Neu5Ac, and Neu5Gc (4.5 mM) were also determined. K , values of 2.6,2.7, and 1.7 mM and V, , , values are 0.014, 0.013, and 0.0097 mM/min were obtained, respectively.
In order to determine whether the observed two enzyme activities, i.e. CMP-KDN and CMP-Neu5Acyl synthetase activities, are due to the same active center of a single enzyme or due to two different enzymes, we used the mixed substrate method (33). It is shown that if two enzymatic activities are due to a single enzyme protein, the rate ut at which it will act on a mixture of the two substrates, KDN and Neu5Ac, is given by the following equation (33). Alternatively, if the observed two enzyme activities are due to the individual two enzyme proteins, the total velocity ut must be the simple sum of the reaction velocities of the individual enzymes at the respective substrate concentrations. A comparison between the experimental ut and the theoretical ut values expected for the two possibilities listed in Table  VI allowed us to conclude that the observed results are consistent with the view of a single enzyme being involved in the synthesis of CMP-KDN and CMP-Neu5Acyl.

TABLE VI Enzyme kinetic data discriminating between a single enzyme or two different enzymes which are responsible for the synthesis of CMP-KDN and CMP-Neu5Ac by the partially purified CMP-nonulosonate synthetase preparation obtained from rainbow trout testis
The theoretical values estimated from the reaction by a single enzyme were obtained using the Equation 1 in the text. Those estimated from the reactions catalyzed b y two different enzymes were obtained merely by the sum of the reaction velocities for two individual enzymes at each substrate concentration. The values of K,,, and VmaX in Table V

Synthesis and Characterization of CMP-Nonulosonates
Characterization of CMP-Neu5Ac and CMP-Neu5Gc, synthesized by rainbow trout CMP-KDN synthetase, was carried out by comparing their chromatographic mobilities on thin layer plates to authentic samples of CMP-Neu5Ac and CMP-Neu5Gc (Fig. 2). Similarly, the chromatographic mobility of the product of the synthetase with ["CIKDN was that expected for CMP-['4C]KDN. This was confirmed by 'H NMR, as described below.

Synthesis and Purification of CMP-KDN and CMP-Neu5Gc
Alkaline phosphatase was used to decompose the excess amounts of CTP, CDP, and CMP present in incubation mixtures containing CMP-KDN. The resulting products were first purified by paper chromatography on Whatman 3MM before gel filtration on Sephadex G-25 as described under "Experimental Procedures." Approximately 0.5 mg of a purified CMP-KDN was obtained. About 0.2 mg of CMP-Neu5Gc was also obtained by the methods similar to those described for CMP-KDN.

400-MHz Proton NMR Spectroscopy
The 'H NMR spectra of CMP-KDN and CMP-Neu5Gc are shown in Fig. 3. In each case, the signals for the H-3,, and H-3,, protons and the observed spin-spin coupling between the phosphorus and H-3,,, when compared to the reported values for CMP-Neu5Ac and the p epimers for KDN and Neu5Gc, clearly showed the characteristic resonances expected for the @-glycosidic configuration. The remaining chemical shifts for CMP-KDN and CMP-Neu5Gc were also essentially identical to that reported for each sugar nucleotide (20, 21), thus confirming the structure of the enzymatically synthesized products as CMP-KDN and CMP-NeuSGc.

HPLC Analyses of CMP-Nonulosonates
As shown in Table VII, CMP-KDN, CMP-Neu5Ac, and CMP-Neu5Gc were readily separated by HPLC, using a cation-exchange column. The elution times (CMP-KDN > CMP-Neu5Gc > CMP-Neu5Ac) provided sufficient resolution to carry out definitive lability studies, as described below. and 100% of CMP-KDN were degraded after 10 min of incubation at 43 and 80 "C (pH 7), respectively. Thus, in accord with our earlier findings (23), CMP-KDN appears to be more labile than CMP-Neu5Ac.
Effect of pH on the Stability of CMP-KDN and CMP-Neu5Ac-The acid lability of CMP-KDN was compared with that of CMP-Neu5Ac by incubating these CMP-nonulosonates at pH 5.0,6.5, and 8.0. The CMP-nonulosonates remaining after incubation were determined by HPLC, and the results are shown in Table VIII. At pH 5.0, CMP-KDN was considerably more labile than CMP-N~U~AC, since 74% was hydrolyzed in 1 h, compared to only 40% hydrolysis of CMP-Neu5Ac. In contrast, the rate of hydrolysis of the two sugar nucleotides was comparably slow at pH 6.5 and 8.0. In fact, CMP-KDN appeared to be relatively more stable at p H 8.0 than CMP-Neu5Ac (11 uers'sus 22% hydrolysis, respectively, after 4 h).

Developmental Changes of CMP-Nonulosonate Synthetase Activity
The developmental expression of CMP-KDN and CMP-Neu5Ac synthetase activities was determined in freshly isolated testis at different stages of spermatogenesis. Rainbow trout testes were dissected from 1-year-old fish, harvested at about 30-day intervals between May and November, 1991, as described under "Experimental Procedures." The level of enzyme activity in the testis per single fish rose rapidly with maturation of testis, assessed by an increase in weight, until the middle of August (Fig. 4A and 23). The developmental expression of CMP-KDN synthetase activity paralleled the expression of CMP-Neu5Ac activity (Fig. 4B). We do not know if this is because a single protein catalyzes synthesis of both sugar nucleotides or, alternatively, because distinct enzymes are coexpressed. It has not been possible to purify to homogeneity the enzyme activities. To determine the fate of CMP-nonulosonate synthetase, the enzyme activity in the mature sperm (spermatozoa) after spermiation was measured and found to be undetectable (data not shown). No activity was detected in the seminal plasma, suggesting that the regulated expression of the CMP-nonulosonate synthetase activity may be restricted to a relatively narrow stage of development.

Stability and Storage of CMP-Nonulosonate Synthetase Activity
The soluble CMP-nonulosonate synthetase activity found in the 60% (NH4)&304-ppt fraction remained stable for at a The 7% recovery of CMP-KDN and CMP-Neu5Ac is shown after incubation at the three different pH values for 30 or 60 min at 37 "C. Recovery was based on determining the amount of sugar nucleotide remaining after HPLC analysis, as described under "Experimental Procedures." least 2 weeks when stored at 4 "C. This fraction also did not lose activity on dialysis against 5 mM Tris-HC1 (pH 7.6) containing 0.1% 2-mercaptoethanol and 1 mM MgClz followed by storage at 4 "C for 2 weeks. 2-Mercaptoethanol was found not to affect the activity of the 60% (NH4)2SO4-ppt fraction.
The same enzyme preparation was completely stable at -20 "C, but about 50% of the activity was lost when incubated at 37 "C for 20 min. When the CS supernatant, obtained after centrifugation at 15,000 rpm for 20 min of the trout testis homogenate (Table 11) was lyophilized, the CMP-nonulosonate synthetase was stable for at least 6 months at -20 "C without any loss of activity.

DISCUSSION
Recent finding of the unique nonulosonic acid residue, KDN, in glycoconjugates (1-10, see Table I) has raised several interesting questions regarding the biosynthetic pathway and how KDN may be activated and transferred to pre-existing endogenous acceptors. Although KDN is not commonly reported as a constituent of most known glycoconjugates, it is likely that KDN-containing glycoconjugates will turn out to be of more widespread occurrence in other animal tissues and cells when more sensitive and specific methods for its detection become readily available. We have therefore considered it particularly important to determine how these new classes of glycoconjugates are synthesized. The first question we addressed was if KDN was activated with CTP to form CMP-KDN, prior to being transferred to endogenous oligosaccharide acceptors. Nothing is known about the mechanism of KDN activation or formation of KDN-glycan units. An alternative pathway would be that Neu5Ac or Neu5Gc residues were first transferred from CMP-Neu5Ac or CMP-Neu5Gc into glycan chains, then deacylated, and deaminated. These two alternative hypotheses could be tested because the first would predict the existence of a CMP-nonulosonate synthetase with a preference for KDN, while the absence of such an activating enzyme would support the latter hypothesis. The results of our studies show that rainbow trout testis contains a CMP-KDN synthetase that catalyzes the formation of CMP-KDN from CTP and KDN, as shown.

Me
CTP + KDN 4 CMP-KDN + PPi REACTION 3 On the basis of these results, we hypothesize that the subsequent reaction leading to formation of KDN glycans will be The CMP-KDN synthetase was purified 116-fold from trout testis and found to catalyze the preferential activation of KDN, although Neu5Ac and Neu5Gc could also be activated. CMP-KDN synthetase showed optimal activity between pH 9-10, and at 25 "C. The presence of either Mg2+ or Mn2+ was essential for CMP-KDN synthetase activity. The enzyme activity with KDN as substrate was stimulated 10fold by 25 mM M$' relative to the activity a t 1 mM M e , whereas activity with Neu5Ac or Neu5Gc was stimulated only 4-fold. Although we have been unable to purify the CMP-KDN synthetase to homogeneity, we determined if the different nonulosonate activating activities are catalyzes by the same or different enzymes by using the mixed substrate method (33). It should be remarked that, as shown in Table  VI, the agreement of the observed vr and those expected for "one and the same enzyme" provides strong evidence in favor of the view that a single enzyme protein is involved in the synthesis of CMP-KDN, CMP-Neu5Ac, and CMP-Neu5Gc. To obtain more detailed protein chemical data on the CMP-KDN synthetase, we have to await further purification of the enzyme. This point does not detract, however, from the importance of our present major finding of a CMP-KDN synthetase activity that has not been previously described. Also of major importance is our finding that the development,al expression of CMP-KDN activity is temporally correlated with development during spermatogenesis and with the developmental expression of (KDN)G"s. From Fig. 4, it can be seen that the level of enzyme activity exhibits characteristic developmental changes, increasingly rapidly from June and reaching a maximum level in mid August. Thereafter, spermiation begins (34) and this is correlated with a dramatic decrease in enzyme activity that then tapers off slowly until November. No CMP-KDN synthetase activity was found in spermiated mature sperm.
Although we have yet to ascertain where the synthetase is localized within spermatozoa, we presume that like CMP-Neu5Ac, CMP-KDN is also synthesized in the cell nucleus, diffuses into the cytosol (15), and is imported into the Golgi to be available for the KDN transferases (35). These aspects of how KDN glycoconjugates are synthesized await further studies that are now possible with the availability of CMP-[I4CJKDN. Since this sugar nucleotide is more labile than CMP-[14C]Neu5Ac, added precautions will have to be taken in carrying out biosynthetic studies.