Tissue-specific Expression of Sialyltransferases”

Three sialyltransferases which attach terminal sialic acids to glycoprotein sugar chains are shown to exhibit striking differential expression in seven tissues of the rat. Using cDNA probes for the Gavl,4GlcNAc cy-2,6-sialytransferase which forms a NeuAccu2-6Galj31-4GlcNAc sequence on N-linked sugar chains, three different sized mRNAs are detected, two of which (4.7 and 4.3 kilobases (kb)) have high homology along the full length, and a third (3.6 kb, in kidney) which is missing the 5’ region corresponding to 45% of the NH2-terminal coding sequence. The 4.7- and 4.3-kb mRNAs exhibit differential expression of over 50-fold with the highest levels in liver and lowest in brain and heart. Assays for enzyme activity in tissue homogenates show high correspondence to the levels of mRNA. Evidence of tissue-specific expression was also obtained for two other sialyltransferases which form the NeuAca2-3Gal/31-4/3GlcNAc and NeuAca2-3GalB1-3GalNAc sequences on N-linked and 0-linked sugar chains, re-spectively. Comparison of the ratios of the three enzymes in several tissues suggests that they are expressed independently. The results are discussed for their relevance to cellular control of terminal glyco-sylation sequences on glycoproteins and glycolipids. Tissue and cell type-specific expression of terminal glyco-sylation sequences of glycoproteins

Tissue and cell type-specific expression of terminal glycosylation sequences of glycoproteins and glycolipids is well documented (1-3). Indeed, antibodies directed to specific carbohydrate structures detect cell surface antigens on some normal tissues but not others and on tumor tissues where the same carbohydrate antigens are not present on surrounding normal tissue (1, 2). Some of these carbohydrate structures are described as onco/fetal antigens because they are most abundant in early fetal development, and their expression is developmentally regulated (1-3).
Recent reports implicate tissue-specific and developmentally regulated carbohydrate structures as important mediators of cell-cell recognition (4-11) and differentiation (12)(13)(14). One of the best characterized stage-specific embryonic antigens (SSEA-1) is a fucosylated structure (GalBl-4(Fucal-3)GlcNAc-R) that appears to be required for normal compaction of the mouse embryo at the 16-cell stage (4,s). Similarly, * This work was supported by United States Public Health Service Grant GM-27904. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. expression of ganglioside G D~~ in the "undifferentiated" mesenchyme of the developing kidney appears to be required to signal the invasion and branching of the ureter bud into the mesenchyme (6).
The basis for the developmentally regulated and cell typespecific synthesis of the terminal carbohydrate sequences of glycoproteins and glycolipids is at present poorly understood. Because the carbohydrate structures of these glycoconjugates are largely determined by the specificities of the glycosyltransferases that synthesize them (E), it is presumed that differential expression of these enzymes is most likely the way in which this is regulated (1,(16)(17)(18). However, few reports have systematically investigated the degree to which specific glycosyltransferases are differentially expressed in normal tissues (19) or how such expression might be regulated in tissuespecific and developmentally regulated fashion.
Cloned cDNAs for several terminal glycosyltransferases have recently become available (20)(21)(22)(23) permitting molecular approaches to the analysis of glycosyltransferase expression. In this report, differential expression of a Gal@l,4GlcNAc a-2,64alyltransferase (23) in various rat tissues by Northern analysis and direct enzyme assays. Enzyme assays of two other glycoprotein sialytransferases also reveal differential expression in rat tissues and a lack of coordinate expression with the Gal@1,4GlcNAc a-2,6-sialyltransferase. The results support the concept that expression of glycosyltransferases is a predominant factor in regulating cell type-specific sugar structures.
Northern Analysis-Total mRNA was prepared from freshly dissected rat tissues as previously described (31). Gel electrophoresis of total RNA (15 pg/lane) was done in 1% agarose gels containing formaldehyde, and Northern hybridizations were performed as reported earlier (23). Radiolabeled cDNAs ((32) 1 X IO9 cpm/bg) were used as probes, prepared from agarose gel-purified fragments of the sialytransferase cDNA (32) subcloned into Bluescript (Stratagene). Blots were exposed to x-ray film for 2-5 days at -80 "C. Sizes of mRNAs were determined by co-electrophoresing total RNA samples mixed with commercial RNA standards (Bethesda Research Laboratories) and comparing the migration of standards with mRNAs detected with a cDNA probe.
Tissue-specific Expression of Sialyltransferases volumes (w/v) of 20 mM sodium cacodylate, pH 6.0,O.l M NaCI. The mucosal epithelium of the small intestine was separated from the intestinal wall by scraping gently with a razor blade and was homogenized immediately. Homogenates were either assayed immediately or aliquoted and stored frozen a t -80 "C and thawed immediately before use. Frozen homogenates were thawed only once before discarding. Alternatively, dissected tissues were frozen a t -80 'C and homogenized immediately before use. Control experiments showed that sialyltransferase activities were equivalent in fresh homogenates of either fresh or frozen tissue and in homogenates stored a t -80 "C and thawed once. However, substantial losses of activity were sometimes observed if homogenates were frozen and thawed several times. Protein contents of homogenates were determined using the Amido Schwarz assay (33).
Based on the strict acceptor substrate specificity of known sialyltransferases, the transfer of sialic acid to lacto-N-tetraose was taken to be solely from the Gal8l,3(4)GlcNAc a-2,3-sialyltransferase (15,35). Of the three acceptor substrates used, the Galj31,4GlcNAc a-2,6sialyltransferase transfers sialic acid only to asialo-a,-acid glycoprotein (35). However, the Gal81,3(4)GlcNAc a-2,3-sialyltransferase also used asialo-al-acid glycoprotein as an acceptor substrate (35). Accordingly, to determine activity of the Gal81,4GlcNAc a-2,6-sialyltransferase, transfer to asialo-al-acid glycoprotein was corrected by subtraction of a factor amounting to 0.4 times the observed transfer to lacto-N-tetraose, a factor determined experimentally using the purified Galpl,3(4)GlcNAc a-2,3-sialyltransferase. This correction typically amounted to less than 10% of the total transfer to asialoal-acid glycoprotein. The validity of this correction was confirmed by examining inhibition of transfer to asialo-a,-acid glycoprotein with affinity-purified antibodies to the Gal81,4GlcNAc a-2,6-sialytransferase, which do not inhibit the a-2,3-sialytransferase ((28) data not shown). Antifreeze glycoprotein is potentially an acceptor substrate of both the Gal@1,3GalNAc a2,3 sialyltransferase and the GalNAc a-2,6-sialyltransferase (15,34). However, direct assay of each homogenate for the latter enzyme with the specific acceptor substrate asialoovine submaxillary mucin (GalNAcaThr/Ser) indicated negligible activity in all cases (not shown). Analysis of the 8-eliminated products obtained for reactions with antifreeze glycoprotein and rat liver homogenate showed NeuAca2-3Gal~1-3GalNAc-ol as the only product (36): Thus, antifreeze glycoprotein was taken be a specific substrate for the Gal81,3GalNAc a-2,3-sialyltransferase in these experiments.
To evaluate the potential for inhibition of sialyltransferase activity by components of the crude homogenate, the activity of purified sialyltransferases was determined separately in duplicate assays containing 10 pl of various homogenates. The extent of the inhibition of the added enzyme was then deduced by comparison with control assays containing purified enzyme but no homogenate and homogenate but not purified enzyme. The results showed that inhibition of the rat Gal81,4GlcNAc a-2,6-sialyltransferase and the porcine Galj31,3GalNAc a-2,3-sialyltransferase was low, generally 0-20%. Accordingly, no corrections were applied. However, inhibition of the Gal81,3(4)GlcNAc a-2,3-sialyltransferase (rat) ranged from 30 to 90% in all cases. Thus, the level of inhibition was determined individually for each homogenate, and the level of activity reported in Tables I   L. Haber, E. Crooke, and J. Paulson, unpublished results. and I1 were adjusted to reflect the amount of activity that would be present without inhibition.

RESULTS
Characterization of Tissue-specific mRNAs of the GalB1, 4GlcNAc a-2,6-Sialyltransferase-Preliminary Northern analysis surveying mRNAs from different rat tissues showed that a cDNA probe corresponding to the coding sequence of the GalB1,4GlcNAc a-2,6-sialyltransferase (Gal a-2,6-ST) hybridized to mRNAs of different size (4.7,4.3, and 3.6 kb) in a tissue-specific manner. An abundant mRNA of 4.3 kb was detected only in liver: one of 4.7-4.8 kb was seen in spleen, lung, ovary, heart, kidney, and brain, and a predominant mRNA of 3.6 kb was found only in kidney (see Fig. 1). All three species are polyadenylated mRNAs evidenced by their purification on oligo-dT columns (not shown).
To examine the relationship of the different sized mRNAs to the cDNA of the enzyme originally cloned from a rat liver library (23), five identical Northern blots containing liver, spleen, and kidney mRNAs were hybridized to probes corresponding to different regions along the length of the sialyltransferase cDNA. The results, shown in Fig. 1, indicate that the 4.3-and 4.7-kb mRNAs are highly homologous along their entire length and likely represent alternative processing of the sialyltransferase mRNA in the various tissues. In contrast, the 3.6-kb mRNA showed no homology to two probes stemming from the 5' end of the liver sialyltransferase cDNA. These probes covered regions corresponding to the 5'-untranslated sequence and approximately 45% of the coding sequence of the enzyme (Fig. 1). The results indicate that the predominant kidney mRNA (3.6 kb) cannot code for the Gal a-2,6-ST. Preliminary evidence suggests that the 3.6-kb kidney mRNA appears to be derived from the same sialyltransferase gene as the other two messages.4 However, whether or not it codes for a functional protein remains to be determined. Tissue-specific Expression of the GalB1,4GlcNAc a-2,6-Sia-1yltransferase"From the results in Fig. 1, it is immediately apparent that liver and spleen contain very different levels of sialyltransferase mRNA. Comparison of the level of mRNAs liver, spleen, and kidney were probed with cDNAs representing different regions of the liver sialyltransferase mRNA. At the top is a diagram representing the fulllength cDNA of the Gal a2,6-ST (4.3 kb) cloned from rat liver cDNA libraries in Xgtll (23). The coding region of the enzyme is indicated in the cross-hntched box. The thick black lines beneath numbers 1-5 show the origin of the cDNA fragments used to make radiolabeled probes for the corresponding five Northern blots at the bottom. cDNA probes were derived from subcloned fragments of the Xgtll clones (23) as follows: I , 5"EcoRI fragment of ST3; 2, 5'PstI fragment of ST1; 3, 3'PstI fragment of ST1; 4, ST2; and 5, ST5.
3The size of the sialyltransferase mRNA in liver was previously ' E. Svensson, J. Weinstein, X. Wen, and J. Paulson, unpublished estimated to be approximately 4.7 kb (23).  2. Differential expression of the Gal a-2,6-ST in various rat tissues. Northern blot of total RNA from liver, brain, ovary, kidney, spleen, and lung probed with a radiolabeled cDNA probe corresponding to the 5' end of the sialyltransferase mRNA in liver (same as probe 1 used in Fig. 1

FIG. 3. Terminal sequences formed by the three glycoprotein sialyltransferases examined in this report.
from several additional tissues is shown in Fig. 2 using a probe from the 5' end of the sialyltransferase cDNA which detects only the 4.3-and 4.7-kb mRNAs containing the complete coding sequence. The blot shown in overexposed to detect minor amounts of message (compare Figs. 1 and 2). Again, it is apparent that these tissues differ dramatically in the amount of sialyltransferase mRNA detected with the lowest levels seen in brain, intermediate levels in ovary, kidney, spleen, and lung, and the highest levels in liver. In a separate experiment, sialyltransferase mRNA was found to be barely detectable in total RNA from heart (not shown). The relative levels of mRNA in liver, spleen, and brain are approximately 502.5:1, as determined from different exposure times of blots from several different experiments.

Comparison of the Levels of Activity of Three Siulyltransferases in Rat
Tissues-To directly examine the relationship between the levels of mRNA and enzyme activity for the tissues examined here, tissue homogenates were assayed for the Gal a-2,6-ST. At the same time, two other sialyltransferases were also examined, the Galpl,3(4)GlcNAc a-2,3-sialyltransferase (Gal a-2,3,-ST) and the GalB1,3GalNAc a-2,3sialyltransferase (Gal a-2,3(0)-ST). The sequences elaborated by these enzymes are given in Fig. 3.
The levels of sialyltransferase activities observed in homogenates of seven rat tissues are summarized in Table I. Specific activities for the Gal a,2,6-ST varied by over lOO-fold, with the highest activity seen in liver and the lowest (undetectable) in brain and heart. Qualitative comparison of the results in Table I with the Northern analysis in Fig. 2 shows a good correlation between the levels of Gal a-2,6-ST activity and the levels of the sialyltransferase mRNA in each tissue. This suggests that the primary reason for variation in levels of sialyltransferase activity is the regulation of the steady state levels of sialyltransferase mRNA.
Another striking finding from the results in Table I is that the three sialyltransferases examined are not coordinately expressed. This is demonstrated by comparison of the ratios of the activities of two enzymes. Indeed, the Gal a-2,3-ST and a Details of homogenate preparation and sialyltransferase assays are given under "Experimental Procedures." Results shown are the average of duplicate assays of one or more experiments. Duplicate assays were typically within 5% of the average, and activities in separate homogenates varied within 25% except at the limits of detection where the variation in background values becomes significant. Thus, while the lowest activities are not as precise, they accurately reflect the differential distribution of the sialyltransferases in different tissues. A value of 0 indicates no activity detected a t a level of 25 pmol/h/mg of protein. Gal a-2,6-ST activities in liver yield a ratio of 0.3, while in ovary the same ratio is 33, representing a differential expression of over 100-fold in these two tissues. Similar comparisons suggest that the three enzymes are expressed independently of each other.

Sialvltransferase activities in rezwns of the small intestine
Sialyltransferase Activities in the Mucosa and Wall of the Small Intestine-In an earlier report we noted lack of expression of the Gal a-2,6-ST and Gal a-2,3-ST in the small intestine (37) while others have described the presence of sialyltransferase activities in this tissue (38,39). To examine the possibility that differential expression of sialyltransferases might account for these discrepancies, sialyltransferase activities were assessed for homogenates of the mucosal surfaces of the duodenum, jejunum, and ileum separated from the intestinal wall by gentle scraping. In addition, a homogenate of the wall of the ileum was also examined. The Gal a-2,3-ST was either not detected or present at very low levels in all homogenates ((37) not shown). Activities observed for the other two sialyltransferases are summarized in Table 11.
While the Gal a-2,6-ST was undetectable in all of the mucosal homogenates, the Gal a-2,3(0)-ST showed a graded distribution, with highest levels in the duodenum, intermediate levels in the jejunum, and not detectable in the ileum. Perhaps the most striking finding is that neither the Gal a-2,6-ST nor the Gal a-2,3(0)-ST sialyltransferases were detectable in ileum mucosa but were found in relatively high levels in the remaining ileum wall. In this case the two enzymes are differentially expressed within two regions of the same tissue, each having separate functions and unique cell types. Together the results provide another example of tissuespecific and differential expression of sialyltransferases.

DISCUSSION
What is the consequence of the differential expression of terminal glycosyltransferases? Most terminal glycosyltransferases compete for common acceptor substrates (15). Thus, the type of terminal glycosyltransferases expressed by a cell should influence the types of terminal sequences found on the carbohydrate groups it produces.
To take a relevant case for the enzymes studied here, the Gal a-2,6-ST and the Gal a-2,3-ST both compete for the Galpl-4BlcNAc-R sequence on the terminal branches of Nlinked oligosaccharides of glycoproteins to form the NeuAca2-6Galpl-4GlcNAc-R and NeuAccu2-3Ga1(31-4GlcNAc-R sequences, respectively. In vitro, either enzyme can completely resialylate asialo-glycoproteins which contain di-, tri-, and tetra-branched oligosaccharides (35), although the Gal a-2,6-ST has been shown to have considerable branch specificity, sialylating some branches easier than others (41). Structure analysis shows that the N-linked oligosaccharides of some glycoproteins have exclusively the NeuAca2-6Gal linkage, others have the only NeuAccuZ-3Gal linkage, and still others have mixtures of both linkages on alternate branches of the same carbohydrate group (40,(42)(43)(44)(45). Taken at face value, these observations suggest that the type of sialic acid linkages produced results from the ratio of the two enzymes expressed by a cell. This principle is supported by recent results examining the carbohydrate structures of recombinant glycoproteins expressed in cultured mammalian cells. While the glycoprotein hormone erythropoietin isolated naturally contains about 40% of its sialylated sequences in the NeuAca2-6Gal linkage, the same glycoprotein expressed in Chinese hamster ovary (CHO) cells has exclusively the NeuAcaZ-3Gal linkage (43,44). In contrast, while the N-linked oligosaccharides of natural human interferon-81 and the recombinant interferon produced in CHO cells both have only the NeuAcaZ-3Gal linkage, the same glycoprotein produced in C127 cells has the NeuAca2-6Gal linkage as the only sialylated sequence and in addition has an alternative terminal sequence Gala1-3Galpl-4GlcNAc-R (40). Thus, it appears that CHO cells have only the Gal a-2,3-ST while the C127 cells have only the Gal a-2,6-ST. It is of interest in this regard that CHO cells are capable of producing N-linked oligosaccharides with the NeuAcaZ-6Gal linkage providing the enzyme is present, as demonstrated by the expression of the Gal a2,6ST in these c e k 5 In this report only three glycosyltransferases have been examined of the 100 or so thought to be required to form known carbohydrate structures, and these three form very common carbohydrate groups. Since many terminal carbohydrate sequences are formed by enzymes that act on both glycoprotein and glycolipid substrates (1, 15), the regulation of the expression of terminal glycosyltransferases will affect the structures of both classes of glycoconjugates. As additional glycosyltransferase cDNAs become available it will be of interest to establish the degree to which the expression of these enzymes is regulated in terminally differentiated and developing cells. This information will be required to understand the cellular control of recognition events mediated by carbohydrate groups in embryogenesis, development, and differentiation (4-14).
E. U. Lee and J. C . Paulson, unpublished results.