Phosphorylation and O-glycosylation sites of human chromogranin A (CGA79-439) from urine of patients with carcinoid tumors.

Because of their water-soluble properties, chromogranins (CGs) and chromogranin-derived fragments are released together with catecholamines from adrenal chromaffin cells during stress situations and can be detected in the blood by radiochemical and enzyme assays. It is well known that chromogranins can serve as immunocytochemical markers for neuroendocrine tissues and as a diagnostic tool for neuroendocrine tumors. In 1993, large CGA-derived fragments have been shown to be excreted into the urine in patients with carcinoid tumors and the present study deals with the characterization of the post-translational modifications (phosphorylation and O-glycosylation) located along the largest natural CGA-derived fragment CGA79-439. Using mild proteolysis of peptidic material, high performance liquid chromatography, sequencing, and mass spectrometry analysis, six post-translational modifications were detected along the C-terminal CGA-derived fragment CGA79-439. Three O-linked glycosylation sites were located in the core of the protein on Thr163, Thr165, and Thr233, consisting in di-, tri-, and tetrasaccharides. Three phosphorylation sites were located in the middle and C-terminal domain, on serine residues Ser200, Ser252, and Ser315. These modified sites were compared with sequences of others species and discussed in relation with the post-translational modifications that we have reported previously for bovine CGA.

Chromogranins/secretogranins (CGs/Sgs) 1 constitute a family of acidic secretory glycoproteins widely expressed in a large number of endocrine and neuroendocrine cells and in neurons (1)(2)(3)(4). Chromogranin A (CGA), the major member (40% of total soluble granule proteins) of this family, has been studied extensively. At the subcellular level, chromogranins are exclusively found in the soluble core of hormone and neurotransmit-ter storage vesicles and are released during exocytosis. Chromogranins have been proposed to play multiple roles in the secretory process. An intracellular function as a "helper" protein in the packaging of peptides, hormones, and neuropeptides by virtue of their ability to aggregate in the low pH and high calcium environment of the trans-Golgi network and as modulators of the processing of these components has been suggested (3). Extracellularly, different members of the chromogranin family are now considered as precursor proteins, which are actively processed into peptides within the secretory granules (see Refs. 1 and 5 for reviews). Previously, we reported a detailed study of the intracellular and extracellular processing of CGA and CGB/SgI (6, 7) and a preliminary analysis of the post-translational proteolysis of CGC/SgII (8) in bovine chromaffin granules.
The proteolytic processing of CGA is a topic of growing interest, as biological activities have been attributed to peptides located along the sequence of CGA. For example, in the Nterminal domain, a peptide corresponding to the sequence 1-113 has been shown to inhibit hormone secretion in the bovine parathyroid gland (9); a homologous peptide, ␤-granin, corresponding to the sequence 1-115 has been isolated from rat pancreas, but its function has not yet been defined (10). Vasostatins are peptides containing the N-terminal sequence (1-76/113) (11) that exhibit vasoinhibitory activity of isolated human blood vessels (12,13). As early as 1988, it was established that CGA is the precursor of a peptide that inhibits the secretory activity on chromaffin cells (14), and recently, catestatin, a novel CGA fragment (344 -364), was characterized as a noncompetitive nicotinic cholinergic antagonist (15). In addition, pancreastatin (248 -293) is a peptide with multiple properties, since it negatively modulates insulin secretion from endocrine pancreatic islets (16,17), amylase release from exocrine pancreas (18), and acid secretion from parietal cells (19). Parastatin (347-419) is another CGA-derived peptide located in the C-terminal domain of CGA that inhibits parathyroid cell secretion (20). In addition to the autocrine or paracrine role in hormone secretion of these CGA-derived peptides, we have shown recently that numerous peptides present as water-soluble components of bovine chromaffin granules and released during secretion display antibacterial activity (7,(21)(22)(23)(24).
Human CGA is a single polypeptide chain of 439 residues, with an apparent molecular mass of 70 kDa as estimated by SDS-polyacrylamide gel electrophoresis gel and a pI of 4.7-5.2. The amino acid sequence of human CGA (25,26) indicates a real molecular mass of 48 kDa for the unmodified form of this protein. The difference between the apparent (70 kDa) and theoretical molecular mass (48 kDa) probably results from post-translational modifications (i.e. glycosylation, phosphorylation) (27,28) and the abundance of acidic residues (25%), which cause a slower migration during electrophoresis in the presence of sodium dodecyl sulfate (see Ref. 1, for review).
In 1997, using mild proteolysis, peptide separation, microsequencing, and mass analysis techniques, seven post-translational modification sites were detected in bovine CGA (29). Two glycosylation sites, each consisting of the trisaccharide NeuAc␣2-3Gal␤1-3GalNAc␣1-O-linked to Ser 186 and Thr 231 . The former residue is present in the antibacterial peptide named chromacin (22). Five phosphorylation sites were located on serine residues at positions Ser 81 , Ser 307 , Ser 372 , Ser 376 , and on Tyr 173 , this latter residue being the N-terminal amino acid of chromacin. Furthermore, studying the new antibacterial bovine CGA-derived peptides G-and PG-chromacin (CGA , we demonstrated that the two post-translational modifications (Tyr 173 and Ser 186 ) are both necessary for the antibacterial activity of chromacin.
Since 1989, it is well known that chromogranins can serve as immunocytochemical markers for neuroendocrine tissues and as a diagnostic tool for neuroendocrine tumors (3, 4, 30 -32). Because of their water-soluble properties, chromogranins and chromogranin-derived fragments are released together with catecholamines from adrenal chromaffin cells during stress situations and can be detected in the blood by radioimmunoassay techniques and enzyme assays (33)(34)(35)(36)(37)(38). Previously, we have shown that large fragments of CGA are excreted into the urine in some patients with carcinoid tumors (38). The present paper deals with the determination of the phosphorylation and carbohydrate binding sites of a large natural C-terminal CGAderived fragment, CGA 79 -439 , present into the urine of these patients. The strategy consists in characterizing the primary structure of modified phosphorylated and O-glycosylated peptides, which were isolated after proteolytic cleavage of CGA 79 -439 with endoproteinase Lys-C. Then, using successively separation by reverse phase HPLC, enzymatic modification of phosphorylated peptides, and complete analysis by sequencing and mass spectrometry (liquid chromatography/mass spectrometry and matrix-assisted laser desorption ionization timeof-flight), a detailed study was carried out. These post-translational modifications were located along the polypeptidic chain, compared with sequences of others species and discussed in relation with biological activity of natural CGA-derived fragments.

Isolation of Excreted CGA-derived Fragments
Urine was collected during a 24-h period from a patient with a histologically verified carcinoid tumor and multiple liver metastasis. The sample was collected, after informed consent, when the patient was on a clinical trial at the Endocrine Unit of Uppsala University Hospital. The study was also approved by the local Ethical Committee. The urine sample was first filtered through a 0.22-m membrane and then concentrated about 100 times in a dialysis tube (Spectra/Por; cutoff value, 6 -8 kDa). CGA-derived fragments were isolated in a one-step separation on an anion exchange column (Mono Q, FPLC; Amersham Pharmacia Biotech) using a linear gradient of 0.2 M ammonium acetate buffer at pH 6.0 to 1.0 M ammonium acetate buffer at pH 6.0 containing 1.0 M sodium chloride. Fractions containing CGA-derived fragments were concentrated on a Minicon concentrator (Amicon) and stored in Ϫ70°C before further analysis.

Purification of CGA-derived Peptides after Digestion with
Endoproteinase Lys-C CGA 79 -439 (2.5 nmol) was digested for 18 h at 37°C with endoproteinase Lys-C at a protein to enzyme weight ratio of 1000:1 in 100 mM Tris-HCl, pH 8.3. Generated peptides were then separated by HPLC, using the SMART system (Amersham Pharmacia Biotech), on a Macherey Nagel 300 -5C18 column (4 ϫ 250 mm; particle size 5 m and pore size 100 Å). Absorbance was monitored at 215 nm, and the solvent system consisted of 0.1% trifluoroacetic acid in water (solvent A) and 0.09% trifluoroacetic acid/acetonitrile (solvent B). Material was eluted at a flow rate of 550 l/min using, successively, a gradient of 0 -30% solvent B in solvent A over 46 min, followed by a gradient of 30 -50% over 13 min, and achieved by a gradient 50 -100% over 10 min. Each peak fraction was automatically collected by the SMART system and concentrated by evaporation, but not to dryness.

Purification of CGA-derived Glycopeptides after Digestion with
Endoproteinase Glu-C CGA 209/210 -245 mixture (100 pmol) was digested for 18 h at 37°C with endoproteinase Glu-C at a protein to enzyme weight ratio of 100:1 in 100 mM Tris-HCl, pH 8.3. Released fragments were then separated by HPLC, using the SMART system (Amersham Pharmacia Biotech) on a RPC-C2/C18 (2.1 ϫ 100 mm; particle size 3 m and pore size 120 Å). Absorbance was monitored at 215 nm, and the solvent system consisted of 0.1% trifluoroacetic acid in water (solvent A) and 0.085% trifluoroacetic acid, 60% acetonitrile, 39.915% water (solvent B). Elution was performed at a flow rate of 200 l/min using successively, a gradient 0 -65% solvent B in 60 min, followed by a gradient 65-100% over 10 min.

Sequence Analysis
The sequence of purified CGA-derived peptides was determined in our laboratory by automatic Edman degradation on an Applied Biosystems 473 A microsequencer. Samples (10 -20 pmol) were loaded onto polybrene-treated and precycled glass fiber filters (6).

Mass Spectrometry Analysis
Liquid Chromatography/Mass Spectrometry (LC/MS)-To isolate and characterize glycopeptides, we have performed LC/MS analysis of CGA-derived peptides obtained after endoproteinase Lys-C digestion of CGA. Then, CGA (500 pmol) was digested for 2 h at 37°C with endoproteinase Lys-C at a protein-to-proteinase weight ratio of 1000:1 in 100 mM Tris-HCl, pH 8.3. Then, peptides were separated with an HPLC system (Applied Biosystems 140 A Solvent Delivery System) equipped with a UV detector (UV Waters Detector 386) on a Narrowbore Macherey Nagel Nucleosil 300 -5C18 column (2 ϫ 150 mm). Absorbance was monitored at 214 nm, and the solvent system consisted of 0.1% trifluoroacetic acid/water (solvent A) and 0.1% trifluoroacetic acid/acetonitrile (solvent B). Material was eluted at a flow rate of 250 l/min using a gradient of 0 -80% solvent B in solvent A over 80 min. A major part of the eluent (90%) was analyzed by UV detection and an aliquot (10%) was measured by LC-MS. The mass spectrometer was calibrated under conditions using a mixture of polyethylene glycols (average masses 400 and 2000 Da). Spectra were scanned over m/z 320 -1800 for 6 s, and the total ion current was recorded.
Matrix-assisted Laser Desorption (MALDI-TOF)-The mass spectrometry analysis was carried out on a Brucker BIFLEX TM matrixassisted laser time-of-flight mass spectrometer equipped with the Scout TM high resolution optics with X-Y multisample probe, a gridless reflector, and the HIMAS TM linear detector. This instrument has a maximum accelerating potential of 30 kV and may be operated either in the linear or reflector mode. Ionization was accomplished with a 337-nm beam from a nitrogen laser with a repetition rate of 3 Hz. The output signal from the detector was digitized at a sampling rate of 250 MHz in linear mode and 500 MHz in reflector mode using a 1-GHz digital oscilloscope (Lecroy model). The instrument control and data processing were accomplished with software supplied by Brucker using a Sun Sparc workstation. These studies were realized according to the procedure previously described (22).

Sequence Comparisons
Sequence alignment of bovine CGA sequences with corresponding fragments of CGA from different species was performed using the Cameleon sequence alignment program using default parameters (41). Chromogranin sequences were retrieved from the Swiss-Prot data base.

RESULTS
In this study, we have isolated a major fragment of human CGA corresponding predominantly to the domain CGA 116 -439 and a minor larger CGA-derived fragment CGA 79 -439 , which are both excreted in urine of patients with carcinoid tumors (38). To determine phosphorylation and O-glycosylation sites included within tumoral CGA, the large C-terminal fragment was digested by endoproteinase Lys-C and the generated fragments were separated by HPLC on a reverse phase C18 column (Fig. 1A). The different peaks of the chromatogram were directly submitted to automatic Edman degradation and mass spectra analysis to detect post-translational modifications. It is important to note that all the sequences determined in this study are in accordance with the primary structure proposed by Konecki et al. (26); in contrast, they diverge on 15 points ( Fig.  2) from the primary structure reported by Helman et al. (25).

Identification of O-Glycosylation Sites
Carbohydrate analysis was performed using gas chromatography (22), and the carbohydrate content was evaluated to 5% (m/m, carbohydrate/protein): NeuAc, Gal, and GalNAc were detected in a molar ratio 1.2: 1.7: 1, suggesting short glycans with different structures.
Fragments resulting from endoproteinase Lys-C digestion of CGA 79 -439 were analyzed by LC-MS to detect areas containing O-glycosylated peptides (Fig. 1B). The HPLC chromatogram (a), the single ion recording of specific ions characteristic of glycosylation sites (b), and the total ionic current of the chromatogram (c) are indicated. In b, the presence of glycans was recovered in peaks of regions I-IV (lineated in Fig. 1A), and each area was analyzed to characterize the O-glycosylated sites. Area I included peaks corresponding to free saccharides, since areas II-IV contained glycopeptides.
Structural Characterization of O-Glycosylated Peptides Contained in Area II Peaks-Sequencing of material included in the two major peaks (30,31) eluting in region II (Fig. 1A) indicates the presence of peptides with N-terminal end located in position 145 and 124, respectively (Table I). MALDI-TOF MS analysis (negative mode) of peptide included in peak 31 reveals an experimental molecular mass of 2216 Da corresponding to the oxidized form of 124 -144 (oxidation of Met 140 ; calculated molecular mass 2200 Da). MALDI-TOF MS analysis of material included in peak 30 (Fig. 3A) shows four different major molecular species with respective molecular masses of 3682, 3974, 4048, and 4264 Da (Table I), indicating the presence of several different glycans. By comparison with the expected molecular mass of CGA 145-175 (3321 Da), the two molecular masses of 3682 and 3974 Da might be attributed to the Oglycosylated peptide CGA 145-175 with the disaccharide Gal␤1-3GalNAc␣1, corresponding to the antigens T described previously as characteristic of human adenocarcinoids (42), and the trisaccharide NeuAc␣2-3Gal␤1-3GalNAc␣1 reported previously for bovine CGA (22,29), respectively (Fig. 3A). In addition, the two other molecular masses of 4048 and 4264 Da are likely to correspond, respectively, to the O-glycosylated peptide CGA 145-175 with two disaccharides Gal␤1-3GalNAc␣1 and the tetrasaccharide with an additional NeuAc linked in 2-6 on GalNAc (Fig. 3A) as reported previously for fetuin (43). To obtain confirmation of the structure of these glycans, material included in peak 30 was slot-blotted onto nitrocellulose sheet and immunodetected with a panel of three lectins (MAA, SNA, and PNA), the specificity of which was reported previously (22). Experimental data have shown unambiguously the simultaneous presence of the following linkages: Gal␤1-3GalNAc␣1-O (PNA), NeuAc␣2-3Gal␤ (MAA), and NeuAc␣2-6GalNAc. Thus, our data revealed the presence of four O-glycosylated moities on the peptide CGA 145-175 , including complete forms with a tri-or a tetrasaccharide and two truncated glycans corresponding to disaccharides (Fig. 3A).
Primary structure of CGA 145-175 (Fig. 3A) includes four potential O-glycosylated residues corresponding to Thr 163 , Thr 165 , Ser 170 , and Ser 173 . The sequence in the vicinity of the serine residues Ser 170 and Ser 173 (PPAS 170 LPS 173 QKYPGP) fits with the sequence patterns described by Wilson for Oglycosylation sites and characterized by high proline, serine, and threonine content (44). In addition, the presence of clusters of several closely spaced glycosylated residues was reported by these authors. In contrast, predictions of mucin type O-glycosylation sites in mammalian proteins, according to Hansen method (45) suggest the presence of glycans on residues Thr 165 , Ser 170 , and Ser 173 . To determine the two O-glycosylation sites we have submitted the peptidic material included in peak 30 to a proteolytic digestion with endoproteinase Glu-C and automatic Edman degradation. Sequencing of the derived peptides indicates the presence of two unmodified serine residues Ser 170 and Ser 173 , while threonine residues Thr 163 and Thr 165 were undetected. Furthermore, the cleavage of the fragment CGA 145-175 by endoproteinase Glu-C was very weak, indicating that the glycans are located in the vicinity of the cleavage point Glu 161 . In contrast, the cleavage by endoproteinase Lys-C of the peptidic linkage Lys 175 -Tyr 176 is achieved with a good yield, suggesting that polysaccharide chains are distant. Then, we proposed that two glycans are linked on residues Thr 163 and Thr 165 .
Structural Characterization of O-Glycosylated Peptides Contained in Area IV Peaks-Sequencing of material included in peak 63 eluting in region IV (Fig. 1A) indicates the presence of CGA-derived peptide beginning at position 198 (Table I) A, Isolation by HPLC of CGA-derived peptides generated after endoproteinase Lys-C digestion. Elution profile of the endoproteinase Lys-C digest on a Macherey Nagel reverse phase Nucleosil 300 -5C18 column (4 ϫ 250 mm). Absorbance was monitored at 215 nm, and elution was performed with a linear gradient as indicated on the right-hand scale. Areas I-IV contained carbohydrate molecules. B, on-line LC/MS of the endoproteinase Lys-C digest of human CGA (500 pmol). In a, the y axis represents the absorbance at 214 nm (0.2 absorbance units as full scale) and x axis indicates the elution time of the different fractions. b, single ion recording specific to glycopeptides (SIR). c, specific total ion current (TIC). verify the location of the O-glycosylation site, a mixture of CGA 209/210 -245 peptides has been treated with endoproteinase Glu-C.
Digestion of CGA 209/210 -245 with Endoproteinase Glu-C-After digestion, the generated CGA-derived peptides were isolated by HPLC (Fig. 4A). Three peaks were isolated, sequenced, and submitted to MALDI-TOF MS analysis. Material included in peak 1 contained a mixture of unmodified CGA 225/228/230 -245 , while material included in peak 3 contained nondigested CGA 209/210 -245 . Sequencing of peptidic material included in peak 2 shows the presence of two CGA-derived peptides with the N-terminal ends located at residues 219 and 221 (Table II). The six peaks obtained by MALDI-TOF MS (Fig. 4B) (43). In addition, we have observed that endoproteinase Glu-C is unactive toward the linkage Glu 230 -Glu 231 , in contrast to endoprotease Lys-C, which cleaves the large CGA-derived fragment CGA 79 -439 after the lysine residue Lys 245 , which suggests that glycans are located close to Glu 230 -Glu 231 . Thus, we propose that threonine Thr 233 is glycosylated.
At this stage, we have determined three O-linked carbohydrate attachment sites on residues Thr 163 , Thr 165 , and Thr 233 located in the middle part of the whole protein (Fig. 5). It is important to note that a fraction (25%) of carcinoid CGA was not completely O-glycosylated since unmodified forms CGA 145-175 and CGA 210 -245 were recovered in peaks 38 and 59, respectively (Fig. 1A).

Identification of Phosphorylation Sites
Structural analysis by sequencing and MALDI-TOF MS of each HPLC peak ( Fig. 1A; Table I) 3. MALDI-TOF mass spectrometry of the modified glycopeptides. A, the glycopeptide included in peak 30 (Fig. 1A) has been identified as CGA 145-175 . B, the glycopeptide included in peak 50 (Fig. 1A) has been identified as CGA 210 -245 . C, glycopeptide included in peak 63 (Fig. 1A) (Fig. 6A) and CGA 246 -277 (Fig. 6B). After alkaline phosphatase treatment we observed the removal of a mass of 80 Da on each peptide, confirming the presence of a phosphate group on either Ser 252 or Ser 254 and Ser 304 or Ser 315 . To characterize the exact location of the two phosphorylated residues, peptidic material included in peak 37 was digested by trypsin and analyzed by MALDI-TOF. The two modified serine residues were identified as Ser 252 and Ser 315 (data not shown).
Sequence and mass spectra analysis of material included in peak 63 show the presence of an additional increment of 80 Da in the sequence CGA 198 -245 (Table I). As this modification is not present in CGA 210 -245 , the location of the phosphorylated residue is likely to be in the region CGA 198 -209 . A unique potential phosphorylated residue is Ser 200 (Fig. 2).
In conclusion, we have identified three phosphorylated sites along the polypeptidic fragment CGA 79 -439 located on residues Ser 200 , Ser 252 , and Ser 315 (Fig. 5).  1 ϫ 100 mm). Absorbance was monitored at 215 nm, and elution was performed with a linear gradient as indicated on the right-hand scale. B, MALDI-TOF mass spectrometry of the glycopeptide included in peak 2 (Fig. 4A) and identified as CGA 219/221-245 .

DISCUSSION
Chromogranins A and B are present in multiple secretory cell types of numerous species within the animal kingdom (1,(46)(47)(48)(49). On the basis of secondary and tertiary structures predicted from its sequence, CGA possesses a "random coil" structure (1). In addition, according to Kyte and Doolittle predictions (50), this protein is very hydrophilic throughout the length of its polypeptidic chain (6). Besides the 70-kDa molecular species, several observations have reported the presence of a molecular mass 80-90 kDa diffuse form of chromogranin A immunoreactivity, as full-length chromogranin A-core proteoglycan, in secretory granules from bovine adrenal medulla and from PC12 cells (51). Multiple neuroendocrine sources other than the adrenal medulla appear to contribute to the high basal circulating CGA concentration in man (1). The widespread occurrence of CGA is indicative of some important biological functions for this protein. Despite the fact that CGA has been largely studied since its discovery thirty years ago, the characterization of these functions is still an open question. However, in addition to the autocrine or paracrine role in hormone secretion of CGA-derived peptides, our recent work has shown that numerous peptides with antibacterial activity are present as water-soluble components of bovine chromaffin granules and are released during secretion (22)(23)(24). In addition, it was recently found that mixtures of CGA-derived fragments can inhibit fibroblast adhesion (52). In contrast, pro-adhesive effects were observed with recombinant N-terminal fragments corresponding to residues 1-78 and 1-115 and with a synthetic peptide corresponding to the domain CGA 7-57 (52). Furthermore, it is established that within the brain CGA is localized in neurodegenerative areas associated with reactive microglia, and it was found that both recombinant human CGA and natural bovine CGA were able to induce an activated phenotype in rat microglial cells maintained in primary culture (53). More recently, it was shown that the recombinant human N-terminal fragment CGA 1-78 stimulated microglial cells in primary culture to secrete heat-stable diffusible neurotoxic agents (54). Furthermore, antibodies against chromogranins have been widely used for immunohistochemical staining of endocrine tissues and tumors of neuroendocrine origin, i.e. in pancreatic tumors, pheochromocytoma, midgut carcinoid tumors, prostate, etc. Antibodies have also been raised to develop specific radioimmunoassays for plasma measurements of the different chromogranins. Measurement of CGA level in plasma has been proved as a useful tool in the diagnosis of peptide-producing endocrine neoplasms. Recent investigation has shown that the 24-h urine measurement of catecholamine and their related metabolites, together with serum CGA and CGB values, are proportional to pheochromocytoma mass and provide reliable diagnosis markers (55).
Knowledge of CGA primary structure from tumoral tissue should provide information with regards to the use of this protein as a tumoral marker (32). In the present paper we report for the first time the characterization of six post-translational modifications of a large CGA 79 -439 present in the urine of patients with carcinoid tumors. Three phosphorylated serine residues were identified as Ser 200 , Ser 252 , and Ser 315 and three O-glycosylation sites were found on residues Thr 163 , Thr 165 , and Thr 233 . In a recent work, we have identified five phosphorylated residues on the polypeptidic chain of bovine CGA located on residues Ser 81 , Tyr 173 , Ser 307 , Ser 372 , and Ser 376 (29). Comparison of human, bovine (56), pig (57), mouse (58), and rat (59) sequences in the vicinity of post-translationnaly modified residues is shown in Fig. 7. Although regions are highly conserved, particularly those bearing phosphorylation modifications (Ser 307 , Ser 372 , and Ser 376 in bovine sequence), differences can be pointed out. The two phosphorylated residues Ser 200 and Ser 315 are strictly conserved and included in homologous sequences with the typical pattern E-K/R-G-L/P-S 200 -A-E/Q-P/Q-G/Q-and E/Q-E/Q-E-E-R/Q-L-S 315 -R/K/E-E-W-E-D/N. In contrast, Ser 252 is included in a nonconservative sequence. These differences may however result from the carcinoid nature of CGA, the modifications on the human normal protein yet remain to be studied. In addition, Ser 315 in human protein, which corresponds to Ser 307 in bovine sequence, is also phosphorylated. The other residues Ser 80 , Tyr 176 , Ser 380 , and Ser 384 in human sequence are not modified. In contrast, the phosphorylated residue Ser 200 in the human sequence is located within a conservative region, but the corresponding residue in bovine sequence is not phosphorylated. The recent characterization of the phosphorylated sites of bovine CGA has established the presence of a phosphorylated tyrosine residue Tyr 173 (22). Tyrosine phosphorylation is not a common post-translational modification, since it represents only 0.03% of the phosphorylated amino acids in normal cells (60). The significance of this tyrosine phosphorylation is not known yet, although we have reported recently that chromacin, the CGA-derived peptide 173-194, displays antibacterial activity when the N-terminal Tyr 173 residue is phosphorylated (22). On the basis of protein consensus domain specific to kinases (61), it is possible to predict that protein kinase C may introduce phosphate group on residues Ser 200 , Ser 252 , and Ser 315 , while cGMP-dependent protein kinase may modify residues Ser 200 and Ser 252 . All present and previous data (29) show that phosphorylation of CGA may be a tissue-specific process resulting from the absence of specific enzymatic material necessary for phosphorylation reactions or the presence of specific phosphatases acting  during the transfer of CGA 79 -439 from chromaffin and enterochromaffin cells to plasma and urine. Along this line, processing of CGA has been shown to be tissue-specific (62), and the carcinoid nature of the biological material may induce variation in protein processing as compared with normal tissues (62,63).
In the present study we have identified three O-glycosylation sites located on Thr 163 , Thr 165 , and Thr 233 . The presence of the three glycans (Fig. 5) on human CGA 79 -439 gives rise to a calculated sugar/protein ratio of 5% in accordance to the sugar experimental content that we have obtained. The attachment O-glycosylation sites Thr 163 and Thr 165 correspond to prolinerich sequences. However, they are not used as such in bovine sequence, and it is not yet known whether they are genuine glycosylation sites in CGAs from rat, mouse, and pig. Therefore, it is probable that post-translational modifications in the human carcinoid CGA are not representative of the modifications present on the non-carcinoid normal protein. In contrast, the residue Thr 233 is present in bovine, human, mouse, and rat CGAs. In pig this threonine is changed into a serine residue, which might well represent a glycosylation site. With regard to the Wilson consensus sequence (44), the proline residues are present upstream and downstream in the bovine, human, and pig sequences but are scarse in rat and mouse CGAs. Concerning the O-glycosylation sites included on bovine adrenal medullary CGA, we had determined previously that an unique carbohydrate moiety composed of the trisaccharide NeuAc␣2-3Gal␤1-3GalNAc␣1 was found to be located on Ser 186 and Thr 231 (29). By comparison with the present study it appears that the glycosylation on Thr 233 is a conservative process, whereas the glycosylation located on residues Thr 163 and Thr 165 might be specific to the human protein or to the carci-noid protein. The identification of the disaccharide Gal␤1-3GalNAc␣1 and the monosaccharide GalNAc␣1, corresponding to the antigens T and Tn, respectively, is in agreement with the human adenocarcinoid source (42).
Concerning the O-glycosylation and phosphorylation modifications, we have reported recently that they are necessary for the full antibacterial activity of chromacin peptides. The natural CGA-derived antibacterial peptide, pro-chromacin CGA 79 -431 contains these seven modifications. The antibacterial activity of chromacin CGA 173-194 is correlated with the presence of O-glycosylation modification on Ser 186 or/and phosphorylation on Tyr 173 (22). In the present study we have established that these two modifications are missing in the human CGA sequence homologous to bovine chromacin. However, this absence may be related to the carcinoid nature of CGA, and these modifications may be well present in the protein expressed in normal human tissues.
The three post-translational glycosylations of CGA probably have a structurally related function. As recently the most antigenic sites of recombinant human CGA have been characterized (64), it seems interesting to correlate their location with regard to the modified residues. The modified residues Thr 233 , Ser 252 , and Ser 315 appear to be preferentially located into or near domains with high antigenic feature 222-230 and 315-330, whereas three modified residues in position 163, 165, and 200 are located in the 163-210 region described to possess low antigenicity. These results indicate that the two glycosylated residues clustered in the N-terminal moiety of the natural excreted CGA (Thr 163 and Thr 165 ) may induce important structural modifications of the protein. For instance, these two moieties may alter proteolytic cleavage of CGA in tumoral tissues and explain the recovery of the large CGA-derived fragment in urine of patients, whereas it is not detectable in normal fluids. In addition, it is important to discuss our present data in relation with a new and accurate sandwich immunoassay (32). To develop this assay, 24 monoclonal antibodies were raised against human CGA. The better combination was that involving two monoclonal antibodies directed against contiguous epitopes located in the mid CGA 145-245 domain. Therefore, taking into account the location on Thr 163 , Thr 165 , and Thr 233 of O-glycosylation sites, it is clear that these glycans prevent the natural proteolytic degradation of this central domain; the quantification of this central domain allows the accurate determination of circulating CGA level (32). This mid-molecule fragment has been shown to be the major circulating constituent in case of renal dysfunction, kidney representing the site where CGA immunoreactivity is removed or destroyed in normal subjects (65). Some years ago, one of us (38) determined four natural proteolytic cleavage sites of human CGA from carcinoid (38). Among them, the two major cleavage sites correspond to dibasic amino acid sites (115-116 and 209 -210), whereas the two minor are located in position Asp-Pro (272-273) and Arg-Gly (394 -395). These four proteolytic cleavage sites have also been described earlier for CGA isolated from neuroendocrine tumors (51, 62, 63, 66 -68), and it has thus been suggested that this natural degradation occurs in tumors or serum but not in urine.
Mid-molecule fragments of CGA are recovered in the urine of patients with carcinoid tumors, whereas CGA is not detectable in the urine of normal subjects; nevertheless, plasma measurement of CGA has been shown to be a more reliable diagnostic marker than the corresponding urine analysis, since CGA was present in urine of one-third of patients (69). The presence of CGA in urine may result from pharmacological treatment (69), dysfunction (65), or structural abnormalities. The characterization of specific short carbohydrate structures (antigens T and Tn) from carcinoid CGA allow us to envisage the development of new tools, including MALDI-TOF or carbohydrate dot blotting, to detect and measure these CGA forms specific to tumors. Subsequently, these tools may also be used to establish corre-