Structural divergence of human ghrelin. Identification of multiple ghrelin-derived molecules produced by post-translational processing.

Ghrelin, a novel 28-amino acid peptide with an n-octanoyl modification at Ser(3), was isolated from rat stomach and found to be an endogenous ligand for the growth-hormone secretagogue receptor (GHS-R). This octanoyl modification is essential for ghrelin-induced GH release. We report here the purification and identification of human ghrelin from the stomach, as well as structural analysis of the human ghrelin gene and quantitation of changes in plasma ghrelin concentration before and after gastrectomy. Human ghrelin was purified from the stomach by gel filtration and high performance liquid chromatography, using a ghrelin-specific radioimmunoassay and an intracellular calcium influx assay on a stable cell line expressing GHS-R to test the fractions. In the course of purification, we isolated human ghrelin of the expected size, as well as several other ghrelin-derived molecules. Classified into four groups by the type of acylation observed at Ser(3); these peptides were found to be non-acylated, octanoylated (C8:0), decanoylated (C10:0), and possibly decenoylated (C10:1). All peptides found were either 27 or 28 amino acids in length, the former lacking the C-terminal Arg(28), and are derived from the same ghrelin precursor through two alternative pathways. The major active form of human ghrelin is a 28-amino acid peptide octanoylated at Ser(3), as was found for rat ghrelin. Synthetic octanoylated and decanoylated ghrelins produce intracellular calcium increases in GHS-R-expressing cells and stimulate GH release in rats to a similar degree. Both ghrelin and the ghrelin-derived molecules were found to be present in plasma as well as stomach tissue. Plasma levels of immunoreactive ghrelin after total gastrectomy in three patients were reduced to approximately half of their pre-gastrectomy values, after which they gradually increased. This suggests that the stomach is the major source of circulating ghrelin and that other tissues compensate for the loss of ghrelin production after gastrectomy.

Growth hormone (GH) 1 secretion from the pituitary gland is regulated by two hypothalamic hormones, growth hormonereleasing hormone and somatostatin (1,2). A third independent pathway responsible for regulation of GH release has recently emerged from studies of artificial GH secretagogues (GHSs) (3,4). GHSs are synthetic compounds that are potent stimulators of pituitary GH release, acting through the GHS receptor (GHS-R) (5)(6)(7). Previously, we identified ghrelin, an endogenous ligand for GHS-R, from rat stomach (8). Ghrelin, a 28-amino acid peptide capable of stimulating GH release in vitro and in vivo, has a unique n-octanoyl modification at its third serine residue (Ser 3 ), which is essential for this function (9 -11). Subsequently, des-Gln 14 -ghrelin, also isolated from rat stomach, was identified as a second endogenous ligand for GHS-R (12). Des-Gln 14 -ghrelin is produced from the ghrelin gene by an alternative splicing mechanism and is also octanoylated at Ser 3 .
In the present study, we purified human ghrelin from the stomach, using a ghrelin-specific radioimmunoassay (RIA) (13) and an intracellular calcium influx assay on a stable cell line expressing GHS-R (8,12). During the course of purification, we noticed several minor peptides with characteristics different from standard ghrelin that displayed ghrelin-like activity. We identified these stomach peptides as ghrelin-derived molecules and examined the levels of ghrelin as well as these ghrelinderived molecules in human plasma.
The ghrelin gene is abundantly expressed in rat (8) and human (14) stomach, and in the rat, no other major sources of ghrelin production have been observed (13,15). These results prompted us to question whether ghrelin should be drastically reduced following gastrectomy. To address this question, we examined the change in plasma levels of immunoreactive ghrelin (ir-ghrelin) in humans before and after total gastrectomy.
Both types of antisera exhibited 100% cross-reactivity with human and rat ghrelins. The anti-rat ghrelin-(1-11) antiserum specifically recognized the n-octanoylated portion at Ser 3 of ghrelin and did not recognize des-acyl ghrelin. The anti-rat ghrelin-(13-28) antiserum equally recognized n-octanoyl and des-acyl ghrelins. In the following sections, the RIA system using antiserum against the N-terminal fragment of rat ghrelin-(1-11) is termed N-RIA; the RIA system using antiserum against the C-terminal fragment (13-28) is termed C-RIA.
Detection of Ghrelin Activity by Calcium Mobilization Assay-CHO-GHSR62 cells, which stably express rat GHS-R (8,12), were plated in flat-bottom, 96-well black-wall plates (Corning Inc., Corning, NY) at 4 ϫ 10 4 cells/well for 12 h prior to the assay. The cells were loaded with 4 M Fluo-4-AM fluorescent indicator dye (Molecular Probes, Inc., Eugene, OR) for 1 h in assay buffer (Hanks' balanced salts solution, 20 mM HEPES, 2.5 mM probenecid, 1% fetal calf serum) and washed four times in assay buffer without fetal calf serum. Intracellular calcium concentration ([Ca 2ϩ ] i ) changes were measured using a fluorometric imaging plate reader (FLIPR, Molecular Devices, Sunnyvale, CA). Maximum changes in fluorescence compared with the baseline were used to quantitate the agonist responses.
Purification of Human Ghrelins from Stomach-A human stomach mucosa (27 g) was minced and boiled for 5 min in 5ϫ volumes of water to inactivate intrinsic proteases. The solution was adjusted to 1 M acetic acid (AcOH)-20 mM HCl. The stomach tissue was homogenized with a Polytron mixer. The supernatant of the extract, obtained after 30-min centrifugation at 11,000 rpm, was concentrated to ϳ25 ml by evaporation. The residual concentrate was subjected to acetone precipitation in 66% acetone. After removal of the precipitate, the supernatant was evaporated to remove the acetone and then loaded onto a 10-g cartridge of Sep-Pak C18 (Waters, Milford, MA), pre-equilibrated in 0.1% trifluoroacetic acid. The Sep-Pak cartridge was washed with 10% CH 3 CN/ 0.1% trifluoroacetic acid, and the peptides were eluted in 60% CH 3 CN/ 0.1% trifluoroacetic acid. The eluate was evaporated and lyophilized. The lyophilized materials were then redissolved in 1 M AcOH and applied to a Sephadex G-50 gel-filtration column (1.8 ϫ 130 cm, Amersham Biosciences, Uppsala, Sweden). Five-milliliter fractions were collected. A portion of each fraction was subjected to ghrelin-specific RIA and the intracellular calcium influx assay using CHO-GHSR62 cells. The active fractions (#43-46) were separated by carboxymethyl (CM) ion-exchange high performance liquid chromatography (HPLC) on a TSK CM-2SW column (4.6 ϫ 250 mm, Tosoh, Tokyo, Japan) using an ammonium acetate (HCOONH 4 ) (pH 4.8) gradient of 10 mM to 1 M in the presence of 10% CH 3 CN and a flow rate of 1 ml/min for 100 min. Onemilliliter fractions were collected and subjected to ghrelin-specific RIAs and intracellular calcium influx assays. The six active fractions (fractions A-F) separated by CM-HPLC were finally individually purified using C18 reverse-phase HPLC (RP-HPLC) columns (Symmetry 300, 3.9 ϫ 150 mm, Waters). The amino acid sequences of the purified peptides were analyzed with a protein sequencer (494, Applied Biosystems, Foster City, CA).
Mass Spectrometric Analysis of Human Ghrelins-Electrospray ionization mass spectrometry (ESI-MS) was performed on a quadrupole mass spectrometer SSQ7000 (Finnigan, San Jose, CA) equipped with a Finnigan ESI source. A needle capillary was heated to 150°C to evaporate the samples. Samples (ϳ20 pmol) were dissolved in 50% (v/v) methanol, 1% AcOH and introduced into the ϩ4.5 kV (positive ionization) ion source at a flow rate of 5 l/min by direct infusion with a syringe pump. Molecular masses of the purified peptides were calculated using ICIS software Bioworks provided by Finnigan.
Human Ghrelin Peptide Synthesis-Peptide synthesis of acylated human ghrelin was performed as previously described for rat ghrelin (8,12). Fully protected 27-and 28-amino acid peptides (with the exception of the exposed hydroxyl group of Ser 3 ) were synthesized by the Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid-phase method using a peptide synthesizer (433A, Applied Biosystems). The Ser 3 hydroxyl groups were acylated with n-octanoic acid or n-decanoic acid by the action of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide in the presence of 4-(dimethylamino)pyridine.
Cloning of Human Prepro-ghrelin cDNA-Based on the amino acid sequence determined for purified rat ghrelin, we searched the Gen-Bank TM expressed sequence tag data base. One rat expressed sequence tag sequence (accession number AI549172) contained the rat ghrelin sequence. Based on this expressed sequence tag sequence, we designed sense and antisense primers (5Ј-TTGAGCCCAGAGCACCAGAAA-3Ј and 5Ј-AGTTGCAGAGGAGGCAGAAGCT-3Ј, respectively) and performed PCR on a rat stomach cDNA library. The conditions for the PCR involved 35 cycles of 98°C for 10 s, 55°C for 30 s, and 72°C for 1 min. A rat ghrelin cDNA of 501 bp was obtained. The amplified fragment was labeled with [ 32 P]dCTP and used as a screening probe for a human stomach cDNA library, which was constructed from 1 g of human stomach poly(A) ϩ RNA (Clontech, Palo Alto, CA) using a cDNA synthesis kit (Amersham Biosciences). By this method, a full-length human ghrelin cDNA was isolated. Several positive phages were isolated and subcloned into the plasmid pBS. Both strands of cloned cDNAs were sequenced.
In Vivo Assay of Growth Hormone-releasing Activity-Male Wister rats (270 -300 g) were prepared with a single indwelling jugular catheter under sodium pentobarbital. Each rat received a 0.5-or 2-nmol injection of synthetic human ghrelins. Blood samples were collected at 0, 5, 10, 15, 20, 30, and 60 min after injection. All samples were centrifuged immediately, and the plasma samples were assayed for GH using the Biotrak rat GH enzyme immunoassay system (Amersham Biosciences, Buckinghamshire, UK).
Human Blood Sample Analysis-To study the changes in plasma levels of ir-ghrelin after gastrectomy, we followed three patients (one male and two females), 50 -61 years of age, who underwent total gastrectomy due to gastric cancer. Blood samples were obtained before and within 30 min after gastrectomy, and then 3, 7, 14, 30, 150, and 240 days after gastrectomy for measurement of plasma ir-ghrelin levels. Four control males (60 -70 years of age) who underwent standard partial colectomy due to cancer, were also examined. Their blood samples were obtained before and 1 day after surgical operation.
Preparation of Human Plasma Samples-Whole blood samples were collected with Na 2 EDTA (2 mg/ml) and aprotinin (500 KIU/ml). Plasma, after centrifugation at 4°C, was diluted with an equal volume of 0.9% saline. The samples were loaded onto a Sep-Pak C18 cartridge (Waters) pre-equilibrated with 0.9% NaCl, washed with 0.9% NaCl and 5% CH 3 CN/0.1% trifluoroacetic acid, and then eluted with 60% CH 3 CN/ 0.1% trifluoroacetic acid. After lyophilization of the eluates, they were subjected to RIAs for ghrelin as described above.

RESULTS
Purification of Human Ghrelins-We purified human ghrelin from stomach tissue extracts by gel-filtration chromatography and HPLC, using an intracellular calcium influx assay with a stable cell line expressing rat GHS-R (CHO-GHSR62) and two ghrelin-specific RIAs to screen fractions for the presence of ghrelin. Ir-ghrelin obtained by N-RIA specifically represents active acylated ghrelin, whereas ir-ghrelin acquired by C-RIA represents the total immunoreactivity of both acylated and des-acyl ghrelin. Fig. 1a depicts the gel-filtration chromatographic separation of stomach peptide extracts. Fractions possessing ghrelin immunoreactivity and promoting intracellular calcium influx were eluted at a molecular weight of roughly 3000. Active gel-filtration fractions were further separated by CM ion-exchange HPLC into six fractions, A-F (Fig.  1b). Fractions B-F induced intracellular calcium influxes and possessed ir-ghrelin as assessed by both C-RIA and N-RIA. Fraction A possessed only C-RIA ir-ghrelin and did not induce intracellular calcium influx. Each of the five factions, B-F, were separately purified to homogeneity by RP-HPLC and subjected to the calcium-mobilization assay (Fig. 1c). Each of the five fractions contained one active peak, except for fraction E, which had two (peaks E-I and E-II). Fraction A was also separated by RP-HPLC into two peaks (peaks A-I and A-II), which were found to possess ir-ghrelin by C-RIA (Fig. 1d). Each of the eight resulting purified active peaks contained a single peptide, and these peptides were then subjected to further analysis.
Structural Analyses of Human Ghrelins-The eight purified peptides were subjected to protein sequencer, showing that peaks C, E-I, E-II, and F shared the 28-amino acid sequence GSXFLSPEHQRVQQRKESKKPPAKLQPR, whereas peaks B and D were of the sequence GSXЈFLSPEHQRVQQRKESKKP-PAKLQP, identical except for the lacked of the C-terminal arginine. Complementary DNA analysis of human ghrelin indicated that the third X and XЈ residues should be serine, and the serine residues were acyl-modified as described below. Moreover, the amino acid sequences of peak A-I and A-II were GSSFLSPEHQRVQQRKESKKPPAKLQPR and GSSFLSPE-HQRVQQRKESKKPPAKLQP, respectively, the same as the acyl-modified ghrelin peptides. We did not detect des-Gln 14ghrelin in this human stomach tissue.
To determine whether the purified peptides were also modified by n-octanoic acid at Ser 3 as is rat ghrelin, we subjected the peptides to ESI-MS and measured their molecular masses ( Table I). The measured molecular mass of peak C, the major active peptide, was 3371.3 Ϯ 0.1, and the calculated molecular mass of the 28-amino acid sequence is 3244.6. The discrepancy, 126.7 mass units, strongly suggests that the hydroxyl group of the Ser 3 in this peptide is indeed replaced by an n-octanoyl moiety (C8:0). The same was found for peak B, which had a measured mass (3214.6 Ϯ 0.6) that was 126.2 mass units higher than the calculated molecular mass of the 27-amino acid peptide (3088.4), indicating modification by n-octanoic acid. The measured molecular masses of the peptides from peaks D and F were ϳ154 molecular mass units higher than the calculated molecular masses, indicating that these two peptides were modified by n-decanoic acid (C10:0). Peaks E-I and E-II were both 152.6 molecular mass units higher than the calcu-lated molecular masses, 2 mass units smaller than what would be expected for decanoyl modification. Based on this result and the fact that peaks E-I and E-II were eluted at a time between the octanoyl (peaks C and E) and decanoyl-modified ghrelins (peaks D and F) by RP-HPLC, it is most likely that the peptides from peaks E-I and E-II are modified by decenoic acid (C10:1). The amounts of peptide purified from peaks E-I and E-II were very low, preventing a determination of the double-bond site of the decenoic acid. In conclusion, we were able to divide the collected ghrelins into four groups on the basis of acyl modification at Ser 3 : non-acylated, octanoylated, decanoylated, and possibly decenoylated.
To    ." The yield of purified human ghrelin was ϳ300 pmol from 27 g of stomach mucosa, and the molar ratio of the various subsets are shown in Table II.
Pharmacological Characterization of Ghrelins Using GHS-R-expressing Cells- Fig. 2  In Vivo Effects of Human Ghrelins on GH Secretion-To confirm that human ghrelins possessed GH-releasing activity, we intravenously injected synthetic ghrelins into anesthetized rats and measured plasma GH concentrations. After injection of the each of the four synthetic ghrelins, plasma GH concentrations increased and reached a maximum within 10 -15 min (Fig. 3). Each of the peptides displayed nearly identical doseresponse relationships, confirming that the newly identified human ghrelin-derived molecules are endogenous GH-releasing peptides with similar potency to human ghrelin.
Structure of the Human Prepro-ghrelin cDNA-Using a rat ghrelin cDNA, we screened a human stomach cDNA library under low stringency conditions and obtained positive phages. Analysis of these clones yielded a deduced amino acid sequence for human prepro-ghrelin (a 117-amino acid precursor) (Gen-Bank TM accession number AB029434), depicted in Fig. 4. The putative initiation codon ATG is located at nucleotides 34 -36, preceded by the consensus initiation sequence, whereas a terminal codon TAG is found 117 codons downstream at position 385-387. A typical polyadenylation signal, AATAAA, is found at position 494 -499.
Although nearly all of the cDNA clones isolated from human stomach encoded the prepro-ghrelin precursor, a few cDNA clones encoded the prepro-des-Gln 14 -ghrelin precursor. Also, although we were not able to isolate des-Gln 14 -ghrelin from the stomach extracts during this study, this result indicates that des-Gln 14 -ghrelin is indeed present in very low amounts in the human stomach.
Characterization of Human Plasma Ghrelin Immunoreactivity-To confirm the presence of multiple molecular forms of ghrelin in human plasma in addition to the stomach, Sep-pak extracts of normal human plasma were fractionated by CM ion-exchange HPLC in exactly the same manner as those from the stomach. The HPLC pattern of plasma extracts in terms of the presence of ir-ghrelin was observed to be similar to that of the stomach extracts (Fig. 5), with peaks a-f emerging at positions identical to that of peaks A-F in Fig. 1b. Thus, it can be concluded that both ghrelin and the ghrelin-derived molecules circulate in human blood.
A minor unknown peak of ir-ghrelin (peak g) detected only by C-RIA was observed in human plasma. This peak accounted for ϳ15% of all ir-ghrelin by C-RIA. By RP-HPLC, this ir-ghrelin peak was eluted earlier than that of des-acyl ghrelin (data not shown), suggesting that the unknown ir-ghrelin in peak g results from digested ghrelin.
Plasma Ghrelin Levels in Gastrectomized Patients-To clar-  ify whether circulating ghrelin is indeed drastically reduced after gastrectomy as would be expected, plasma ir-ghrelin was measured before and after total gastrectomy in three patients (Fig. 6). Within 30 min after gastrectomy, plasma ir-ghrelin was found to decrease to approximately half of its pre-surgery levels. The levels remained depressed for roughly a week, but after that they began to increase. By the end of the day 240, two of the patients had ir-ghrelin levels that were two-thirds of their original levels, and one patient's ghrelin levels had completely normalized. In contrast, the subjects who underwent par-tial colectomy showed no change in plasma ir-ghrelin before (75.4 Ϯ 22.0 fmol/ml, mean Ϯ S.D.) and 1 day after operation (69.6 Ϯ 17.8 fmol/ml). These results suggest that the stomach is the major source of circulating ghrelin, and the other tissues compensate to maintain circulating ghrelin levels after gastrectomy. DISCUSSION Here, we reported the purification and characterization of human ghrelin and the other minor ghrelin-derived molecules from the stomach. The major active form of human ghrelin is a 28-amino acid peptide with an n-octanoyl modification at Ser 3 . This peptide is identical to rat ghrelin with the exception of two residues (Arg 11 -Val 12 ). The ghrelin-derived molecules we observed include octanoyl ghrelin- , decanoyl ghrelin, decanoyl ghrelin- , and decenoyl ghrelin. Moreover, the nonactive forms des-acyl ghrelin and des-acyl ghrelin-  were also present in the human stomach. As described, we were able to classify human ghrelin and the ghrelin-derived molecules into two groups on the basis of amino acid length and into four groups by type of acylation at Ser 3 . Furthermore, all of these molecular forms of ghrelin were found in human plasma as well as in the stomach. In human stomach, the processing product ratio of 27-amino acid to 28-amino acid ghrelins was observed to be ϳ1:3.
It is likely that the 27-and 28-amino acid ghrelin molecules isolated in this study are produced through alternative Cterminal processing of the same ghrelin precursor. It is well known that peptide hormones are cleaved by processing proteases to product multiple forms, such as the enkephalins (16), endorphins, dynorphins (17), corticotropins, and ␤-lipotropins (18). Many of the known proteolytic precursor cleavage events occur at pairs of basic amino acid residues (Lys or Arg), and both basic residues are usually absent from the resultant products (19). However, some proteolytic cleavages, as in the case of cholecystokinin, occur immediately after a C-terminal single basic residue (especially Arg) (20). Ghrelin-(1-28) may fit into this category, because cleavage to produce this peptide occurs following the C-terminal Pro 27 -Arg 28 . Interestingly, this basic arginine residue remains at the C terminus of ghrelin-(1-28) but is removed from ghrelin-(1-27) (which terminates in proline). A similar cleavage profile is seen in the case of ␣-neoendorphin (YGGFLRKY-Pro-Lys) (21) and ␤-neo-endorphin (YGGFLRKY-Pro) (22), whose precursor possesses the Lys-Arg basic pair followed by a C-terminal proline. It is thought that production of these peptides occurs through cleavage at the C terminus of paired basic Lys-Arg residues, followed by removal of the C-terminal basic residue by a carboxypeptidase B-like enzyme (23). Peptide bonds involving proline are resistant to common proteases such as this. However, the removal of C-terminal lysine in ␣-neo-endorphin occurs partially to generate ␤-neo-endorphin. In a similar manner, ghrelin-(1-27) may be produced by removal of the C-terminal arginine of ghrelin-(1-28) by a carboxypeptidase B-like enzyme. Although ghrelin-  was present only at a very low level in rat stomach, it is likely that these processing mechanisms control of the maturation of human ghrelin.
The human prepro-ghrelin we isolated is predicted to encoded a 117-residue precursor peptide. We previously reported that there are two types of ghrelin precursors from rat stomach cDNA analysis, a 117-amino acid precursor (prepro-ghrelin) and a 116-amino acid precursor (prepro-des-Gln 14 -ghrelin) (12). Des-Gln 14 -ghrelin, a splice variant of ghrelin, is the second endogenous ligand for the GHS-R. Only a small percentage of the ghrelin clones isolated from the human stomach library encoded the des-Gln 14 -ghrelin precursor, and the des-Gln 14 -ghrelin peptide was not identified in human stomach. The ratios observed between the two precursor populations, ghrelin and des-Gln 14ghrelin, was 5 to 1 in rat stomach, and 6 to 5 in mouse stomach (24). These differences are likely species-specific. Ghrelin was the first example discovered of a bioactive peptide modified by an n-octanoic acid moiety. Although acyl modification of many proteins has been observed, including G-proteins and some G-protein-coupled receptors, the modifications are most often myristoylations (25) and palmitoylations (26). In this study, we further showed that ghrelin can be modified by n-decanoic acid. All of the acyl-modified ghrelins and ghrelin-derived molecules studied here have the same potency to induce an increase of [Ca 2ϩ ] i in the GHS-R-expressing cells and stimulate GH release in anesthetized rats, and de novo synthesis of molecules concretely demonstrated that octanoic acid is not the only Ser 3 modification that will confer full activity to ghrelin (27,28). We recently reported the identification of bullfrog ghrelin acylated with n-decanoic acid and found that this ghrelin species comprises 33% of total isolated bullfrog ghrelin (29). In the human stomach, the ratio of octanoylated to decanoylated ghrelin was found to be roughly 3:1. Because acylation of ghrelin is essential for its activity, the enzyme that catalyzes this modification step should be an important regulator of ghrelin biosynthesis. However, the mechanism by which ghrelin is acylated during post-translational processing is still unclear.
We also analyzed the change in human plasma ir-ghrelin levels before and after total gastrectomy. We already reported that plasma levels of ir-ghrelin in totally gastrectomized patients were reduced to 35% of those in normal controls (14). In this study, C-RIA was used due to the instability of acylated ghrelin relative to its des-acylated counterpart, making its measurement from stored plasma samples unreliable. Plasma levels of ir-ghrelin were promptly reduced by approximately half within 30 min after total gastrectomy. Half-lives after intravenous administration of human ghrelin was about 10 min (30). Interestingly, the levels then began to increase in these cases, and in one patient even completely normalized. Significant amounts of ir-ghrelin were detected in the rat duodenum, jejunum, ileum, and colon (13), suggesting that these organs may be responsible for this compensation.
Further work will be required to determine the physiological significance of the various different forms of human ghrelin and ghrelin-related peptides discovered during the course of this study. Although all of these molecules displayed similar dose-response profiles from the tests performed in this study, it is possible that the various ghrelin forms have different signaling properties or stability.