Location of Disulfide Bonds within the Sequence of Human Serum Cholinesterase *

Human serum cholinesterase was digested with pepsin under conditions which left disulfide bonds intact. Peptides were isolated by high pressure liquid chromatography, and those containing disulfide bonds were identified by a color assay. Peptides were characterized by amino acid sequencing and composition analysis. Human serum cholinesterase contains 8 half-cystines in each subunit of 574 amino acids. Six of these form three internal disulfide bridges: between Cy@"Cys", Cy~~~"-Cys~'', and C y ~ " ~ C y s ~ ' ~ . A disulfide bond with Cyse6 rather than Cysee was inferred by homology with Torpedo acetylcholinesterase. Cys"" forms a disulfide bridge with Cys"'l of an identical subunit. This interchain disulfide bridge is four amino acids from the carboxyl terminus. A peptide containing the interchain disulfide is readily cleaved from cholinesterase by trypsin (Lockridge, O., and La Du, B. N. (1982) J. Biol. Chern. 267,12012-12018), suggesting that the carboxyl terminus is near the surface of the globular tetrameric protein. The disulfide bridges in human cholinesterase have exactly the same location as in Torpedo californica acetylcholinesterase. There is one potential free sulfhydryl in human cholinesterase at Cyse', but this sulfhydryl could not be alkylated. Comparison of human cholinesterase, and Torpedo and Drosophila acetylcholinesterases to the serine proteases suggests that the cholinesterases constitute a separate family of serine esterases, distinct from the trypsin family and from subtilisin.

by antibodies aroused suspicion that the amino acid sequences and the protein folding might be significantly different in the two kinds of cholinesterases. Therefore, it was a surprise to find that the amino acid sequence of human serum cholinesterase (7) is 53.8% identical with the amino acid sequence of acetylcholinesterase from Torpedo culifornica (8). The chain lengths are nearly the same at 574 and 575 amino acids/ subunit.
In this report we locate the disulfide bonds in human cholinesterase and compare the results with the disulfide bonds in Torpedo acetylcholinesterase (9). We found that the disulfide bonds in human cholinesterase are in the same location as in Torpedo acetylcholinesterase and have the same number of amino acids within each disulfide loop. This suggests that protein folding in the two enzymes is similar.
To explain the lack of recognition by antibodies, amino acid sequence differences could still be invoked. The human cholinesterase but not the human acetylcholinesterase sequence is known; the Torpedo acetylcholinesterase but not the Torpedo cholinesterase sequence is known. Therefore, it is not yet possible to compare the two enzymes from a single species. Drosophila acetylcholinesterase is the only other cholinesterase which has been sequenced to date (lo), but Drosophila has only one type of cholinesterase which has specificities intermediate between acetylcholinesterase and cholinesterase. Another factor which may affect antibody recognition is the number and location of carbohydrate chains. Human cholinesterase has nine asparagine-linked carbohydrate chains (7), whereas Torpedo acetylcholinesterase has four potential asparagine-linked carbohydrate chains (9), only two of which have the same location in cholinesterase.

EXPERIMENTAL PROCEDURES
Purification of Cholinesterase-Outdated human plasma was a gift from the Michigan Department of Public Health, Lansing, MI. 7.5 litera of plasma were used for each cholinesterase purification procedure, which consisted of three steps: ion-exchange chromatography at pH 4.0, followed by affinity chromatography on procainamide-Sepharose 4B, and finally by ion-exchange chromatography at pH 7.0 (3,11,12). The yield was 8-12 mg of highly purified cholinesterase.
Digestion with Pepsin-Cholinesterase disulfides were not reduced and alkylated prior to digestion with pepsin, because we wanted the disulfide bonds intact. 22 mg of cholinesterase in a volume of 6.5 ml was used for the digestion. The pH was adjusted to 1.3 by adding 1. 6 ml of 88% formic acid. Then, 0.5 mg of pepsin (porcine mucosa, Sigma No. P6887) was added. Digestion was at 37 "C for 48 h.
HPLC' Purification of Peptides-Pepsin-digested cholinesterase was injected onto a Synchropak RP-P reverse-phase column on a Varian model 5060 HPLC equipped with UV and fluorescence detectors. The Synchropak RP-P column, with its 300-A pore size, gave high recoveries of even large peptides. Elution was with 0.1% heptafluorobutyric acid and acetonitrile. Further purification was achieved by using a phenyl pBondapak reverse-phase column (Waters Co.) The abbreviation used is: HPLC, high pressure liquid chromatography or high pressure liquid chromatograph. eluted with 0.1% trifluoroacetic acid and acetonitrile, or the Synchropak RP-P column eluted with 0.1 M hexafluoroacetone-ammonia, pH 7.3, and acetonitrile. Hexafluoroacetone (Aldrich) is at present the only available neutral pH buffer that is volatile and UV transparent (13).
Identification of Disulfide-containing Peptides-Thannhauser et al, (14) have devised a rapid sensitive method for detecting disulfide bonds. 50-p1 aliquots of HPLC fractions were mixed with 120 pl of Thannhauser's cleavage buffer, incubated for 10 min, and then reacted with 5 pl of 25 mM disodium 2-nitro-5-thiosulfobenzoate assay solution. A yellow color indicated the presence of disulfide bonds.
Reduction and Alkylation with Vinylpyridine-Peptides containing a disulfide bond were dissolved in 0.5 ml of 6 M guanidine HCl (Pierce Chemical Co., Sequanal grade) containing 50 mM Tris-C1, pH 8.0, and 1 mM EDTA. Dithiothreitol was added to a final concentration of 26 mM. The solution was blanketed with nitrogen. After 1-4 h, 4 pl of 4-vinylpyridine (Aldrich) was added to a final concentration of 72 mM. Alkylation with vinylpyridine was allowed to proceed under nitrogen for 2-24 h before the sample was injected into the HPLC. Vinylpyridine was used because it gives a derivative that is readily detected during sequencing and amino acid analysis (15).
Amino Acid Sequencing-Peptides were sequenced by the manual method of Tarr (16). Tarr's modifications of the Edman degradation allowed sequencing 5 to 50 peptides at the same time using 200 pmol to 2 nmol of each peptide. Phenylthiohydantoins were identified on a Waters HPLC using an Ultrasphere ODS 5-pm column (Altex) at 50 "C. The column was equilibrated with 100 mM ammonium acetate, pH 4.5, containing 25% acetonitrile. Immediately after sample injection, the elution buffer was switched to 100 mM ammonium acetate, pH 4.5, containing 50% acetonitrile (17). One analysis was completed every 15 min.
Amino Acid Analysis-The University of Michigan Sequencing Facility under the direction of George Tarr performed amino acid composition analysis by the pico-tag method (15). Salt-free peptides were hydrolyzed for 4 h at 150 "C in 6 N HCI, the amino acids were derivatized with phenylisothiocyanate, and the phenylthiocarbamyl derivatives were identified by HPLC.
Attempted Alkylation of Free Sulfhydryl-1) 6.5 mg of cholinesterase in 1.5 ml of 6 M guanidine HC1,50 mM Tris-Cl,l mM EDTA, pH 8.4, was made anaerobic by blowing nitrogen for 1.5 h. 25 p1 of 6.66 mM [3H]iodoacetic acid (150 Ci/mol, Amersham Corp.) was added, and the reaction was allowed to proceed for 1 h. Dialysis was used to remove unbound iodoacetic acid. The stoichiometry of labeling was 0.22 mol bound per subunit. Tryptic peptides were prepared from the labeled protein and separated on HPLC. It was found that the radioactive counts were distributed throughout 80 ml of the chromatogram and were not associated with any particular peptide. 2) 5.5 mg of cholinesterase was digested with 0.11 mg of pepsin at pH 1.3 for 44 h at 37 "C. The volume was reduced under vacuum and the digest dissolved in 6 M guanidine HCI, 1 mM EDTA. The pH was adjusted to 7.5 by adding solid Tris base. 100 p1 of 6.66 mM [3H]iodoacetic acid was added. After overnight reaction, the entire sample was injected onto an HPLC column. All of the radioactivity eluted in the breakthrough volume between 2 and 10 min. No peptide appeared to be alkylated. 3) 4.8 mg of cholinesterase was incubated in 8 M urea, 0.1 M Tris-C1, 40 mM EDTA, pH 8.2, for 2 h at room temperature to denature the protein and then diluted to 2 M urea. One sample was digested with trypsin, another with both trypsin and S. aureus protease. 0.92 pmol of [3H]iodoacetic acid was added to each digest and allowed to react for 12 h. The entire sample was injected into the HPLC. The HPLC chromatogram showed that digestion was extensive. All of the radioactivity eluted in the breakthrough volume.
tional Biomedical Research Foundation, Georgetown University Computer Analysis-The Protein Identification Resource, Na-Medical Center, Washington, D. C. contains continuously updated databanks of protein and nucleic acid sequences in databanks NBRF, NEW, KABAT, JAPAN, PGTRANS, EMBL, and GENBANK. Cholinesterase was examined for the presence of kringle, finger, growth factor, catalytic, and vitamin K-dependent domains by use of these databanks. The ALIGN program was used to compare homologies of three cholinesterase sequences. The PRPLOT program was used to plot the hydropat.hy index.

RESULTS
Our strategy for identifying disulfide-linked peptides was based on knowledge of the complete amino acid sequence of cholinesterase (7) and on our observation that the protein contains no detectable free sulfhydryls. The protein was digested under conditions which left the disulfide bonds intact. HPLC fractions containing disulfide were identified by a color assay (14). After peptide purification and sequencing, a disulfide link could be inferred by reference to the complete amino acid sequence. To have additional proof of disulfide linkage the purified disulfide peptides were reduced, the products alkylated with vinylpyridine and rechromatographed on HPLC, and finally subjected to amino acid composition analysis. This additional proof was considered necessary because neither sequencing nor amino acid composition analysis detected cystine or nonalkylated cysteine. Finding the expected number of alkylated cysteines and the expected amino acid composition led to the conclusion that two cysteine residues were disulfide-linked. T h e possibility that peptides contained free sulfhydryls rather than disulfide bonds was tested by alkylation experiments with radiolabeled iodoacetic acid or with vinylpyridine. Pepsin was chosen for digestion of cholinesterase because the optimum pH for pepsin activity is p H 1-2. Disulfide bonds are stable at acid pH. Disulfide interchange, a possible occurrence in denatured protein at alkaline pH (18), was thereby avoided. We found that pepsin preferred to cleave at the carboxyl side of leucine and phenylalanine when digestion was carried out at pH 1.3. When digestion was at pH 2.0, pepsin was less specific. Fig.   1 shows HPLC separation of peptides produced by digesting cholinesterase with pepsin at pH 1.3. The Thannhauser color reaction (14) showed that disulfide-containing peptides eluted at 41, 43, 45, 47, 49, 51, and 59 min. Fractions were purified by additional H P L C runs a n d the identity of the peptides determined by amino acid sequencing. The HPLC fractions in Fig. 1 are mixtures of nondisulfide peptides as well as one or more disulfide-containing peptides. The four disulfide-containing peptides obtained in highest yield are discussed below. Other disulfidecontaining peptides were subfragments of the four peptides in Fig. 2.

FIG. 2. HPLC purification of disulfide-containing peptides and of the same peptides after reduction and alkylation. Panels A l , A 5 A3, and A4
show HPLC chromatograms which yielded purified disulfide-containing peptides. Panels B1, B2, B3, and B4 show HPLC chromatograms of the same peptides after reduction with dithiothreitol in the presence of 6 M guanidine chloride and alkylation. In BI alkylation was with iodoacetic acid. In B2, B3, and B4 alkylation was with vinylpyridine. The peptide peaks are Ib, IIb, IIIb, IIIc, and IVb; all other peaks in panel B are due to excess reagent and alkylation side products. In panel A the large peak between 2 and 5 min is formic acid, the solvent in which peptides were dissolved. HPLC for A I , BI, B2, B3, and B4 used a phenyl column equilibrated with solvent A (0.1% trifluoroacetic acid) and eluted with a gradient increasing in solvent B (acetonitrile containing 0.075% trifluoroacetic acid) at 1%/ min. For A2 the phenyl column was equilibrated with 80% solvent A, 20% solvent B, and eluted with a gradient increasing in solvent B at a rate of O.B%/min. For A3 and A4 a Synchropak RP-P column was equilibrated with 75% solvent C (0.1 M hexafluoroacetone/ammonia, pH 7.31, 25% acetonitrile, and eluted with a gradient increasing in acetonitrile at a rate of l%/min. (Table I). The identity of this peptide was determined by sequencing 26 cycles and by amino acid composition analysis of peptide Ib, the reduced and alkylated peptide. The amino acid Composition analysis (Table 11) agreed with the sequence shown, except that the yield of aspartic acid was low, a frequent result in the pico-tag method of amino acid analysis. Peptide Ia contains three cysteines. The disulfide bond was inferred to be between Cy@' and CysQz. The reason for choosing Cy@' rather than CysW is that Cysm is conserved in the amino acid sequences of Torpedo acetylcholinesterase and Drosophila acetylcholinesterase. cysW is present only in human cholinesterase (Fig. 3). Furthermore, Cys6' but not C y P , gives a disulfide loop exactly the same length as in Torpedo acetylcholinesterase.

Cys-Leu
Peptide IIa of Fig. 2 is Tyr-Glu-Ala-Arg-Asn-Arg-Thr-Leu-Asn-Leu-Ala-Lys-Leu-Thr-Gly-Cys-Ser-Arg-Glu-Asn-Glu-Thr-Glu-Ile-Ile-Lys-Cys-Leu. Its identity was determined by complete amino acid sequencing of IIa and amino acid composition analysis of IIb, the reduced and alkylated peptide. Amino acid composition analysis agreed with the sequence shown. The disulfide bond was between C~S~'~ and CyP3.
Peptide IVa of Fig. 2 contained Phe-Ile-Cys-Pro-Ala-Leu-Glu and Asn-Thr-Glu-Ser-Thr-Arg-Ile-Met-Thr-Lys-Leu-Arg-Ala-Gln-Gln-Cys-Arg-Phe-Trp-Thr-Ser-Phe linked via a disulfide bond. The sequence of both peptides was obtained simultaneously when the purified disulfide peptide IIIa was subjected to amino acid sequencing. After reduction and alkylation of IVa, the peptide separated into 11% and IIIc as shown in panel B3 of Fig. 2. Both fragments were sequenced and subjected to amino acid composition analysis. Results agreed with the sequence shown and led to the conclusion that the disulfide bond is between C y s ' @ ' and CysS1'.

Peptide IVa ofFig. 2 is Phe-Pro-Lys-Val-Leu-Glu-Met-Thr-Gly-Asn-Ile-Asp-Glu-Ala-Glu-Trp-Glu-Trp" Phe-His-Arg-Trp-Asn-Asn-Tyr-Met-Met-Asp-Trp-Lys-Asn-Gln-Phe-Asn-Asp-Tyl-Thr-Ser-Lys-Lys-Glu-Ser-Cys-Val-
Gly-Leu. The identity of this peptide was established by sequencing 16 cycles of IVa and by amino acid composition analysis of IVb, the reduced and alkylated peptide. The results supported the sequence shown. This peptide contains only one cysteine. For it to be disulfide-bonded it would have to be linked to an identical peptide from another subunit. To confirm that there was a disulfide bond between subunits and that these results were not due to the presence of a free sulfhydryl in this peptide, the following experiment was done. Peptide IVa was dissolved in 6 M guanidine, 0.1 M Tris-C1, pH 8.0,l mM EDTA, and allowed to react with vinylpyridine. It was then repurified on HPLC and subjected to amino acid composition analysis. The analysis showed zero vinylpyridinederivatized cysteine. When the same peptide was reduced with 26 mM dithiothreitol before being reacted with vinylpyridine, the amino acid composition analysis showed one vinylpyridine-derivatized cysteine. These results show that there was no free sulfhydryl in peptide IVa. We had established in earlier work that there was one disulfide bridge between two subunits and that this interchain disulfide was located very near the subunit terminus (3, 12). The disulfide bond in peptide IVa is consistent with these earlier results. We conclude that the disulfide bond is between e y~~~~ and Cys6" of two identical subunits.
Other disulfide-containing HPLC peaks, in addition to those in Fig. 2, were sequenced. These were subfragments produced by pepsin cleavage at additional sites. For example,

65-92
LGIPYAQPPLGRLRFKKPQSLTKWSDIWNATX   the disulfide peptide in panel A1 was isolated not only as the intact 65-residue peptide but also as the 60-residue peptide covalently attached to Ser-Glu-Asp-Cys-Leu. The peptide in panel A2 was also found as the subfragment Ala-Lys-Leu-

Thr-Gly-Cys-Ser-Arg-Glu-Asn-Glu-Thr-Glu-Ile-Ile-Lys-Cys-
Leu. The peptide pair in p a n e l A3 was also found as the subfragment Arg-Ala-Gln-Gln-Cys-Arg-Phe disulfide-bonded to Phe-Ile-Cys-Pro-Ala-Leu. The peptide in panel A4 was found not only as the intact 49-residue peptide but also as a disulfide-linked mixture whose amino acid sequence started with Phe-Pro-Lys-Val-Leu-Glu-Met and Met-Thr-Gly-Asn-Ile-Asp-Glu and Thr-Gly-Asn-Ile-Asp-Glu-Ala.
Residue 66 may be a free sulfhydryl. However, we were unable to specifically alkylate it with [3H]iodoacetic acid. Cysm was successfully alkylated only under conditions which alkylated all 8 cysteines, that is following reduction with dithiothreitol in the presence of 6 M guanidine chloride, 1 mM EDTA, pH 8.0.
Steric hindrance from the adjacent disulfide bond at position 65 could be a reason for lack of reactivity of the free sulfhydryl. To lessen this effect and make residue 66 more accessible, the protein was digested with various proteases before alkylation was attempted with radioactive iodoacetic acid. To test for incorporation of iodoacetic acid, the digests were chromatographed on HPLC and the HPLC fractions counted for radioactivity. No peptide was found to contain significant radioactivity. It is possible that the adjacent disulfide bridge blocked access to the sulfhydryl even in relatively short peptides.
Another explanation for the lack of reactivity of a free sulfhydryl could be that the free sulfhydryl is unstable once the protein is digested and that it becomes oxidized during the digestion period. With this possibility in mind, the protein was digested under conditions that stabilized free sulfhydryls, that is under a blanket of nitrogen, in the presence of 4 mM EDTA, for the short time of 4 h. [3HJIodoacetic acid was added at the end of the 4-h digestion period. The result was that even with these precautions, no alkylation was detected. Another possibility is that the free sulfhydryl became oxidized during protein purification or storage.
The possibility was considered that the free sulfhydryl was not free but was disulfide bonded to a collagen tail fragment or to some other peptide. NH-terminal sequencing of the intact protein showed only one sequence, whereas two sequences would have been expected if this explanation were correct. Peptide Ia of Fig. 2, which includes Cysa and Cys %, showed only one NHz-terminal sequence, whereas two NHzterminal sequences would have been expected if a peptide were disulfide-linked to Cysm. Amino acid composition analysis of peptide Ia revealed no unknswn peptide that might have been covalently attached to Cys". Chymotryptic as well as tryptic subfragments of this peptide were sequenced (7), and all subfragments were found to fit the known 574-residue sequence. In conclusion there was no evidence to suggest that Cys= is disulfide-linked to a collagen tail fragment or to any other peptide. Similarly, MacPhee-Quigley et al. (9) found no evidence for linkage to a collagen tail via C y~~~l , the free sulfhydryl of Torpedo acetylcholinesterase.
Other possible explanations for the lack of reactivity of Cys= are that Cys@ is disulfide-bonded to a low molecular weight sulfhydryl compound such as glutathione, or that it is in a thioester linkage, or that Cys" is a sulfenic or sulfinic acid (19). The possibility that Cys" is protected by ligation to metal is considered unlikely because EDTA was present in all alkylation experiments.

DISCUSSION
There are 8 cysteines/subunit in the complete amino acid sequence of human serum cholinesterase (7). The present results suggest that six of these cysteines are involved in forming three internal disulfide bridges within one subunit. The intrachain disulfide bonds are at C y~~-C y s '~ (Cys" rather

--Ju
Hum"n ChE   (7). T. californica and Drosophila melanogaster acetylcholinesterase (AChE) sequences were deduced from the nucleotide sequences of cDNA clones (8,10). Sequences start at the amino terminus for the human and Torpedo enzymes but include the signal peptide for the Drosophila enzyme. The residue numbers correspond to the sequence of human cholinesterase. The disulfide bonds are known for the human and Torpedo enzymes but are speculative for Drosophila acetylcholinesterase. The disulfide bridge at Cys&71-CysS7l is the interchain disulfide which covalently links two identical subunits. Human cholinesterase has nine glycosylated (CHO) asparagines. The number of glycosylated asparagines in Torpedo and Drosophila acetylcholinesterase could be four and five, respectively. Two histidines, His 423 and His 438 , are conserved in all three cholinesterases. The active site serine is Ser ' 98 • The sequences have been lined up to maximize homology with human cholinesterase. Torpedo acetylcholinesterase has 309 identical residues, and Drosophila acetylcholinesterase has 216 identical residues when compared to human cholinesterase. than Cys66 being inferred from homology with Torpedo acetylcholinesterase), and Cys252_Cys263, and Cys400_Cys619. A fourth disulfide bridge involves Cys571 which appears to be covalently attached to Cys671 of an identical subunit. One potential free sulfhydryl was suggested, though the inaccessibility of Cys66 to alkylation left open the possibility that it was not free.

S S K L P WP E WMG V i E I E F V F G L P L E R R O~Y T K A E E I L S R S I V K R WA N F A K Y GN P N E T Q N N S T S WP V F K S T E QK Y L T L N T E S T -R I M T K L R -A
Comparison with Disulfide Bonds of Torpedo Acetylcholinesterase-Acetylcholinesterase from the electric organ of T. californica has been cloned and sequenced (8), and its disulfide bonds have been established (9). Acetylcholinesterase has a chain length of 575 amino acids. It has 8 cysteines. Its disulfide bonds are at C ys67 -C ys 94, C ys254 _C ys266, C ys402 _C ys621 , and Cys572_Cys572. Cys672 is the interchain disulfide bond. The free sulfhydryl is Cys23l . Comparison of the disulfide bonding in acetylcholinesterase and cholinesterase shows that the disulfide bonds are in identical positions and involve the same cysteines (Fig. 3). Both proteins contain three internal disulfide loops of exactly the same length, and both are linked to an identical subunit via a cysteine that is 4 residues from the carboxyl terminus. The amino acid sequences are 53.8% identical. It is expected that the secondary and tertiary structures will have much in common.
One significant difference in the structures of these two proteins is in the number of carbohydrate chains. Human serum cholinesterase has nine asparagine-linked carbohydrate chains, at Asn", Asn 67, Asn 106, Asn U l , Asn 256, Asn 34" Asn 456, Asn 48 " and Asn 486 (7). Torpedo acetylcholinesterase has the possibility of four asparagine-linked carbohydrate chains, at Asn 59, Asn 416, Asn 467, and Asn 633 (9), where the sequence Asn-X-Thr/Seroccurs. This sequence is common to all N-glycosidically linked carbohydrate chains, though not every asparagine in such a sequence needs to be glycosylated (20). Cholinesterase and acetylcholinesterase have only two common glycosylation sites. Carbonhydrate chains are known to be directed toward the surface of the molecule (21). The differences in number and location of carbohydrate chains may account for some structural differences near the surface and may partly explain why antibodies to one enzyme do not recognize the other enzyme.
The comparison can be extended to bovine thyroglobulin because 544 amino acids at the carboxyl terminus of this protein have 28% sequence identity with both Torpedo acetylcholinesterase and human cholinesterase (7,8). cosyiation sites in this region, but neither site is present in the cholinesterases.
Comparison to Drosophila Acetylcholinesterase-The cDNA of Drosophila acetylcholinesterase, including the signal peptide, indicates a total of 11 cysteines (10). Six cysteines are located in positions which would allow the same three internal disulfide bridges in Drosophila that are present in human cholinesterase and in Torpedo acetylcholinesterase (Fig. 3). The carboxyl-terminal portion of Drosophila acetylcholinesterase has very little homology with the Torpedo and human enzymes, and, therefore, it is unclear whether either of the two cysteines in that region is likely to have the function of linking two subunits. Drosophila acetylcholinesterase has five potential N-glycosylation sites, none of which is in exactly the same position as in the Torpedo or the human enzymes.
Disulfide at Cy~~~~--Identification of CysS7l as the interchain disulfide is supported by earlier work where we had shown that there was one disulfide bond between two subunits (3) and that this interchain disulfide is located very near the subunit terminus (12). At that time we did not know whether this disulfide is near the amino or the carboxyl terminus. Cyd7' is 4 residues away from Le~''~at the carboxyl terminus. Our earlier work showed that a peptide containing the interchain disulfide was easily removed by trypsin digestion. There are two lysines near the carboxyl terminus. Cleavage by trypsin at either Lys567 or L Y S~~ would remove 800-900 daltons and would explain why the cleaved subunit has the same apparent subunit weight of 85,000-90,000 as the intact subunit. Cholinesterase has a broad band on sodium dodecyl sulfate gels, probably because of its nine carbohydrate chains. The broadness of the band does not allow discrimination of differences less than 5,000 daltons. Anglister et al. (22) also concluded that the interchain disulfide bridge for eel acetylcholinesterase was at one end of the catalytic subunit. For both eel acetylcholinesterase and human cholinesterase it has been observed that the interchain disulfide bridges are not essential for maintenance of quaternary structure under nondenaturing conditions.

SOQ 6on
The finding that the interchain disulfide is at C y P supports our earlier interpretation regarding the sodium dodecyl sulfate gel electrophoresis pattern of purified cholinesterase preparations. A preparation that has both monomer and dimer bands in the absence of reducing agent can be concluded to have experienced some proteolysis during the purification procedure. A preparation that has only a dimer-sized band of approximately 170,OOO-180,000 daltons in the absence of reducing agent can be concluded to be free of proteolytic cleavage. The effect of proteolysis is similar to the effect of mercaptoethanol insofar as both yield a monomer-sized band on sodium dodecyl sulfate gel.
Hydropathy Index-The hydropathy (23) profiles for human cholinesterase and Torpedo acetylcholinesterase are similar in Fig. 4, suggesting that folding in the two proteins is similar. The region from residue 538 to 569 falls below the midpoint line and is, therefore, predicted to be on the exterior in both human cholinesterase and Torpedo acetylcholinesterase. This prediction agrees with the results of limited proteolysis discussed above. The evidence strongly supports the conclusion that the subunits in the globular tetrameric serum cholinesterase protein are arranged in such a way that the carboxyl ends are near the surface of the molecule.
The hydropathy profile of Drosophila acetylcholinesterase has less resemblance to those of the other two cholinesterases. Two regions stand out as different: the region from residue 107 to 140, and the region at the carboxyl terminus. The 33residue peptide starting at 107 is a nonhomologous extra peptide (see Fig. 3), and when it is omitted from the hydropathy figure the profile becomes more similar to the others. This peptide is below the midpoint line in Fig. 4 and, therefore, is likely to be near the enzyme surface. The carboxyl terminus of Drosophila acetylcholinesterase has a prominent peak above the line which contrasts strongly with the carboxyl-terminal profiles of the other two cholinesterases. The hydropathy profile suggests that the carboxyl terminus of the Drosophila enzyme is buried.
Comparison to Serine Proteases-The trypsin family of serine proteases includes trypsin, chymotrypsin, the blood coagulation factors, and other hydrolytic enzymes (24). The serine proteases are irreversibly inhibited by organophosphate esters such as diisopropyl fluorophosphate, due to acylation of the active site serine (25). Another common feature is the charge relay system serine, aspartic acid, histidine (26 Subtilisin is a bacterial serine protease with a different three-dimensional structure than the trypsin family of serine proteases (28). Though subtilisin has serine at the active site and has a charge relay system, the aspartic acid of the charge relay system is at a different location in the sequence. Subtilisin has no disulfide bonds.
The cholinesterases are similar to the serine proteases. They have serine at the active site, and this serine is irreversibly labeled by diisopropyl fluorophosphate. A charge relay system may exist (29), since histidine appears to be essential for catalysis (30, 31). However, neither aspartic acid nor histidine of such a possible charge relay system has been unequivocally labeled or identified.
Unlike the serine proteases, the cholinesterases are primarily esterases. There is controversy regarding the possibility that cholinesterases may also be peptidases (32-39). If cholinesterases hydrolyze esters exclusively, then this would clearly distinguish them from the serine proteases, because the latter hydrolyze esters, peptides, and proteins. Cholinesterases can also be distinguished from the serine proteases by the chemical nature of the stable serine derivative obtained after reaction with diisopropyl fluorophosphate (40). The cholinesterases initially form diisopropyl phosphorylserine, but this quickly ages to yield the monoisopropyl phosphorylserine (41). In contrast, the serine proteases form only the diisopropyl phosphorylserine derivative.
The serine protease family and the cholinesterase family also show structural differences. Fig. 5 compares the locations of the active site serine, the charge relay aspartic acid and histidine, and the disulfide bonds. When these proteins are lined up so that the active site serines are on the same line, it becomes apparent that the cholinesterases differ from the serine proteases. The active site serine in the cholinesterases is closer to the amino terminus than the carboxyl terminus.
A charge relay histidine in cholinesterase is likely to involve His423 or His438, because these are the only conserved histidines in the three cholinesterases. This location for histidine is very different from the charge relay His6' of chymotrypsinogen. The three cholinesterases have a conserved aspartic acid at Asp' l near the charge relay Asp'02 of chymotrypsinogen. This makes Aspg1 a candidate for the charge relay aspartic acid, but there are five other conserved aspartic acids that need also be considered for this function.

Cholinesterase Disulfides
At the domain level there are also significant structural differences. The amino acid sequence of human serum cholinesterase was examined for the presence of kringle, finger, growth factor, and vitamin K-dependent domains. Tissue plasminogen activator, prothrombin, and urokinase were used for comparison. Cholinesterase has no significant sequence homology with any of these domains nor is any significant sequence homology found around the active site serine. Furthermore, the pattern of disulfide bonding in cholinesterase does not resemble the pattern in these serine proteases.
These findings lead to the conclusion that the cholinesterases constitute a separate family of serine esterases that differs from the trypsin family of serine proteases and from subtilisin.