Inr. J . Pepride Protein Res. 30, 1987, 808-821

Propensity for spontaneous succinimide formation from aspartyl and asparaginyl residues in cellular proteins

STEVEN CLARKE

Department of Chemistry and Biochemistry and the Molecular Biology Institute, University of California, Los Angeles. CA. USA

Received 31 March, accepted for publication 1 July 1987

One mechanism for the spontaneous degradation of polypeptides is the intramole- cular attack of the peptide bond nitrogen on the side chain carbonyl carbon atom of aspartic acid and asparagine residues. This reaction results in the formation of succinimide derivatives and has been shown to be largely responsible for the racemiza- tion, isomerization, and deamidation of these residues in several peptides under physiological conditions (Geiger, T. & Clarke, S . J . Biol. Chem. 262,785-794 (1987)). To determine if similar reactions might occur in proteins, I examined the sequence and conformation about aspartic acid and asparagine residues in a sample of stable, well-characterized proteins. There did not appear to be any large bias against dipep- tide sequences that readily form succinimides in small peptides. However, it was found that aspartyl and asparaginyl residues generally exist in native proteins in conforma- tions where the peptide bond nitrogen atom cannot approach the side chain carbonyl carbon to form a succinimide ring. These orientations also represent energy minimum states, and it appears that this factor may account for a low rate of spontaneous damage to proteins by succinimide-linked reactions. The presence of aspartic acid and asparagine residues in other conformations, such as those in partially denatured, conformationally flexible regions, may lead to more rapid succinimide formation and contribute to the degradation of the molecule. The possible role of isoimide inter- mediates, formed by the attack of the peptide oxygen atom on the side chain carboxyl group, in protein racemization, isomerization, and deamidation is also considered.

Key w o r d : asparagine; aspartic acid; deamidation; isomerization; succinimide formation

A class of enzymes has recently been isomerized L-aspartyl residues and the beta- described that catalyzes the S-adenosyl- carboxyl group of D-aspartyl residues, but methionine-dependent methyl esterification not the side chain carboxyl of normal L- of the carboxyl groups of abnormal aspartic aspartyl residues. The exact origin of these acid residues in proteins (for a review, see ref. residues, which occur normally at low 1). These reactions apparently involve the frequencies, is not clear but two possibilities recognition of the alpha-carboxyl group of have been suggested. In the first place, altered

808

Succinimide formation in proteins

specific three-dimensional conformations. Initial studies with synthetic peptides indi- cated that succinimides are key intermediates in the deamidation of asparagine residues, and the isomerization and racemization of both asparagine and aspartyl residues (6). These studies also established the magnitude of the difference in the rates of succinimide formation from homologous aspartyl and asparaginyl sequences and confirmed the influence of the side chain of the following residue upon these rates (12-14). However, the expected conformational flexibility of these peptides would not allow direct comparisons to be made for the rate of succinimide formation in intact proteins where the atoms may not be properly aligned for this reaction (Fig. 2).

The purpose of this study was to examine the sequences and conformations of aspartyl and asparaginyl residues in a group of proteins with known three-dimensional struc- tures. The results of this work suggest that the conformational restrictions on succinimide formation may be very important in proteins and it may be difficult to predict the rate of succinimide formation from a knowledge of the sequence alone.

aspartyl residues may be produced during protein synthesis itself. For example, if the tRNA molecule is attached to the side chain beta-carboxyl group instead of the alpha- carboxyl group, an isoaspartyl residue would result. Similarly, if racemization, a very common side-reaction in chemical synthesis, occurs, D-aspartyl residues may be produced. The second possibility is that these altered residues are formed by spontaneous chemical processes during the aging of the protein. Although one can explain the racemization of aspartyl residues by simple acid-base chemis- try at the alpha-carbon (2), more complicated chemistry, generally involving a succinimide intermediate, is required to explain the formation of isoaspartyl residues by a spon- taneous mechanism (3-7; cf. Fig. 1). It has been suggested that the physiological role of the enzymatic methylation reaction is the recognition of these damaged sites for either proteolysis or repair (8-1 1).

In an attempt to predict the sites of deamidation, racemization, and isomeriza- tion in proteins, I was interested in examining the factors that control the rate of succini- mide formation from protein sequences con- taining aspartyl and asparaginyl residues in

-L-Asn-Gly-

I tH.1 4 d

-L-Asp-Gly- -L-Asp&ccinirnidy I-Gly -

FIGURE 1 Spontaneous formation of peptide succinimides and hydrolysis products from aspartyl and asparaginyl residues. The half times for

-D-Asp-succinimidyl-Gly- the reactions shown are given for the model peptides Val-Tyr-Pro- Asn-Gly- Ala

t,,2= 0.19d and Val-Tyr-Pro-Asp-Gly- Ala at 37", pH 7.4 (6). When the glycine residue is

0 73% 27% 73% - 27% replaced in the asparagine- n - L - A ~ ~ - G ~ ~ - - D - i s o ~ s p - ~ l y - - D - A ~ ~ - G ~ ~ - containing peptide by a

C H ; " N H ~ ~ ? / leucine residue, the rate of imide formation is I

c*O approximately 50 times XNH/CH. ,oe 0

-L- ~soAsp-Gly- slower (6).

809

S. Clarke

Optimal conformation for lmlde formation

* = -120" x , = 120'

lrnide

1328) HdkX1 N- (N-Cdistance / H-- -C+C%

N b o /

Least optimal conformation

p-60" ,

c3 No Reoction

H

FIGURE 2 Productive and non-productive conformations about asparaginyl residues for succinimide formation. When the dihedral angle psi is - 120" and the angle chi, is + IZO", the nitrogen atom of the peptide bond makes its closest approach to the gamma carbonyl atom (1.89A) and succinimide formation is favored. On the other hand, when psi is +60" and chi, is -W, these atoms are 4.89A apart and no reaction is possible. Although the peptide bond nitrogen is shown in the protonated form, it is very likely that the reactive species is actually the deprotonated nitrogen anion (7). It should also be noted that the attack on the gamma carbonyl group would be expected to be directed from an angle perpendicular to the plane of the amide group; this would require that the chi, dihedral angle be either + W or -90" (see text). Similar considerations would apply for succinimide formation from an aspartyl residue; in this case it would be expected that the reactive species would be the carboxylic acid rather than the carboxylate anion.

METHODS

Selection of sample of intracellular and extracellular proteins for conformational analysis A representative sample of vertebrate proteins with well defined three-dimensional structures were selected from the September 1984 version of the Brookhaven Protein Data

Bank (1 5 ) (Table 1). The selection was limited to vertebrates because it is in these organisms that most of the work on protein racemiza- tion, isomerization, and carboxyl methyla- tion of abnormal proteins has been per- formed. For comparative purposes, the sample was divided into intracellular proteins, defined by their localization in the cytosolic fraction of the cell, and extracellular proteins. This latter class includes secreted proteins that function outside the cell cytosol. Because there is little information on the presence of protein carboxyl methyltrans- ferases in the various intracellular compart- ments of the cell (mitochondria, nucleus, lys- osomes, etc.), proteins in these fractions were not considered in this study. The six intra- cellular proteins chosen contained a total of 61 asparagine and 98 aspartyl residues out of a total of 1654 residues. The eight extra- cellular proteins contained lo0 asparagine re- sidues and 68 aspartyl residues out of a total of 1625 residues (Table 1). The relative prevalence of aspartyl residues in intracell- ular proteins and asparaginyl residues in extracellular proteins has been noted previously in an examination of a larger sample of proteins (16).

Extraction of dihedral angles from the Brookhaven Data Bank The three-dimensional positions of each of the atoms of the proteins in the sample in Table 1 are given in the Brookhaven Data Bank as xyz coordinates. These coordinates were used to calculate the dihedral angles phi, psi, chi,, and chi, (17). For this analysis, a modification of the GEOM78 program (Cambridge Crystallographic Data Center; ref. 18) was written by Dr. Pinak Chakrabarti (Department of Chemistry and Biochemistry, UCLA) for use with a VAX 1 1-780 computer.

Calculation of atomic distances from dihedral angles The distance from the peptide nitrogen atom to the side chain gamma carbonyl carbon was calculated from the values of the psi and chi, dihedral angles assuming standard bond lengths and geometry.

810

Succinimide formation in proteins

TABLE I Sample of intracellular and extracellular proteins

Protein source Brookhaven Number of file residues

In tracellular Adenylate kinase Carbonic anhydrase D-Glyceraldehyde

Hemoglobin (deoxy)

Lactate dehydrogenase

Triose phosphate isomerase

Extracellular Carboxypeptidase A Alpha chymotrypsin Immunoglobin Fab

Immunoglobin Fc LYmzYme Ovomucoid, third

domain Phospholipase A, Ribonuclease A

phosphate dehydrogenase

Porcine muscle Human erythrocyte. Human muscle

Human erythrocyte

Dogfish muscle

Chicken muscle

Bovine pancreas Bovine pancreas Human plasma

Human plasma Hen egg white Hen egg white

Bovine pancreas Bovine pancreas

2ADK 2CAB 3GPD

4HHB

4LDH

lTIM

5CPA 2CHA 3FAB

1 FCI 6LYZ 1 ovo 1 BP2 4RSA

194 260 334

141 (A chain) 146 (B chain)

33 1

248

307 245

214 (L chain) 220 (H chain)

207 129 56

123 124

RESULTS AND DISCUSSION

Spontaneous succinimide formation from aspartyl and asparaginyl residues and the subsequent formation of deamidated, race- mized, and isomerized residues would be expected to be detrimental to protein function. Although it is possible that such damage in intracellular proteins may be par- tially repaired by the D-aspartyl/L-isoaspartyl protein methyltransferase system (9-1 I), neither the methyltransferase (19) nor its substrate, S-adenosylmethionine (20), is present in the extracellular plasma and damage to proteins not facing the cytosol may be essentially irreversible.

From what is known about the formation of succinimides in peptides (5-7,9-10, 13-14, 21-22), there would appear to be at least three stages at which cells may design protein struc- tures that would minimize this side reaction. The simplest mechanism is to avoid using

aspartyl and asparaginyl residues in proteins. Secondly, the rate of succinimide formation appears to depend upon the nature of the residue on the carboxyl side of the aspartyl or asparaginyl residue. While small, or poten- tially reactive, residues such as glycine or serine lead to increased rates, the inclusion of a bulky residue slows succinimide formation. Finally, the three-dimensional conformation of the aspartyl or asparagine residue can be positioned so that attack of the peptide bond nitrogen atom on the side chain carbonyl to form a succinimide is not possible. In this paper, I will focus on these three possibilities in turn.

Are asparaginyl and aspartyl residues selected against in intracellular and extracellular proteins? Of the total of 64 messenger RNA codons for amino acid incorporation into proteins, two code for aspartic acid and two code for asparagine. If these codons are

811

S. Clarke

utilized on a random basis, we would expect about 2 aspartyl residues and 2 asparaginyl residues per 64 residues, or about 3.1 mol% of each amino acid in proteins. Any selection against these residues might be reflected in values below this figure, or in increased utilization of the homologous residues glutamic acid and glutamhe*. Nevertheless, previous data has shown no evidence for the exclusion of asparagine or aspartyl residues from proteins. For example, in a study of 28 intracellular proteins, the mole fraction of asparagine residues was found to be actually greater than that of glutamine residues (3.4% versus 3.0%), and a similar situation oc- curred with aspartyl and glutamyl residues (6.3% versus 6.1%). For a sample of 39 extracellular proteins, comparable results were found (asparagine, 5.8%; glutamine, 4.1%; aspartic acid, 5.0%; glutamic acid, 3.1%) (15). In all cases, the fraction of aspartyl and asparaginyl residues was greater than expected from codon assignments. Furthermore, the proportion of the more succinimide-prone asparagine residue was actually greater in the extracellular proteins than in the cytoplasmic proteins. It thus appears that the importance of both aspartyl and asparaginyl residues for catalytic and/or structural roles in proteins override the potential for age-dependent damage at these sites.

Are speciJic sequences that have more potential to form succinimides avoided in intracellular and extracellular proteins? Previous work with denatured proteins and synthetic peptides has indicated that the rate of succini- mide formation is dependent upon the amino acid sequence. Under physiological condi- tions, asparagine residues in one peptide sequence have been found to be approximate- ly 40 times more susceptible to this reaction that the corresponding sequence containing an aspartyl residue (6). The presence of residue with a large unreactive side chain on

*Although analogous reactions can generate glutarimide residues from glutamic acid and glutamine residues, the rate of these reactions are much less than those form succinimides (cf. ref. 13).

the carboxyl side of aspartic acid and asparagine residues and their derivatives can slow imide formation and the presence of a small residue with hydrogen bonding poten- tial in the side chain (such as serine) can enhance the reaction (5-7, 12-14, 22). These results would suggest that imides would be expected to form fastest with AsnGly, AsnAla, or AsnSer sequences and slowest with AspLeu or AspTrp**. To ask whether there might be fewer of the former dipeptides in proteins than the latter, I compared the actual frequency of these and other dipep- tides with the distribution expected from the amino acid composition alone (Table 2). The data shown here clearly indicate that the fre- quency of aspartyl dipeptides, both in intra- cellular and extracellular proteins, is not sig- nificantly different from what one would expect from the bulk amino acid com- position. This was also found to be the case for asparagine residues in the sample of intra- cellular proteins. However, the situation with asparagine residues in extracellular proteins, where the succinimide reaction products are not metabolized by protein carboxyl methyl- transferases, is more complex. Although the sample size is limited, it does appear that the frequency of dipeptides in extracellular proteins where a bulky residue follows is higher than expected and the frequency of asparagine dipeptides followed by a small residue is lower than expected (Table 3). These would be the results expected if there was a selection in extracellular proteins against succinimide formation. When this type of analysis was carried out on a larger collection of 16 intracellular and 24 extra- cellular proteins from bacteria, plants, and

+*It is interesting to note, however, that the major site of L-isoaspartyl formation in glucagon is from an aspartyl- tyrosine sequence (21). Since this would appear to be a much less likely site for succinimide formation than the asparaginyl-threonine or aspartyl-serine residues present in this hormone, other factors may be important in suc- cinimide formation. In this case, neighboring groups may participate as either general bases to deprotonate the peptide nitrogen or as general acids to activate the side chain carbonyl group of the aspartyl residue (cf. refs. 7, 14, 22).

812

Succinimide formation in proteins

TABLE 2 Occurrence of aspartyl and asparaginyl dipeptides in intracellular and extracellular proteins

The number of aspartyl and asparaginyl dipeptides in the sample of proteins in Table 1 is compared with the number expected (in parentheses) on the basis of the amino acid composition. These latter values were calculated by multiplying the total number of aspartic acid and asparagine residues in the intracellular or extracellular group of proteins given in Table 1 by the mole fraction of the second amino acid in a representative sample of intracellular

or extracellular proteins (16)'

First residue Aspartyl Asparaginyl Second

residue Intracellular Extracellular Intracellular Extracellular

10 (8.1) 12 (9.9) 6 (6.2) 1 (1.9) 5 (4.8) 6 (7.4) 6 (6.2) 7 (3.2) 2 (1.6) 9 (7.7) 7 (4.9) 5 (6.0) 0 (2.9) 7 (7.8) 4 (4.1) 2 (2.7) 1 (2.3) 1 (2.6) 2 (1.5) 5 (4.2)

7 (7.1) 4 (5.8) 7 (6.5) 2 (2.4) 5 (5.2) 5 (5.0) 3 (3.4) 4 (3.9) 0 (0.8) 3 (4.2) 4 (3.3) 2 (2.1) 1 (2.8) 5 (3.3) 5 (2.1) 0 (1.0) 4 (2.9) 5 (2.0) l(1.0) 1 (2.7)

4 (5.1) 9 (6.2) 3 (3.8) 3 (1.2) 0 (3.0) 8 (4.6) 3 (3.8) 2 (2.0) 0 (1.0) 5 (4.8) 4 (3.1) 4 (3.7) 2 (1.8) 2 (4.9) 4 (2.6) 2 (1.7) 0 (1.4) l(1.6) 1 (0.9) 4 (2.6)

7 (10.1) 6 (8.6) 5 (9.6) 7 (3.6) 8 (7.6) 3 (7.3) 2 (5.0) 9 (5.8) 0 (1.2) 4 (6.2) 2 (4.8) 2 (3.1) 5 (4.1)

3 (3.1) 5 (1.5) 7 (4.3) 5 (2.9) 7 (1.5) 4 (3.9)

7 (4.8)

GlY Ala Ser CYS Thr Val ASP Asn Met Leu Ile

Glu Gln LYS Phe His TYr Are Trp Pro

TABLE 3 Effect of the size of the carboxyl residue side chain on the frequency of aspartyl and asparagine dipeptides

The number of dipeptides is taken for either the sample of vertebrate proteins in Table 1 or for a larger set of proteins from a wider range of organisms (asterisks)'. The number in parentheses is the number predicted from the average

amino acid composition of extracellular and intracellular proteins (see Table 2)

Asparagine dipeptides In tracellular

Extracellular

Aspartyl dipeptides Intracellular

16 (15.1) 32 (39.6). 18 (28.6) 66 (78.7)*

28 (24.2) 70 (71.7)* 18 (19.4) 90 (72.4).

6 (5.5) 12 (14.9). 24 (10.2) 31 (28.0).

6 (9.1) 25 (27.0)* 10 (6.9) 26 (25.8)*

Extracellular

813

S. Clarke

animals', a similar distribution was found, although there was no apparent selection against asparagine-small residue dipeptides or selection for asparagine-large residue dipeptides. All in all, these results indicate that there is no large bias in proteins against dipeptide sequences that can readily form succinimides in small peptides.

Are three-dimensional conformations of proteins that allow succinimide formation at aspartyl and asparaginyl residues selected against? The attack of the peptide nitrogen atom on the gamma carbonyl group of a peptide asparagine or aspartic acid side chain requires that the dihedral angles psi and chi,, which define the rotation around the alpha- carbon/peptide carbonyl carbon and the alpha carbon/beta carbon, approach values of about - 120" and 120" respectively (Fig. 2). In addition, the value of chi,, defining the rotation of the acid or amide group, should be about 90" or -90" to minimize steric repulsion from the oxygen and nitrogen side chain atoms. Although there is relatively free rotation around these bonds in small peptides in solution, in proteins the overall structure of the main polypeptide chain constrains the psi value and the generally fixed position of the side chains similarly constrains the chi, and chi, values.

It has been well established that main chain conformations that are conducive to suc- cinimide attack (psi = - 120") are relatively rare in proteins (for a review see ref. 23). However, there is at least one family of regular secondary structure, type 11' beta-

'Data was taken from the Brookhaven Protein Data Bank. The intracellular proteins included liver alcohol dehydrogenase, cytochrome b,, cytochrome c, dihy- drofolate reductase, ferrodoxin, glutathione reductase, glyceraldehyde phosphate dehydrogenase, hemoglobin, myoglobin, phosphoglycerate mutase, rhodanase, thioredoxin reductase, lactate dehydrogenase, superoxide dismutase, carbonic anhydrase C, and triose phosphate isomerase. The extracellular proteins included trypsin, elastase, proteinase A and B, acid proteinase, prealbumin, themolysin, subtilisin, Staphylococcal nuclease, ribonuclease A, phospholipase A,, papain, lysozyme, insulin, immunoglobin, concanavalin A, alpha-bungarotoxin, arabinose binding protein, wheat germ agglutinin, and actinidin.

814

hairpin turns, where this conformation does occur for the residue in the second position (24). Although aspartyl and asparaginyl residues, often followed by a glycyl residue, commonly occur at this position in other types of beta-hairpin turns, these residues have not been found to date in this specific type of turn structure (Table 4). For example, Asn-Gly or Asp-Gly sequences occur at high frequency in type I' turns where the psi angle for the aspartyl/asparaginyl residue is 30" and succinimide formation is not favored (Table 4). These results suggest that aspartyl and asparaginyl residues in intact proteins, par- ticularly those followed by small residues such as glycine, may be protected from suc- cinimide formation by exclusion from three- dimensional structures where imide forma- tion is possible.

Considering these results in light of the previous proposal that the bulk of protein succinimide formation may occur at Asn-Gly sequences (25, 26), I examined the conforma- tion of each of the chemically susceptible Asn-Gly sequences in the proteins shown in Table 1 (Table 5). With the exception of Asn-384 in the immunoglobulin Fc region, it was apparent that imide formation at any of these sites would require major conforma- tional changes in both the main chain and the side chain. For optimal imide formation, an average change of 124" in the rotation of the main chain psi angle and 112" in the side chain chi, angle would be required. These results suggest that intracellular and extracellular Asn-Gly sequences may be generally protected against succinimide formation by having three-dimensional con- formations where attack of the peptide nitrogen cannot take place without large- scale conformational change, such as would occur with protein denaturation. Therefore, it may not be possible to predict the sites of succinimide formation in folded proteins on the basis of the rate of succinimide formation in small peptides.

Fig. 3 shows the psi and chi, dihedral angles for the complete set of Asp and Asn residues in the proteins listed in Table I . The results are plotted on contour maps showing the distance between the peptide nitrogen and

Succinimide formation in proteins

TABLE 4 Presence of aspartyl and asparaginyl residues in two types of beta-hairpin turns"

Turn type Total sequences Sequences with Asp or Asn in position 2

Type 11' beta-hairpin turns (psidd,, = - 120"; favorable for imide formation)

Type I' beta-hairpin turns (psi,,, = +30"; unfavorable for imide formation)

10 0

15 7b

'Data from Sibanda & Thornton (24) bThese seven sequences include three Asn-Gly sequences in elastase, chymotrypsin, and Streptomyces griseus protease A; three Asp-Gly sequences in Staphyloccocal nuclease and Dglyceraldehyde phosphate dehydrogenase, and one Asn-Asn sequence in phospholipase A,.

gamma carbonyl carbon atom as an indica- residues are in ideal positions for imide tion of the propensity to succinimide forma- formation and any such reaction, as with the tion (cf. Fig. 2). The most striking feature of Asn-Gly sequences discussed above, would these maps is that essentially none of these require displacements of both the main chain

A. lntracellular Asn 9+ g- t

ASP 1 I I

120 120 3. FIGURE3 Distribution of dihedral angles psi and chi, for asparagine and aspartic acid residues in the proteins listed in Table 1. The contours indicate distances

-120 -120 in angstroms between the peptide bond nitrogen and the carbonyl carbon atom (cf. Fig. 2). The area shaded represents distances equal to or less than the Van der

Asn Waals separation of 3.0A between these atoms. Circled points represent

120 120 3' Asn-Gly and Asp-Gly sequences. The approximate range of psi values for alpha helix and beta sheet struc- tures are indicated, along with the chi, values corre- sponding to g+, g-, and t configurations. A. Data for intracellular proteins. B.

-120 0 120 -120 0 120 Data for extracellular

0

X1 B. Extracellular

0

la -120 -120

Xl proteins.

815

S. Clarke

TABLE 5 Conformation of asparagine residues in Asn-Gly sequences: How much bond rotation is required for succinimide

formation?

Protein Phi Psi Chi, Chi, Distance Rotation sequence N to Cgamma needed for

(A) imide formation Psi, Chi,, Chi,

A ) Intracellular Proteins Triose phosphate isomerase Lys-Met-Asn,,-Gly-Lys-Arg

- 145 - 174 Glyceraldehyde 3-phosphate dehydrogenase

Asp-Phe-Asn,,, -Gly-Ser- Asn 55 77

Carbonic anhydrase B Asp-Lys-Asn,, -Gly-Pro-Glu

Ile-Ala-Asn,, -Gly-Asn-Asn - 139 16

- 97 13

B ) Extracellular Carboxypeptidase A

Asn-Pro-Am, ,, -Gly-Phe-Ala - 68 - 47

Ovomucoid third domain Glu-Ser- Asn,, -Gly-Thr- Leu

26 64 Phospholipase A,

Gln-Phe-Am,-Gly-Met-Ile - 62 - 38

Ribonuclease A Cys-Lys-Asn,, -Gly-Gln-Thr

- 84 5 Alpha chymotrypsin

Ile-Val- Asn,, -Gly-Glu-Glu

Lys-Lys-Asn,, -Gly-Ala-Trp 65 31

45 31 Immunoglobulin Fc

Glu-Ser- Asn,, -Gly-Gln-Pro - 160 - 123

- 88

- 140

64

- 58

- 67

- 80

167

79

- loo

- 58

- 165

79

127

11

133

171

- 172

- 106

- 18

- 70

3

22

4.07

4.36

3.66

4.77

4.47

4.85

3.56

3.43

4.84

4.84

2.91

54". 152", I I "

163", loo", 37"

136", 56", 79"

133", 178", 43"

73", 173", 81"

176", 160", 82"

82". 47", 16"

125", 41", 72"

151". 140". 20"

151", 178", 87"

3". 75". 68"

and side chain conformations. However, it is possible to determine from this type of analy- sis which residues would require the least conformational change in order to form an imide. The aspartyl and asparaginyl residues in conformations where the peptide nitrogen atom is within 3 A of the side chain carbonyl carbon atom are shown in Table 6. In these cases, the reactive atoms approach their van der Waals separation (27), and these residues might be the best candidates for intramole-

816

cular imide formation. The presence of several of these sites in the constant region of immunoglobulins may reflect the effective- ness of this protein as a methyl-acceptor for the L-isoaspartyl/D-aspartyl methyltrans- ferase (4). It should be noted, however, that the class of residues where the distance is less than 3 A represents only 13 of the 327 aspartyl and asparaginyl residues in the sample tested.

To determine if these results indicate that

Succinimide formation in proteins

TABLE 6 Aspartyl and asparaginyl residues in conformations closest to that required for succinimide formation'

Protein sequence

Psi Chi, Chi, Distance Rotation N to C,,, needed for

(A) imide formation Psi, Chi,, Chi,

D-Glyceraldehyde phosphate dehydrogenase Asn,,,-His - 55 88 Asp,,-Thr - 77 74

Asp,, -Val I57 - 180

Asn,,,-Glu - 26 101

AsR,-Ile -40 109 Asp,,,-Asn - 62 150

ASP,, -LYS - 142 60

(light) Am,,-Ile - 125 50 (heavy) Asb-Leu 168 - 174

Asn,,, -Glu - 117 48 Lactate dehydrogenase

Ash-Tyr - 140 98

Triose phosphate isomerase

Alpha chymotrypsin

Immunoglobin Fab

Immunoglobin Fc Am,,-Gly - 123 - 165 Asp,,-Ser - 179 - 165

- 102 -91

96

86 114 - 78

- 69 - 150

- 151

- 77 - 94

22 26

2.47 2.33 2.85

2.92 2.20 2.96

2.77 2.91

2.90

2.89 2.85

2.91 2.86

65", 32", 12" 43", 46", I"

30, 72", 6"

83". 60". 4" 20°, 22", 24" 94", 19", 12"

SO", II",2I0 58", 30", 60"

22", 60", 61"

5",70", 13" 72", 66", 4"

3", 75", 68" 59", 75", 64"

'The criterion used for these residues in the proteins listed in Table 1 is a distance from the peptide nitrogen atom to the side chain carbonyl carbon of less than 3.OA (cf. Figs. 2 4 ) .

conformations of asparagine and aspartic acid residues that can potentially form suc- cinimides are selected against in stable proteins, we compared the psi and chi, values for other residues that also contained an unbranched beta carbon and a gamma carbon atom. These residues (Glu, Gln, Arg, His, Lys, Leu, Tyr, Trp, Phe, and Met) are identical to Asp and Asn residues through the side chain gamma carbon atom but are chem- ically unreactive at this position to attack by the peptide bond nitrogen atom. We found that the distribution of dihedral angles was not significantly different between these 10 residues and the aspartyl and asparaginyl residues shown in Fig. 3 (data not shown). In Fig. 4, we have summarized these data from measurements of the distances between the peptide nitrogen and gamma carbon. We find no significant difference in the average distance from groups of intracellular Asp, Asn, or the collection of Glu, Gln, Arg, His,

Lys, Tyr, Trp, Phe, and Met residues, and only perhaps a marginal increase in the average distance for Asp and Asn residues in the extracellular proteins (cf. Fig. 4 legend). For both classes of proteins, the distribution pattern also appears to be similar (Fig. 4).

Thus, the factors that lead to a general exclusion of conformations that promote succinimide formation do not appear to be specific for Asp and Asn residues. These factors appear to involve the energetics of protein folding. For example, minimal con- formational energy states for protein side chains occur for values of chi, of - 60" (g+), - 180(180") (t), and 60" (g-) (28). The lowest energy occurs with the g+ conformation that is 180" away from the chi, value of 120" needed for imide formation. In fact, this latter value represents an energy maximum. Studies of the chi, angle distribution in a variety of protein residues, including aspartyl and asp- araginyl residues, indicate that these energetic

817

S. Clarke

A. I ntracel lula r

B. Ext racell ula r - -’ i) As’p

n o r

0 Distance N - C, ( A )

FIGURE 4

Distribution of distances between the peptide nitrogen atom and the gamma carbon atom for aspartyl, aspara- ginyl, and a group of residues including Glu, Gln, Arg, Lys, His, Trp, Tyr, Phe, Met, and Leu (other) for the proteins in Table I . Data from Asp-Gly or Asn-Gly sequences are distinguished by dots. A. Intracellular proteins. The average distance for aspartyl residues is 4.14A, for asparaginyl residues 4.05 A, and for the group of other residues 4.12 A. B. Extracellular proteins. The average distances for Asp, Asn, and “other” residues are 4.14A, 4.22A, and 4.10A, respectively.

considerations are primary determinants in the final structure (28, 29; cf. ref. 30).

In general, there also do not appear to be large differences in the distribution of main chain dihedral angles phi and psi for Asp, Asn, and other amino acid residues in proteins (23, 3 1). However, model-building studies have suggested the possibility of intra- molecular hydrogen bonds between the Asn side chain amide and the backbone carbonyl atom that may stabilize a class of structures that generally do not occur for other residues (31). It is of interest that when this structure does occur in proteins, the peptide nitrogen is effectively excluded from attacking the amide and forming an imide.

The overall conclusion from these studies is that the most stable conformations of Asn and Asp residues in proteins do not allow for facile succinimide formation and that a special selection is not required to prevent proteins from undergoing this spontaneous degradation reaction at rates comparable to those observed in model peptides (6). This conclusion, of course, leaves the question of where isoaspartyl and D-aspartyl residues originate from. It should be stressed that the proteins we have examined in this study, as well as those studied previously, may represent a particularly stable class of proteins that can thus be purified, crystal- lized, and subjected to and X-ray beam without major alterations to the structure. It is possible that the cellular proteins that readily generate succinimides would not stand up to these treatments. It is of interest to note that the protein in human eryth- rocytes with the lowest methyl-acceptor activ- ity to the L-isoaspartyl/D-aspartyl methyl- transferase is the stable hemoglobin molecule (32).

These results also suggest the possibility that succinimides may largely form in dena- tured or partially denatured proteins in which rotations around the main chain and side chain dihedral angles allow intramolecular imide formation at sites which are conforma- tionally unfavorable in the native structure. Evidence for this idea was obtained by Bornstein & Balian (33) who showed that a succinimide could be formed in v i m at Asn-

818

Succinimide formation in proteins

Clear evidence has only been obtained from a succinimide deamidation mechanism for one of the proteins shown in Table 7, seminal plasma ribonuclease. In this instance, chemical analysis of the deamidated alpha polypeptide indicated the presence of both normal and isoaspartyl peptides in the ratio expected for a succinimide intermediate (36). The situation for the other proteins is less certain. The presence of a carboxypeptidase resistant bond in the deamidated form of rabbit muscle aldolase suggests that an isoaspartyl residue may be present. On the

67 of bovine pancreatic ribonuclease, but only after the protein was unfolded. In the native structure, this site is poorly positioned for imide formation (Table 5). In intact cells, succinimides and their racemized and isomerized products may represent one of the signals for the specific proteolytic degrada- tion of denatured proteins. Finally, con- formational restraints that exist in native proteins and that can regulate the rate of imide formation make it difficult to interpret studies of succinimide formation in proteins and peptides treated under basic conditions (4) or studies of asparagine deamidation in proteins incubated at elevated temperatures (34, 35) in terms of the reactions that may occur under physiological conditions.

Relevance of these results to the deamidation, isomerization, and racemization of cellular proteins. Although there is little direct evidence in the literature for the presence of succinimide or isoaspartyl residues in intact proteins in vivo (36, cf. refs. 1, lo), and the literature on aspartyl/asparaginyl racemiza- tion is limited to studies where D-aspartic acid is isolated from acid hydrolysates (thus losing any sequence information or knowledge of whether the D-residue was present as a D- aspartyl, D-asparaginyl, or D-isoaspartyl residue; cf. ref. 37), there is a body of know- ledge on protein deamidation. As a result of, the charge difference that occurs when an asparagine residue is converted into a car- boxylic acid residue, it has been possible to follow deamidation reactions that do occur in vivo in several systems by analyzing the elec- trophoretic mobility of the proteins. The results from a number of studies in which it has been possible to identify the specific sequences involved in deamidation in intact cells are summarized in Table 7. Unfortunate- ly, there are no three-dimensional structures reported yet for these proteins. Although the structure of triose phosphate isomerase from chicken muscle is known, this enzyme is not known to undergo spontaneous deamidation (46). In the chicken enzyme, Asn-71 is replaced by a lysine residue and the conformation of the remaining Asn-15 site does not favor succinimide formation (psi = - 174"; chi, = - 88").

TABLE 7 Specific in vivo deamidation sites in cellular proteins

Protein Site of Product Reference deamidation identification

Human triose phosphate isomerase Asn,,Gly ? 38 Asn,, Gly ? 38

Human hemoglobin Providence (Beta 82 Lys + Asn)

Bovine seminal ribonuclease

Human growth hormone

Human erythrocyte hypoxanthine-guanine phospho- ribosyltransferase

Bovine cytochrome c

Rabbit muscle aldolase

Mouse IgG Kappa chains

Bovine lens alphacrystallin

Asn,,Gly Asp' 39

A s ~ , G l y Asp, isoAspb 36

AS"I,2 ASP ? 40

Asn,, Asp Asp" 41

Asn,,,Glu ? 42

Asn,,,His IsoAs~(?)~ 43

Asn, Ile ? 44

Asn,,Met ? 45

'Evidence for a normal aspartyl linkage from the semi- quantitative recovery of Asp and the following residue upon Edman sequence analysis. The presence of an isoaspartyl residue would result in the termination of the Edman sequence. b T ~ o tryptic peptides were obtained containing this site in the deamidated alpha species of this protein. The identification of the isoaspartyl and aspartyl linkage was based on both chemical analyses and the susceptibility of the peptides to methylation by the L-isoaspartyl/D- aspartyl protein rnethyltransferase. 'The "unusual" failure of carboxypeptidase A to cleave the Asp-His linkage is consistent with the presence of an isoaspartyl linkage (cf. ref. 3).

819

S. Clarke

other hand, Edman sequencing results suggest that the deamidated product in both hemoglobin Providence and hypoxanthine- guanine phosphoribosyltransferase contains a normal aspartyl residue. If a significant amount of isoaspartyl residues were present at this position, a decrease in the yield of the aspartyl derivative would be expected and this was not found in either case. The apparent absence of isoaspartyl residues from these deamidation reactions suggests two possibilities - either these processes do not occur via succinimide intermediates (perhaps as the result of an unfavorable conformation) or the aspartyl residue is formed by the repair of an original isoaspartyl residue by protein carboxyl methyltransferase activity (9-1 1). Further work will be necessary to resolve this question.

Alternative deamidationlracemization path- ways involving isoimide intermediates from asparagine and aspartyl residues. Succinimide formation is dependent upon the nucleophilic attack of the peptide nitrogen atom on the carbonyl carbon of the side chain of aspartyl and asparaginyl residue. Attack at this posi- tion can also occur with the peptide bond carbonyl oxygen as the nucleophile. This reaction results in the formation of a five- membered 4,5-diamino-2-oxolanone ring which we have previously termed an “anhydride” (47, 48) and has also been termed as “isoimide” (49, 50). Several isoimides have been synthesized and have been shown to be readily hydrolyzed to the normal carboxylic acid (49, 50). Isoimides may also share the racemization-prone property of succinimides (6). Although there is considerable evidence for succinimide inter- mediates in peptide aspartyl ester hydrolysis reactions (7, 9-10, 14, 25, 51-52), this evidence does not preclude the initial formation of an isoimide because it is known that rearrangement reactions may convert isoimide products to succinimides (49, 50). Such isoimide intermediates may therefore be involved in protein deamidation, isomeriza- tion, and racemization reactions at asparagine and aspartyl residues.

The optimal conformation for the attack

820

by the peptide bond oxygen to form an iso- imide differs from that needed for succini- mide formation because the peptide bond must be rotated by 180” to bring the oxygen atom in proximity with the carbonyl side chain carbon (cf. Fig. 2). Examination of the conformation of the aspartyl and asparaginyl residues in the proteins in Table 1 reveals that there are no cases in which near optimal values for isoimide formation (psi = 60°, chi, = 120”) are present (Fig. 3). In fact, if rotations of psi and chi, of up to 60” from the optimal values are allowed, only about 12% of the asparagine residues and 5% of the aspartyl residues would be candidates for isoimide formation. Although these data would suggest that isoimides might not form in this group of proteins, such derivatives remain viable candidates for intermediates of deamidation and racemization reactions in other types of folded proteins and in unfolded proteins.

ACKNOWLEDGMENTS

Special thanks are due to Dr. Pinak Chakrabarti for writing the programs to extract dihedral angles from the set of Brookhaven atomic coordinates. 1 am also indeb- ted to Drs. Chakrabarti, Douglas Rees, David Sigman and Clare OConnor for very helpful discussions. Thanks are also due to Terrence Geiger and to William Wensil, who assisted with the initial studies on the distribution of asparagine and aspartyl residues in proteins. Part of this work was performed in the Department of Molecular Biology at Princeton University while the author was a Visiting Fellow in the laboratory of Dr. Jeffry Stock; his interest in this work is greatly appreciated. This work was supported by grants from the National Science Founda- tion (DMB-8602102) and the National Institutes of Health (GM-26020).

1. 2. 3.

4.

5 .

6.

7.

REFERENCES

Clarke, S . (1985) Ann. Rev. Biochem. 54, 479-506 Bada, J.L. (1984) Merhodr Enzymol. 106, 98-1 I5 Murray, E.D., Jr. & Clarke, S. (1984) J. Biol. Chem.

Aswad, D.W. (1984) J. Biol. Chem. 259, 10714-10721 Meinwald, Y.C., Stimson, E.R. & Scheraga, H.A. (1986) Int. J. Peptide Protein Res. 28, 79-84 Geiger, T. & Clarke, S . (1987) J. Biol. Chem. 262,

Bernhard, S.A., Berger, A., Carter, J.H., Katchal-

259, 10722-10732

785-794

Succinirnide formation in proteins

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

ski, E., Sela, M. & Shalitin, Y. (1962) J. Am. Chem.

McFadden, P.N. & Clarke, S. (1982) Proc. Narl. Acad. Sci. US 79, 2460-2464 McFadden, P.N. & Clarke, S. (1987) Proc. Narl. Acad. Sci. US 84,2595-2599 Johnson, B.A., Murray, E.D., Jr., Clarke, S., Glass, D.B. & Aswad, D.W. (1987) J. Biol. Chem. 262,

Johnson, B.A., Langmack, E.L. & Aswad, D.W. (1987) J. Biol. Chem. 262, 12283-12287 Bornstein, P. & Balian, G. (1977) Merhoak Enzymol. 47, 132-145 Blodgett, J.K., Loudon, G.M. & Collins, K.D. (1985) J. Am. Chem. SOC. 107, 4305-4313 Bodanszky, M. & Kwei, J.Z. (1978) Int. J. Pepride Protein Res. 12, 69-74 Bernstein, F.C., Foetzle, T.F., Williams, G.J.B., Meyer, E.F., Jr.. Brice, M.D., Rodgers, J.R., Kennard, O., Shmanouchi, T. & Tasumi, M. (1977) J. Mol. Biol. 112, 535-542 Clarke, S. (1985) in Cellular and Molecular Aspects of Aging: The Red Cell as a Model (Eaton, J.W., White, J.W. & Konzcn, D., eds.), pp. 91-103, Alan R. Liss, New York IUPAC-IUB Commission on Biochemical Nomenclature (1970) J. Mol. Biol. 52, 1-17 Allen, F.H., Bellard, S., Brice, M.D., Cartwright, B.A., Doubleday, A., Higgs, H., Hummelink, T., Hummelink-Peters, B.G., Kennard, O., Mother- well. W.D.S., Rodgers, J.R. & Watson, D.G. (1979) Acra Crysr. B35, 2331-2339 Kim, S., Galletti, P. & Paik, W.K. (1980) J. Biol. Chem. 255, 338-341 Barber, J.R., Morimoto, B.H., Brunauer, L.S. &. Clarke, S. (1986) Biochim. Biophys. Aclu 886,

Ota, I.M., Ding, L. & Clarke, S. (1987) J. Biol. Chem. 262, 8522-8531 Shalitin, Y., & Bernhard, S.A. (1964) J. Am. Chem.

Nemethy, G. & Scheraga, H.A. (1977) Qwrr. Rev. Biophys. 10, 239-352 Sibanda, B.L. & Thornton, J.M. (1985) Nature 316, 170-174 Johnson, B.A. & Aswad, D.W. (1985) Biochemistry 24, 2581-2586 Johnson, B.A., Freitag, N.E. & Aswad, D.W. (1985) J. Biol. Chem. 260, 10913-10916 Ramachandran, G.N., Ramakrishnan, C. & Sasisekharan, V. (1963) J. Mol. Biol. 7, 95-99 Janin, J., Wodak, S., Levitt, M. & Maigret, B.

SOC. 84,242 1-2434

5622-5629

361-372

SOC. 86, 2291-2292

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

Peptide Protein Res. 18, 121-126 OConnor, C.M. & Clarke, S. (1984) J. Biol. Chem. 259,2570-2578 Bornstein P. & Balian, G. (1970) J. Biol. Chem. 245, 48544856 Ahern, T.J. & Klibanov, A.M. (1985) Science 228, 1280-1284 Ahern, T.J., Casal, J.I., Petsko, G.A. & Klibanov, A.M. (1987) Proc. Natl. Acad. Sci. VS84,675-679 Di Donato, A., Galletti, P. & DAlessio, G. (1986) Biochemistry 25, 8361-8368 Brunauer, L.S. & Clarke, S. (1986) J. Biol. Chem. 261, 12538-12543 Yuan, P.M., Talent, J.M. & Gracy, R.W. (1981) Mechan. Ageing Develop. 17, 15 1-1 62 Moo-Penn, W., Jue, D.L., Bechtel, K.C., Johnson, M.H., Schmidt, R.M., McCurdy, P.R., Fox, J., Bonaventura, J., Sullivan, B. & Bonaventura, C. (1976) J. Biol. Chem. 251, 7557-7562 Lewis, U.J., Singh, R.N.P., Bonewald, L.F. & Seavey, B.K. (1981) J. Biol. Chem. 256, 11645-1 1650 Wilson, J.M., Landa, L.E., Kobayashi, R. & Kelley, W.N. (1982) J. Biol. Chem. u7, 14830-14834 Flatmark, T. (1966) Acra Chem. Scand. 2%

Midelfort, C.F. & Mehler, A.H. (1972) J. Biol. Chem. 247, 3618-3621 Svasti, J. & Milstein, C. (1972) Biochem. J. 128, 427444 Voorter, C.E.M., Mulders, J.W.M., Bloemendal, H. & de Jong, W.W. (1987) in Posrrranslational Modijications of Proreins and Ageing (Zsppia, V., Galletti, P., Porta, R. & Wold, F., eds.), Plenum Press, New York, in press Yuan, P.M., Talent, J.M. & Gracy, R.W. (1981) Biochim. Biophys. Acra 671, 21 1-218 Barber, J.R. & Clarke, S. (1985) Biochemistry 24, 4867-487 I Clarke, S. (1986) in Biological Methylarion and Drug Design (Borchardt, R.T., Creveling, C.R. & Ueland, P.M., eds.), pp. 3 1 4 , Humana Press, New York Sauers, C.K., Marikakis, C.A. & Lupton, M.A. (1973) J. Am. Chem. SOC. 95, 67924799 Ernst, M.L. & Schmir, G.L. (1966) J. Am. Chem. SOC. 88, 5001-5009 Murray, E.D., Jr. &Clarke, S. (1986) J. Biol. Chem. 261,306-312 McFadden, P.N. & Clarke, S. (1986) J. Biol. Chem. 261, 11503-11511

1487-1496

(1978) J. MoI. Bi01. 125, 351-386 Bhat, T.N., Sasisekharan, V. & Vijayan, M. (1979) Int. J. Peptide Protein Res. 13, 170-184 Benedetti, E., MoreUi, G., Nemethy, G. & Department of Chemistry and Biochemistry Scheraga, H.A. (1983) Inr. J. Peptide Protein Res. 22, 1-15

Steven Clarke

University of California Los Angeles, CA 90024

31. Ravichandran, V. & Subramanian, E. (1981) Inr. J . (USA)

82 1