Eur. J. Biochem. 227, 524-528 (1995) 0 FEBS 1995

Site-specific N-terminal auto-degradation of human serum albumin Bernard CHAN', Neil DODSWORTH', John WOODROW', Alan TUCKER2 and Roy HARRIS'

' Delta Biotechnology Limited, Nottingham, England Birkbeck College, 1Jniversity of London, London, England

(Received 22 September 1994) - EJB 94 1443/3

Human serum albumin prepared by blood fractionation for clinical purposes was found to degrade when stored at or above 30°C. Mass spectrometry and N-terminal sequencing of the protein identified degradation corresponding to the loss of the first two residues, aspartic acid and alanine. The reaction was shown to be dependent upon temperature and the N-terminal a-amino group. In addition, comparison with serum albumins derived from other species showed that the instability of the N-terminus was specific to the human albumin sequence. An intact aspartyl-alanyl dipeptide, purified from degraded albumin solutio'ns, differed substantially from a synthetic dipeptide on amino acid analysis, N-terminal sequencing and NIMR. It is suggested that the released dipeptide may be cyclic, implying a novel cleavage mecha- nism.

Keywords. Human serum albumin ; N-terminus ; degradation ; stability ; dipeptide.

Human serum albumin (HSA) is the most abundant protein in blood (40 g/l) and is widely distributed throughout the body. The single polypeptide chain consists of 585 amino acids (66438 Da) folded into three homologous domains and the three-dimensional structure has been determined by X-ray crys- tallography (He and Carter, 1992). The main role of HSA in the interstitial and blood compartments is the maintenance of osmolarity, but the protein is also involved in the binding and transport of many ligands including fatty acids, breakdown prod- ucts of haem, small drugs and metals (Brown and Shockley, 1982; Kragh-Hansen, 1990; Carter and Ho, 1994).

Extensive studies of the metal-binding properties of HSA have revealed that metal ions bind to a wide variety of sites (Yongqia et al., 1992; Quinlan et al., 1992; Carter and Ho, 1994). Damage to t[SA protein structure resulting from binding of metals such as vanadium, copper and iron has been detected by changes to the thiol groups and the tryptophan residue (Quin- lan et al., 1992). The best characterised metal-binding site is located at the N-tei-minus and has particular high affinity for copper and nickel (ILaussac and Sarkar, 1984; Marx and Chev- ion, 1985; Predki et al., 1992). The groups that participate in this binding have been shown to be the a-amino group, the two intervening peptide nitrogen atoms, the 8-imidazole nitrogen from His3 and the side-chain carboxyl group of Asp1 (Predki et al., 2992).

Stability of a commercial parenteral such as HSA is clearly of major importance in ensuring the long-term safety and effi- cacy of the product. Under current American and European reg- ulations, the use of preservatives is not allowed [US Pharmaco- poeia (USP) and European Pharmacopoeia (EP), 19931. To en- sure stability, appropriate stabilisers (N-acetyl tryptophan and/or

Correspondence ~ G N B. Chan, Delta Biotechnology Limited, Castle

Fax: +44 115 955 1299. Abbreviations. HS.4, human serum albumin ; Des-Asp'-Ala*-HSA,

HSA lacking the first two N-terminal amino acid residues; ES-MS, electrospray mass spectrometry ; RF-HPLC, reverse-phase HPLC ; PhNCS, phenylisothiocyanate; MCO, metal catalysed oxidation.

Court, 59 Castle Boulevard, Nottingham, England NG7 1FD

octanoate) are added to the purified protein, which must then be filter sterilised and heated at 60°C for 10 h in the final product container. Under these conditions, the regulations allow for a shelf life of 3 years at 25°C (EP) or 3 years at 37°C (USP). Despite the apparent stability of this product, we have observed considerable N-terminal degradation in solutions of clinical grade HSA. This study investigates the nature of this degrada- tion.

MATERIALS AND METHODS Albumin sources and preparations. Serum albumins were

purchased from Sigma, except for clinical-grade solutions of HSA which were purchased from Armour Pharmaceutical Co., Baxter Healthcare Cop., Blood Products Ltd, Cutter Biological and Immuno. All HSA solutions had been heated at 60°C for 10 h by the manufacturers during commercial production. Albu- mins were diluted to 50 g/l in 141 mM sodium chloride and 4 mM sodium octanoate. All samples were stored at 4°C.

Albumin solutions were incubated at temperatures of 30"C, 37°C or 57°C for varying periods. For assessing the effects of metal, HSA solutions were brought to 50 mM metal salts (CuSO,, CoCl,, FeSO,, ferric citrate, NiCl,, TiC1, and ZnC1,). Samples used to examine the effect of metal chelation were in- cubated either with 1 mM EDTA or were previously incubated with Chelex (Bio-Rad) at 4°C for 2 h. Anti-oxidants [Trolox (Aldrich) or butylhydroxytoluene] were added at 10 mM where indicated.

Reaction of proteins with dicarboxylic anhydrides, such as citraconic anhydride or succinic anhydride, results in modifica- tion of lysine residues and the N-terminal a-amino group. HSA was citraconylated essentially by the method of Yarwood (1989) to block the N-terminal a-amino group. HSA (50 g/l) was incu- bated at room temperature in 0.2 M TrisEIC1, pH 8.5, 6 M gua- nidine HC1 (3 ml) and 2x10 p1 citraconic anhydride was added at 15 min intervals. The reaction was stirred continuously for 2 h and the pH was maintained in the range 8.3-8.6 by the addi- tion of 1 M NaOH. After citraconylation, the HSA was desalted

Chan et al. ( E m J. Biochem

by passage through a PD-10 column (Pharmacia) equilibrated with water, and concentrated to 50 g/l by ultrafiltration using a PM30 membrane (Amicon).

Isolation of dipeptide. Albumin samples were desalted using PD-10 columns equilibrated in water. The dipeptide was isolated from the salt fraction by reverse-phase HPLC (RP- HPLC) on a 250 mmX4.6 mm Vydac C-18 column using a Gil- son HPLC system. A linear gradient of l % to 100 % solvent B was used [where solvent A was 0.1 % (by vol.) F,CCOOH and solvent B was 0.09% (by vol.) F,CCOOH in 70% (by vol.) acetonitrile), with a flow rate of 0.5 ml/min over 20 min, moni- toring absorbance at 214nm. The dipeptide was collected di- rectly as it eluted from the detector.

Analytical techniques. Composition analysis was per- formed on an Applied Biosystems (ABI) 420H amino acid ana- lyser using automated hydrolysis (except where indicated) and phenylisothiocyanate (PhNCS) pre-column derivatisation. N-ter- minal sequencing was performed by Edman degradation using an ABI 477A protein sequencer with an on-line 120A phenyl- thiohydantoin amino acid analyser. Albumin samples were sub- jected to molecular mass determination by electrospray mass spectrometry (ES-MS) using a VG BIO-Q and a VG Quattro mass spectrometer ; the instruments were calibrated using horse heart myoglobin (16951 Da; Sigma) over m/z range 950- 1750 Da. Samples were prepared for ES-MS by either dialysis against MilliQ grade water followed by lyophilisation, or by RP- HPLC using an ABI 140A separation system with a 30mm X2.1 mm Brownlee RP8 cartridge. A Bruker AM500 spectrom- eter was used to obtain proton-NMR spectra of the dipeptide. Samples were prepared in D,O at pH 5.3.

RESULTS

Detection of N-terminal degradation. Analysis of each of the clinical-grade HSA solutions by ES-MS revealed the presence of two major species, corresponding to the expected masses of HSA with a free thiol at Cys34 (66438 Da) and HSA with the thiol blocked with cysteine (66557 Da); an example spectrum is shown in Fig. 1 a. In addition, a species approximately 185 Da smaller than the unblocked HSA was present to varying extents ; a corresponding species for blocked HSA, approximately 186 Da smaller, was also observed. Incubation of HSA samples at 37°C for 8 weeks resulted in an increase of the smaller spe- cies from approximately 15 % to 35 % (determined by peak- height ratios) of the total (Fig. 1 b). N-terminal sequence analysis showed the presence of a secondary sequence, His-Lys-Ser-Glu- in addition to the major sequence Asp-Ala-His-Lys- representing the expected N-terminus (Fig. 2). This indicated that the smaller species seen on ES-MS represented HSA lacking the first two amino acids, which would produce a theoretical decrease in mass of 186 Da. A released dipeptide could not be detected by ES-MS analysis of this sample due to the high background at low-mass range.

The concentration of the Des-Asp'-Ala'-HSA (HSA lacking the first two N-terminal amino acid residues), measured by both ES-MS and quantification of the secondary sequence on N-ter- minal sequencing, was seen to increase with time when incu- bated at 30"C, 37°C and 57"C, the rate of degradation increas- ing with temperature (data not shown). No increase in the con- centration of degradation product was observed during storage at 4OC.

Dipeptide analysis. The loss of the two N-terminal residues of HSA could also be detected by purification of the Asp-Ala dipeptide. HSA stored at 4°C or incubated at 57°C for 12 days

100 a

h s U U c 1

i 4

0

'27)

b IoC

n s W U c a

4 i

525

3

66000 67000 66000 67000

Mass (Da) Mass (Da)

Fig.1. Electrospray mass spectrometry of HSA. HSA was stored either at 4°C (a) or incubated at 37°C for 8 weeks (b). Peak 1 (a, 66252Da; b, 66252Da) is Des-Asp'-Ala*-HSA; peak 2 (a and b ~ 6 6 3 5 6 Da) is Des-Asp'-Ala'-HSA with blocked thiol; peak 3 (a, 66437 Da; b, 66436 Da) is HSA; peak 4 (a, 66554 Da; b, 66550 Da) is HSA with blocked thiol. The spectra shown are typical examples of each HSA from the different sources.

I

0 8 16 24 Time (min)

Fig. 2. N-terminal sequence of degraded HSA. The first four cycles of N-terminal sequencing are shown for HSA. DAHK is the major se- quence, but a minor sequence, HKSE, was also detected; the minor se- quence was most noticeable by the appearance of Lys at cycle 2.

was desalted and the salt fraction of each was subjected to RP- HPLC. A single peak at 214 nm, appearing soon after the injec- tion peak, was detected for both samples (Fig. 3). The size of this peak was much increased by the incubation at 57°C. The peaks were collected and the composition of each determined by amino acid analysis; both showed identical patterns (the 57 "C incubated sample is shown in Fig. 4a). Prior to hydrolysis, no amino acids were detected at above background levels, whereas after hydrolysis, two major species corresponding to Asp and Ala were present. This suggested that the N-terminal degrada- tion involved the loss of an intact dipeptide rather than the indi- vidual amino acids. The presence of dipeptide in the 4°C sample confirmed that the starting HSA solution was already degraded, as shown previously by ES-MS.

Role of metal ions in N-terminal degradation. The known ca- pacity of the N-terminus of HSA to bind metal ions and of metal ions to damage protein structure (see above) suggested that the

526 Chan et al. (EUK J. Biochem. 227)

----.- I

b

0 4 8 12 0 4 8 12

Time (min) Time (min) Fig.3. RP-HF'LC of HSA-derived dipeptide. Analysis of HSA after either (a) storage at 4"C, or (b) incubation at 57°C for 12 days, by RP- HPLC of PD-10 salt fractions showed that a single peak was detected at 214 nm, eluting immediately after the injection peak. Incubation at the higher temperature resulted in an increase in this species.

observed degradation might be metal dependent. The binding site is partly dependent upon coordination of the metal ion with the N-terminal a-amino group (Predki et al., 1992). Using citra- conic anhydride, the N-terminal a-amino group was blocked as described in Materials and Methods, and complete blockage confirmed by sequencing. During the first cycle no sequence was detected. However, removal of the blocking group by F,CCOOH used in the Edman degradation during this first cycle allowed subsequent sequencing and the detection of Asp in the second cycle. The 14-terminally blocked HSA was incubated at 57°C for 12 days to1 examine the effect of blocking on N-termi- nal degradation. Afier incubation, the sample was re-sequenced and showed no increase in the level of secondary sequence (data not shown), as observed with the unblocked molecule. This indi- cated that the a-amino group plays a crucial role in the reaction and supported the involvement of metals ions.

HSA was incubated with a mixture of metal solutions, metal chelators and anti-oxidants, such as Trolox and butylhydroxytol- uene (see Materials and Methods), to assess the possibility that degradation was caused by a metal-catalysed oxidation (MCO) reaction. In all cases, mass spectrometry, N-terminal sequencing and dipeptide analysis of the samples after incubation at 57°C showed no difference in the rate of N-terminal degradation as a result of any of the treatments (data not shown). Therefore, de- spite showing a role for the a-amino group, degradation was not the result of oxidative damage by metals.

N-terminal degradation of albumins from other species. Since metal binding was not the cause of degradation, albumins of species other than human were examined, as some maintain the metal-binding site, but have evolutionarily diversified se-

Table 1. N-Terminal sequences of serum albumins.

Serum albumin from

Amino acid at residue number

1 2 3 4 5

Human Bovine Chicken

Goat Horse Pig Rabbit Rat Sheep

Dog

Ala Thr Ala Ala Thr Thr Thr Ala Ala Thr

His His Glu S r His His TYr His His His

LYS Ser LYS Ser

LYS Ser LYS Ser LYS Ser LYS Ser LYS Ser LYS Ser LYS Ser

His LYS

quences (Table 1). Many of these differ from HSA at the N- terminus only by one or two residues and offer an ideal opportu- nity to study the specificity of the N-terminal degradation reac- tion. A number of albumins, all with N-terminal sequences dif- ferent to HSA, were incubated at 50 gll for 2 weeks at 57°C. The samples were examined for both the degraded protein and the released dipeptide by N-terminal sequencing and amino acid analysis, respectively.

The degraded protein was measured as an increase in the lysine peak at cycle 2 (cycle 3 in chicken serum albumin) during N-terminal sequencing. This residue is highly conserved since all of the albumins have lysine as the fourth residue except for chicken serum albumin, in which glutamic acid at position 3 shifts the lysine to position 5. No significant degradation at the N-terminus was detected by N-terminal sequencing in any of the non-human albumins. The assay of dipeptide by amino acid analysis confirmed that the non-human albumins showed no sig- nificant N-terminal degradation. This indicated that the cleavage reaction was highly specific for the HSA N-terminal sequence.

Characterisation of HSA-derived dipeptide. The purified dipeptide from HSA was found to differ significantly from a synthetic Asp-Ala dipeptide (Sigma). Firstly, the synthetic dipeptide added to the HSA sample was not recovered in the salt fraction upon desalting on a PD-10 column in water, the method used to isolate the HSA-derived dipeptide. Secondly, amino acid analysis of the synthetic dipeptide without hydrolysis showed a major species assumed to be PhNCS-aspartyl-alanine (Fig. 4 b), which was absent with the HSA-derived dipeptide (Fig. 4 a). Finally, N-terminal sequencing of the synthetic dipep- tide gave Asp and Ala as expected, whereas the HSA-derived dipeptide yielded no sequence. These results indicated that de- gradation of HSA released a modified N-terminal dipeptide with a blocked N-terminus.

Purified and synthetic dipeptides were analysed by proton NMR (Fig. 5). Examination of the Asp PCH, signals showed considerable inequivalence for the synthetic dipeptide, giving an eight-line system (Fig. 5 b), but little inequivalence for the HSA- derived dipeptide (Fig. 5 a). This indicated that the two dipep- tides were of different structure.

DISCUSSION

Here we report the site-specific N-terminal degradation of HSA. Internal fragmentation of HSA and other albumins has been described previously (Max and Chevion, 1985 ; Davies and Delsignore, 1987; Uchida and Kawakishi, 1988; Meucci et al., 1991). However, in these studies the degradation was the

Chan et al. ( E m J. Biochern. 227) 527

A

D I

Time (min)

Time (min)

Fig. 4. Amino acid analysis of dipeptides. HSA-derived dipeptide (a) collected after RP-HPLC (as shown in Fig. 3) and synthetic Asp-Ala dipeptide (b) were examined by amino acid analysis, before hydrolysis (. . . . .) and after hydrolysis (-). Asp (D) and Ala (A) were detected before hydrolysis for both dipeptides, but a species (X) detected after hydrolysis was only present for the synthetic dipeptide.

result of in vitro MCO generated by the addition of abnormally high concentrations of exogenous metal ions. Here, we have identified a fragmentation that is highly specific for the HSA N- terminal sequence, known to be a strong metal-binding site. The sequence specificity of the reaction for the N-terminus indicated the involvement of metal binding and an MCO reaction, but this was not supported by our data. Metal chelators such as EDTA and Chelex and anti-oxidants such as Trolox and butylhydroxy- toluene both of which should prevent MCO reactions, were inef- fective. Additionally, if metals were catalysing the reaction, an increase in degradation with higher concentrations of metal ions would be expected, but this was not observed.

Mass spectrometry and N-terminal sequencing have consis- tently shown that a specific dipeptide is released from the pro- tein. The data suggest the reaction produces an alteration of the expected dipeptide, changing its chromatographic properties and blocking the accessibility of the a-amino group to PhNCS, but still yielding Asp and Ala upon hydrolysis. The signals obtained from NMR for the purified dipeptide and the synthetic Asp-Ala standard confirm that these are two different molecules. One potential rearrangement into a PAsp-Ala form (Pistorius et al., 1993) would possess a free amino group that would react with PhNCS allowing both N-terminal sequencing and detection of a species by amino acid analysis without hydrolysis, neither of which were observed in practice.

A novel reaction mechanism is therefore proposed to account for the observed reaction. In this model, cleavage of the Ala2- His3 peptide bond is a hydrolysis, catalysed by proton with- drawal from the a-amino group by the carboxyl group of Aspl and possibly proton donation to the peptide-bond carbonyl from the imidazole of His3. This results in a cyclic dipeptide, as shown in Fig. 6. Such a reaction would not be metal dependent and the proposed cyclic dipeptide would not possess a free N-

AspolCH n AlaolCH

dioxan AlaSCH, n

4.4 3.0 Chemical Shift (ppm) 1.4

I I I

4.4 3.0 Chemical Shift (ppm)

Fig. 5. NMR spectra of dipeptides. (a) HSA-derived dipeptide; (b) syn- thetic dipeptide. Dioxan was used as an internal reference in the HSA- derived dipeptide sample.

N H - - -

Fig. 6. Model for N-terminal degradation. Nucleophilicity of the a- amino group is increased by proton withdrawal by the Aspl carboxyl. Nucleophilic attack by the a-amino nitrogen on the Ala2-His3 peptide carbonyl may result in cleavage of the peptide bond and release of a cyclic dipeptide. The electrophilic nature of the carbonyl may possibly be enhanced by proton donation from the His3 imidazole.

terminus. Therefore it could not react with PhNCS, preventing the labelling of the dipeptide and N-terminal sequencing as ob- served. Amino acid analysis after hydrolysis, however, would still yield Asp and Ala despite cyclisation. The NMR spectrum of the cyclic dipeptide would also differ considerably from that of a linear form of Asp-Ala.

This proposed mechanism for N-terminal degradation would partly explain the high specificity of the reaction for the HSA sequence since the reaction of bonds depends upon the Aspl carboxyl group and the His3 imidazole group. However, it does

528 Chan et al. (EUK J. Biochem. 227)

not explain why four non-human albumins with Thr in place of the human Ala at residue 2, fail to show the same reaction, since it is not clear how Ala2 would contribute to the specificity. Pos- sibly other structural parameters may also play a role in the reac- tion, such as the flexible nature of the N-terminus (He and Car- ter, 1992) and the effects of residues further into the HSA se- quence on N-terminal conformation, e.g. the free thiol of Cys34 (Davies et al., 19938; Sadler et al., 1994).

Protease contamination has been eliminated as the cause of this degradation reaction for three reasons. Firstly, the same de- gradation was evident in each of the different commercial sources of HSA. Secondly, all HSA solutions had been heated at 60°C for 10 h by the manufacturers which would eliminate most potential protease activity. Finally, degradation resulting from proteases would produce a linear Asp-Ala dipeptide, which is clearly not the product of this reaction.

N-terminal degradation occurred under physiological condi- tions of pH and temperature in vitro and may therefore take place in vivo. The results indicate that the process would take place at 37°C during the life-time of the HSA molecule (half- life of approximately 21 days in circulation). It would be inter- esting to determine the rate of auto-degradation in vivo and the concentration of the degraded species, where factors in circula- tion with HSA may influence the reaction. N-terminal auto-de- gradation might represent the initial stages of HSA breakdown in viva If this is the case, the fate of the released dipeptide should be evaluated since some dipeptides are known to be active neuropeptides in the central nervous system (Sandberg et al., 1994). Clearly this degradation warrants further investigation to confirm the proposed mechanism and to elucidate the impor- tance of this reactio'n in vivo.

We would like to thank Mr J. Christodoulou (Birkbeck College, Lon- don, U. K.) for help with the NMR analysis.

REFERENCES Brown, J. R. & Shockley, P. (1982) Serum albumin: Structure and char-

acterisation of its ligand binding sites, in Lipid-protein interactions I (Jost, P. C. & Griffith, 0. H., eds) pp. 25-68, Wiley, New York.

Carter, D. C. & Ho, J. X. (1994) Structure of serum albumin, Adv. Pro- tein Chern. 45, 153-203.

Davies, K. J. A. & Delsignore, M. E. (1987) Protein damage and degra- dation by oxygen radicals, J. Biol. Chem. 262,9908-9913.

Davies, M. J., Gilbert, B. C. & Haywood, R. M. (1993) Radical-induced damage to bovine serum albumin: role of the cysteine residue, Free Radical Rex Commun. 18, 353-367.

He, X. M. & Carter, D. C . (1992) Atomic structure and chemistry of human serum albumin, Nature 358, 209-215.

Kragh-Hansen, U. (1990) Structure and ligand binding properties of hu- man serum albumin, Dan. Med. Bull. 37, 57-84.

Laussac, J.-P. & Sarkar, B. (1984) Characterisation of the copper(I1)- and nickel(I1)-transport site of human serum albumin. Studies of cop- per(I1) and nickel(I1) binding to peptide 1-24 of human serum albu- min by and 'H NMR spectroscopy, Biochemistry 23,2832-2838.

Marx, G. & Chevion, M. (1985) Site-specific modification of albumin by free radicals, Biochem. J. 236, 397-400.

Meucci, E., Mordente, A. & Martorana, G. E. (1991) Metal-catalyzed oxidation of human serum albumin : conformational and functional changes, J. Biol. Chem. 266,4692-4699.

Pistorius, A. M. A., Groenen, J. T. A. & De Grip, W. J. (1993) Infrared analysis of peptide succinimide derivatives, Int. J. Pept. Protein Res.

Predki, P. F., Harford, C., Brar, P. & Sarkar, B. (1992) Further characteri- sation of the N-terminal copper(I1)- and nickel(I1)-binding motif of proteins, Biochem. J. 287, 211 -215.

Quinlan, G. J . , Coudray, C., Hubbard, A. & Gutteridge, J. M. C. (1992) Vanadium and copper in clinical solutions of albumin and their potential to damage protein structure, J. Pharm. Sci. 81, 611- 614.

Sadler, P. J., Tucker, A. & Viles, J. H. (1994) Involvement of a lysine residue in the N-terminal Ni2+ and CuZ+ binding site of serum albu- mins, EUK J. Biochem. 220, 193-200.

Sandberg, M., Li, X., Folestad, S., Weber, S. G. & Orwar, 0. (1994) Liquid chromatographic determination of acidic P-aspartyl and y- glutamyl peptides in extracts of rat brain, Anal. Biochem. 21 7, 48 - 61.

Uchida, K. & Kawakishi, S. (1988) Selective oxidation of tryptophan and histidine residues in protein through the copper-catalyzed autooxidation of 1-ascorbic acid, Agric. Biol. Chem. 52, 1529- 1535.

Yarwood, A. (1989) Manual methods of protein sequencing, in Protein sequencing, a critical approach (Findlay, J. B. C. & Geisow, M. J., eds) pp. 119-145, IRL Press, Oxford.

Yongqia, Z., Xuying, H., Chao, D., Hong, L., Sheyi, W. & Panwen, S. (1992) Structural studies on metal-serum albumin. IV. The interac- tion of Zn(II), Cd(1I) and Hg(I1) with HSA and BSA, Biophys. Chem.

42, 570-577.

42, 201-211.