Comparative Mapping of Recombinant Proteins and Glycoproteins by

Schering-Plough Research Institute,2015 Galloping Hill Road, Kenilworth, NewJersey 07033, and Institut f ur. Medizlnlsche Physik and Bbphyslk, Univers...
0 downloads 0 Views 1009KB Size
Anal. Chem. 1994,66, 2062-2070

Comparative Mapping of Recombinant Proteins and Glycoproteins by Plasma Desorption and Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Anthony Tsarbopoulos,’” Michael Karas,* Kerstln Strupat? Birendra N. Pramanlk,’it Tattanahalll L. Nagabhushan,t and Franz Hillenkamp* Schering-Piough Research Institute, 20 15 Galloping Hili Road, Kenilworth, New Jersey 07033, and Institut f ur Medizinische Physik and Biophysik, Universitat M aster, Robert-Koch-Strasse 3 1, 0-4400 M unster, Germany

The mass spectrometric (MS) techniques of =‘Cf-plasma desorption (PD) and matrix-assisted laser desorptiodionization (MALDI) are compared in the molecular weight determination and the mapping analysis of several recombinant proteins and glycoproteins. MALDI MS analysis exhibited better sensitivity and mass measurement accuracy and a remarkably short analysistime compared witb PD MS analysis. The latter was not successful in the analysis of rhIFN-r and the higher mass mammalian cell-derived IL-5 glycoproteins. Mappingof the Escherichia cobderived rhIFN a-2b and rhU-4 proteins, by direct PD or MALDI MS analysis of the trypsingenerated peptide mixtures provided signals for ca. 95% and 88% of the expected tryptic peptides, respectively. Peptide signals below m/z I500were generally more intense in the PD mass spectra, while higher mass signals were more intense in the MALDI mass spectra. Both PD and MALDI MS analyses provided a rapid confirmation of the existing two and three disulfide bonds in the rhIFN a-2b and rhIL-4 proteins, respectively. In the mapping of the CHO I G 4 glycoprotein, detection of the trypsin-generated glycopeptides was only possible by MALDI, where their detection was greatly improved by using the super-DHB (sDHB) matrix, a 9 1 mixture of 2,S-dihydroxybenzoic acid (DHB) with 2-hydroxy-5-methoxybenzoic acid. This sDHB matrix also generated significantly enhanced and better resolved MALDI peptide signals, which in turn resulted in a much improved mass measurement accuracy. In recent years, the rapid evolution in recombinant DNA methods and DNA sequencing techniques has led to the elucidation of a large number of gene sequences, which, in turn, has made it possibleto produce biologically active proteins and their modified counterparts in large quantities. The development and approval of these protein products as therapeutic agents must satisfy certain criteria of safety, quality, and efficacy. In that respect, the quality control and structure characterization of recombinant protein products presents a formidable task due to the sheer complexity of these protein molecules. The presence of several structural variations, which can arise during the different steps in the protein production process, could severely affect the protein’s biological and immunological properties and thus alter the safety, potency, and stability of the protein product. The + Schcring-Plough Research Institute. 8

Universitat Mfinster.

2062

Analytical Chemktty, Vol. 66, No. 13, Ju& 1, 1994

development of sensitive analytical procedures for the analysis of proteins is, therefore, necessary to provide structural information essential for the drug’s approval. Within the past decade, the development of several new ionization methods combinedwith advances in instrumentation has made significant contributions in the field of biological mass spectrometry (MS). Not only have these ionization methods allowed the direct analysis of intact biomolecules previously not amenable to mass spectrometry,14 but they have also contributed to the development of new mapping approaches for peptides and proteins, as well as carbohydrates and To date, the recently introduced methods of matrix-assisted laser desorption/ionization (MAL~~-~~ DI)’ 1,12 and electrospray/ion spray (ES) i o n i z a t i ~ n have been successfully demonstrated for protein molecular weights (MW) over 25016 and 100 kDa,I7 respectively, at the low picomole-to-femtomolerange. Nevertheless, protein mapping employingdirect MS analysis of the enzyme-generated peptide mixtures has so far been the territory of plasma desorption (PD)18 and fast atom bombardment/liquid secondary ion (FAB/LSI) MS;19,20these two techniques have been extensively used to support or complement data obtained by classical methods used in protein chemistry. To date, the use of the ES and MALDI methods in the analysis of these (1) Sundqvist, B.; Macfarlanc, R. D. Mass Spectrom. Rev. 1985, 4, 421-460. (2) Biemann, K.; Martin, S.A. Mass Spectrom. Rea 1987,6,1-76 and references cited therein. (3) Barber, M.; Green, B. N. Rapid Commun. Mass Spectrom. 1981. I , 80-83. (4) Laccy, M. P.; Keough, T. Rapid Commun. MassSpectrom. 1989,3,323-328. ( 5 ) Moms, H. R.; Panico, M.; Taylor, G. W. Biochem. Biophys. Res. Commun. 1983, I 17, 299-305. (6) Gibson, B. W.; Biemann, K. Proc. Narl. Acad. Sei. U.S.A. 1984,81, 19561960. (7) Carr, S.A. Adu. Drug Deliu. Reu. 1990, 4, 113-147. (8) Tsarbopoulos, A.; Bccker,G. W.;Occolowitz, J. L.;Jardine, I. AMI. Blochem. 1988, 171, 113-123. (9) Rocpstorff, P. Acc. Chem. Res. 1989, 22, 421427. (10) MassSpcctrometry; McCloskcy,J. A., Ed.;MethodsinEnzymologv;Academic Press: San Diego, CA, 1990; Vol. 193. (1 1) Karas, M.; Hillcnkamp, F. AMI. Chem. 1988,60, 2299-2301. (12) Hillcnkamp, F.; Karas, M.; Bcavis, R. C.; Chait, B. T. Anal. Chem. 1991,63, 1193A-1203A. (13) Whitchoust, C. M.; Drcyer, R. N.; Yamashita, M.; Fcnn, J. B. AMI. Chem. 1985,57,675-679. (14) Fenn, J. B.; Mann. M.; Mcng, C. K.; Wong, S.F.; Whitchow, C. M.Mass Spectrom. Reo. 1990, 9, 37-70. (15) Bruins, A. P.; Covey, T. R.;Henion. J. D. AMI. Chem. 1987,59,2642-2646. (16) Karas, M.;Bahr, U.; Ingendoh, A.; Hillcnkamp, F. Angew. Chem.. In?. Ed. Engl. 1989, 28, 760-761. (17) Loo, J. A.; Udscth, H.R.; Smith, R. D. AMI. Biochem. 1989,179,404412. (18) Macfarlanc, R. D.; Torgemon, D. F. Science 1976, 191, 920-925 (19) Barber, M.; Botdoli, R. S.;Scdgwick, R. D.; Tyler, A. N. J . Chem. Soc., Chem. Commun. 1981, 325-327. (20) Aberth, W. H.; Straub, K. M.; Burlingame, A. L. AMI. Chem. 1983, 54, 2029-2034. 000~2700/94/03682062$04.50/0

0 1994 American Chemical Society

protein digests has been limited, due to the complexity of the resulting ES mass spectra and the limited availability of MALDI instruments. The former ionization method usually generates a series of multiply charged ions per peptide fragment present in the mixture, thus giving rise to a highly complex mass spectrum which is extremely difficultto interpret without prior chromatographic separation. In this report, we discuss the use of 252Cf-PDand UVMALDI in the analysis and mapping of recombinant proteins and glycoproteins. The general properties of these mass spectrometric techniques (sensitivity, selectivity, mass accuracy, speed and ease of analysis, etc.) are compared, and their analytical utility in the mapping of recombinant protein products is assessed. EXPERIMENTAL SECTION Materials. Recombinant human interferon a-2b (rhIFN a-2b), recombinant human granulocyte-macrophagecolonystimulating factor (rhGM-CSF), recombinant human interferon-y (rhIFN-y), and recombinant human interleukin-4 (rhIL-4) were purified from Escherichiacoli (shering-Plough Research Institute, Kenilworth, NJ). Interleukin-4 (IL-4) and interleukin-5 (IL-5) werederived and purified from CHO cells (Schering-Plough Research Institute, Kenilworth, NJ/ DNAX Research Institute, Palo Alto, CA) as previously described,21whereas IL-5 was also derived from the nonsecreting mouse myeloma (NS-1) cell line.22 Ammonium bicarbonate and 2-hydroxy-5-methoxybenzoicacid were purchased from Aldrich Chemical Co. (Milwaukee,WI), while dithiothreitol (DTT) and 2,5-dihydroxybenzoic acid were obtained from Sigma Chemical Co. (St. Louis, MO). L-1(Tosylamido)-2-phenylethyl chloromethyl ketone (TPCK)treated trypsin was purchased from Worthington Biochemical Co. (Freehold, NJ). High-performance liquid chromatography (HPLC)-grade trifluoroacetic acid (TFA) was obtained from Pierce Chemical Co. (Rockford, IL), and HPLC-grade water and acetonitrile were purchased from Burdick and Jackson Lab Inc. (Muskegon, MI). All the other chemicals were of the highest purity commercially available and were used without further purification. Plasma Desorption Mass Spectrometry. Plasma desorption mass spectra were obtained on a Bioion 20 californium-252 plasma desorption time-of-flight mass spectrometer (Bio-Ion Nordic, Uppsala, Sweden) equipped with a IO-pCi sample of californium-252 emitting ca. 2000 fission fragments/s and using an accelerating voltage of 18 kV. The mass resolution was approximately 300 at 50% of peak base. The time-offlight spectra were acquired over a 1-3-h period and then converted to mass spectra using thecentroids for H+ and NO+ as calibration peaks. All PD mass spectra illustrated in this paper have been previously subjected to background subtraction. Protein (0.5-2 nmol) and protein digest mixtures (400 pmol-1 nmol) were dissolved in 20 pL of a 1:2 mixture of ethanol:O.l% aqueous TFA. Typically, 5-10 pL of the resultingmixture was appliedon a nitrocellulose- (NC-) coated (21) Le,H. V.; Ramanathan, L.; Labdon, J. E.; Mays-Ichinco, C.; Syto, R.; Ami, N.; Hoy, P.; Takcbe, Y.;Nagabhushan, T. L.; Trotta, P. P. J . Biol. Chem. 1988, 263, 10817-10823. (22) Cowan, N. J.; Sechcr, D. S.; Milstein, C. J. Mol. Biol. 1974, 90, 691-701.

aluminum f0iF3 and ~ p i n - d r i e d before ~ ~ , ~ ~insertion into the source of the mass spectrometer. The electrospray method26 was used for coating the aluminum foil with a nitrocellulose film. An additional step involving thorough rinsing of the NC-adsorbed sample with 0.1%TFA or deionized water was carried out. Thorough rinsing of PDMS samples improves the signal-to-noise ratio of the PD signals, thus enhancing the mass measurement accuracy of proteins and high-mass enzymatic fragments.27 Matrix-Assisted Laser Desorption Mass Spectrometry. Laser desorption mass spectra were obtained on a Lamma 1OOO reflector-type time-of-flight microprobe mass spectrometer (Leybold-Hereaus,Kaln, Germany). This instrument is equipped with both a Q-switched frequency-quadrupled Nd: YAG laser (wavelength 266 nm) and a nitrogen laser (wavelength 337 nm), whose pulse of light is focused by a microscope quartz lens system onto the sample at 4 5 O . Z 8 The MALDI spectra were obtained at an instrument resolution of m/Am 500-800 (50% valley), and generally 20-30 singleshot spectra were accumulated corresponding to a measurement time of several minutes. All MALDI mass spectra shown in this work are the acquired raw data without any smoothing or background subtraction. Protein samples and enzymatic digests were dissolved in 0.1% aqueous TFA to a concentrationof 0.1-0.5 pg/pL, while the 2,5-dihydroxybenzoicacid (DHB) matrix was dissolved in water or 10% (v/v) aqueous ethanol to a concentration of 5-10 g/L. A 9:l mixture of DHB with 2-hydroxy-5methoxybenzoic acid, referred to as super-DHB (sDHB), was also used for protein analysis. It has recently been demonstrated that using the sDHB matrix results in considerable spectral improvement and better mass accuracy.29 To 1 pL of the protein solution was added 10pL of the matrix solution, and 0.5-1 pL thereof was deposited onto a polished stainless steel target and air-dried prior to its transfer into the source of the mass spectrometer. The MALDI mass spectra were usually calibrated externally using calibrant peptides and proteins (human angiotensin I, melittin from beevenom, bovine insulin, and horse heart cytochrome c) from a second preparation on the same sample support; this yielded a mass accuracy of f0.7 Da in the mass range of 1-5 kDa and up to 100 ppm for proteins up to 30 kDa. A more accurate mass assignment of the proteolytic fragments was achieved by using two of the identified peptide fragments as internal calibrants (T19/Ts and T I ~ / T ~ - Tfor I O the tryptic digest of rhIFN a-2b and rhIL-4, respectively); the results of these mass measurementsare listed in Tables 3 and 4. Theanalyte peaks employed as internal calibrants did not contain any potential sites for deamidation, which would cause a 1-Da mass shift. The ~~

~

(23) Jonsson. G. P.; Hcdin. A. B.; Hhkanaeon, P. L.; Sundqvist. B. U. R.; SWc, B. G.S.;Niclscn,P. F.;Roepstorff,P.;Johanason,K.-E.;Kamcnsky.I.;Lindberg, M.S.L.Anal. Chem. 1986.58, 1084-1087. (24) Jardinc, I.; Scanlan, G. F.; Tsarbopouloo, A.; Liberato, D. J. Anal. Chem. 1988.60, 10861088. (25) Niclscn, P. F.; Klarskov, K.; Hojrup, P.; Roepstorff, P. Bfomed. Emiron. Mass Spectrom. 1988,17, 355-362. (26) McNeal, C. J.; Macfarlane, R. D.; Thurston, E. L. Anal. Chem. 1979, 51, 20362039, (27) Tsarbopoulos,A.; Pramanik, B. N.; Rcichcrt, P.;Sicgel, M.M.;Nagabhushan, T.L.; Trotta, P. P. Rapid Commun. Mass Spectrom. 1991, 5, 81-85. (28) Hillenkamp, F.; Karas, M.In MethodsinEnzymology;McCIoskey, J. A., Ed.; Academic Press: Ssn Diego, CA, 1990; Vol. 193, pp 280-295. (29) Karas, M.;Ehring, H.; Nordhoff, E.; Stahl, B.; Strupat, K.; Hillcnkamp, F.; Gnhl, M.;Krebs, B. Org. Mass Spectrom. 1993 28, 14761481.

Amlytlcel m&tVd. t~ 66, , No. 73, Ju& 1, 1994

2063

Tabk 1. M888 M@asuremonlAccuachs ol W fmd MALDI MS

protein

calcd MWO

E. coli IFN a-2b E. coli GM-CSF E. coli IL-4 CHO IL-4

19 265 14 474 14 957 14 957c

CHO IL-5 NS-I IL-5 E. coli IFN-.,

26 294e 26 294c 16 3 W

mtasd MW PDb MALDI* 19 250 14 459 14 952 17 637

19 259 14 475 14 957 16 711d 17 019 17 319 30 840 31 520 16 300

0 The calculated isotopically averaged mass values reflect the disulfide bonds present in the protein molecule. The measured mass value is an average of the determination for the PD +1, +2, and +3 and the MALDI +1and +2 molecular ions (in sDHB matrix). The calculatedmass value correspondsto the rotein part of theglycoprotcin. Theobwed MALDI values correspontto three glycoforms of CHO IL-4. e The calculated mass value reflects the incorporation of an initiator methionine at the N-terminus.

absence of a significant amount of deamidation is also indicated by the random deviation observed in the measurement of the other peaks in the spectra (Tables 3 and 4). Enzymatic Digestion. Protein samples of rhIFN a-2b, rhIL-4, and CHO IL-4 were dialyzed against several changes of 0.1 M ammonium bicarbonate buffer (pH 7.8) and then lyophilized. A 200-pg aliquot was dissolved in 1%ammonium bicarbonate buffer (pH 8.4 adjusted with 10% ammonium hydroxide), and then the sample was incubated with TPCKtreated trypsin at a 1 5 0 (w/w) trypsin:protein ratio for 18 h at 37 OC. The reaction was stopped by flash-freezing followed by lyophilization, and the crude digest mixture was subjected to PD and MALDI mass spectrometric analysis. Reduction of the disulfide bonds was carried out by adding a 30-fold molar excess of DTT over the cysteine present and reacting under nitrogen at 40 OC for 1 h.

RESULTS AND DISCUSSION The results of the molecular weight measurement of the recombinant protein samples by PD and MALDI MS are shown in Table 1. Since both these techniques do not resolve the natural stable isotope pattern of proteins, the provided mass values can only be compared to their isotopically averaged molecular weight (MW). In the PDMS experiments, the isotopicallyaveraged MW of the protein was determined from the observed series of the singly and multiply charged protonated molecular ions (Figure 1 and Table l), whereas only the singly and doubly charged ions were used in the MALDI measurements. The MALDI mass spectra were dominated by singly charged ions, while doubly and triply charged ions usually dominated the PD mass spectra, as clearly shown in the PD and MALDI mass spectra of rhIL-4 (Figures 1and 2). For proteins where analyses by both PD and MALDI techniques were successful the mass measurement accuracy was better than 0.1% and 0.0596, respectively. The latter was the result of the use of the sDHB matrix,29since use of the DHB matrix resulted in lower accuracy mass measurements (-0.1-0.2%). Theuseof thesDHB matrix reduces theusually intense matrix background ions and results in more intense, relatively narrow and sharp molecular ion signals in the spectrum.29 The beneficial effect of the 2-hydroxy-5-meth2064

Anelytlcel ChemLstty, Vd. 66, No. 13, July 1, 1994

MH: 7479

1

R

1 4r li MH:+

MH* 1 4966

R miz Flguro 1. Positive Ion PD mass spectrum of rhIL-4. 100

c

E

50

Flguro 2. Positive ion UVMALDI mass spectrum of rhIL-4 using the SM.19 matrix (a 9: 1 mlxture of DHB with 2-hydroxy-5-methoxybnzoic acid).

oxybenzoic acid additive was attributed to a disorder in the 2,5-DHB lattice resulting in enhanced ion desorption and reduced internal energy, which, in turn, results in reduced metastable fragmentation. It should be noted that the improved mass measurement accuracy obtained by MALDI MS using the sDHB matrix is comparable to that usually achieved by ES mass spectrometric analysis.30 The improvement of the MALDI mass spectra arising from the use of the sDHB matrix was more prominent in the analysis of CHO IL-4 (Figure 3A). For this glycoprotein, both PD and MALDI mass spectrometrictechniques provided far more accurate mass measurements than those obtained by the classical method of SDS-PAGE,31 with comparableor better, in the case of MALDI MS, sensitivity. The singly and doubly charged molecular ion regions of CHO IL-4 in the MALDI mass spectrum (Figure 3A) showed three distinct signals corresponding to the major glycoforms present in the sample (Table 1). These signals indicated the presence of mono- and disialylated oligosaccharide components since their mass separation approximately corresponded to the incremental mass of the N-acetylneuraminic/sialicacid (NeuAc) unit (291 Da). When the MALDI mass spectrometric analysis of the same CHO IL-4 sample was carried out in neat DHB, only the asialo and monosialylated glycoprotein signals were resolved (Figure 3B). Moreover, there was a shift in the observed glycoform pattern toward the asialo form and the (30)Smith,R.D.;Loo,J.A.;Edmonds,C.G.;Barinaga,C.J.;Udscth,H.R.A~1. Chem. 1990,62,882-899. (31) Tsarbopoulos,A.; Her, G.R.; Pramanik, 8. N.; Trotta, P. P.; Nagabhushan, T. L.AMI. Chem. 1992,64, 2303-2305.



z

E

50.

0

8Ooo

1

m

Mi2

MI2

Flguo 4. Molecular ion regbn in the positive bn PD mass spectrum of the^ F”f3ase F-released oligosaccharide of CHO 11-4 after chromatographic isolation and peracetylatbn.

1/l”

Tabk 2. CHO IL-4 oygOrroohrrld.r 0k.n.d by PD MS

I \

1

0

MH

17013

‘MJ ‘

l

2

m



,-*

lo0l

t

2

15OOO

CalCd composn mass valud (Hex)~-(HexNAc)4-dHex-NeuAc 3302.0 (Hex)5- (HexNAc),-dHcx-( NeuAc)z 3659.3 (Hex)6-(HexNAc)s-dHex-NeuAc 3877.5 (Hex)6-(HexNAc)s-dHex-(NeuAc)~ 4234.8 (Hex)6-(HexNAc)~-dHex-(NeuAc)3 4592.2 (Hex)7-( HexNAc)s-dHex-( NeuAc)z 4810.3 (Hex)t-(HexNAc)6-dHex-(NeuAc)3 5167.7 (Hex)s-( HexNAc)7-dHex-(NeuAc)3 5743.2

i 8Ooo

lodo0

15dOO

2obo

M/z

ObSd

PD value 3301.9 3659.0 3877.2 4234.6 459 1.9 4808.4 5167.3 5738.1

Flguro 5. Positive bn UVMALDI mass spectrum of CHO IL-4 in (A, top) eDHB matrlx and (B, bottom) DHB matrlx. Asterisks in the WE derived mass spectrum (panel B) indlcate signals that are broad and shifted in mass. The slgniRcent peak broadening observed for the adeb cOmpOnent is presumably due to small neutral losses (see text for discmion).

*The calculated isotopically averaged mass value corresponds to the peracctylated and cationized (+Na) molecular ion.

molecular ion peaks were two to three times broader than those obtained in the sDHB matrix. This extensivemetastable loss of sialic acid(s) and/or small neutral molecules (such as NH3 and H2O) observed in neat DHB (Figure 3B) resulted in less accurate mass assignments. Although the sDHB matrix showed a better performance than that of neat DHB, the distribution of the differently sialylated protein components in the sDHB matrix (shown in Figure 3A) does not represent the distribution present in the sample as shown by ES MS;32 this still indicates prominent metastable loss of one and two sialic acids from the major disialylated component. This observation is in agreement with recently reported results,33 and means to overcome this problem will be discussed in a separate publication. The heterogeneity of the carbohydrate component in CHO IL-4 was more apparent in the PD mass spectrum, where the presence of more higher order glycoforms resulted in a higher mass measurement of the glycoprotein (Table 1). Despite the low resolution of the PDMS time-of-flight instrument, insights on the type and variations of the oligosaccharide components were provided by the partially resolved signals of the doubly charged molecular ion, where the most intense signal corresponded to the disialylated component.31 Additional PD signals indicated higher order structures whose signals were either very weak or (surprisingly) absent in the

MALDI mass spectra of CHO IL-4. These additional higher order branching structures were clearly observed in the PD mass spectrum of the PNGase F-released oligosaccharideafter chromatographic isolation and peracetylation,” where distinct signals indicated the presence of additional lactosamine units with mono- and disialylated galactose end residues (Figure 4). The signals at m/z 3301.9 and 3659.0 corresponded to the cationized molecular ions of the mono- and disialylated biantennary oligosaccharide, whereas additional signals indicated the presenceof tri- and tetraantennary structures with three sialic acids as shown in Table 2. In the case of the CHO IL-5 glycoprotein, a disulfidelinked homodimer of -30 kDa, PD mass spectrometric analysis was not successful possibly due to the larger size and/or the higher, by weight, percentage of the carbohydrate component. On the contrary, the MALDI mass spectrum of CHO IL-5 was dominated by intense signals corresponding to the singly and doubly charged molecular ions (Figure 5 ) . The significantbroadening of these signals, compared to those obtained for the CHO IL-4 sample (Figure 3A), reflects the large amount of heterogeneity in the carbohydrate portion of the glycoprotein. Since the individual glycoform signals cannot be differentiated due to the low resolution of the time-offlight instrument used, the measured mass value for CHO IL-5 corresponded to the weighted average MW of all glycoform components composing the molecular ion signals

~

(32) Tsarbopouloa, A.; Pramanik, 8. N.; Huang, E.; Covey, T. In Proceedings of the 4lst ASMS Conference on Mass Spectmmtry and Allied Topics, San Francisco, CA; American Society for M a s Spectrometry: Santa Fe, NM, 1993; pp 101a-101b. (33) Hukrty, M.C.; Vath, J. E.; Yu, W.; Martin, S.A. Anal. Chem. 1993, 65,

2791-2800.

(34) Her, G. R.; Rsmanik, 8.; Kumarasamy, R.; Das,P.; Nagabhushan, T. L; Trotta, P. P.; Tindall, S. H.; Tsarbopoulos, A. In Proceedings of the 38th ASMS Conference on Mass Spectrometry and Allied Topics, Tucson, AZ: American Society for Mass Spectrometry: East Lansing, MI, 1990; pp 13411342.

AnaEyticelChemWY, Vd. 66, No. 13, July 1, 1994

2085

n 0

I

loo00

30000

Figure 5. Posltlve ion UV-MALDI mass spectrum of CHO 11-5 in sM.18 matrix.

(Table 1). Nevertheless, the difference between the MALDIdeduced average mass value and the DNA-calculated mass value for the protein portion (26 294 Da) gave an estimate for the average size of the carbohydrate component attached onto the protein backbone, i.e., 4550 Da. Similarly, the observed MALDI mass value for the IL-5 protein derived from the NS-1 cell line was 31 520 Da (Table l), thus indicating that the size of the attached carbohydrate component was ca. 700 Da higher than that in CHO IL-5, possibly due to an additional O-glycosylation. The other sample where PD MS failed to generate a signal was that of rhIFN-y, a polypeptide containing 138 amino acid residues. The MALDI-derived mass value for rhIFN-y (16 300 Da; Table 1) clearly confirmed the incorporation of an N-terminal initiator methionine to the expected amino acid sequence, while a weak MALDI signal at m / z 15 494 indicated the presence of a truncated form lacking the seven amino acid residue sequence SQMLFRG at the C-terminus. The failure for mass measurement of rhIFN-y by PD MS can be attributed toan “extended” tertiary structure due to the absence of disulfide bonds, or the presence of a considerable C-terminal heterogeneity as reported p r e v i ~ u s l y .In ~ ~support of the second argument, PD mass spectrometric analysis of the rhIFN-y sample was successful, albeit the low intensity of the PD signals, only after the removal of five to seven residues from the C-terminal sequence of rhIFN-y. Even though mass spectrometric analysis of the intact protein is a rapid way not only to assess the integrity and purity of recombinant products, but also to reveal the presence of any post-translational modifications, further examination of the protein’s primary structureis required prior to evaluating its therapeutic utility. Mapping the protein sequence by mass spectrometric analysis of enzyme-generated peptide mixtures provides a rapid, as well as accurate, confirmation of the expected sequence and identification of any existing modifications. In the mapping of the E. coli-derived rhIFN CY-2 and rhIL-4, the proteins were cleaved at the C-terminal side of the arginine and lysine residues with trypsin and the resulting peptide mixtures were analyzed by PD and MALDI MS without any prior removal of buffer or salts. Tryptic mapping of rhIFN 12-2 by PD and MALDI MS revealed about 95% (35) Grcer, F. M.; Morris, H. R.; Fallon, T.; Brewer, S. J. In Proceedings ofthe 35th ASMS Conference on Mass Spectrometry and Allied Topics, Denver, CO; American Society for Mass Spectrometry: East Lansing, MI, 1987; pp 942-943.

2066

Analytical Chemistry, Vol. 66. No. 13, July 1, 1994

of the amino acid sequence, Le., 157 of the 165 residues were mapped (Table 3). As shown in Figures 6 and 7 and Table 3, almost all of the expected tryptic peptide fragments of rhIFN a-2 were observed in the PD and MALDI mapping except those peptide fragments containing one or two amino acid residues, i.e., T2, T4, T6, T12,T15, T16, T20 and T21. However, information on these low-mass fragments was often present as signals arising from incompletecleavage between two contiguoustryptic sites, such as the T6.7 and T12,13 signals (Table 3). Comparison of the mass spectra depicted in Figures 6 and 7 shows that the PD mass spectrum contained slightly more information in the 400-1000-Da mass range (Figure 61, whereas all signals beyond m / z 1000 were more intense in the corresponding MALDI mass spectra (Figure 7). The use of sDHB (Figure 7B) definitely lowered the high baseline level usually observed with DHB (Figure 7A) and enhanced the intensity of the low-mass tryptic peptide signals, thus revealing several of the previously undetectable mixture components. In addition, the much narrower peaks allowed the more accurate measurement of the peptide fragments present in the mixtures, especially when two analyte peaks were used as internal standards (see Experimental Section). One of the striking similarities between the PD and MALDI mass spectra was the presence of a weak Ti4 signal. The weak signal of this highly hydrophilic peptide (Bull and Breese hydrophobicity index of -5430), which yields an intense signal when the tryptic mixture is analyzed by FAB MS,36suggests that the relative or total suppression of peptide signals in the analysis of mixtures by MALDI and PD MS is not related to their relative hydrophobicity. Another similarity between the PD and MALDI (with sDHB) analyses was the high tolerance to the presence of salts, as indicated by the low intensity (and sometimes absence) of cationized molecular ion signals. If the salt and buffer content of the sample is high, its diminishing effect on the PD and MALDI mass spectral signals can be eliminated by rinsing the NC-adsorbed analyte and the dried analyte-matrix deposit, respectively, with deionized ~ a t e r . ~ J ’ The identification and location of the existing disulfide pairing among the four cysteine residues of rhIFN a-2b was readily provided by the signals at m/z 21 19 and 4616 (Table 3), which confirmed the previously established C 1-C98, C29C 138 disulfide arrangement.36.38 Additional confirmation of the former disulfide bond was provided by the T1-T9,lo signal (calculated MW 6047.8) (Table 3) arising from an incomplete cleavage at K83. Moreover, an unexpected cleavage at the C-terminal side of H7 of peptide T1 gave rise to a pair of signals corresponding to the disulfide-bonded peptides T’ITlo and T’l-T9,10 (in combination with the incompletecleavage at K83),thus providing another supportive evidence for the existing disulfide pattern. This unusual cleavage after H7 was further confirmed by Edman sequencing results of the chromatographically isolated fractions containing these peptide fragments.39 From our experience,this cleavage was more prominent when the tryptic reaction was carried out in 0.1 M (36) Pramanik,B. N.; Tsarbopoulos, A.; Lsbdon, J. E.; Trotta, P. P.;Nagabhwhan, T.L. J. Chromatogr. 1991.562, 377-389. (37) Bcavis, R. C.; Chait, B. T. AMI. Chem. 1990, 62, 18361840. (38) Lydon, N. B.; Favre, C.; Bove, S.; Neyrct, 0.;Benurcau, S.; Levine, A. M.; Scclig, G. F.; Nagabhushan, T. L;Trotta, P. P. Biochemistry 1985,24,41314141. (39) Tsarbopoulos, A,; Pramanik, B. N. Personal communication.

Table 3. PD and MALDI MS Analyrk ot the Tryptk Digod d rh-IFN a 9

obsd mass value posn

sequence

calcd mass valu@

PDMS

MALDIMS

1-12 1-7 13 14-22 23 24-3 1 32-33 34-49 32-49 50-70 71-83 84-1 12 1-12 and 84-112 1-12 and 71-112 1-7and84-112 1-7 and 71-1 12 113-120 121 122-125 121-125 126-131 132-133 134 135-144 24-31 and 135-144 24-31and134-144 145-149 150-162 163-164 165

CDLPQTHSLGSR CDLPQTH R TLMLLAQMR R ISLFSCLK DR HDFGFPQEEFGNQFQK DRHDFGFPQEEFGNQFQK AETIPVLHEMIQQIFNLFSTK DSSAAWDETLLDK FYTELYQQLNDLEACVIQGVGVTETPLMK TISS-TIO TISS-Tg,io T'iSS-Tio T'iSS-Tg,i o EDSILAVR K YFOR KYFQR ITLYLK EK

1314.5 813.9 175.2 1077.4 175.2 911.2 290.3 1956.1 2227.4 2460.9 1451.5 3304.8 4616.3 6048.8 4115.7 5548.2 903.0 147.2 613.7 741.9 751.0 276.3 147.2 1210.4 21 18.6 2246.7 619.8 1482.6 234.3 148.1

1313.9 813.1

1314.6

1077.4

1077.5

a Isotopically average

K YSPCAWEVVR TS-Tl7 TSeT16.17 AEIMR SFSLSTNLQESLR SK E

911.4 1955.0 2227.7 2460.8 1451.6

1956.1 2227.0 2461.0 1451.3

4616.5 6050.5 4114.3 5550.8 903.1

4616.2 6049.2 4116.2 5542.7 903.1

613.6 741.9 750.4

613.5 741.7 750.1

1209.6 21 19.0 2246.5 619.4 1483.3

1210.2 2118.6 2247.0 619.4 1482.5

MH+ mass values; no entry means not observed.

4 2

M/Z

Flgure 6. Positive ion PD mass spectrum of the tryptic digest of rhIFN a-2b. The asteriskdenoted peak corresponds to the TrT',, peptlde (Table 3).

phosphate buffer (pH 6.5) instead of ammonium bicarbonate (pH 8.4); in that case the T'I-Tm and T'l-Tg,lo signals, rather than Tl-Tlo and Tl-T9,10, dominate the high-mass region of the mass spectrum. It should be noted that weak ion signals corresponding to the MH+ of the constituent cysteinecontaining peptides, e.g., TI,Tlo, Ts, and TI,, were also present in both PD and MALDI mass spectra. Similar fragmentation has been previously observed in PD8and FAB& mass spectra of disulfide-linked peptides. Tryptic mapping of the E. coli-derived rhIL-4 by PD and MALDI MS produced roughly the same results, since. signals for 16 out of 22 of the expected tryptic peptides were observed (40)Yazdanparast, R.;Andrews, P.;Smith, D.L.; Dixon J. E.A w l . Wochrm. 1986,153, 348-353.

MR Flgure 7. Posltive ion UV-MALDI mass spectrum of the tryptic digest of M F N a-2b in (A) DHB matrix and (B) sDHB matrix.

(Table 4), which represents an 88% coverage of the rhIL-4 sequence(Figure8). Asin thecaseoftherhIFNa-2mapping, all peptide fragments consisting of one up to three amino acid residues were not detected except when incompletecleavages at the expected sites have formed larger peptides, such as the TIZ-14 and T14,15 peptides (Table 4). The cysteinecontaining peptides of rhIL-4 gave rise to "unique" signals at m/z 1735.1,2125.2, and 2942.9 (MALDI massva1ues;Table 4) that did not match with any of the MH+ values for the anticipated peptides or combinations thereof. These signals disappeared upon subsequent treatment with DTT giving rise AnatWCaIChemlsby, Vol. 66,No. 13, Ju& 1, 1994

2067

~

Table 4. PD and MALDI MS A n a m of the Tryptk Dlg.rt of I L 4

peptide

Ti,rT2z T3 T4

TI T; T7 Ts

Tg Ts.9 Tio

TrTio Ti1 Ti2 TI3 TI4

Tiz-14 Ti4.i~

TIS TI6 T6-Ti6 Ts.6-Ti6 TI7 T6-Ti6.17

Ts,6-T16.17

Tis Ti9 T20

Tzi T22 0

posn

calcd mass value

1-12 and 127-129 13-21 22-37 38-42 43-47 48-53 54-61 62-64 54-64 65-75 22-37and65-75 76-77 78-8 1 82-84 85 78-85 85-8 8 86-88 89-102 43-47and89-102 38-47and89-102 103-1 15 43-47 and 89-1 15 38-47 and 89-1 15 116-117 118-121 122-123 124-126 127-129

1735.0 1034.2 1714.0 592.6 655.7 630.8 1076.2 391.4 1448.5 1232.4 2943.4 284.3 529.7 407.5 175.2 1074.4 559.7 403.5 1472.7 2125.5 2699.1 1551.7 3658.1 4231.8 260.4 520.7 276.3 397.4 296.3

obsd PD mass value E. coli IL-4 CHO IL-4 1735.3 1033.8 1713.4

obsd MALDI mass value E. coli IL-4 CHO IL-4

1735.2 1033.9

1735.1 1034.1b

655.6 631.0 NR

Glyco 655.2 630.8 NR

655.4 630.5 1O75.gb

1735.1 1034.0 1713.9 Glyco 655.5 630.4 1076.1

1232.2 2944.1

1232.4 2943.7

1448.4 1231.9 2942.9

1448.4 1232.1 2943.1

529.8

529.8

529.7 407.9

529.3 407.3

1073.9 559.7

1073.6 559.8

559.2b

559.56

1473.1 2125.7 2699.8 1551.9 3658.3 4232.2

1471.9 2125.4 Glyco 1552.2

520.7

Glyco

1472.6 2125.2 2698.8 1551.8 3657.6 4228.9

1471.3 2125.6 Glyco 1551.6 3658.6 Glyco

520.7

520.5

520.2

Isotopically average MH+ mass values; NR means signal was not resolved; no entry means not observed. b Weak signal. 'p

Z?

HKCDITLQEIIKTLNSLTEQKTLCTELWTDIFAASKNTTEKET -Ti-

T2 T -3

TI

Tg -T6-

T4

PEPllDE HK~DITLQEIIK S-S

6? 90 FCRAATVLRQFYSHHEKDTRCLGATAQQFHRHKQLIRnKRL -Ts

-T9

T2a

v

TI,

to the signals of the half-cystine-containing peptides, thus revealing the existing disulfidebond arrangement in the protein molecule as C34127, C24465, and C46499, as shown in Figure 9. It has been previously reported that this disulfide pattern probably contributes to the high thermal stability of rhIL-4,41142 by linking the most spatially distant and flexibIe regions of the molecule. Further confirmation of the C46C99 disulfide linkage was provided by an additional PD and MALDI signal at m/z 2700 (Figure 10). This signal corresponded to the disulfide-linked peptide T5,6-T16 (calculated MW 2698.1) resulting from an incomplete cleavage after K42, which was found to be minimized after extending the tryptic digestion to 22 h with further 1 5 0 (w/w) addition of trypsin (4 pg) after the first 18 h. Another partial cleavage ~~

~

R.;Le, H. V.; Trotta, P.P.Btoehemisrry 1991, 30,

1259-1 264. (42) Walter, M. R.; Cook, W. J.; Zhao, B. G.; Cameron, R. P., Jr.; Ealick, S.E.; Walter, R.L.,Jr.; Reichert, P.; Napbhushan, T. L; Trotta, P.P.;Bum, C. E. J. Bioi. Chem. 1992,267, 20371-20376.

2068

2124.5

2124.7

2124.2

2942.4

2943.1

2941.9

99 2?

T,

TI0

~

1734.1

I NLWGLAGLNSCPVK

T17-TlS.cT19gT2PT21n22~

~

1734.3

S-S

Fiqure 8. Amino acid sequence of rhIL-4 from E. col,indicating ail the expected tryptic peptldes T,

~

ETFFR

im

DRNLWGLAGLNSCPVKEANQSTLENFLERLKTIMREKYSKCSS

(41) Windsor, W. T.; Syto,

1734.0

lZ7

T6

-T15.C--T16

MUD1

I CSS

-T10~ll~T12gT13~14-

le4

PD

3

T , ~

-T7

EXPECTED MW

Am&ticaIChemis&y, Vol. 66, No. 13, Ju& 1, 1004

TL TELTVTDIFAASK

Fs-s

CLCATAQQFHR 65

Figwo B. Amino acid sequence along wlth calculated and observed average molecular weight values of the disuMdelinkedtryptic peptides in rhIL-4.

at K 102 yielded additional signals corresponding to disulfidelinked peptides T6-T16,17 and T5.6-Tl6.17 (Figure lo), which further corroborated the C46C99 disulfide bond. In the tryptic mapping of the IL-4 glycoprotein derived from CHO cells, PD MS provided a spectrum almost identical (Figure 11) with that of the E. coli-derived IL-4 protein. The disulfide information inferred from the PD mapping analysis was the same as that for the E. coli-derived IL-4 and also consistent with the FAB mapping results reported previou~ly.~~ The only mass spectral difference was the absence of the T5.6T16 and T5,6-T16,17 signals in the PD mapping of the glycoprotein. These signals were also absent in the MALDI spectrum of the CHO IL-4 tryptic digest (Figure 12A), thus indicating potential glycosylation at the N38 residue. This

A 5288

ME Flguro 10. Positive Ion UV-MALDI mass spectrum of the tryptic digest of E. colMerived IL-4 in sDHB matrix.

7doo

MR Figure 12. PosMve ion UV-MALDI mass spectrum of the tryptic digest of CHO IL-4 In sDHB matrix (A). Panel B shows the higbmass reglon of the spectrum containing the glycopeptideddved signals (sum of 200 laser shots). olycopeptkle signals A-C are defined In the text,

wh#etheasterisk-andtrie~signalslndicatethelncorporatkn of additional NeuAc and lactosamine units, respectively (seetext for discussion).

ME

Flguro 11. Positive lon PD mass spectrum of the tryptlc digest of CHO IL-4. The intensity of the asterlsk-denoted signal (TIS),as well as its mass assignment, was improvedafter thorough washing of the sample at the expense of the low-mass peptide signals.

residue is part of the triplet N-T-T which fulfills the usual carbohydrate acceptor consensus sequence N-XS/T (where X can be any amino acid except possibly proline and aspartic acid) for N-linked carbohydrates. The glycosylation of N38, rather than the other potential glycosylation site of N 105, has been shown previously34by a comparative FAB and PD mass spectrometric analysis of the V8 digest mixture prior to and after enzymatic digestion with PNGase F. The glycosylation of the N38 residue was nicely corroborated by the appearance of new glycopeptide signals in the high-mass (>4000 Da) region of the MALDI mass spectrum (Figure 12B). It is notable that these signals were not observed in the PD mass spectrum of the CHO IL-4 tryptic digest, thus indicating that desorption and/or ionization of the glycopeptides were totally suppressed when these digest mixtures were analyzed by PD MS. It seems that the presence of carbohydrate units on the N38 residue of peptide T5 may partially shield the adjacent trypticsite (K37) from proteolysis, thus resulting in the incorporation of the Ts glycopeptide (referred to as Tp) into the adjacent peptide T4, which in turn is part of the T4-Tlo disulfide-linked peptide. The MALDI signal A at m / z 5286 (Figure 12B) corresponded to the T4.5.Tlo disulfide-linked glycopeptide containing the asialo biantennary oligosaccharide (calculated MW 5286.6), and this signal was still present in the analysis of samples exposed to longer digestion times. Similarly, partial inaccessibility of K42 to proteolytic cleavage led to the incorporation of the Ts*

glycopeptide into the other neighboring disulfide-linked peptide; the resultant T5*,6-T16 disulfide-linked glycopeptide (calculated MW 4468.7) gave the MALDI signal B at m/z 4468 (Figure 12B). When both the aforementioned incomplete cleavages at the lysine residues K37 and K42 occurred simultaneously due to the presence of the neighboring sugar, combination of the Ts. glycopeptide with both the adjacent disulfide-bonded peptides led to the formation of a disulfide "core" glycopeptide with calculated MW 7393.1 (signal C in Figure 12B). Each of theseglycopeptide-derivedsignals were accompanied by several low-intensity peaks indicating the presence of additional glycoforms for this N-linked oligosaccharide. For example, theasterisk- and triangle-denoted weak signals ca. 291 and 365 Da higher than the corresponding A-C signals, respectively, indicated the addition of NeuAc and lactosamine units. Even though the low signal-to-noise ratio of these signals prevented a more accurate mass assignment, these findings are in agreement with the results inferred from the PD mass spectrum of the isolated carbohydrate component after peracetylation (Table 2). It should be noted that the glycopeptide-relatedMALDI signals shown in Figure 12B were summed over 200 laser shots, rather than the normally accumulated 20 laser shots, in order to improve the statistics of the resulting signals. Nevertheless, the use of the sDHB matrix definitely improved the peak shape of these glycopeptide signals, and enhanced their intensity especially in the case of the previously undetectable satellite signals of A-C. Similar enhancement of both resolution and intensity of peptide signals has also been demonstrated by using a carbohydrate matrix4' and a carbohydrate comatrix with DHB or other UV-absorbing compound." Even though the generation of these weak glycopeptide signals provides only an insight on the carbohydrate chains (43) K(bter,C.;Cartoro, J. A.; Wilkins, C. L.J. Am. Chem. Soc. 1992,114,15127514. (44) Billeci, T.hi.; Stults, J. T.And. Chem. 1993,65, 1709-1716.

AnaWcal Chemistry, Vol. 66, No. 13, Ju& 1, 1994

2069

attached to the protein backbone, it clearly demonstrates the analytical utility of MALDI MS in detecting glycopeptides from direct analysis of glycoprotein digest mixtures.

CONCLUSION MALDI was successful in mass measuring all the recombinant proteins of this study, whereas PD failed in the case of rhIFN-y and the higher mass protein IL-5. The MALDI mass measurement accuracy was two to five times better than that typically achieved by PD MS, at least for the samples where both techniques were successful. The higher mass capabilities of MALDI were demonstrated in the analysis of the IL-5 glycoproteins,and the spectral improvementresulting from the use of the sDHB matrix was illustrated in the analysis of the CHO IL-4 and IL-5 glycoproteins, as well as for several other highly glycosylated protein receptor molecules (>35 kDa) analyzed in our laborat~ry.‘~A significant advantage of the MALDI technique over that of PD was the remarkably short analysis time, usually requiring only a few minutes of analysis in order to obtain a good-quality MALDI mass spectrum. Even though sensitivitieswere highly compound dependent, the amount of material required for MW determination by PD MS was in the 300 pmol to 1 nmol range, whereas low femtomole to low picomole quantities were required for the MALDI mass spectrometric analyses. These amounts of proteins were also required per sequence map, i.e., tryptic digestionof the protein followed by mass spectrometricanalysis of the digest mixture. A notable difference between the PD and MALDI mass spectra was that peptidesignals below m / z 1500were more intense in the PD mass spectra, whereas those beyond m / r 1500 were more intense in the MALDI mass spectra. Higher mass signalscorresponding to disulfide-linked peptides or glycopeptides,usually in the m/z4000-8000 range, were significantly more intense in the MALDI rather than the PD mass spectra. The latter spectra were devoid of glycopeptide signals, thus indicating that desorption and/or ionization of the glycopeptide components were totally suppressed when these digest mixtures were analyzed by PD MS. Both the PD and MALDI techniques tolerate high concentrations of salts and buffers, as evidenced by the low (45) Tsarbopoulos, A.; Kam, M. Unpublished rcaults. (46) Bcavis, R. C.; Chait, B. T. Proc. Natl. Acad. Sei. U S A . 1990,87,6873-6877. (47) Caprioli, R. M.;Whaley. B.; Mock,K. K.; Cottrell, J. S.In Techniques in Protein Chemistry M Villafranca, J. J., Ed.;AcademicPress: San Diego, CA, 1991; pp 497-510. (48) Cotter, R. J. Biomed. Enoiron. Mass Spectrom. 1989, 18, 513-532. (49) Ingendoh,A.; Karas, M.; Hillenkamp,F.;Giarwnann,U. In?.J. MassSpectrom. Ion Processes 1994, 131, 345-354. (50) Chait, B. T.;Chaudhary, T.;Field, F. H. Biochem. Biophys. Res. Commun. 1986, 134,420-426. (51) Tsarbopoulos. A. Peptide Res. 1989, 2. 258-266.

intensity of the cationized molecular ion signals. The use of the sDHB matrix in the protein mapping experiments resulted in an enhancement of the MALDI mass spectral signals and an improvement in the mass measurement accuracy of the peptide signals compared to that obtained with the DHB matrix. The latter was the result of thegeneration ofrelatively narrow and intense peaks (when sDHB was used as MALDI matrix), which also allowed the use of two analyte peak centroids as internal standards. In addition, the role of sDHB on the spectral improvement was more clearly illustrated in the analysis of the CHO IL-4 digest mixture, where several glycopeptide-relatedsignals were successfullydetected; these signals were barely detected with the DHB matrix while they were absent in the PD mass spectra. The mapping of CHO IL-4 from direct analysis of the tryptic digest showed that MALDI MS represents an extremely useful analytical tool for detecting high-mass glycopeptides from direct analysis of glycoprotein digest mixtures, which has been a limitation for the other desorption techniques of PD and FAB/LSI. The high-mass capability of the MALDI MS is also an invaluable asset in deducing the sequence order of the peptide fragments from time-course proteolytic reactions of the protein, as demonstrated previously.46,4’ All the advantages of the MALDI technique, Le., high sensitivity, high mass range, and the good mass accuracy especially when coupled to a reflector anal~zer,4**~~ combined with the recent introduction of severalcommercial instruments will ultimately render the MALDI approach ideal for mapping and microheterogeneity checking of recombinant protein products at the low femtomole level. Furthermore, the nondestructive nature of the MALDI technique allows the already analyzed sample to be further subjected to another chemical or enzymatic analysis, such as DTT reduction or glycosidasedigestion, either after the sample’sextraction from the MALDI target or in situ as previously reported for PDanalyzed ~ a m p l e s . ~ ~ * ~ ~

ACKNOWLEDGMENT We thank P. Reichert and A. Hruza for providing the samples of purified proteins and Dr. G. R. Her for the sample of the peracetylated oligosaccharide isolated from CHO IL4. Preliminary reports of this work were presented at the 4th Sanibel Mass SpectrometryConference,January 25-28,1992, Sanibel Island, FL, and at the “Desorption ’92” Conference, September 6-9, 1992, Burg Waldeck, Germany. Recehmd for review December 6, 1993. Accepted March 22, 1994.@ *Abstract published in Advance ACS Abstracts, May 1, 1994.