Protein sorting by high performance liquid chromatography. 2

Cynthia P. Quan, Sylvia Wu, Nancy Dasovich, Chung Hsu, Tom Patapoff, and Eleanor Canova-Davis. Analytical Chemistry 1999 71 (20), 4445-4454...
0 downloads 0 Views 2MB Size
Anal. Chem. 1994,66, 335-340

Protein Sorting by High Performance Liquid Chromatography. 2. Separation of Isophosphorylates of Recombinant Human DNase I on a Polyethylenimine Column J. Frenz,'*t#sC. P. Quan,t J. Cacia,t C. Dernocko,*~llR. Brldenbaugh,s and T. McNerneys Departments of Medicinal Analytical Chemistry, Quality Control, and Manufacturing Science, Genentech, Inc., 460 Point San Bruno Boulevard, South San Francisco, California 94080

Anion exchange HPLC with a polyethylenimine(PEI) column separatesrecombinant human deoxyribonucleaseI (&DNase) glycoforms according to the extent and positions of phosphorylation of mannose residuesin N-linked oligosaccharides.The separationprovides a selectivityunavailable by anion exchange HPLC with other columns or by isoelectric focusing gel electrophoresis and can be used to quantify the phosphate content of preparations of &DNase. Tryptic mapping of fractions collected from the column and treated with alkaline phosphatase was used to identify the sites of phosphorylation. Unnatural forms of &DNase, bearing oligosaccharide structures at only one of the two sites of glycosylation,were prepared by cleaving the phosphate-containinghigh mannose and hybrid structuresfrom the purified isophosphorylatesfractionatedon the PEI column. The separationof &DNase isophosphorylates on the PEI column mimicsthe relative affinitiesfor the mannose 6-phosphatereceptor that traffics acid hydrolasesto lysosomes and provides a useful example of protein sorting by biomimetic interaction chromatography. Improvementsin resolutionof a separation technique accrue from gains in both the efficiency of the method, as represented for example by high performance liquid chromatography (HPLC), and the selectivity of the separation medium. Classical affinity chromatography' is one example of the achievement of high resolution separations by exploiting a highly selective interaction, that of a protein for its biological ligand. In certain cases, separations of interest can only be achieved by exploiting such interactions, or the interactions between a protein and an analog of the natural ligand or between a protein and an ersatz ligand that mimics a particular biological interaction. These modes of separation,which can be termed biomimetic interaction chromatography (BIC),2 facilitate the resolution of closely related proteins that differ on the basis of a specific, defined structural feature that is the object of interest. Since proteins are heterogeneous amalgams of diverse chemical groups, selective separation systems such as BIC simdifv in many instances and make tractable the + Medicinal and Analytical Chemistry. t

Quality Control.

I Manufacturing Science. I Current address: Department of Regulatory Affairs, Genentech, Inc., 460

Point San Bruno Blvd., South San Francisco, CA 94080. (1) Cuadrecasas, P.; Wilchek, M. Biochem. Biophys. Res. Commun. 1968, 33, 235-9. (2) Cacia. J.; Quan, C. P.; Vasser, M.; Sliwkowski, M.; Frenz, J. J. Chromatogr. 1993, 634, 229-39. 0003-2700/94/0366-0335$04.50/0 0 1994 American Chemical Society

desired fractionation of complex mixtures. Chromatography with relatively promiscuous nonselective stationary phases, on the other hand, may not provide sufficient resolution of the structural feature of interest to yield rapid analysis or purification. This report describes a useful separation of glycoforms of recombinant human deoxyribonuclease I (rhDNase) differing in extent and position of phosphorylation of mannose residues and illustrates the power of protein sorting by BIC. rhDNase, an endonuclease that nonspecificallyhydrolyzes double-stranded DNA to oligonucleotides, is currently undergoing clinical investigation for the treatment of cystic f i b r ~ s i s .The ~ bovine form of DNase I (bDNase) has been the object of many investigations that are classics of protein c h e m i ~ t r y . ~The , ~ human protein expressed in Chinese hamster ovary (CHO) cells differs from the bovine form in being glycosylated at asparagine- 106, a site that is not glycosylated in bDNase although it includes the consensus Asn-Xxx-Ser ~ e q u e n c efor ~ . ~N-linked glycosylation. The oligosaccharides of rhDNase contain sialic acid as do those on bDNase, but rhDNase also bears mannose 6-phosphate (Man-6-P) residues that have not been reported on b D N a ~ e .Mannose ~ residues in N-linked carbohydrates are phosphorylated in the Golgi apparatus: for targeting to ly~osomes.~*~ Man-6-P is the ligand that binds to the two Man-6-P receptors1&12involved in this well-characterized cellular protein trafficking system. The presence of both sialic acid and phosphate at both of the glycosylation sites in rhDNase complicates the characterization of the charge heterogeneity of the purified enzyme. Sialic acid in glycoproteins is found to varying extents on complex-type and hybrid oligosaccharides: while Man-6-P occurs, also to varying extents, on high-mannose and hybrid glycan^.^.^ rhDNase contains two glycosylation sites, either of which can, in principle, be occupied by complex oligosac(3) Shak, S.;Capon, D. J.; Hellmiss, R.; Marsters, S.A.; Baker, C. L. P m . Narl. Acad. Sci.(U.S.A. 1990, 87, 9188-92. (4) Liao, T.-H.; Sslnikow, J.; Moore, S.;Stein, W. H. J. Biol. Chem. 1973, 248, 1489-95. ( 5 ) Moore, S.In The Enzymes, 3rd ed.; Boyer, P. D., Ed.; Academic Prcas: N e w York, 1981; Chapter IS. (6) Kornfeld, R.; Kornfeld, S.Annu. Reu. Biochem. 1985, 54, 631-64. (7) Frenz, J.; Cacia, J.; Quan, C. P.; Pai, R. Submitted for publication in J. Bfol. Chem. (8) von Figura, K.; Hasilik, A. Annu. Reu. Biochem. 1986, 55, 167-93. (9) Kornfeld, S.FASEB J . 1987, I , 462-8. (10) Kornfeld, S . Annu. Reo. Biochem. 1992, 61, 307-30. (11) Tong, P. Y.; Gregory, W.; Kornfeld, S.J . B i d . Chem. 1989, 264, 1962-9. ( 1 2 ) Tong, P. Y.; Kornfeld, S. J . B i o l . Chem. 1989, 2 6 4 , 7970-5.

Ana&ticalChemistty, Vol. 88, No. 3,February 1, 1994 335

charides containing from zero to four sialic acid residues, high mannose glycans with zero to two phosphate groups, or hybrid sugars containing some combination of the two. A large number of permutations are possible, engendering the considerablecharge heterogeneity of rhDNase that is a feature of many glycoproteink. In this report, we describe the separation of glycoforms of rhDNase that differ in the extent and position of mannose phosphorylation. The separation was achieved by anion exchange HPLC with a polyethyleneimine bonded phase, a column that we have found to resolve the isophosphorylates of rhDNase with high efficiency.

EXPERIMENTAL SECTION Materials. rhDNase was expressed in CHO cells and purified at Genentech. Details of the cloning and properties of rhDNase are described el~ewhere.~ Anion Exchange HPLC. DEAE. Samples of 100 pg of rhDNase were loaded onto a TSK DEAE-5PW column (7.5 X 75 mm, 10pm particle size) using a Hewlett-Packard 1090M HPLC system with a column temperature of 40 OC and a flow rate of 1 mL/min. The column was equilibrated with a buffer consisting of 50 mM NaCl, 5 mM 2-(N-morpholino)ethanesulfonic acid (MES), and 1 mM CaCl2 at pH 6.0. Following injection of the sample, a linear gradient increasing the NaCl concentration to 200 mM over 100 min was started. The absorbance of the column effluent was monitored at 280 nm. PEL Samples of 0.02-1.0 mg of rhDNase were loaded onto an Interaction Chemicals, Inc. HP-PEI column (7.8 X 100 mm, 10 pm particle size) using a Hewlett-Packard 1090M HPLC system and run at ambient temperature and a flow rate of 1 mL/min. The column was equilibrated with a buffer consisting of 100 mM NaCl, 10 mM N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid (HEPES), and 1 mM CaC12 at pH 7.0. Following injection of the sample, a linear gradient increasing the NaCl concentration to 1.5 M NaCl over 47 min was started. The absorbance of the column effluent was monitored at 280 nm. Microphosphorous Analysis. Phosphate in rhDNase samples was determined by a modification of Bartlett'sI3 procedure for phosphorus determination. rhDNase samples were hydrolyzed in 10N sulfuric acid and oxidized with 30% hydrogen peroxide (Sigma) at high temperature (>220 "C). Inorganic phosphate was complexed with an ammonium molybdate solution (Sigma) using an aminonaphtholsulfonate reagent (Orbeco). Sample color development was assayed in a spectrometer at 820 nm and compared to a standard curve generated from a phosphorus standard solution (Sigma). Isoelectric Focusing Gel Electrophoresis. A 0.4 mm thick gel was poured onto a plastic Gel Fix backing (Serva, Heidelberg, Germany), using an ultra thin layer horizontal casting tray (Bio Rad, Richmond, CA). The solution contained 4 mL of 29: 1 acrylamide-bisacrylamide mixture (Bio Rad), 24.25 mL of water, 300 pL of pH 3-10, 600 pL of pH 3-4, and 600 pL of pH 3-5 Ampholytes (Serva), 200 pL of 10% ammonium sulfate (Bio Rad), and 25 pL of N,N,N',N'-tetramethylethylenediamine(TEMED) (Bio Rad). hDNase samples (20 pg) were loaded, and the gel was focused in a Model 2217 Ultrophor (LKB) electrofocusing unit for 2

-

(13) Bartlett, G. R. J. Biol. Chem. 1959, 234, 466-8.

336 AnalyticalChemistry, Vol. 66,No. 3, February 1, 1994

h at 1600 mV with a 10-W limit. Following electrophoresis, the gel was fixed with 10% trichloroacetic acid and 5% sulfosalicylic acid prior to staining in 0.1% R250 Brilliant Blue dye (Bio Rad), 40% methanol, and 0% acetic acid for 30 min. Ampholytes were removed from the gel by soaking for 20 min in 40% methanol, 8% acetic acid, and 0.5% copper sulfate solution. The gel was further destained in 20% ethanol and 8% acetic acid in the presence of polystyrene foam blocks. The intensity of staining of bands in the gel is not sensitive to the latter treatment. Tryptic Mapping. Aliquots of rhDNase were diluted 1:l with 40 mM Bis-Tris-10 mM [ethylenebis(oxyethylenenitri1o)ltetraacetic acid (EGTA), pH 6.0, and incubated at 37 OC for 1 h. After incubation, 1 mg of rhDNase was exchanged by PD-10 (Pharmacia) into 100 mM Tris-HC1, pH 8.0. The samples were then incubated at 37 OC for 4 h with additions of 10 pg of TPCK-treated trypsin (Worthington) at 0 and 2 h. Samples were frozen at -70 "C until analyzed by HPLC. The tryptic peptides were chromatographed by reversed-phase HPLC using a Hewlett-Packard 1090M with a Nucleosil C 18 (Alltech) column (2.1 X 150 mm, 5 pm particle size, 100 A pore size) equilibrated in 0.1% aqueous TFA. A 0-60% linear gradient of 0.1% TFA in acetonitrile was run in 1 h at a flow rate of 0.25 mL/min. The effluent was monitored at 214 nm with the column compartment maintained at 40 OC. Neuraminidase Digestion. Samples were exchanged into digest buffer consisting of 200 mM sodium acetate-2 mM CaC12, pH 5.6, using a Centricon-10 device. Neuraminidase (Sigma) was added to a ratio of 0.3 unit/mg of rhDNase. Samples were incubated at 37 OC for 18 h, and the digestion stopped by freezing at -20 OC. Alkaline Phosphatase Treatment. rhDNase samples were exchanged into digest buffer consisting of 50 mM Tris-HCI-2 mM CaC12, pH 8.2 (37 "C), using a Centricon-10 (Amicon) device. Alkaline phosphatase (Sigma) was added to a ratio of 7.4 units/mg of rhDNase. Samples were incubated at 37 OC for 6 h and frozen at -20 "C to stop digestion. Endo H Digestion. Samples were exchanged into digest buffer containing 100 mM sodium acetate-1 mM CaC12, pH 6.0, using a Centricon-10 (Amicon) device. Endo-0-Nacetylglycosaminidase H (Boehringer-Mannheim) was added to a ratio of 50 milliunits of enzyme/l mg of rhDNase. Samples were incubated at 37 OC for 18 h, when the digest reaction was stopped by freezing and storage at -20 OC.

RESULTS AND DISCUSSION IEFof DNase. The occurrence of varying amounts of sialic acid and/or Man-6-P on the oligosaccharides attached at two positions in rhDNase imparts considerable charge heterogeneity to the purified enzyme. Figure 1 illustrates the charge heterogeneity, showing, in lane c, the banding profile of rhDNase in a narrow-range isoelectric focusing (IEF) gel. rhDNase is a relatively acidic glycoprotein, as is bDNase,14 with an isoelectric point (pl) between 3 and 4. Both sialic acid and Man-6-P have pK,'s less than 2.5 and so contribute to the acidic character of rhDNase. The banding ladder observed in IEF analysis of rhDNase is due to the presence of both sialic acid and phosphate, as shown by the analyses of enzymatically treated rhDNase shown in lanes d-f of Figure

I

121

4.55

4.15

33

Time. min Flgure 2. Analysis of rhDNaSe samples by anion exchange HPLC whh a DEAE column. PTior to injection, samples were (A) untreated, (6) digested with neuraminidase,or(C)digestedwith alkaline phosphatase. Chromatographic conditions are described In the text.

a

a

h

50-

E

C

d

e

l

Flgure 1. IEFgelelectrophoresisofrhDNasesamples.Priortoloading the gel. rhDNase aliquotswere (lanec)untreated, (laned)digested with neuraminidase. (lane e ) digested with alkaline phosphatase. or (lane f) serially digested with neuraminidaseand alkaline phosphatase. Lane acontains 1 pgofalkalinephosphatase.Lanebcontainstheplrnarkers designated at the left of the gel.

I . Treatment of rhDNase with neuraminidase, an exoglycosidasethatcleaves sialic acid residues fromoligosaccharides, results in loss of the most acidic bands in the profile obtained in IEF, and a concomitant intensification of the more basic bands, as shown in lane d of Figure 1. Alkaline phosphatase treatment of rhDNase does not affect the most acidic bands in the pattern, as shown in lane e of Figure 1, but intensifies the more basic bands in the pattern. The basic buffer conditions employed for alkaline phosphatase digestion also cause deamidation of rhDNase,2 an artifact evinced by the doubling of the bands in the IEF pattern since deamidation adds to the charge heterogeneity of the protein. Lane f shows the IEF analysis of rhDNase that has been treated with both neuraminidase and alkaline phosphatase, substantially eliminating the residual banding multiplicity exhibited by rhDNase treated with only one enzyme. These results indicate that (i) both sialic acid and phosphate contribute to the charge heterogeneity of rhDNaseand (ii) that IEFgel electrophoresis does not resolve rhDNase isoforms containing Man-6-P from thosecontaining sialic acid, although it does provide sufficient resolution to separate isoforms differing in the number of charged residues attached to oligosaccharides. Anion Exchange HPLC of DNase. The banding pattern observed for rhDNase in IEF analysis is not reproduced by anion exchange HPLC with a diethylaminoethyl (DEAE) column, as shown in Figure 2. rhDNase elutes as a broad peak from the DEAE column, shown in Figure 2A. Neuraminidase-treated rhDNase elutes from the DEAE column as two broad, overlapping peaks, seen in Figure 28, the less retained of which does not entirely coelute with untreated rhDNase. rhDNase treated with alkaline phos-

$

40-

C

0

m N

30 -

mu c

2

20-

2

10-

E

A 00

10

20

30

40

50

Time. min Flgure 3. Analysis of rhDNase samples by anion exchange HPLC wlth aPEl column.Priutoinjection,samples were(A)untreated,(B)di~sted with neuraminidase, or (C) digested with alkaline phosphatase. Five peaks were manually collected from the untreated sample and numerically designated as shown In (A). Chromatographicconditions are described in the text. phatase, shown in Figure 2C, exits the column as 11or more overlapping peaks, suggesting that retention on the column is sensitivetovariations in thestructuresof the oligosaccharides on the protein that are more subtle than simply the number of sialic acid residues present. Comparison of the retention behaviors of treated and untreated rhDNase samples on the DEAE column indicates that thiscolumn exhibits no particular selectivity for either sialic acid or Man-6-P residues, and that it is probably selective for subtle structural differences among glycoforms of rhDNase, but that it lacks sufficient efficiency to resolve individual species. Another weak anion exchange HPLC column, however, separates rhDNase into a simpler series of peaks according to a selectivity that can be established by enzyme treatments, as shown in Figure 3. Figure 3A shows the five partially resolved peaks obtained by analysis of rhDNase with a polyethylenimine (PEI) anion exchange column. Treatment of rhDNase with neuraminidase has little effect on the profile obtained, inducing only a shift to slightly earlier retention of the first three peaks in the pattern, as shown in Figure 3B. rhDNase treated with alkaline phosphatase, however, elutes from the PEI column as a single peak, shown in Figure 3C, Anal)rtical Chemistry, Vol. 66,NO. 3, February I . 1994 337

Table 1. Phorphorus Content ot rhDNase Imphoophorylater Separated by PEI Chromatography

sample

mol of phosphate/mol of rhDNase

column load fraction 1 fraction 2 fraction 3 fraction 4 fraction 5

0.8 0.0 0.9 1.0 1.8 3.0

that coelutes with the earliest eluting peak in the pattern obtained for untreated rhDNase. These results suggest that the PEI column is relatively insensitive to sialic acid differences among rhDNase glycoforms, but that it exhibits a strong selectivity for phosphate differences among these species. Microphosphorous Analysis of PEI Fractions. The separation of rhDNase variants according to phosphate content was examined by collection of the individual peaks .labeled 1-5 in Figure 3A and quantitative analysis by chemical means of phosphate in the fractions. Table 1 shows the results of phosphate analyses, confirming that retention of rhDNase on the PEI column is governed by phosphate content. The earliest eluting peak was found to contain no phosphate, the next two peaks to each contain 1 mol of phosphate/mol of rhDNase, the fourth peak to contain -2 mol of phosphate/mol of rhDNase, and the fifth peak to contain 3 mol of phosphate/ mol of rhDNase. Hence, the retention order of isophosphorylates on the PEI column parallels the binding affinities of these species for the Man-6-P receptor.'&'* Assignment of these phosphate contents to the individual peaks in the chromatogram permitted the calculation, by integration of peak areas, of the average phosphate content of the sample of rhDNase analyzed in Figure 3A. The resulting value, 0.80 mol of phosphate/mol of rhDNase, is in exact agreement with thevalue determined by chemical analysis of the sample shown as the starting material in Table 1. Hence, the analysis of rhDNase on the PEI column is one method for quantitative estimation of the phosphate content of rhDNase samples. IEF of DNase and PEI Fractions. The separation of phosphate variants of rhDNase provides a selectivity that is not obtainable by IEF gel electrophoresis. Figure 4 shows the analysis by IEF of fractions collected from the PEI column. Lane c of Figure 4 shows the IEF analysis of peak 1, which contains sialylated, nonphosphorylated forms of rhDNase and exhibits a banding pattern similar to that of unfractionated rhDNase, shown in lane b. The monophosphorylated fractions, peaks 2 and 3, show a narrower range of bands in IEF analysis, as seen in lanes d and e. These species esentially lack both the most basic band in the pattern that evidently contains no charged sugars, as well as the most acidic, heavily sialylated glycoforms. The absence of the uncharged band is due to the presence of at least one Man-6-P residue in these fractions from the PEI column. The absence of the most acidic bands reflects that the presence of a high mannose or hybrid structure at one of the glycosylation sites (indicated by the presence of phosphate) precludes the occupation of that site by a complex, highly sialylated structure. The most acidic bands in the IEF pattern obtained for rhDNase thus represent protein in which both glycosylation sites are occupied by complex oligosaccharides, each bearing several residues of sialic acid. Peak 4 from the PEI column, shown in lane f of Figure 4, contains

-

-

338 Analytical GhetnMy, Vol. 66,No. 3, Februery 1, 1994

2 mol of phosphate/mol of rhDNase and lacks the two most basic bands in IEF analysis by comparison with untreated rhDNase. It also exhibits a narrower distribution of bands than the samples shown in lanes b-e, owing to the reduced potential for sialylation entailed by the increased extent of phosphorylation observed in this subpopulation of rhDNase glycoforms. Finally, lane g of Figure 4 shows that peak 5 forms a single prominent band in IEF, reflecting the apparent absence of charged sugars other than Man-6-P on these rhDNase glycoforms, which results from occupation of both glycosylation sites, apparently, by phosphorylated oligosaccharides. Lanes i-m of the IEF gel shown in Figure 4 contain neuraminidase-treated aliquots of the fractions collected from the PEI column. The desialylated samples exhibit a prominent band that represents the particular rhDNase glycoform containing the specified number of phosphate groups, with no additional sialic acid. Since the only charge differences among the samples shown in lanes i-m of Figure 4 arise from the defined number of phosphate groups, we can assign numbers of carbohydrate-associated charges to each of the bands occurring in the IEF pattern of rhDNase. The most basic band in the pattern of the sample shown in lane i represents the glycoform lacking both sialic acid and phosphate, since fraction 1 contains no phosphate (Table 1) and has been disialylated with neuraminidase. The next more acidic band in the pattern represents rhDNase forms containing only one charged sugar residue, corresponding to the desialylated samples shown in lanes j and k. The next more acidic band then contains two charged sugars, the next more acidic band three charged sugars, and so on down the ladder of bands observed in the unfractionated rhDNase pattern. The combination of fractionation on the PEI column and neuraminidase treatment, which yields rhDNase species exhibiting a defined number of charged carbohydrate residues, thus permits the assignment of identities to the bands obtained in IEF analysis and facilitates characterization of the charge heterogeneity observed in the glycoprotein. Tryptic Mapping of Fractions. Tryptic mapping has been employed to identify the sites of glycosylation in rhDNase and, together with endo-j3-N-acetylglycosaminidaseH (endo H) digestion, is a useful tool to determine the sites of phosphorylation of the fractions collected from the PEI column. Digestion of rhDNase with trypsin yields, in principle, the peptides shown in Table 2. Two of the peptides, T3 and T9, are glycosylated, containing asparagine residues 18 and 106, respectively. In practice, T3 is observed in the tryptic map, but the peptide bond between T9 and T10 is not cleaved under the conditions employed for mapping of rhDNase, so the longer glycopeptide T9-10 is encountered. In order to identify the sites of phosphorylation of the fractions collected from the PEI column, tryptic maps were madeof the individual fractions with and without prior endo H treatment. Since endo H removes high mannose and hybrid oligosaccharides, including those containing Man-6-P, it causes a change in the peak shape of phosphate-containing peptides that can be observed by comparison of the tryptic maps of treated and untreated samples. Figure 5 shows, for illustration, the changes in peak shape observed in phosphate-containing variants of the T3 and T9-10 glycopeptides following endo H treatment. The rhDNase sample shown in Figure 5 was the diphosphorylated

4.55

4.15

3.5

a

b

c

d

e

f

g

h

i

j

I

k

m

n

Flgure 4. IEF gel electrophoresis of peaks collected from PEI anion exchange separation of rhDNase isoforms shown in Figure 3. Lanes band h contain unfractionated rhDNase. Lanes c-g contain untreated aliquots of peaks 1-5. respectively, collected from the PEI column. Lanes I-m contain neuraminidasedigested aliquots of peaks 1-5. respectively. Lanes a and n contain the p l markers indicated. 150

Table 2. Peptldes Expected TO Be Produced on Digestion ol rhDNaM wRh Trypsin.

identifier T1 T2

residue no.

T8 T9

1-2 3-15 16-31 32-41 42-50 51-13 14-11 18-79 80-111

T10 T11 TI2 T13

112-111 118-121 122-126 121-151

T3 T4 T5 T6 TI

T14-Tl5

158-185

T16 TI1

186213 214-222 223-260

sequence 600

LK IAAFNIQTFGETK MSN'ATLVSYIVQILSR YDIALVQEVR DSHLTA~~GK

LLDNLNQDAPLi ""'HYI NSYK

ll

1 T9-lO(+GicNAC)

I

I I

0

WSEPLGR

N

ER

YLFWRPDQVSAVDSYYYDDGCEPCGN*DTFNR

LWTSPTFQWLIPDSADTTATPTHCAYDR

WAGMLLR GAVVPDSALPFNFQAAYGLSDQLAQAISDHYPVEVMLK

a Amino acid residues in the sequence are denoted by the singleletter code. Disulfide bonding patterns are indicated. Sites of N-linked glycosylation are denoted by asterisks.

protein that elutes as peak 4 from the PEI separation shown in Figure 3A. After endo H treatment, the tryptic map in Figure 59 reveals that both glycosylation sites contain endo H susceptible, Le., high mannoseand hybrid, oligosaccharides. Bothsites in thediphosphorylated sample alsocontain complex structures as shown by the residual unsusceptible glycopeptides. Since complex structures are not phosphorylated and a diphosphorylated rhDNase sample was employed for this example, Figure 5 shows that each glycosylation site can contain up to two phosphate structures. By this approach, the sites of phosphorylation of fractions collected from the PEI column shown in Table 3 were assigned. The results show that thePEI column resolves rhDNase isophosphorylates according to the position of phosphorylation as well as

50

45

40

Time. mi" Flgure 5. Portions of the tryptlc maps of (A) untreated and (E) endo H treated rhDNase samples illustratlng the approach used to Mentlfy sites of phosphorylation. The maps shown here were made from the diphospholylatedrhDNase collected as peak 4 from a separation such as that shown in Figure 3A. The elution positions of the T9-10 and T3 glycopeptidesthatelutefromthecoiumnasbroadpeaksareindicated. The peaks corresponding to the T9-10 and T3 peptMes containing only a single GicNAc residue, formed by endo H treatment. are also shown. Table 3. Dlstrlbutlm of Man-6-P Resldues In rhDNaM Isophosphorylates Separated by PEI Chromatography . . .

no. of Man-6-P residues-

peak no. 1 2 3 4 5

at Asn-18

at Asn-106

0

0

1

0 1 0-2 1-2

0 0-2 1-2

total 0 1 1 2 3

according to the number of Man-6-P residues. Hence, peak 2 in the rhDNase chromatogram shown in Figure 3A contains Man-6-P in theoligosaccharidestructure at Asn-1 8 only, while peak 3 has a Man-6-P residue in the glycan at the other glycosylation site, Asn-106. As noted, peak4, which represents diphosphorylated rhDNase, includes species with diphospho-

40 35

1 -

30 25

-

20

-

15

-

10 -

C

L

B

h

5-

A 00

5

IO

Time, mln

Figure 8. Analysis by anion exchange HPLC on the PEI column of (A) untreatedrhDNase, (B) rhDNase collected as peak 2 from a separation such as that shown in Figure 3A and subsequently treated with endo H, and (C) monophosphorylated rhDNase collected as peak 3 from the I . same separation and subsequently treated with endo !

rylated oligosaccharides at either of the two glycosylation sites. Oligosaccharides containing two residues of Man-6-P have been shown to bind the Man-6-P receptor more tightly than monophosphorylated glycans and are thought to be more effectively targeted to lysosomes.8 Site-DirectedDeglycosylation. The separation of isophosphorylates of rhDNase further allows the selective removal of oligosaccharides from either site of glycosylation. This can be achieved since Man-6-P is found only in high mannose and hybrid carbohydrate structures; these structures are susceptible to removal by endo H digestion and the PEI column permits the facile isolation of peaks 2 and 3, which carry Man-6-P at defined sites. Endo H cleaves high mannose and hybrid structures from glycoproteins, leaving only a single N-acetylglucosamine (GlcNAc) residue attached to the asparagine side ~ h a i n . 1It~ can thus be employed to pare the oligosaccharide at Asn-18 down to a single GlcNac in peak 2 from the PEI column, without significantly altering the complex glycan at Asn- 106. Similarly, the oligosaccharide at Asn-106 in peak 3 can be pared, without substantially altering that at Asn-18. As noted above, bDNase is not glycosylated at Asn-106, so the product of endo H digestion of peak 2 is a human enzyme similar in this respect to bDNase. Conversely, the product of endo H digestion of peak 3 carries an oligosaccharide only at Asn-106 and so differs from both rhDNase and bDNase. Figure 6 shows the chromatograms of the analyses on the PEI column of (B) peak 2 and (C) peak 3 following endo H treatment. Peak 2 contains Man-6-P, and therefore endo H susceptible oligosaccharides, at Asn- 18. As expected, cleavage of the oligosaccharide, and concomitant removal of the Man-6-P residue, shifts the retention time of the fraction to the position of the nonphosphorylated forms of rhDNase, as shown in Figure 6B. Tryptic mapping of the endo H treated fraction (data not shown) confirmed the presence of a peptide with a mass of 1997, that is, the mass (15) Tai, T.; Yamashia, K.; Kobata, K. Biochem. Eiophys. Res. Commun. 1977, 78, 434-41. (16) Lobel, P.; Dahms, N. M.; Korfeld, S. J . B i d . Chem. 1988, 263, 2563-70.

340

Analytical Chemistry, Voi. 86, No. 3, February 1, 1994

+

of T3 GlcNAc. Peak 3 contains rhDNase with Man-6-P in the glycan at Asn-106, but following digestion with endo H, the protein elutes from the PEI column near the retention time of the untreated peak 2, not to the retention time expected of aphosphoryl rhDNase. Tryptic mapping (data not shown) confirmed the removal of all but a GlcNAc residue from Asn106 in the treated fraction. The later than expected retention time of the endo H treated peak 3 may be due to the exposure to the anion exchange column surface of the aspartic acid side chain at residue 107 following cleavage of the adjacent bulky oligosaccharide. Note that removal of the phosphate alone shifts the peak to the position of peak 1, as shown in Figure 3. Electrostatic interaction of the acidic Asp107 side chain with the surface would be precluded, in this hypothesis, by the presence of the glycan at Asn-106, but on pruning of the oligosaccharide down to a single GlcNAc residue, this site may interact with the surface to cause the shift to longer retention on the column. The shift to earlier retention times of both peaks 2 and 3 following endo H treatment suggests that the site-directed deglycosylation procedure has produced the singly glycosylated variants of rhDNase. The role of glycosylation on the function of the enzyme can therefore be examined using variants prepared by this approach.

CONCLUSIONS Characterization of the carbohydrate heterogeneity of rhDNase is facilitated by the separation of isophosphorylates by anion exchange HPLC with the PEI column. The column resolves rhDNase glycoforms according to both the extent and sites of mannose phosphorylation. This selectivity is unavailable by either anion exchange HPLC with a DEAE column or IEF gel electrophoresis. The separation can be employed to quantify the molar phosphate content of a preparation of rhDNase. Fractionation of isophosphorylates also permitted the identification of the sites of phosphorylation. Site-directed deglycosylation of isophosphorylates allows the selective removal of oligosaccharides to investigate the effects of glycosylation on the enzymatic activity of rhDNase. The separation of rhDNase glycoforms according to the position and number of Man-6-P residues in N-linked oligosaccharides mimics the binding of phosphorylated oligosaccharides to the Man-6-P receptor. The strength of binding of a ligand to the receptor is also governed by the position and number of residues involved in binding.' Hence, while there is probably little structural homology between immobilized PEI and the Man-6-P receptor,16the biomimetic interaction can be exploited to sort rhDNase isoforms into subpopulations on this basis, providing a useful tool for protein characterization. ACKNOWLEDGMENT The authors acknowledge Mike Spellman of Genentech and Csaba Horvdth of Yale University for many helpful discussions in connection with this work. Received for revlew July 7, 1993. 1993.' e

Accepted November 16,

Abstract published in Adunnce ACS Abstrocfs. January 1 , 1994.