- Email: [email protected]
[45] Glycosylphosphatidylinositol-phospholipase D: A tool for glycosylphosphatidylinositol structural analysis
630
GPI-ANCHOREDPROTEINS
[45]
Acknowledgments We thank Chris Redman and Achim Treumann for permission to use unpublished data shown in Figs. 4 and 5. This work was supported by The Wellcome Trust. M.A.J.F. is a Howard Hughes International Research Scholar. P.S. is an EMBO long-term fellow.
[451 G l y c o s y l p h o s p h a t i d y l i n o s i t o l - P h o s p h o l i p a s e A Tool for Glycosylphosphatidylinositol Structural Analysis
By MARK A.
D:
DEEG and MICHAEL A. DAVITZ
Introduction A fundamental problem in molecular and cellular biology is how lipids and proteins assemble and interact to create biological membranes. The phospholipid bilayer is not static; rather, it is a dynamic entity composed of over 1000 distinct lipid species that are constantly being remodeled. 1 Lipids and lipid metabolites can also act as important second messengers. Because of their chimeric nature, glycosylphosphatidylinositol (GPl)-anchored proteins offer a unique window into lipid biosynthesis, transport, and degradation. GPI-anchored proteins comprise an extraordinarily diverse class of membrane molecules. They protect cells from complementmediated lysis, control cell to cell adhesion, activate T cells, and play a role in the etiology of slow viral diseases. 2 Despite the functional diversity, GPIanchored proteins are all attached to the plasma membrane by a glycolipid anchor, ethanolamine-(mannose)3-glucosamine-phosphatidylinositol, that is conserved from parasites to humans. The GPI-specific phospholipase D (GPI-PLD) is a phospholipase D present in all mammalian plasma. Because it recognizes a common core element of the GPI anchor, consisting of glucosamine-phosphatidylinositol, the G P I - P L D has proved to be a useful analytical tool for characterizing the structure of GPI-anchored molecules. Over the past several years, we and other have discovered, purified, cloned, and characterized one of the key enzymes of GPI anchor degradation, the GPI-PLD. In this chapter, we discuss the use of the enzyme as a tool for structural analysis of GPI-anchored proteins. 1 R . H . H j e l m s t a d a n d R . M. Bell, Biochemistry 30, 1731 (1991). 2 G . A . M. C r o s s , Annu. Rev. Cell Biol. 6, 1 (1990).
METHODS IN ENZYMOLOGY,VOL. 250
Copyright © 1995 by Academic Press, lnc. All rights of reproductionin any form reserved.
[451
GPI-PHOSPHOLIPASED
631
Protein
!
Ethanolamine
N
Glucosamine
~an3~ (Galactosen
Inositol
)
Plasma Membrane
FIG. 1. Conserved core structures of GPI anchors.
S t r u c t u r e of G l y c o s y l p h o s p h a t i d y l i n o s i t o l A n c h o r The basic glycan core structure of the GPI anchor is highly conserved. The detailed carbohydrate structure of the anchor was first established for the membrane form of the variant surface glycoprotein of the African trypanosome (mfVSG). In mfVSG, the C-terminal amino acid is coupled via phosphoethanolamine to a core glycan structure consisting of (mannose)3glucosamirle, with a branched side chain linked to one of the mannose sugars containing a variable number of galactose residues. 3 The glycan is linked to the 6-hydroxyl group on the inositol ring of dimyristoylphosphatidylinositol (Fig. 1). Although the core structure is conserved, there can be a number of structurally significant modifications to the carbohydrate backbone. 4 These include the presence of additional a-mannose residues, ethanolamine phosphate groups, as well as palmitoylation of the inositol ring. 4 The nature of the lipid attachment on the glycerol backbone can also vary, with O-alkyl rather than O-acyl linkages at the sn-1 position. 4 Discovery a n d Purification of Glycosylphosphatidylinositol-Specific Phospholipase D
Discovery of Enzyme The presence of a GPI anchor in mammalian cells immediately implied the existence of a corresponding anchor-specific phospholipase; however, 3 M. A. J. Ferguson, M. G. Low, and G. A. M. Cross, J. Biol. Chem. 260, 14547 (1985). 4 M. J. McConville and M. A. J. Ferguson, Biochem. J. 294, 305 (1993).
632
GPI-ANCHOREDPROTEINS
[45]
TABLE I PURIFICATION OF PLASMA GLYCOSYLPHOSPHATIDYLINOSITOL--PHosPHOL1PASEDa
Fraction
Total protein (mg)
Total units b
PF ~
%R d
Serum MonoQ Phenyl-5PW W h e a t germ lectin MonoQ
1785 60 1.59 0.176 0.020
34,550 20,360 21,060 5600 1700
-18 700 1660 4500
-60 60 16 5
a A d a p t e d from M. A. Davitz, J. H o m , and S. Schenkman, J. Biol. Chem. 264, 13760 (1989). b O n e unit of enzymatic activity is defined as the a m o u n t of G P I - P L D activity required to hydrolyze 0.5 /zg of [3H]mfVSG completely in 30 min at 37 ° in the presence of excess substrate. c Purification factor (PF) = (total units/total protein)~raction/(total units/total protein) . . . . . . d Recent recovery (%R) = total unitSfraction/34,550.
at first no conclusive evidence of a cellular phospholipase fitting the description was found. Using [3H]myristate-labeled mfVSG ([3H]mfVSG) as a GPI substrate, three groups simultaneously discovered the existence of a mammalian GPI-specific phospholipase. 5-7 Surprisingly, the enzyme was present in plasma not, as would have been predicted, in the cells. Moreover, the only phospholipid product observed after incubation of unfractionated serum with the GPI-anchored substrate [3H]mfVSG was phosphatidic acid, indicating that, contrary to expectations, the enzyme was a phospholipase D rather than a phospholipase C.
Purification of Enzyme The GPI-PLD was purified to homogeneity from human serum using a four-step procedure involving fractionation first over an anion-exchange
Mono Q column (Pharmacia, Piscataway, N J), followed by passage over a hydrophobic phenyl-5PW column,s The eluate from the phenyl column was passed over a wheat germ lectin affinity column and finally purified to homogeneity by passage over an anion-exchange Mono Q column. Using the approach a 4500-fold purification was achieved (Table I). The purified 5 M. A. Davitz, D. Hereld, S. Shak, J. Krakow, P. T. Englund, and V. Nussenzweig, Science 238, 81 (1987). 6 M. G. Low and A. R. S. Prasad, Proc. Natl. Acad. Sci. U.S.A. 85, 980 (1988). 7 M. L. Cardoso de Almeida, M. J. Turner, B. B. Stambuk, and S. Schenkman, Biochem. Biophys. Res. Commun. 150, 476 (1988). 8 M. A. Davitz, J. H o m , and S. Schenkman, J. Biol. Chem. 264, 13760 (1989).
[451
GPI-PHOSPHOLIPASED
633
G P I - P L D has an Mr of 110,000 and appears to consist of a single polypeptide chain. Using a different chromatographic sequence another group purified the G P I - P L D to homogeneity from bovine plasma. 9 The eight-step procedure consisted of a 9% polyethylene glycol precipitation, Q-Sepharose anionexchange chromatography, S-300 gel filtration, wheat germ lectin-Sepharose chromatography, hydroxylapatite agarose chromatography, zinc chelate matrix chromatography, M o n o Q [high-performance liquid chromatography (HPLC)], and Superose 12 (gel filtration). As was the case with the enzyme from human serum, the bovine G P I - P L D has a molecular weight of 100,000.
Protein S t r u c t u r e of G l y c o s y l p h o s p h a t i d y l i n o s i t o l - S p e c i f i c P h o s p h o l i p a s e D: A n a l y s i s of t h e cDNA A c D N A encoding a G P I - P L D has been isolated from bovine liver, 1° h u m a n liver, u and human u and mouse 12 islets. The nucleotide sequences of the liver and pancreatic enzymes are very similarl°,u; however, whether there is a related family of G P I - P L D type enzymes is not known. The bovine liver c D N A predicts an 816-amino acid protein with a 23-residue signal peptide, suggesting that the protein is preferentially secreted. That conclusion is supported by transfection studies with Chinese hamster ovary ( C H O ) cellsJ ° After transfection with the bovine G P I - P L D c D N A , the majority of the enzymatic activity was found in the m e d i u m J ° Analysis of the primary amino acid sequence from the bovine liver c D N A reveals eight potential N-glycosylation sites (Fig. 2). T r e a t m e n t of the bovine serum G P I - P L D with N-glycosidase F results in a decrease in the mobility of G P I - P L D on sodium dodecyl sulfate-polyacrylamide gel electrophoresis ( S D S - P A G E ) from 115 to 9l kDa, 13 consistent with the predicted molecular mass from the c D N A . Four regions of internal sequence similarity are present at residues 357379, 426-448, 489-511, and 694-716 (Fig. 2). The domains have similarity to the metal binding sites in the a subunits of integrins and consist of 9 K. S. Huang, W. J. C. Fung, J. D. Hulmes, L. Reik, Y. C. E. Pan, and M. G. Low, J. Biol. Chem. 265, 17738 (1990). 10B. Scallon, W. J. Fung, C. Tsang, S. Li, H. Kado-Fong, K. S. Huang, and J. Kochan, Science 252, 446 (1991). u T. C. Tsang, W. Fung, J. Levine, C. N. Metz, M. A. Davitz, D. K. Burns, K. Huang, and J. P. Kochan, manuscript in preparation. 12C. N. Metz, Y. Zhang, Y. Guo, T. C. Tsang, J. P. Kochan, N. Altszuler, and M. A. Davitz, J. Biol. Chem. 266, 17733 (1991). 13M. C. Hoener, S. Stieger, and U. Brodbeck, Eur. J. Biochem. 190, 593 (1990).
634
GPI-ANCHOREDPROTEINS
[451 oaJciumbinding II domain
Domains of GPI-PLD
potential heparin bindingsite
8
I'
IF potential N-glyco6ylationsite
PKC ~ Casein KinaseII
Potential phosphorylatlon sites
rl.i
11111
T Tyrosine Kinase ¢ PKA
Kinase
FIG. 2. Structural featurs of G P I - P L D derived from the deduced amino acid sequence of the enzyme from bovine liver.
asparagine-rich regions seen in other proteins that bind calcium or magnesium. t4 Given that calcium or another similar divalent cation is required for the catalytic activity of GPI-PLD, the calcium-binding regions are probably involved in substrate recognition and/or the catalytic site. In addition, two regions have a high density of positively charged amino acids, residues 274-303 and 382-390, and sequence similarity to a consensus sequence for heparin binding. 15,16It is unknown whether G P I - P L D binds heparin and, if so, what potential role that might have on G P I - P L D function. Sequences that match the phosphorylation motifs for a number of intracellular kinases including protein kinase C (PKC), MAP kinase, cdc2, casein kinase II, tyrosine kinases, and cAMP-dependent protein kinase are present. Although no experimental evidence suggests that G P I - P L D is phosphorylated either intra- or extracellularly, or that the phosphorylation state of the enzyme affects the catalytic activity of GPI-PLD, phosphorylation of G P I - P L D represents an intriguing possibility for controlling the biological behavior of the enzyme. Both primary and secondary amino acid sequence analyses of the protein structure predict G P I - P L D to be an amphipathic protein, and this has been demonstrated experimentally, myEvaluation of the hydropathy of the primary sequence using the Kyte-Doolittle algorithm illustrates the amphipathic nature of GPI-PLD. Analysis of the predicted protein secondary 14 D. M. E. Szebenyi, Nature (London) 294, 327 (1981). 15 A. D. Cardin, D. A. D e m e t e r , H. J. R. Weintraub, and R. L. Jackson, this series, Vol. 203, p. 556. 16 D. Alan, Cardin, and H. J. R. Weintraub, Arteriosclerosis (Dallas) 9, 21 (1989). 17 M. C. H o e n e r and U. Brodbeek, Eur. J. Biochem. 206, 747 (1992),
[451
GPI-PHOSPHOLIPASED
635
structure by the method of Chou and Fasman TM reveals primarily a helices in the N-terminal half. Conversely, the C-terminal half consists primarily of/3 sheets. Four of the helical regions are predicted to have amphipathic properties at residues 8-18, 289-302, 319-341, and 541-556. Amphipathic helices have been identified in a number of proteins that interact with lipids, including hormones, transmembrane proteins, and apolipoproteins.19 The amphipathic nature of G P I - P L D was directly demonstrated by sucrose density gradient centrifugation and gel filtration. 17 Addition of 0.1% Triton X-100 to partially purified human or bovine serum G P I - P L D resulted in a decrease in the sedimentation coefficient during gradient centrifugation and the Stokes radius on gel filtration, suggesting that G P I - P L D binds detergent. E n z y m o l o g y of Glycosylphosphatidylinositol-Specific Phospholipase D Substrate Specificity
The G P I - P L D exhibits absolute specificity for the GPI anchor, that is, it does not cleave phosphatidylethanolamine or phosphatidylcholine, which are known substrates for other phospholipase D-type enzymes (Fig. 3). Moreover, the G P I - P L D does not cleave phosphatidylinositol that has been derived from HNO2-cleaved mfVSG. 5 The enzyme is, however, capable of cleaving a glycolipid biosynthetic precursor of the GPI anchor. 2° The biosynthetic precursor contains a portion of the common core GPI anchor structure, consisting of glucosamine-phosphatidylinositol. 4,2°,21This appears to be the minimum carbohydrate unit required for substrate recognition by the enzyme. None of the known naturally occurring structural modifications of the GPI anchor affect the susceptibility of the GPI anchor to the G P I - P L D : these include palmitoylation of the inositol ring [which renders the GPI anchor resistant to the action of phosphatidylinositol-inositol phospholipase C (PI-PLC)], the presence of an ether linkage in the sn-1 position on the glycerol backbone, and the presence of a phosphoethanolamine on the first mannose residue closest to the phosphatidylinositol anchor. 4 Thus, all GPI-anchored molecules are potential substrates for the G P I - P L D . 18p. y. Chou and G. D. Fasman, Adv. EnzymoL 47, 45 (1978). 19p. Jere, H. D. L. Segrest,J. G. Dohlman, G. B. Christie, and G. M. Anatharamaiah, Proteins: Struct. Funct. Genet. 8, 103 (1990). 20T. L. Doering, W. J. Masterson, P. T. Englund, and G. W. Hart, J. Biol. Chem. 264, 11168 (1989). 21W. J. Masterson, J. Raper, T. L. Doering, G. W. Hart, and P. T. Englund, Cell (Cambridge, Mass.) 62, 73 (1990).
636
GPI-ANCHORED PROTEINS
[45]
Protein II Core Recognition Site Ethanolamlne for the GPI-PLD
I (Man3 )
o
ii
co
li
Site of GPI-PLD
~ oI
Cleavage
AIkylacyIglycerol or
Dlacylglycerol FIG. 3. Cleavage specificity of G P I - P L D . The minimal carbohydrate recognition site for G P I - P L D is shown. The inositol ring is displayed as palmitoylated in order to illustrate that the modification does not affect the sensitivity of the GPI anchor to the action of G P I - P L D .
Dependence on Divalent Cations
The enzymatic activity of the G P I - P L D could be inhibited by incubation with EGTA, indicating that it is a divalent cation-dependent phospholipase. 5-7 The specific activity of the G P I - P L D increased with increasing Ca 2+. From 1 to 100 tzM Ca 2+, there was a slow increase in specific activity (0-25 nmol of [3H]mfVSG hydrolyzed/min/mg protein), followed by a sharp increase to a specific activity of 75 at 200 tzM or greater Ca 2+ concentration. 8 The enzymatic activity could be distinguished from other divalent cationdependent cellular phosphatidylinositol-specific phospholipase C-type enzymes by the fact that the activity could be inhibited by incubation with 1,10-phenanthroline, a transition metal chelator. 6'22 22 M. G. Low, in "Lipid Modification of Proteins: A Practical Approach" (N. M. Hooper and A. J. Turner, eds.), p. 117. Oxford Univ. Press, Oxford, 1992.
[45]
GPI-PHOSPHOLIPASED
637
Effects of Detergents Although the G P I - P L D effectively hydrolyzes GPI-anchored proteins in detergent solutions, the enzyme is unable to cleave and release GPIanchored proteins from intact cell membranesfi 6 This is in direct contrast to PI-PLC, which cleaves and releases most GPI-anchored proteins from the cell membrane. The likely explanation for the observation is that the action of the phospholipase D is restricted by the physical state of the phospholipid bilayer. 23 Low and Huang 23 demonstrated that incubation of cells with low concentrations of Nonidet P-40 (NP-40) (0.1%, v/v) rendered membrane alkaline phosphatase susceptible to the action of the G P I - P L D . Similarly, NP-40 stimulated GPI-mediated degradation of GPI-anchored proteins that had been reconstituted into phospholipid vesicles. The latter results suggested that the physical interaction of the G P I - P L D with the phospholipid bilayer, rather than binding of an enzymatic inhibitor, was controlling the action of the G P I - P L D on the cell. Given that the enzymatic activity of several other phospholipases is affected by amphipathic agents, such as lipids and detergents, a it is not surprising that alone the soluble G P I - P L D does not cleave GPI-anchored proteins on intact cells. Low and Huang 2a have shown that phosphatidic acid, lysophosphatidic acid, and lipid A are all inhibitors of G P I - P L D enzymatic activity. Inhibition appeared to depend on the presence of a phosphomonoester group and seemed to involve a direct interaction of phosphatidic acid or lipid A with the G P I PLD. 24 As the authors point out, it is difficult to extrapolate the in vitro effects in detergent solutions to potential physiological control of enzymatic activity on or inside the cell.
Using Glycosylphosphatidylinositol-Speeifie Phospholipase D as Analytical Tool: Practical Advice
Buffers, pH, Salt, and Detergents A solution containing the GPI substrate is incubated in a total volume of 100/xl. The G P I - P L D is active in a variety of different buffer systems. Typically, we use a buffer system containing 50 mM Tris, pH 7.4, 10 m M NaCI, and 2.0 mM CaCI2. We have also utilized Bicine, Bis-Tris, HEPES, and phosphate buffers with equal success. The one exception to this observation is a bicarbonate buffer mixture; here, the G P I - P L D appears to be 23M. G. Low and K. S. Huang, Biochem. J. 279, 483 (1991). 24M. G. Low and K. S. Huang, J. Biol. Chem. 268, 8480 (1993).
638
GPI-ANCHORED PROTEINS
[45]
inhibited via a reverse carbamylation reaction. 25 Because the enzyme is uniformly active over a broad p H range (i.e., p H 5-8), s p H is not a critical factor. However, because the G P I - P L D is inhibited at high salt concentrations (>0.2 M), the final salt concentration should be kept at a minimum. 8'9 We have tested the G P I - P L D activity with a variety of nonionic detergents, such as NP-40 and Triton X-100. Although enzymatic activity is stable over a range of detergent concentrations (0.01-0.1%, v/v), we typically perform the assays in solutions having a final NP-40 or Triton X-100 concentration of 0.01% (v/v). It should be noted that, in contrast to the situation observed for nonionic detergents, ionic detergents such as sodium dodecyl sulfate (SDS) significantly inhibit G P I - P L D enzymatic activity.
Substrate As much as 1-2/~g of the potential GPI substrate can be added to each assay, although 0.1 to 0.5 /~g of the GPI substrate is more usual. After addition of the G P I - P L D to the reaction mixture, the sample is briefly vortexed and incubated at 37 °. The mixture is incubated for up to 60 min and then extracted with 500 /.d of n-butanol saturated with either 1 M N H 4 O H or water.
Source of Enzyme Either unfractionated serum or purified G P I - P L D can serve as a source of the enzyme. When using serum, 1-2/~l of serum, which is equivalent to 1 - 2 units of enzymatic activity, is used in each assay. Because the G P I - P L D is not commercially available, and because the activity in serum exhibits properties identical to those of the purified enzyme, 8,9 it is more convenient to use serum as a source of the G P I - P L D . Bovine or human serum is convenient and can be used as long as the blood is not collected with chelating agents such as E D T A or citrate, which is commonly used in clinical situations. It is important to note, however, that if human serum is used, all appropriate safety measures for working with human blood products must be observed. Storage of serum or plasma leads to dissociation of the G P I - P L D from the native complex with apolipoprotein A1, which causes aggregation of the enzyme. 26 However, this does not affect use of the enzyme as an analytical tool. In addition to the G P I - P L D , serum contains a number of other lipases that investigators must be aware of. 25 S. Stieger, S. Diem, A. Jakob, and U. Brodbeck, Eur. J. Biochem. 197, 67 (1991). 26 M. A. Deeg, M. A. Davitz, A. C. Wolf, E. L. Bierman, and M. C. Cheung, submitted for publication (1995).
[45]
GPI-PHOSPHOLIPASE i~
639
Protein
TLC
I
Ethanolamin¢
DAG
Ja Inositol ( n~ Glucosamlne~ o
4D FFA
°p""° PA _
II °
Organic Extraction with n-butanol. Analysisof Lipid Products by TLC,
0 Origin
Plate developed in CHCI3:CH3OH:KCI FIG. 4. Sample product analysis, n-Butanol is evaporated under an N2 stream. The lipid products are resuspended in CHC13/CH3OH/water (10:10:3, v/v/v) and then applied to a TLC plate. The plate is developed in CHC13/CH3OH/0.25% KCI (55:45:10, v/v/v) [M. A. Davitz, J. Horn, and S. Schenkman, J. Biol. Chem. 264, 13760 (1989)]. The relative positions of various lipid cleavage products are illustrated. PA, Phosphatidic acid; FFA, free fatty acid; DAG, diacyl- or alkylacylglycerol.
These include sphingomyelinase which is present in bovine serum, 27 platelet-activating factor (PAF) hydrolase, a phospholipase A2,28 and a phosphatidylcholine-specific phospholipase C which has been found in rabbit serum. TM In addition, under certain conditions apolipoprotein B, a major constituent of very low and low density lipoprotein, has phospholipase A1 and A2 activity.29'3°However, the only enzyme in serum that is active against GPI anchors is the GPI-PLD. 5-7
Product Analysis Examination of the cleavage products is a crucial part of the analysis. GPI-anchored molecules can be biosynthetically labeled with a variety of 27 M. W. Matthew, D. M. Byers, F. B. S. C. Palmer, and H. W. Cook, J. Biol. Chem. 264, 5358 (1989). 28 D. M. Stafforini, S. M. Prescott, and T. M. Mclntrye, J. BioL Chem. 2629 4223 (1994). 28a M. Deeg, A. Mendez, C. Furlong, and A. Chait, manuscript in preparation. 29 N. Reisfeld, D. Lichtenberg, A. Dagan, and S. Yedgar, F E B S Lett. 315, 267 (1993). 30 S. Parthasarathy and J. Barnett, Proc. Natl. Acad. Sci. U.S.A. 87, 9741 (1990).
640
GPI-ANCHOREDPROTEINS
[45]
different anchor constituents, including [3H]ethanolamine, [3H]glucosamine, [3H]inositol, and 3H-labeled fatty acids (e.g., myristic or palmitic acid). R o s e n b e r r y and colleagues have developed several different chemical methods for chemically tagging the G P I molecules, including use of [azsI]TID [3-(trifluoromethyl)-3-(m-[125I]iododiazirine], which labels the hydrophobic fatty acid m e m b r a n e anchors, 31 and H l a C H O , which radiomethylates free amino groups. 3e It should be noted that radiomethylation of the glucosamine renders the G P I anchor resistant to the action of the G P I PLD. 33 Once the putative G P I molecules have b e e n treated with the G P I PLD, the organically soluble cleavage products can be analyzed by thinlayer c h r o m a t o g r a p h y (TLC) or H P L C . After cleavage, the hydrophilic anchor c o m p o n e n t s can be examined by gel filtration using a BioGel P4 (Bio-Rad, Richmond, C A ) sizing column. 34-37 A complete examination of the analytical procedures is beyond the scope of this chapter, and the subject has been thoroughly reviewed elsewhere4'34-37; Fig. 4 illustrates one possible analytical sequence. Summary Cleavage by the G P I - P L D provides definitive evidence of a minimal G P I structure: glucosamine-phosphatidylinositol. Unlike the case for PIPLC, cleavage by the G P I - P L D is unaffected by acylation of the inositol ring. Thus the G P I - P L D provides an excellent simple enzymatic tool for analyzing the basic core structure of G P I anchors. Acknowledgments Mark A. Deeg is a Pfizer Postdoctoral Fellow. Michael A. Davitz is a Lucille P. Markey Scholar in the Biomedical Sciences. This work was supported by grants from the Lucille P. Markey Charitable Trust and the Council for Tobacco Research.
31W. Roberts and T. Rosenberry, Biochemistry 25, 3091 (1986). 32R. Haas and T. Rosenberry, Anal, Biochem. 148, 154 (1985). 33M. A. Deeg, N. R. Murray, and T. L. Rosenberry, J. Biol. Chem. 267, 18581 (1992). 34A. K. Menon, S. Mayor, M. A. J. Ferguson, M. Duszenko, and G. A. M. Cross, J. Biol, Chem. 263, 1970 (1988). 35A. K. Menon, R. T. Sehwarz, and S. Mayor, EMBO J. 9, 4249 (1990). 36m. K. Menon, S. Mayor, M. J. Ferguson, M. Duszenko, and G. A. M. Cross, J. Biol. Chem. 263, 1970 (1988). 37S. Mayor, A. K. Menon, and G. A. M. Cross, J. Biol. Chem. 265, 6174 (1990).
GPI-ANCHOREDPROTEINS
[45]
Acknowledgments We thank Chris Redman and Achim Treumann for permission to use unpublished data shown in Figs. 4 and 5. This work was supported by The Wellcome Trust. M.A.J.F. is a Howard Hughes International Research Scholar. P.S. is an EMBO long-term fellow.
[451 G l y c o s y l p h o s p h a t i d y l i n o s i t o l - P h o s p h o l i p a s e A Tool for Glycosylphosphatidylinositol Structural Analysis
By MARK A.
D:
DEEG and MICHAEL A. DAVITZ
Introduction A fundamental problem in molecular and cellular biology is how lipids and proteins assemble and interact to create biological membranes. The phospholipid bilayer is not static; rather, it is a dynamic entity composed of over 1000 distinct lipid species that are constantly being remodeled. 1 Lipids and lipid metabolites can also act as important second messengers. Because of their chimeric nature, glycosylphosphatidylinositol (GPl)-anchored proteins offer a unique window into lipid biosynthesis, transport, and degradation. GPI-anchored proteins comprise an extraordinarily diverse class of membrane molecules. They protect cells from complementmediated lysis, control cell to cell adhesion, activate T cells, and play a role in the etiology of slow viral diseases. 2 Despite the functional diversity, GPIanchored proteins are all attached to the plasma membrane by a glycolipid anchor, ethanolamine-(mannose)3-glucosamine-phosphatidylinositol, that is conserved from parasites to humans. The GPI-specific phospholipase D (GPI-PLD) is a phospholipase D present in all mammalian plasma. Because it recognizes a common core element of the GPI anchor, consisting of glucosamine-phosphatidylinositol, the G P I - P L D has proved to be a useful analytical tool for characterizing the structure of GPI-anchored molecules. Over the past several years, we and other have discovered, purified, cloned, and characterized one of the key enzymes of GPI anchor degradation, the GPI-PLD. In this chapter, we discuss the use of the enzyme as a tool for structural analysis of GPI-anchored proteins. 1 R . H . H j e l m s t a d a n d R . M. Bell, Biochemistry 30, 1731 (1991). 2 G . A . M. C r o s s , Annu. Rev. Cell Biol. 6, 1 (1990).
METHODS IN ENZYMOLOGY,VOL. 250
Copyright © 1995 by Academic Press, lnc. All rights of reproductionin any form reserved.
[451
GPI-PHOSPHOLIPASED
631
Protein
!
Ethanolamine
N
Glucosamine
~an3~ (Galactosen
Inositol
)
Plasma Membrane
FIG. 1. Conserved core structures of GPI anchors.
S t r u c t u r e of G l y c o s y l p h o s p h a t i d y l i n o s i t o l A n c h o r The basic glycan core structure of the GPI anchor is highly conserved. The detailed carbohydrate structure of the anchor was first established for the membrane form of the variant surface glycoprotein of the African trypanosome (mfVSG). In mfVSG, the C-terminal amino acid is coupled via phosphoethanolamine to a core glycan structure consisting of (mannose)3glucosamirle, with a branched side chain linked to one of the mannose sugars containing a variable number of galactose residues. 3 The glycan is linked to the 6-hydroxyl group on the inositol ring of dimyristoylphosphatidylinositol (Fig. 1). Although the core structure is conserved, there can be a number of structurally significant modifications to the carbohydrate backbone. 4 These include the presence of additional a-mannose residues, ethanolamine phosphate groups, as well as palmitoylation of the inositol ring. 4 The nature of the lipid attachment on the glycerol backbone can also vary, with O-alkyl rather than O-acyl linkages at the sn-1 position. 4 Discovery a n d Purification of Glycosylphosphatidylinositol-Specific Phospholipase D
Discovery of Enzyme The presence of a GPI anchor in mammalian cells immediately implied the existence of a corresponding anchor-specific phospholipase; however, 3 M. A. J. Ferguson, M. G. Low, and G. A. M. Cross, J. Biol. Chem. 260, 14547 (1985). 4 M. J. McConville and M. A. J. Ferguson, Biochem. J. 294, 305 (1993).
632
GPI-ANCHOREDPROTEINS
[45]
TABLE I PURIFICATION OF PLASMA GLYCOSYLPHOSPHATIDYLINOSITOL--PHosPHOL1PASEDa
Fraction
Total protein (mg)
Total units b
PF ~
%R d
Serum MonoQ Phenyl-5PW W h e a t germ lectin MonoQ
1785 60 1.59 0.176 0.020
34,550 20,360 21,060 5600 1700
-18 700 1660 4500
-60 60 16 5
a A d a p t e d from M. A. Davitz, J. H o m , and S. Schenkman, J. Biol. Chem. 264, 13760 (1989). b O n e unit of enzymatic activity is defined as the a m o u n t of G P I - P L D activity required to hydrolyze 0.5 /zg of [3H]mfVSG completely in 30 min at 37 ° in the presence of excess substrate. c Purification factor (PF) = (total units/total protein)~raction/(total units/total protein) . . . . . . d Recent recovery (%R) = total unitSfraction/34,550.
at first no conclusive evidence of a cellular phospholipase fitting the description was found. Using [3H]myristate-labeled mfVSG ([3H]mfVSG) as a GPI substrate, three groups simultaneously discovered the existence of a mammalian GPI-specific phospholipase. 5-7 Surprisingly, the enzyme was present in plasma not, as would have been predicted, in the cells. Moreover, the only phospholipid product observed after incubation of unfractionated serum with the GPI-anchored substrate [3H]mfVSG was phosphatidic acid, indicating that, contrary to expectations, the enzyme was a phospholipase D rather than a phospholipase C.
Purification of Enzyme The GPI-PLD was purified to homogeneity from human serum using a four-step procedure involving fractionation first over an anion-exchange
Mono Q column (Pharmacia, Piscataway, N J), followed by passage over a hydrophobic phenyl-5PW column,s The eluate from the phenyl column was passed over a wheat germ lectin affinity column and finally purified to homogeneity by passage over an anion-exchange Mono Q column. Using the approach a 4500-fold purification was achieved (Table I). The purified 5 M. A. Davitz, D. Hereld, S. Shak, J. Krakow, P. T. Englund, and V. Nussenzweig, Science 238, 81 (1987). 6 M. G. Low and A. R. S. Prasad, Proc. Natl. Acad. Sci. U.S.A. 85, 980 (1988). 7 M. L. Cardoso de Almeida, M. J. Turner, B. B. Stambuk, and S. Schenkman, Biochem. Biophys. Res. Commun. 150, 476 (1988). 8 M. A. Davitz, J. H o m , and S. Schenkman, J. Biol. Chem. 264, 13760 (1989).
[451
GPI-PHOSPHOLIPASED
633
G P I - P L D has an Mr of 110,000 and appears to consist of a single polypeptide chain. Using a different chromatographic sequence another group purified the G P I - P L D to homogeneity from bovine plasma. 9 The eight-step procedure consisted of a 9% polyethylene glycol precipitation, Q-Sepharose anionexchange chromatography, S-300 gel filtration, wheat germ lectin-Sepharose chromatography, hydroxylapatite agarose chromatography, zinc chelate matrix chromatography, M o n o Q [high-performance liquid chromatography (HPLC)], and Superose 12 (gel filtration). As was the case with the enzyme from human serum, the bovine G P I - P L D has a molecular weight of 100,000.
Protein S t r u c t u r e of G l y c o s y l p h o s p h a t i d y l i n o s i t o l - S p e c i f i c P h o s p h o l i p a s e D: A n a l y s i s of t h e cDNA A c D N A encoding a G P I - P L D has been isolated from bovine liver, 1° h u m a n liver, u and human u and mouse 12 islets. The nucleotide sequences of the liver and pancreatic enzymes are very similarl°,u; however, whether there is a related family of G P I - P L D type enzymes is not known. The bovine liver c D N A predicts an 816-amino acid protein with a 23-residue signal peptide, suggesting that the protein is preferentially secreted. That conclusion is supported by transfection studies with Chinese hamster ovary ( C H O ) cellsJ ° After transfection with the bovine G P I - P L D c D N A , the majority of the enzymatic activity was found in the m e d i u m J ° Analysis of the primary amino acid sequence from the bovine liver c D N A reveals eight potential N-glycosylation sites (Fig. 2). T r e a t m e n t of the bovine serum G P I - P L D with N-glycosidase F results in a decrease in the mobility of G P I - P L D on sodium dodecyl sulfate-polyacrylamide gel electrophoresis ( S D S - P A G E ) from 115 to 9l kDa, 13 consistent with the predicted molecular mass from the c D N A . Four regions of internal sequence similarity are present at residues 357379, 426-448, 489-511, and 694-716 (Fig. 2). The domains have similarity to the metal binding sites in the a subunits of integrins and consist of 9 K. S. Huang, W. J. C. Fung, J. D. Hulmes, L. Reik, Y. C. E. Pan, and M. G. Low, J. Biol. Chem. 265, 17738 (1990). 10B. Scallon, W. J. Fung, C. Tsang, S. Li, H. Kado-Fong, K. S. Huang, and J. Kochan, Science 252, 446 (1991). u T. C. Tsang, W. Fung, J. Levine, C. N. Metz, M. A. Davitz, D. K. Burns, K. Huang, and J. P. Kochan, manuscript in preparation. 12C. N. Metz, Y. Zhang, Y. Guo, T. C. Tsang, J. P. Kochan, N. Altszuler, and M. A. Davitz, J. Biol. Chem. 266, 17733 (1991). 13M. C. Hoener, S. Stieger, and U. Brodbeck, Eur. J. Biochem. 190, 593 (1990).
634
GPI-ANCHOREDPROTEINS
[451 oaJciumbinding II domain
Domains of GPI-PLD
potential heparin bindingsite
8
I'
IF potential N-glyco6ylationsite
PKC ~ Casein KinaseII
Potential phosphorylatlon sites
rl.i
11111
T Tyrosine Kinase ¢ PKA
Kinase
FIG. 2. Structural featurs of G P I - P L D derived from the deduced amino acid sequence of the enzyme from bovine liver.
asparagine-rich regions seen in other proteins that bind calcium or magnesium. t4 Given that calcium or another similar divalent cation is required for the catalytic activity of GPI-PLD, the calcium-binding regions are probably involved in substrate recognition and/or the catalytic site. In addition, two regions have a high density of positively charged amino acids, residues 274-303 and 382-390, and sequence similarity to a consensus sequence for heparin binding. 15,16It is unknown whether G P I - P L D binds heparin and, if so, what potential role that might have on G P I - P L D function. Sequences that match the phosphorylation motifs for a number of intracellular kinases including protein kinase C (PKC), MAP kinase, cdc2, casein kinase II, tyrosine kinases, and cAMP-dependent protein kinase are present. Although no experimental evidence suggests that G P I - P L D is phosphorylated either intra- or extracellularly, or that the phosphorylation state of the enzyme affects the catalytic activity of GPI-PLD, phosphorylation of G P I - P L D represents an intriguing possibility for controlling the biological behavior of the enzyme. Both primary and secondary amino acid sequence analyses of the protein structure predict G P I - P L D to be an amphipathic protein, and this has been demonstrated experimentally, myEvaluation of the hydropathy of the primary sequence using the Kyte-Doolittle algorithm illustrates the amphipathic nature of GPI-PLD. Analysis of the predicted protein secondary 14 D. M. E. Szebenyi, Nature (London) 294, 327 (1981). 15 A. D. Cardin, D. A. D e m e t e r , H. J. R. Weintraub, and R. L. Jackson, this series, Vol. 203, p. 556. 16 D. Alan, Cardin, and H. J. R. Weintraub, Arteriosclerosis (Dallas) 9, 21 (1989). 17 M. C. H o e n e r and U. Brodbeek, Eur. J. Biochem. 206, 747 (1992),
[451
GPI-PHOSPHOLIPASED
635
structure by the method of Chou and Fasman TM reveals primarily a helices in the N-terminal half. Conversely, the C-terminal half consists primarily of/3 sheets. Four of the helical regions are predicted to have amphipathic properties at residues 8-18, 289-302, 319-341, and 541-556. Amphipathic helices have been identified in a number of proteins that interact with lipids, including hormones, transmembrane proteins, and apolipoproteins.19 The amphipathic nature of G P I - P L D was directly demonstrated by sucrose density gradient centrifugation and gel filtration. 17 Addition of 0.1% Triton X-100 to partially purified human or bovine serum G P I - P L D resulted in a decrease in the sedimentation coefficient during gradient centrifugation and the Stokes radius on gel filtration, suggesting that G P I - P L D binds detergent. E n z y m o l o g y of Glycosylphosphatidylinositol-Specific Phospholipase D Substrate Specificity
The G P I - P L D exhibits absolute specificity for the GPI anchor, that is, it does not cleave phosphatidylethanolamine or phosphatidylcholine, which are known substrates for other phospholipase D-type enzymes (Fig. 3). Moreover, the G P I - P L D does not cleave phosphatidylinositol that has been derived from HNO2-cleaved mfVSG. 5 The enzyme is, however, capable of cleaving a glycolipid biosynthetic precursor of the GPI anchor. 2° The biosynthetic precursor contains a portion of the common core GPI anchor structure, consisting of glucosamine-phosphatidylinositol. 4,2°,21This appears to be the minimum carbohydrate unit required for substrate recognition by the enzyme. None of the known naturally occurring structural modifications of the GPI anchor affect the susceptibility of the GPI anchor to the G P I - P L D : these include palmitoylation of the inositol ring [which renders the GPI anchor resistant to the action of phosphatidylinositol-inositol phospholipase C (PI-PLC)], the presence of an ether linkage in the sn-1 position on the glycerol backbone, and the presence of a phosphoethanolamine on the first mannose residue closest to the phosphatidylinositol anchor. 4 Thus, all GPI-anchored molecules are potential substrates for the G P I - P L D . 18p. y. Chou and G. D. Fasman, Adv. EnzymoL 47, 45 (1978). 19p. Jere, H. D. L. Segrest,J. G. Dohlman, G. B. Christie, and G. M. Anatharamaiah, Proteins: Struct. Funct. Genet. 8, 103 (1990). 20T. L. Doering, W. J. Masterson, P. T. Englund, and G. W. Hart, J. Biol. Chem. 264, 11168 (1989). 21W. J. Masterson, J. Raper, T. L. Doering, G. W. Hart, and P. T. Englund, Cell (Cambridge, Mass.) 62, 73 (1990).
636
GPI-ANCHORED PROTEINS
[45]
Protein II Core Recognition Site Ethanolamlne for the GPI-PLD
I (Man3 )
o
ii
co
li
Site of GPI-PLD
~ oI
Cleavage
AIkylacyIglycerol or
Dlacylglycerol FIG. 3. Cleavage specificity of G P I - P L D . The minimal carbohydrate recognition site for G P I - P L D is shown. The inositol ring is displayed as palmitoylated in order to illustrate that the modification does not affect the sensitivity of the GPI anchor to the action of G P I - P L D .
Dependence on Divalent Cations
The enzymatic activity of the G P I - P L D could be inhibited by incubation with EGTA, indicating that it is a divalent cation-dependent phospholipase. 5-7 The specific activity of the G P I - P L D increased with increasing Ca 2+. From 1 to 100 tzM Ca 2+, there was a slow increase in specific activity (0-25 nmol of [3H]mfVSG hydrolyzed/min/mg protein), followed by a sharp increase to a specific activity of 75 at 200 tzM or greater Ca 2+ concentration. 8 The enzymatic activity could be distinguished from other divalent cationdependent cellular phosphatidylinositol-specific phospholipase C-type enzymes by the fact that the activity could be inhibited by incubation with 1,10-phenanthroline, a transition metal chelator. 6'22 22 M. G. Low, in "Lipid Modification of Proteins: A Practical Approach" (N. M. Hooper and A. J. Turner, eds.), p. 117. Oxford Univ. Press, Oxford, 1992.
[45]
GPI-PHOSPHOLIPASED
637
Effects of Detergents Although the G P I - P L D effectively hydrolyzes GPI-anchored proteins in detergent solutions, the enzyme is unable to cleave and release GPIanchored proteins from intact cell membranesfi 6 This is in direct contrast to PI-PLC, which cleaves and releases most GPI-anchored proteins from the cell membrane. The likely explanation for the observation is that the action of the phospholipase D is restricted by the physical state of the phospholipid bilayer. 23 Low and Huang 23 demonstrated that incubation of cells with low concentrations of Nonidet P-40 (NP-40) (0.1%, v/v) rendered membrane alkaline phosphatase susceptible to the action of the G P I - P L D . Similarly, NP-40 stimulated GPI-mediated degradation of GPI-anchored proteins that had been reconstituted into phospholipid vesicles. The latter results suggested that the physical interaction of the G P I - P L D with the phospholipid bilayer, rather than binding of an enzymatic inhibitor, was controlling the action of the G P I - P L D on the cell. Given that the enzymatic activity of several other phospholipases is affected by amphipathic agents, such as lipids and detergents, a it is not surprising that alone the soluble G P I - P L D does not cleave GPI-anchored proteins on intact cells. Low and Huang 2a have shown that phosphatidic acid, lysophosphatidic acid, and lipid A are all inhibitors of G P I - P L D enzymatic activity. Inhibition appeared to depend on the presence of a phosphomonoester group and seemed to involve a direct interaction of phosphatidic acid or lipid A with the G P I PLD. 24 As the authors point out, it is difficult to extrapolate the in vitro effects in detergent solutions to potential physiological control of enzymatic activity on or inside the cell.
Using Glycosylphosphatidylinositol-Speeifie Phospholipase D as Analytical Tool: Practical Advice
Buffers, pH, Salt, and Detergents A solution containing the GPI substrate is incubated in a total volume of 100/xl. The G P I - P L D is active in a variety of different buffer systems. Typically, we use a buffer system containing 50 mM Tris, pH 7.4, 10 m M NaCI, and 2.0 mM CaCI2. We have also utilized Bicine, Bis-Tris, HEPES, and phosphate buffers with equal success. The one exception to this observation is a bicarbonate buffer mixture; here, the G P I - P L D appears to be 23M. G. Low and K. S. Huang, Biochem. J. 279, 483 (1991). 24M. G. Low and K. S. Huang, J. Biol. Chem. 268, 8480 (1993).
638
GPI-ANCHORED PROTEINS
[45]
inhibited via a reverse carbamylation reaction. 25 Because the enzyme is uniformly active over a broad p H range (i.e., p H 5-8), s p H is not a critical factor. However, because the G P I - P L D is inhibited at high salt concentrations (>0.2 M), the final salt concentration should be kept at a minimum. 8'9 We have tested the G P I - P L D activity with a variety of nonionic detergents, such as NP-40 and Triton X-100. Although enzymatic activity is stable over a range of detergent concentrations (0.01-0.1%, v/v), we typically perform the assays in solutions having a final NP-40 or Triton X-100 concentration of 0.01% (v/v). It should be noted that, in contrast to the situation observed for nonionic detergents, ionic detergents such as sodium dodecyl sulfate (SDS) significantly inhibit G P I - P L D enzymatic activity.
Substrate As much as 1-2/~g of the potential GPI substrate can be added to each assay, although 0.1 to 0.5 /~g of the GPI substrate is more usual. After addition of the G P I - P L D to the reaction mixture, the sample is briefly vortexed and incubated at 37 °. The mixture is incubated for up to 60 min and then extracted with 500 /.d of n-butanol saturated with either 1 M N H 4 O H or water.
Source of Enzyme Either unfractionated serum or purified G P I - P L D can serve as a source of the enzyme. When using serum, 1-2/~l of serum, which is equivalent to 1 - 2 units of enzymatic activity, is used in each assay. Because the G P I - P L D is not commercially available, and because the activity in serum exhibits properties identical to those of the purified enzyme, 8,9 it is more convenient to use serum as a source of the G P I - P L D . Bovine or human serum is convenient and can be used as long as the blood is not collected with chelating agents such as E D T A or citrate, which is commonly used in clinical situations. It is important to note, however, that if human serum is used, all appropriate safety measures for working with human blood products must be observed. Storage of serum or plasma leads to dissociation of the G P I - P L D from the native complex with apolipoprotein A1, which causes aggregation of the enzyme. 26 However, this does not affect use of the enzyme as an analytical tool. In addition to the G P I - P L D , serum contains a number of other lipases that investigators must be aware of. 25 S. Stieger, S. Diem, A. Jakob, and U. Brodbeck, Eur. J. Biochem. 197, 67 (1991). 26 M. A. Deeg, M. A. Davitz, A. C. Wolf, E. L. Bierman, and M. C. Cheung, submitted for publication (1995).
[45]
GPI-PHOSPHOLIPASE i~
639
Protein
TLC
I
Ethanolamin¢
DAG
Ja Inositol ( n~ Glucosamlne~ o
4D FFA
°p""° PA _
II °
Organic Extraction with n-butanol. Analysisof Lipid Products by TLC,
0 Origin
Plate developed in CHCI3:CH3OH:KCI FIG. 4. Sample product analysis, n-Butanol is evaporated under an N2 stream. The lipid products are resuspended in CHC13/CH3OH/water (10:10:3, v/v/v) and then applied to a TLC plate. The plate is developed in CHC13/CH3OH/0.25% KCI (55:45:10, v/v/v) [M. A. Davitz, J. Horn, and S. Schenkman, J. Biol. Chem. 264, 13760 (1989)]. The relative positions of various lipid cleavage products are illustrated. PA, Phosphatidic acid; FFA, free fatty acid; DAG, diacyl- or alkylacylglycerol.
These include sphingomyelinase which is present in bovine serum, 27 platelet-activating factor (PAF) hydrolase, a phospholipase A2,28 and a phosphatidylcholine-specific phospholipase C which has been found in rabbit serum. TM In addition, under certain conditions apolipoprotein B, a major constituent of very low and low density lipoprotein, has phospholipase A1 and A2 activity.29'3°However, the only enzyme in serum that is active against GPI anchors is the GPI-PLD. 5-7
Product Analysis Examination of the cleavage products is a crucial part of the analysis. GPI-anchored molecules can be biosynthetically labeled with a variety of 27 M. W. Matthew, D. M. Byers, F. B. S. C. Palmer, and H. W. Cook, J. Biol. Chem. 264, 5358 (1989). 28 D. M. Stafforini, S. M. Prescott, and T. M. Mclntrye, J. BioL Chem. 2629 4223 (1994). 28a M. Deeg, A. Mendez, C. Furlong, and A. Chait, manuscript in preparation. 29 N. Reisfeld, D. Lichtenberg, A. Dagan, and S. Yedgar, F E B S Lett. 315, 267 (1993). 30 S. Parthasarathy and J. Barnett, Proc. Natl. Acad. Sci. U.S.A. 87, 9741 (1990).
640
GPI-ANCHOREDPROTEINS
[45]
different anchor constituents, including [3H]ethanolamine, [3H]glucosamine, [3H]inositol, and 3H-labeled fatty acids (e.g., myristic or palmitic acid). R o s e n b e r r y and colleagues have developed several different chemical methods for chemically tagging the G P I molecules, including use of [azsI]TID [3-(trifluoromethyl)-3-(m-[125I]iododiazirine], which labels the hydrophobic fatty acid m e m b r a n e anchors, 31 and H l a C H O , which radiomethylates free amino groups. 3e It should be noted that radiomethylation of the glucosamine renders the G P I anchor resistant to the action of the G P I PLD. 33 Once the putative G P I molecules have b e e n treated with the G P I PLD, the organically soluble cleavage products can be analyzed by thinlayer c h r o m a t o g r a p h y (TLC) or H P L C . After cleavage, the hydrophilic anchor c o m p o n e n t s can be examined by gel filtration using a BioGel P4 (Bio-Rad, Richmond, C A ) sizing column. 34-37 A complete examination of the analytical procedures is beyond the scope of this chapter, and the subject has been thoroughly reviewed elsewhere4'34-37; Fig. 4 illustrates one possible analytical sequence. Summary Cleavage by the G P I - P L D provides definitive evidence of a minimal G P I structure: glucosamine-phosphatidylinositol. Unlike the case for PIPLC, cleavage by the G P I - P L D is unaffected by acylation of the inositol ring. Thus the G P I - P L D provides an excellent simple enzymatic tool for analyzing the basic core structure of G P I anchors. Acknowledgments Mark A. Deeg is a Pfizer Postdoctoral Fellow. Michael A. Davitz is a Lucille P. Markey Scholar in the Biomedical Sciences. This work was supported by grants from the Lucille P. Markey Charitable Trust and the Council for Tobacco Research.
31W. Roberts and T. Rosenberry, Biochemistry 25, 3091 (1986). 32R. Haas and T. Rosenberry, Anal, Biochem. 148, 154 (1985). 33M. A. Deeg, N. R. Murray, and T. L. Rosenberry, J. Biol. Chem. 267, 18581 (1992). 34A. K. Menon, S. Mayor, M. A. J. Ferguson, M. Duszenko, and G. A. M. Cross, J. Biol, Chem. 263, 1970 (1988). 35A. K. Menon, R. T. Sehwarz, and S. Mayor, EMBO J. 9, 4249 (1990). 36m. K. Menon, S. Mayor, M. J. Ferguson, M. Duszenko, and G. A. M. Cross, J. Biol. Chem. 263, 1970 (1988). 37S. Mayor, A. K. Menon, and G. A. M. Cross, J. Biol. Chem. 265, 6174 (1990).