Cytoplasmic Tail Deletion Converts Membrane Immunoglobulin to a Phosphatidylinositol-linked Form Lacking Signaling and Efficient Antigen Internalization Functions*

Membrane-bound immunoglobulin (mIg) is the anti- gen receptor on B lymphocytes mediating early events in antigen presentation and signal transduction. Wild- type human mIgM constructs transfected into the murine B-cell lymphoma A20 are expressed as transmembrane proteins with antigen presentation and signaling functions comparable to the endogenous mIgG2A; the transfected wild-type mIgM is internalized rapidly after anti-Ig cross-linking. Transfected constructs lacking the normal three-amino acid cytoplasmic tail are expressed exclusively as phosphatidylinositol- linked proteins, lack both antigen presentation and signal transduction functions, and are internalized slowly following anti-Ig binding. The molecular mass of the cytoplasmic tail-deleted phosphatidylinositol- linked Ig molecule is consistent with cleavage of the transmembrane residues during processing. Cyto- plasmic domains may therefore regulate the mode of expression of membrane proteins and thereby influ- ence their functional capabilities.

All transmembrane proteins contain hydrophobic sequences that traverse the lipid bilayer, with charged intraand extracellular domains. By comparison, it is thought that prior to PI linkage, immature precursor proteins are processed by deletion of a number of hydrophobic residues at the carboxyl terminus (1, 2, 7). A unique signal sequence for the post-translational modification has not been identified, and it appears likely that a variety of sequences differing in length and primary structure (albeit predominantly hydrophobic) will suffice (1, 7). Of note, a single Asp + = Val amino acid substitution converts the PI-anchored Qa-2 antigen to a transmembrane molecule with a three-amino acid (Asn-Arg-Arg) cytoplasmic tail (8). It is likely, however, that the mutated Qa-2 residue is distant from the ultimate PI-linkage site, and the putative cytoplasmic and transmembrane domains, including mutated residue, may be deleted prior to attachment of the PI anchor (2). In comparison, a single Ser + Phe substitution in the transmembrane domain of FcyRIII determines membrane topology either by directing PI-linkage processing or by altering the interaction of FcyRIII with the y-chain of FctRI (9). In this case, the SerZo3 residue in the immediate extracellular domain may be at or very near the site of PI attachment (9).
Membrane-bound immunoglobulins (mIg) are integral transmembrane proteins on B-cells and are capable of binding and internalizing specific antigen and directing its subsequent processing and presentation to T-cells (10,11). In addition, cross-linking mIg with specific antigen or anti-Ig antibodies induces a cascade of intracellular signals including an increase in cytoplasmic calcium, altered inositol phosphate metabolism, and expression of oncogene mRNA, resulting in movement of resting B-cells into the cell cycle (reviewed in Ref.

12).
All isotypes of mIg have (a) extracellular antigen binding domains; ( b ) highly conserved, largely hydrophobic transmembrane sequences; and ( c ) cytoplasmic tails of variable lengths (3-28 residues). IgM and IgD, the antigen receptors on mature, resting B-cells, have conserved three-amino acid (Lys-Val-Lys) cytoplasmic domains (13). Given the short cytoplasmic sequence, it is not surprising that mIgM and mIgD have no known kinase or other enzymatic activity, and are therefore likely to interact with other transmembrane or cytoplasmic molecules to mediate their known biological effects.
We previously demonstrated that cotransfection of an antigen-specific IgM heavy chain construct (containing wildtype human transmembrane and cytoplasmic domains), with the appropriate antigen-specific K chain into the A20 murine B-cell lymphoma line, faithfully reproduced antigen presentation and signaling functions of mIg (14). Nonconservative mutations affecting the transmembrane Tyr587 ( T~r~~~-S e r~~ + Val-Val, mutant TM: YS/VV) abrogated both presentation and signaling activities (14). We also showed that, although mutants with deletion of the cytoplasmic tail (Cyto: A) continued to be expressed a t high levels on the cell surface, these mutants also lost both antigen presentation and signaling functions (14). We sought to investigate whether the loss of functions in the mutant mIg could be attributed to a change in the form of their cell surface expression.

EXPERIMENTAL PROCEDURES
DNA Constructs-The heavy chain constructs have been described in detail previously (14) and are membrane and cytoplasmic mutants of a phosphorylcholine-specific human p heavy chain which is exclusively membrane-bound. Briefly, the rearranged phosphorylcholinespecific, heavy chain variable region, derived from the S107 plasmacytoma, was a gift from Dr. s.-P. Kwan. This sequence, along with the murine heavy chain enhancer and the human heavy chain promoter, was inserted into the ppA14 plasmid which contained the human p heavy chain exons modified by Bal3I digestion to delete the polyadenylaton cleavage site for secreted p (15). To facilitate mutagenesis, a unique KpnI restriction enzyme site was generated at the location of an SstI site just 5' of the membrane exons (14). I n vitro site-specific mutagenesis was performed as described by Kunkel et al. (16), using oligonucleotide sequences synthesized on an Applied Biosystems 340A synthesizer (14). The rearranged, phosphorylcholinespecific murine K light chain, also a gift of Dr. S.-P. Kwan, was ligated into a plasmid containing a G418 resistance gene (14). All plasmid constructs for transfection were sequenced to confirm correct placement of mutations (14).
Cells and Transfections-A20 murine B-cell lymphoma cells (y2A+, K + , H-2d) were grown in RPMI 1640 medium containing 10% fetal calf serum, 2 mM L-glutamine, 50 p M 2-mercaptoethanol, 50 units/ ml penicillin, and 50 pg/ml streptomycin (medium). Transfection was performed by electroporation (17) using 10-15 pg of linearized heavy chain and 1 pg of linearized light chain plasmids. Transfected cells were selected and continuously maintained in medium containing 0.6 mg/ml G418 (corrected activity, GIBCO); resistant cells expressing levels of transfected mIgM comparable to the levels of endogenous mIgG2A were selected by repeated rounds of fluorescent staining and cell sorting on an Ortho Cytofluorograf 11s. Single-cell clones were derived by limiting dilution. For all mutations independent clones from separate transfections gave comparable results. High stringency RNase protection assays (18) were used to confirm accurate construct transcription.
Analysis of mlg Internalization by Flow Cytometry-Clones were incubated at 2 X 10'/ml in 100 pg/ml F(ab'), rabbit anti-mouse IgG2A or F(ab'), rabbit anti-human IgM for 30 min on ice, washed twice, and incubated at 106/ml at 37 "C for varying lengths of time. Endocytosis was stopped by adding ice-cold PBS/O.l% azide. Cells were then stained with fluorescein isothiocyanate-goat anti-rabbit Ig, fixed, and analyzed by flow cytometry as described above.
PI-specific Phospholipase C (PI-PLC) Treatment-PI-PLC prepared from Bacillus thuringiensis (20) was a generous gift from Dr. Martin Low. PI-PLC treatment was performed by incubation for 1 h at 37 "C at lo7 cells per ml in RPMI/l% bovine serum albumin (BSA) containing 2 units/ml PI-PLC. Washed cells were then either stained for flow cytometry or solubilized for SDS-polyacrylamide gel electrophoresis. Supernatants from PI-PLC-treated cells were DreDared bv treating cells a t 10"/ml in RPMI 1640/1% BSA with 30 &i&/ml Pi-PLC for 1 h a t 37 "C.
Enzyme-linked Immunosorbent Assay Protocol-The enzymelinked immunosorbent assay procedure was adapted from Boom et al. (21). Microtiter plates (Microtest 111 flexible plates, Becton Dickinson, Oxnard, CA) were coated overnight at 4 "C with 10 pg/ml polyvalent goat anti-mouse IgG2A or anti-human IgM (Southern Biotechnology Associates, Birmingham, AL), washed, and subsequently blocked with 1% BSA in PBS. Nonidet P-40 cell lysates (22) or supernatants from 1-or 24-h incubations were added to the wells for 2 h a t room temperature. Adherent Ig was detected by incubation with 10 pg/ml alkaline phosphatase-conjugated goat anti-mouse IgG2A or anti-human IgM (Southern), addition of substrate, and measuring on a Bio-Rad model 2550 EIA reader. Standard curves were generated using myeloma proteins of the appropriate isotypes (ICN Biomedicals); no crossreactivity of the standards was detected and the assay was sensitive to 20 ng/ml of Ig protein.
SDS-Polyacrylamide Gel Electrophoresis, Western Blot, and Autoradiographic Analyses-Cells were washed and lysed a t 107/ml in 1% SDS containing 0.6 M 2-mercaptoethanol. 5 X lo5 cell-equivalents were applied to each gel lane. Cell supernatants prepared as described above were recovered and boiled 5 min in the above SDS sample solution; IO6 cell-equivalents were applied to each lane. Samples were electrophoresed on 10% SDS-polyacrylamide gels with appropriate prestained molecular mass standards (Bio-Rad) and transferred to nitrocellulose using a Bio-Rad trans-blot cell. After blocking overnight a t 4 "C with 5% powdered milk, 0.1% Tween 20, 0.1% azide in PBS, pH 8, the nitrocellulose was washed with 0.05% Tween 20, 1% BSA, PBS (TBS) and probed with a 4-pg/ml solution of alkaline phosphatase-conjugated goat anti-human IgM (Southern) in TBS for 1 h at room temperature. The blot was then washed with TBS and PBS, and the bound alkaline phosphatase was visualized by the method of Leary et al. (23).
For [y-32P]ATP metabolic labeling of cells, 2 X lo7 cells of each clone were cultured for 24 h with 330 pCi of [y-"PIATP (Amersham Corp.) in medium containing G418, washed, and then incubated for 1 h on ice with 50 pg of goat anti-human IgM (Southern) in 1 ml of medium. The cells were washed three times in medium containing 0.1% azide and incubated for 2 h with PI-PLC as described above. Supernatants were recovered and gently mixed overnight at 4 "C with 4 mg/ml rabbit anti-goat Ig covalently bound to microspheres (Kierkegaard and Perry Laboratories, Inc., Gaithershurg, MD). The microspheres were washed five times with PBS/l% BSA and then boiled 5 min in the above SDS sample solution which also contained 1 mg/ml goat Ig (Sigma). lo7 cell-equivalents were applied to each lane, and the samples were separated by SDS-polyacrylamide gel electrophoresis as described above. After fixing for 1 h in 25% methanol, 10% acetic acid, the gel was dried and exposed to Kodak X-OMAT film with an enhancer screen a t -70 "C.

RESULTS AND DISCUSSION
To test the possibility that certain mutant mIg molecules are expressed as PI-anchored proteins, transfected cells were cultured for varying times with and without PI-PLC, and the quantity of released and cell-associated Ig measured by enzyme-linked immunosorbent assay. As shown in Table I, treatment of the Cyto: A cells with PI-PLC for 1 h leads to quantitative release of the transfected human IgM. In comparison, the transfected wild-type and TM: YS/VV human ND 520 " ND, not determined. mIgMs are not released by enzymatic treatment; the endogenous mouse mIgG2A is also unaffected. Interestingly, the Cyto: A IgM is also released spontaneously during a 24-h culture, perhaps due to endogenous phospholipase activity. The wild-type and TM: YS/VV transfected clones show no accumulation of human IgM in 24-h supernatants because the transfected human mIgM construct was generated with a deletion of the polyadenylation cleavage site necessary for synthesis of secreted Ig (14). In contrast, the endogenous murine IgG2A is detected in 24-h supernatants from all the cell lines over 24 h, presumably because it can be expressed as a membrane or secreted Ig. These results suggest that the Cyto: A human IgM construct is PI-linked, whereas the wildtype and TM: YS/VV clones are insensitive to PI-PLC treatment, and are therefore transmembrane proteins analogous to the endogenous mIgG2A.
To determine the proportion of Cyto: A cells with PI-linked mIgM, we measured the PI-PLC-induced loss of mIg by immunofluorescent staining and flow cytometry. Fig. 1 shows a uniform reduction in fluorescent staining for human IgM on the entire Cyto: A population after PI-PLC treatment. Wild-type and TM: YS/VV clones show identical staining with or without PI-PLC exposure, confirming that these Clones were incubated with (-) or without (control, dotted lines) PI-PLC. After washing, cells were incubated with F(ab')2 human IgM or with F(ab'), rabbit IgG (background, dashed lines), followed by fluorescein isothiocyanate-goat anti-rabbit Ig. The clones are A, wildtype, R, Cyto: A; and C, TM: YS/VV. The x axis is logarithmic, representing three logs of fluorescence intensity.
constructs are PI-PLC-insensitive. Likewise, the endogenous IgG2A is insensitive to PI-PLC on all clones (results not shown). Loss of staining of the PI-linked glycoprotein J l l d (24) following PI-PLC hydrolysis confirms that all clones are uniformly capable of PI-linkage, and are comparably sensitive to PI-PLC (results not shown).
By Western blot analysis (Fig. 2 A ) , wild-type (lunes b and f ) and TM: YS/VV (lunes d and h ) clones have a similar PI-PLC-insensitive doublet corresponding to the human IgM heavy chain, with apparent molecular masses of 74 and 71 kDa. It is not clear whether the doublets represent variably glycosylated molecules or mature cell surface and immature intracellular Ig, although we favor the latter possibility. By comparison, the Cyto: A clone (lunes c and g) has a doublet with apparent molecular masses of 72 and 70 kDa, both of which are released after PI-PLC treatment. A band a t 72 kDa is present in the supernatant of PI-PLC-treated Cyto: A cells only (lune h). Western blots probed for the endogenous murine IgG2A show doublets a t 63 and 62 kDa which are insensitive to PI-PLC in all clones (results not shown). The difference of approximately 2 kDa in the apparent molecular mass between the wild-type and Cyto: A Ig p heavy chains is consistent with deletion of most, if not all, of the putative transmembrane 26-amino acid residues prior to PI-linkage.
T o verify that the Cyto: A IgM is directly PI-anchored, rather than associated indirectly with another PI-linked molecule, wild-type, Cyto: A, and TM: YS/VV clones were metabolically labeled with [-/-"'PP]ATP. After PI-PLC treatment, the human IgM was immunoprecipitated from supernatants and analyzed by SDS-gel electrophoresis. As shown in Fig.  2B, only the Cyto: A cells (lune b) have anti-IgM precipitable, PI-PLC-sensitive phosphorylated proteins with the appropriate molecular mass, confirming that the transfected construct is directly PI-linked.
It is likely that the Cyto: A mutant is incapable of signal transduction (14) because it can no longer appropriately associate with other transmembrane or cytoplasmic molecules which mediate the response. Similarly, a PI-anchored protein is likely to be either internalized poorly or aberrantly targeted upon internalization, so that the antigen presentation function of mIg may be rendered less efficient (14).
To further characterize the nature of the defect in the nonfunctional mutants, we adapted flow cytometric techniques to analyze mIg internalization (Fig. 3). mIg on the clones was cross-linked with either F(ab'), rabbit anti-mouse 72A (RAMy2A, internal control) or F(ab'), rabbit anti-human p (RAHp). The bound ligand was then allowed to inter-  nalize for increasing periods of time prior to staining for residual surface rabbit Ig using fluoresceinated goat antirabbit Ig. For all clones, the endogenous IgG2A is rapidly internalized after cross-linking (represented by a decrease in fluorescence staining intensity; maximum shift occurs by 15 min), in agreement with other published data (10,11). In control experiments with 1% azide, no internalization is seen after cross-linking (not shown). Wild-type and TM: YS/VVtransfected IgM are also rapidly internalized with kinetics virtually identical to the endogenous IgG2A. In contrast, the PI-anchored Cyto: A IgM is slowly internalized, with little diminution of fluorescence intensity at 15 min (Fig. 3) or even at 60 min (not shown). This slow rate of internalization may be largely due to constitutive membrane turnover (25,26). Comparable results have been obtained with radioiodinated antigens which are bound by the transfected mIgM (results not shown). Of note, while the PI-linked Cyto: A mIgM is much less efficient than the transmembrane wild-type mIgM in terms of antigen internalization and subsequent presentation, it is still more efficient than fluid-phase pinocytosis in mock-transfected cells lacking any form of mIgM (14).
Thus, we have shown that tail-deleted mIg mutants, which are defective in intracellular signaling and antigen presentation (14), have been converted from transmembrane to PIanchored molecules, and are slowly internalized at a rate attributable to normal membrane turnover. These results demonstrate that deletion of the cytoplasmic tail of integral transmembrane proteins may facilitate the processing necessary to generate P I linkage. These results also suggest that P I anchorage may not be dependent on specific signal sequences; in the absence of a charged cytoplasmic tail, the alternative PI linkage may be preferentially used for cell surface expression. As a practical application, cytoplasmic tail-deleted constructs of a variety of surface molecules (e.g. T-cell antigen receptors) may permit relatively easy generation of soluble forms after PI-PLC treatment.
Furthermore, conversion of mIg from a transmembrane to a PI-linked form interferes with normal signal transduction and efficient antigen internalization, presumably since the mIg can no longer interact with other transmembrane or cytoplasmic proteins important for these functions. Although it has been suggested that PI-linked proteins (specifically Ly-6 and Thy-1) can directly mediate T-cell activation following antibody cross-linking (27,28), alternative explanations for such "activating" PI-linked molecules include a role in T-cellaccessory cell interactions (29) or surface interaction with other, transmembrane signal-transducing molecules.
Finally, the mIgM mutants Cyto: A and TM: YS/VV, both of which are inferior to wild-type constructs in antigen presentation, show distinct cell surface linkages and internalization kinetics. In light of the observation that wild-type and TM: YS/VV clones internalize bound ligand at comparable rates, it is interesting to speculate that the TM: YS/VV mutant may have defective antigen presentation function (14) because of abnormal intracellular targeting after internalization. We are currently investigating such a possibility using subcellular fractionation techniques.