Characterization of a Human Recombinant Receptor-linked Protein Tyrosine Phosphatase*

The receptor-linked tyrosine phosphatase RPTPa from human brain (Kaplan, R., Morse, B., Huebner, K., Croce, C., Howk, R., Ravera, M., Ricca, G., Jaye, M., and Schlessinger, J. (1990) Proc. Natl. Acad. Sei. U. S. A. 87, 7000-7004) was expressed in insect cells following infection with recombinant baculovirus. Two major forms of the enzyme, with molecular sizes of 98 kDa and 114 kDa, were detected by immunoblot analysis. This heterogeneity could be ascribed to N-linked glycosylation on the basis of two lines of evidence; namely, blockage of glycosylation with tunica- mycin in vivo and removal of carbohydrates by endoglycosidase F in vitro. The 114-kDa form was purified to homogeneity by chromatography on Superose 12 and Mono Q. Compared to the low M, placenta and T- cell tyrosine phosphatases, RPTPa displayed a low optimum pH of 6 and a high K,,, in the micromolar range for TCAll.PTP,

gency hybridization. All except one have two highly conserved PTP domains (4). The role of the second domain is still unclear; it might serve a regulatory rather than a catalytic function ( 5 ) . On the other hand, there are considerable differences in the size and structure of the extracellular segments of these receptors suggesting that they might play different parts in the regulation of cellular processes (reviewed in Refs. 4 and 6 ) . Expression of CD45 is restricted to hematopoietic cells with the exception of mature erythrocytes and their immediate progenitors (1). Its external segment is heavily 0and N-glycosylated and contains cysteine-rich regions. Several "leucocyte common antigen-related transmembrane molecules (LAR ( 6 ) , DLAR, and DPTP ( 7 ) ) have structural similarities with the neural cell adhesion molecules in that they contain type I11 fibronectin and IgG-like domains, while HPTPp has only type I11 fibronectin repeats (4). Their structural features suggest that these molecules might be involved in cell-cell interaction. By contrast, the short external segments of R P T P a (123 residues (4,9,10,11)) and HPTPt (27 residues (4)) show no homology to the other receptor-linked PTPs or to any other transmembrane molecules so far identified.
RPTPa has been cloned from human (4, 10) and murine (9,11) cDNA libraries. The enzyme contains eight putative sites for N-linked glycosylation plus many seryl and threonyl side chains that could be 0-glycosylated. Northern blot analysis indicates that RPTPa is present in many tissues and cell lines suggesting that it might have a more general function in signal transduction.
This report describes enzymatic properties of purified RPTPn following its expression in baculovirus-infected insect cells. Construction of the pVL 941-RPTPn Expression Vector-The full length cDNA of RPTPa (10) was cloned into pML2D (13) using a unique XhoI restriction site found within a multiple cloning site inserted between nucleotides 23 and 375 of the vector. Digestion of this construct with DraI and BglII resulted in a 2.5-kilobase fragment that was ligated into the BamHI site of the pVL 941 expression vector (14) which had first been treated with Klenow fragment and deoxynucleotides to generate blunt ends. The construct pVL 941-RPTPa was purified by repeated CsCl gradient centrifugations; correct orientation of the insert was verified by DNA sequence analysis (15).

Materials-Restriction
Preparation and Purification of the Recombinant Ac-RPTPa Virus-All standard procedures were carried out according to the protocol of Summers and Smith (16). Monolayer cultures of Sf9 cells were grown in Grace's Antheraea medium (12) containing 3.3 g/liter yeastolate, 3.3 g/liter lactalbumin hydrolysate, 10% fetal calf serum (17), 100 units/ml penicillin, 100 pg/ml streptomycin, and 0.25 pg/ ml fungizone. A recombinant baculovirus (Ac-RPTPa) was produced by co-transfecting Sf9 cells with 1 pg of Ac-NPV DNA and 2 pg pVL 941-RPTPa plasmid DNA (16, method I). The method of serial end point dilution was used to purify the recombinant virus (16). Blots were screened by hybridization with [ a-"YP]ATP-labeled RPTPa cDNA. The purity of the final virus suspension was checked by hybridization with a '"P-labeled oligonucleotide consisting of nucleotides 37-66 of the Ac-NPV polyhedrin gene (14). The pVL 941 vector lacks this sequence; therefore, a failure of hybridization of Sf9 cell extracts with this oligonucleotide indicates that these cells had been infected with the recombinant virus only.
Purification of RPTPa from Sf9 Cells Infected with Ac-RPTPa-Three 150-cm' flasks containing 3-4 X 10' cells each were infected with recombinant virus at a high (>3) multiplicity of infection and incubated at 27 "C. Cells were harvested 36 h postinfection by a 5min centrifugation at 3,000 X g. The supernatant containing the virus was kept at 4 "C for further infections. All of the following steps were carried out at or below 4 "C. The cells were homogenized in 10 ml of buffer E containing 0.6 M KC1 in a Dounce homogenizer. The homogenate was centrifuged for 45 min at 100,000 X g (50 Ti rotor, Beckman). The pellet was extracted again in 10 ml of buffer E and centrifuged for 45 min at 100,000 X g. The supernatant was discarded, and the pellet was solubilized with 1.2 ml of buffer E containing 1% Triton X-100 (v/v). The membrane extract was first cleared by centrifugation at 12,000 X g for 10 min and then filtered through a 0.45-pm filter (Acro LC13, Gelman Sciences). The extract was applied to a Superose 12 (preparation grade) column (HR 16/50, Pharmacia LKB Biotechnology Inc.), equilibrated in buffer AT containing 100 mM NaC1. The chromatography was performed a t a flow rate of 1 ml/min; 1-ml fractions were collected and assayed. Tyrosine phosphatase-containing fractions were pooled, dialyzed against buffer A T for 1 h, and then applied to an FPLC-Mono Q column (HR 5/5, Pharmacia). The column was subsequently washed with buffer AT and buffer A T containing 50 mM NaC1. Tyrosine phosphatase activity was eluted with a 20-ml linear salt gradient (50-350 mM NaCl in buffer AT, flow rate: 0.5 ml/min, fraction size: 0.5 ml). Active fractions were pooled and dialyzed for 1 h against buffer AT containing 20% glycerol (v/v). Aliquots (50 pl) of the protein solution were frozen in liquid nitrogen and stored at -70 "C.
Antibodies Raised against RPTPa Peptides-A peptide corresponding to the C terminus of RPTPa (residues 785-802) was cross-linked to keyhole limpet hemocyanin with l-ethyl-3-(3-dimethylaminopro-py1)carbodiimide (18). Three New Zealand white male rabbits were immunized three times at 21-day intervals with that complex containing 150 pg of peptide plus complete Freund's adjuvant for the first injection only. Subsequently, rabbits were immunized at 14-day intervals until antibody titers decreased. A 1:lOO dilution of serum was used for immunoblotting.
Phosphatase Assays-These were performed in buffers M6 and H7.5 (or AT for screening column fractions). Aliquots of 60 pl containing the substrate and, where indicated, effectors were incubated for 10 min at 30 "C. The reaction was stopped by adding 180 pl of 20% trichloroacetic acid (w/v). The suspension was spun in a microcentrifuge for 10 min, and the radioactivity of 200 p1 of the supernatant was determined. As substrates, RCML and MBP were prepared and phosphorylated in the presence of [Y-"~P]ATP as de- scribed previously (19,20). One unit is defined as the release of 1 nmol of phosphate/min. Substrate concentration represents the concentration of phosphotyrosyl residues. Protein concentration was determined according to Bradford (21). Purification-The 114-kDa form of RPTPa was purified to electrophoretic homogeneity from a membrane fraction of Sf9 cells (Fig. 1A). As summarized in Table I  " PTP activities were determined with 0.67 p~ MBP in buffer AT.

Properties of a Receptor-linked
Tyrosine Phosphatase 12213 0.14 mg of pure RPTPty. The activity of the membrane fraction eluted from the Superose 12 column in a broad peak (Fig. 2) corresponding to proteins with molecular masses ranging from 160 to 1000 kDa (void volume of the column). Immunoblot analysis showed that most of the 98-kDa form eluted ahead of the ll4-kDa species (compare lane b and c in Fig. 1B). In order to isolate the 114-kDa protein, only the four peak fractions (40-43) were pooled and subjected to FPLC Mono Q chromatography.
The PTP activity emerged in an almost symmetrical peak (Fig. 3). The remaining 98-kDa species eluted from t,he Mono Q column in the trailing fractions of the activity peak together with traces of proteins of lower M,, which were probably generated in the course of purification (compare lane d and p in Fig. IR). To avoid contamination by these lower M, species, only five fractions (14-18) were pooled. The enzyme could be stored at -70 "C for months without significant loss of activity.
Glycosylation--It was likely that the two RPTPti forms observed resulted from a difference in carbohydrate content, since the receptor displays eight putative N-linked glycosylation sites in its external domain (10). To address this question, cells were grown in the presence or absence of tunicamycin, an inhibitor of N-linked glycosylation (22). As shown in Fig. 4A Only endoglycosidase F, which removes Nlinked glycosyl residues (23), was able to convert the 114-kDa form into the 98-kDa species (Fig. 4B, lane c) detected. This was mainly due to the acidic optimum pH of 6 (with half-maximum activity observed at pH 5 and 6.7, respectively) exhibited by RPTPn when RCML and MBP were used as substrates. Therefore, all further characterizations were carried out at both pH 6 and 7.5 to cover the physiological range.
Substrate Specificity and Effect of Divalent Cations and EDTA-Like all other members of the protein tyrosine phosphatase family, RPTPtv was specific for tyrosyl residues, displaying no activity toward MBP or histones phosphorylated by the CAMP-dependent protein kinase. Such substrates are readily dephosphorylated by serine/threonine-specific protein phosphatases (24). RPTPrv showed a complex behavior since its response toward different effecters varied with pH and the nature of the substrate. As shown in Fig. 5A, at pH 6 and with RCML as substrate, stimulation of activity was the largest with Mn')+ followed by Mg" > Ca" > EDTA. At. pH 7.5, however, the enzyme was unaffected or slightly inhibited by these compounds. By contrast, with MBP as substrate, all effecters were inhibitory regardless of pH (Fig.  5R). Note that the optimum concentration of Mn2+ is far above the physiological range. Calmodulin (10 PM) in the presence of 1 mM calcium was without effect. The kinetic parameters of RPTPm determined with RCML and MBP are summarized in Table II. The low activity toward RCML observed for the enzyme at its optimum pH of 6 and in the absence of effecters can be attributed to its high K,,, of 80 pM; Mn" decreases this value 7-fold. At pH 7.5 or with MBP as substrate, Mn'+ causes a 3-to 5-fold increase in K,,,. Under all conditions, V,,,., remained essentially unaffected by Mt?+ showing variations less than 2-fold. Effect of Polycations-As shown in Fig. 6, the influence of polycations such as spermine and spermidine on RPTPtv was similar to that observed with Mn"+: activity toward RCML was enhanced 12-to 14-fold at pH 6 but not at pH 7.5, while dephosphorylation of MBP was inhibited at either pH. A lofold higher concentration of spermidine than spermine was required for activation as seen by comparing the scales of when RCML rather than MBP was used as substrate except at micromolar concentrations (Fig. 7). Likewise, R P T P a was 10 times more susceptible to Zn2+ inhibition in the presence of RCML, displaying ICso values of 40 and 400 PM with RCML and MBP, respectively (Fig. 8).

Expression of the receptor-linked tyrosine phosphatase
RPTPa in Sf9 cells occurred in a narrow time frame, with 90% of the protein being produced between 24 and 36 h postinfection. Approximately 90% of the phosphatase was present in a 114-kDa form with the remainder at 98 kDa, the difference being due to N-linked glycosylation. However, the size predicted on the basis of the amino acid sequence (10) is 88 kDa. This discrepancy could result from the presence of 0-linked carbohydrates, since the extracellular segment exhibits a serine/threonine content of approximately 30%; such modification would not be affected by tunicamycin or endoglycosidase F. Alternatively, there could be an asymmetry in the transmembrane molecule. The murine analog of RPTPa, with 794 residues versus 802 for the human receptor and 95% sequence identity, has recently been expressed in COS cells (9). Its higher than expected molecular mass of 130 kDa could be due to a greater extent of glycosylation. release appreciable amounts of the phosphatase but removed contaminating proteins, thereby providing a valuable purification step. Although R P T P a represents 40-50% of total protein in Triton X-100 extracts, two chromatography steps were needed to obtain the pure 114-kDa form. The low yield of 5-10% resulted from ( a ) a loss of 80% in the Mono Q chromatography step probably due to a tendency of the receptor to stick to the filter of the column, and ( b ) pooling only a few of the peak fractions in order to minimize contamination. There is no evidence that the low yield is due to the loss of a n activator during the purification, since the calculated enrichment of 2.6-fold is consistent with the proportion of the phosphatase in the membrane extract.
A comparison between the properties of the intracellular and transmembrane PTPs with RCML as substrate (taking into account that the low M , enzymes (20)' and CD45 (25) were assayed at pH 7.2 and RPTPa at pH 6) leads to the following conclusions. The following model, consistent with the above data, can be suggested. R P T P a exists in two conformations in vitro, determined by the pH of the system (see Fig. 9). In the absence of "positive" effectors, such as Mn'+, the phosphatase is almost inactive toward RCML. At its optimum pH of 6, it exhibits a high V,,, but a low affinity for substrate (K,,, -80 PM). At pH 7.5, the opposite is true; that is, the phosphatase displays a low V,,, but a relatively high affinity for substrate (K,,, -10 p~) . Addition of Mn'+ switches the high affinity form into the low affinity form and vice versa. This results in an enzyme that is maximally active a t p H 6 and almost inactive at pH 7.5. This model also accounts for the MBP data since that particular substrate can act as its own "positive" effector. tion between the divalent cation and MBP for the "activator" binding site.
This study demonstrates that RPTPa has the capacity to dephosphorylate tyrosyl residues in proteins. However, the low optimum pH it displays in vitro raises the question as to its activity within the cell. An interesting possibility is that binding of a ligand leads to a change in conformation that converts RPTPa to the high affinitylhigh V,,, state i n vivo.
Another possibility, of course, is that RPTPa exhibits different properties when acting on its physiological substrate(s).