Internalization of the Lymphocytic Surface Protein CD22 Is Controlled by a Novel Membrane Proximal Cytoplasmic Motif*

CD22 is a key receptor on B-lymphocytes that modulates signaling during antigenic stimulation. We have defined a novel cytoplasmic motif in human CD22 that controls its unusually rapid turnover at the plasma membrane. Chimeric and mutated CD22α cDNA vectors were constructed and stably transfected in CD22-negative Jurkat T-lymphocytic cells. Two assays were employed to measure CD22α internalization: first, cytoplasmic uptake of radioiodinated anti-CD22 monoclonal antibody; and second, lethal targeting of a toxin, saporin, into cells via CD22 using bispecific F(ab′)2([anti-CD22 × anti-saporin]) antibody. Results showed that CD22α lacking a cytoplasmic tail was not internalized and that replacement of the cytoplasmic tail of CD19 with that of CD22α resulted in a chimeric molecule that behaved like CD22α and internalized rapidly. Step-wise deletion of the cytoplasmic tail of CD22α located the internalization motif to a polar region of 11 residues (QRRWKRTQSQQ) proximal to the plasma membrane, a part of the molecule predicted to form a coil or turn structure. Interestingly, additional CD22 mutants showed that the two glutamine residues sandwiching the serine are critical to internalization but that the serine itself is not.

CD22 is a 135-kDa B-lymphocyte-specific glycoprotein and a member of the recently described sialoadhesin family of molecules (1)(2)(3). It is a key accessory molecule that first appears at the late pro-B-cell stage of differentiation as a cytoplasmic protein and is then expressed simultaneously with IgD as a surface membrane receptor on most mature B-cells (2). Although the nature of the CD22 ligand(s) is not fully defined, it is known to bind sialoglycoconjugate NeuAc␣2-6Gal␤1-4GlcNAc, which is widely distributed on N-linked carbohydrates (3,4). The principal function of CD22 is to regulate B-cell responses, which is probably achieved by recruiting key signaling molecules to the antigen receptor complex (5,6). The cytoplasmic domain is rapidly tyrosine-phosphorylated upon ligation of the B-cell antigen receptor and associates with a range of intracellular signaling molecules that includes tyrosine phosphatase SH2-domain containing tyrosine phosphatase-1, tyrosine kinases Lck and Syk, phospholipase C-␥1, and phosphatidylinositol 3-kinase (1,2,7). The importance of CD22 in modulating B-cell responses has recently been confirmed in knockout mice, which show augmented antibody responses, expanded peritoneal B-1 cell populations, and increased levels of circulating autoantibodies (8 -10).
In humans, two isoforms of CD22 have been identified (11,12). A larger species, CD22␤, which has seven extracellular Ig-like domains and a cytoplasmic domain containing 141 amino acids, and a smaller isoform, CD22␣, which has five extracellular Ig-like domains (domains 3 and 4 are deleted) and a cytoplasmic region that is 23 amino acids shorter than that of CD22␤. Although CD22␤ is by far the predominant species, and the only form identified in mouse, a smaller immunoprecipitate has been found that may correspond to the CD22␣ in human (13). Both isoforms contain tyrosine residues in their cytoplasmic tails, 3 residues in the ␣ form and 6 in the ␤ form. Most of these tyrosine residues are arranged to form immunoreceptor tyrosine-based inhibition motifs (ITIM), consisting of a 6-amino acid stretch with the consensus sequence (I/L/V)XYXX(L/V), and/or potential immunoreceptor tyrosine-base activation motifs (ITAM) (12, 14 -16). In addition, CD22 contains a YXXM sequence recognized by the NH 2 -terminal SH2 domain of the p85 subunit of phosphoinositide 3-kinase. Thus, CD22 has the capacity to modulate B-cell behavior via a wide range of intracellular, tyrosine-dependant signaling molecules.
In addition to any signaling role, CD22 shows an unusual and unexplained capability to internalize rapidly from the cell surface to the cytoplasm. Recent studies have shown that CD22 internalizes constitutively on unstimulated B-cell lines, and this is followed by degradation without detectable recycling (17). This internalization probably explains, at least in part, the success achieved when CD22 has been employed as a target molecule for delivering toxins into neoplastic B-cells using antibody immunotoxins (18,19) and bispecific antibodies (BsAbs) 1 (20 -22). We find BsAbs with specificity for CD22 and the ribosome-inactivating protein, saporin ([anti-CD22 ϫ antisaporin]), is unusually effective at delivering saporin into neoplastic B-cells and thereby inhibiting protein synthesis and eradicating tumors in animals (21) and patients (22). Most other membrane molecules do not work as efficiently (20). In the present study, we have investigated the molecular structure of the cytoplasmic domain of CD22 and identified a new motif that controls its internalization from the cell surface and may account for its unusual immunotherapeutic properties.

MATERIALS AND METHODS
Human Cell Lines-The Burkitt's lymphoma cell line Daudi and the acute T-cell leukemia cell line Jurkat were used in these studies. Both lines were maintained in RPMI 1640 medium containing 10% fetal calf serum (Life Technologies, Inc.) and supplemented with 1 mM glutamine, 1 mM sodium pyruvate, 100 IU/ml benzyl penicillin, and 100 g/ml streptomycin sulfate. Cells were maintained continuously in the * This work was supported with funds from the Tenovus Cancer Charity, Cardiff, United Kingdom and the Leukaemia Research Fund. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
cDNA and Construction of CD22 Mutants-cDNA clones of human CD19 and CD22␣ were kind gifts from Dr. Brian Seed. The CD19 clone provided, which will be called CD19 BS throughout this paper, is not the full-length, wild type transcript, but a truncated version that has its cytoplasmic domain reduced from the normal 242 to 148 amino acids long (27). Fortunately, for the present study we were only interested in CD19 BS as a control surface protein that could be engineered into a chimeric protein containing the extracellular and transmembrane region of CD19 BS and the cytoplasmic tail of CD22␣.
As discussed above, the CD22␣ molecule is a minor isoform of CD22 that may not be expressed in normal B-cells. Because of its truncated cytoplasmic domain and reduced number of tyrosines, it is likely to have a number of signaling properties that are different from those of the predominant CD22␤ isoform. For the purposes of this study, the membrane proximal regions of the molecule that control internalization (see under "Results") are identical in CD22␣ and CD22␤.
Both CD19 BS and CD22␣ cDNA clones were subcloned into pcDNA3 (Invitrogen, San Diego, CA) using the HindIII-NotI sites for CD19 BS and the XhoI site for CD22␣. Deletion of the CD22␣ cytoplasmic tail was accomplished by PCR using primer pairs 22␣F (TTGAATCTCGAGAC-GCGGAAACAGGCTTGCAC) and 22␣⌬cytR (TCGCTAGATCTAGAGC-CCACAGATTGCCAGGA). To delete the whole cytoplasmic tail, a stop codon (TAG) was introduced after Leu 510 in the transmembrane region. PCR was performed in a Protocol Thermal Cycler (AMS Biotechnology Ltd.) using CD22␣ full-length cDNA as template. All reagents were from the Repli-Pack kit (Boehringer Mannheim). The PCR construct generated contained both the extracellular and transmembrane domains and was then subcloned into pGEM-T vector (Promega, Madison, WI) for sequencing and into expression vector pcDNA3 for transfection.
To replace the cytoplasmic region of CD19 BS with that of CD22␣, a two-step recombinant PCR strategy was used (28). In this method, a pair of chimeric PCR primers was designed to span the chimeric junctions. The 5Ј-half of one chimeric primer, 19/22␣F (CAACGTCGCTG-GAGCTTAAGATGAAGAATGCCCA), bound to the 3Ј-end of the transmembrane region of the antisense strand of CD19 BS cDNA between base pairs 939 and 954, whereas the 3Ј-half of the primer bound to the 5Ј-end of the cytoplasmic region of the antisense strand of CD22␣ cDNA (base pairs 1622-1638). A second chimeric primer, 19/22␣R (TGGGC-ATTCTTCATCTTAAGCTCCAGCGACGTTG), the complimentary oligonucleotide to 19/22␣F, was made so that its 5Ј-end bound to the sense strand of CD22␣ between base pairs 1622 and 1638 and the 3Ј-end recognized the sense strand of CD19 BS between base pairs 939 and 954. Appropriate end primers, 19F (TCATCAAGCTTGGAGAGTCTGACCA-CCATGC) and 22␣R (TCGGAGATCTCCCTGGCCCCCGCTGCCCCAG) were also designed to hybridize with the 5Ј-end of the antisense strand of the full-length CD19 BS cDNA (base pairs 1-30) and the 3Ј-end of the sense strand of CD22␣ cDNA (base pairs 2072-2093), respectively. In the first PCR, 2 separate products corresponding to the CD19 BS extracellular region and transmembrane domain (CD19 BS ⌬cyt) and the CD22␣ cytoplasmic domain (CD22␣cyt) were generated using the respective primer pairs of [19F ϩ 19/22␣R] and [19/22␣F ϩ 22␣R]. These PCR fragments were extracted from agarose gels using Qiaex gel extraction kit (Qiagen GmbH, Hilden, Germany), and then 100 ng of each was used in a second PCR. Initially, this reaction was carried out for 10 cycles without end primers (19F and 22␣R) or DNA template. End primers were then added, and a further 10 cycles were completed to give a full-length chimeric CD19 construct with CD22 cytoplasmic region. The chimeric construct (CD19⌬cyt/22␣cyt) was then cloned into pGEM-T vector via HindIII and BamHI sites for sequencing and further subcloned into eukaryotic expression vector pcDNA3 via HindIII and NotI.
To delete individual tyrosine residues Tyr 556 , Tyr 566 , and Tyr 600 , 3Ј primers containing stop codons were used to amplify CD22␣ template with the 5Ј primer 22F via PCR. This yielded deletion mutants (CD22␣Y⌬556, CD22␣Y⌬566, and CD22␣Y⌬600) in which stop codons replaced the respective tyrosine residues. Deletion mutants CD22␣⌬-Q513, CD22␣⌬G524, and CD22␣⌬E527 were constructed in a similar manner using 3Ј primer to respectively replace the corresponding amino acid residues with stop codons. Mutations of Ser 521 to Ala 521 and Gln 520 /Gln 522 to Leu 520 /Leu 522 were carried out using 3Ј primers containing the corresponding changes in the codons. The sequences of all constructs were confirmed by dideoxy DNA sequencing (Sequenase, version 2.0, Amersham Pharmacia Biotech). All constructs were then subcloned into pCI NEO mammalian expression vector (Promega) for transfection.
Transfection of Jurkat Cells-Jurkat cells were transfected with cDNA constructs by electroporation based on the procedure described by Andreason and Evans (29). Approximately 5 ϫ 10 6 cells were pelleted, resuspended in a total volume of 0.8 ml with RPMI, and loaded into a 4-mm electroporation cuvette together with 50 g of DNA. After standing for 5 min on ice, cells were electroporated at 300 V, 960 microfarads using a Gene Pulser (Bio-Rad), and then the cuvette was returned to an ice bath for a further 10 min without mixing. The cells were then left at room temperature for 10 min before adding 30 ml of complete RPMI medium. For the next 48 -72 h, the cells were cultured in 75-cm 2 flasks under standard conditions at 37°C in a moist atmosphere of 5% CO 2 before being transferred to 96-well plates in complete medium containing 1 mg/ml Geneticin (Life Technologies, Inc.). Emerging colonies were tested for surface expression by flow cytometry (described previously; Ref. 24) after 3-4 weeks, and positive cells were cloned by limiting dilution.
Internalization of 125 I-Labeled mAbs by Transfected Cells-Anti-CD19 and -CD22 mAbs were trace radiolabeled using carrier free 125 I (Radiochemical Center, Amersham Pharmacia Biotech) and IODO-BEADS (Pierce) as oxidizing reagents according to the manufacturer's instructions. The radioimmunoassays for evaluating surface binding and internalization of the 125 I-labeled mAbs were carried out by a method based on that described by Pelchen-Mattews et al. (30). Cells (10 7 /ml) were incubated with 125 I-labeled mAb (2 g/ml) at 4°C for 1 h to allow binding and then warmed to 37°C to allow internalization. Aliquots were taken at various intervals, washed twice in 10% supplemented RPMI medium before centrifugation at 13,000 rpm for 45 s through a cold solution of 5% bovine serum albumin (in phosphatebuffered saline) and counting the radioactivity bound to the pellet. For the internalized count, the medium in the second wash was replaced by a HCl-titrated buffered RPMI medium (pH2.0), which removes surface bound activity and thus allows the internalized 125 I-antibodies to be measured and surface binding estimated by subtraction of the internalization counts from the total counts.
[ 3 H]Leucine Uptake by Cells Cultured with BsAb⅐Saporin Complexes-We have established that a BsAb that is designed to recognize a toxin, such as saporin, and a suitable cell surface molecule will target the toxin to the cells and allow efficient internalization so that the cells are killed. Protein synthesis levels in target cells exposed to BsAb⅐saporin complexes were assessed by measuring incorporation of [ 3 H]leucine uptake according to methods described previously (20). Briefly, transfected Jurkat cells (1 ϫ 10 5 /well) were cultured for 24 h in 96-well, flat-bottomed plates with 200 l of leucine-free RPMI medium containing 1 g/ml BsAb and the required concentration of saporin. Each well was then pulsed with [ 3 H]leucine (1 Ci/well) for 16 h, and the incorporation of radioactivity into protein was assessed by harvesting the cells onto glass fibers, washing, and counting. Throughout the work, a mixture of two BsAbs was used as described previously (25). Each pair of reagents was selected to bind to the cell via a single epitope on CD19, CD22, or CD38 and to the saporin via two different epitopes. This maneuver allows the BsAb⅐saporin complexes to be captured bivalently and, due to their increased avidity, bind 20 times more avidly to the target cells than with a single BsAb (25).

Transfection of Jurkat T-cells with Human CD19 BS and
CD22␣-We decided to investigate the molecular basis of CD22␣ internalization and in particular the structural features of the cytoplasmic tail that control delivery into the cell. Our strategy was to construct a panel of CD22␣ cDNA containing vectors in which the cytoplasmic tail of CD22␣ had been selectively altered. These vectors were used to make stable tranfectants of CD22-negative Jurkat cells, which were then investigated, after cloning, for their ability to internalize the newly expressed CD22␣.
Details of all the constructs in this investigation are shown in Fig. 1. A total of 12 cDNA constructs were made, including molecules in which the cytoplasmic tail of CD22␣ was spliced onto the extracellular and transmembrane region of CD19 (CD19⌬cyt/22␣cyt), together with a range of CD22␣ mutants containing truncated cytoplasmic tails and molecules in which certain amino acids were changed.
The Jurkat T-cell leukemia line was chosen as a suitable CD22-negative lymphoid cell line in which to express these molecules. Jurkat cells were stably transfected and then screened by flow cytometry to isolate clones that expressed the various mutated CD22␣ molecules. As shown in Fig. 2, both wild type CD19 BS ( Fig. 2A) and CD22␣ (Fig. 2C) were expressed at easily detectable levels, showing that neither surface B-cell marker required associated B-cell restricted molecules for their expression. The CD22␣ clone was employed as a positive control for assessing CD22␣ internalization throughout the remainder of this investigation, whereas CD19 BS Jurkat cells provided our negative controls (see below). Fig. 2 also shows the levels of mutated CD22␣ expression for the other transfectants. With the exception of one clone, CD22␣⌬cyt (Fig.  2J), which gave only weak immunofluorescence staining, relative to other transfectants, all cells expressed levels of surface CD22␣ that were as high as or even higher than that of CD22␤ generally expressed on normal B-cells and B-cell lines (data not shown). The expression of all the constructs was also confirmed by immunoprecipitation of Nonidet P-40 lysates of the radioiodinated transfectants (results not shown).
Cells were incubated with radiolabeled anti-CD22 or anti-CD19 mAb for 1 h at 4°C and then warmed to 37°C for various intervals. The cells were then acid-washed to remove surface bound mAb, and therefore, any increase in the cell-associated radioactivity must be due to the endocytosis of the antigenantibody complex. The results show that both Daudi cells (CD22␤ ϩ ) and CD22␣-expressing Jurkat cells accumulated 125 I-labeled anti-CD22 mAb at a similar rate (Fig. 3A). There was a sharp initial uptake of mAb in the first 2 h, after which the level remained quite steady. The accumulation of intracellular 125 I-labeled mAb by these cells was accompanied by a corresponding loss of surface-associated mAb. In contrast, the Jurkat cells expressing the CD22␣⌬cyt, in which the cytoplasmic region of the molecule had been removed (Fig. 1), were unable to accumulate significant levels of intracelluar radiolabel, and the mAb remained on the surface of the cells. This result is consistent with the cytoplasmic tail playing a central role in CD22␣ turnover at the cell surface and with immunoelectron microscopy data that showed that CD22␣ is strongly clustered around coated pits in transfected Jurkat cells but that following removal of the cytoplasmic tail the truncated molecules are evenly distributed along the plasma membrane (data not shown).
The importance of the CD22␣ cytoplasmic tail in mediating internalization was confirmed using cells expressing the chimeric CD19⌬cyt/22␣cyt molecule in which the CD22␣ cytoplasmic tail had been grafted onto the extracellular and transmembrane domains of CD19 BS (Fig. 3B). Whereas CD19 (endogenous expression on Daudi cells) and CD19 BS (transfected Jurkat) were relatively poor at taking up radioiodinated anti-CD19 mAb, once the cytoplasmic tail of CD19 BS was replaced by that of CD22␣, it behaved like the intact CD22␣ and internalized mAb very effectively. In fact, for reasons that are not understood, uptake of 125 I-labeled mAb by this chimeric molecule was quicker and more extensive than by the transfected CD22␣. These results suggest that there is an internalization signal within the cytoplasmic region of CD22␣.
The Cytoplasmic Tail of CD22 Controls Effective Delivery of Toxin into Target Cells-For the next stage of this investigation we used the Jurkat transfectants as target cells in an assay which measures the cytotoxicity of a toxin, saporin, when delivered to the cells via CD19 BS or CD22␣. In the assay, cells are exposed to a BsAb ([anti-CD22 ϫ anti-saporin] or [anti-CD19 ϫ anti-saporin]) at 1 g/ml together with saporin at various concentrations for 24 h at 37°C, during which period the BsAb⅐saporin complexes are taken up by the transfectants according to the readiness of the target antigen (CD22␣ or CD19 BS ) to internalize. The cultures were then pulsed for a further 12 h with [ 3 H]leucine to establish the level of protein synthesis in any surviving cells. As a positive control to show that all Jurkat cells, transfected or not, were sensitive to killing with BsAb⅐saporin complexes, we used a BsAb directed at CD38 ([anti-CD38 ϫ anti-saporin]), which is expressed constitutively by these T-cells. Fig. 4 shows a typical set of results with transfected Jurkat cells. The CD22␣ molecule is highly effective at delivering BsAb⅐saporin complexes, giving an IC 50 for saporin of approximately 2 ϫ 10 Ϫ10 M, a value close to that achieved with the anti-CD38 BsAb. This showed that saporin delivered by both anti-CD22 and anti-CD38 BsAbs was approximately 2500-fold more toxic than saporin alone (5 ϫ 10 Ϫ7 M). The increase in the toxicity of CD22-specific BsAbs on the CD22␣-transfected Jurkat cells was similar to that observed when B-cell lines were treated with this reagent (20), suggesting that the transfected CD22␣ molecules behave similarly to native CD22␤ on B-cells. In contrast, when the tail of CD22␣ was removed and the For the CD22␣ molecules, the cytoplasmic region starts at amino acid position 510, and for CD19 BS it starts at 295. The final amino acid number is also given. In two CD22␣ constructs (CD22␣S521A and CD22␣Q520L/Q522L), the cytoplasmic region was terminated at amino acid 526, and then a serine at position 521 was converted to an alanine (CD␣22S521A), or two glutamines at positions 520 and 522 were changed to leucines (CD␣22Q520L/Q522L). a, clone name used to represent the Jurkat cells transfected and selected with each molecular construct. b, IC 50 values were taken as the concentration of saporin at which the uptake of [ 3 H]leucine was inhibited by 50% in the cytotoxicity assays (e.g. Fig. 4). The IC 50 values shown are representative results (Ϯ S.D. of triplicate samples) from two or three cytotoxicity assays for each transfectant. IC 50 values varied by no more than 10-fold between cytotoxicity assays. truncated molecule expressed in the CD22␣⌬cyt-transfected cells, then the same anti-CD22 BsAb was completely unable to increase saporin toxicity. This result is important because it is consistent with the internalization data shown in Fig. 3 and shows that the rate at which a molecule is internalized from the cell surface is a key feature in determining whether it will make an effective target for antibody delivery of toxins.
In Fig. 4, we also see that the cytoplasmic tail of CD22 will transform CD19 BS , a molecule that is unable to deliver saporin with BsAb (20) into a useful vehicle for delivering BsAb⅐saporin complexes for cytotoxicity. Thus, whereas Jurkat cells expressing CD19 BS were not susceptible to anti-CD19-specific BsAb⅐saporin complexes, those carrying CD19⌬cyt/22␣cyt were hypersensitive, with an IC 50 for saporin that was below 10 Ϫ11 M.
A Turn Structure Motif Is Essential for Internalization-To investigate whether the cytoplasmic tail tyrosine residues of CD22␣ are involved in the internalization, CD22␣-mutants were constructed in which the cytoplasmic tail was truncated fractionally by introducing a stop codon at the tyrosine residues (CD22␣⌬Y556, CD22␣⌬Y566, and CD22␣⌬Y600) using PCRdirected mutagenesis. The three constructs were expressed at a high level in Jurkat cells (Fig. 2). Cytotoxicity assays showed that these truncated CD22␣ molecules were all equally as good as the unmutated CD22␣ at delivering BsAb⅐saporin complexes into the cells (Fig. 1), showing that the carboxyl-terminal end of the cytoplasmic tail is not directly involved in the internalization motif.
The results using tyrosine mutants described above suggested that the region of cytoplasmic tail nearer to the cell membrane might be critical in the endocytosis process. Crystallographic work (31) and NMR studies (32) suggest that a tight-turn structural motif is involved in the clustering signals for surface molecules, and individual specificity is conferred by side-chain differences. It is thus interesting to note that two tight-turn structures are predicted near to the membrane within the CD22␣ cytoplasmic tail (Gln 513 -Gln 523 and Asn 528 -Gln 532 ). To determine whether a turn motif is involved in the endocytosis process, truncated CD22␣ mutants were constructed, using PCR-directed mutagenesis, in which a stop codon was introduced at positions Gln 513 (CD22␣⌬Q513), Gly 524 (CD22␣⌬G524), and Glu 527 (CD22␣⌬Q527) to yield mutants with two amino acids in the cytoplasm or just one of the turns (Fig. 1). These mutants were transfected and selected in Jurkat cells (Fig. 2), and the resulting clones were tested in cytotoxicity assays. The results (Fig. 5) showed that with two amino acids remaining in the cytoplasmic tail (CD22␣⌬Q513), saporin toxicity was not enhanced by the presence of CD22specific BsAbs, demonstrating that the surface CD22 molecule was no longer active for delivery. This result was the same as that obtained using the cytoplasmic deleted mutant, CD22␣⌬cyt, even though the expression levels were different (Fig. 2.). This indicated that the motif for internalization lies further to the COOH-terminal end of the molecule. Interestingly, the deletion mutants CD22␣⌬G524 and CD22␣⌬E527, which each contain one predicted turn in the cytoplasm, were fully active in this assay. Thus, in the presence of BsAbs, the IC 50 values for saporin toxicity were 1.5 ϫ 10 Ϫ10 and 5.5 ϫ 10 Ϫ10 M, respectively, with these two transfectants (Fig. 5.), comparable to those seen with the CD22␣ transfectant (Fig. 4) or Daudi cells (20) and indicate quite clearly that the internalization motif of CD22 lies between residues 512 and 523, just under the membrane.
Mutagenesis of the Internalization Motif-The 11-amino acid turn structure motif in CD22 is unusual in that, unlike most other internalization signals described, it contains no tyrosine residues (33). Examination of the sequence (QR-RWKRTQSQQ), however, showed the presence of a serine bearing a hydroxyl group at position 521. It is possible that serine replaces a tyrosine residue in this motif and that the hydroxyl group of both amino acids may play a similar role in the internalization signal. To investigate this, a mutant was constructed in which Ser 521 was mutated to Ala 521 (Fig. 1), and its ability to endocytose was determined in the cytotoxicity assay. Surprisingly, this substitution did not affect delivery of BsAb⅐saporin complexes (Fig. 6), indicating that although serine might have replaced tyrosine in the position of the internalization motif in CD22␣, the mechanism of internalization/ endocytosis of CD22␣ involves interaction with structures other than the side group of a determinant amino acid.
It has been proposed that charged or polar amino acids surrounding the crucial tyrosine residue in an internalization motif may also be important for function (34,35). Of the 11 amino acid residues in the CD22␣ motif, 8 are polar or positively charged (Fig. 7). Such a cluster of polar amino acids clearly suggests that a charged interaction (hydrogen bonding or electrostatic interaction) is likely to be important for internalization. Comparison with tyrosine-containing motifs has shown that amino acids either side of the tyrosine tend to be more prominently polar and also important for function. It was FIG. 5. BsAb⅐saporin cytotoxicity assay on Jurkat cells transfected with CD22␣ tight turn deletion mutants CD22␣⌬Q513, CD22␣⌬G524, and CD22␣⌬E527. The cytotoxicity assays were performed as described in Fig. 4 using anti-CD22-specific (q) and anti-CD38-specific (छ) BsAbs to deliver saporin or with saporin alone (Ⅺ). Jurkat cells were transfected with the three constructs shown. Triplicate samples were assayed for each concentration of saporin investigated, and the mean values are plotted. Two other experiments yielded similar results.
FIG. 6. Cytotoxicity assay for BsAb⅐saporin complexes on Jurkat cells carrying CD22␣ with amino acid substitutions in the newly identified internalization motif. Jurkat cells were transfected and selected for expression of two CD22␣ constructs, CD22␣S521A and CD22␣Q520/522L (Fig. 1). The cytotoxicity assays were performed as described in Fig. 4 using anti-CD22-specific (q) or anti-CD38-specific (छ) BsAbs to deliver saporin or saporin alone (Ⅺ). Triplicate samples were assayed for each concentration of saporin investigated, and the mean values are plotted. Two other experiments yielded similar results. thus decided to make a CD22␣ mutant with amino acid changes on either side of the serine. This mutant, CD22Q520L/Q522L (Fig. 2), had a cytoplasmic tail truncated at amino acid 526 with Gln 520 and Gln 522 mutated to leucine residues. Cytotoxicity assay results with this mutant showed that it had a greatly reduced ability to deliver saporin in the presence of BsAb (Fig. 6). The IC 50 value for saporin toxicity was only 3.3 ϫ 10 Ϫ8 M in the presence BsAb, compared with 2.1 ϫ 10 Ϫ10 M for the CD22␣ transfectants, confirming the previous proposal that polar amino acids are important.

DISCUSSION
CD22 plays a key role in the regulation of B-cell responses to antigen (1,2). One relatively unexplored mechanism of controlling CD22 function and ensuring that it provides accessory signals at the appropriate time is by restricting its cellular localization, i.e. making sure that CD22 is at the right place at the right time. Recently, Sherbina et al. (36) have shown that following stimulation of the B-cell antigen receptor, intracellular CD22 moves rapidly from the cytoplasm to the plasma membrane, resulting in a 50 -100% increase in surface expression within 5 min of stimulation. In addition to membrane delivery, it is clear that CD22 also undergoes rapid and constitutive internalization from the plasma membrane (17). Shan and Press (17) have reported that CD22 is turning over at the cell surface with a t1 ⁄2 of less than 1 h, that internalization occurs regardless of anti-CD22 mAb binding, and that the CD22, once internalized, is transported to lysosomes for degradation rather than being recycled back to the plasma membrane as occurs with many lymphocyte receptors (17). Although it is unclear why CD22 transport is controlled in this way, it obviously has enormous potential for regulating its function.
In the current study, we have investigated the structural basis of CD22␣ internalization using chimeric and mutated constructs expressed in Jurkat T-cells. The results show that the cytoplasmic tail of CD22␣ controls its endocytosis into the cells. CD22␣ molecules without a cytoplasmic tail were not internalized and hence failed to deliver radioactive mAb or BsAb⅐saporin complexes inside the cell. Furthermore, when the cytoplasmic region of a second B-cell marker, CD19, was replaced by that of CD22␣, the resulting chimeric molecule behaved like CD22␣ and internalized efficiently (Fig. 1). Stepwise deletion located the endocytosis motif to an 11-amino acid sequence (QRRWKRTQSQQ) proximal to the plasma membrane: provided this short peptide remained on the truncated CD22␣, it performed perfectly well in delivering toxin BsAb⅐saporin complexes into the transfected cells. A complete list of all the constructs, together with data on deliver BsAb⅐saporin complexes into transfected cells for cell killing, is summarized in Fig. 1. Because the predominant CD22␤ isoform has the same membrane proximal 11-amino acid sequence, it seems reasonable to assume that this motif controls internalization of both CD22␣ and CD22␤ isoforms. Interestingly, the mouse CD22 molecule shows a very similar sequence in this region (37), with just four conserved amino acid differences, and consequently may perform the same function.
Typically plasma membrane receptors are moved into early endosomes via clathrin-coated pits and vesicles (38). Our immunoelectron microscopy data confirm that CD22␣ is strongly clustered around coated-pits. Identified signal sequences for entry into or formation of clathrin-coated pits are degenerate and quite variable, making it difficult to define precise motifs (33,36,38). However, generally they fall in to two classes. The most common are tetra-or hexapeptides containing an aromatic residue, usually tyrosine or phenylalanine (33), placed in the context of one or more amino acids with largely hydrophobic side chains. Typical examples would include FDNPVY for low-density lipoprotein receptor (39), YKYSKV for CD-mannose 6-phosphate receptor (40), and YXRF for the transferrin receptor (31). The second general class of control signals involves adjacent leucine-type residues, which include dileucine or leucine plus a small hydrophobic residue near the carboxyl terminus of the cytoplasmic domain. The current list of receptors using these motifs is quite short and tends to include leukocyte receptors, such as Fc receptor and CD74 (invariant chain of MHC class II) (41). Clearly the CD22 control sequence, QRRWKRTQSQQ, does not fall easily into either group and may represent a novel class, or more likely a modified member of the tyrosine-based signaling motifs. A comprehensive investigation by Ktistakis et al. (34) has defined two salient features required for function of these structures: first, they contain residues that show a preference to "break" regular structure in protein domains; and second, they have a high concentration of polar or positively charged residues on both sides of the tyrosine. A schematic representation of the endocytic tyrosinebased signals is shown in Fig. 7A. This generic internalization signal consists of 8 -10 cytoplasmic amino acids with the tyrosine residue present in the membrane distal portion of the sequence. There appears to be a strong preference for certain types of amino acid at specific positions, suggesting that the orientation of the motif with respect to the membrane is critical and that it is independent of polarity of the protein, and, for both type I and II transmembrane proteins, extends 6 amino acids membrane proximal and 2 residues membrane distal to the tyrosine. The most critical amino acids appear to be located at positions ϩ2, ϩ1, Ϫ1, Ϫ4, and Ϫ6 with respect to the tyrosine. It is thought that the amino acids that disturb regular structure may be required to disrupt the polypeptide chain after it leaves the transmembrane sequence and provide a fairly extended structure immediately proximal to the membrane. Structural predictions on these internalization motifs suggest that the polar groups either side of the tyrosine generate hydrogen-bonded tight turns, membrane proximal to the tyrosine. It is thus predicted that the recognition sequence for the clathrin adapter protein complex-2 is simply a small surface loop (34).
Using the Garnier-Gibrat-Robson prediction algorithm (42), we found that the CD22 motif (Gln 513 -Gln 523 ) showed a preference for a turn structure that was strikingly similar to that for a known tyrosine-based signal (Fig. 7B). It is highly polar and basic and differs from the schematic tyrosine-based motif only in a switch of serine for tyrosine. Mutagenesis studies on tyrosine-based motifs have shown that the tyrosine residue can be substituted with phenylalanine or tryptophan, suggesting that any aromatic residue could suffice for the internalization signaling (43). Surprisingly, in the present study, investigation to determine whether the serine was critical to the CD22 motif showed that it was not, and substituting the serine with alanine did not influence internalization of BsAb⅐saporin complexes in cytotoxicity assays (Fig. 6). However, in contrast, mutating the polar groups either side of the serine completely destroyed function and prevented saporin toxicity, suggesting that polarity was critical for internalization. Numerous questions remain about this newly defined internalization motif and its interaction with the endocytic pathway. However, it appears that internalization is more dependent on charge and secondary structure than on linear sequence, and this is consistent with CD22 associating with the clathrin-associated adapter, clathrin adapter protein complex-2, or an clathrin adapter protein complex-2-like molecule that recognizes the membrane proximal, charged loop for constituent delivery of CD22 into the cell.