This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Nolan, K. F. Articles by Junghans, R. P. Search for Related Content PubMed PubMed Citation Articles by Nolan, K. F. Articles by Junghans, R. P.

Bypassing Immunization: Optimized Design of "Designer T Cells" against Carcinoembryonic Antigen (CEA)-expressing Tumors, and Lack of Suppression by Soluble CEA¹

Biotherapeutics Development Lab, Harvard Institute of Human Genetics, Harvard Medical School, and Division of Hematology-Oncology, Beth Israel Deaconess Medical Center, Boston, Massachusetts 02215 [K. F. N., C.-O. Y., Y. A., J. C. M., E. J. B., R. P. J.]; and Immunomedics, Morris Plains, New Jersey 07950 [S.-O. L.]

	ABSTRACT

Top ABSTRACT Introduction Materials and Methods Results Discussion REFERENCES

 
Tumor-associated antigens are typically nonimmunogenic in cancer patients, "immune surveillance" having manifestly failed. The fact that most tumor antigens are normal human proteins presents significant obstacles to current cancer immunization approaches that researchers are presently striving to overcome. An alternative strategy bypasses immunization altogether by direct genetic alteration of autologous patient T cells, to create "designer T cells" specific to a particular antigen. Chimeric immunoglobulin-T cell receptors (IgTCR) with a specificity for carcinoembryonic antigen (CEA) were created to evaluate the optimal IgTCR structure for cancer therapy. Antigen-binding domains of a humanized antibody were combined with TCR signaling chains to yield four different chimeric IgTCR: single chain Fv fragment (sFv)-, fragment antigen-binding (Fab)-, sFv-, and Fab-. All of the IgTCR were well expressed on T cells, and all showed specific binding and activation, as demonstrated by IL-2 production on contact with immobilized or cellular CEA, excepting sFv- alone which was inert solely against cellular targets for steric reasons unique to this construct. In contrast to prior studies of isolated TCR chains that related increased tyrosine-based activation motifs in as a reason for superior signaling potency, these tests are the first to show that and are indistinguishable for T cell signaling when assayed in the context of the intact TCR complex. Further, Fab was equivalent to sFv as an IgTCR component for expression and antigen binding, establishing an important alternative for IgTCR antigen recognition because sFvs may often lose antigen affinity. When IgTCR was expressed on normal human T cells, cytotoxic potency was demonstrated at low E:T ratios, with T cell recycling and progressive tumor cell destruction. Contrary to recent speculations, these observations prove that high affinity TCR interactions are not an impediment to serial target engagement and disengagement by cytotoxic T cells. The multivalent intercellular interactions of target cell binding, activation, and cytotoxicity were resistant to inhibition by soluble CEA. These studies establish a potentially important new immunotherapeutic modality for the treatment of CEA-expressing tumors.

	Introduction

Top ABSTRACT Introduction Materials and Methods Results Discussion REFERENCES

 
A major focus of modern cancer research has been to explore the potential of immune system components to combat malignant disease. This strategy has exploited, on the one hand, the humoral immune system as represented by antibodies, and, on the other, the cellular immune system as represented by lymphokine-activated killer cells and tumor-infiltrating lymphocytes. In general, antibodies are able to supply specificity but suffer from lack of potency (1) . In contrast, cellular therapies have the requisite potencies, sufficient to induce lethal consequences if not closely regulated, but typically lack specificity for the self-proteins that predominate among tumor-associated antigens (2) . The IgTCR³ approach, demonstrated first by Kuwana et al. (3) in 1987, attempts to combine the specificity of antibodies with the efficient cytotoxic effector properties of T cells, essentially bypassing immunization. Model systems subsequently showed that IgTCR can direct T cells to respond to antigen targets in a specific, MHC-unrestricted manner, generating IL-2 secretion, cellular proliferation, and cytotoxicity, the hallmarks of an effective, self-sustaining immune response (reviewed in Refs. 4 and 5) . It came to be an appropriate juncture at which to extend these studies to a human system of wide clinical relevance, to explore the clinical potential of this technology for cancer therapy.

The TCR is an oligomeric complex composed of six chains in a canonical structure, ()()₂ (6) . The clonotypic, heterodimeric / is responsible for peptide-MHC recognition whereas T cell signaling is conferred by the remaining chains: plus the noncovalently linked, homologous TM proteins , , and that form CD3. Association of all components is necessary for correct assembly, transport, and surface expression of the TCR complex, in which the chain is limiting (7 , 8) . Studies using chimeric molecules have demonstrated that the cytoplasmic tails of all signaling chains of the TCR can independently transduce signals leading to cellular cytotoxicity and/or cytokine production, bypassing the recognition modality of the TCR (9, 10, 11, 12) .

In preparation for direct human application, several features of the design of chimeric IgTCR were identified as subjects for optimization. The concept of IgTCR was established using antibody V genes fused to TCR and constant domains (3 , 13, 14, 15) . Subsequently, other IgTCR joined antibody V regions directly to signaling chains, TCR or the homologous FcR-I chains (see Refs. 4 and 5) . Two basic studies showed that was 300% more potent than , both by IL-2 secretion and cytotoxicity [Refs. 9 and 10 ; recently reconfirmed 16) ], for which reason we considered only native TCR chains for our chimeric receptors. Several studies (9, 10, 11 , 16 , 17) created chimeras with modified or irrelevant TM domains that prevent assembly into the complete TCR (18) . We wished instead to exploit the available multichain TCR apparatus in our constructs and therefore preserved the native TM domains. Additionally, no direct functional comparisons between TCR- and - chains have been reported in the context of an intact TCR complex. This is of particular interest because TCR- has the longest history as the principal target for antibody-mediated T cell activation (e.g., via OKT3; Ref. 19) and for redirected killing via bifunctional antibodies (20) , to which the IgTCR is structurally and functionally analogous.

Further, all recent IgTCR used sFv antibody V gene constructs (see Refs. 4 and 5) . The preparation of an sFv requires its own engineering, expression in soluble form, and Scatchard analysis to ascertain affinity preservation, and then molecular fusion to the TCR chain. Many times this effort fails, and appropriately stable sFvs may not be generated (21) . It was therefore of interest to examine alternatives for integrating antibody binding domains into the IgTCR:Fab constructs, by contrast, maintain the monovalent binding affinity of the original antibody because the native juxtaposition of V_H and V_L is undisturbed, stabilized through C_H1:C_L interactions.

A potential concern for all antibody and IgTCR therapies, regardless of structure, is the presence of high levels of soluble antigen that frequently accompany expression of cancers that may inhibit specific interactions with target cells. It was therefore pertinent to examine whether the multivalent binding between designer T cells and tumor targets would overcome this potential source of interference.

Finally, the early IgTCRs against human tumor antigens all used mouse antibodies (see Refs. 4 and 5) , which could render them immunogenic in vivo, leading to immune destruction of the IgTCR-modified T cells themselves. Clinical experience with humanized antibodies, by contrast, shows that the small murine CDRs that mediate antigen binding, when grafted onto a human immunoglobulin framework, rarely induce immune responses in humans (22, 23, 24) , suggesting an advantage for IgTCR applications as well.

For our first clinical application of this technology, we selected CEA as the target antigen. One of the best characterized of human tumor-associated antigens, CEA is expressed on tumors in 60–94% of patients with metastatic and recurrent colorectal cancer and in 30–60% of metastatic carcinomas of the breast, pancreas, and other organs (25) . In the United States alone, 150,000 patients die per year with CEA-expressing malignancies (25 , 26) . Therefore, a therapy effective against such tumors could have a major impact on the clinical and financial consequences of cancer in this country.

To address these several hypotheses, we used a humanized antibody as sFv or Fab joined via IgG 1 hinge or CD8 hinge domains to CD3 and chains to derive an optimal IgTCR structure for future anticancer designer T cell therapies. Specific gene modifications were implemented to enhance IgTCR stability and surface expression. We show that all combinations of anti-CEA IgTCRs are well expressed on the surface of T cells, they recognize CEA, they bind tumor cell targets, and they efficiently transmit TM signals for T cell activation in an antigen-specific manner. Potent cytotoxicity was demonstrated by chimeric IgTCR when expressed in normal human T cells, with persistent T cell killing and recycling over a period of at least several days. The T cell responses were independent of soluble CEA at concentrations that far exceed in vivo levels in patients. This report represents the completed preclinical optimizations for a new therapy directed against CEA-expressing tumors in humans. Phase I clinical trials are currently under way.

	Materials and Methods

Top ABSTRACT Introduction Materials and Methods Results Discussion REFERENCES

 
Cell Lines.
TCR-positive human T cell line Jurkat and colon cancer cell lines MIP-101 and MIP-CEA clone 8 (Ref. 27 ; gift of Dr. P. Thomas, Beth Israel Deaconess Medical Center) were cultured in RPMI 1640 (BioWhittaker, Walkersville, MD) plus 10% heat-inactivated FCS. The amphotropic retroviral packaging cell lines Bing (28) , Phoenix⁴ (gifts of Dr. W. Pear, MIT), or PA-317 (Ref. 29 ; gift of Dr. A. D. Miller, Fred Hutchinson Cancer Center), were grown in DMEM or RPMI supplemented as above.

Construction of Anti-CEA Chimeric Receptor Genes in Expression Vectors.
Initial gene cloning was performed in p2.1, a vector derived from the pCDLSRSXm plasmid under the T cell tropic promoter SR (Ref. 30 ; gift of Dr. R. Klausner, National Cancer Institute) by removal of the m insert NotI/SmaI, Klenow fill-in, and religated) followed by removal of a BamHI site by Klenow fill-in and religation. The human chain was amplified by reverse transcription-PCR from Jurkat mRNA using primers, forward: 5'-AAAAATCTAGACCTGCTGGATCCCAAAC-3', incorporates XbaI and BamHI sites, and reverse: 5'-AAAAACCTCGAGTTAGCGAGGGGGCAGG-5', incorporates a XhoI site. The chain was amplified from pDJ4 (Ref. 31 ; gift of Dr. C. Terhorst, Beth Israel Deaconess Medical Center), using primers, epsilon forward: 5'-TTTTTTTCTAGAGGGCAACCCGGGAATGAAGAAATG-3', incorporates XbaI and SmaI sites, and reverse: 5'-CCCCCCCTCGAGTCAGATGCGTCTCTGA-3' incorporates a XhoI site. Products were cloned into p2.1 using XbaI/XhoI (p2.1- and p2.1-). A DNA fragment encoding amino acids 114–159 of the CD8 hinge (32) , avoiding Cys-160 and mutating Cys-143 to Ala, was generated by annealing synthetic primers 5'-TCGTTCTAGACTTCCTGAGCGCTAAGCCCACCACGACGCCAGCG-CCGCGACCACCAACACCGGCGCCCACCATCGCGT-CGCAGCCCCTGT-3' (incorporates XbaI and Eco47 III sites) and 5'-TTTGGGATCCAGGGCGAAGTCCAGCCCCCT-CGTGTGCACTGCGCCCCCCGCCGCTGGCCGGCACG-CCTCTGGGCGCAGGGACAGGGGCTG-3' (incorporates a BamHI site), followed by 10 cycles of PCR. The 174-bp fragment was digested with XbaI and BamHI and ligated to XbaI-BamHI-digested p2.1- to obtain p2.1-CD8 hinge-.

To clone DNA encoding the antigen-binding region of hMN14, RNA was extracted from 5 x 10⁷ hMN14/7/g9 transfectoma cells (Immunomedics) with an RNA isolation kit (Stratagene), and used as a template for reverse transcription-PCR. Amplification of variable regions and sFv construction were previously described (33) . The 0.8-kb XbaI-ScaI fragment excised from a pUC19-sFv clone was ligated into the XbaI-SmaI- or XbaI-Eco47III-digested p2.1- or p2.1-CD8 hinge- to generate complete sFv-TCR.

To engineer the Fab, separate plasmids were generated for expression of the L chain and the H chain TCR. For the H-TCR construct, a DNA fragment encoding amino acids 217–230 of the C1 hinge with Cys-226 and Cys-229 mutated to Ala, was first cloned at the 5' end of the TCR chain. Two synthetic DNA fragments, 5'-CTAGAAAAGGATCCCAAATCTTGTGACAA-AACTCACACAGCCCCACCGGCCCCA-3' and 5'-TGGGGC-CGGTGGGGCTGTGTGAGTTTTGTCACAAGATTTGGGAT-CCTTTT-3' BamHI site is underlined), were annealed and ligated into the p2.1- digested with XbaI and Klenow filled BamHI, to obtain p2.1-C1 hinge/. For p2.1-C1 hinge/, the annealed hinge fragment was cloned into the XbaI-SmaI site of p2.1-. DNA fragments encoding V_HC of hMN-14 were cloned by reverse transcription-PCR, V_H forward primer (see above) and C1 reverse primer, 5'-CCCCCCGTTAACTCTCTTGTCCACCTTG-3' which incorporates a HpaI site and cloned into XbaI-SmaI cleaved p2.1-hinge/ or p2.1-hinge/. The resulting SalI-XhoI SR-chimeric TCR cassettes were subcloned to the SalI site of the retroviral vector pBabe puro (Ref. 34 ; gift of Dr. W. Pear, MIT).

The expression vector pBSpGKneo was constructed by subcloning a 1.6-kb XhoI/ClaI Klenow filled-in DNA fragment encoding the pgk promoter and neo^r gene from pEFPGKneo (Ref. 35 ; gift of Dr. S. Orkin) to the EcoRV site of pBlueScript SK+ (Stratagene). The hMN14 V_LC fragment was amplified using V forward (see above) and C reverse primer, 5'-GGGGGGCTCGAGCTAACACTCTCCCCTGTTG-3', incorporating a XhoI site, and cloned into the XbaI-XhoI-digested p2.1 vector (p2.1V_LC). The 1.9-kb SR L chain SalI cassette of p2.1V_LC was blunted using Klenow and cloned into the HincII site of pBSpGKneo (pBSMNLpGKneo). The DNA sequences of all fragments cloned by PCR were confirmed using Sequenase (Amersham, Arlington Heights, IL).

Expression of the Chimeric Genes.
The chimeric receptors were expressed on CD4+ human Jurkat T cells, for binding studies and for activation by IL-2 secretion, and on normal human T cells, which include a CD8+ fraction, for cytotoxicity studies. sFv-/ fusion genes were introduced by retroviral transduction. Vector producer cells were derived from amphotropic packaging cell lines (above): Bing or Phoenix were transfected by CaPO₄ method in the presence of 25 µM chloroquine, whereas PA-317 was infected by ping-pong from 293 cells cotransfected with the ecotropic helper pPEM5 (gift of Dr. A. D. Miller) and vector DNAs. Viral supernatants were applied to recipient cells with 8 µg/ml Polybrene. Retroviral vectors for Jurkat transduction carried the pur^R marker, and were selected in 0.8 µg/ml puromycin. For Fab-TCR expression, pBSMNLpGKneo L chain plasmid was transfected by electroporation using Gene Pulser (Bio-Rad, Richmond, CA) and selected with 2 mg/ml G418 (Life Technologies, Inc., Grand Island, NY). H-/ fusion constructs were then introduced by retroviral transduction and selection of G418-puromycin double resistant clones. For sFv- in normal human T cells, T cells were activated in 20 ng/ml OKT3 and infected for two cycles as above with high titer supernatants of MFG-MPSV or kat vectors without selection markers to yield 30–60% IgTCR+ T cells (36) . Selection of transduced normal T cells for 100% IgTCR+ was by panning against immobilized CEA or WI₂ anti-idiotype antibody.

Identification of Chimeric Receptors by Western Blot.
Membrane fractions of transfected cells were analyzed in reduced or nonreduced condition by SDS-PAGE (in 4–20% polyacrylamide). After electrophoresis, protein was transferred to Immobilon P (Millipore, Bedford, MA) and incubated with the first antibody, either TIA-2 (antihuman chain antibody, gift of Dr. S. Schlossman) or rabbit antihuman antibody (Dako, Glostrup, Denmark), followed by horseradish peroxidase-conjugated second antibody. Detection was by enhanced chemiluminescence (ECL; Amersham) and X-ray film exposure.

Flow Cytometry.
Surface expression of IgTCR was determined using flow cytometric analysis with WI₂ (anti-idiotype antibody of the hMN14) and developed with goat antimouse IgG phycoerythrin (Tago, Burlingame, CA). Normal T cell subsets were assessed by staining with anti-CD4-FITC and anti-CD8-FITC (Tago).

Conjugation Assay.
MIP-101 and MIP-CEA cells as targets were seeded on the chamber slide several days before staining to form colonies. Rhodamine-stained Jurkat cells (1 x 10⁶/ml) were coincubated with hydroethidine-stained target cells as previously (37) and then washed to remove unbound Jurkat cells. Cells were observed under the UV fluorescent microscope. To test the effect of soluble CEA, 10,000 ng/ml CEA were added to Jurkat stained with rhodamine, preincubated for 15 min at 37°C before adding the mixture to the target cells, and coincubated as above. (Initial attempts to duplicate these target binding tests with modified normal human T cells were hampered by the rapid lysis of targets under the assay binding conditions by the CD8+ CTLs in the preparations; however, binding to immobilized antigen was readily demonstrated. This was not pursued further.)

IL-2 Production Assay.
For stimulation of transfectants, microtiter plates were coated with antibodies and purified antigens [OKT3, antihuman CD3 antibody (Ortho); WI₂, anti-idiotypic antibody against hMN-14 (Immunomedics); UPC 10, purified mouse antibody, (Sigma, St. Louis, MO); CEA (Calbiochem, San Diego, CA), and BSA (Sigma)] were coated on microtiter plates at 5 µg/ml in 0.1 M sodium bicarbonate (pH 8.0) and then washed twice with growth medium. The transfected clones were then added (5 x 10⁴/well) in RPMI supplemented with 10% FCS, 10 ng/ml phorbol 12-myristate 13-acetate, and 10 µM -mercaptoethanol. Activation cellular targets was performed at an E:T ratio of 1:1. After 24 h in culture, IL-2 in the supernatant was determined by proliferation bioassay using [³H]thymidine incorporation versus a standard curve on IL-2-dependent HT-2 cells (13) . Maximal stimulations of IL-2 production by OKT3 antibody or ionomycin were in the range of 200–400 IU/ml, which is normal for the number of Jurkat T cells used in these assays.

Cytotoxicity Assay.
Adherent tumor cell targets MIP-CEA (CEA-positive) and MIP-101 (CEA-negative) were plated at 1 x 10⁴ cells/well into 6-well culture dishes that grew to 5 x 10⁵ cells/well after 24 h. IgTCR+ T cells were then added at specific E:T ratios. At harvest intervals, wells were washed to remove unbound T cells, trypsinized, and counted for viable tumor cells by trypan blue exclusion and cell morphology. Cell counts were averaged, and surviving tumor cells plotted versus time. Killing rate per day per effector CTL is derived from the difference in rate constants on the log linear plots of target cell growth curves in the presence and absence of effectors (36) .

	Results

Top ABSTRACT Introduction Materials and Methods Results Discussion REFERENCES

 
Design of Anti-CEA IgTCR.
As part of the optimization plan, different IgTCRs were designed to test hypotheses related to their potential for activity in in vivo therapies. Chimeric receptors were prepared using sFv and Fab antigen-binding domains of humanized anti-CEA antibody hMN-14 joined with both and chains of the TCR (Fig. 1) .

View larger version (22K): [in this window] [in a new window]   Fig. 1. Representation of IgTCR molecules. sFv-, Fab-, sFv-, and Fab- are shown in the context of complete TCR complexes. This schematic is not intended to reflect the actual organization of the complex, which is still unsettled. Adapted from Weiss (8) , with permission.

For the sFv- IgTCR construct, we introduced 46 amino acids of the CD8 hinge between the sFv and the TCR chain to add an extra spacer between the antigen-binding moiety and the membrane surface (sFv-CD8 hinge-). Because has a very small extracellular domain (nine amino acids),⁵ a spacer was previously shown to be critical for effective interaction with some antigens on the surface of target cells for both single-chain Fv and single-chain TCR (32 , 38 , 39) . With the aim to increase surface expression of CD8 hinge-containing molecules, we engineered the hinge to remove the two cysteines that are normally involved in CD8 and CD8 dimerization. We reasoned that free cysteines could interfere with transport of single sFv-hinge-TCR molecules out of the ER to the cell surface (40, 41, 42) , permitting surface expression only for the homodimeric (sFv-)₂ and loss of all (sFv-)() heterodimeric forms, which necessarily have such exposed cysteines. The primary reference on use of the CD8 hinge in IgTCR demonstrated a strong predominance of surviving dimeric over heterodimeric molecules, despite the vast molar excess of the endogenous that should favor heterodimeric forms (43) .⁶ By removing these cysteines, we would remove this constraint, and heterodimeric (sFv-)() forms could be expressed also, thereby permitting a much higher net surface expression, by simple binomial predictions (below).

For the Fab- construct, the CD8 hinge was not inserted because the Fab constant region (C_H1 and C_L) was thought to serve as a natural spacer, analogous to the C and C of the TCR (Fig. 1) . A natural human IgG H chain 1 hinge region (14 amino acids) was inserted to link the C_H1 domain of the Fab H chain to the chain (H-1 hinge-). The 1 hinge provides sufficient spacing between the terminal -sheet hydrophobic domain of the C_H1 and the membrane attachment to allow suitable immunoglobulin-domain protein folding. Like the CD8 hinge, the 1 hinge includes two cysteine residues that are normally involved in inter-H chain disulfide links. As for the CD8 hinge, we modified-to-remove these two free cysteines to increase net expression by permitting surface transport of heterodimeric molecules of (Fab-)() in addition to the homodimeric (Fab-)₂ form; only the one cysteine involved in L chain bonding was retained in the modified 1 hinge.

For the constructs, it was supposed that no extra spacer would be required because the immunoglobulin-like extracellular domain of could serve as a natural spacer (Fig. 1) . Hence, the sFv was directly linked to the chain without the CD8 hinge, with 6 amino acids from the NH₂ terminus directly joined to the sFv COOH terminus as it exits the sheet; this corresponds to the length of the normal linker between the V and C domains of antibody. Fab- used a 1 hinge linker as for Fab- joining, but it now contained two membrane spacer domains: extracellular domain and the Fab constant domain. In contrast to Fab-, there is no opportunity for Fab- dimerization, because forms only noncovalent complexes with TCR- and chains in the complete TCR and does not normally self-associate with other chains (44) . Hence, it was quite possible that free cysteines in the 1 hinge of Fab- might have drastically reduced its surface expression, because no covalently linked Fab-1 hinge homodimers would be expected to form that could transport to the surface. This was not independently assessed.

To express Fab- and Fab- chimeric proteins, genes for H-TCR and L chain were coexpressed from different vectors for the purpose of these experiments ("Materials and Methods"). There is an ER retention sequence in the CH1 domain of the Fab that prevents surface transport of the H-/ constructs unless associated with L chain that masks this sequence (45) . Hence, no H-TCR forms should reach the surface as incomplete Fab lacking L chain.

Finally, the nucleotides surrounding the ATG initiation codons of the cloned V_H and V_L were examined for the presence of Kozak consensus sequences (GCCACCATGG; 46 ), and both were optimal as obtained at the critical -3 and +4 residues, with minor divergence at the 5' end of the motif. In another instance of reconstructing an antiganglioside antibody for IgTCR, nonoptimal contexts were reengineered to optimal.⁷ The designs of all constructs ensured 20 nucleotides between the 7mG cap and the AUG for optimal translational initiation (46) .

Expression of Chimeric Receptors.
The human T helper cell line Jurkat was used as a model system in which to express anti-CEA receptors and test their ability to initiate T cell activation in response to specific antigen. Gene transfer to create IgTCR transductants was performed as in "Materials and Methods." To confirm the expression of each molecule, membrane fractions were analyzed by Western blotting using either anti- or anti- antibody (Fig. 2) . In the reduced condition using anti- antibody, monomers of sFv- chain (including the CD8 hinge), and the H- molecule from the transgene were detected (Lanes 2 and 3) as well as the 16-kDa endogenous chain. On the other hand, only the endogenous chain was detected for sFv/Fab- transformants (Lanes 4 and 5) and for Jurkat cells transfected with empty vector (V, Lane 1). When the same reduced samples were probed with anti- antibody, the sFv- chain, and the H- molecule (of similar mass when L chain is stripped from Fab-) as well as endogenous chain of 23 kDa were detected (Lanes 9 and 10). For sFv/Fab- transformants (Lanes 7 and 8), and the negative control (Lane 6), only the signals corresponding to the endogenous chain were observed.

View larger version (17K): [in this window] [in a new window]   Fig. 2. Membrane-associated IgTCR show expected structures. Membrane fractions of Jurkat transformants were separated on 4–20% SDS-PAGE under reduced (R) or nonreduced (NR) conditions. Proteins were detected with anti- (Lanes 1–5 and 11–15) or anti- (Lanes 6–10) antibody after blotting. Mass scales are in kDa. A, apparent molecular masses of sFv-, H-, sFv-, and H- molecules were 53, 47, 51, and 51 kDa (46, 40, 47, and 45 kDa by calculation). B, homodimer and heterodimer of sFv- were detected as 100- and 65-kDa bands, respectively (Lane 12, 93 and 63 kDa by calculation). Fab- homodimer and heterodimer were observed with an apparent molecular mass of 145 and 89 kDa (131 and 82 kDa by calculation) in Lane 13.

To confirm expression of IgTCR on the surface of the modified T cells, Jurkat transductants were also examined by flow cytometry. In Fig. 3 , cells were stained with anti-idiotypic antibody to the hMN14 anti-CEA antibody, termed WI₂. All IgTCR-modified Jurkat clones showed comparably high expression, although this feature was variable in different transductions. (Specifically, the lower staining of sFv- in Fig. 3 was not seen in other comparisons.) Cells expressing Fab-TCR were stained similarly by both anti-idiotype and anti- L chain antibodies (not shown), indicating that the chimeric molecules retain L chain and form appropriate antigen-binding sites, corroborating the results of Western blotting (Fig. 2 , nonreduced). Cells expressing L chain only are not detected because L chain is secreted to the culture medium in the absence of H-TCR to retain it on the membrane (not shown).

View larger version (24K): [in this window] [in a new window]   Fig. 3. Chimeric IgTCR molecules are expressed at high levels on cell surfaces. Transduced and control nontransduced (·····) Jurkat cells were analyzed by FACS staining with anti-hMN14 idiotypic antibody WI2 as in "Materials and Methods."

IgTCRs Direct Antigen-specific Binding by Modified T Cells.
To trigger signaling, IgTCR-modified T cells must contact antigen in such a way that the receptors are crosslinked or aggregated. Specific binding of IgTCR-modified T cells was confirmed by microscopy with immobilized target antigens (Fig. 4) . Unmodified Jurkat T cells did not adhere to plastic surfaces coated with BSA or CEA (Fig. 4, A and B) , showing only rounded cells resting on the plate bottom. IgTCR-modified cells did not adhere to BSA-coated plastic (not shown, but appearance like A and B) but did bind to CEA, with firm adhesion that led to progressive cellular flattening on the plate surface over time (Fig. 4C) . Identical effects were observed with all four IgTCR Jurkat transformants.

View larger version (87K): [in this window] [in a new window]   Fig. 4. Anti-CEA IgTCR-modified cells bind to immobilized CEA. A, Jurkat V (transduced with empty vector) with BSA; B, Jurkat V with CEA; C, Jurkat Fab- with CEA. IgTCR-modified T cells did not bind to BSA-coated plates (not shown).

View larger version (94K): [in this window] [in a new window]   Fig. 5. Anti-CEA IgTCR-modified cells bind to CEA-expressing tumor cells. Fluorescence microscopy was used to distinguish targets (orange) from Jurkat effectors (green) prepared as in "Materials and Methods." A, Jurkat V with MIP-CEA; B, Jurkat sFv- with MIP-101; C, Jurkat sFv- with MIP-CEA; D, Jurkat sFv- with MIP-CEA in the presence of 10,000 ng/ml soluble CEA.

Serum CEA to 1000 ng/ml or more may occasionally be observed in patients versus the normal level of <5 ng/ml. Incubation in the presence of soluble CEA at 10,000 ng/ml did not interfere with conjugate formation (Fig. 5D) , or with binding to immobilized CEA (not shown). Conjugate formation was also not inhibited in whole patient sera containing 1,000 ng/ml CEA (not shown). This resistance to soluble CEA is attributed to the avidity enhancement of multivalent cell-cell binding versus that of monovalent ligand ("Discussion").

IgTCR Directs T Cell Activation with IL-2 Secretion by Helper T Cells.
Once it was shown that anti-CEA IgTCR-modified T cells can bind specifically to CEA antigen, the activity of chimeric IgTCRs as functional receptor molecules was examined next (Fig. 6) . Perhaps the most used model for T cell activation, human Jurkat CD4+ helper T cells are typically monitored by IL-2 secretion in response to stimuli. As positive controls, cells were treated with ionomycin as a nonspecific activator, and with immobilized OKT3, an anti-CD3 antibody, to cross-link TCR complexes; each provides the maximum known stimulus to activate all T cells, with equal signals for unmodified and IgTCR-modified Jurkat alike. Negative controls were nonspecific antibody (UPC) or BSA (not shown), which show no stimulation on any of the T cells. IL-2 was detected after incubation with immobilized anti-idiotypic antibody (WI₂) and immobilized CEA for all four of the IgTCR+ constructs but not for unmodified Jurkat. When tested instead on the CEA-expressing cell line, MIP-CEA, however, only the sFv-, Fab-, and Fab- transformants produced IL-2, but not the sFv- transformants, paralleling the lack of binding of the latter construct to CEA+ cells (above). With non-CEA-expressing MIP-101 cells, no T cells produced any detectable IL-2, as expected. Similarly, activation-induced secretion of IL-2 was paralleled by characteristic T cell tyrosine phosphorylation patterns on Western blot.⁸ These data confirm that all constructs assemble and are able to recognize and signal T cells on contact with antigen but that specific steric factors associated with CEA expression on tumor cells appear selectively to hinder access of the sFv- construct.

View larger version (55K): [in this window] [in a new window]   Fig. 6. IgTCR transduce CEA-specific activation of T cells. Jurkat CD4+ T cells with different IgTCR modifications were incubated for 24 h with various stimuli, and the supernatants tested for IL-2 ("Materials and Methods"). IL-2 levels were standardized to OKT3-stimulated secretion for each group to control for variation in cell numbers and cell conditioning, and ranged from 200 to 350 IU/ml. Stimulators were: UPC, nonspecific antibody; OKT3, anti-TCR antibody; ionomycin; CEA; -Id, WI2 anti-idiotypic antibody for hMN14; MIP-101, CEA-negative colon cancer cell line; MIP-CEA, CEA-positive colon cancer cell line. The use of no stimulator or BSA gave absent signals that were equivalent to UPC (not shown). SE ±20%.

View larger version (16K): [in this window] [in a new window]   Fig. 7. Soluble CEA does not block IgTCR-mediated T cell activation. IgTCR-modified Jurkat transformants were stimulated against immobilized CEA (A) or MIP-CEA cells (B), in the presence of soluble CEA at varying concentrations, and IL-2 levels were measured and normalized for graphing purposes. Uninhibited IL-2 levels ranged from 300 to 400 IU/ml. sFv- (•, Fab- (), sFv- (). SE ±30%; differences not significant. sFv- transformants had no activity on cells and are not represented in B. Fab- Jurkat cells were not available during these tests.

For an in vitro model of human tumor therapies, IgTCR-modified normal human T cells were incubated with CEA+ tumor cell targets to generate a CD8+ CTL E:T ratio of 0.15:1, under conditions of sparse seeding [15,000 cytotoxic T cells and 100,000 tumor cells/well (960 mm²) of a 6-well plate]), and surviving tumor cells were visually counted at daily intervals. Although more laborious, this cell counting assay was favored over the ⁵¹Cr release assay for its direct assessment of actual viable cell numbers rather than the indirect measure of radioactivity release, and only direct counting is able to assess multilog changes in surviving cell numbers. Moreover, the long time intervals necessary to show T cell recycling and multiple tumor cell killing per T cell are incompatible with the high spontaneous release rates of ⁵¹CrO₄ by these cell lines (not shown).

Although these T cells had been in continuous ex vivo culture for >3 months without TCR restimulation, they still displayed potent CEA-specific antitumor cytotoxicity. The tumor cells in the coculture were progressively and completely eliminated over a period of 5 days; no single tumor cells survived at the end of the assay period, indicating a reversion rate to CEA negative of <1:10⁵ tumor cells (Fig. 8) . CEA-negative cells incubated in parallel with the same effectors expanded 20-fold, yielding a net >6-log difference in surviving cells at the end of the 5-day assay period. Identical expansion of CEA+ tumor cells was seen in the presence of an equal ratio of unmodified, mock-transduced effector T cells (not shown), indicating that the cytotoxicity is specific. T cells remained approximately constant throughout the assay period.

View larger version (15K): [in this window] [in a new window]   Fig. 8. Anti-CEA IgTCR-modified CTLs efficiently lyse CEA+ tumor targets. MIP-101 (CEA-) and MIP-CEA (CEA+) tumor targets were seeded at 100,000 cells/well of a 6-well plate, to which were added 15,000 anti-CEA IgTCR+ CD8+ CTLs. Surviving tumor cells were enumerated daily from separate wells. At 5 days, total CEA—cells were 2 x 106; graph truncated at 250,000 cells for graphing purposes.

	Discussion

Top ABSTRACT Introduction Materials and Methods Results Discussion REFERENCES

 
We have undertaken a chimeric IgTCR approach to target CEA-expressing tumors for human cancer therapy. In the native setting, the clonotypic T cell receptor and chains bind MHC-presented peptide, and the CD3, -, -, and - chains transduce a signal to the nucleus to trigger particular responses in the cognate T cell. By introducing the IgTCR into this apparatus, a new, "designed" specificity is achieved. If we consider that the evolved function of T cells is to destroy our own cells that are infected with virus or other intracellular pathogens, the "designer T cell" approach in essence reeducates those T cells by "fooling" them into "thinking" that the tumor cells are virally infected, thereby leading to tumor rejection and, ideally, tumor cure.

To apply these concepts to clinical studies in humans, it was desirable to determine the optimal structure for the IgTCR. Among the features examined were choice of TCR signaling chain ( versus ), antibody structure (sFv versus Fab), structure of the V_H-V_L linker in the sFv, soluble antigen sensitivity, use of hinges and spacers with modifications to enhance surface expression, confirmation of optimal translation initiation sites, expression under high activity promoters, and use of a humanized version of the antibody to avoid host reactions against the modified autologous T cells in human therapies. Procedures were concurrently developed to generate high titer supernatants for high efficiency retroviral vector transduction without accompanying drug selection, to be reported elsewhere.

Jurkat human T cells were initially used as a model system because they are well characterized with regard to signal transduction, and retain a number of the functional characteristics associated with normal resting human CD4⁺ T helper cells, including the ability to secrete IL-2 upon stimulation of the native T cell receptor (50) . We substituted normal human T cells for study of cytotoxicity because comparably functional human CD8+ immortal CTL lines for long-term study were not available. This use of normal human T cells additionally provided key information on transduction efficiencies to support the translation of these efforts to clinical application.

Finally, an ever important consideration in tumor immunology is the judicious choice of antigen for immunotherapeutic targeting.

Selection of Antigen: CEA.
CEA is a PI-linked glycoprotein of 180–220 kDa with a normal expression that is confined to the microvilli on the luminal surface of certain cells of the large and small intestine (25 , 51) . Because the anti-CEA IgTCR-modified T cells will not detect CEA as MHC class I-presented peptide present on all surfaces of the enterocyte but recognizes only the native three-dimensional surface antigen, the lumenal localization of normal CEA may isolate these normal cells from attack. By contrast, cancers do not have the benefit of such a protected geometry for their CEA expression. In addition, tumor cells typically express quantitatively higher levels of CEA, averaging 35-fold more than normal colonic mucosa (52) , which should further enhance immune discrimination between normal and tumorous expression of the protein (53) . The preferential expression of this antigen in tumor of advanced disease patients may be related to the observation that nonmetastatic human cancer cell lines became aggressively metastatic in xenograft models when transfected with the gene for CEA, suggesting a role for CEA in metastasis (54) .

CEA is a member of a family of related proteins, including nonspecific cross-reactive antigen (NCA), biliary glycoprotein and others, among which anti-CEA antibodies may be cross-reactive, depending on the epitope recognized (51) . From the many (>50) antibodies to CEA presently available, MN14 was selected for these studies. MN14 is a Primus class III antibody; i.e., it reacts exclusively with CEA in the family of CEA-related proteins. We acquired MN14 in a humanized version (hMN14; Ref. 55 ), which provides the further considerable advantage of reducing possible host responses against the IgTCR-modified cells.

Structural Design.
A rationale was offered for the utility of molecular spacers to enhance contact between IgTCR and antigen on the target cells, using natural ( extracellular or C_H1:C_L of Fab) and/or introduced (CD8 and IgG1 1 hinge) domains. We hypothesized that cysteines that normally dimerize CD8 and IgG 1 hinge might, as unpaired residues in monomeric form, hamper transport and surface expression, leading to ER degradation (40 , 41) . Precedent for protein underrepresentation by this phenomenon is seen in osteogenesis imperfecta in which a mutation that substitutes a cysteine in the glycine-X-Y repeat in the ₁ chain yields unpaired cysteines that form abnormal interchain disulfide bonds that are thought to block intracellular transport (42) . With the aim of increasing the net surface expression of IgTCR, we genetically removed these hinge cysteines.

The logic of this approach is demonstrated by simple binomial statistics (1 = a² + 2ab + b²) based on the chain dimerization. For example, if sFv- is 20% of total cellular a = 0.2) and all chains associate freely as it appears, the sFv- is represented in 4% a²) of total TCR as homodimer [(sFv-)₂] but increases to 36% (a²+2ab) of total TCR, or 9-fold, by the possibility of expression as a heterodimer [(sFv-)()] as well. The expression of hinge-containing IgTCRs in heterodimeric format with other TCR chains, and the high net surface expression of IgTCR by Western blot and flow cytometry appear to validate this strategy. Prior sFv-CD8 hinge-TCR constructs using a native CD8 hinge (38 , 43) generally showed lower expressions on the surface of transduced cells, in which it appeared that only dimeric forms reliably achieved Golgi passage and surface expression, with a net preferential underrepresentation of heterodimers; and the heterodimers that were present would be susceptible to intramolecular disulfide bonding and CD8 cyclization (32) that would negate the purpose of the extended hinge.

Whereas low expression of IgTCR has not been a strict barrier to activity of prior constructs against cellular targets with abundant antigen (4 , 5 , 38 , 56 , 57) , it appeared prudent to elect for high IgTCR expression if it was feasible, and this aim was pursued on several levels as outlined above. High IgTCR expression could potentially compensate by "mass action" for low antigen expression on some tumors that might otherwise fail to trigger IgTCR signaling (58) , and high TCR expression has been shown to correlate with improved sensitivity and more intense signaling at all levels of antigen expression (59) . Finally, readily detectable IgTCR greatly facilitates pharmacokinetic monitoring of IgTCR+ T cells during in vivo therapies: without the ability to identify IgTCR by FACS analysis, it is problematic to identify IgTCR-modified cells among the endogenous, unmodified T cells in patients’ blood.

Anti-CEA sFv-TCR and Fab-TCR Are Equivalently Functional.
For this study, we created both sFv and Fab versions of the humanized antibody, hMN14. Our anti-CEA sFv retains binding affinity equivalent to Fab (5 x 10⁷versus 6 x 10⁷M^-1) against cellular CEA and is significantly stable, as demonstrated by prolonged (24-h) incubation at 37°C in the form of a soluble fusion protein (33) . Anti-idiotype antibody, purified immobilized CEA, and CEA on tumor cell targets all triggered stimulation of sFv-TCR-expressing T cells. The failure of sFv- to activate solely against cellular antigen is considered to be related to steric factors hindering IgTCR access to the epitope of CEA on the surface of cells; this is the shortest of all four constructs by our molecular structural analysis.⁹ This hindrance of sFv- for cellular CEA may not generalize to other surface antigens with more exposed epitopes and would likely be ameliorated by an additional IgTCR spacer in the present instance, inasmuch as Fab-, with its one additional spacer domain (Fig. 1) , is fully active.

Fab-TCRs were expressed as well as or better than the sFv-TCR on the surface of modified T cells, and were similarly potent in activation assays, and, in contrast to sFv-TCRs, both Fab-TCRs activated against cellular and immobilized CEA alike. The principal practical disadvantage of Fab constructs is that H chains and L chains must be coexpressed in the transduced cells to form a functional antigen-binding domain. Yet this is not a serious obstacle because the L- and H-TCR chains can be expressed from single mRNA transcripts by incorporating an intervening internal ribosome entry site that yields efficient expression of both genes under a single promoter (60) .

These studies establish the utility of Fab-TCR constructs. This may be an important alternative strategy, particularly when an antibody sFv proves to be unstable or loses affinity for antigen. Accordingly, one could reasonably bypass the preparation and testing of sFv and go directly to Fab-TCR with confidence.

and Chains Are Equipotent for Signaling in the Whole TCR Complex.
We, as others (17 , 38) , confined our IgTCR studies to chains of the TCR instead of the commonly used FcR-I chain (see Refs. 4 and 5 ). TCR- is more potent for signaling than the homologous FcR chain (9 , 10 , 16) , and, although can also assemble into TCR complexes on T cells, it is much less efficient at doing so (61) and is associated with T cell immune dysfunction when substituting for (62) .¹¹ In our comparison of TCR chains, all IgTCR and IgTCR were potently activated when stimulated with immobilized antigen, and Fab- was as potent as sFv- and Fab- against cellular CEA (Fig. 6) . The selective inactivity of sFv- against cellular antigen but not against immobilized antigen was considered a consequence of steric factors in the epitope exposure in the cellular expression of the antigen rather than a consequence of activity in the chain itself (above).

Published studies using isolated TCR chains i.e., not assembling into the TCR) fused to irrelevant extracellular domains, showed that was weaker than with respect to both IL-2 production and T cell proliferation and that both were weaker than engagement of the complete TCR (12 , 65) . This was explained by the number of ITAMs in the complete TCR (ten) which is greater than (three) which is greater than (one) (65, 66, 67) . Similarly, reduction or amplification of the number of ITAMs in had corresponding effects on activation potency of the chain (68 , 69) . Accordingly, we left undisturbed the assembly motifs of the and chains (18) in our chimeras to promote integration into the intact TCR complex. Engagement of antigen via the chain in our studies was as effective as engagement via the chain for the stimulation of the T cell, irrespective of the number of ITAMs in the individual chimeric chains. In either instance, the stimulation will be transmitted to the whole assembly, 10 ITAMs per TCR complex. Inasmuch as TCR- and TCR- have no ITAMs yet are the normal elements to mediate T cell recognition and activation, this conclusion of equivalence of different chains in the context of the complete TCR is not entirely surprising. The biological relevance of the intact TCR was recently shown by subtle deficiencies of T cell signaling in transgenic animals with single or constructs engineered to express on cell surfaces but not to assemble with other TCR chains (65) . Accordingly, prior studies of isolated TCR chains would seem to provide no compelling basis for preferring over other TCR chains for IgTCR design: the complete TCR context, in which all IgTCR are expressed, equalizes all.

Gene-modified T Cells Mediate Potent Killing against CEA-expressing Targets.
In vitro reconstructions of tumor therapies with an excess of rapidly growing tumor cells showed that the IgTCR+ T cells are able to bind, kill, release, and kill again repeatedly over an extended period. The cells used for this test were applied after 3 months of continuous culture, without restimulation, showing persistence of IgTCR expression, cell viability, and cytotoxic potency. Even gene-modified normal T cells when rested without IL-2 maintained cytotoxic potency for prolonged periods of time (36) .

It was recently speculated that the high affinity of typical antibodies in IgTCRs could be a hindrance to the activity of the gene-modified T cells (4) . The affinity of our constructs were 50- to 500-fold above that of native TCR for MHC-peptide, yet maintained the capability of serial targeting of tumor cells, thus apparently voiding this concern. In any case, it has been concluded that most of the cell:cell binding energy is supplied by adhesion molecules rather than by the TCR interactions (70) , and the additional energy of IgTCR interactions is apparently not an impediment to serial engagement and disengagement of targets by single T cells. Furthermore, high affinity TCR interactions have been shown to reduce the numeric threshold for antigen expression to attract T cell killing (58) , and thus high affinity may in fact confer an advantage to offset any relative down-regulation of antigen expression by tumor cells during therapy. Correspondingly, we noted an escape rate of <1:10⁵ tumor cell targets (Fig. 8) despite heterogeneous surface expression of CEA (not shown).

The activation and cytotoxicity of these designer T cells occur in the absence of a source of costimulation on the tumor targets. Costimulation of the CD28 T cell molecule by B7 on antigen-presenting cells is regarded as important mainly for the education of naive T cells during primary immune responses, but not so for subsequent memory or effector T cell responses (71 , 72) . Our procedure for modifying patient T cells ("Materials and Methods") begins with activation that provides both TCR stimulation (by OKT3) and CD28 costimulation (by B7 on autologous monocytes), promoting even naive T cells to become active effector cells. While our initial patient studies will examine designer T cell efficacy without further modification, there is evidence nevertheless for quantitative improvements in activation thresholds (73) and in antitumor efficacy with CD28 costimulation of mature effector cells (74) that may be profitably examined in later clinical tests.

Soluble CEA Does Not Inhibit Binding and Activating of Chimeric Receptor.
We reasoned that the polyvalent reaction of cellular CEA with IgTCR on effector cells would be resistant to competition by the necessarily monovalent reaction with soluble CEA, in accord with the well-known affinity enhancement available to multivalent cell-cell interactions (75, 76, 77) . In tests with an anti-CEA immunotoxin with the same hMN14-sFv, cytotoxicity was inhibited with 5,000 ng/ml (25 nM) but not with 1,000 ng/ml (5 nM) soluble CEA (33) . In the context of the multivalent binding interactions between modified T cells and targets, the inhibitory capabilities of soluble antigen are further blunted, in which the presence of even 10,000 ng/ml (50 nM) of soluble CEA does not block designer T cells for purposes of binding, activating IL-2 secretion or directing cytotoxicity against tumor targets. Soluble CEA in patient sera only infrequently exceeds 1,000 ng/ml (5 nM), and thus would be unlikely to be a factor in inhibiting anti-CEA designer T cells in in vivo therapies.

We have pursued a strategy to optimize the design of chimeric IgTCR for use in treatment of a wide range of CEA-expressing adenocarcinomas. We have shown that the sFv and Fab antibody forms are equally effective in engaging target antigen on tumor cells and that the and TCR signaling chains are equivalent when activated in the context of the TCR under the functional tests of this study. With all four constructs in hand, the hMN14 sFv-CD8 hinge-TCR was selected based on its single-chain structure as the construct for CEA-expressing cancer therapy. However, the effort to create and test the sFv and modified CD8 hinge made this the most laborious of all to engineer, and Fab- or Fab- might have been preferred for de novo preparations. Other studies not shown indicate that CD4+ helper and CD8+ cytotoxic normal human T cells are gene modified by retroviral transduction with equal efficiency and equal levels of IgTCR expression (36) . With the optimization of IgTCR structure for expression and function, demonstrating IL-2 secretion and cytotoxicity, this opens the potential for a self-sustaining autogenic response against CEA-expressing tumors as a new therapeutic modality against this major class of cancers.

	ACKNOWLEDGMENTS

 
First and foremost, we acknowledge the vision and generosity of Dr. Glenn D. Steele, Jr., who, as Chief of Surgery at the former New England Deaconess Hospital, afforded us the initial resources that made possible the early phases of this work. We thank Drs. Robert Sharkey, David Goldenberg, and Hans Hansen of Immunomedics and Garden State Cancer Center for providing hMN14 and its anti-idiotypic antibody, WI_2; Dr. Cornelius P. Terhorst of Beth Israel Deaconess Medical Center for cloned human CD3 chain gene and for helpful discussions; Dr. Patrick Hwu (Surgery Branch, National Cancer Institute) and Dr. Zelig Eshhar (Weitzmann Institute) for discussions and expert advice in the application of IgTCR; Dr. Warren Pear (MIT), Dr. Mitch Finer (Cell Genesys, Inc.), and Dr. Richard Mulligan (Children’s Hospital, Boston, MA) for generously providing vectors, cell lines, and expertise in retroviral gene transduction; Gang Zheng for performance of the cell conjugation assays; and Jennifer Watters for constructing the modified CD8 hinge. We also thank Shilpa Mhatre, Kristy Johnson, Kelli Chapin, Shiara Ortiz-Pujols, and Daniel Hagg for technical assistance.

	FOOTNOTES

 
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.

¹ This work was supported by research grants to R. P. J. from the Surgery Department of the former New England Deaconess Hospital, from the Skin Cancer Foundation, from the American Cancer Society, and from the National Cancer Institute, and by a Clinical Oncology Career Development Award from the American Cancer Society and an National Cancer Institute Research Career Development Award (K04) to R. P. J.

² To whom requests for reprints should be addressed, at HIM 403, Beth Israel Deaconess Medical Center, Boston, MA 02215. Phone: (617) 432-7004; Fax: (617) 432-7007; E-mail: junghans{at}hms.harvard.edu

³ The abbreviations used are: IgTCR, chimeric immunoglobulin-T cell receptor; sFv, single chain Fv fragment of antibody; Fab, fragment antigen-binding of antibody; H-TCR, heavy chain of Fab joined to TCR chain; CDR, complementarity-determining region; Jurkat V, Jurkat cells transduced with empty vector; BSA, bovine serum albumin; ER, endoplasmic reticulum; TM, transmembrane; CEA, carcinoembryonic antigen; L chain, light chain; H chain, heavy chain; ITAM, immunoreceptor tyrosine-based activation motif.

⁴ G. N. Nolan, unpublished observations.

⁵ Further truncated to three amino acids by our engineering strategy ("Materials and Methods").

⁶ A further article greatly underrepresents the endogenous by a detection method that labels according to the size of the extracellular domain (38) .

⁷ C.-O. Yun and R. P. Junghans, manuscript in preparation.

⁸ Y. Akamatsu and R. P. Junghans, data not shown.

⁹ K. Johnson and R. P. Junghans, unpublished calculations.

¹¹ Early IgTCR suggesting equivalence or superiority of (63 , 64) may have reflected instead the now evident stability problems of sFv with in the absence of a spacer hinge (42) .

Received 6/18/99; revised 9/ 8/99; accepted 9/21/99.