Deficiency of Transporter for Antigen Presentation (TAP) in Tumor Cells Allows Evasion of Immune Surveillance and Increases Tumorigenesis¹

Tumor cells express peptide/class I MHC (MHC-I)³ complexes containing peptides derived from tumor-specific Ags. This allows recognition and killing of tumor cells by CD8 T cells. CTL that can recognize and kill tumors in vitro have been isolated from patients (1, 2, 3). Despite this potential for immune detection and destruction of tumor cells, however, the immune system rarely serves as a significant barrier against tumorigenicity in vivo (4).

Multiple mechanisms may contribute to the ability of tumor cells to evade antitumor immunity. Some tumors express cytokines or surface receptors, e.g., FAS-ligand (5, 6), that interfere with T cell responses. In other cases, expression of tumor Ags may be limited. Tumor cells that lack expression of costimulator molecules may be less immunogenic (7, 8). Finally, tumor cells may evade T cell responses by decreasing expression of peptide-MHC-I complexes. This could be accomplished by decreased synthesis and expression of MHC-I, as observed in some human tumors (9, 10), or by decreased processing of tumor Ags and production of peptide-MHC-I complexes, as addressed in this study. Deficiencies in components of the MHC-I Ag processing pathway have been shown in a variety of human tumor tissues and human tumor cell lines (11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31), and some studies have correlated the presence of these deficiencies with tumor progression (13, 16, 18, 19, 25, 27). These studies, however, are only correlative, and it is still unclear whether deficits in Ag processing components have any consequence for tumor progression. A well-controlled in vivo study is still needed to determine whether alterations of Ag processing confer a survival advantage upon tumor cells and increase tumorigenesis.

In the MHC-I Ag processing pathway, peptides are generated from cytosolic proteins by proteasomes and then transported into the lumen of the endoplasmic reticulum via the TAP, an ATP-dependent peptide transporter that is composed of two subunits, TAP1 and TAP2. Deficiency of either TAP1 or TAP2 blocks TAP function. TAP-deficient cells fail to transport cytosolic peptides into the endoplasmic reticulum, reducing the supply and repertoire of peptides available for binding to MHC-I. The decrease in peptide binding reduces stability and surface expression of MHC-I on TAP-deficient cells.

One hypothesis is that deficiency of TAP1 expression by tumor cells should reduce MHC-I processing and presentation of cytosolic tumor Ags, decrease susceptibility of tumor cells to antitumor immunity mediated by CD8 T cells, and increase the tumorigenic capacity of these cells. T cells, however, are not the only cells of the immune system that recognize MHC-I. Recognition of MHC-I by NK cells decreases killing of target cells, and reduced expression of MHC-I increases killing of targets by NK cells. Thus, an alternative hypothesis states that deficiency of TAP1 expression by tumor cells should decrease MHC-I expression, making the tumor cells more susceptible to killing by NK cells and less tumorigenic. In fact, a previous in vivo study showed that deficiency of TAP in a lymphoma cell line increased susceptibility to NK cells and decreased tumorigenicity of these cells (32). It is unlikely, however, that TAP deficiency prevents growth of all tumors, because this deficiency occurs frequently in at least some kinds of solid tumors. Deficits in positive regulators of NK cytotoxicity or the presence of other unknown negative regulators may limit susceptibility to attack by NK cells for some tumors with low MHC-I expression.

The studies presented here address the impact of TAP deficiency on tumor progression in a novel tumor model system. Matched sets of TAP1-positive and TAP1-negative cell lines were derived from a common parental transformed murine fibroblast cell line. When mice were inoculated with these cells, TAP1-positive cells were rejected by T cell-dependent mechanisms, whereas TAP1-negative cells survived and produced tumors. These studies provide the first data from well-controlled in vivo tumorigenesis experiments to demonstrate that loss of TAP1 function allows some tumors to avoid immune recognition and T cell-dependent elimination without an overriding increase in susceptibility to NK cells. Thus, deficiency in expression of TAP1 by tumor cells results in increased tumorigenicity.

Six-week-old C57BL/6 females were obtained from The Jackson Laboratory (Bar Harbor, ME). Six-week-old HSD/athymic nude-nu/nu females were bred at Case Western Reserve University. Cells were maintained in standard medium composed of DMEM with 10% FCS and antibiotics. Transfected cells were maintained in standard medium plus 1 μg/ml puromycin (Sigma, St. Louis, MO) or 1 μg/ml puromycin and 400 μg/ml hygromycin B (Life Technologies, Rockville, MD). To isolate mouse embryonic fibroblasts (MEF), embryos (<18 days old) were removed from a pregnant TAP1 knockout mouse backcrossed onto C57BL/6 background (generous gift from Dr. Luc Van Kaer, Vanderbilt University, Nashville, TN) (33). The head, limbs, and liver were removed from each embryo, and remaining tissue was minced with razor blades and incubated for 15 min at 37°C in 0.05% trypsin/0.53 mM EDTA. Material from the embryos was pooled, passed through a strainer, pelleted, plated in 100-mm tissue culture plates in standard medium, grown for 8 days to achieve confluent cultures, and split 1 day before transfection to achieve cultures that were 30% confluent.

MEF were transformed by cotransfection with SV40 large T Ag expressed from the SvlacT plasmid (34), ras expressed from the Ras-Zip6 plasmid (35), and puromycin resistance gene encoded by the pBABEpuro plasmid (36). The TAP1 gene was cut from pcDNA1neoHAM1 (generous gift from John Monaco, University of Cincinnati, Cincinnati, OH) by a double digestion with XbaI (New England Biolabs, Beverly, MA) and HindIII (Boehringer Mannheim, Indianapolis, IN). The fragment containing the TAP1 gene was isolated by gel electrophoresis with 1% low melt agarose, cut from the gel and ligated in the solid gel phase into the pcdna3.1hygro vector (Invitrogen, San Diego, CA), which was prepared by digestion with XbaI and HindIII and purified on an agarose gel (37). The ligated plasmid, pcdna3.1hygro-TAP1, was introduced into competent Escherichia coli by electroporation. Plasmids were purified from bacteria using a Qiagen Maxi-Prep Kit (Qiagen, Santa Clarita, CA). MEF and transformed fibroblast cell lines were transfected by the calcium phosphate precipitation method (38). Five hundred microliters of filter-sterilized 2× HEPES-buffered saline (4.09 g NaCl, 2.97 g HEPES, and 0.19 g Na₂HPO₄ in 250 ml H₂O, pH 7.1) was bubbled with a 1-ml pipette. Plasmid DNA in 250 mM filter-sterilized CaCl₂ was added dropwise. The mixture was vortexed for 5 s and incubated at room temperature for 20 min. Calcium phosphate-DNA precipitates were dripped over cells in 100-mm plates containing 9 ml standard medium. The cells were cultured for 4–8 h, washed, cultured for 48–60 h, and split into standard medium containing either 1 μg/ml puromycin alone or 400 μg/ml hygromycin plus 1 μg/ml puromycin. After 2–3 wk, isolated colonies were removed, expanded, and frozen in liquid nitrogen. A large stock of vials was generated in a single freeze to provide a consistent source of cells for subsequent experiments.

To prepare cells for PCR of genomic DNA, confluent cultures from T25 flasks were harvested with 0.05% trypsin/0.53 mM EDTA, washed three times with ice cold PBS and resuspended in 0.5 ml autoclaved digestion buffer (50 mM Tris (pH 8.0), 10 mM EDTA (pH 8.0), 100 mM NaCl, and 1% Triton X-100) with 0.2 mg/ml proteinase K (Life Technologies). The samples were digested overnight at 55°C, boiled for 5 min, and diluted 1:5 in water to generate the DNA samples. The PCR reaction contained 2.5 μl DNA sample, 0.75 μl 50 mM MgCl₂, 0.05 μl of each 100 mM dNTP, 0.25 μl Taq DNA polymerase (Life Technologies), 0.5 μl of each 50 μM primer, 2.5 μl 10× PCR buffer (Life Technologies), and water to achieve a final volume of 25 μl. Primers for TAP1 were GGACTGTCAGCAGCGGCAACC and CAAGGCCTTTCATGTTTGAGGG. Primers for TAP2 were CAGGATGCAGTGGCCAGGGCG and TAGATACACGTCTTTTTCCAGG. PCR samples were incubated in a PTC-100 Thermal Controller (MJ Research, Watertown, MA) at 94°C for 1 min, followed by 25 of the following cycles: 30 s at 94°C, 30 s at 60°C, and 1.5 min at 72°C. The samples were then incubated at 72°C for 5 min. PCR products were electrophoresed at 90 V in 1.5% agarose gels with 0.5× Tris-borate-EDTA buffer solution (10× TBE: 108 g Tris base, 55 g boric acid, and 40 ml 0.5 M EDTA (pH 8.0), in 1 liter of water).

Adherent cells were removed with Versene (Life Technologies) or 0.05% trypsin/0.53 mM EDTA, which give similar results for MHC-I staining. Cells were pelleted, washed in flow cytometry buffer (PBS/1% rabbit serum/0.1% BSA), plated in 96-well round-bottom plates at 2.5 × 10⁵ cells/well, incubated for 30 min at 4°C with Y-3 anti-K^b (American Type Culture Collection, Manassas, VA) or isotype control Ab (mouse IgG2b; Caltag Laboratories, Burlingame, CA) at 5 μg/ml in flow cytometry buffer, washed three times, incubated for 30 min at 4°C with a 1:50 final dilution of FITC-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA), washed three times, and analyzed on a FACScan flow cytometer (Becton Dickinson, Franklin Lakes, NJ). MHC-I on explanted tumors was analyzed on a Coulter Epics flow cytometer (Coulter, Miami, FL).

For each experiment a fresh aliquot of cells was thawed, cultured, and screened by flow cytometry for MHC-I expression to confirm TAP1-positive or TAP1-negative phenotype. Confluent cells were harvested using 0.05% trypsin/0.53 mM EDTA, pelleted in standard medium, washed three times with DMEM, counted using trypan blue to exclude dead cells, and adjusted to a concentration of 5 × 10⁷ cells/ml in DMEM. Cells (0.1 ml) were inoculated s.c. in the flank of each 6- to 8-wk-old mouse. Mice were followed for 3–5 wk, and orthogonal diameters of the tumors were measured externally using calipers. The presence or absence of tumors was verified by postmortem dissection and histopathologic examination.

Tumors were removed from each mouse, minced, and digested for 30 min at 37°C with 1 mg/ml collagenase I, 0.2 mg/ml trypsin, and 0.5 mg/ml DNase I in DMEM (all enzymes were from Worthington Biochemical, Lakewood, NJ). The resulting mixture was triturated and passed over a cell strainer to remove large pieces of debris. Cells were pelleted, loaded onto step gradients with three layers composed of 70% Percoll, 30% Percoll, and PBS, and centrifuged for 1 h at 1300 × g at 25°C. Cells were removed from the interface of the 70% and 30% Percoll layers, washed with DMEM, and cultured overnight in standard medium. The transfected tumor cells were then isolated by culture for 7 days with selection media containing 400 μg/ml hygromycin and 1 μg/ml puromycin to deplete other host-derived cells. Cells were then analyzed by RT-PCR for TAP1 and TAP2 or by flow cytometry for MHC-I expression. To simplify the distinction between MHC-I expression on TAP1-positive and TAP1-negative cells, MHC-I expression was enhanced by stimulation of the cells with 100 U/ml recombinant murine IFN-γ (Genzyme, Cambridge, MA) for 2 days before flow cytometry analysis.

Tumors were removed and cells were selected in vitro as described above. Three different cultures were derived from different mice in the same experimental group, and 9 × 10⁵ cells from each culture were pooled. RNA was prepared using the Trizol protocol (Life Technologies). The pooled cells were lysed in 0.5 ml Trizol, the samples were incubated at room temperature for 5 min, and 0.1 ml of chloroform was then shaken with the homogenate. After 2 min the phases were separated by centrifugation. The upper aqueous phase was then mixed with an equal volume of isopropyl alcohol and incubated for 10 min at room temperature. RNA was pelleted by centrifugation and washed once with 75% ethanol. The pellet was resuspended in diethyl pyrocarbonate-treated water, and OD at 260 nm was used to determine RNA concentration. For reverse transcription (RT), 2 μg of RNA was combined with 5 μl of 50 μM random primers (Life Technologies), 5 μl of 5× RT first strand buffer (Life Technologies), and 11.5 μl of diethyl pyrocarbonate-treated water. This mixture was incubated at 70°C for 5 min and then combined with 5 μl of 200 U/μl of Moloney murine leukemia virus RT (Life Technologies), 5 μl of 0.1 M DTT, 0.25 μl of each 100 mM dNTP (Life Technologies), and 1.5 μl of 33 U/μl RNasin RNase inhibitor (Promega, Madison, WI). The RT mixture was incubated at room temperature for 10 min, 42°C for 50 min, and 99°C for 5 min. The resulting cDNA was amplified by PCR as described above with 28 amplification cycles.

A novel murine tumor model was created to compare the tumorigenicity of matched sets of TAP1-positive and TAP1-negative tumor cell lines. MEF were obtained from TAP1-knockout mouse embryos. These cells were transformed in vitro with SV40 large T Ag and ras, cloned, and tested for tumorigenicity in mice. Some clones were tumorigenic only in athymic mice, whereas others, such as MEF1TAP1–27, formed tumors in both athymic mice and normal C57BL/6 mice. Clones were transfected with an expression vector containing both TAP1 and the hygromycin B resistance gene, or a control vector containing only the hygromycin B resistance gene. After selection with hygromycin B, numerous subclones were isolated and screened for the presence of TAP1 by genomic PCR (Fig. 1 A). PCR primers to the TAP1 gene produced a specific product only in those subclones transfected with TAP1. None of the subclones tested had deficiencies in RNA expression for TAP2, low molecular mass polypeptide (LMP2), or LMP7 by RT-PCR (data not shown).

MHC-I molecules are stabilized on the cell surface by their association with peptide. Because TAP1-negative cells do not efficiently load peptide onto MHC-I, a large proportion of MHC-I molecules in these cells is unstable, and MHC-I expression is decreased. Thus, the presence of functional TAP1 correlates with increased cell surface MHC-I. Accordingly, MHC-I expression was measured on cell lines by flow cytometry to determine whether functional TAP1 protein was expressed (Fig. 1 shows labeling for K^b, and analysis of D^b expression gave similar results). Six TAP1-transfected subclones were found to have increased MHC-I expression (Fig. 1), demonstrating functional expression of TAP1. Two of these (27.V+I-10 and 27.V+I-24) were chosen for additional studies. In contrast, flow cytometry analysis of six subclones that were transfected with the control vector showed that none had significant expression of MHC-I (Fig. 1). Two of these (27.V-26 and 27.V-29) were randomly selected for additional studies.

Transfection with TAP1 increased the expression of MHC-I molecules, although MHC-I levels were still lower on the TAP1-positive fibroblasts than some other cell types (e.g., RMA cells). This apparently reflects low natural synthesis and expression of MHC-I molecules by these cells. To exclude the possibility of persisting TAP-related constraints in TAP1-transfected cells, the cells were cultured at 26°C in medium containing the SIINFEKL peptide that binds to K^b, but K^b levels were increased only modestly and remained much lower than those found on RMA cells (data not shown). Thus, K^b expression remained low even under conditions that should bypass TAP-related restraints, indicating that the level of TAP function was not a major limitation for MHC-I expression on TAP1-transfected cells. This suggests that low constitutive cell surface expression of K^b reflected low constitutive synthesis of MHC-I molecules in these cells.

Tumorigenicity of the TAP1-positive and TAP1-negative cell lines was studied by injection of the cells s.c. in the flank of 6- to 8-wk-old C57BL/6 mice. The mice were monitored for 3–5 wk. Tumors occurred at the site of injection only and as a single mass. Orthogonal diameters of the tumor were measured externally in each mouse at regular intervals, and all mice were ultimately dissected to verify the presence or absence of tumors. No metastatic tumors were found with this protocol. As shown in Fig. 2 and Table I, none of 47 mice injected with the cell lines expressing functional TAP1 (27.V+I-10 and 27.V+I-24) developed long-lasting tumors, whereas 44 of the 47 mice injected with TAP1-negative cell lines (27.V-26 and 27.V-29) developed long-lasting tumors. Some animals injected with TAP1-positive cells did show transient tumors that were grossly evident (detected by external measurement from approximately day 6 to day 9), but these tumors regressed spontaneously and were absent after day 12. The kinetics with which these tumors regressed is consistent with the development of an antitumor T cell response. Some mice that were injected with TAP1-positive cells were followed for as long as 150 days without subsequent evidence of tumor.

Autopsy and histopathologic examination showed a spindle cell sarcoma at the site of injection in mice that received TAP1-negative cells (Fig. 2 C). No metastases were identified. In contrast, mice that received TAP1-positive cells showed only a residual chronic inflammatory infiltrate and no remaining tumor at the site of injection after 3–4 wk. In other experiments, autopsy and histologic examination was performed at earlier time points. Tumors were present from day 4 to day 6 in animals that were injected with either TAP1-positive or TAP1-negative cells, but TAP1-positive tumors were regressing by day 9 and were completely eliminated by day 12, whereas TAP1-negative tumors persisted and continued to grow.

The difference in growth of the TAP1-positive and TAP1-negative tumors could not be explained by different inherent growth rates of the cell lines. All of the cell lines showed similar rates of proliferation in vitro, as assessed by uptake and incorporation of tritiated thymidine, metabolism of an indicator dye (Alamar blue), and cell proliferation manifested by cell counts (data not shown). In addition, all of the cell lines, both TAP1-positive and TAP1-negative, grew progressively in vivo to produce tumors in 100% of inoculated athymic mice, which lack functional T cells (Table I).

Although the tumor studies described above involved injection of cells with either homogeneous expression of TAP1 or lack thereof, TAP-deficient tumor cells must arise in vivo in the vicinity of TAP-positive tumor cells. It is possible that immune responses to TAP-positive tumor cells could activate other cells or effector mechanisms, e.g., NK cells, which could attack TAP-negative tumor cells, circumventing potential immune evasion. To examine these issues, mice were inoculated with one of several different cell preparations consisting of 5 × 10⁶ TAP1-negative cells, 5 × 10⁶ TAP1-positive cells, or a mixture of 2.5 × 10⁶ TAP1-positive cells and 2.5 × 10⁶ TAP1-negative cells. As noted above, the TAP1-negative and TAP1-positive cell lines had similar rates of growth in vitro. Of the animals that were injected with the mixture of TAP1-positive and TAP1-negative cells, tumors grew in 23 of 25 C57BL/6 mice and 24 of 24 athymic mice (Table I). Thus, tumors resulted when TAP1-negative cells were included in the inoculum, regardless of the presence or absence of accompanying TAP1-positive cells.

The development of tumors from inocula containing a mixture of TAP1-positive and TAP1-negative cells suggested preferential growth and selection of TAP1-negative cells in C57BL/6 mice. To test this hypothesis, expression of TAP1 was assessed in cells derived from these tumors. Tumors were removed from animals, and cells were cultured for 7 days in selective medium to eliminate nontumor host cells and allow the selective outgrowth of TAP1-positive and/or TAP1-negative tumor cells. TAP1 RNA expression was then analyzed by RT-PCR. TAP1 RNA was detected in cells from tumors in athymic mice that received TAP1-positive cells or a mixture of TAP1-positive and TAP1-negative cells (Fig. 3). In contrast, TAP1 RNA was not detected in cells from tumors in C57BL/6 mice that received the same mixture of TAP1-positive and TAP1-negative cells (Fig. 3). Thus, TAP1-positive cells were selectively eliminated following inoculation of immunocompetent mice with a mixture of TAP1-positive and TAP1-negative cells. TAP1 was not detected in cells from tumors in athymic mice or C57BL/6 mice that received TAP1-negative cells, demonstrating that the protocol eliminated any host-derived contaminating TAP1-positive cells. TAP2 was detected in all tumor cell cultures (Fig. 3), demonstrating that all RNA preparations contained intact mRNA that could be amplified. As expected, TAP1-positive cells alone did not produce tumors in C57BL/6 mice, and RT-PCR analysis of tumor-derived cells was not possible for this group. In summary, inoculation of C57BL/6 mice with mixtures of TAP1-positive and TAP1-negative cells resulted in selective loss of TAP1-positive cells and outgrowth of TAP1-negative cells to produce tumors.

In addition, tumor cell expression of MHC-I was assessed by flow cytometry to determine whether TAP1 mRNA expression, as assessed by RT-PCR, correlated with functional manifestations of TAP1 protein expression. To simplify the distinction between MHC-I expression on TAP1-positive and TAP1-negative cells, MHC-I expression was enhanced by stimulation of the cells with IFN-γ for 2 days. Inoculation of C57BL/6 mice with mixtures of TAP1-positive and TAP1-negative cells produced tumors that were composed of cells with uniformly low MHC-I expression, consistent with a uniform lack of TAP function (Figs. 4,C and 5 and Table II). In contrast, inoculation of athymic mice with the same mixture produced tumors that contained many cells with higher MHC-I expression, consistent with expression of functional TAP by many tumor cells (Figs. 4,B and 5 and Table II). In these tumors, TAP-positive cells actually outnumbered TAP-negative cells, suggesting a relative growth disadvantage for TAP-negative cells in athymic mice (perhaps due to selection pressure by NK cells). This observation does not alter the interpretation of the main mixing experiment in immunocompetent C57BL/6 mice, which showed that the TAP1-negative cells selectively grew to produce tumors, whereas TAP1-positive cells were eliminated (if the selection mechanism noted in athymic mice were significant in C57BL/6 mice, it would have produced the opposite result). In summary, tumor cells that lacked TAP1 expression and TAP function were selected during tumor growth in C57BL/6 mice, whereas cells with TAP1 expression and TAP function were selectively eliminated. Furthermore, these observations confirm the ability of TAP1-negative tumor cells to produce tumors even in the context of rejection of adjacent TAP1-positive tumor cells by an immunocompetent host.

This controlled in vivo study directly demonstrates that lack of TAP1 expression can increase tumor cell survival and tumorigenesis in an immunocompetent host. In addition, the difference between the growth of TAP1-positive and TAP1-negative tumors was found to be T cell dependent. Thus, TAP deficiency can allow tumor cells to evade immune surveillance by T cells. Moreover, in this model the evasion of T cell responses predominated over possible enhancement of NK cell activity, because TAP1-deficient cells had increased tumorigenicity.

Our results differ from those of Franksson et al. (32), who demonstrated that a TAP-deficient murine lymphoma cell line was less tumorigenic than a TAP-positive control cell line due to increased recognition of TAP-deficient tumor cells by NK cells. NK cells were apparently not the major mediators of tumor killing in our system, in spite of the fact that the TAP1-negative tumor cells had low MHC-I surface expression. Although NK cell sensitivity of a target correlates with decreased MHC-I expression, MHC-I expression is not the sole determinant of sensitivity to NK cells and does not fully predict the importance of NK cell effects in vivo. Different cells vary in their relative susceptibilities to killing by T cells and NK cells, and this may influence the evolution of tumors to evade immune surveillance. Defects in Ag processing would provide an effective evasion mechanism primarily for tumors that express immunogenic tumor Ags and are recognized by T cells, but not for other tumors that do not elicit T cell responses and are susceptible to killing by NK cells. Thus, the incidence and impact of TAP deficiency and other Ag processing defects may vary from one type of tumor to another.

In theory, effective evasion of both T cells and NK cell responses could occur if the expression of tumor Ag-derived peptide-MHC-I complexes were decreased sufficiently to inhibit T cell recognition, but total surface expression of MHC-I was not reduced so drastically as to elicit NK cell killing. Alterations in Ag processing may accomplish this in some cases, because the production of specific peptide-MHC-I complexes from tumor Ags may be blocked, yet some expression of MHC-I may be maintained to inhibit killing by NK cells. Deficiencies in the regulated proteasome subunits do not decrease overall MHC-I expression, although the production of certain specific peptide-MHC-I complexes is inhibited (39). Even deficiency of TAP does not entirely eliminate MHC-I expression, despite the blockade in processing of many cytosolic Ags. Thus, for some tumors inhibition of tumor Ag processing may block CD8 T cell responses but may still allow sufficient enough MHC-I expression to decrease killing by NK cells.

The fact that deficiency of TAP1 can enhance survival of tumor cells in vivo suggests that defects in the expression of TAP and other Ag processing components by human tumors actually confer advantages for survival and growth of these cells in the host. In turn, this suggests that T cells kill some tumor cells, causing selection for defects in Ag processing and presentation. Therefore, it should be possible for TAP-negative tumor cells to survive, even as adjacent TAP-positive tumor cells are recognized and eliminated, in the face of any other cellular responses or effector mechanisms that are recruited by the T cell responses. To test this hypothesis, C57BL/6 mice were inoculated with a mixture of TAP1-positive and TAP1-negative cells. Tumors derived from these cell mixtures uniformly lacked TAP expression, as judged by RT-PCR of cultured tumor cells, and TAP function, as judged by decreased expression of MHC-I by these cells. Thus, TAP1-negative tumor cells selectively survived and produced tumors in immunocompetent mice.

The selective survival of TAP1-negative tumor cells in this experimental system supports a model wherein tumor progression is accompanied by the development of deficiencies in Ag processing by some tumor cells and the selective survival of cells with these deficiencies. A tumor cell that loses TAP expression in the course of natural tumor progression will initially still express peptide-MHC-I complexes, but the expression of these previously formed complexes will decay. The decay of tumor Ag presentation may occur before killing of the tumor cells by CTL, because the initial CTL response may not suffice to kill all tumor cells. Rather, T cells may attack a tumor over days to weeks (or longer in some systems), providing a prolonged period of selection pressure. In a spontaneous tumor, this process may be even more prolonged, because such tumors progress over weeks, months, or years, providing time for immune selection to occur. Thus, immune selection of tumors may occur over a period sufficient to allow loss of peptide-MHC-I complexes and selection of cells with decreased expression of these complexes.

Although the effector phase of antitumor CD8 T cell responses involves Ag presentation by tumor cells, in at least some cases, the priming of antitumor CD8 T cell responses involves the presentation of tumor Ags by bone marrow-derived or phagocytic APCs (40, 41), which must internalize tumor Ags and process them via alternate MHC-I Ag processing pathways (42, 43, 44, 45). The mechanisms of alternate MHC-I Ag processing remain controversial. Some studies have suggested cytosolic processing, wherein internalized Ags escape from vacuolar compartments into the cytosol and are processed in a TAP-dependent manner (46, 47). Other studies have suggested vacuolar processing and binding of peptides to MHC-I molecules within vacuolar compartments (45, 48, 49). Although vacuolar alternate MHC-I processing was found to be TAP-independent in some studies, other studies have shown that even vacuolar alternate MHC-I processing can be TAP-dependent (50), because deficiency of TAP may decrease the availability of peptide-receptive MHC-I molecules in vacuolar compartments (50, 51, 52). Regardless of the exact mechanism of alternate MHC-I processing in bone marrow-derived APCs, the priming of CD8 T cell responses by these cells should not be altered by deficits in MHC-I Ag processing in the tumor cells. Thus, we predict that deficiencies in the expression of Ag processing components by tumor cells should not affect the priming of antitumor CD8 T cell responses by other APCs. These deficiencies should reduce any contribution of tumor cells to priming of CD8 T cell responses and, regardless of the priming mechanism, should inhibit the ability of T cells to recognize and kill tumor cells.

The ability of tumor cells to evade T cell responses has important implications for both our understanding of tumorigenesis and the potential efficacy of immunotherapy for cancer. If mutations in TAP and other Ag processing components allow tumor cells to escape antitumor surveillance by T cells, this may compromise some immunotherapy strategies based on T cell responses. Further understanding of the role of Ag processing defects in tumor immunity will provide important insight into potential problems with immunotherapy strategies and may suggest ways to circumvent these problems.

We thank Luc Van Kaer for generous gifts of knockout mice; John Monaco for generous gifts of TAP1 vectors; and Jui-Han Huang, Mark Tykocinski, Robert Plenge, Tera McCool, and John Schreiber for providing valuable advice, assistance, and reagents.