Loss of Functional Beta₂-Microglobulin in Metastatic Melanomas From Five Patients Receiving Immunotherapy

Abstract

Background:

In a subset of patients with metastatic melanoma, T lymphocytes bearing the cell-surface marker CD8 (CD8⁺ T cells) can cause the regression of even large tumors. These antitumor CD8⁺ T cells recognize peptide antigens presented on the surface of tumor cells by major histocompatibility complex (MHC) class I molecules. The MHC class I molecule is a heterodimer composed of an integral membrane glycoprotein designated the α chain and a noncovalently associated, soluble protein called beta₂-microglobulin (β₂m). Loss of β₂m generally eliminates antigen recognition by antitumor CD8+ T cells.

Purpose:

We studied the loss of β₂m as a potential means of tumor escape from immune recognition in a cohort of patients receiving immunotherapy.

Methods:

We successfully grew 13 independent tumor cell cultures from tumor specimens obtained from 13 patients in a cohort of 40 consecutive patients undergoing immunotherapy for metastatic melanoma and for whom tumor specimens were available. These cell lines, as well as another melanoma cell line (called 1074mel) that had been derived from tumor obtained from a patient in a cytokine–gene therapy study, were characterized in vitro cytofluorometrically for MHC class I expression and by northern and western blot analyses for messenger RNA (mRNA) and protein expression, respectively, and ex vivo by immunohistochemistry.

Results:

After one melanoma cell line (1074mel) was found not to express functional β₂m by cytofluorometric analysis, four (31%) of the 13 newly established melanoma cell lines were found to have an absolute lack of functional MHC class I expression. Northern blot analysis of RNA extracted from the five cell lines exhibiting no functional MHC class I expression showed that these cells contained normal levels of α-chain mRNA but variable levels of β₂m mRNA. In addition, no immunoreactive β₂m protein was detected by western blot analysis. When human β₂m was transiently expressed with the use of a recombinant vaccinia virus, cell-surface MHC class I expression was reconstituted and the ability of these five cell lines to present endogenous antigens was restored. Immunohistochemical staining of tumor sections revealed a lack of immunoreactive MHC class I in vivo, supporting the notion that the in vitro observations were not artifactual. Furthermore, archival tumor sections obtained from patients prior to immunotherapy were available from three patients and were found to be β₂m positive. This result was consistent with the hypothesis that loss of β₂m resulted from immunotherapy.

Conclusions:

These data suggest that the loss of β₂m may be a mechanism whereby tumor cells can acquire immunoresistance. This study represents the first characterization of a molecular route of escape of tumors from immune recognition in a cohort of patients being treated with immunotherapy.

A subset of T lymphocytes bearing the cell-surface marker CD8 can be shown to directly lyse tumor cells in vitro (1,2). These T cells, designated CD8⁺ T cells, can be expanded to large numbers ex vivo and adoptively transferred back to patients together with interleukin 2 (IL-2) where they can, in some cases, effect the regression of even large tumors (3-5). In patients with a number of different human malignancies, antitumor T cells can be elicited that are capable of recognizing autologous tumor cells as measured by cytolytic- and cytokine-release assays (6-10). Tumor deposits from patients with metastatic melanoma lesions yield tumor-infiltrating lymphocytes (TILs), with antitumor specificity in approximately 30% of the cases, making melanoma the most tractable of human cancers to T-cell-based immunotherapy. Objective clinical responses are observed in approximately 35% of patients with metastatic melanoma who are treated with TIL-based therapy; some responses are complete and long lasting [reviewed in (11)].

The reasons why the metastatic lesions of some patients yield successful. TIL cultures ex vivo while others do not are unknown. It is also unknown why some patients with apparently specific and lytic TILs often fail to respond. Even more perplexing is why some tumors escape after an initial response to therapy and why only some, but not all, of the lesions in an individual respond to treatment.

Abnormalities in T-cell signal transduction induced by tumor cells are potential mechanisms for tumor escape from immune recognition (12). However, since most patients with melanoma do not have measurable immunosuppression, the tumor cell has become the target of investigation to determine the mechanisms of escape from immunologic recognition (13). The steps required for the processing and presentation of antigens for recognition by CD8⁺ T cells, in particular, may be involved in tumor escape, since CD8⁺ T cells do not recognize intact antigens on the surfaces of tumor cells but generally recognize peptide fragments of protein antigens presented by major histocompatibility complex (MHC) class I molecules (14-16).

MHC molecules are known as human leukocyte antigens (HLAs) in humans and H-2 antigens in mice (17). MHC class I molecules are heterodimers composed of a 44- to 46-kd integral membrane glycoprotein designated the α chain and a noncovalently associated, soluble 12-kd protein called beta₂-microglobulin (β₂m). All nucleated cells in the body, with the exception of germline cells and some neurons (18), express class I–peptide complexes, the ligands for CD8+ T cells. A molecular understanding of the structure of MHC molecules has made clear their true function with respect to antigen recognition by T cells: MHC molecules are receptors for peptide antigens (19). MHC molecules are physically associated with peptide antigens, and x-ray crystallographic data have indicated precisely how such an interaction occurs. The solution of the crystal structure of MHC class I and II molecules has clarified the molecular structure of MHC molecules and specifically the way MHC molecules bind antigen (20-22). MHC molecules have peptide-binding domains, consisting of a deep groove that runs between two long α helices found on the outward-facing surface of the MHC molecule. In the case of class I molecules, x-ray crystallographic findings have since been refined and extended to include x-ray images of particular peptides lying in the cleft of MHC molecules (23-26). These findings reveal that the peptides are bound in an extended conformation. Antigenic peptides bound by MHC class I molecules are generally eight to 10 amino acids in length, usually resulting from proteolytic activity in the cytoplasm.

Peptides are transported into the endoplasmic reticulum by a specialized adenosine triphosphate-dependent transporter called TAP, which is related to the multidrug resistance protein, Mdrl. Peptides then assemble together with α chains and β₂m to form a trimolecular complex that is then transported out of the endoplasmic reticulum, through the Golgi apparatus, to the cell surface for potential recognition by CD8+ T cells. With few exceptions (27), recognition by CD8⁺ T cells does not occur in the absence of β₂m (28,29), since MHC class I α chains not associated with β₂m are retained in the endoplasmic reticulum (30).

Bicknell et al. (31), Bodmer et al. (32), and Momburg and Koch (33) have shown that human cancer cells, especially cells of adenocarcinomas of the colon, may fail to express β₂m; Wang et al. (34) and D'Urso et al. (35) have shown that two human melanoma cell lines do not express MHC class I as a result of mutations in the genes encoding β₂m. The loss or mutation of the genes encoding β₂m is a highly efficient mechanism for tumor escape from immune recognition, since stable presentation of peptide antigen by MHC class I molecules does not occur in the absence of β₂m. The purpose of our study was to investigate the loss of β₂m as a potential means of tumor escape from immune recognition in a cohort of patients undergoing immunotherapy.

Materials and Methods

Patients

Tumors were obtained according to institutional guidelines, and National Institutes of Health (NIH) Clinical Center standards for informed consent were adhered to for all patients. Thirteen independent tumor cell cultures were successfully grown from a cohort of 40 consecutive patients receiving immunotherapy for whom tumor specimens were available. These patients were enrolled in clinical trials at the National Cancer Institute (NCI), Bethesda, MD; the results of one of these trials was recently published (4).

Tumor Cells

Melanoma cell lines (1074mel, 1106mel, 1180mel, 1174mel, 1259mel, 526mel, 888mel, 397mel, and 624mel) were established in our laboratory, and the last four of these cell lines have been previously described (36). One tumor cell culture (1074mel; i.e., melanoma cell culture from patient No. 1074) was not part of the cohort of patients undergoing immunotherapy for metastatic melanoma and was derived for a different purpose [a cytokine–gene-therapy study (37)]. The β₂m-deficient melanoma cell line FO-1 has been previously described (35). Daudi (a B-lymphoblastoid cell line derived from a patient with Burkitt's lymphoma), H82sclc [a small-cell lung cancer cell line (14)], and K562 (human chronic myelogenous leukemia) were obtained from the American Type Culture Collection (ATCC), Rockville, MD. Tumor cells were maintained as a monolayer culture in complete medium containing RPMI-1640, 10% heat-inactivated fetal calf serum (both from Biofluids, Rockville, MD), and 0.03% (100 mM) glutamine (NIH Media Unit, Bethesda, MD).

Cytofluorography

Cultured tumor cell lines were harvested with 0.05% trypsin–0.02% versene without calcium and magnesium (Biofluids), washed, then incubated with hybridoma culture supernatant containing the W6/32 monoclonal antibody (MAb) for 30 minutes (Sera Labs, Westbury, NY) (mouse immunoglobulin G2a [IgG2a] isotype). W6/32 reacts with monomorphic determinants on the HLA-A, -B, and -C molecules. In all cases, cells were stained with the appropriate isotype-matched control antibody, nonspecific mouse IgG2a (Becton Dickinson Immunocytometry Systems, San Jose, CA). MAb binding to cells was followed by binding with goat anti-mouse fluorescein isothiocyanate-conjugated antibody (Boehringer Mannheim Biochemicals, Indianapolis, IN) and detected by fluorescence-activated cell sorting (FACS) using a FACScan 440 (Becton Dickinson, Mountain View, CA).

Cytotoxic T Lymphocytes

Cytotoxic T lymphocytes (CTLs) were generated from excised tumor specimens by culturing suspension cells in complete medium containing IL-2 (6000 U/mL) (Chiron Therapeutics, Emeryville, CA) for 30-70 days as previously described (38).

Experiments that use mouse CTLs to study antigen processing in human tumors have been described elsewhere (14). In brief, polyclonal CD8⁺ T-cell populations were generated from 6- to 8-week-old female BALB/c mice by intravenous injection of 5 × 10⁶ plaque-forming units of vaccinia virus. After at least 2 weeks, spleens were removed, dispersed to single-cell suspensions, and stimulated in vitro with vaccinia-infected BALB/c splenocytes at a ratio of 2:1. Cells were then cultured in complete medium to generate CD8⁺ T cells. The care of animals was in accord with NIH guidelines.

Microcytotoxicity Assays

^5lCr release assays were performed as previously described (39). Briefly, 5000 target cells labeled with 200 μCi of Na^5lCrO₄ (Du Pont NEN, Boston, MA) were mixed with various numbers of effector cells and incubated for 6 hours. In experiments involving infection of cells with vaccinia virus, target cells were infected with 10 plaque-forming units per cell of the designated recombinant vaccinia virus(es) for 60-90 minutes, then labeled with ^5lCr for 1 hour. Target cells were then washed three times and mixed with CD8⁺ T cells. In all micro-cytotoxicity assays, the amount of released ⁵¹Cr was determined by γ-photon counting of aliquots of supernatants of cells after lysis experiments. Percent specific lysis was calculated as follows: ([experimental cpm–spontaneous cpm]/[maximal cpm–spontaneous cpm]) × 100.

Immunohistochemical Staining of Archival Tissue Specimens From Patients

Immunohistochemistry was performed on archival tissue sections where available. Formalin-fixed, paraffin-embedded sections, 5 μm thick, were deparaffinized in xylene and stained as previously described (40). Murine MAb, ATCC HB159 (anti-human β₂m, also known as SF1-1.1.1), supernatant was brought to pH 7 by the addition of HC1 and used at a 1:5 dilution in phosphate-buffered saline (PBS) (Biofluids). Biotinylated horse anti-mouse IgG at 1:200 was used as the secondary antibody (Boehringer Mannheim Biochemicals). Immunoperoxidase staining was completed with the use of the avidin–peroxidase complex (Vector Laboratories, Inc., Burlingame, CA). Diaminobenzidine-hydrogen peroxide (Sigma Chemical Co., St. Louis, MO) was used as the developing solution, and the sections were counterstained with hematoxylin–eosin.

Western Blot Analyses

Tumor cells were lysed by adding a buffer consisting of PBS containing 2 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, and 1 mM dithiothreitol (Sigma Chemical Co.) at 4 °C. The cells were then sonicated for 10-15 seconds, and the lysate was centrifuged at 3000 rpm for 5 minutes at 4 °C. The concentration of protein in the cell homogenate supernatant was determined by the use of the Bio-Rad protein assay (Bio-Rad Laboratories, Richmond, CA). Equal amounts of protein (20 μg) from each cell lysate were electrophoretically resolved on a 16% sodium dodecyl sulfate–polyacrylamide gel and transferred to a nitrocellulose membrane with the use of a Bio-Rad Transblott Mini cell (Bio-Rad) at 250 mA for 1 hour. The nitrocellulose sheet was blocked overnight with Blocking Buffer (5 Prime → 3 Prime, Inc., West Chester, PA) and first incubated with the anti-β₂m MAb HB159 (ATCC) for 2 hours and then with horseradish peroxidase-labeled secondary goat–anti-mouse antibody for 2 hours. Immunofluorescence was detected with the use of the Enhanced Chemi Luminescence (Amersham Life Science Inc., Arlington Heights, IL). Protein sizes were estimated with the use of low-molecular-weight standards (14 000-70 000; Sigma Chemical Co.).

Northern Blot Analysis

Specific probes for β₂m were generated by the use of the reverse transcription-polymerase chain reaction (PCR) method. The total RNA was isolated with the use of guanidine isothiocyanate (Fluka Chemical Corp., Ronkonkoma, NY) then purified with the use of the cesium chloride (Serva Biochemicals) centrifugation method, as previously described (14) from the Epstein-Barr virus-transformed B-cell line designated 501 that was established in our laboratory (41). First-strand complementary DNA (cDNA) was synthesized from 10 μg total RNA using an oligo dT primer and murine leukemia virus reverse transcriptase (Life Technologies [GIBCO BRL], Gaithersburg, MD). The specific 5′-side oligonucleotide primer for β₂m that was used had the sequence 5′ CAC GTC ATC CAG AGA ATG G 3′, and the 3′-side oligonucleotide primer had the sequence 5′ CGA TCC CAC TTA ACT ATC TTG G 3′. For PCR amplification, samples were initially denatured for 2 minutes at 94 °C, then 30 cycles were performed with the use of the following conditions: 94 °C for 30 seconds, 60 °C for 30 seconds, and 72 °C for 1 minute followed by an extension cycle of 10 minutes at 72 °C. PCR products were electrophoresed on 1.5% agarose gels with an acetate buffer, and the predicted fragment sizes of bands (260 bp) were isolated, purified by the glass-powder method (GeneClean, La Jolla, CA), and used as hybridization probes for northern blots. The HLA-A2.1 full-length cDNA was used as a probe template for MHC class I (42). Human β-actin cDNA was purchased from Clontech Laboratories, Inc., Palo Alto, CA. The probes were labeled with [³²P]-deoxycytidine triphosphate (Amersham Life Science Inc.) by the random-priming method as previously described (14). For the northern blot analysis, 10 μg of total RNA from each tumor was electrophoresed in a 1% agarose–formaldehyde gel, transferred to a nylon membrane (Zeta-Probe, Bio-Rad Laboratories), and fixed to the membrane by UV cross-linking. Prehybridization was done for 30 minutes with the use of the 5 Prime → 3 Prime Northern Hybridization buffer. Hybridization was done in a 40% formamide hybridization solution (Northern hybridization buffer, 5 Prime → 3 Prime, Inc.) at 42 °C overnight. Membranes were washed three to four times in 2× standard saline citrate at 60 °C for 30 minutes for each wash, and autoradiography was then performed.

Gene Replacement Studies

The production of a recombinant vaccinia virus expressing the K^d gene (K^d-Vac) has been described (14), as has the production of the recombinant vaccinia virus containing the β₂m gene (β₂m-Vac) (42).

Results

Steady-state cell surface expression of MHC class I was studied in a group of patients with metastatic melanoma at the NCI. All of the patients entered in the T-cell-based immunotherapy protocol had either failed standard therapy or had disease for which no effective therapy was available. All patients had clinically assessable disease and received no other therapy during the 30 days before the protocol treatment or during treatment. Patient characteristics, including the immunotherapy received, are shown in Table 1.

Table 1.

Patient characteristics

Patient No.	Age, y; sex	Specimen site*	Comments†	Tumor-infiltrating lymphocyte growth	Class I, histology	Response to immunotherapy‡
1106	46; male	SC	Previous response to treatment with IL-2 and IFN α; recurred and did not rerespond	No	No	—
		SC	Same as above	Yes	Yes§	Mixed
		SC∥	Treatment with TILs from above, no response	No	No¶,#	—
		SC	Additional biopsies	No attempt	No	—
1180	46; female	SC	No previous treatment of metastasis	Yes	Yes	No
		SC	New nodule after no response to treatment with TILs/IL-2/IFN α	Yes	Low**	No
		SC∥	New nodules after no response to TILs/IL-2/IFN α	No	No††,‡‡	—
1074§§	26; female	SC∥	Previous mixed response to IL-2	Yes	No	No
1174	53; male	LN∥	No response to previous IL-2	No	No	—
1259	30; male	SC	No previous immunotherapy	Yes	Yes	Yes
		SC∥	Previous response to TILs; given cryopreserved TILs	No	No	No

^*SC = subcutaneous; LN = lymph node.

^†IL-2 = interleukin 2; IFN α = interferon alpha; TILs = tumor-infiltrating lymphocytes.

^‡For definitions of clinical responses, see (4).

^§Immunohistochemical staining, shown in Fig. 6, panel B.

^∥Tumor from which melanoma cell line was derived.

^¶Hematoxylin–eosin stain, shown in Fig. 6, panel A.

^#Immunohistochemical staining, shown in Fig. 6, panel C.

^**Immunohistochemical staining, shown in Fig. 6, panel E.

^††Hematoxylin–eosin stain, shown in Fig. 6, panel D.

^‡‡Immunohistochemical staining, shown in Fig. 6, panel F.

^§§From another study (42).

Of the 40 attempts to grow tumor cells obtained from patients from whom melanoma deposits were surgically resected, 18 tumor cell lines could not be grown at all and nine could be derived, but did not grow adequately, and therefore were not studied for their expression of cell surface HLA. Of the 13 tumor cell lines that could be derived into tumor cell cultures that could be grown to quantities sufficient for in vitro evaluation, four were found to be devoid of MHC class I and were included in the present study for the characterization of their MHC class I deficits. One additional tumor cell culture (from patient No. 1074) was not part of this group of 40 and was derived for a different purpose [a cytokine gene-therapy study (37)], but is nevertheless included here because it was observed to be MHC class I deficient.

From this screening analysis, five independently established melanoma cell cultures from patients with metastatic melanoma were devoid of cell-surface HLA class I molecules as measured cytofluorometrically (Fig. 1). The loss of MHC class I expression on these cell lines was complete, even when assessed by the MAb W6/32 (that can recognize virtually all human HLA class I alleles).

Fig. 1.

Absence of major histocompatibility complex class I on the surfaces of metastatic human melanoma cell lines and restoration by transient expression of beta₂-microglobulin (β₂m). Fluorescence-activated cell sorter analysis was performed on tumor cell lines alone (heavy solid line), infected with wild-type vaccinia virus (dashed line), and infected with the recombinant vaccinia virus expressing β₂m (light solid line, shaded gray under the curve); cell lines were sequentially incubated with either the control immunoglobulin G2a antibody (left panels) or the W6/32 monoclonal antibody (right panels) and then with goat anti-mouse fluorescein isothiocyanate-conjugated Fab fragments. The cell lines analyzed were the positive control cell line 888mel (that is normal in its antigen-presenting capacity), the negative-control H82sclc (a small-cell lung cancer cell line that does not process antigens), and the five melanoma cell lines (1259mel, 1074mel, 1180mel, 1106mel, and 1174mel) that have lost functional β₂m. For two cell lines (H82sclc and 1180mel), only two curves are shown in this experiment because the tumor-cell-alone control was not performed because of inadequate numbers of cells.

To characterize the mechanisms underlying the lack of MHC class I expression, we considered the possibility that HLA α chains were poorly loaded with peptides in the endoplasmic reticulum because of down-regulation or mutation of the TAP peptide transporters or proteasome components, since we have observed this mechanism to be functioning in small-cell lung cancer (14). However, defects in the above-named components of the antigen-processing machinery have generally resulted in MHC class I expression that is decreased, but not absent, since some peptides can originate from peptides or proteins translocated across the endoplasmic reticulum membrane by TAP-independent signal sequence mechanisms, exposing the proteins to proteolytic activities in the endoplasmic reticulum (43,44).

The completeness of the loss observed in the melanoma cell lines suggested the structural loss or mutation of either all of the heavy-chain MHC class I molecules or of β₂m. The former would require disabling mutations of six independent genomic transcripts (two alleles each of HLA-A, -B, and -C) or the complete loss of approximately 1500 bp of the MHC class I region on the short arm of both copies of chromosome 6 (since MHC class I molecules are codominantly expressed). Another, more likely, possibility was the loss or mutation of both copies of β₂m, a structural component of cell-surface MHC class I.

To explore the possibility that the deficit in functional β₂m occurred at the level of decreased gene transcription or increased degradation of messenger RNA (mRNA), northern blot analysis of the melanoma tumor cell lines was performed. The α-chain mRNA levels were found to be normal (data not shown). However, expression of β₂m mRNA was variable and lower than that seen in control cell lines (Fig. 2). The controls used in this experiment were the antigen-presentation normal melanoma cell lines 624 and 397 and the FO-1 cell line, which is known to have a defect in the expression of functional β₂m, due to a lack of β₂m gene expression (35). A western blot analysis, however, failed to detect any immunoreactive protein (Fig. 3).

Fig. 2.

Northern blot analysis of total RNA extracted from the indicated cells (624[mel], 397[mel], FO-1, 1106[mel], 1074[mel], 1l74[mel], 1180[mel], 1259[mel]) and probed for the expression of beta₂-microglobulin (β₂m) messenger RNA. The northern blot probed for β₂m messenger RNA (top panel) was stripped and probed for β-actin to ensure sufficient loading of all lanes (bottom panel).

Fig. 3.

Western blot analysis of total protein extracts for expression of immunoreactive beta₂-microglobulin (β₂m). Lysates of melanoma cell lines 1074mel, 1106mel, 1180mel, 1174mel, and 1259mel do not express any measurable immunoreactive β₂m. Line l074mel is shown after infection with recombinant vaccinia virus expressing β₂m (left-most panel) or after infection with wild-type control vaccinia virus (second from left).

To test the hypothesis that β₂m was responsible for the absence of MHC class I on the cell surfaces, we sought to replace functional β₂m within the cells. Since MHC class I molecules are generally retained in the endoplasmic reticulum if they are not complexed with β₂m, the addition of functional β₂m would likely reconstitute cell-membrane expression of MHC class I molecules. One rapid, although transient, method of β₂m gene expression involved infection of the cells with a recombinant vaccinia virus that expresses β₂m. As shown in Fig. 1, expression of β₂m mRNA by a recombinant vaccinia virus elicited MHC class I expression on the surfaces of the five human melanoma cell lines under study. Indeed, there may be some suppression of endogenous MHC class I expression by control vaccinia virus in line 1259M, a phenomenon that has been described by others (45). The melanoma cell line 888 was used as a positive control, since it exhibits normal antigen presentation; the small-cell lung cancer cell line H82 was used as a negative control, since it is defective in antigen processing, as previously reported (14).

The results of the western blot analysis, together with the cytofluorographic evidence, suggested that the genetic insertion of functional β₂m would fully restore the ability of the cell lines to process and present endogenous antigens for recognition by CD8+ T cells. On the other hand, if there were additional abnormalities in antigen processing, a full restoration of function in the antigen-presenting capabilities of the cell lines being studied would not be seen with the replacement of β₂m. To test this question directly, we employed a method for the study of antigen processing that was independent of both the HLA type of the tumor and the presence or absence of specific cellular proteins. With the use of a recombinant vaccinia virus encoding the mouse H-2 Kd MHC class I molecule (K^d-Vac), we tested human tumor cell lines for presentation of viral antigens to mouse K^d-restricted, vaccinia-specific CD8⁺ T-cell populations and thus studied antigen-processing capabilities of human tumor cells per se. As shown in Fig. 4, restoration of the β₂m molecule in the tumor cells completely restored the antigen-processing capabilities of all five of the melanoma cell lines tested. The expression of β₂m had no effect on H82, the negative control in this experiment (14).

Fig. 4.

Microcytotoxicity assay evaluating the antigen-presenting capacities of beta₂-microglobulin (β₂m)-deficient cell lines. Cell lines were screened for their capacities to present vaccinia virus antigens in the context of a transiently expressed H-2 K^d molecule. Effector cells (E) were vaccinia virus-specific, H-2 K^d-restricted murine cytotoxic T lymphocytes. Target cells (T) were either uninfectcd (▲), infected with vaccinia virus expressing H-2 K^d alone (●), or in combination with recombinant vaccinia virus expressing β₂m (△). The E:T ratios for the six points in each curve, reading from left to right, were 100:1, 50:1, 25:1, 12.5:1, 6.25:1, and 3.125:1. The melanoma cell line 526 (the positive control) is an antigen-processing intact cell line that processes vaccinia antigens efficiently in the absence of β₂m. The small-cell lung cancer cell line H82sclc fails to process antigens even in the presence of β₂m, since the antigen-processing lesions in this cell line are elsewhere (14). Note that 1259mel, 1180mel, 1106mel, 1174mel, and 1074mel all fail to present vaccinia virus antigens but can do so efficiently when functional β₂m is expressed endogenously.

To test whether lack of functional β₂m was the sole mechanism of escape of these tumors from recognition by CD8+ T cells, we studied whether restoration of β₂m in these cell lines was associated with the restoration of killing of these cell lines by antitumor CD8+ T cells. This was a critical question because the loss of β₂m could be accompanied by the loss of tumor-associated antigens, making the restoration of β₂m in these cell lines a less relevant aspect of therapy. The problem with testing this question was that autologous TILs were not available from patients whose tumors gave rise to these cell lines. Since one of the five patients lacking functional β₂m expression was HLA-A2.1 positive (No. 1074) and since we have generated CD8+ T cells that were capable of recognizing ubiquitously expressed melanoma tumor-associated antigens in the context of HLA-A2.1 (46), we could directly test whether restoration of β₂m led to immune recognition by CD8+ T cells. As shown in Fig. 5, expression of functional β₂m in the 1074mel cell line restored its susceptibility to lysis by HLA-A2.1-restricted antimelanoma CD8+ T cells.

Fig. 5.

Microcytotoxicity assay evaluating the capacity of a beta₂-microblobulin (β₂m)-deficient cell line to present tumor-associated antigen after restoration of β₂m. Melanoma cell lines were tested for recognition by HLA-A2.1-restricted antimelanoma effector (E) cells. Target cells (T) were the HLA-A2.1-positive, β₂m-positive 526 cell line (◆); the HLA-A2.1-positive but β₂m-negative 1074 cell line infected with control vaccinia virus (▲) or infected with recombinant vaccinia virus-expressing β₂m (△); the HLA-A2.1-negative but β₂m-positive 888 melanoma cell line with control (□) or β₂m vaccinia virus (■); and the HLA-A2.1-negative and β₂m- negative 1259 cell line with control (●) or β₂m vaccinia virus (○).

Despite the lack of expression of cell-surface MHC class I molecules, all five of the β₂m-deficient melanoma cell lines were capable of being lysed by lymphokine-activated killer cells (data not shown).

To test the possibility that these observations were due simply to an artifact of the culture of the melanoma cells, we sought to assess the status of MHC class I expression of these tumors in vivo. Since a portion of each of the fresh melanoma specimens from which the melanoma cultures were developed was sent for pathologic analysis, we procured the paraffin-embedded tissue blocks and performed immunohistochemical analysis for the presence of β₂m. Tissue sections in all five of the cases in question were found to be uniformly devoid of β₂m (Fig. 6 and data not shown). Note that each tumor cell line contains the positive control of infiltrating cells that make up the vasculature of the tumor. Thus, the cells of the lines under scrutiny did not express β₂m in vivo that was capable of complexing with MHC class I α chains and stabilizing them on the tumor cell surfaces.

Fig. 6.

Immunohistochemical staining of sections of melanoma in situ. Sections from tumor specimens described in Table 1 from patient Nos. 1106 (panels A-C) and 1180 (panels D-F) are shown. Immunohistochemical staining in panels B, C, E, and F was done at the same time, under exactly the same conditions with anti-human beta₂-microglobulin (β₂m) monoclonal antibody followed by immunoperoxidase staining (see “Materials and Methods” section for details). Strongly positive and weakly positive staining for β₂m is shown in panels B and E, respectively. Panels C and F are examples of specimens that stain negative for β₂m. Hematoxylin–eosin stains are shown in panels A and D, where melanin deposits that appear dark brown can be seen.

Finally, archival tumor sections obtained from patients prior to immunotherapy were available for three patients (Nos. 1106, 1180, and 1259) and were found to be β₂m positive (Table 1, Fig. 6, and data not shown). This result was consistent with the hypothesis that the loss of β₂m resulted from immunotherapy.

Discussion

The five cell lines described in this article were independently derived from metastatic melanomas obtained from five patients who had undergone immunotherapy. The tumors from which these five cell lines were derived may escape immune recognition by CD8⁺ T cells via the loss of expression of functional β₂m, since replacement of β₂m fully restores the antigen-processing capabilities of all of these cell lines (Fig. 4) and can restore the ability of antitumor CD8+ T cells to recognize the cell line in the tumor (1074mel) that was matched for HLA-A2.1. Importantly, the loss of β₂m by these tumors was also observed in vivo by immunohistochemical analysis.

The loss of β₂m may be an important mechanism for the escape of some melanoma cells from recognition by CD8+ T cells and could explain why some patients experience a recrudescence of tumor growth, often in aggressive and ultimately lethal forms, after excellent responses to T-cell-based therapy. Indeed, none of the patients responded to immunotherapy subsequent to the observed loss of functional β₂m, and all of these five patients (Table 1) have died of disease that was noted to be clinically very aggressive. Perhaps most importantly, antitumor CD8+ T cells could not be grown from any of these lesions once there was loss of β₂m. However, the sample is too small to enable us to definitively draw the statistical conclusion that patients bearing tumors deficient in functional β₂m have poor outcomes compared with patients whose tumors do not suffer such a loss. Furthermore, we do not have enough illustrative examples to conclude that immunotherapy, or T-cell-based immunotherapy, was responsible for the loss of β₂m. Many patients whose tumors do express MHC class I fail to respond, and thus tumors are likely to use multiple different mechanisms to evade immune destruction. Nevertheless, the findings presented here could be used as the basis for prospective studies of much larger cohorts of patients receiving immunotherapy that could provide more definitive answers to these questions.

The incidence of the loss of functional μ₂m in this study, four (31%) of 13, is strikingly high. However, loss of MHC class I expression in a cohort of patients undergoing immunotherapy with IL-2 has not yet been performed. Abnormalities in MHC class I expression in the absence of immunotherapy may be much higher than has been described in the earlier literature, as indicated by more recent studies examining particular allospecificities (36,47). Furthermore, the loss of β₂m has recently been associated with a mutator phenotype, frequently an early event in the development of carcinoma. The molecular basis for the loss of β₂m in these patients, particularly as it relates to microsatellite instability, is now being explored (48).

We have previously reported poor expression of MHC class I molecules by human small-cell lung cancers and an experimental mouse sarcoma (14,38,49). In our previous studies of deficient MHC class I expression, all of the small-cell lung cancer cell lines could be induced to express MHC class I by the addition of interferon gamma (IFN γ). We could not induce MHC class I expression with the use of IFN γ in the five human melanoma cell lines reported here (data not shown). Thus, there appears to be histologic specificity in the mechanisms underlying a failure of recognition by CD8⁺ T cells: Six of six small-cell lung cancers previously studied presented antigen poorly because of a down-regulation of molecules important in antigen processing, while loss of functional β₂m was responsible for this effect in five of five melanomas in the present study.

The loss of β₂m is a highly effective mechanism of tumor escape from recognition by CD8+ T cells, since MHC class I molecules are not stable in its absence. The loss of expression of both copies of β₂m is also an efficient mechanism of escape, since loss of expression of any two other genes involved in antigen processing, such as TAP or proteasome components, would cause a much less profound loss of antigen-presenting capacity (13,50). Loss of β₂m may result in increased susceptibility to lysis by natural killer (NK) cells (51-53). However, the aggressive lethality of each of the five β₂m-deficient tumors reported in this study suggests either that lysis of these tumors by NK cells is ineffective or that the tumor cells have additional mechanisms of escape from immune recognition by NK cells.

Positive expression of β₂m mRNA by northern blot analysis suggested that the deletion of both copies of the gene was not responsible for the loss of functional ν₂m and that mutation of one or both copies was probably responsible for the complete loss of functional β₂m. The absence of β₂m shown by western blot analysis indicated that the epitopes recognized by the MAb were not present to any measurable extent. Since western blot analysis measures the steady-state level of antigen, either the epitope was not produced because of mutation or premature termination or the protein produced had a vastly reduced half-life because of enhanced proteolysis due to a change in sequence or to a misfolding of the protein.

Studies in experimental murine models indicate that loss of β₂m expression by germline transmission of a homozygous disruption of the β₂m gene is not lethal (54). The β₂m-deficient mice express little or no functional class I antigen, and while other T-cell subsets were found in their normal distribution, the mice were found to have no mature CD8+ T cells (29). It is interesting that these mice were still capable of rejecting skin grafts (55) and were able to clear some viral infections (56,57). These findings are consistent with a view of the immune system as highly redundant. Indeed, all of the β₂m-loss mutants remained sensitive to lysis by lymphokine-activated killer cells, suggesting that immunotherapy based on non-MHC-restricted effector cells may be more effective in selected patients.