Understanding immune mechanisms influencing cancer regression, recurrence, and metastasis may be critical to developing effective immunotherapy. Using a tumor expressing HIV gp160 as a model viral tumor Ag, we found a growth-regression-recurrence pattern, and used this to investigate mechanisms of immunosurveillance. Regression was dependent on CD8 T cells, and recurrent tumors were resistant to CTL, had substantially reduced expression of epitope mRNA, but retained the gp160 gene, MHC, and processing apparatus. Increasing CTL numbers by advance priming with vaccinia virus expressing gp160 prevented only the initial tumor growth but not the later appearance of escape variants. Unexpectedly, CD4 cell depletion protected mice from tumor recurrence, whereas IL-4 knockout mice, deficient in Th2 cells, did not show this protection, and IFN-γ knockout mice were more susceptible. Purified CD8 T cells from CD4-depleted mice following tumor regression had more IFN-γ mRNA and lysed tumor cells without stimulation ex vivo, in contrast to CD4-intact mice. Thus, the quality as well as quantity of CD8+ CTL determines the completeness of immunosurveillance and is controlled by CD4 T cells but not solely Th2 cytokines. This model of immunosurveillance may indicate ways to enhance the efficacy of surveillance and improve immunotherapy.
Improvement of T cell-mediated immunity for immunotherapy and development of vaccines has been one of the major strategies against cancer in this decade. Although many tumor-associated Ags were found in human cancer cells (1, 2, 3, 4, 5, 6, 7, 8, 9, 10) and several trials have been conducted in a number of cancer patients (11, 12, 13, 14, 15, 16), with a few exceptions (16), most of these therapies failed to induce an immune response sufficient to prevent further development of disease. The reasons for this poor success rate should be considered from the perspectives of both the factors in the tumor and those in the host. Lessons may be learned from the natural immunosurveillance against tumors expressing tumor Ags that could be applied to immunotherapy.
Despite the induction of a specific CTL response and appropriate help by helper T cells, it has been difficult to eradicate all of the tumor cells. There is accumulating evidence for escape mechanisms of tumor cells (17, 18, 19), including loss of Ag or class I expression, production of suppressive cytokines by tumor cells, and expression of Fas ligand on tumor cells (20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37). Considering the fact that escape variants expanded again after nearly complete rejection (20), it is important to determine how to prevent these tumor escape mechanisms to obtain durable remissions.
In this study, we examine whether CTL induced against a model viral tumor Ag can control the growth of tumor, how tumor cells escape from this immunosurveillance, and how we can prevent those escape variants. As a model tumor with a well-characterized Ag as a model viral tumor Ag, we used a BALB/c 3T3 fibroblast line transfected with HIV gp160 and with mutant ras and myc for tumorigenicity. Immunosurveillance of tumors expressing viral Ags may succeed, as is often the case for EBV-transformed B lymphocytes, or sometimes fail, as in the case of cervical carcinoma expressing human papillomavirus Ags, even in individuals who are not immunodeficient. We have previously characterized in depth the murine CTL response to an immunodominant determinant of gp160, called P18, contained within the V3 loop (38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48), that facilitates use of this Ag as a model viral tumor Ag.
There have been reports indicating that Th2-type cytokines down-regulated antitumor immunity (49, 50) and the activation of type 1 T cell responses produced antitumor immunity (51, 52, 53, 54). In a cross-sectional epidemiological study of papillomavirus-related cervical neoplasia, we observed an inverse correlation between the fraction of individuals making a Th1 cytokine response and the degree of progression of disease (55). Thus, a shift to Th1-type cytokine production may be one goal for effective immunotherapy for tumors as well as virus infection. Therefore, we also address a question whether reduction of Th2-type cytokine production could enhance immunosurveillance. We found a novel striking enhancement of surveillance by CD4 cell depletion that cannot be explained primarily by elimination of Th2 cells.
Thus, we have taken advantage of the intriguing pattern of growth, spontaneous (immune-mediated) regression, and recurrence of this novel model tumor to investigate the relation between different cellular immune responses and tumor growth. We examined the role of CD8 and CD4 cells in the initial tumor regression and in preventing or facilitating tumor recurrence. We also examined the molecular mechanism of tumor escape from CTL in vivo. Unexpectedly, we discovered a novel important role for the quality of CTL, with respect to ex vivo activity and the amount of IFN-γ mRNA, in prevention of recurrence of tumor, and the regulation of this CTL quality by CD4 cells.
Materials and Methods
Mice and reagents
Female BALB/c mice were purchased from Charles River Breeding Laboratories (Frederick, MD). IL-4 knockout (KO)2 and IFN-γ KO mice having the BALB/c background were obtained from The Jackson Laboratory (Bar Harbor, ME). All mice were kept under pathogen-free conditions and used at 6–10 wk old. Animal experiments were all approved by the National Cancer Institute (NCI) Animal Care and Use Committee.
Anti-CD4 mAb (clone GK1.5) and anti-CD8 mAb (clone 2.43) for in vivo depletion were obtained from the Frederick Cancer Research and Development Center, NCI (Frederick, MD). FITC-conjugated anti-H-2Dd mAb (clone 34-2-12) was purchased from PharMingen (San Diego, CA). Mouse CD8 T cell subset enrichment columns were obtained from R&D Systems (Minneapolis, MN). Recombinant vaccinia vPE16, expressing the gp160 envelope protein of HIV-1 strain IIIB and control vaccinia vSC8, expressing β-galactosidase, were kindly contributed by Drs. Patricia Earl and Bernard Moss (National Institute of Allergy and Infectious Disease, Bethesda, MD) (56, 57). Trizol Reagent, Super Script cDNA synthesis reagents, and PCR SuperMixture were purchased from Life Technologies (Rockville, MD). NorthernMax, Stripe-EZ RNA Kit, and pT7 mouse IFN-γ probe template were obtained from Ambion (Austin, TX).
15-12 are BALB/c 3T3 cells transfected with gp160 envelope protein of HIV-1 IIIB (38). 18Neo are BALB/c 3T3 cells transfected with the neomycin resistance gene alone as a control. 15-12RM was made from 15-12 by transfection with Myc and mutant Ras genes, containing the substitution of glycine to valine at position 12 of K-ras p21. All cells were maintained in T cell complete medium containing 0.2 mg/ml of geneticin (Sigma, St. Louis, MO).
Splenocytes from BALB/c mice previously immunized with 1 × 107 PFU of vPE16 were stimulated with irradiated BALB/c splenocytes pulsed with 1 μM peptide18 (P18) IIIB in a 24-well culture plate in complete T cell medium supplemented with 10% T-stim (Collaborative Biomedical Products, Bedford, MA). After 7 days of culture, viable cells were harvested and a CTL line against P18-IIIB was established by several restimulations with P18-IIIB-pulsed splenocytes. A CTL line for 15-12RM was derived from splenocytes of 15-12RM tumor-bearing mice taken on day 50 after inoculation of 15-12RM cells. These splenocytes were stimulated with 15-12RM cells, which were treated with mitomycin C (Sigma) (100 μg/ml for 45 min), plus irradiated splenocytes of normal BALB/c mice. The CTL line specific for 15-12RM was induced after several such restimulations. T cell complete medium is Biofluids (Rockville, MD) R2E Medium (a 50:50 mixture of RPMI 1640 and EHAA media) supplemented with l-glutamine, sodium pyruvate, nonessential amino acids, penicillin, streptomycin, 5 × 10−5 M 2-ME, and 10% FCS.
Cytolytic activity against several target cells was assayed by a 4-h 51Cr-release assay, as described elsewhere (58). Tumor cells harvested from tumor-bearing mice were used for target cells either on the same day when they were resected or after 1 wk of culture in complete T cell medium containing geneticin. Where indicated, CD8+ T cells were purified from splenocytes using a mouse CD8+ T cell subset enrichment column (R&D Systems) and used as effector cells. The percentage of specific 51Cr release was calculated as: 100 × (experimental release − spontaneous release)/(maximum release − spontaneous release). Maximum release was determined from supernatants of cells that were lysed by addition of 5% Triton X-100. Spontaneous release was determined from target cells incubated without added effector cells.
A total of 1 × 106 15-12RM cells in 200 μl of PBS were injected s.c. on the right flank of the mouse. Where indicated, 1 × 107 PFU of vPE16 or vSC8 were injected i.v. at 5 wk and 7 days before tumor inoculation. Some mice were treated i.p. with 0.2 ml of PBS containing either 0.5 mg of anti-CD4 mAb (clone GK1.5), anti-CD8 mAb (clone 2.43), or control rat IgG (ICN Pharmaceuticals, Costa Mesa, CA) starting 3 days before tumor cell injection and then twice a week.
Detection of mRNA by RT-PCR and Northern blot analysis
Total RNA was extracted from 5 × 106 tumor cells in Trizol reagent (Life Technologies). cDNAs were synthesized by extension of oligo(dT) primers using the Super Script Preamplification System (Life Technologies), according to the manufacturer’s instructions. PCR of the cDNA was performed in a final volume of 50 μl containing each primer at 0.2 μM and Supermixture (Life Technologies) using the GeneAmp 9700 PCR system (Perkin-Elmer, Norwalk, CT). The amplification cycles were 94°C for 30 s, 55°C for 1 min, and 72°C for 1 min. After 25 cycles, PCR products were separated by 10% TBE gel electrophoresis and stained with Vista Green (Amersham, Arlington Heights, IL). The sequences of primers are as follows: hypoxanthine-guanine phosphoribosyl transferase (HPRT) (sense), 5′-GTTGGATACAGGCCAGACTTTGTTG-3′; HPRT (antisense), 3′-GAGGGTAGGCTGGCCTATGGCT-5′; gp160 v3 loop (sense), 5′-GCTGTTAAATGGCAGTCTAGC-3′; gp160 v3 loop (antisense), 3′-CGTTAGGAGTCCTCCCCTGGG-5′. Northern blot analysis was performed using NorthernMax and Strip-EZ RNA Kit (Ambion), according to the manufacturer’s instruction. Total cellular RNA was used, and the integrity of RNA was tested by electrophoresis in 2% agarose gel. A total of 15 μg of RNA was loaded per lane and transferred onto nylon membranes. After prehybridization, the 32P-labeled probe for IFN-γ was hybridized at 65°C overnight. Then the membrane was washed with low- and high-stringency solution. Autoradiography was conducted at −70°C for up to 4 days by using Kodak (Rochester, NY) BioMax-MS film. The same membrane was stripped and used to rehybridize with a GAPDH probe as an internal control.
18Neo and tumor cells were stained with FITC-conjugated anti-H-2Dd mAb to detect the expression of class I molecules. They were analyzed by FACScan using CellQuest software (both from Becton Dickinson, Mountain View, CA).
Tumor tissues were excised on the indicated days after inoculation of tumor cells, fixed in 10% v/v formalin solution, and processed for paraffin embedding. Sections were cut according to standard procedures and stained with hematoxylin and eosin.
15-12RM tumor regressed after initial growth and recurred in vivo
15-12RM cells were injected s.c. on the right flank at day 0. Tumors initially started growing within 5 days after inoculation, and they reached about 8–10 mm diameter at approximately day 7. The tumors then began to regress spontaneously and disappeared after ∼10–12 days. The growth rate decreased at this time, even in the occasional small tumors that remained. However, the mice in which tumors had regressed initially, even beyond the point of detection, developed tumors again between 20 and 30 days after inoculation. These did not regress in this second growing stage (Fig. 1, a and b).
Vaccinia viruses were tested to induce tumor-protective immune responses in vivo. Vaccinia virus vPE16, which expresses gp160 in infected cells (56), was injected i.v. twice with a 30-day interval, and tumor was inoculated 7 days after the second immunization. Though 15-12RM tumor cells injected into control vaccinia vSC8-immunized mice did not always behave identically to normal mice, they shared a similar pattern of development of tumors, including the three phases of initial growth, regression, and regrowth. In contrast to normal and vSC8-immunized mice, mice immunized with vaccinia virus vPE16 showed a different process for growth of tumors. They did not manifest the primary development of tumors, but the tumors appeared only late, after day 30, and behaved like the tumors seen in the recurrence phase of normal and vSC8-immunized mice, without first undergoing regression (Fig. 1 c). These vPE16-immune mice also had CD8+ CTL specific for HIV-1 gp160 peptide P18-IIIB (data not shown, and Refs. 38 and 48). Thus, immunization against gp160, inducing specific CTL, could protect mice from the initial growth of tumors, but not from the later development observed 30 days after inoculation of tumor cells.
Regression depends on the existence of CD8 T cells
From the above results, we hypothesized that CD8+ CTL might play a role in the initial regression of tumor. To investigate the role of CD4 and CD8 T cells in the regression and growth of tumor, nonimmunized mice were treated with Abs against CD4 or CD8 molecules. Flow cytometric analysis showed that >98% of CD4 and CD8 T cells were depleted by the treatment with the respective Abs (data not shown). In the anti-CD8 Ab-treated groups, tumors grew initially in the same time period as in control mice, but they continued growing without regression (Fig. 2,a). Their growth rate did not decrease at 10–12 days, when the tumors of control mice regressed. Although anti-CD4 Ab treatment did not result in any significant change in the process of growth and regression of tumors up to day 30 after the challenge with tumor cells, depletion of CD4 cells unexpectedly protected the mice from later regrowth of tumor (Fig. 2,a). When CD8 cells were depleted simultaneously with depletion of CD4 cells, the tumors continued to grow without regression, as in the anti-CD8 Ab-injected group (Fig. 2 b). Thus, CD8 cells were necessary for initial rejection of the tumor, and the depletion of CD4 cells somehow prevented the regrowth of tumors after regression. When anti-CD8 Ab were given to the CD4-depleted group only after regression (starting from day 21), tumors developed in only some of the mice, despite the fact that there were no CD8 cells (data not shown). This observation is consistent with the interpretation that, in anti-CD4 treated mice, 15-12RM tumor cells did not survive after the initial regression, so the absence of CD8 cells at the later time point was no longer relevant.
Histological examination confirmed infiltrating lymphocytes
Since it was suggested that the growth and regression of tumors were regulated by immunosurveillance, especially CD8 T cells, we performed a histological examination of tumors to see whether they were compatible with the observation described above. Specimens were harvested from 15-12 RM tumor cells injected into normal mice at different time points after inoculation. At day 9, when initial tumors were growing, focal sparse mixed cellular (chronic) inflammatory infiltrates were noted within the tumor mass. At day 13, when growth stopped, the number of infiltrating lymphocytes seen around tumor cells increased (Fig. 3), and, when there were no tumors on the surface of the flank at day 19, the lymphocytic response reached a maximal level (Fig. 3). These findings raise the possibility that these infiltrated lymphocytes contributed to the regression of tumors. In contrast, when tumor reappeared and grew again after regression, the recurrent tumors had a more spindle-shaped morphology and were relatively devoid of infiltrating lymphocytes at days 33 and 46 (Fig. 3). Thus, the histology was consistent with a lack of a cellular immune response to the recurrent tumors.
Recurrent tumors developing after regression were resistant to 15-12RM-specific CTL
CTL specific for 15-12 RM tumor could also lyse target cells presenting P18-IIIB from the v3 loop of the HIV-1 envelope (Fig. 4,b). Likewise, P18-IIIB-specific CTL, which were induced from a vPE16-immunized mouse by stimulation with 1 μM P18-IIIB-pulsed spleen cells, could recognize 15-12 RM tumor cells (Fig. 4,a). Therefore, we utilized CTL against both 15-12 RM tumor cells and P18-IIIB to determine the susceptibility of the tumor cells to cytotoxicity. Our objective was to examine how tumors recurred after initial regression, despite the rejection mediated by CD8 T cells or the immune protection against initial growth observed in vPE16-immunized mice. The possible explanations of this observation could be either alteration of tumor cells (escape variants) or induction of tolerance in responding T cells. To examine the former possibility, tumor cells were recovered from mice at different stages in vivo and used as target cells in the CTL assay. While the tumor cells that grew initially retained sensitivity for 15-12 RM CTL, tumor cells from the recurrent stage could not be lysed by the same CTL (Fig. 5, a and b). Both freshly isolated cells and cultured cells, which were selected by growth in culture medium containing geneticin (to kill cells that had lost the NeoR gene), showed the same results. As mentioned above, vPE16-immunized mice developed tumors late without an early growth and regression phase. Tumor cells from these animals were not killed by 15-12 RM CTL (Fig. 5,a). However, tumor cells from normal or vSC8-immunized mice treated with anti-CD8 mAb were killed by 15-12 RM CTL (Fig. 5 c). Moreover, 15-12RM-specific CTL could kill tumor cells harvested from vPE16 immunized, anti-CD4 plus anti-CD8 mAb-treated mice (data not shown).
To test the possibility that contamination of CTL-resistant tumor cells in the cultured 15-12RM clone was the origin of recurrent tumor in vivo, 15-12RM cells were subjected to killing once or twice in vitro by P18 CTL, and then surviving cells were used as target cells for the CTL assay after several days of culture. We hypothesized that, if there are very few resistant tumor cells in the tumor cell suspension used for injection, these preexisting resistant tumor cells should gradually become the majority of target cells after selection by killing of sensitive cells by CTL in vitro. As shown in Fig. 6, 15-12RM were lysed by P18 CTL, even after two rounds of selection, in three of four experiments. Moreover, in the one case when cells became resistant after killing by CTL, although they were lysed after first round killing at the same level as original 15-12RM, they showed <5% of lysis after the second round. Therefore, the data suggest that it is unlikely that preexisting resistant tumor cells grew after regression of susceptible tumor cells in vivo. In summary, 15-12 RM cells before the tumor regression caused by CD8 T cells were susceptible to lysis by 15-12 RM CTL, but they did not maintain this character when they recurred after initial regression. Although vPE16-immunized mice showed early protection against tumor growth instead of growth and regression of tumor, 15-12 RM tumors that grew out in these mice after day 30 were resistant to gp160-specific CTL.
Detection of mRNA encoding the gp160 v3 loop in resistant and nonresistant tumors
To explore the mechanism of tumor resistance to CTL lysis, we examined the expression of the class I molecule H-2Dd on the resistant tumor cells because P18-IIIB requires Dd to be presented to T cells. As shown in Fig. 7,a, resistant tumor cells harvested from mice expressed Dd at the same level as the original 15-12RM cells. Also, resistant tumor cells could be lysed by P18 CTL when they were pulsed with P18-IIIB or infected with vPE16 (Fig. 7,b). These results showed that class I expression on the resistant tumor cells is intact and that resistant tumor cells can process and present the endogenously expressed Ag normally. Therefore, we asked if they had lost expression of the Ag. It was impossible to detect the expression of gp160 on the 15-12RM cells by Western blot analysis and flow cytometry. Thus, we compared the expression of mRNA in resistant and nonresistant tumor cells by RT-PCR. Fig. 8,a shows that mRNA for the gp160 v3 loop was absent in the resistant tumor cells, while nonresistant tumor cells and the original 15-12RM cells retained clear message. Next, we examined if this mRNA defect resulted from the loss of the transfected gene or a mutation at the either of the primer regions of v3 loop. DNA was isolated from 5 × 106 tumor cells and amplified by PCR. There were the clear bands of amplified DNA by v3 loop primers for the resistant tumors, as well as nonresistant tumors (Fig. 8 b). These PCR products were cloned and then sequenced from three resistant tumors and two nonresistant tumors harvested from mice and the original 15-12 RM cells. There was neither deletion of genes nor point mutations in the v3 loop region (data not shown). These results suggested that the tumors developed resistance against CTLs for P18-IIIB by decreasing the expression of mRNA encoding the gp160 v3 loop.
Loss of tumor-specific CTL does not account for the tumor regrowth
As mentioned earlier, the regrowth of tumor could be due either to the development of escape variants of the tumor, or to loss of CTL activity when the tumors reappear. Although the evidence above pointed to the outgrowth of resistant variants of the tumor, we also wanted to examine the possibility of CTL loss. Mice with large recurrent tumors were examined on day 62 after tumor inoculation. Spleen cells were restimulated for 6 days with the original tumor cells 15-12RM as stimulators, and then tested in a lytic assay on both 15-12RM targets and on 18Neo BALB/c 3T3 fibroblasts either with no peptide as a negative control or pulsed with 1 μM peptide P18-IIIB (Fig. 9). As can be seen from the two mice shown, a very vigorous CTL response was still present, specific for both HIV-1 gp160 peptide and tumor cells, even after substantial outgrowth of recurrent tumor. Thus, tolerance induction cannot account for the observed outgrowth of tumor.
These results, combined with the observation mentioned earlier that mice treated with anti-CD4 from the start but then treated with anti-CD8 only from day 21, after tumor had regressed, still mostly did not grow recurrent tumor, suggest that the protective effect of anti-CD4 is to allow the complete regression of tumor, so that no tumor is left to regrow once CD8 cells are eliminated. In the CD4-intact mice, the tumor regression is incomplete, and the tumor recurs despite the continued presence of CTL. Thus, we wondered whether there was some qualitative functional difference in the CTL that can complete the elimination of tumor in the absence of CD4 cells but not in their presence.
Production of IFN-γ by CD8 cells contributes to protection against the regrowth of tumors in CD4-depleted mice
As described above, depletion of CD4 cells, including Th cells, led to the protection against regrowth of 15-12RM tumors after regression. Since previous literature suggested that Th1 cells played an important role in protection against viral infection and tumor development (53, 54, 59), we utilized IFN-γ KO mice and IL-4 KO mice as models of Th1/Th2 imbalance to examine the influence of these cytokines on tumor growth and recurrence and on the prevention of regrowth by anti-CD4 mAb injection. Though some of the IFN-γ KO mice rejected tumors after initial growth, tumors regrew within 20 days after the inoculation, statistically significantly more rapidly than the control wild-type mice (Fig. 10,b) (p < 0.005, wild type vs IFN-γ KO, Log-Rank test). On the other hand, surprisingly, IL-4 KO mice did not significantly differ from control wild-type mice with regard to the pattern of tumor development. They were not protected against regrowth of tumors as observed with anti-CD4 treatment, even though they had a skewed balance toward Th1 caused by the absence of IL-4 (Fig. 10,c compared with Fig. 10,a). When CD4 cells were depleted before tumor inoculation, all the mice showed better protection than mice injected with control rat IgG Ab (ICN Pharmaceuticals) in each group (Fig. 10, a–c). The results using IL-4 KO mice indicate that the complete loss of IL-4 and the great diminution of other Th2 cytokines that are in part dependent on IL-4 for their production (66, 67, 68, 69), does not mimic the effect of CD4 depletion and, therefore, does not explain the striking protection in CD4-depleted mice. However, since tumors could regrow in CD4-depleted IFN-γ KO mice (Fig. 10 b), these results suggested that IFN-γ produced by CD8 cells or NK cells played an important role in protection against regrowth of tumors in CD4-depleted mice.
To investigate further the mechanism of the protection produced by CD4 depletion, we purified CD8 T cells from splenocytes of untreated and CD4-depleted mice on day 14 after the inoculation of 15-12RM cells to compare the CTL activity. Tumors grown on the right flank initially disappeared at this time point in all the mice in both groups. As shown in Fig. 11,a, CD8+ T cells from CD4-depleted mice, without restimulation in vitro, could kill 15-12RM target cells, as well as 18Neo pulsed with 1 μM of P18-IIIB, while CD8+ cells from tumor-inoculated, normal mice could not kill 15-12RM cells without restimulation. Also, purified CD8 T cells from CD4-depleted mice injected with tumor could lyse 18Neo pulsed with 1 μM of P18-IIIB better than those from normal mice injected with tumor (about 15% vs 40% specific lysis at 80:1 E:T ratio). To further explore the functional difference between these CTL, we examined the expression of IFN-γ mRNA in these CD8+ T cells. Consistent with the results of the CTL assay, we could detect four times as much IFN-γ mRNA from CD8 T cells in CD4-depleted mice as from CD8+ cells in tumor-inoculated untreated mice (Fig. 11 b). These results indicated that the quality of CD8 T cells in CD4-depleted mice was different from that in normal mice with regard to IFN-γ mRNA expression, as well as CTL activity without restimulation, properties that may be important for the protection against recurrence of tumor.
In this study, we present a novel manipulable animal model of tumor immunosurveillance in which we can evaluate escape mechanisms of tumor cells from immunosurveillance, the role of CD8+ T cells in causing regression and preventing local recurrence, and the adverse role of CD4+ T cells in local recurrence. As a model of soft tissue tumor in vivo, we utilized the transfectant fibroblast cell line expressing myc and the K-ras point mutant along with HIV-1 gp160 as a tumor Ag. Although not a natural tumor Ag, this viral Ag is not unlike other viral Ags expressed in certain human tumors, such as human papillomavirus Ags in cervical carcinoma. This tumor has the histological appearance of and invades muscle tissue like a soft-tissue sarcoma. It may serve as a model of human tumors that express viral Ags, such as cervical carcinoma induced by human papillomavirus or lymphoid malignancies induced by HTLV-I. CTL induced against this tumor cell line, 15-12RM, could recognize target cells pulsed with an immunodominant CTL epitope peptide P18-IIIB of the gp160 v3 loop (38), as well as tumor cells themselves. Prevention of regression by CD8+ cell depletion in vivo indicated that the regression was CD8-dependent. However, in unmanipulated mice, the tumors virtually always recurred, despite the continued presence of CTL (Figs. 1 and 9). Thus, this system serves as a model of tumor immunosurveillance. In the clinical situation in humans, many tumors express potential tumor Ags, and many subclinical tumors may undergo this type of spontaneous regression, but only the ones that escape immunosurveillance and reach the recurrent stage, as in our model, are detected clinically. If we could understand what prevents immunosurveillance from achieving complete regression and allows tumor recurrence and later metastasis, we might be able to overcome this obstacle and prevent tumor recurrence. This understanding is also critical to designing optimal immunotherapy of cancer.
A potentially important clue to this problem came from the surprising effect of CD4 depletion. Depletion of CD4 T cells did not change the pattern of initial tumor growth and regression, but it inhibited recurrence of the tumor. How could one explain the prevention of recurrence of the tumor in CD4-depleted mice? In accordance with the fact that CTL induction is associated with a Th-1 type immune response, it has been shown that CD4 T cells that can secrete Th1-type cytokines have a beneficial role in protection against tumor development (53, 54). Likewise, we previously reported that IL-2 production by human peripheral lymphocytes in response to human papillomavirus Ags is inversely associated with disease status (55). McAdam et al. (60) showed that murine carcinoma cells transfected with IL-2 and IFN-γ were more likely to be rejected than parental cells when implanted in BALB/c mice. On the other hand, the shift from Th1-type to Th2-type cytokine production was found in progressive cancer patients (61, 62), and T cells harvested from tumor-bearing hosts produced only Th2-type cytokines when they were stimulated in vitro (63). In addition to these findings, Th2-type cytokines could even accelerate the experimental pulmonary metastasis of melanoma (49).
Therefore, we hypothesized that the shift to a Th2-type response to gp160 could be the cause of failure of complete tumor regression, allowing recurrence of tumors after regression. We used IFN-γ KO mice and IL-4 KO mice as a model of Th1/Th2 imbalance (64) to address this question. Although the IFN-γ KO mice could suppress the initial growth of tumor, these tumors finally developed earlier than those in control BALB/c mice, suggesting a role for IFN-γ. However, other data suggested a critical role for CD8 T cells: 1) vPE16-immunized mice that have P18-IIIB-specific CTL before inoculation with tumor did not have initial growth of tumors; 2) CD8-depleted mice could not inhibit tumor growth at all; 3) Recurrent tumors had all become resistant to lysis by CTL; and 4) We could induce a CTL response even from 15-12RM tumor-bearing IFN-γ KO mice by stimulation with P18-IIIB in vitro (data not shown). Thus, we concluded that both lytic activity by CD8 T cells and production of IFN-γ are necessary for the regression of the initial tumor in our model system.
Prior immunization with vaccinia virus vPE16, which induces gp160-specific CTL, could prevent the initial growth of tumor completely, but not its recurrence later. This result indicated that the increase of CTL precursors for P18-IIIB contributed to clearance of tumor cells, but could not eradicate them. In contrast, in CD4-depleted mice, initial growth and regression of tumors were still observed but the subsequent recurrence was prevented. Thus, the increase in CD8 T cell numbers alone cannot explain the benefit of CD4 depletion. When CD8 T cells were depleted by the injection of anti-CD8 mAb starting from day 21 after inoculation of tumor cells in mice treated with anti-CD4 mAb from the beginning, the tumors developed in only 25% of these CD4-depleted mice (data not shown). This result indicated that CD4-depleted mice might eliminate all of the tumor cells before the critical point for recurrence, so that once the tumor cells were gone, the CD8+ cells were no longer necessary. In contrast, in the presence of CD4 cells, the clearance of tumor by CD8 cells was incomplete, even though CTL remained present during tumor recurrence (Fig. 9).
However, unexpectedly, IL-4 KO mice did not mimic CD4-depleted mice in that they could not stop the development of tumor after regression even though they had a shift to Th1-type response in general (Fig. 10) (67, 65). IL-4 is necessary for the normal development of Th2 responses and production of other Th2 cytokines (66, 67, 68), although some production of IL-5 and IL-10 can still occur (69). The IL-4 KO mice thus indicate that IL-4 is not required for the CD4-mediated prevention of complete regression, and probably other Th2 cytokines, which are greatly diminished in the IL-4 KO mice, also do not account completely for the inhibitory effect of CD4 cells on the elimination of tumor. Therefore, although a Th1-type response could contribute to the rejection of initial tumors and a shift to a Th2-type response could interfere with this protection, even a substantial skewing toward Th1-type response by elimination of IL-4 was not sufficient to prevent recurrence of tumors after regression. CD4 depletion must accomplish more than just Th2 depletion.
Koeppen et al. (70) observed that anti-CD4 treatment of mice increased the frequency of rejection of an allogeneic tumor expressing a foreign class I MHC molecule. Rakhmilevich and North (71) showed that elimination of CD4 T cells augments the antitumor effect of IL-2 therapy in mice bearing an advanced sarcoma by releasing CD8 T cell-mediated immunity from T cell-mediated suppression. Martinotti et al. (72) reported that tumor infiltration by the CD8 T cells was inhibited by CD4 T cells, but the tumors were unusual in being transduced with the gene for IL-12. In that report, they postulated that CD4-mediated suppression is exerted on CD8 expansion and on the ability of CD8 T cells to infiltrate tumor nodules. However, we could observe infiltration of lymphocytes in the tumor tissue not only in CD4-depleted mice but also in untreated mice. Also, we found several lines of evidence for a qualitative difference in CD8 cells from CD4-depleted mice that appeared to contribute to more effective tumor elimination. First, CD8 T cells purified from splenocytes of CD4-depleted and tumor-injected mice had higher CTL activity specific for 15-12RM cells than CD8 T cells from tumor-inoculated CD4-intact mice, even without any stimulation in vitro. This result indicated that CD8 T cells from CD4-depleted mice were already activated to kill tumor cells efficiently in vivo. Second, this better lytic activity of CTL was correlated with the expression of IFN-γ mRNA. Thus, CD8+ T cells from CD4-depleted mice, which could secrete more IFN-γ, could play an important role in prevention of recurrence of tumor in our model system. We conclude that a qualitative alteration of CD8 T cells following depletion of CD4 T cells could account for protection from recurrence of tumor, whereas the reduction in Th2-type cytokines in IL-4 KO mice was not sufficient.
We previously reported the importance of the quality of CTL as well as the quantity of CTL for viral clearance, in a study in which high-avidity CTL specific for P18-IIIB could protect better against virus challenge than low-avidity CTL (48). We are now investigating whether the quality of CTL has any correlation with high- or low-avidity CTL and whether there is any relation between CD4 depletion and the appearance of escape variant tumor cells.
Even during the recurrence phase of new tumor growth, a CTL response was detected in spleen cells from these tumor-bearing mice after the stimulation with P18-IIIB-pulsed spleen cells or 15-12RM cells (Fig. 9). This observation indicated that the exhaustion or tolerization of responding cells against P18-IIIB was not the cause of tumor recurrence after regression. There are several possible remaining explanations for this escape mechanism of tumor cells from immunological destruction (19). Besides the host factors that can suppress immune defenses against tumor cells already discussed, there were several reports of Ag loss of tumor variants (17, 18, 20). Selection of Ag loss, epitope loss, or class I loss variants might lead to recurrence of tumor. In this study, recurrent tumor cells became resistant (resistant tumor cells) to CTL that could lyse both the original tumor cells and also the tumor cells recovered from the initial growth stage before regression (nonresistant tumor cells). The resistant tumor cells remained class I-positive at the same level as the original tumor cells by FACScan, and could process and present endogenous Ag, gp160, as shown by their ability to be lysed by P18-IIIB-specific CTL when infected with recombinant vaccinia virus expressing gp160. Since gp160 on 15-12RM cells could not be detected by Western blot and flow cytometric analyses, we investigated the presence of DNA and the expression of mRNA. We could detect amplified DNA of the v3 loop (containing P18-IIIB) of gp160 from resistant tumors as well as from the original 15-12RM and nonresistant tumors, and the DNA did not contain any point mutation at any coding position within or near the v3 loop. However, only resistant tumor cells lost the expression of mRNA for the v3 loop. Moreover, tumor cells recovered from CD8-depleted mice could be lysed by CTL specific for P18-IIIB. Thus, this acquisition of resistance against CTL occurred only in the presence of immune pressure by CD8 T cells. These resistant tumor cells could grow in normal BALB/c mice without regression and were not killed by CTL after several weeks of culture in G418-containing medium, which means that escape variants selected by CD8 T cells were stable both in vivo and in vitro, in contrast to the situation in another report (73). Our results make contamination of escape variants in the original 15-12RM cells unlikely. In vivo, the tumor is removed from the G418-containing selection medium used to select for neomycin-resistant cells. However, this removal is unlikely to play a role in the escape for several reasons. First, the original 15-12 cells were made by cotransfection with NeoR and gp160 genes on different plasmids, so that NeoR could be used to select for cells that took up DNA in the original transfection, but selective pressure from G418 should not affect retention of the gp160 gene (38). Second, the original 15-12 transfectants continue to express gp160 for at least 3 mo in culture without G418. Third, the failure to develop resistant tumors in the CD8-depleted mice (in the absence of G418), as just mentioned, implies that the selection for resistance requires CD8-mediated immune pressure. Fourth, the gene for gp160 is retained, but just the mRNA expression is lost. Thus, it is very important to know how CTL pressure can cause the decrease of mRNA expression of the epitope region in tumor cells to allow escape from immunological surveillance.
In conclusion, this model has allowed us to begin to dissect some of the mechanisms mediating and regulating tumor immunosurveillance. CD8+ CTL appear to be critical for causing tumor regression, but quantity of CTL alone is not sufficient. Rather, qualitatively different CTL that produce more IFN-γ and remain activated in vivo may be critical. The qualitative difference in CTL is influenced by CD4+ cells that regulate the CD8+ response, but this regulation cannot be explained simply by the Th1/Th2 balance. Further studies to determine the mechanism of this regulation will be important for designing optimal immunotherapy. It may be valuable to induce CTL producing high amounts of IFN-γ, as well as having high avidity, to obtain good quality CTL for immunotherapy. Such CTL could be a useful component of a strategy to prevent escape variants of tumor cells and to prevent recurrence of tumors to control malignant disease. Concurrent abrogation of the inhibitory effects of CD4 cells without eliminating IFN-γ production may provide a successful concerted approach to cancer immunotherapy.
We thank Drs. Gene Shearer and Alan Sher for critical reading of the manuscript and helpful suggestions; Drs. Igor Belyakov, Richard Bamford, and Felicita Hornung for discussions and technical advice; and Terri Emerson for help during preparation of the manuscript.
Abbreviations used in this paper: KO, knockout; HPRT, hypoxanthine-guanine phosphoribosyl transferase.