The Mycobacterium bovis bacillus Calmette-Guérin cell-wall skeleton (BCG-CWS) activates Toll-like receptor (TLR) 2 and TLR4, but unlike the typical TLR4 agonist bacterial lipopolysaccharide barely induces type 1 IFN. BCG-CWS has been used for adjuvant immunotherapy for patients with cancer. We investigated the adjuvant potential of BCG-CWS for induction of CTLs subsequent to TLR-mediated dendritic cell (DC) maturation, using a syngeneic mouse tumor model (B16 melanoma in C57BL/6). We evaluated the retardation of tumor growth and cytotoxic response in wild-type and MyD88−/− mice immunized with tumor debris and/or BCG-CWS. Delays in tumor growth and cytotoxic response were induced by immunization with a mixture of BCG-CWS emulsion and the tumor. BCG-CWS was capable of activating DCs ex vivo by the criteria of CD80/CD86 up-regulation and cytokine (interleukin-12, tumor necrosis factor-α) induction. Efficient tumor suppression and ex vivo cytokine induction did not occur in MyD88-deficient mice and cells, suggesting that the MyD88 adapter is crucial for induction of tumor cytotoxicity. Because TLR4 is involved in both MyD88-dependent and -independent pathways and the latter affects DC maturation, our findings indicate that both pathways cooperate to induce CTL-based tumor immunity.
Microbial components that activate the host immune system have been designated as adjuvants. Adjuvants have often been used for immunization with pure antigen for potential induction of antibody production and CTL and natural killer (NK) cell activation (1 , 2) . Variations in the output responses appear to depend on the properties of each adjuvant and the target immunocompetent cells. One representative adjuvant is dead mycobacteria conjugated with mineral oil, which is called Freund’s complete adjuvant (3) .
Because cancers become established and clinically detectable presumably by circumventing the host immune surveillance, tumor cells generally possess poor immunogenicity by themselves (4 , 5) . Enhancing host in vivo immunity and/or increasing tumor antigenicity has been a goal in the design of immunotherapy. Selective manipulation of immune cells, particularly dendritic cells (DCs), has been attempted for immunotherapy with vectors and reagents (6) . Although DCs sensitized with a targeting antigen migrate to lymphoid tissues and induce a strong and efficient T-cell response, this tailored therapy requires cell purification and culture. Furthermore, manipulation of patient tumor cells or DCs and the identification of CTL-defined tumor-associated antigens are highly crucial for the routine application of this method to patients. In this regard, cell-free vaccines would be more suitable for clinical purposes. Adjuvants should be an alternative tool for tuning up host immunocompetent cells for cancer immunotherapy.
The role of adjuvants in effective immune potentiation had not been identified at the molecular level until Toll-like receptor (TLR) was discovered in mammals (7) . TLR is a receptor family consisting of >10 members in humans and mice (8 , 9) . At present, the evidence is accumulating that each TLR is a receptor for a specific adjuvant. Adjuvants, here named pathogen-associated molecular patterns based on the nature of their receptors (10) , have been found to interact with professional antigen-presenting cells (11 , 12) , including DCs, via TLRs on their membranes.
We have conducted immunoadjuvant therapy by s.c. administration of agonists of TLRs. Bacillus Calmette-Guérin cell-wall skeleton (BCG-CWS), which has been used as a potent adjuvant therapy in patients with cancer, has been identified as an agonist of TLR2 and TLR4 (13 , 14) . BCG-CWS exerted tumor regression activity within a dose that demonstrates no toxicity. The application trial of BCG-CWS to >600 patients with postoperative cancer largely (>60%) brought about good prognosis for the patients in our hospital (15 , 16) . In ex vivo analysis of the patients’ blood mononuclear cells, IFN-γ was produced in response to exogenously added BCG-CWS in most of the patients with good prognosis (15, 16, 17) . The mechanism of BCG-CWS-mediated host immune activation is at least in part attributable to the maturation of DCs, which is induced through TLR2 and TLR4 (13 , 14) and putative BCG-CWS uptake receptors (16) on DCs prepared from blood from normal volunteers. The costimulators CD80/CD83/CD86 and the cytokines tumor necrosis factor (TNF)-α and interleukin (IL)-12 p40 are up-regulated in human DCs by experimental stimulation with BCG-CWS (13) . None of the lymphocyte populations (NK, NKT, B, and T cells) are directly activated in response to BCG-CWS (13 , 17) . Hence, BCG-CWS acts as a potent inducer of ex vivo DC maturation via its TLR agonist activity. However, the mechanisms by which BCG-CWS induces in vivo effector activation, including IFN-γ production and CTL response to tumor-associated antigens, and the resultant suppression of tumor growth have not been identified.
TLR signaling pathways, on the other hand, are being elucidated. The cytoplasmic domain of each TLR recruits distinct sets of adapter molecules that in turn activate specific downstream signaling molecules (8, 9, 10, 11) . To date, four adapters have been identified, and selection of the adapters appears to determine the particular TLR signaling pathway leading to the activation of specific TLR-defined transcription factors such as nuclear factor-κB, c-Jun (AP-1), or IRF-3 (8 , 9 , 18) . MyD88 is a pivotal adapter that activates nuclear factor-κB, leading to induction of the cytokines TNF-α, IL-6, IL-8, and IL-12 (19 , 20) . However, the up-regulation of costimulators and induction of IFN-β are largely independent of MyD88 (21 , 22) . Lipopolysaccharide (LPS)-stimulated TLR4 reportedly activates both MyD88-dependent and -independent pathways (21 , 22) . BCG-CWS, despite acting as an agonist of TLR4 (13) , poorly induces LPS-like MyD88-independent responses. The molecular mechanism whereby BCG-CWS serves as a TLR4 agonist with properties distinct from those of LPS and has the potential for exerting antitumor immunity also remains unresolved.
Here we show evidence, using MyD88-deficient mice, that MyD88 is a critical adapter for induction of tumor-specific cytotoxicity and subsequent active immunity for tumor suppression by BCG-CWS. These results provide insight into the mechanism of the BCG-based antitumor potential (23) and may be useful for testing adjuvant immunotherapies for cancer presently under study.
MATERIALS AND METHODS
BCG-CWS was prepared in our laboratory as described previously (24) . The lot used in this study (Lot 10-2) consisted of mycolic acid, arabinogalactan, and peptidoglycan with >97% purity and no LPS contamination: the results of chemical analysis of this lot were published previously (14 , 24) . Minimal amounts of phospholipid (∼0.2%) and amino acids (<2%) contaminated this preparation. No mannan or glucose was detected. Because BCG-CWS is insoluble in water and organic solvents, oil-in-water emulsion forms of BCG-CWS micelles (BCG emulsion) were used throughout the in vivo study. Dried BCG-CWS was resuspended at a concentration of 1 mg/ml in emulsion buffer (PBS containing 1% drakeol and 1% Tween 80) with a Potter homogenizer and sterilized by heating for 30 min at 60°C (24) . In some in vitro experiments, we also used BCG-CWS that was homogenized in PBS without oil or solubilizer (BCG-PBS), which acted more potently on DCs in culture than the BCG emulsion.
Mice and Cell Lines.
Breeding pairs of MyD88−/− mice were provided as reported previously (21) . Wild-type male and female C57BL/6 mice were purchased from Japan Clea (Tokyo, Japan). Mice were maintained in our institute under specific pathogen-free conditions. All animal experiments were approved by the committee in our institute.
B16D8 was established in our laboratory as a subline of the B16 melanoma cell line (25) . This subline was characterized by its low or lack of metastatic properties when injected s.c. into syngeneic C57BL/6 mice (25) . Mouse cell lines 3LL, EL-4 (C57BL/6 origin), Colon-26 (BALB/c origin), and YAC-1 (BALB/c origin) were provided by Sumitomo Co. Ltd., as described previously (26) . These cell lines were cultured in RPMI 1640 containing 10% FCS.
Reagents and Antibodies.
The following materials were obtained as indicated: Fetal bovine serum was from Bio Whittaker (Walkersville, MD), mouse granulocyte-macrophage colony-stimulating factor (GM-CSF) and mouse IL-2 were from PeproTech EC, Ltd. (London, United Kingdom), polymyxin B and LPS (Escherichia coli O111:B4) were from Sigma (St. Louis, MO), [51Cr]sodium chromate was from Amersham Biosciences (Piscataway, NJ), and Lympholyte-M was from Cedarlane (Ontario, Canada). The ELISA kits for IL-12 p40 and TNF-α were from Amersham Biosciences.
The following antibodies were used: FITC-conjugated goat antirat, rabbit, mouse IgG F(ab′)2 from American Qualex (San Clemente, CA), antimouse CD40 (phycoerythrin labeled), antimouse CD80 (FITC), antimouse CD86 (FITC) and isotype control antibodies [American hamster IgG (FITC), rat IgG2a (FITC), rat IgG2a (phycoerythrin), and rat IgG1 (FITC) from eBioscience (San Diego, CA)], antimouse CD8 (FITC) and antimouse CD4 (phycoerythrin) from Immunotech (Marseille, France), anti-H-2Kb (FITC) and anti-H2Db (FITC) from Cedarlane, and anti-Qa-Ib from PharMingen (San Diego, CA).
A peptide of melanocyte differentiation antigen tyrosinase-related protein-2 [SVYDFFVWL; TRP-2 (180–188)] was obtained from Biologica Co. (Aichi, Japan). TRP-2 (180–188) contains the epitope for rejection of B16 melanoma (27) and is mounted on H-2Kb.
Flow-Activated Cell-Sorting (FACS) Cytometric Analysis of Cell Surface Antigens and ELISA.
The practical methods were described previously (13) . Briefly, for FACS analysis, cells were suspended in PBS containing 0.1% sodium azide and 1% FCS and then incubated for 30 min at 4°C with fluorescently labeled monoclonal antibodies. Cells were washed, and the fluorescence intensity was measured by FACS.
For ELISA, culture supernatants of DCs were collected after removal of insoluble material by centrifugation and stored at −30°C. The levels of IL-12 p40 and TNF-α were measured by commercially available ELISAs.
Bone Marrow-Derived DCs (BM-DCs) of Mice.
BM-DCs were prepared by a reported method (21 , 28) with minor modifications. Briefly, BM cells were cultured overnight in 24-well plates at 0.5–1 × 106 cells/2 ml/well in RPMI 1640 containing 50 μm 2-mercaptoethanol, 10 mm HEPES, and 10% FCS. Nonadherent cells were harvested, resuspended in the same medium supplemented with 10 ng/ml GM-CSF, and cultured in the GM-CSF-containing medium. On day 3, adherent cells were cultured in fresh medium with 10 ng/ml GM-CSF. On day 6, nonadherent cells and loosely adherent cells were harvested and used for experiments as immature DCs. Immature DCs were resuspended in fresh RPMI 1640 containing 10 ng/ml GM-CSF and cultured for 24 h for ELISA or 48 h for FACS. To exclude the possible effect of contaminating LPS, BCG-CWS and macrophage-activating lipopeptide (MALP)-2 were pretreated with polymyxin B at 37°C for 60 min (14) . The stimulating reagents were then added to the culture medium of immature DCs as indicated in the text (final concentration, 15 μg/ml BCG-CWS, 100 ng/ml LPS, 100 nm MALP-2, 10 μg/ml polymyxin B).
For ex vivo DC stimulation, we used BCG-PBS in place of BCG emulsion because BCG emulsion was not suitable for culture cell stimulation because of its micelle formation (data not shown), which interferes with the easy access of BCG-CWS to cells.
Immunization and Tumor Challenge.
On days −28, −21, −14, −1, and +7 relative to the day of B16D8 challenge (Fig. 1B) ⇓ , 5 × 105 B16D8 cells (in 10 μl) were irradiated in PBS to prepare “debris,” and the debris was mixed with 20 μl of 1 mg/ml BCG-CWS in emulsion buffer (BCG-emulsion-tumor). Wild-type and MyD88−/− mice each received s.c. immunizations containing 30 μl of this mixture at the base of the tail. The administration protocol is shown in Fig. 1B ⇓ . As controls, tumor debris only or emulsion only, but not BCG-CWS in emulsion buffer (BCG emulsion), was used for the reason described in the “Discussion.” We also checked the activity of this reagent for induction of tumor cytotoxicity. At tumor challenge, C57BL/6 mice were shaved at the flank and received s.c. injections of 300 μl of 6 × 105 syngeneic B16D8 melanoma cells in PBS. After 3 weeks, tumor volumes were measured at regular intervals by a caliper. The mouse 3LL cell line was used as an irrelevant control tumor. Tumor volume was calculated using the formula: tumor volume (cm3) = long diameter (cm) × short diameter (cm) × short diameter (cm) × 0.4.
Generation of Tumor-Specific CTLs by in Vitro Tumor Stimulation.
Immunized mice were sacrificed on day 0 (Fig. 1B) ⇓ , and lymph node cells were isolated by use of Lympholyte-M. Lymph node cells (5 × 106) were cultured with 1 × 105 irradiated (160 Gy) B16D8 cells (27 , 29) , pretreated with or without 100 units/ml IFN-γ for 24 h, in a 24-well culture plate in 2 ml of RPMI 1640 supplemented with 50 μm 2-mercaptoethanol, 10 mm HEPES, and 10% FCS. Mouse IL-2 was not added except where indicated. After 5 days, the cytolytic activity of tumor cells was tested with the cultured lymph node cells. In some cases, the lymph node cells were restimulated with irradiated B16D8 cells and 20 units/ml mouse IL-2.
Tumor-specific cytotoxic activity was enhanced by an alternative method. Splenocytes from naïve C57BL/6 mice were homogenized, incubated with 100 μg/ml of the TRP-2 (180–188) peptide for 4 h at 37°C, and irradiated with 30 Gy. For in vitro stimulation, these peptide-pulsed, irradiated splenocytes (2 × 106 cells) were mixed with lymph node cells or splenocytes (2 × 106 cells) isolated from BCG-emulsion-tumor-immunized mice in 2 ml of medium (see above) and cultured for 5 days at 37°C. Repetitive restimulation was performed an additional four times with these cells, and effector cells were prepared with use of Lympholyte-M. The cytotoxic activity of the effector cells toward B16 melanoma cells was evaluated by the 51Cr release assay (29) .
Assessment of in Vitro Cytolytic Activity.
Target cells were labeled with 51Cr for 3 h at 37°C, then washed and coincubated with effector cells at the indicated lymphocyte-to-target cell ratio in V-bottomed 96-well plates in a total volume of 200 μl of RPMI 1640. Cytotoxicity was determined by measuring the 51Cr radioactivity released in 100 μl of the supernatant harvested from the plate after 8 h of incubation at 37°C (29) . The percentage of specific lysis was calculated using the formula: specific lysis (%) = [(experimental release − spontaneous release)/(total release − spontaneous release)] × 100.
Establishment of Mouse Model for Evaluation of Adjuvant Potential.
The B16 melanoma cell line (1 × 103−3 × 106 cells) was implanted in a syngeneic C57BL/6 host to determine an appropriate tumor burden. The dose of 6 × 105 cells/8-week-old female mouse was determined to be appropriate based on the injection of different numbers of cells s.c. into wild-type mice (not shown). This dose yielded 100% tumor manifestation within 4 weeks from the time of tumor injection in mice, with 93% survival of the mice for >6 weeks. We next examined immunotherapy protocol conditions with various doses of tumor debris and/or oil-in-water emulsion containing BCG-CWS (BCG emulsion; Fig. 1, A and B ⇓ ). Emulsion buffer or BCG emulsion only was used as a control. These controls elicited minimal tumor regression in mice (Fig. 1A) ⇓ . We decided to use emulsion or BCG emulsion as a control for subsequent experiments. In initial trials, we found that four repeats of immunization with B16 tumor debris conjugated with BCG emulsion (BCG-emulsion-tumor) were required for suppression of tumor growth in mice: three immunizations with BCG-emulsion-tumor or more than four immunizations with tumor debris or emulsion alone was insufficient for suppression of tumor progression (data not shown).
A large-scale study was then performed according to the protocol established (Fig. 1B) ⇓ . The tumor indices in each group are shown in Fig. 1C ⇓ . Tumor cells grew progressively in mice treated with emulsion buffer (n = 30). Tumors grew similarly when tumor debris alone was used instead of emulsion (data not shown). Compared with the control, the growth kinetics of the tumor were significantly retarded in mice immunized with BCG-emulsion-tumor (n = 30; Fig. 1C ⇓ ). In the mice treated with emulsion control and BCG-emulsion-tumor, 30 of 30 and 18 of 30 died, respectively, of tumor progression 7 weeks after tumor inoculation.
The levels of IFN-γ in mice treated with tumor debris, BCG emulsion, or both were measured by ELISA (Fig. 1D) ⇓ . IFN-γ was detected in mouse serum at low levels in the BCG-emulsion-only group and at high levels in the BCG-emulsion-tumor group. No IFN-γ was detected in mice treated with saline or tumor debris only. These results to some extent resemble those observed in BCG-CWS-treated patients with cancer (15 , 16) .
Mice vaccinated with B16 debris in BCG emulsion showed progression of implanted 3LL cells, an irrelevant syngeneic tumor line (Fig. 1E) ⇓ . In mice preimmunized with BCG emulsion, specificity appeared to be exerted on the tumor species initially immunized as antigens.
Antitumor Response in Wild-Type and MyD88-Deficient Mice.
TLR2 and TLR4 activate nuclear factor-κB and p38 mitogen-activated protein kinase through two adapters, Mal/TIRAP and MyD88 (30 , 31) . MyD88 is the relevant effector because it directly binds IRAK family proteins via their death domains (20) . Using the same protocol, we examined tumor growth in age-matched wild-type (n = 8) versus MyD88−/− mice (n = 8). The reduction in tumor growth achieved by immunization with BCG-emulsion-tumor was almost abolished in MyD88−/− mice (Fig. 2A) ⇓ . The tumor sizes of the wild-type mice were significantly smaller than those of MyD88−/− mice, with the effect being evident at early time points. Tumor growth was delayed in wild-type mice immunized with BCG-emulsion-tumor compared with the MyD88−/− mice, in which tumor grew irrespective of immunization (Fig. 2B) ⇓ . All MyD88−/− mice died within 6 weeks after tumor challenge (Fig. 2C) ⇓ , although 50% of the wild-type mice were surviving at the 6-week time point. It is notable that in wild-type mice, delayed hypersensitivity-like skin reactions developed only in the group immunized with BCG-emulsion-tumor, whereas no skin lesions were observed in MyD88−/− mice.
Properties of DCs Matured in Response to BCG-CWS.
We then analyzed the surface expression of costimulatory molecules and the responses to cytokines of BM-DCs prepared from wild-type and MyD88−/− mice and of BM-DCs that were stimulated with BCG-CWS (Fig. 3) ⇓ . LPS and MALP-2, representative ligands for TLR4 and TLR2, respectively (21 , 32) , were used as control DC maturation inducers. The appropriate doses of these TLR stimulators were determined by the ability to induce IL-12 p40, and 15 μg/ml BCG-CWS was found to functionally correspond to 100 nm MALP-2 and 100 ng/ml LPS. Wild-type BM-DCs responded to LPS and MALP-2 by showing up-regulation of surface CD40 (not shown), CD80, and CD86 (Fig. 3 ⇓ , left panel). BCG-CWS enhanced the surface expression of these maturation markers similarly to the TLR2 and TLR4 stimulators.
The effects of the TLR stimulators on MyD88−/− BM-DCs were next tested (Fig. 3 ⇓ , right panel). LPS, but not MALP-2, induced costimulator up-regulation in MyD88−/− cells with a FACS profile similar to that of MyD88+/+ cells. Enhancement of the costimulator levels was also induced by BCG-PBS in MyD88−/− BM-DCs as well as in wild-type BM-DCs.
In general, costimulators were induced on the DC surface by stimulation with either LPS or MALP-2. In MyD88−/− cells, only the MALP-2-mediated DC maturation was abrogated. These profiles are in accord with previous findings (30 , 31) that MyD88 is the only adapter that governs TLR2-dependent DC maturation, whereas TLR4 additionally activates a MyD88-independent pathway that may participate in DC maturation. Because BCG-CWS up-regulated DC maturation markers even in MyD88−/− cells, BCG-CWS-mediated activation of TLR2 and TLR4 appears to induce BM-DC maturation via the MyD88-dependent and -independent pathways. If this is the case, MyD88 and other adapters may participate in functional maturation of BM-DCs. Although the outputs are different between BCG-CWS- and LPS-stimulated DCs, particularly with respect to induction of IFN-inducible genes (33) , both of these stimulators rely on either MyD88 or an alternative adapter for the DC maturation signal.
The cytokine production profiles of BCG-CWS-stimulated BM-DCs are shown in Fig. 4 ⇓ ; LPS was used as a positive control. BCG-PBS induced IL-12 p40 and TNF-α production in wild-type BM-DCs but not MyD88−/− DCs (Fig. 4) ⇓ . Hence, MyD88−/− cells lose the ability to produce cytokines but retain the ability to up-regulate DC maturation markers in response to BCG-CWS. This difference was also observed in LPS-stimulated wild-type versus MyD88−/− DCs, which is consistent with a finding reported by Kaisho et al. (21) . Addition of IL-4 to the cells barely affected this tendency. Of note, expected levels of IL-12 p40 and lower levels of TNF-α were detected in LPS-treated BM-DCs compared with BCG-PBS-treated DCs. In conclusion, BCG-CWS induces up-regulation of costimulators but fails to induce TNF-α and IL-12 p40 in MyD88−/− DCs.
Tumor-Specific CTL Induction in Wild-Type and MyD88-Deficient Mice.
The BCG-CWS-based vaccines elicited cytotoxic responses, presumably CTLs, against B16 melanoma cells (Fig. 5) ⇓ . To test the relationship between the BCG-CWS-based cytotoxic response and antitumor potential, lymph node cells were recovered from individual wild-type mice administered BCG-emulsion-tumor (B16 cells) and restimulated with irradiated B16 cells, and their ability to lyse B16 targets was assessed. For logistical reasons, it was not possible to quantify the CTL response in all mice from each experiment. A total of 12 mice in two experiments were tested according to the protocol (Fig. 1B) ⇓ .
We measured the number of draining lymph node cells from wild-type mice to evaluate the effect of immunization in vivo. The number of lymph node cells recovered was increased in response to BCG-CWS immunization (Fig. 5A) ⇓ , although the number of spleen cells was not affected (not shown). Lymph node cells from mice immunized with tumor debris only, BCG emulsion only, or BCG-emulsion-tumor were stimulated in vitro one round with tumor debris and stained with CD8 for FACS analysis (Fig. 5B ⇓ , left panel). CD8+ cells were augmented only in lymph node cells stimulated with BCG-emulsion-tumor in response to the in vitro tumor challenge. No cytolytic activity against B16 tumor cells was found in lymph node cells stimulated with BCG emulsion only (Fig. 5, C and D) ⇓ . After a second round of in vitro tumor challenge, CD8+ T cells in lymph node cells of the BCG-emulsion-tumor group were further increased, whereas CD4+ cells did not expand (Fig. 5B ⇓ , right panel). However, CD8+ T cells were only marginally expanded in similarly treated MyD88−/− cells (data not shown).
A significant cytotoxic response was detected in lymph node cells from wild-type mice challenged with BCG-emulsion-tumor after the first round of in vitro stimulation with tumor (Fig. 5C ⇓ , left panel). In contrast, only a minimal cytotoxic response was detected in lymph node cells from MyD88−/− mice preimmunized with BCG-emulsion-tumor (Fig. 5C ⇓ , right panel). No other wild-type or MyD88−/− groups pretreated with either BCG emulsion or tumor debris exhibited a cytolytic response (Fig. 5C) ⇓ . The results paralleled those of tumor regression in wild-type versus MyD88-deficient mice. Thus, tumor immunity appeared to be linked with cytotoxicity in conjunction with signaling via MyD88.
We next examined whether the BCG-CWS-induced tumor lysis was specific to the tumor, an initially challenged subline of B16 melanoma. The lymph node cells harvested from the four groups of wild-type mice immunized with BCG emulsion and/or B16 tumor debris were treated as in Fig, 5C. The cytolytic response toward various types of syngeneic tumor cells was measured in these four groups (Fig. 5D) ⇓ . A significant response was detected only in the group stimulated with BCG-emulsion-tumor. Cytolysis was enhanced if IFN-γ was added to the B16 target cells (Fig. 5D) ⇓ . The cytotoxicity was directed against B16, but not Colon 26 or YAC1 cells. Thus, the BCG-CWS-induced tumor lysis is likely to be CTL dependent. CTLs induced by BM-DCs via TLR-MyD88 signaling thus appear to be specific for the antigens used for the initial sensitization.
To show tumor reactivity, specificity, and MHC restriction of the CTLs more clearly, we tried to enrich the CTL population specific to B16 melanoma by repetitive restimulation of lymph node cells from mice immunized with B16 debris. However, when we repetitively stimulated lymph node cells from wild-type mice immunized with BCG-emulsion-tumor, harvested lymphocytes largely died, as reported previously (34) . Thus, we used an alternative way to expand the CTL population toward B16 melanoma. Finally, we obtained a CTL-rich fraction through peptide-pulsed restimulation: lymph node cells from immunized mice restimulated were four times with splenocytes pulsed with TRP-2 (180–188). CTLs against B16 melanoma proliferated and exhibited robust cytotoxic activity (Fig. 5E) ⇓ . Thus, initially immunized tumor confers specificity on CTLs in terms of proliferation and cytotoxicity.
In the present study using a mouse syngeneic model and MyD88−/− mice, we demonstrated that (a) immunizing a tumor with BCG-CWS induces effective tumoricidal response; (b) the response is specific to the immunized tumor species, suggestive of CTL induction; (c) BCG-CWS-mediated TLR2/4 stimulation leads to the induction of both DC maturation markers and cytokines (IL-12 p40 and TNF-α), only the latter being impaired in MyD88−/− cells; and (d) BCG-CWS-dependent CTL induction and tumor regression are abolished in MyD88−/− mice. Hence, MyD88-dependent cellular responses involve BCG-CWS-mediated cytokine production and tumor cytotoxicity induced by DCs. The results may reflect the finding that TLR2 and TLR4 share the same adapter, MyD88 (30 , 31) , which is activated by BCG-CWS.
BCG in emulsion buffer (BCG emulsion) elicited minimal regression of the control tumor in mice (Fig. 1A) ⇓ , which was inconsistent with the nature of this adjuvant in humans (15) . Because we have kept the mice under high specific pathogen-free conditions, the mice may not be sensitized to human pathogens. At present, we consider that most Japanese are vaccinated with BCG, leading to robust response to BCG emulsion and eliciting antitumor immunity. Other reasons, including differences between the human and mouse immune systems or the properties of the tumors implanted, may have led to this discrepancy. In most experiments, we used emulsion-only as a control.
The first and second of our findings provide molecular-based evidence for adjuvant activity that corroborate previous experimental findings about BCG-mediated antitumor immune potential (15 , 16) . Human DC maturation is induced by BCG-CWS in a TLR2/4-dependent manner (13 , 14) . Immunizing tumor debris with BCG adjuvants subsequently induced a specific cytolytic response to the immunized tumor species (Fig. 5) ⇓ . These findings suggest that CTLs are responsible for the antitumor cytotoxicity induced by BCG-CWS. Final confirmation of this issue, however, will be needed to show the parallelism between CD8+ T-cell depletion and loss of antitumor responses in vivo. CTL induction toward specific tumors has not been experimentally verified in the BCG therapy. We therefore first demonstrated that TLR signaling in DCs participates in inducing tumor-specific cytotoxicity, most likely reflecting CTLs.
The third and fourth findings demonstrate that the TLR adapter MyD88 plays a key role in DC-mediated CTLs. TLR2 recruits MyD88, whereas TLR4 recruits other adapters in addition to MyD88. Hence, DC maturation is supported by MyD88 in TLR2 and by both the MyD88-dependent and -independent pathway in TLR4. Because BCG-CWS is a ligand for TLR2/4 (13) , our interpretation is that the surface markers of DC maturation are up-regulated by either the MyD88-dependent or -independent pathway in BCG-CWS-stimulated cells. The BCG-CWS-mediated DC maturation should be crucial for tumor-specific CTL induction.
The pathways sustaining CTL response appear to differ from those supporting the allostimulatory mixed lymphocyte response. The mixed lymphocyte response is provoked even in MyD88−/− DCs if they are stimulated with LPS or BCG-CWS (Ref. 21 ). 6 For CTL induction, in contrast, MyD88-dependent cellular responses are essential in addition to MyD88-independent responses. In line with this, TLR4-mediated pathways including MyD88 and other adapters (35 , 36) are important for induction of the LPS-mediated lymphoproliferative response. Therefore, activation of TLR4 by BCG-CWS in MyD88-dependent and -independent manners would be essential for CTL induction. Alternatively, unidentified receptors for the uptake of BCG-CWS (14 , 16) may participate in the observed antitumor response.
What molecule is responsible for the MyD88-independent DC activation response is the next question to be addressed. It has been accepted that at least four adapters, MyD88, Mal/TIRAP, TICAM-1, and TICAM-2, are linked to TLR4 to deliver signals leading to the activation of nuclear factor-κB, c-Jun (AP-1), and IFN-β (18 , 30 , 31 , 36 , 37) . These signaling pathways are known to mature DC in different ways and stages. IFN-β expression is a major outcome of the MyD88-independent pathway in TLR3 and leads to a unique DC maturation via the TICAM-1 adapter (37) . TICAM-1 recruited to the TLR4-TICAM-2 complex was identified as an effective adapter in TLR4-mediated IFN-β promoter activation (36 , 38) . Unexpectedly, however, BCG-CWS activates TLR4, but no IFN-inducible genes are induced (33) . As reported recently, IFN type 1 (a main product of the TICAM-1 pathway) and STAT-1, rather than TNF-α-mediated cellular responses (39) , cause LPS-mediated endotoxic shock, and LPS acts on TLR4 in a manner that activates both MyD88 and TICAM-1 (36 , 38) . This may be the reason that BCG-CWS is far less toxic than LPS and suitable for clinical use as an adjuvant.
The important point is that TICAM-1-mediated DC activation, unlike the MyD88-dependent DC response, sustains DC motility to lymph nodes, which is supported by CCR7 (40 , 41) . Thus, factors induced by TICAM-1 (37) cause the maturation of DCs in a fashion distinct from those induced by MyD88 (42) . It remains possible that TICAM-1 contributes to full DC maturation in concert with MyD88-dependent signaling. This issue could be clarified by testing TICAM-1 knockout mice in the future.
It has been accepted that MyD88 is shared as an adapter with receptors for IL-1β, IL-18, and some members of the TLR family (18, 19, 20 , 43) . For example, type 1 IFN directly activates the gene expression of IL-18 receptor components (AcPL), IL-1 receptor-related protein, and MyD88 in NK and T cells (44) . MyD88 may support the so-called danger signal induced by tumors or tumor-disrupted tissues (45) . This interpretation is reminiscent of the properties of the danger signal in the suppression of tumor cell progression. It seems possible that most danger signals cause the delay of tumor growth through activation of the adapter MyD88.
BCG-CWS was found to induce the expression of IL-23 but not IL-12 p70 simultaneously with DC maturation (33) . IL-23 participates in the production of IFN-γ in lymphocytes and is relatively weakly induced in the activation of NK cells (33 , 46) . In vitro analysis suggests that BCG-CWS activity induces DC maturation and IL-23 production, leading to effective Th1 polarization. The importance of IL-23 has not been clarified, but we favor the interpretation that the resultant induction of IFN-γ directly activates CD4+ T cells or cancels CD4+/CD25+ regulatory T-cell activity (47) . Recent studies using Serex suggested that CTL induction is sustained in the absence of regulatory T-cell activity (48) . In fact, IFN-γ-positive patients who underwent repetitive administration of BCG-CWS have enhanced long-term survival, which may reflect the induction of memory T function, presumably attributable to IL-23 (49) . Of course, possible differences between the mouse and human TLR systems need to be determined in this regard.
Live BCG has been reported to be effective for reduction of bladder tumor growth by activating host immunity (23) . BCG-CWS has also been administered to patients for postoperative treatment of cancer, producing good prognoses (15 , 17 , 18) . In addition to being a cost-effective therapy, such BCG-CWS treatment is simple, highly useful, and applicable to patients with various cancers. Patients with postoperative lung cancer receiving this therapy exhibited a high quality of life index of ∼70% and 5-year survival of ∼60% in our clinic (15 , 17) . The most intriguing idea is to use these adjuvant functions to establish an optimal immunotherapy strategy for the host: combined activation of CTLs and NK cells ought to more potently elicit antitumor potential because both MHC-negative and -positive tumor cells can be eliminated as their targets (50 , 51) . Such a therapy could provide immune system-activating signals, eliminate inhibitory factors, and avoid the emergence of immunoresistant phenotypes. These aims may be achieved by a combination of adjuvant modalities that induce CTL- and NK-mediated tumor elimination.
We are grateful to Drs. H. Koyama and M. Tatsuta (Osaka Medical Center for Cancer, Osaka, Japan) for support of this work and to Dr. N. Inoue, Dr. N. A. Begum, Dr. H. Oshiumi, S. Kikkawa, M. Kurita-Taniguchi, and K. Shida in our laboratory for many useful discussions.
Grant support: Supported in part by Core Research for Evolutional Science and Technology (CREST), Japan Science and Technology Corporation and Grants-in-Aid from the Ministry of Education, Science, and Culture (Specified Project for Advanced Research and Grant-in-Aid for Young Scientists); the Ministry of Health and Welfare of Japan; and the Takamatsunomiya Princess Memorial Foundation.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Tsukasa Seya, Department of Immunology, Osaka Medical Center for Cancer and Cardiovascular Diseases, Higashinari-ku, Osaka 537-8511, Japan. Fax: 81-6-6973-1209; E-mail:
↵6 Our unpublished data.
- Received May 29, 2003.
- Revision received October 20, 2003.
- Accepted November 6, 2003.
- ©2004 American Association for Cancer Research.