Targeting the Lymphotoxin-β Receptor with Agonist Antibodies as a Potential Cancer Therapy

Receptors of the tumor necrosis factor (TNF) receptor (TNFR) family have been pursued as attractive therapeutic targets in oncology since the early studies showing antitumor activities of TNF and later FAS/Apo-1 antibody (1, 2). TNFR1 and FAS belong to a group of cell death–inducing TNFRs that also includes nerve growth factor receptor, TNF-related apoptosis-inducing ligand receptor (TRAILR) 1/2, death receptor (DR) 3, DR6, and possibly ectodermal dysplasia receptor (EDAR). These TNFRs harbor signaling adaptor motifs termed death domains that can initiate the extrinsic apoptosis program. In addition, TNFRs of this group can exert antitumor effects via other mechanisms that include tumor sensitization to chemotherapeutic agents, activation of antitumor immunity, and disruption of tumor-associated microvasculature (1, 2). The therapeutic potential of TNFR agonists has been well documented in many preclinical studies; however, excluding the current TRAILR-based therapies, clinical development of TNF and other death receptor agonists has been hampered by the danger of nonspecific cytotoxicity, vascular activation, and inflammation.

The TNFR family also includes several members, including lymphotoxin-β receptor (LTβR), RANK, and EDAR, which regulate developmental programs in lymph nodes, hair follicles, teeth, bone, and mammary epithelium (3–5). Signaling through these receptors may offer novel means for therapeutic modulation of tumors via signals that control cellular proliferation, survival, and differentiation. In addition, LTβR, CD27, CD30, CD40, 41BB, OX40, HVEM, RANK, and others in this group serve immunoregulatory functions that may stimulate tumor immunity (6). For example, activation of CD40 inhibits the growth of hematopoietic and epithelial tumor cells and can promote antitumor immune responses (7). LTβR plays a central role in the formation of lymphoid organs (3). During lymph node development and in the mature system, LTβR induces stromal cells to specialize leading to the expression of surface adhesion molecules and the secretion of chemokines that attract and position lymphocytes, thereby maintaining certain microenvironments (8, 9). Likewise, LTβR contributes to the morphogenesis of specialized epithelial cells termed M cells in the Peyer's patches that develop from the stem cell compartment of the intestinal crypt (10, 11). Continual LTβR signaling is also crucial for the specialization of the high endothelial vasculature in the lymphoid organs (12). The lymphotoxin system is integrally involved in the formation of organized ectopic lymphoid structures in chronically inflamed sites (13). These activities may be important in the tumor microenvironment to facilitate immunologic involvement.

We have reported previously that LTβR activation induced by its ligand lymphotoxin-α/β or an agonistic monoclonal antibody (mAb) triggers an IFN-γ-dependent death in HT29 colon carcinoma cells (14, 15). More recently, expression within a tumor of a second ligand for LTβR called LIGHT was able to trigger an immunomediated tumor response (16, 17). Here, we show that an agonistic anti-LTβR mAb blocks tumor growth in models of colon and cervical carcinoma, including s.c. and orthotopic xenografts of human colorectal tumors. These findings indicate that agonists of LTβR may prove useful as treatments for malignant diseases.

Cell lines and reagents. The HT29 clone 14 was described previously (14). All other human cultured tumor lines and the murine CT26 colon carcinoma line were obtained from the American Type Culture Collection (ATCC) or the National Cancer Institute tumor cell line collection. Monoclonal antibodies, including the murine CBE11 and BDA8 anti-human LTβR, 1E6 anti-human LFA3 (all IgG1), and the hamster Ha4/8 anti-keyhole limpet hemocyanin and ACH6 anti-murine LTβR mAbs, have been described (14, 18). The anti-EpCAM mAb was HT29/26 (IgG2a, ATCC; antigen defined by Cathy Hession at Biogen Idec, Cambridge, MA). A chimeric V/C mouse/human IgG1/κ variant of CBE11 was constructed and produced at Biogen Idec. An IgM version of the chimeric CBE11 was engineered by extending the heavy chain COOH terminus with 18 amino acids of human mu chain to force pentamerization as described (19). This pentameric antibody was purified by protein A affinity followed by size exclusion chromatography and called CBE11p. A fully humanized version of CBE11 was created by humanization of the variable domains of the mouse/human chimeric CBE11. A chimeric V/C human/mouse IgG1/κ CBE11 was also generated. The purified mAbs contained <2 endotoxin units/mg of protein. Size exclusion chromatography showed that the murine CBE11 preparations contained <2% aggregated forms.

In vitro cell viability assays. In vitro HT29 and WiDr viability assays were done as described previously (14). CT26 growth assays were carried out with the anti-murine LTβR agonist mAb ACH6 immobilized on plastic cell culture plates coated with anti-hamster capture antibodies (BD Biosciences, San Jose, CA). For long-term assays of cellular responses to LTβR agonistic mAbs, HT3 cells were plated onto collagen gels in six-well tissue culture plates at 1,000 per well and grown for 21 days. Cell viability was scored visually by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) staining of the colonies. Soft agar colony formation was carried out by standard methods and MTT-stained colony numbers were quantitated using Eagle Eye imager (Stratagene, La Jolla, CA). CBE11, CBE11p, and the control mAb 1E6 were used in these assays at 100 ng/mL.

Tumor models. S.c. tumor models were carried out by s.c. injection of 2 × 10⁶ WiDr or 5 × 10⁶ HT3 into 6- to 8-week female athymic nude mice (Harlan Sprague-Dawley, Madison, WI). Size measurements were done with LABCAT Tumor Tracking software (Innovative Programming Associates, Inc., Princeton, NJ). All xenografts were implanted on the flanks, except for the orthotopic experiments where implantation was on the cecum. The xenograft studies were conducted as randomized, double-blinded studies, where both identification of the therapies and previous tumor size measurements were blinded from the technical staff. Tumor volume was calculated as (L × W²) / 2. Typically, tumors were pregrown without treatment to a minimum size of 5 mm and then randomized into test and control groups. Antibodies were injected i.p. based on the half-life of each mAb construct (mCBE11 dosed every 14 days; hCBE11 dosed every 7 days). Formal assessment of synergism between hCBE11 and chemotherapeutic agents in vivo used calculations using CalcuSyn software version 1.1 (Biosoft, Cambridge, United Kingdom). Mouse CT26 carcinoma cells (1 × 10⁶) were injected s.c. on the flank of BALB/c mice, tumors were grown to 100 to 200 mm³, and the mice were randomized before being treated with 2 mg/kg hamster anti-mouse LTβR ACH6. Orthotopic models were carried out as described in both blinded and nonblinded formats (20). Briefly, animals were implanted by surgical orthotopic implantation using 3 to 5 tumor fragments (2 mm³) harvested from minimally passaged s.c. tumor stock animals. Fragments were implanted onto the ascending colon after the serosa had been stripped away and the implants were sutured onto the colon (21). Statistics were calculated using ANOVA values for the tumor sizes and log-rank analysis for the survival curves. All tumor studies were carried out in accordance with the respective institutional animal care committee protocols.

Gene expression analysis. RNA from cultured cells or xenograft tumor samples was isolated by the standard Trizol protocol (Invitrogen, Carlsbad, CA) followed by additional purification on RNeasy columns (Qiagen, Valencia, CA). Hybridization probes were prepared and hybridized to U95Av2 GeneChips according to the Affymetrix (Santa Clara, CA) protocols. Transcript profiles of the clinical colorectal tumor samples were extracted from a commercially available data set (Gene Logic, Gaithersburg, MD) as U95A GeneChip data and converted to the U95Av2 format for further analysis. Signal intensity data were exported using MAS5 software, filtered to remove genes marked as absent across all samples, and imported for further analysis into GeneSpring (Agilent, Palo Alto, CA). Prediction of sensitivity to the LTβR agonist mAb was done using the k-nearest neighbor class prediction algorithm implemented in the GeneSpring gene expression analysis software (Agilent) on log-transformed data sets normalized to median chip intensity. Initial size of each predictor gene list was set to 50 genes using cross-validated genes with the class prediction P cutoffs of 0.2. For hierarchical clustering analysis, expression values of the selected predictor genes were averaged for each group of samples representing a given cell line or class of clinical tumor samples and standardized using the mean and SD values of the resulting groups of averaged samples.

Immunohistochemistry. Formalin-fixed paraffin-embedded tissue sections (Imgenex slides, San Diego, CA) were deparaffinized and blocked with 2% horse serum in PBS. Due to poor performance of CBE11 in the staining of paraffin sections, another murine mAb BDA8 was used as a substitute to detect human LTβR by immunohistochemistry. Bound BDA8 was detected with the Vectastain avidin-biotin complex method mouse IgG kit (Vector Laboratories, Burlingame, CA) according to the manufacturer's protocol. Specificity of the LTβR staining was monitored in control experiments using a mixture of BDA8 with a 10-fold excess of human LTβR-Ig protein. To examine murine B-cell and T-cell infiltrates, paraffin sections of the CT26 tumor grafts were stained with anti-CD45R (B220) or anti-CD3.

Oligomeric agonist mAb-induced LTβR activation inhibits growth of tumor cell lines in vitro. TNFR-induced signaling was originally thought to result solely from ligand-induced receptor trimerization. However, it is now apparent that efficient transduction of such signals may require a higher degree of receptor aggregation (22, 23). Consistent with this, anti-LTβR agonistic mAbs induced cell death more efficiently when immobilized on the surface of cell culture plates than when applied to cells in the soluble form (14, 24). To enhance the potency of agonistic mAb-mediated LTβR activation, we engineered an expression construct combining the COOH-terminal tail of the IgM heavy chain with the heavy chain of CBE11 and thereby converting this LTβR agonist mAb into an IgM-like oligomer (19). The resulting pentameric form of CBE11 (CBE11p) added to the culture medium inhibited cell proliferation more potently than the original monomeric mAb provided in solution or immobilized on the cell culture substrate (Fig. 1A). The CBE11p-induced HT29 cell death occurred within 3 to 4 days in the presence of IFN-γ. Further experiments have revealed that IFN-γ potentiated the effect of CBE11p but was not essential for it. For example, as shown in Fig. 1B, HT29 colony formation in soft agar was still efficiently inhibited by CBE11p in the absence of IFN-γ. In contrast to HT29 cells, the cervical carcinoma cells HT3 displayed a morphologic change (flattening) but showed no signs of impaired viability within the first 3 to 4 days of treatment with CBE11p. However, prolonged exposure of HT3 cells to CBE11p inhibited their growth (Fig. 1C) in vitro and in xenograft tumors (see below).

LTβR agonist mAb CBE11 inhibits tumor growth in xenograft models. Effects of CBE11 on the growth of established tumors were studied in xenograft models of human colon, breast, lung, and cervical carcinomas and melanoma. Both HT29 and WiDr lines have been investigated, although it is questionable whether these two lines are actually distinct (14). Both lines were similarly sensitive in vitro and in vivo and the in vivo sensitivity of HT29 has been reported by us earlier (14). Because of the more uniform performance in xenograft experiments, we preferred WiDr for in vivo work. CBE11 efficiently inhibited the growth of WiDr colon carcinoma xenografts at doses as low as 1.25 mg/kg given once every 2 weeks (Fig. 2A). Similar inhibition of tumor growth and frequent tumor regression were also observed with the cervical carcinoma cell line HT3 (Fig. 2B). As shown in the previous section, the soluble monomeric CBE11 used in our in vivo experiments showed little activity in vitro, whereas its pentameric form was efficacious. Therefore, we sought to determine whether small amounts of the aggregated monomeric CBE11 could account for its efficacy in vivo. Aggregate-free batches of CBE11 were prepared and, when tested in vivo, showed activity virtually indistinguishable from that of the original aggregate-containing preparations, thus showing that nonspecific aggregation of CBE11 was unlikely to contribute to its antitumor efficacy (data not shown). The inhibition of tumor growth by CBE11 could potentially result from the activation of LTβR signaling as well as from the activation of complement-dependent cytotoxicity (CDC) or antibody-dependent cellular cytotoxicity (ADCC). Therefore, we have tested the potential contribution of ADCC in experiments using an isotype-matched (IgG1) antibody against the human LFA-3 (mAb 1E6) and a nonmatched anti-EpCAM mAb (HT29/26). Both 1E6 and CBE11 were found to detect comparable numbers of binding sites per cell as determined by fluorescence-activated cell sorting (FACS) analysis (data not shown) and therefore were equally likely to promote ADCC responses against the engrafted tumor cells. In contrast to CBE11, 1E6 showed no antitumor activity and the anti-EpCAM mAb HT29/26 efficiently bound to HT29 cells but only slightly delayed the growth of HT29 xenografts. In addition, we have identified several cell lines that expressed similar levels of LTβR yet did not respond to CBE11 as xenografts. Therefore, direct cytotoxicity due to ADCC or CDC was unlikely to mediate the tumor inhibition by the anti-LTβR mAb. A summary of the in vitro and in vivo data is presented in Supplementary Table S1.

Humanization of CBE11 produced an IgG1 mAb that retained the original binding affinity for LTβR. However, the antitumor activity of the humanized CBE11 as well as that of the chimeric mouse/human CBE11 was significantly lower than the activity of the original murine mAb, suggesting potential functional importance of the Fc domain. To test this possibility, we have modified the humanized antibody construct by replacing its Fc portion with the murine IgG1 Fc domain. This modification restored in vivo activity of the humanized CBE11 (Fig. 2C), thereby suggesting that the antitumor efficacy of this mAb could be modulated by host-derived factors or by bystander cells interacting with the IgG1 Fc domain of CBE11 and presenting the mAb to the tumor cells in an oligomerized Fc form.

CBE11 enhances tumor sensitivity to chemotherapeutic agents. Agonists of several TNFRs, including TNFR1, FAS, and TRAILRs, have been shown to increase sensitivity of tumor cells to chemotherapy (25). Similar to these observations, we have observed enhanced WiDr xenograft responses to combinations of CBE11 with several chemotherapeutic agents. Synergistic tumor inhibition was observed using combinations of CBE11 with Camptosar (Fig. 2D), Taxol, and gemcitabine. In addition, additive potentiation of tumor inhibition was observed in experiments combining CBE11 with cis-platinum and Adriamycin. Similar results were obtained with another colorectal tumor cell line KM20L2 (data not shown).

Anti-murine LTβR agonist mAb induces tumor necrosis and lymphocyte infiltration in a syngeneic colon carcinoma model. Immunoregulatory functions of LTβR include regulation of leukocyte trafficking and function via the induction of chemokines and cell adhesion molecules in stromal cells, suggesting that activation of this receptor could promote antitumor immune responses. Furthermore, recent studies have shown that expression of the dual LTβR/HVEM ligand LIGHT in tumor cells can induce leukocyte infiltration into the tumors accompanied by development of antitumor immunity (16). To determine whether selective activation of LTβR could lead to similar consequences, we have tested cellular responses to LTβR activation in several murine tumor cell lines. LTβR activation with the agonist mAb ACH6 caused rounding and growth inhibition of mouse CT26 colon carcinoma cells, which, like the similar responses of the HT29 human colon carcinoma cells, were amplified in the presence of IFN-γ (Fig. 3A). A single treatment of BALB/c mice bearing established CT26 tumors with an agonist LTβR mAb (2 mg/kg ACH6) led to pronounced tumor necrosis at day 3 (Fig. 3B and C). Staining for CD3⁺ T cells revealed increased numbers of tumor-infiltrating lymphocytes in the tumor capsule and in the adjacent tissue (Fig. 3D). In addition, we have observed increased numbers of B cells in the tumor-adjacent tissue of ACH6-treated tumors; however, in contrast to the T cells, the B cells did not infiltrate the tumor compartment (data not shown). Although the limitations of the hamster-derived agonist mAb precluded examination of long-term tumor growth, our findings indicate that selective activation of LTβR may trigger events similar to those induced by simultaneous activation of LTβR and HVEM with LIGHT (16).

Prediction of CBE11 efficacy in tumor models derived from human colon carcinoma cell lines and clinical tumor samples. Activation of LTβR has been shown to suppress tumor cell viability in vitro by an ill-defined mechanism that differs from classic apoptosis (15). Because molecular determinants of the LTβR-induced suppression of tumor cell growth remain unknown, we sought to identify surrogate molecular features characteristic for CBE11-sensitive and CBE11-resistant tumor models. For this purpose, we have searched for patterns of gene expression that would allow prediction of cellular and tumor response to the activation of LTβR. This approach has been successfully used in several recent studies to predict the course of malignant disease and tumor response to therapy (26). We have done transcript profiling of a panel of colon carcinoma cell lines grown both in vitro and as xenograft tumors. A set of genes whose differential expression pattern was predictive of the cellular response to CBE11 was identified using a training sample set composed of cell lines that were experimentally determined to be sensitive (HT29 and WiDr) or resistant (DLD1, HCT15, SW480, and LS174T) to the cytotoxic effects of CBE11. This analysis has yielded a panel of 48 genes (Fig. 4) whose expression was largely invariant to experimental differences in growth conditions but reliably distinguished the sensitive and resistant models included in the training set (Supplementary Table S2). This gene set was further used to classify an additional panel of four colon carcinoma cell lines with unknown sensitivity to LTβR activation. One cell line (KM20L2) was predicted to be sensitive and three cell lines (Geo, KM12, and Colo205) were classified as resistant. Subsequent experimental validation of the prediction in both in vitro and in xenograft models has showed complete concordance of the predicted and experimentally observed properties of the test panel of cell lines (Supplementary Fig. S1).

Following the experimental validation of the predictor gene set, we applied it to assess the likelihood of clinical tumor response to CBE11 using a commercially available gene expression data set derived from 40 clinical colorectal tumor samples. Tumors were grouped into three classes, CBE11 sensitive, CBE11 resistant, and unclassifiable. Within this collection, ∼35% were predicted to be potentially sensitive to CBE11 and exhibited an average gene expression profile resembling those of the CBE11-sensitive cell lines HT29, WiD, and KM20L2 (Fig. 4).

To verify this analysis, we have tested antitumor efficacy of CBE11 in studies using orthotopic xenografts of tumor explants independently isolated from surgical specimens of six different human colorectal tumors and minimally expanded by serial transplantation in nude mice. In the first study, CBE11 reduced tumor growth in two of the six orthotopic colorectal tumor xenografts (AC3717 and AC3609; Supplementary Table S3). In the second study using the AC3717 tumor explants, treatment with CBE11 was shown to improve long-term survival when used as a monotherapy and to further improve it in combination with 5-fluorouracil (5-FU; Fig. 5). Early treatment with CBE11 in combination with 5-FU resulted in an 80% survival rate at 350 days compared with 30% and 65% for 5-FU or CBE11 alone. The second responder tumor AC3609 had lost its original growth and histologic properties on further serial expansion; hence the observation could not be repeated. The originally observed tumor response rate of two of six tumors in these studies was consistent with the frequency of tumor response predicted by the transcript profiling and class prediction analyses, thus suggesting that CBE11 therapy could affect a significant percentage of colorectal tumors.

LTβR is expressed in a wide range of tumor types. To assess the distribution of LTβR in solid tumors of different origin, we have done an immunohistologic survey of clinical tissue samples using the anti-LTβR mAb BDA8 that was found in preliminary experiments to work in immunohistochemistry considerably better than CBE11. In a survey using human tumor arrays, 13% of breast, 66% of colorectal, 44% of lung, 63% larynx/pharynx, 86% stomach, and 37% of melanoma tumors were classified as 2 to 3+ LTβR positive. When all positive tumors were included, 87% to 96% of tumors in all groups, except breast, were LTβR positive and 50% of breast tumors showed at least 1+ staining. Many tumors showed preferential expression of LTβR in the tumor nests relative to the surrounding stroma. Most of this expression was punctate within the cytoplasm and such intracellular staining was noted in earlier studies both with LTβR and some other TNF family receptors (27–29). Well-defined cell surface staining in the histologic survey was rare, probably reflecting the relatively low density of this receptor on the cell surface as determined by FACS analyses (data not shown). Examples of receptor-positive breast, colon, and lung tumors are shown in Fig. 6. The images show that LTβR expression is not restricted to the tumor compartments but can also be detected in the normal breast ductal epithelial cells as well as in the normal colon epithelium. These observations were in agreement with a previous study showing expression of LTβR in lung carcinomas and in hyperplastic bronchial lining cells (30). Based on these findings, the high frequency of LTβR expression in a broad range of solid tumors suggests tumors of diverse tissue origins and histotypes as potential additional candidates for CBE11 therapy.

Studies of TNF family receptors as drug targets for the treatment of cancer have focused primarily on the death receptor subfamily, including TNF-R, Fas, and TRAILR (1). More recently, induction of caspase-independent cell death mechanisms has received considerable attention (31). Here, we have shown that activation of the death domain-less TNFR family member LTβR can inhibit the growth of established tumors derived from cultured cell lines and from explants of clinical tumor samples. The results of our studies using xenograft models show that an agonist anti-LTβR mAb can suppress tumor growth via direct effects on tumor cells. Similarly, several groups have reported that LTβR activation can directly inhibit growth and survival in cultured cell lines (14, 24, 28, 32–37). The molecular mechanisms of this inhibition remain largely undefined; however, reports have implicated TNFR-associated factor (TRAF) 2 and TRAF3 (38–40), caspase-independent mechanisms (15, 36, 41), superoxide radical formation (42), kinase activation (41, 43), IAP1 and Smac (39), and down-modulation of bcl-2 (36) as elements of LTβR signaling. LTβR as well as BAFF-R, RANK, Fn14, and CD40 have been shown to activate the alternate nuclear factor-κB (NF-κB) pathway (44). Although the prosurvival role of canonical NF-κB signaling is widely recognized, it is unclear how components of the alternate NF-κB pathway may interact with mechanisms that control cell survival and differentiation. Nonetheless, several observations suggest that NF-κB2 can be involved in the regulation of tumorigenesis (45), possibly via the regulation of cell death mechanisms (46, 47).

Another potential mechanism for LTβR-induced tumor inhibition may involve cell cycle arrest. We have observed that stimulation of HT29 with CBE11 causes accumulation of the cells in the G₁-G₀ phase of the cell cycle (data not shown). Consistent with this observation, analysis of gene regulation by CBE11 revealed repression of a group of G₂-M cell cycle regulators (data not shown). Therefore, modulation of cell cycle progression could contribute to the CBE11-induced enhancement of tumor chemosensitivity observed in this study. We have shown that treatment with CBE11 can markedly increase in vivo responses to gemcitabine, Camptosar, Taxol, cis-platinum, and Adriamycin in the WiDr and KM20L2 colon carcinoma tumors that are normally resistant to many common chemotherapeutic agents. Therefore, our data indicate that activation of LTβR may facilitate the treatment of chemoresistant tumors.

In addition to the direct antitumor effects, LTβR activation may interfere with tumor growth via induction of host-mediated immunologic mechanisms. LTβR is a receptor for two ligands, lymphotoxin-α1/β2 and LIGHT. Expression of LIGHT in a breast carcinoma cell line has been shown to block growth of the tumor in vivo and its expression in a syngeneic tumor model induced an immunologic response to the tumor leading to systemic tumor immunity (16, 17, 32). Transgenic expression of LIGHT can also induce severe inflammation in nonlymphoid tissues (9). LIGHT also binds to a second TNF family receptor, HVEM, and although HVEM does not contribute to the direct effects of lymphotoxin-α/β on tumor growth in vitro, HVEM is expressed on T and natural killer (NK) cells (8, 9, 17, 34). The effects of LIGHT are mediated by interaction with LTβR expressed by the tumor and host stroma as well as through activation of HVEM on T and NK cells. More recent work indicated that this eradication of established tumors was mediated by CD8 cells in a manner dependent on NK cell activation by LIGHT (17). In our experiments with the murine syngeneic CT26 tumor model, we observed induction of lymphocyte infiltration and tumor necrosis by an LTβR-specific agonist mAb. Therefore, activation of LTβR may promote some of the aspects of the antitumor host responses similar to those inducible with LIGHT. One potential mechanism involved in this response to the LTβR agonist mAb may be the induction of proinflammatory chemokines. Expression of CXCR3-binding chemokines CXCL9-11 (MIG, IP-10, and I-TAC) can be induced in carcinoma cells by LTβR agonists.⁴

Formation of ectopic organized lymphoid aggregates is believed to rely on LTβR signaling in chronically inflamed settings (3, 13). Anti-LTβR agonist mAb therapy could potentially trigger some of the events that nucleate the formation of such ectopic centers, notably, release of chemokines, such as CCL9, CCL10, and perhaps CXCL13 and CCL21. These chemokines could attract various immune cell subsets into the tumor. Likewise, LTβR is critical for the differentiation of flat endothelium into the high endothelium capable of mediating lymphocyte trafficking in the lymph nodes (12). Therefore, this therapy could potentially alter the leukocytic composition of the tumor due to effects on both chemokine and endothelial trafficking points. Indeed, the expression of both CCL21 and the addressin MAdCAM were elevated within tumors expressing LIGHT (2). The combination of direct effects on tumor cell cycling and survival with the types of changes that could enhance an immunologic response to the tumor may be beneficial in some settings.

Our data suggest, but do not prove, that inhibition of tumor growth by the LTβR agonist mAb was primarily due to the activation of LTβR signaling rather than to the induction of CDC or macrophage- or NK-mediated ADCC mechanisms. We have shown that antitumor efficacy of the anti-LTβR mAb is likely to be modulated by the tumor host via an unidentified mechanism involving the Fc domain of the mAb. Comparison of matched anti-TRAILR2 antibodies with murine IgG1 or IgG2a Fc domains revealed profound Fc-dependent differences in the in vivo potency of the antibodies (50). This observation may reflect mechanisms similar to those causing the species- and Fc-dependent modulation of the in vivo efficacy of CBE11 observed in these studies. It is plausible that this mechanism involves aggregation of the mAb resulting in its enhanced signaling capacity. Potential mediators of such enhancement include Fc receptor engagement or the complement system.

Therapeutic activation of TNF family receptors for the treatment of cancer has been complicated by concerns of dose-limiting toxicity, which have hampered the clinical development of TNFR- and FAS-activating agents. Our analyses by FACS and gene expression profiling of normal primary human endothelial cells, fibroblasts, and monocytes exposed to CBE11p revealed no signs of a major proinflammatory response similar to that induced by TNF (data not shown). Furthermore, no toxic effects following short-term (weeks) exposure of mice or nonhuman primates to LTβR agonist mAbs were observed in preclinical studies (18).⁴ Two studies have suggested that LTβR activation may actually promote tumors. In one study, cell lines transfected to express a LTβR-immunoglobulin fusion protein (i.e., a lymphotoxin/LIGHT inhibitor) grew more slowly than control lines, suggesting that lymphotoxin or LIGHT can promote tumor growth and angiogenesis (51). In a second report, cultured NIH3T3 fibroblasts transfected with LTβR showed signs of focus formation indicating a transforming activity and similar observations were made with constitutive CD40 activation (52, 53). We had also noted that LTβR activation could increase proliferation of a human nontransformed fibroblastoid cell line (54). Whether this effect is limited to fibroblastoid cells as opposed to this work with epithelial lineages is unclear; however, in our experiments, we have not observed exacerbation of tumor growth. Mice bearing a LIGHT transgene develop autoimmune disease; hence long-term activation of LTβR by an agonist mAb could lead to detrimental effects (9). In this regard, clinical experience with activation of CD40, another non-death domain receptor, may provide some guidance (7).

Successful clinical development of novel cancer therapeutics can be facilitated by the availability of prognostic information about the spectrum of potentially sensitive and resistant tumors; however, limitations of the mouse tumor models allow only empirical evaluation of tumor sensitivity. To further assess potential clinical relevance of a CBE11-based tumor therapy, we developed a gene expression profile that was predictive of the cellular response to LTβR activation. Interestingly, profiles of gene expression similar to that of CBE11-sensitive cell lines were observed in ∼35% of actual colorectal tumors. Moreover, the response rate in the orthotopic xenograft studies using minimally passaged tumor fragments derived from surgical samples of human colorectal tumors was consistent with this prediction. Despite the small overall sample size, the frequency of tumor response to CBE11 observed in the orthotopic experiments was higher than the frequency of CBE11-sensitive cell lines identified thus far. The apparent discrepancy between cultured tumor cell lines and actual tumors likely reflects changes in the biological properties of tumor cells, which invariably occur during their adaptation to growth in vitro and can modify or eliminate original cellular phenotypes sensitive to LTβR activation. Application of a sensitivity prediction approach similar to that used in this study has proven to be feasible in clinical settings and could facilitate the identification of patients likely to benefit from therapeutic targeting of LTβR (26).

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.

We thank Joe Davie for his support of this investigation in the early stages; Greg Thill for helpful discussions; Humphrey Gardner and Lawrence Zuckerberg for the help with the assessment of tumor histology; members of the Biogen Idec animal facility; Tausha Pico, Fangxian Sun, and Ray Magana of AntiCancer, Inc. (San Diego, CA); the preclinical group at Biogen Idec for the nonhuman primate testing of humanized CBE11; and Paula Hochman and Ann Ranger for critical reading.