Human CD4⁺ T Cells Present Within the Microenvironment of Human Lung Tumors Are Mobilized by the Local and Sustained Release of IL-12 to Kill Tumors In Situ by Indirect Effects of IFN-γ¹

The role of CD4⁺ T cells in controlling solid tumor growth has been rather narrowly defined and less well understood compared with that of CD8⁺ CTL. The requirement for CD4⁺ T cells in antitumor response has been largely attributed to providing help to naive CD8⁺ T cells, leading to their activation and differentiation into tumor-specific effector CTL. CD4⁺ T cells have also been shown to be required for CD8⁺ T cells to: 1) establish long-term radio-resistant, specific memory immunity against tumors (1); 2) sustain CTL responses during chronic viral infections (2); and 3) maintain proliferation and viability of adoptively transferred CD8⁺ T cells in patients (3).

Work in animal tumor models, however, has suggested that CD4⁺ T cells have a much broader role in mediating antitumor effector functions that are independent of CD8⁺ T cells (4, 5, 6). For example, it has been established in mice that CD4⁺ T cells can reject tumors without coparticipation of CD8⁺ T cells, and in the absence of direct interaction between effector T cells and tumor cells. In such studies, adoptive transfer of CD4⁺ T cells (in the absence of CD8⁺ T cells) led to the complete elimination of tumor cells that failed to express MHC class II Ags, indicating an indirect mechanism of tumor killing (7, 8, 9, 10). The indirect effects of CD4⁺ T cells appear to be mediated through cytokines produced by both Th1 and Th2 cells, including IFN-γ (11), which leads to the activation of accessory cells including eosinophils and tumoricidal macrophages (12). CD4⁺ T cells may be particularly important in the immune response to tumors that have lost MHC class I, because direct CD8⁺ T cell tumor recognition is precluded in the absence of MHC class I expression. This has been demonstrated in mice in which vaccination with tumor cells lacking MHC class I resulted in tumor rejection that was dependent upon CD4⁺ T cells and NK cells, but not CD8⁺ T cells (13). Although these in vivo animal models have been used effectively to establish and define a significant role for CD4⁺ T cells in the antitumor response in mice and rats, it has heretofore not been possible to design and conduct similar in vivo studies on human cancer patients for obvious ethical reasons.

We have recently evaluated the antitumor activity of cancer patients’ lymphocytes in vivo by coengrafting PBL with the patient’s autologous tumor cells into SCID (3) mice. Using this human-mouse chimeric (SCID-Winn) model, it was established that CD4⁺ T cells mediated an IL-12- and IFN-γ-dependent suppression of tumor growth in vivo that was independent of CD8⁺ CTL (14). Although the SCID-Winn model does provide valuable insights with respect to the role of human CD4⁺ T cells in tumor immunity that cannot be studied using strictly in vitro systems, there are several limitations of this approach. Because the tumor cells and PBL are injected together as single cell suspensions, the antitumor activity of the human leukocytes is initiated in the absence of a true tumor microenvironment. As a result, the SCID-Winn model fails to account for the many complex interactions among the tumor-associated inflammatory leukocytes, stromal fibroblasts, and endothelial cells, as well as the potential tumor-induced immunosuppression that may evolve from the production and release of angiogenic factors, chemokines, cytokines, or other bioactive molecules such as matrix metalloproteinases (MMPs)³ and PGs. To overcome these limitations, a more physiologically relevant model has been developed in which nondisrupted pieces of human tumor biopsy tissues (obtained from patients with primary lung tumors) are surgically implanted into SCID mice. The resulting xenografts maintain an intact human tumor microenvironment consisting of tumor cells, inflammatory cells, fibroblasts, an extracellular matrix, and blood vessels of human origin (15). Previous work in this primary tumor xenograft SCID model has established that the patients’ tumor-associated inflammatory cells remain viable and responsive to cytokine stimulation for prolonged periods (16).

In this study, we demonstrate that CD4⁺ T cells present within the tumor microenvironment of human tumor xenografts are activated by the local and sustained release of human rIL-12 (rhIL-12) to secrete human IFN-γ, which results in the suppression of tumor growth in vivo. Changes in gene expression within the microenvironment of rhIL-12-treated tumor xenografts have been characterized, which suggest several indirect IFN-γ-dependent mechanisms contribute to the observed tumor suppression, including the activation of inducible NO synthase (iNOS).

C.B-17 scid/scid mice were obtained from the breeding colony at Roswell Park Cancer Institute. The original breeding stock was provided by B. Phillips (Hospital for Sick Children, Toronto, Ontario, Canada) with permission from M. Bosma (Fox Chase Cancer Center, Philadelphia, PA). SCID mice were maintained in microisolation cages (Lab Products, Maywood, NJ) under pathogen-free conditions. Animals of both sexes were used for tumor engraftment at 8–12 wk of age; however, all mice within a single experiment were age and sex matched. SCID mice were depleted of NK cells with a mAb (TM-β1) to the murine IL-2R β-chain (17). A single i.p. injection of TM-β1 ascites fluid (100 μl, diluted 1/2 in PBS) was given ∼30 min before surgical implantation of tumor tissue. The TM-β1-producing hybridoma was kindly provided by T. Tanaka (Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan).

Fresh surgical specimens of human lung tumors (non-small cell lung carcinomas) were obtained from the Tissue Procurement Service at Roswell Park Cancer Institute under sterile conditions. All tumor specimens were from patients who gave their informed consent before surgery, and definitive diagnoses for each specimen were obtained from the postsurgery pathology report. Each experiment was repeated with at least three different patient tumors with similar results; however, in some of the figures, representative data from a single patient’s tumor are presented.

rhIL-12 (1.7 × 10⁷ U/mg) was a generous gift from S. Wolf (Genetics Institute, Cambridge, MA). A phase-inversion nanoencapsulation technique was used for encapsulation of cytokines into biodegradable microspheres, as previously described (18). Briefly, BSA (RIA grade; Sigma-Aldrich, St. Louis, MO), polylactic acid (m.w. 24,000; Birmingham Polymer, Birmingham, AL), with or without rhIL-12 in methylene chloride (Fisher, Pittsburgh, PA), was rapidly poured into petroleum ether (Fisher) for formation of microspheres (1–5 μm). Microspheres were filtered and lyophilized overnight, for complete removal of solvent. Two formulations containing 10% BSA (w/w) were produced: 1) control (no cytokines), and 2) IL-12 (1.0 mg (1.7 × 10⁷ U)/mg polylactic acid). For injection into mice, microspheres were weighed out into sterile microfuge tubes, resuspended in PBS + 0.5% BSA, and sonicated twice for 15 s to achieve hydration. Suspended, rehydrated microspheres were then injected directly into tumor xenografts (see below).

Human lung tumor tissue (from a fresh surgical specimen for primary engraftment, or harvested from SCID mice for secondary passage) was carefully examined, and all macroscopically normal or necrotic tissue was removed. Tissue was cut into pieces (1–2 mm³ using a surgical scalpel), and kept moist with sterile PBS. Mice were injected with Avertin (Aldrich Chemical, Milwaukee, WI; 0.012 g/ml, i.p.) to induce anesthesia. A small incision was made on the abdominal skin (ventral midline) and undermined to produce a s.c. pocket (∼5–7 mm). Tumor pieces (4–5 pieces for a total of ∼50 mg/mouse) were implanted into the pocket, and the incision was closed with a surgical staple. Each mouse received pieces of tumor tissue chosen randomly from different areas of the original tumor specimen to ensure that the distribution of tumor and tumor-infiltrating lymphocytes among mice was equivalent. Samples of the original patient tumor specimen (i.e., pre-engraftment) were processed for analysis of gene expression and/or immunohistochemical staining (see below). Secondary (i.e., tumor-infiltrating lymphocyte-depleted) xenografts were established by serial passage of tumor tissue from one SCID mouse to another (19). Tumor tissue was passaged three to four times before use in an experiment, and the absence of human CD45⁺ cells was confirmed by immunohistochemistry (see below).

Seven days after surgical implantation of tumor tissue, mice were randomized into control and treatment groups (4–5 mice/group) and given a single intratumoral injection of microspheres (2.0 mg/100 μl/mouse) using a 281/2-gauge needle attached to a 0.5-ml insulin syringe. Following treatment on day 0, s.c. tumors were measured every 5–7 days (using engineer’s calipers) for 90 days or until tumor diameter reached >10 mm in two dimensions, at which point all mice in the experiment were sacrificed. Tumor volumes were calculated using the formula: (A² × B)/2, in which A and B are the length (in mm) of the shortest and longest dimensions, respectively. Statistical significance among the different treatment groups was determined using the unpaired Student’s t test to compare mean final tumor volumes.

Mice were sacrificed at different time points postmicrosphere treatment (e.g., 6 h, 24 h, 5 days, etc.), and xenografts were removed for subsequent analyses. Tissue to be used for gene expression studies (i.e., for RNA isolation) was transferred to a sterile cryopreservation vial containing 1 ml RNAlater (Ambion, Austin, TX) and stored at −20°C. Tissue was saved for histologic analysis (immunohistochemistry) by fixation in 10% neutral buffered Formalin. Tumor tissue to be analyzed by flow cytometry was harvested from SCID mice and immediately subjected to enzymatic digestion for generation of single cell suspensions (see below). In some experiments, xenografted mice were bled at different times (before and after microsphere treatment) using capillary pipettes to obtain 100–150 μl blood from the orbital sinus under anesthesia. Mouse sera were stored frozen at −80°C until analysis by ELISA (see below).

Global patterns of gene expression within the tumor microenvironment were analyzed using the human cytokine gene expression (macro) array (R&D Systems, Minneapolis, MN; GA001), which represents a collection of 375 different human genes encoding cytokines, chemokines, immunostimulatory molecules, activation markers, angiogenic factors, growth factors, proteases, etc., and their corresponding cell surface receptors. The protocols used for array experiments were based upon the manufacturer’s recommendations, and unless otherwise stated, all reagents used were from R&D Systems. A brief description of the experimental protocols, including cDNA synthesis, hybridization, imaging, and data analysis, is given below.

Total RNA was isolated from the original patient tumor specimen or the surgically excised xenografts using the TRIzol reagent (Life Technologies, Grand Island, NY), according to the manufacturer’s protocol. Following isolation, the RNA pellet was redissolved in the RNA storage solution (Ambion), quantified by measuring the absorbance at 260 nm, diluted to a final concentration of 0.25 μg/μl, and was stored at −80°C. For cDNA synthesis, 8.0 μl human cytokine-specific primers (R&D Systems; GAC11; an equimolar mixture containing an antisense primer corresponding to each cDNA on the human cytokine gene expression array) were first added to total RNA (4.0 μg/22.0 μl) and denatured for 3 min at 90°C, cooled at room temperature for 5 min, and then equilibrated to 42°C in a water bath for 5 min. A master mix for the reverse transcription (RT mix) was prepared with the final concentration of each reagent as follows: 1× reverse-transcriptase buffer, 333 μM dGTP, 333 μM dTTP, 1.67 μM dCTP (unlabeled), 1.67 μM dATP (unlabeled), 100 U/75 μl anti-RNase (Ambion), 25 μCi [α-³³P]dCTP, 25 μCi [α-³³P]dATP, and 250 U/75 μl avian myeloblastosis virus reverse transcriptase. A total of 30 μl RT mix was added to each 30 μl RNA/primer sample. Samples were then incubated in a 42°C water bath for 3 h to generate ³³P-labeled cDNA. Unincorporated radioactive nucleotides were removed from the cDNA by centrifugation (1100 × g for 4 min) through Sephadex G-25 spin columns (R&D Systems). A 2.0-μl aliquot of column-purified cDNA was removed to assess efficiency of ³³P labeling using a scintillation counter (Beckman LS58001; Beckman Instruments, Fullerton, CA). The activity of a 2.0 μl cDNA sample was typically ∼1.5–2.0 × 10⁶ cpm.

Macroarrays were rinsed for 5 min in 2× standard saline citrate phosphate/EDTA buffer, sealed individually in plastic bags (Seal-A-Meal; Dazey, Industrial Airport, KS) with ∼1-cm margins, and prehybridized (10 ml hybridization solution per bag) for 1–2 h at 65°C. Column-purified cDNA (∼75–80 μl) was denatured in a boiling water bath for 5 min and then added to each array in a total volume of 6 ml hybridization solution. Hybridization was performed for 12–14 h in a rotating water bath set to 65°C with continuous rotation (35 rpm). All posthybridization washes were performed in a plastic wash container according to the manufacturer’s instructions. Washed arrays were placed face up on blotting paper and air dried for 5 min, after which they were wrapped individually in plastic saran wrap and exposed to a low energy phosphor screen (Molecular Dynamics, Sunnyvale, CA) for a period of 4 days.

Images of each array were generated by scanning the exposed phosphor screen with the STORM 860 PhosphorImager (Molecular Dynamics) at 50 μm pixel size resolution and creating a .gel file using ImageQuant software version 5.1 (Molecular Dynamics). To quantify hybridization signal intensity, a template grid consisting of rectangles drawn around each pair of duplicate spots was constructed using ImageQuant. (Note that the template grid covered all areas of the array and included blank areas in which no DNA was spotted on the membrane.) For each rectangle in the template, the average pixel number per unit area (area) was calculated, and raw data were exported to Microsoft Excel for analysis.

Data were analyzed by first dividing the hybridization-signal intensity of each gene (averaged from duplicate spots) by the average background value for that particular array to generate values representing fold expression over background. Next, 1× background was subtracted to calculate the relative hybridization-signal intensity (RSI) of each gene. A ratio of total signal intensity (of all spots on the array) divided by average background was used to normalize RSIs between individual arrays generated from control and treated specimens. Changes in gene expression between arrays were considered significant if the magnitude of increase or decrease was >2-fold (i.e., ratio of normalized RSIs >2.0 or <0.5). The minimum threshold intensity used to differentiate expressed genes from nonexpressed genes was determined empirically to be 2 SDs above average background. In some experiments, a particular gene was not expressed above threshold in control-treated tissue, but was highly expressed (i.e., >3-fold over background) after IL-12 treatment. To calculate the ratio of gene expression in these cases, the minimum threshold value (i.e., background + 2 SD) was used in place of 0 for the value of the nonexpressed gene. Because human tissue was maintained in a SCID mouse host, macroarrays were tested for cross-reactivity with murine cDNA generated from both SCID mouse spleen tissue and murine lung tumor tissue (line 1 alveolar carcinoma) treated with murine rIL-12. A total of nine cross-reactive genes (of 375 total genes) were identified in these experiments, which were excluded from all analyses of human tumor xenografts.

The concentration of human IFN-γ in serum of tumor-bearing mice was determined using a sandwich ELISA, as described (19, 20). Briefly, microtiter plates were coated with an anti-IFN-γ mAb (M700A against human IFN-γ; Endogen, Cambridge, MA). Mouse sera and human IFN-γ standards (Pierce Endogen, Rockford, IL) were added to the plates, followed immediately with biotinylated secondary anti-IFN-γ Ab (M700B; Endogen). The bound Ab was detected with 1 μg/ml streptavidin-conjugated HRP (Sigma-Aldrich; A3151) and 3, 3′, 5, 5′-tetramethylbenzidine (Sigma-Aldrich). This ELISA was determined to detect human IFN-γ and not murine IFN-γ (data not shown).

For in vivo neutralization of human IFN-γ, SCID mice bearing established primary lung tumor xenografts were given i.p. injections of the anti-human IFN-γ mAb, B133.3.1 (21), starting the day of microsphere treatment. Control mice received the same dose of the mAb, 2C3 (mouse IgG, γ1, κ, anti-hapten phthalate), as an isotype control (22). The first dose of Ab was given ∼6 h before intratumoral microsphere injections. Subsequent i.p. injections of Ab were given daily thereafter for a total of 10 consecutive injections. Abs were prepared as ascites fluid diluted 1/4 in a total of 200 μl sterile PBS for injections. The monoclonal anti-human IFN-γ-secreting hybridoma, B133.3.1, was a gift from G. Trinchieri (Schering-Plough Laboratory of Immunological Research, Dardilly, France).

To deplete human CD4⁺ cells from established primary xenografts, tumor-bearing SCID mice were injected i.p. with Abs to human CD4 (γ-globulin purified from OKT4 ascites; 200 μg/200 μl PBS/mouse) 1 day before microsphere treatment (i.e., 6 days after tumor engraftment). A second group of mice bearing established primary xenografts received 2C3 as an isotype control.

For inhibition of iNOS activity in vivo, SCID mice bearing established primary lung tumor xenografts were injected i.p. with N-nitro-l-arginine methyl ester (l-NAME), or the inactive isoform N-nitro-d-arginine methyl ester (both from Sigma-Aldrich) as a control. Mice received 10 daily injections (0.2 mg/200 μl PBS), with the first injection ∼6 h before microsphere treatment (23).

Tissues (from the pre-engraftment specimen or tumor xenografts removed from mice) were fixed in 10% buffered Formalin, processed in an automated tissue processor, and embedded in paraffin. Serial sections were cut at 5 μm, placed on charged slides, and dried at 60°C for 1 h. Slides were deparaffinized in xylene and rehydrated through graded alcohol. Quenching of endogenous peroxidase was done using 3% aqueous H₂O₂ for 30 min. Slides were then rinsed in PBS and loaded onto a DAKO (Carpenteria, CA) automated Immunostainer. Slides were incubated for 1 h with a mAb to human CD45 (Zymed Laboratories, South San Francisco, CA; 18-0166) or mouse IgG (Sigma-Aldrich; M-5284) as a negative control (2.5 μg/ml). Biotinylated secondary Ab (Vector Laboratories, Burlingame, CA; PK-6200) was added for 30 min, followed by a PBS wash, and then Elite ABC Reagent (Vector; 6200) for 30 min. After a final PBS wash, diaminobenzidine chromagen (DAKO K3466) was added for 5 min. Slides were counterstained with Harris hematoxylin (Poly Scientific, Bay Shore, NY; s212) for 2 min, dehydrated, cleared, and mounted. Digital images were obtained using an Axioskop 2 microscope (Carl Zeiss, Thornwood, NY) with a SPOT camera (Diagnostic Instruments, Sterling Heights, MI). All images were taken under the ×10 objective lens (×100 magnification). The anti-human CD45 Ab was tested for cross-reactivity with murine CD45 (using a SCID mouse spleen) and was found not to stain murine leukocytes (data not shown).

Four-color flow cytometry was used to quantify human leukocyte subsets in single cell suspensions generated from primary lung tumor xenografts. Microsphere-treated xenografts were surgically excised from mice and digested using an enzyme mixture (10 ml/0.25 g tissue) consisting of 0.5 mg/ml collagenase A (Boehringer Mannheim, Indianapolis, IN; 1088 785), 0.2 mg/ml hyaluronidase type V (Sigma-Aldrich; H-6254), and 0.1 mg/ml DNase I (Sigma-Aldrich; D-5025), in PBS + 0.5% BSA (filter sterilized and stored at −20°C as 10× stock). After incubation at 37°C for 30 min on rotating platform, undigested pieces of tumor were allowed to settle by gravity, and the remaining supernatant (containing cells in suspension) was carefully transferred to a new 15-ml tube. Cells were washed twice in cold PBS + 0.5% BSA + 2 mM EDTA, resuspended in 14 ml cold PBS, placed on ice, and taken to the Laboratory of Flow Cytometry, Roswell Park Cancer Institute, for Ab staining and FACS analysis.

Cell suspensions were stained with different Ab panels consisting of four different Abs (each with a different fluorochrome label) to human leukocyte surface markers, including CD45, CD3, CD4, CD8, CD19, and CD33. Anti-human CD45 (present on all human leukocytes except plasma cells) (24) was included in every panel so that data could be gated on human leukocytes. For detection of intracellular IFN-γ, single cell suspensions of tumor tissue were incubated for ∼2 h in brefeldin A (20 μg/ml final concentration), and then surface stained with Abs to human CD45, CD3, and CD4 or CD8. After fixation with 2% formaldehyde, cells were permeabilized with 100 μl permeabilization reagent (Caltag, Burlingame, CA), stained with an anti-human IFN-γ Ab (clone MCH-IFG01; Caltag), and then analyzed by flow cytometry. All data were acquired on a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA), and analyzed by C. Stewart (Director, Laboratory of Flow Cytometry, Roswell Park Cancer Institute) using the WinList software package (Verity Software House, Topsham, ME). Between 10,000 and 20,000 events were collected for each sample analyzed. In preliminary experiments, the Ab panels used were found to reliably stain human PBL subsets after short (∼20-min) treatment with enzyme mixture, and did not demonstrate any cross-reactivity with SCID mouse spleen cells (data not shown).

The implantation of small nondisrupted pieces of human lung tumor biopsy tissues into the subcutis of SCID mice results in the engraftment of a tumor microenvironment that includes both tumor cells and human CD45⁺ inflammatory cells (Fig. 1,A, inset). To determine whether the inflammatory cells within the xenograft retain functional capacity, their response to the proinflammatory cytokine IL-12 was assessed. One week after engraftment, biodegradable polymer microspheres loaded with rhIL-12 were injected directly into the tumor xenografts. Cytokine-free microspheres were used as a control. As shown in Fig. 1,A, the local and sustained release of rhIL-12 completely suppressed tumor growth over a period of 45 days postengraftment. In contrast, tumor xenografts inoculated with control microspheres grew progressively (Fig. 1 A). Similar results were observed in 18 of 20 different patient tumors, including three different histological types of non-small cell lung cancer (i.e., squamous cell carcinomas, adenocarcinomas, and adenosquamous cell carcinomas).

Because human IL-12 is not active on murine cells (25) and human tumor cells of epithelial origin do not express the IL-12R (26), one would predict that it is the human tumor-associated inflammatory cells that are responding to rhIL-12 therapy in this model system. To confirm that the observed antitumor activity was mediated by human inflammatory leukocytes, lung tumor xenografts were depleted of human leukocytes by serial passage in SCID mice, and then treated with rhIL-12 or control microspheres. As illustrated in Fig. 1,B, rhIL-12 treatment had no effect on the growth of leukocyte-depleted xenografts (established from the same patient’s tumor as in Fig. 1,A). The absence of human CD45⁺ cells was confirmed by immunohistochemistry in Fig. 1 B inset. These results are representative of experiments conducted with five different patients’ tumors. Thus, we conclude that the human CD45⁺ inflammatory cells present within the lung tumor xenografts are responsive to cytokine stimulation, and are necessary for tumor suppression following local rhIL-12 therapy.

IL-12 is known to stimulate the proliferation of activated lymphocytes (27, 28), and may also serve to protect cells from apoptosis (29, 30). To evaluate the effects of local IL-12 treatment on the cellular events occurring within the tumor microenvironment, xenografted tissue was removed from control and experimental mice for histological and immunohistochemical analyses. The presence of human inflammatory cells in each tissue sample was assessed by staining with a mAb to human CD45, and a serial section of each specimen was stained with mouse IgG as a control (data not shown). As illustrated in Fig. 2, a greater number of CD45⁺ cells was observed in the rhIL-12-treated xenograft, compared with control-treated tumor, as early as 5 days after treatment (Fig. 2, A and B). By day 24, human CD45⁺ cells were essentially absent from the control-treated xenograft (Fig. 2,C), which was now densely packed with tumor cells. In contrast, the rhIL-12-treated xenograft (Fig. 2 D) was heavily infiltrated by CD45⁺ cells and showed little evidence of viable tumor. Data from additional patient tumors have yielded similar results, and human CD45⁺ cells have been detected up to 90 days postengraftment in IL-12-treated xenografts (data not shown).

Previously, we have demonstrated by immunohistochemical staining of untreated tumor xenografts that the majority of human CD45⁺ tumor-associated leukocytes were CD3⁺ T cells with fewer numbers of plasma cells, B cells, and macrophages (31). To gain further insight into the specific subsets of CD45⁺ cells affected by IL-12 treatment, four-color flow cytometry was performed on single cell suspensions isolated from treated and control xenografts after staining with different panels of Abs to human leukocyte differentiation Ags. Data from three independent experiments confirmed that human CD3⁺ T cells accounted for the highest percentage of CD45⁺ cells in day 7 xenografts (72–80%), with an average CD4-CD8 ratio of 3.5:1. Fewer numbers of CD19⁺ B cells (5–12%), and CD33⁺ myeloid cells (1–7%) were also detected in primary tumor xenografts. Within 5 days of rhIL-12 treatment, the frequency of CD4⁺ and CD8⁺ T cell subsets isolated from primary tumor xenografts increased by an average of 4.1- and 4.5-fold, respectively, compared with control-treated tumors. IL-12 treatment also increased the frequency of CD19⁺ B cells an average of 9-fold. In contrast, the total number of CD33⁺ myeloid cells was not significantly altered by IL-12 treatment.

We conclude that IL-12 has a significant effect upon the proliferation and longevity of the tumor-associated inflammatory cells and may contribute to their infiltration into the tumor parenchyma. Although each of these IL-12-associated effects may facilitate the human leukocyte-mediated killing, the data presented to date provide little insight with respect to the effector mechanisms responsible for the observed antitumor activity. It is now well established that IL-12 stimulates both activated CD4⁺ and CD8⁺ T cells (and NK cells) to secrete the proinflammatory cytokine IFN-γ, which in turn mediates tumor killing via several different effector mechanisms (28). Therefore, experiments were conducted to determine whether rhIL-12 was inducing the production of IFN-γ in one or more subsets of CD45⁺ human leukocytes within the lung tumor xenografts.

Single cell suspensions were obtained from tumor xenografts 16 h after treatment with either rhIL-12-loaded or control microspheres. The 16-h time point was selected for comparison to allow sufficient time for IFN-γ induction, while minimizing the effects of rhIL-12 on lymphocyte proliferation. Following surface staining with Abs to human CD45, CD3, and CD4 or CD8, the cells were then permeabilized and stained intracytoplasmically with a mAb to human IFN-γ. As shown in Table I, IFN-γ-producing CD3⁺ T cells were identified in both control- and rhIL-12-treated tumors, with CD4⁺ T cells accounting for the majority of IFN-γ production (86 and 87% of all IFN-γ⁺ T cells in control- and IL-12-treated tumors, respectively). IL-12 treatment increased the frequency of IFN-γ⁺ CD4⁺ T cells 3.6-fold (4.2–15.0%), while the frequency of IFN-γ⁺ CD8⁺ T cells increased 2.3-fold (2.5–5.7%) in response to IL-12 (Table I). Similar results were obtained using two additional patient tumors (data not shown). These results indicate that both CD8⁺ and CD4⁺ T cells within the microenvironment of human lung tumors produce IFN-γ, and that these IFN-γ-producing cells are increased in response to local IL-12 stimulation. The data also suggest that CD4⁺ T cells are the more important source of IFN-γ under basal conditions, and following IL-12 treatment.

To determine whether human IFN-γ was detectable in the sera of SCID mice bearing primary lung tumor xenografts, mice were bled at different time points after intratumoral microsphere treatment. As shown in Fig. 3, human IFN-γ was detected in the sera of mice treated with rhIL-12-loaded microspheres, but not in the sera of mice treated with control microspheres. This representative kinetic analysis revealed that human IFN-γ was present in the SCID serum as early as 1 day after injection of rhIL-12 microspheres, continued to increase until day 5, and was no longer detectable by day 10 posttreatment (Fig. 3). Human IFN-γ was not detected in the sera of mice bearing tumor xenografts depleted of inflammatory leukocytes even after treatment with IL-12-loaded microspheres (data not shown). Thus, rhIL-12 treatment of primary xenografts stimulates human inflammatory cells (i.e., CD4⁺ and CD8⁺ T cells) within the tumor microenvironment to produce human IFN-γ protein, which can be detected in the sera of SCID mice bearing the primary xenografts. Additional experiments revealed that the serum level of human IFN-γ 5 days after rhIL-12 microsphere treatment of the xenograft correlated directly with the degree of tumor growth suppression in individual mice that was observed 90 days after engraftment (Fig. 4). These data establish a link between IFN-γ production and tumor suppression.

To determine whether human IFN-γ was required for the IL-12-induced tumor suppression observed in this model, SCID mice bearing primary lung tumor xenografts were injected i.p. with either a neutralizing mAb to human IFN-γ or an isotype-matched control Ab before the intratumoral injection of rhIL-12-loaded or control microspheres. Representative data presented in Fig. 5 demonstrate that anti-IFN-γ Abs reversed the rhIL-12-induced tumor suppression, but had no effect upon the growth of tumors receiving control microspheres. In mice receiving rhIL-12, anti-IFN-γ treatment also reduced serum levels of human IFN-γ to undetectable levels, but did not inhibit the induction of IFN-γ mRNA as detected by RT-PCR (data not shown). These data establish that human IFN-γ is a necessary component in the suppression of human lung tumor xenografts, rather than simply a marker of leukocyte activation, following local rhIL-12 therapy.

Studies using both murine and human (14, 32) tumors have failed to demonstrate any direct effect of IFN-γ upon tumor growth in vivo. Nevertheless, IFN-γ is known to alter the expression of a number of genes that have the potential to indirectly alter tumor growth. To gain further insight with respect to possible antitumor effector mechanisms that contribute to IL-12-induced tumor suppression, gene expression patterns within xenografts were monitored following treatment with either control or rhIL-12-loaded microspheres.

A human gene expression array that monitors the expression of 375 genes encoding cytokines, chemokines, adhesion molecules, other immunomodulatory factors, and their receptors was used to determine which genes, if any, demonstrated altered expression within the tumor xenografts after treatment with rhIL-12. ³³P-labeled cDNA was generated from RNA isolated from tumor xenografts 5 days after treatment with rhIL-12- or BSA-loaded microspheres, and hybridized to the array (see Materials and Methods for details). The RSIs for each gene were determined from phosphor images of the arrays, and values from the rhIL-12 microsphere-treated xenografts were plotted against values obtained from xenografts treated with control microspheres. Representative data presented in Fig. 6 A show that rhIL-12 treatment of a primary xenograft induced significant changes (i.e., ≥2-fold increase or decrease) in 56 of 243 (∼23%) of expressed genes. In contrast, no significant change in gene expression with rhIL-12 was observed in leukocyte-depleted xenografts established from the same patient’s tumor. Similar results were observed with xenografts established from additional patient tumors and at different time points posttreatment (e.g., 6 h, 24 h, 3 days, 10 days; data not shown); however, the most significant changes were seen 5 days after rhIL-12 treatment.

Examples of specific genes whose expression patterns were altered by rhIL-12 in the primary tumor xenografts are shown in Fig. 7. Each of the six symbols represents a different patient’s lung tumor whose gene expression was profiled 5 days after treatment of xenografts with either rhIL-12-loaded microspheres or control microspheres. The data establish that rhIL-12 treatment significantly increased the expression of proinflammatory chemokines and cytokines (e.g., monokine induced by IFN-γ (MIG), IFN-γ-inducible protein 10 (IP-10), monocyte chemoattractant protein-1 (MCP-1), macrophage-inflammatory protein-1α, RANTES, IFN-γ, TNF-α, IL-16, etc.) in the majority of tumors studied (Fig. 7, A and B). In addition, immunostimulatory factors, leukocyte activation markers, and adhesion molecules (e.g., iNOS, CD40, Fas, CD27, VCAM, integrin α₄, etc.) were found to be increased in response to rhIL-12 (Fig. 7,C). By contrast, the expression of immunosuppressive factors such as TGF-β1 and macrophage migration inhibitory factor (Fig. 7,B), as well as proangiogenic and vascular-associated factors (e.g., vascular endothelial growth factor (VEGF), platelet endothelial cell adhesion molecule (PECAM), CD34, epithelial-derived neutrophile-activating peptide-78, platelet-derived growth factor (PDGF)-Ra/b, etc.; Fig. 7,D) decreased after IL-12 treatment. Decreases in the expression of a number of growth factors and tumor-associated molecules (e.g., insulin-like growth factor-I/II, decorin, osteopontin, epithelial cell adhesion molecule, etc.; Fig. 7,E) were observed in IL-12-treated tumors compared with control-treated tumors. Several MMPs (MMP-7, MMP-9, MMP-10, MMP-12, MMP-13) were down-regulated, while the MMP inhibitor, tissue inhibitor of metalloproteinase-1, and the proapoptotic factor, caspase-1, were up-regulated following IL-12 treatment (Fig. 7 F).

The changes in gene expression observed in tumors receiving a sustained dose of human IL-12 reflect a very similar pattern (i.e., the up-regulation of many IFN-γ-responsive genes) in five of the six tumors evaluated (Fig. 7, A–F, open symbols). None of the patients from whom these five IL-12-responsive xenografts were established had been treated with radiation or chemotherapy before surgical resection of their tumor. However, primary xenografts established from a single patient who was treated with high dose chemotherapy before surgical resection of his tumor (i.e., neoadjuvant chemotherapy) demonstrated a markedly attenuated response to IL-12 therapy (Fig. 7, •). Moreover, the poor response of this patient’s xenografts to IL-12 in terms of changes in gene expression was paralleled by a lack of IL-12-induced tumor growth suppression in vivo (data not shown). The cytotoxic effects of chemotherapy may account for the relative unresponsiveness of the inflammatory cells within these xenografts to IL-12 immunotherapy. Interestingly, primary xenografts established from this IL-12-unresponsive tumor appeared to have a similar degree of leukocytic infiltration as that of the tumors obtained from patients who were chemotherapy naive, suggesting that the effect of chemotherapy was not in reducing the number of tumor-associated inflammatory cells, but rather in decreasing their functional capacity.

Among the genes whose expression was most significantly increased by rhIL-12 treatment in the five IL-12-responsive tumors analyzed, several proinflammatory chemokines were identified. Quantitative data presented in Fig. 8 indicate the level of expression of the C-X-C chemokines MIG and IP-10, as well as the C-C chemokines MCP-1 and RANTES under different conditions. Although none of these genes was highly expressed in the pre-engraftment tumor specimen or the control-treated xenografts, all four genes were induced by treatment with rhIL-12-loaded microspheres (Fig. 8). In mice treated with neutralizing Abs to human IFN-γ, the IL-12-mediated induction of MIG, IP-10, RANTES, and (to a lesser extent) MCP-1 was inhibited (Fig. 8). Moreover, in secondary xenografts (i.e., depleted of inflammatory leukocytes) established from the same tumor specimen, rhIL-12 treatment failed to up-regulate the expression of these proinflammatory factors (Fig. 8). Similar results were obtained with xenografts established from additional patient tumor specimens (data not shown). Thus, we conclude that local and sustained delivery of rhIL-12 to the microenvironment of human lung tumor xenografts stimulates leukocyte-dependent and IFN-γ-dependent changes in gene expression, including the induction of the proinflammatory chemokines MIG, IP-10, RANTES, and MCP-1.

Some of the changes in gene expression patterns monitored by gene array analysis have been confirmed and quantified by standard as well as real time RT-PCR. The enhanced expression of IFN-γ has been observed as early as 6 h after treatment with rhIL-12, reaching a peak of 110-fold increase in IFN-γ message 2 days after the cytokine treatment. Similar increases in the expression of MIG and its receptor CXCR-3 were also detected by RT-PCR analysis. No increase in the expression of these genes was observed in xenografts treated with control microspheres (compared with the untreated pre-engraftment specimen), and the IL-12-induced increases in MIG and CXCR-3 expression were abrogated by treatment with Abs to human IFN-γ.

Although the changes in gene expression patterns induced by IL-12 correlate with tumor suppression, they do not define the effector mechanisms responsible for the observed inhibition of tumor growth. Previous studies in mice have demonstrated that the IL-12-induced suppression of established tumors is associated with activated macrophages in the tumor microenvironment that are expressing high levels of iNOS, and the importance of NO as an effector molecule has been confirmed in mouse tumor models (33). Because the gene encoding iNOS was expressed in six of six tumor xenografts tested, and its expression increased in four of six tumors after treatment with rhIL-12 (Fig. 7 C), the possible role of NO in mediating rhIL-12-induced tumor regression was now investigated.

To address the potential role of NO in the suppression of tumor in lung tumor xenografts, mice were treated systemically with l-NAME, an inhibitor of iNOS activity, during rhIL-12 therapy. Data presented in Table II show that treatment of mice with l-NAME resulted in a significant loss of rhIL-12-induced tumor suppression, compared with mice treated with N-nitro-d-arginine methyl ester (an inactive isoform of l-NAME). Despite the loss of rhIL-12-induced tumor suppression, serum levels of human IFN-γ were not altered by l-NAME treatment (Table II). Moreover, l-NAME had no effect on the growth of xenografts treated with control microspheres (Table II). Similar results were observed with one additional patient tumor. These data suggest that following local rhIL-12 treatment, human IFN-γ mediates the suppression of human lung tumor xenografts at least in part through an indirect mechanism requiring the induction of iNOS.

The role of CD4⁺ T cells in tumor growth suppression, IFN-γ production, and changes in gene expression was assessed by the depletion of these cells from the xenograft before treatment of the tumor-bearing mice with rhIL-12-loaded microspheres. To deplete the CD4⁺ T cells from established primary tumor xenografts, mice were injected i.p. with 200 μg of anti-human CD4 (OKT4) Ab 6 days after tumor engraftment. A second control group of tumor-bearing mice received 200 μg of an isotype control Ab. One day after receiving the Ab, both groups of mice were treated intratumorally with rhIL-12-loaded microspheres. Five days later, all mice were bled, and their sera were assayed for levels of human IFN-γ. At this time, two mice from each group were sacrificed, and RNA was isolated for gene expression analysis. The remaining mice were monitored weekly for 11 wk for tumor growth. The results presented in Table III indicate that CD4 depletion reversed the rhIL-12-induced tumor suppression in two of three animals and significantly decreased the serum levels of human IFN-γ in five of five mice tested. The altered gene expression patterns observed with rhIL-12 were partially or completely reversed in xenografts depleted of human CD4⁺ T cells. Following rhIL-12 treatment, message levels for IFN-γ, TNF-γ, MIG, IP-10, MCP-1, and RANTES were 3–5-fold lower in CD4⁺-depleted xenografts compared with nondepleted xenografts. The expression levels of genes that were suppressed as a result of treatment of rhIL-12 in the control Ab-treated group (i.e., MMPs, VEGF, PECAM, and CD34) were found to be higher in the anti-CD4-treated group compared with the control group. Genes encoding the human IL-12R (i.e., IL-Rβ1 and IL-12Rβ2) that were expressed in the control (i.e., nondepleted) group were below the level of detection in xenografts depleted of human CD4⁺ T cells. The CD4⁺ T cell depletion experiment was repeated with a second tumor with similar results. In this repeat experiment, there were four mice in the control group and four mice in the experimental group. Tumor growth was suppressed in three of four mice in the control group treated with rhIL-12-loaded microspheres and an isotype control Ab. In the IL-12-treated CD4⁺ T cell-depleted experimental group, tumor suppression was observed in only one of four animals. No IFN-γ was detected in the sera of mice from the CD4⁺ T cell-depleted group, while three of four mice were positive for IFN-γ in the control group. These data establish a significant role for human CD4⁺ T cells in the rhIL-12-induced tumor growth suppression, IFN-γ production, and associated changes in gene expression pattern.

It is established in this study that CD4⁺ T cells present within the microenvironment of human non-small cell lung tumors produce and secrete IFN-γ in response to a local and sustained release of IL-12, resulting in significant suppression or complete eradication of the tumor. Based upon the rapid response to IL-12 (changes in gene expression patterns are observed as early as 6 h after rhIL-12 treatment) and the dependence upon CD4⁺ T cells, the responding T cells are assumed to have been previously activated (most likely by the tumor itself) and may represent an effector memory subset of CD4⁺ T cell. This assumption is based upon the fact that the functional IL-12R (IL-12Rβ1 and β2) is only expressed on T cells after activation (26, 28). The role of these IL-12-responsive CD4⁺ T cells in mediating the tumor growth suppression was established by cell depletion studies. Adoptive cell transfer experiments in mice have previously established that tumor-sensitized, but not naive T cells are essential for tumor rejection that is induced by IL-12 (34). The findings presented in this work are the first to show that a similar activated T cell exists within human lung tumors and that this cell can be reactivated by rhIL-12.

In addition to mobilizing CD4⁺ T cells to kill tumor cells in situ by the release of IFN-γ, rhIL-12 was observed to enhance the longevity of the CD45⁺ tumor-associated inflammatory cells, which would be expected to have a significant long-term effect upon the ability of these inflammatory cells to control tumor progression. We previously documented proliferation of human leukocytes within established primary xenografts by 5-bromo-2′-deoxyuridine uptake (31) and now have confirmed by FACS analysis the expansion of the CD45⁺ cells, including CD4⁺ T cells, CD8⁺ T cells, and CD19⁺ B cells. IL-12 is known to stimulate the proliferation of activated murine lymphocytes (28) and has been shown to protect these cells from apoptosis (35, 36). Our finding of dense accumulations of human CD45⁺ inflammatory cells in rhIL-12-treated (but not in control) tumor xenografts as long as 90 days after engraftment is consistent with these earlier studies in mice.

Studies in mice have shown that IFN-γ is required for in vivo immune responses against tumors (37), virus-infected cells (38, 39), self Ags in autoimmune diseases (40, 41), and parasite-infected cells (42). A common characteristic of all of these murine studies is that IFN-γ is necessary, but not sufficient for the immune response. By linking the IFN-γ dependency of these T cell-mediated responses to the production of NO by macrophages as an effector mechanism, others have attempted to explain the necessary, but not sufficient, role of IFN-γ because this is only one of several cytokines and factors that are required for the macrophage-mediated NO production (43, 44, 45, 46). Our findings that the gene encoding human iNOS is expressed in human lung tumor xenografts and is elevated after treatment with rhIL-12 suggest that NO may be an antitumor effector molecule in humans also. The data showing that inhibition of iNOS partially reverses rhIL-12-induced tumor suppression provide the first direct evidence of NO-mediated tumor killing in vivo in humans. However, while NO is one of the effector mechanisms, we have not ruled out other factors within the tumor microenvironment that may be contributing to tumor arrest in response to local rhIL-12 therapy.

Changes in gene expression patterns observed in the primary tumor xenografts following rhIL-12 therapy are consistent with what one would expect from inflammatory cells in response to this cytokine, and suggest a number of other mechanisms that may be contributing to the tumor arrest (28). The consistent and significant enhancement of two IFN-γ-inducible chemokines IP-10 and MIG could be contributing to tumor arrest in a number of different ways. In addition to their ability to attract activated T cells, these C-X-C chemokines have been shown to have antiangiogenic properties (47). Both MIG and IP-10 can inhibit tumor growth by preventing the generation of new vessels needed for tumor expansion (48, 49, 50). Although we do not yet have evidence to sustain this possibility directly, our data showing rhIL-12-induced decreases in the expression of genes associated with angiogenesis (i.e., VEGF, PECAM, CD34, epithelial-derived neutrophile-activating peptide-78, growth-related oncogene-α, Ephrin A4, pleiotrophin, PDGFRa, and PDGFRb) suggest that inhibition of tumor neovascularization may be an important effector mechanism in the suppression of human lung tumor growth.

The suppression observed in several other genes that was associated with IL-12 treatment and tumor arrest needs further study and may be of considerable potential importance clinically. Ten different growth factors or lung tumor-associated genes were suppressed, and the expression of five MMP genes was significantly decreased in tumor xenografts treated with rhIL-12. A decrease in tumor-promoting growth factors within the tumor microenvironment could obviously contribute to a decrease in tumor growth, while the reduction in MMPs would be expected to have a number of different inhibitory effects upon tumor progression (51). MMPs have been shown to contribute to virtually all stages of cancer evolution and progression including metastasis (51). It therefore seems likely that the rhIL-12-induced decrease in MMPs and the increase in tissue inhibitor of metalloproteinase-1 (an inhibitor of MMP function) would have a significant impact upon tumor progression.

Our studies using this human/SCID chimeric model sustain the hypothesis that inflammatory cells associated with human lung tumors are functional, and able to respond to a proinflammatory cytokine when studied in situ. Previous studies addressing the potential functional capacity of human leukocytes in tumors have been primarily conducted on cells isolated from disrupted tumor tissues (52, 53, 54, 55, 56, 57, 58). The isolation of these cells requires the disruption of the tumor microenvironment, resulting in the loss of cell-cell and cell-extracellular matrix interactions that may be critical to leukocyte function and response to cytokines. It is becoming increasingly apparent that the extracellular matrix and stromal cells within microenvironments play a key role in programming immunocompetent cells and coordinating T cell activation and migration of both leukocytes and tumor cells (59, 60, 61, 62). Using nondisrupted fresh human tumor tissue explants, it was recently established by others that rhIL-12 induced the production of IFN-γ (63). However, because these studies were done in vitro, the long-term effect of rhIL-12 treatment and the release of human IFN-γ on tumor growth could not be assessed. The one obvious advantage of the human SCID chimeric model used in this study is that one is able to monitor within an intact microenvironment the cellular and molecular events associated with IL-12 treatment and correlate these events with in vivo tumor growth. The advantages, limitations, and pitfalls of using this and other human SCID chimeric models to study the tumor microenvironment and to evaluate therapeutic approaches to human cancer have been reviewed elsewhere (15).

The data presented in this work suggest that a local and sustained release of IL-12 into the tumor microenvironment of a single tumor nodule may be exploited clinically in the treatment of human cancer. The cytokine delivery may be accomplished either by the intratumoral injection of IL-12-loaded biodegradable microspheres, as reported in this work, or by the injection of irradiated cells transfected to produce IL-12, as reported previously (20). An assumption is being made that the IL-12-dependent CD4 T cell-mediated antitumor response locally, results in tumor cell lysis and the release of tumor Ags that induce a systemic tumor-specific immune response. The systemic antitumor response would be expected to eradicate uninjected tumors at other sites throughout the body. These predictions and assumptions have been tested and sustained using a completely murine lung tumor model that is syngeneic to BALB/c mice, i.e., the line 1 tumor. In this experimental model, it was established that the local and sustained release of IL-12 (or a combination of IL-12 and GM-CSF) into a single primary tumor nodule led to the eradication of multiple established metastatic tumors (32, 64). These studies also revealed that the local and sustained release of cytokines into a single tumor microenvironment provoked a tumor-specific systemic and protective antitumor immunity (32). A potential limitation of the in situ tumor therapy as a general therapeutic approach to be used clinically is that at least one accessible tumor nodule is required. In the case of lung cancer, most patients do not have any readily accessible tumor nodules. However, it would be possible to access a primary tumor nodule within the lung with the aid of a stereotactically directed needle. Such an approach is already used routinely for performing lung tumor biopsies and could be easily modified to inject IL-12-secreting cells or IL-12-loaded microspheres into the primary tumor.