Control of Metastasized Pancreatic Carcinomas in SCID/Beige Mice with Human IL-2/TKD-Activated NK Cells¹

Pancreatic carcinoma is currently the fifth leading cause of cancer-related mortality with an increasing incidence, especially in women. The overall 5-year survival rate is <5% (1, 2). The aggressive nature of pancreatic cancer is reflected by nonresectable primary lesions, locally advanced tumors, and a high metastatic potential (3, 4). The orthotopic (o.t.)⁶ implantation of human pancreatic adenocarcinoma cells in immunodeficient mice provides an appropriate tool to determine tumor growth, progression, and metastatic dissemination in vivo (5, 6, 7, 8). Because pancreatic carcinomas are largely refractory to conventional therapies, there is a strong medical need for the development of innovative therapeutic strategies. Within the last decade, the immunostimulatory potential of heat shock proteins (Hsps) for anticancer immune responses has been elucidated. The pioneering work of the group of Srivastava et al. (9) demonstrated the ability of Hsps to elicit cancer immunity. Apart from their carrier function, chaperoning tumor-specific peptides to APCs and thus initiating a CD8-specific T cell immune response (10), Hsps also have been characterized as potent activators of the innate immune system (11, 12, 13, 14, 15). Extracellular localized Hsp70s have been found to mediate activation signals to human monocytes, murine macrophages (16), and murine dendritic cells (17). Screening for tumor-specific markers revealed a cell surface localization of Hsp70, the major stress-inducible member of the Hsp70 group, on a variety of different tumor entities (18, 19, 20), including the adenopancreatic cell line Colo357 (21). Furthermore, 64% of the tested primary pancreatic cancer biopsies, but none of the corresponding normal tissues was found to be Hsp70 membrane-positive (22). In vitro, membrane-bound Hsp70 provides a tumor-specific target structure for CD3⁻ NK cells (15) coexpressing the C-type lectin receptor CD94 at high densities (23). The amino acid sequence of Hsp70, which is exposed to the extracellular milieu of tumors, was identified as the 14-mer peptide TKDNNLLGRFELSG (TKD) residing in the substrate binding domain (24, 25). Incubation of NK cells with the synthetically produced Hsp70-peptide TKD in combination with low dose IL-2 was found to enhance both the cytolytic and proliferative capacity of CD3⁻CD94⁺ NK cells toward Hsp70 surface-positive tumors in vitro (25). In vivo, IL-2/TKD-activated effector cells reduced the tumorigenicity of Hsp70-positive colon carcinoma in immunodeficient mice (26). Other groups reported about an involvement of NK cells in the control not only of primary tumors but also of metastases (27, 28, 29, 30, 31).

In this study, we have established a xenograft tumor mouse model, closely resembling clinical features of locally advanced, metastasized human pancreatic carcinoma. In this model, we compared the ability of NK and T cells that had been stimulated ex vivo either with IL-2/TKD or with IL-2 alone to eradicate established Hsp70-positive pancreatic carcinomas and metastases. After o.t. injection of the adenopancreatic carcinoma cell line Colo357 (32) in SCID/beige mice, lacking functional mature T, B, and NK cells, tumor-bearing, metastasized mice were immunoreconstituted with either IL-2/TKD-activated or IL-2-activated human NK or T cells. A single i.v. injection of CD3⁻CD94⁺ IL-2/TKD-activated NK cells, but not of identically treated CD3⁺ T cells or IL-2-activated NK cells, resulted in a significant reduction in the weight of primary pancreatic carcinoma, a delayed onset of metastatic dissemination, and a significantly prolonged life expectancy. Furthermore, four repeated injections of IL-2/TKD-activated NK cells eradicated primary tumors and hepatic metastases.

Female and male C.B.-17/IcrHsd-scid-bg (SCID/beige) mice were obtained from an animal breeding colony (Harlan Winkelmann) and maintained in pathogen-free, individually ventilated cages (Tecniplast). The animals were fed sterilized, laboratory rodent diet (Meika) and used for experiments at 6–12 wk of age. All animal experiments had been approved by the Regierung der Oberpfalz and were in accordance with institutional guidelines.

After three i.p. passages through SCID/beige mice, a master cell bank (total, 1 × 10⁹ cells) was established from the human pancreatic adenocarcinoma cell line Colo357, provided by Dr. H. Graf (Schering; Ref. 32). Identical aliquots (5 × 10⁶) of this fast-growing (doubling time increased from 40 to 24 h) and highly metastatic variant were frozen in liquid nitrogen. After thawing, Colo357 cells were maintained in RPMI 1640 supplemented with 10% heat-inactivated FCS, 2 mM l-glutamine, 1 mM sodium pyruvate, and antibiotics (100 IU/ml penicillin, 100 μg/ml streptomycin) at 37°C in 5% CO₂. Single-cell suspensions were derived by short-term (<1 min) trypsin-EDTA treatment (all cell culture reagents were purchased from Life Technologies). Before injection viability of Colo357 cells, as determined by trypan blue exclusion assay, was always >95%.

For o.t. implantation, mice were anesthesized by i.p. injection of 0.03–0.04 ml of xylazine (6–8 mg/kg body weight; WDT) and ketamine (90–120 mg/kg body weight; WDT). The abdomen was sheared and sterilized with 2-isopropanol (70%). The skin and peritoneum were opened by a 1-cm incision along the linea alba caudal of the processus xyphoideus under strict aseptic conditions. The o.t. injection of either 25 or 10 μl of the Colo357 stock solution (100 × 10⁶ cells/ml) into the exteriorized pancreatic corpus using a 25-μl glass syringe with a 22-gauge needle (Hamilton) was visually controlled using a 7× stereomicroscope (Zeiss). After replacement of the pancreas into the abdominal cavity, the incision was closed in two layers using a 6-0 vicryl suture for the peritoneum and 7.5-mm suture clips for the skin (Fine Science Tools). This mouse model was found to be technically standardized and feasible.

Hsp70 membrane expression was routinely determined on Colo357 cells derived from cell cultures before injection, from single-cell suspensions of random tumor samples, and from patient-derived pancreatic tumor samples by flow cytometry, using a FITC-labeled mAb directed against the extracellular exposed epitope of membrane localized Hsp70 (cmHsp70.1; Multimmune). Briefly, after incubation of single-cell suspensions of viable tumor cells (0.2 × 10⁶ cells) with Hsp70-specific mAb for 30 min on ice and following two washing steps, nonfixed cells were analyzed on a FACSCalibur instrument (BD Biosciences). Isotype-matched control Abs and a human HLA class I-specific mAb (clone W6/32) were used as negative and positive controls.

CD3⁻CD94⁺ NK cells were selected from leukapheresis products obtained from healthy human volunteers using either CD94-biotin mAb (HP3-D9; Ancell Immunology Research Products) and anti-biotin magnetic microbeads (Miltenyi Biotec) or following a standard CD3/CD19 depletion protocol (Miltenyi Biotec), as described earlier (33). The purity of NK cells as determined by flow cytometry using directly conjugated CD3-FITC/CD16/56-PE (BD Biosciences) and CD94-FITC (Ancell) mAbs on day 2 after separation and stimulation was always >95%. Identical numbers (2 × 10⁶ cells/ml) of CD3⁻CD94+ NK and CD3+/CD94- T cells were incubated separately in RPMI supplemented with low dose IL-2 (100 IU/ml) plus TKD (2 μg/ml) (Bachem) or with IL-2 (100 IU/ml) alone for 4 days. TKD is a GMP grade 14-mer peptide of the C-terminal substrate-binding domain of human Hsp70 (TKDNNLLGRFELSG, aa 450–463) known to exert immunostimulatory activity on NK cells (25).

The cytolytic activity of CD3⁻CD94⁺ NK cells incubated with Hsp70 peptide TKD (2 μg/ml), with IL-2 (100 IU/ml) alone, or with a combination of IL-2/TKD for 4 days was determined in a standard ⁵¹Cr release assay (34) at E:T ratios ranging from 40:1 to 5:1. Cultured single-cell suspensions derived from Colo357 tumors that were surgically removed from SCID/beige mice on day 22 after tumor cell inoculation were used as target cells.

On day 4 after stimulation, both IL-2 and TKD were removed by intensive washing. Cells were then resuspended in RPMI at a cell density of 30 × 10⁶ cells/ml; 100 μl of this stock solution containing 3 × 10⁶ cells/mouse were injected i.v. into the tail vein of tumor-bearing mice at indicated time points. For repeated injections, activated NK or T cells were derived from one leukapheresis donor, but different animals received NK or T cells derived from different donors. Effector cells were injected into tumor-bearing SCID/beige mice only if viability was >95%.

After o.t. implantation of tumor cells, immunodeficient control mice and immunoreconstituted mice were sacrificed by craniocervical dislocation. The weight of the mice was then determined, the peritoneal cavity was inspected morphologically, and primary tumors and livers were excised in total. The weight of each primary pancreatic tumor was measured separately. Organs were kept on ice-cold, sterile PBS until needed for further investigations.

After determination of the weight, parts of the collected tumors and mouse livers, which were cut in equal pieces 4-mm thick, were fixed in Bouin’s solution containing 71.5% picric acid, 23.8% formaldehyde, and 4.7% acetic acid and embedded in paraffin. Consecutive section pairs of the liver and tumors (5 μm) were prepared from the ventral margin of each piece at a distance of 250 μm. The morphology of tumors and organs was visualized by standard H&E and Masson-Goldner staining; nuclei were costained in 1% Mayer’s hematoxylin (Dako).

For immunohistochemistry, the activity of the endogenous peroxidase was blocked with freshly prepared 1% hydrogen peroxide containing 0.1% sodium azide. Sections were then incubated with unconjugated anti-human Hsp70/constitutive Hsp70 (Hsc70) (rat; clone 7F4; provided by Dr. Elisabeth Kremmer, GSF-Institute of Molecular Immunology, Munich, Germany) at 4°C overnight. For the detection of effector cells, sections were heated for 30 min at 97°C and then incubated with anti-human CD94 (mouse; clone DX-22; IgG1; ebioscience), anti-human CD3 (mouse; clone OKT3; Ortho Biotech) or an IgG1 isotype-matched control (mouse; IgG1; Dako) mAb as described before. After a washing, a rabbit anti-rat or rabbit anti-mouse HRP-conjugated secondary Ab (Dako) was incubated for 2 h at room temperature. Diaminobenzidine (Dako) was used as a chromogen. Sections were counterstained with 1% Mayer’s hematoxylin (DakoCytomation) for 30 s and analyzed on an Axiovert 25 microscope (Zeiss).

The significance of the data was determined by using a nonparametric rank sum test (Mann-Whitney).

An o.t. injection of 2.5 × 10⁶ Colo357 pancreatic carcinoma cells suspended in 25 μl of medium resulted in aggressive, rapidly growing pancreatic tumors (Fig. 1,A). On day 30, the average tumor weight was 0.60 ± 0.14 g (n = 6); and due to cachexia, stenosis of the proximal ileum, and/or ductus choledochus, the animals rapidly became moribund. After injection of 1 × 10⁶ Colo357 cells, the average tumor weight on day 30 was 0.20 ± 0.08 g (n = 15), and first signs of disease were detectable from day 50 onwards with an average tumor weight of 0.43 ± 0.09 g (n = 6; Fig. 1,A). Tumor take at any tested time point, for both injection modes was always 100%. A comparative phenotyping of Colo357 cells from tissue culture and from single-cell suspensions derived from tumor-bearing mice on days 15, 22, and 30 revealed no significant differences in their Hsp70 membrane expression pattern; on average 60 ± 9.8% of the Colo357 cells were Hsp70 membrane-positive, as determined on at least three tumor biopsies at each time point. Single-cell suspensions of these Hsp70-expressing Colo357 cells were used as targets in a standard chromium release assay. As summarized in Fig. 1,B, Colo357 cells were lysed most efficiently by CD3⁻CD94⁺ NK cells that had been treated with low dose IL-2 plus TKD (IL-2/TKD) for 4 days. None of the two reagents alone exhibited a comparable stimulatory capacity on NK cells (Fig. 1 B). IL-2/TKD-activated T cells did not exhibit significant lysis of Colo357 cells (data not shown).

In vitro, the combined treatment of NK cells with IL-2/TKD was a prerequisite for the initiation of the migratory capacity toward Hsp70 membrane-positive tumors (21). With regard to these results, the question arose whether these NK cells were also attracted by solid Colo357 pancreatic tumors in vivo. We could show that 24 and 48 h after i.v. injection of ¹¹¹In-labeled IL-2/TKD-activated NK cells (3 × 10⁶) in tumor-bearing mice (day 22), an accumulation of the radioactivity was found in Hsp70 membrane-positive but not in Hsp70 membrane-negative tumors (data not shown). Next, identical numbers of unlabeled, IL-2/TKD-activated CD3⁺ T or CD3⁻CD94⁺ NK cells were injected i.v. into the tail vein of tumor-bearing SCID/beige mice on day 15 after o.t. injection of Colo357 cells (2.5 × 10⁶). The injection protocol is illustrated in Fig. 2,A. One week after the inoculation of the effector cells, mice were sacrificed and inspected for primary pancreatic tumors. Deparaffinized sections of the pancreas of animals before (day 0) and after (day 22) injection of Colo357 cells were incubated with the rat mAb 7F4, directed against human Hsp70/Hsc70. This mAb was highly specific for cells of human origin and showed no cross-reactivity with murine tissues (S. Stangl and G. Multhoff, manuscript in preparation). On day 0, the pancreas of healthy animals revealed no specific staining with 7F4 mAb (Fig. 2,B, upper left graph), whereas on day 22, after o.t. inoculation of 2.5 × 10⁶ Colo357 cells, major parts of the pancreas were positively stained for human Hsp70/Hsc70. In contrast, normal pancreatic and connective tissues of the mice that surrounded human tumor nodules remained unstained (Fig. 2 B, upper right graph).

The presence of NK cells was determined using a mAb directed against the C-type lectin receptor CD94, a surrogate marker for Hsp70-reactive NK cells; T cells were visualized with a CD3-specific mAb. As illustrated in Fig. 2,B, lower right graph, CD94⁺ effector cells were detectable within primary tumor nodules of immunoreconstituted animals. In contrast, an isotype-matched control Ab revealed no staining of tumor nodules and mouse pancreatic tissues (Fig. 2,B, lower left graph). Consecutive sections counterstained for CD3 (data not shown) indicated that predominantly CD3⁻CD94⁺ NK cells had the capacity to infiltrate Colo357 tumors in SCID/beige mice. These findings corroborated previous results of in vitro migration assays (21). A quantification of the number of infiltrating effector cells in tumor sections of mice reconstituted with either 3 × 10⁶ IL-2/TKD-activated NK cells or T cells revealed significant differences (Table I). On day 22, 1 wk after i.v. injection of the effector cells, the cell count showed 40 ± 14 infiltrating CD3⁻CD94⁺ NK cells but only 6.7 ± 9.4 CD3⁺ T cells. Five days later, on day 27, the number of infiltrating CD3⁻CD94⁺ NK cells further increased to 136 ± 79. The CD3⁺ T cells always remained low (20.3 ± 8.5) within tumor sections of a comparable size of 1 cm² (Table I).

To elucidate the cytolytic activity of tumor-infiltrating CD3⁻CD94⁺ NK cells in vivo, tumors were resected on day 22, 1 wk after the adoptive transfer of the effector cells. Representative photographs of resected tumors taken from three independent experiments, were viewed in Fig. 2,C; the average tumor weights of three to five independent experiments were shown below. On day 15, in the absence of injected effector cells, the average tumor weight was 0.11 ± 0.04 g (data not shown) and increased up to 0.31 ± 0.04 g on day 22 (Fig. 2,C). These values were in line with the results derived in the tumor growth kinetic shown in Fig. 1. One week after the injection of IL-2/TKD-activated NK cells, the tumor weight was significantly reduced (68%) to 0.10 ± 0.05 g (n = 3; p < 0.01). In contrast, identically stimulated T cells reduced the tumor weight by only 40% with the average tumor weight being 0.19 ± 0.05 g (n = 3; p < 0.1). Furthermore, NK cells that had been stimulated with IL-2 alone also lacked the capacity to control pancreatic tumors (n = 5; 0.22 ± 0.03 g; p < 0.1), thus indicating that TKD is a prerequisite for an efficient stimulation of the Hsp70-mediated antitumor immune response in NK cells.

Next, life expectancy of tumor-bearing mice was analyzed after a single i.v. injection on day 15 of differentially treated effector cells (3 × 10⁶ T or NK cells). Immunodeficient control mice showed first signs of tumor disease from day 18 onward; the maximum survival rate of animals inoculated with 2.5 × 10⁶ Colo357 cells was 37 days (Fig. 2,D). Injection of preactivated CD3⁺ T cells only marginally improved life expectancy, with T cell-treated mice showing first signs of disease by day 20 and all animals dying from progressive tumor disease on day 42. In contrast, a single injection of IL-2/TKD-activated NK cells significantly prolonged the survival of the mice, with >60% of the mice still alive on day 72 (Fig. 2,D). As already shown earlier, IL-2 alone was less efficient in the stimulation of an Hsp70-mediated antitumor immune response in NK cells; on day 52, all mice died from progressive tumor disease (Fig. 2 D).

Because pancreatic tumors in humans have a high potential for hepatic metastases, mouse livers were inspected by immunohistochemistry using the rat anti-human Hsp70/Hsc70-specific mAb 7F4. The injection protocol was identical with that shown in Fig. 2,A. As expected, the anti-human specific mAb did not stain normal mouse liver on day 0 preceding tumor cell inoculation (Fig. 3,A). However, by day 22 after o.t. transplantation of 2.5 × 10⁶ human Colo357 pancreatic carcinoma cells, liver metastases had developed, and large areas of the mouse liver stained positive for human Hsp70/Hsc70 (Fig. 3 A). Like primary pancreatic tumors, liver metastases were also found to be encapsulated by nonstained connective tissue of mouse origin.

Staining of consecutive sections with mouse anti-human CD94- and CD3-specific mAbs was performed to examine the presence of NK and T cells. As shown in Fig. 3,A, lower right graph, human CD94⁺ NK cells were detectable within metastatic lesions in the liver. In contrast, CD3⁺ T cells were not found in hepatic metastases after immunoreconstitution with IL-2/TKD-activated T cells (data not shown). An isotype-matched control mAb stained neither tumor nor normal liver tissues (Fig. 3,A, lower left graph). To examine the potentially protective effects of ex vivo IL-2/TKD-activated NK cells on metastatic dissemination, the number of immunohistochemically defined metastases was counted in liver sections of SCID/beige mice. For this purpose, the left lateral lobe of the liver was resected on days 5, 10, 15, 20, 25, and 30 after o.t. injection of 2.5 × 10⁶ Colo357 cells and on days 10, 20, 30, and 40 after injection of 1 × 10⁶ Colo357 cells. Each organ was cut into three identical parts to prepare three consecutive section pairs. The size of the counted metastases in the liver ranged between 50 and 4000 μm after o.t. injection of 2.5 × 10⁶ Colo357 cells (Fig. 3,B) and between 20 and 2500 μm after o.t. injection of 1 × 10⁶ Colo357 cells (Fig. 3,C) after staining with the human-specific mAb 7F4. Hepatic metastases were initially detected by day 5 after inoculation of 2.5 × 10⁶ Colo357 cells (Fig. 3,B) and by day 10 after inoculation of 1 × 10⁶ Colo357 cells (Fig. 3,C). On day 30, immunodeficient mice contained on average 4.8 ± 3.5 and 4.1 ± 2.8 liver metastases after injection of 2.5 × 10⁶ and 1 × 10⁶ Colo357 cells, respectively. The adoptive transfer of preactivated T cells on day 15 exhibited no significant inhibitory effect on the development of distant metastases in the liver compared with immunodeficient control animals. In contrast, a single dose of preactivated NK cells on day 15 resulted in a significant reduction of the number of metastases in the liver on days 20, 30, and 40 (Fig. 3, B and C, p < 0.01). The number of examined mouse livers at each indicated time point ranged from 10 to 13.

To further improve the antitumor activity, mice were repeatedly injected up to four times with effector cells. After o.t. injection of 1 × 10⁶ Colo357 cells, mice were reconstituted with either 3 × 10⁶ IL-2/TKD-activated T or NK cells once on day 15; twice on days 15 and 25; three times on days 15, 25, and 35; or four times on days 15, 25, 35, and 45 (Fig. 4,A). Separated T or NK effector cells for repeated injections of one mouse were obtained from one leukapheresis donor; different mice were treated with T or NK cells derived from leukapheresis products derived from different human donors. Every 10 days after the last effector cell infusion, mice were sacrificed and inspected for primary pancreatic tumors and hepatic metastases. As summarized in Fig. 4 B, three to four repeated NK cell infusions into the tail vein of tumor-bearing SCID/beige mice significantly reduced the weight of primary tumors. The minimal tumor load after the fourth injection round of NK cells was completely necrotic, as determined by histological analysis of H&E-stained tumor sections (data not shown). This finding was further supported by the fact that these mice did not develop recurrent disease. Immunohistological examinations of the liver using the human-specific Ab 7F4 revealed no metastatic dissemination in this organ on day 55 (data not shown). In contrast, the reduction of the tumor weights using IL-2/TKD-activated T cells was less pronounced after one to three injection cycles. On day 55, 10 days after the fourth i.v. injection of T cells, tumors showed progressive growth kinetics comparable with that observed in control animals. All control and T cell-treated mice died from progressive tumor disease between days 55 and 60 (data not shown).

Similar results were obtained in mice that had been treated o.t. with 2.5 × 10⁶ Colo357 cells and three repeated effector cell injections. However, due to the very rapid tumor growth, T and NK cells were given on a weekly basis. The experiment was terminated after three injections cycles because all control and T cell-treated animals were dead on day 36 (data not shown).

NK cells have been found to play a crucial role in the control of hematological malignancies (35). As previously described by our group, preactivation of CD3⁻CD94⁺ NK cells with IL-2 plus Hsp70 peptide TKD resulted in an enhanced cytolytic and migratory capacity toward Hsp70 membrane-positive tumor cells in vitro (25). Human pancreatic carcinomas, the fifth leading cause of cancer-related mortality, were frequently found to be Hsp70 membrane-positive (unpublished observation) and thus were susceptible to NK cell-mediated lysis in vitro (15). In this study, we addressed the question of whether Hsp70-reactive NK cells have an impact on highly aggressive pancreatic carcinomas in vivo. Because humans and mice differ not only in the C-terminal localized 14-mer Hsp70 sequence TKD (human: TKDNNLLGRFELSG; mouse TRDNNLLGRFELSG) which is used for the activation of NK cells but also in their NK cell receptor repertoire we established a xenograft SCID/beige mouse model to test the immunotherapeutic potential of human NK cells against human tumors. To mimic the growth characteristics of human pancreatic carcinoma, we initiated aggressive tumor growth and metastatic dissemination in SCID/beige mice by an o.t. injection of a fast growing variant of the human pancreatic adenocarcinoma cell line Colo357. The Hsp70 phenotype of the original and the fast growing variant remained identical. Growth characteristics after o.t. injection of either 2.5 × 10⁶ or 1 × 10⁶ Colo357 cells revealed similar results although severe signs of tumor disease, onset of metastatic dissemination, and mortality were delayed after inoculation of the lower cell numbers. Injection of 2.5 × 10⁶ Colo357 tumor cells resulted in tumor-related mortality from day 30 onwards and liver metastases starting by day 5. After injection of 1 × 10⁶ Colo357 mice died from day 55 onwards, and hepatic metastases were detectable from day 10 onward.

To resemble the clinical situation, an adoptive transfer of differentially activated NK and T cells was performed by i.v. injection on day 15 in mice with persisting tumor disease and hepatic metastases. In vitro experiments demonstrated that only IL-2/TKD-activated NK cells had the capacity to migrate toward Hsp70 membrane-positive tumor cells (21). The presence of CD3⁻CD94⁺ NK cells in tumor nodules and distant metastases confirmed the migratory activity of IL-2/TKD-activated NK cells in vivo. In contrast, IL-2 alone was less efficient in the stimulation of a Hsp70-mediated antitumor immune response in NK cells. Because we have recently shown that Hsp70 membrane-positive tumors actively release Hsp70 surface-positive lipid vesicles (36) with biophysical characteristics of exosomes (37, 38), we hypothesize that exosomal vesicles might be involved in the attraction of IL-2/TKD-activated NK cells. In contrast, 1–2 wk after i.v. injection of identically activated T cells, the amount of intratumoral CD3⁺ T cells was still low and completely absent in liver metastases. These findings are in line with the hypothesis that NK cells provide the first line of defense in cancer immunity.

In vitro, the amount of membrane-bound Hsp70 on tumor cells could be correlated with the sensitivity to lysis by IL-2/TKD-activated NK cells (15, 21, 23). In this study, we addressed the question whether Hsp70-reactive NK cells have the capacity to eradicate primary tumors and distant metastases originating from Colo357 cells. These tumors exhibited an Hsp70 cell surface-positive phenotype on ∼60% of the cells (21). As early as 1 week after adoptive transfer of IL-2/TKD-activated NK cells, weights of primary tumors were significantly reduced. In contrast, identically treated T cells and NK cells stimulated with IL-2 alone showed only a moderate reduction in the mass of primary Colo357 tumors.

Interestingly, a single i.v. injection of IL-2/TKD-activated NK cells not only inhibited primary tumor growth within the first week after injection but also significantly prolonged the life expectancy of tumor-bearing mice. These data indicated that even in immunodeficient SCID/beige mice lacking functional T and B cells but also IL-15, which is required for the NK cell development, human NK cells exerted their protective ability for an extended period of time. Again the Hsp70 peptide TKD was a prerequisite for an efficient stimulation of an antitumor immune response in NK cells.

It has been shown by others that NK cells play important roles in the control of distant metastases in animal models and in humans (27, 28, 29, 30, 31). In the present study, we observed an inhibitory effect of IL-2/TKD-activated NK cells on the metastatic spread in the liver when compared with control and T cell-treated animals. Furthermore, metastasis-free survival rates could be correlated with an enhanced NK cell activity in a clinical phase I trial (28).

NK cell-treated mice had a later onset of severe signs of tumor disease compared with control mice or mice injected with T cells. These data also suggested that the partial immunoreconstitution with IL-2/TKD-activated NK cells has a life-prolonging effect. With regard to recently published data (39), studies are ongoing analyzing as to whether the addition of human DCs and T cells might further improve the antitumor effects in our xenograft mouse model.

To evaluate the relevant dose of NK cells necessary for a complete tumor control the infusions of IL-2/TKD-activated NK cells were repeated. After four injection cycles, primary tumors as well as distant metastases were eliminated. In contrast, repeated T cell injections controlled primary tumors only until day 45; then tumors grew progressively. These findings might be explained in part by the fact that NK cells have higher intracellular glutathione levels than do T cells (40), which might enable a prolonged survival and an enhanced cytolytic activity of NK cells in immunodeficient mice. In summary, our data provide evidence that even in IL-15-deficient hosts IL-2/TKD-activated NK cells control primary pancreatic tumors and hepatic metastases.

In a recently published clinical phase I trial, the tolerability, feasibility, and safety of IL-2/TKD-activated NK cells were tested in patients with colorectal and lung carcinoma. Apart from an excellent safety profile of these effector cells, we were able to successfully demonstrate that repeated treatment cycles resulted in clinical responses even in multiply metastasized patients, refractory to standard chemo/radiotherapy (41). Again, low dose IL-2 alone was inefficient in stimulating the Hsp70-reactivity in patient-derived NK cells.

We thank Lydia Rossbacher for excellent technical assistance, Dr. Hermann Graf (Schering) for fruitful discussions, and Prof. Anne Dickinson (University of Newcastle, Newcastle upon Tyne, United Kingdom) for proofreading of the manuscript.

The authors have no financial conflict of interest.