Tumor-Specific Targeting of Pancreatic Cancer with Shiga Toxin B-Subunit

Pancreatic carcinoma is one of the most aggressive tumor entities with a median survival of 6 months and a 5-year survival rate of less than 5% (1). Curative resection often is not possible due to late diagnosis of the tumor, and the current standard chemotherapy with gemcitabine has increased the overall survival only modestly (2, 3). The targeting of pancreatic cancer for early diagnosis (e.g., by imaging methods) and for therapeutic purposes (e.g., by specific delivery of cytotoxic compounds to pancreatic tumor cells) is therefore of great clinical interest. The nontoxic B-subunit of the bacterial Shiga toxin (STxB) is a promising new option for diagnosis and therapy of gastrointestinal tumors. We and others showed previously that the expression of the cellular receptor of Shiga toxin, the glycosphingolipid Gb₃ (CD77), is significantly increased in colorectal carcinoma and their metastases (4, 5). Studies on genetic mouse models for human digestive cancer showed that STxB is suitable for noninvasive tumor imaging (6). STxB is taken up rapidly by tumor cells that express the receptor Gb₃ on their surface membrane, and STxB then follows the so-called retrograde route (7). The retrograde route bypasses the late endocytic pathway and avoids the degrading environment of the lysosomes. Instead, STxB is transported from the endosome to the trans-Golgi network and then to the Golgi apparatus and is transported from there to the endoplasmic reticulum (8–10).

It is feasible that STxB has naturally evolved to withstand inactivation by intestinal fluids. Moreover, STxB has only low immunogenicity in humans (11, 12). Functionalized STxB can be used efficiently to target gastrointestinal tumors in animal models (6), as well as short-term primary cultures from human colon cancer and liver metastases (5). Moreover, a recent study confirmed the findings on Gb₃ overexpression in colon cancer and reported an upregulation of Gb₃ expression in human pancreatic cancer patients (13). In the present study, we have quantified Gb₃ expression in human pancreatic adenocarcinoma by lipid extraction and thin layer chromatography (TLC), as well as by immunofluorescence microscopy on tissue sections, and tested pancreatic cancer cells for the uptake kinetics of functionalized STxB. Moreover, human astrocytoma xenografts in nude mice are eliminated rapidly, completely, and with long-term efficacy using verotoxin, supporting the concept of therapeutic use of Shiga toxin–like toxins (14). Recently, successful coupling of STxB to 7-ethyl-10-hydroxycamptothecin (SN38), the active metabolite of the topoisomerase inhibitor irinotecan (CPT-11), has been reported (15). SN38 is amenable to biochemical modification, making it a useful candidate drug for a proof-of-principle approach for STxB-mediated tumor targeting. Furthermore, SN38 is several orders of magnitude more active than irinotecan, but its hydrophobicity and insolubility in most physiologically compatible and pharmaceutically acceptable solvents limits its clinical application (16, 17). Irinotecan has already been reported in several clinical trials as second-line chemotherapy of advanced pancreatic cancer either as single agent or in combination with gemcitabine or other agents (2, 18–20). Anticancer cytotoxicity was reported, even though no obvious survival benefit was reported so far. The STxB-SN38 compound has been shown previously to possess highly increased cytotoxicity as compared with irinotecan when tested on a colorectal cancer cell line (15). Moreover, the STxB-SN38 compound showed high receptor specificity, as the cytotoxic effect relied on an intracellular uptake that is mediated by the receptor Gb₃. The cytotoxic effect was inhibited when the glycosphingolipid Gb₃ was blocked (15). Therefore, STxB-SN38 could be a highly useful tool for targeted chemotherapy, with greatly enhanced cytotoxicity and low side effects. The aim of this study was therefore to analyze the potential diagnostic and therapeutic uses of STxB on pancreatic carcinoma. Taken together, we report a significantly increased expression of Gb₃ in pancreatic cancer, as compared with matched normal tissue samples. In addition, we have successfully tested a new compound for targeted chemotherapy, STxB-SN38, on pancreatic cancer cell lines.

Samples were obtained from 28 patients with pancreatic carcinoma admitted to our Surgical Department in 2003 to 2006 (16 male and 12 female, median age: 68 years, range: 42–83). Informed, written consent about the use of tissue samples had been obtained previously, with approval of the local ethics committee. Median follow-up after surgery was 16.9 months. During this period, 3 patients died because of unrelated causes and 17 died because of cancer progression or disease recurrence. Tumors were classified according to the International Union Against Cancer, as summarized in Table 1 (21). Histology-guided sample selection was conducted by a pathologist to identify tumor samples and to exclude cancer cell contamination in the control samples. All samples were snap frozen in liquid nitrogen immediately after resection and stored at −80°C until use.

To allow for chemical coupling of fluorophores or cytotoxic drugs to a defined acceptor site in STxB, a Cys residue was added to the C-terminus of the wild-type protein. The recombinant mutant STxB-Cys protein was produced in endotoxin-free form as described (6). Antibodies and reagents used were as follows: anti-golgin p97 (Molecular Probes); anti-STxB (22); anti-CD31 (BD Biosciences Pharmingen); DAPI, Ki67, and staurosporine (from Sigma-Aldrich); and calnexin and anti-vimentin (both from Santa Cruz Biotechnology). Secondary antibodies coupled to fluorophores were purchased from Jackson Immunoresearch and cell culture reagents from Invitrogen. Irinotecan was obtained from the Pharmacy of the Klinikum rechts der Isar. SN38 was coupled to STxB as described (15).

Immunostaining was conducted as described before (6). Briefly, after paraformaldehyde fixation, cell lines on coverslips were permeabilized with 0.1% Triton X-100, blocked with PBS containing 2% bovine serum albumin, and primary antibodies in blocking buffer were added before counterstaining with secondary antibodies. At this concentration, Triton X-100 did not alter the Gb₃ detection on cryosections. STxB was purified from bacteria, as previously described (23). Covalent coupling of STxB to fluorochrome Cy3 (Amersham Biosciences) was carried out according to the supplier's instructions. For staining of endogenous Gb₃, tissue cryosections were fixed with 3% paraformaldehyde for 20 minutes and incubated with STxB-Cy3 for 30 minutes at a final concentration of 10 μg/mL, in PBS containing 0.2% bovine serum albumin. For counterstaining against the nuclear antigen Ki67, 0.1% Triton X-100 was used on cryosections. For image acquisition, epifluorescence or confocal microscopes (Zeiss) were used. Images were processed using Adobe Photoshop Software CS3 and colocalization was tested with ImageJ version 1.42q (NIH).

Gb₃ expression was quantified as previously described (refs. 5, 24–26; additional details in Supplementary Material). Briefly, unfixed tissue of pancreatic tumors and adjacent tissue were collected immediately after surgery by an experienced pathologist. Tissues were weighed and mechanically homogenized in 1 mL of aqueous buffer. The established pancreatic and colorectal cancer cell lines were collected from cell culture dishes, washed, and analyzed accordingly. The material in aqueous buffer was injected immediately into 3.75 mL of chloroform/methanol (1:2). After mixing, 1.25 mL of chloroform and 1.25 mL of water were added. The hydroalcoholic phase was washed once with 1.5 mL of chloroform. The combined chloroform phases were dried under nitrogen, and lipids were saponified at 56°C for 1 hour in 1 mL of methanol/KOH. The saponification reaction was once again extracted as described above, and the chloroform phase was washed once with methanol/water (1:1). The isolated neutral glycolipids were spotted on high-performance TLC plates (Merck) and separated with chloroform/methanol/water (65:25:4). Dried plates were soaked in 0.1% polyisobutylmethacrylate in hexane, floated for 1 hour in blocking solution, followed by incubation with STxB (20 nmol/L), primary polyclonal antiserum against STxB, and secondary horseradish peroxidase–coupled antirabbit antibodies. Reactive bands were revealed with enhanced chemiluminescence (Amersham Pharmacia Biotech).

Pancreatic (DanG, MiaPaCa2, and BxPC3) and colorectal (DLD1 and HT29) cancer cell lines were kept in Dulbecco's Modified Eagle's Media (DMEM) with fetal calf serum (FCS; 10% for BxPC3, 7% for all other cells), 1% Pen/Strep, and 1% glutamine. STxB covalently labeled with the fluorophore Cy3 was added at final concentration of 2.5 μg/mL as published earlier (5). After various time points, cells were fixed with 3% paraformaldehyde and analyzed by immunofluorescence as specified (see above). Cell lines were authenticated by the German cell line repository (DSMZ).

Cell growth analysis was conducted using the Cell Proliferation Kit II (XTT) from Roche Diagnostics, the amount of dye corresponding to the number of metabolically active cells. The cell lines were split into 6-well plates (at 1 × 10⁵ cells per well) and grown for 3 days. Treatment with irinotecan or STxB-SN38 was carried out for 6 hours in cell medium. Cells were washed with DMEM and substituted with fresh medium for an additional 48 hours. Cells were then seeded onto 96-well plates (in triplicates at 0.5 × 10⁴ cells per well and 1 × 10⁵ cells per well) in medium containing 0.5% FCS. After 12 to 18 hours, the XTT solution was added in medium with 3% FCS, and the cells were incubated for 30 hours. The amount of formazan was quantified with a Mithras LB 940 microplate reader (Berthold Technologies) at 450 to 500 nm. Mean values of 5 independent analyses are shown. A competition assay was conducted with 5 μmol/L STxB-SN38 or 50 μmol/L unlabeled STxB or 50 μmol/L of unlabeled STxB 5 minutes before treatment with 5 μmol/L STxB-SN38.

Statistical analyses were conducted using SPSS (version 16.0), Graph Pad Prism 5, and GraphPad InStat3 (Graph Pad Software). Data are presented as mean ± SD or where specified as median (range). The comparison of Gb₃ expression in pancreatic adenocarcinoma and corresponding nondiseased pancreas was done using paired Student's t test. For this calculation, the 27 pairwise complete observations were used. For evaluating the influence of Gb₃ expression on survival time, Cox regression was calculated. The Mann–Whitney U tests were used for evaluating the correlation of Gb₃ expression with histopathologic data. All statistical comparisons were done at a 0.05 level of significance. The IC₅₀ values were calculated using GraphPad Prism 5, a general-purpose curve fitting and scientific graphics program.

Previously, we have shown that the glycosphingolipid Gb₃ is overexpressed in human colorectal cancer (5). Here, we have analyzed surgically resected pancreatic adenocarcinoma and corresponding nondiseased pancreas by lipid extraction and TLC. Gb₃ expression was detectable in all tumors tested (n = 28; Table 1). Average median expression was 54.9 ng Gb₃/mg tissue as compared with 29.0 ng Gb₃/mg tissue in normal pancreas (n = 27). The mean difference of 31.2 ng Gb₃/mg tissue was highly significant (95% CI: 14.8–47.7, P = 0.0006). In a direct comparison of matched tissue samples from individual patients, we could detect an increased Gb₃ expression in 78% (21 of 27) of the tumors as compared directly with the surrounding normal pancreas (Fig. 1A, Supplementary Fig. S1). In 48% (13 of 27) of the samples, the Gb₃ level was higher than the calculated median expression of 54.9 ng Gb₃/mg tissue. In addition, established pancreatic cancer cell lines were tested and all showed Gb₃ expression (Fig. 1B). However, Gb₃ expression in cell lines was lower than in primary patient samples. Thus, Gb₃ was clearly upregulated in pancreatic adenocarcinoma, in good accordance with recent observations (14). HT29 colon cancer cells, which have been described to bind STxB (16), had 40.9 ng Gb₃/mg wet weight (not shown), whereas DLD1 colon cancer cells completely lacked Gb₃. Our statistical analysis showed no correlation between Gb₃ expression and histopathologic or survival data. In fact, the HR for the level of Gb₃ expression and survival time was 1.005. The correlation with histopathologic data involved the tumor size (pT, P = 0.78) and lymph node metastasis status (pN category, P = 0.52) as well as grading (G, P = 0.21), lymphangiosis (P = 0.19), and angioinvasion (P = 0.48; Supplementary Table SI). For further testing on differential expression, the parameter Gb₃-delta was devised, a calculation based on the level of Gb₃ in carcinomas minus Gb₃ expression level in normal tissue for each of the 27 matched pairs. In normal pancreatic tissue, Gb₃ was detectable by lipid extraction and by STxB overlay on cryosections, albeit at lower levels than in carcinoma samples. However, comparison of Gb₃ delta with histopathology failed to show significant association (Supplementary Table SII). Thus, Gb₃ seems to be broadly expressed in pancreatic neoplasm, irrespective of grading or tumor stage.

Analysis of tissue sections indicated that pancreatic cancer cells, but not stroma components, were the major source of upregulated intratumoral Gb₃ levels. Previous work showed that endothelial, immune, and enteroendocrine cells express Gb₃ (5, 6, 27). Moreover, tumor cells of epithelial origin, as well as tumor-associated blood vessels and stroma are the source of increased Gb₃ expression in colorectal adenocarcinoma (5). In the present study, tissue from patients with high (n = 3) or low Gb₃ expression (n = 3) was analyzed by immunohistochemistry (Fig. 2), and in parallel, by TLC. Despite different sensitivities, both detection methods were in good accordance. In normal pancreas, Gb₃ expression was mainly detected in endothelia (Fig. 2A–C). In tumors with low Gb₃ expression, stromal components like blood vessels were the major source of Gb₃ expression (Fig. 2D–F). However, in tumors with high Gb₃ expression, duct-like cancer cells were strongly stained with fluorescently labeled STxB (Fig. 2G–L). Tumor-associated endothelial cells were only weakly stained for STxB (Fig. 2F and J). No association of tumor cell growth with Gb₃ expression was detected in carcinoma with Gb₃ overexpression (Supplementary Table SIII). However, in carcinoma with low Gb₃ expression, the few cells that were positive for Gb₃ were negative for the proliferation marker Ki67 (Supplementary Table SII).

Next, we investigated whether pancreatic cancer cells actively take up STxB along the retrograde route. Incubation of pancreatic cancer cell lines with fluorescently labeled STxB-Cy3 at 4°C allowed binding to the receptor in the absence of cellular uptake and resulted in STxB staining at the plasma membrane (Fig. 3). Upon incubation at 37°C, STxB-Cy3 was rapidly internalized and colocalized with the Golgi marker Golgin p97 or with the endoplasmic reticulum marker calnexin (Fig. 3, Supplementary Figs. S2 and 3), indicating that STxB followed the retrograde route to the Golgi apparatus. Two days after the initial uptake, STxB was still detectable in punctate structures, colocalizing with the Golgi in DanG and MiaPaCa2 cells. However, no obvious colocalization of STxB could be detected with Golgi structures or endoplasmic reticulum in BxPC3 cells after 48 hours (Supplementary Fig. S2). In accordance with our findings on colorectal cancer cells, we did not observe an induction of apoptosis by STxB in any of the cell lines tested, as evidenced by staining for cleaved caspase-3 (data not shown). The cell line DLD1, which lacks Gb₃, did not show STxB binding and uptake (Fig. 3). The percentage STxB-positive cells was quantified after 60 minutes: 19% of BxPC3, 20% of DanG, 35% of MiaPaCa2, compared with 56% for HT29 colon cancer cells (positive control), and 0% of DLD1 cells (negative control; not shown). Supplementary Fig. S3 shows the distribution of STxB uptake in analyzed cell lines at the tested time points, distinguishing between plasma membrane, Golgi, and vesicular staining.

Because STxB was taken up efficiently by pancreatic cancer cells, we investigated the feasibility of STxB-mediated chemotherapy with a topoisomerase I inhibitor coupled covalently to STxB (15). SN38 is linked to STxB via a cleavable linker arm that allows for SN38 release from STxB in membrane of the endoplasmic reticulum or Golgi apparatus. On pancreatic cancer cells, the cytotoxic effects of this compound were compared with the effects of the standard drug irinotecan (Fig. 4). The left column indicates cell growth in response to increasing concentrations of irinotecan, the right column shows cell growth after treatment with STxB-SN38. All pancreatic cancer cell lines showed significant growth inhibition upon treatment with irinotecan, and cell growth was more strongly inhibited by STxB-SN38 (right column), with IC₅₀ values in the submicromolar range (Supplementary Table SII and Supplementary Fig. S4). The inhibitory concentration 50% (IC₅₀) is the concentration that exerts half the specific cytotoxicity. Moreover, even when cells were treated with the chemotherapeutic agent for only 15 minutes, STxB-SN38 showed a 20-fold greater efficacy than irinotecan (Supplementary Fig. S5). Irinotecan had a cytotoxic effect on DLD1 cells that lack the receptor Gb₃, whereas STxB-SN38 did not significantly inhibit cell growth (Fig. 4). At highest concentrations of STxB-SN38, a reduction of cell growth was observable for DLD1 cells, which may be due to nonspecific pinocytosis. Figure 5 shows the results of a competition assay for MiaPaCa2 cells. Treatment of cells with unlabeled STxB induced only a slight change in cell growth whereas STxB-SN38 had a pronounced cytotoxic effect. After prior treatment with a 10-fold molar excess of unlabeled STxB, the effect of STxB-SN38 was significantly diminished. Even though the competition with unlabeled STxB did not completely block the cytotoxic effects of STxB-SN38, this indicates that the effect of STxB-SN38 is mediated specifically by the Gb₃ receptor.

In the present work, we show the use of the nontoxic STxB for specific targeting of human pancreatic cancer cells. Pancreatic cancer is one of the most aggressive tumor entities and represents the fourth leading cause of cancer-related death in Western countries. So far, the only curative therapy is surgery, but because of late diagnosis, most patients do not qualify for this treatment. The standard chemotherapy with gemcitabine and chemotherapeutic combinations improves cancer-related symptoms but enhances survival only modestly (2, 20). Therefore, new options for diagnostic and therapeutic approaches are urgently needed. Previously, we were able to show that STxB allows efficient targeting of colorectal carcinoma due to an increased expression of its specific receptor, the glycosphingolipid Gb₃ (or CD77; refs. 5, 6). The binding kinetics of STxB to the receptor Gb₃ are complex. The pentameric STxB has 3 binding sites per monomer. Thus, up to 15 Gb₃ molecules per pentamer are predicted to bind to STxB (7, 28). When bound to Gb₃, STxB is internalized and transported to the Golgi apparatus and the endoplasmic reticulum via the retrograde pathway, avoiding degradation in the lysosomes (8–10). The protein is detectable in cancer cells for as long as 5 days (5, 7, 22, 23).

Consistent with recent data, we detected increased levels of Gb₃ in 78% of pancreatic carcinoma compared with nondiseased pancreas (n = 21 of 27 cases; ref. 13). Distler and colleagues reported overexpression of Gb₃ in 62% of pancreatic carcinoma (n = 21) based on a newly developed mass spectrometry method (13). We used a TLC approach, which yielded an absolute quantification of Gb₃ levels, based on wet weight of the tissue samples analyzed (5). The median expression of Gb₃ for nondiseased pancreas was 29 ng/mg of tissue, as opposed to 55 ng Gb₃/mg tumor tissue. However, the data reported here show differences to previous reports, which may be explained by the relatively small sizes of the patient collectives. The difference in expression between normal and carcinoma tissue was statistically highly significant in our case (P = 0.0006) but not in a previous study (13). Gb₃ expression has been reported to lack correlation with any clinicopathologic parameter, except with tumor differentiation (13). This indicated high levels of Gb₃ in less differentiated tissue. However, we could not confirm a correlation between tumor grading and Gb₃ expression on our collective data (P = 0.21). Furthermore, our analysis provided no evidence for a correlation of Gb₃ expression with clinical or pathologic parameters. Survival analysis showed no significant difference concerning Gb₃ levels. This is in accordance with our earlier findings obtained on colon cancer (6) and indicates that elevated Gb₃ expression is a frequent phenomenon in malignant gastrointestinal tumors, irrespective of stage and differentiation. The pathophysiologic role of upregulated Gb₃ expression in tumors is still unclear (29, 30). Recently published data show that cisplatin induces Gb₃ expression in cancer cells and that Gb₃ expression is linked to acquired cisplatin resistance (31). Moreover, increased Gb₃ levels were correlated to increased expression of multidrug resistant 1/permeability glycoprotein (MDR1/PgP), a glycoprotein involved in the biosynthesis of glycolipids such as Gb₃ (29). Gb₃ and MDR1/PgP partially colocalize, with Gb₃-containing lipid rafts being crucial for intracellular MDR1/PgP surface trafficking. Furthermore, elevated levels of Gb₃ were described in drug-resistant cancers (32, 33) and a functional interplay between MDR1/PgP and membrane Gb₃ was suggested.

Importantly, carcinoma cells were the major Gb₃ source within the tumors. In contrast, the stroma was essentially negative for Gb₃. Therefore, the large stroma content that is frequently found in pancreatic cancer may lead to an underestimation of the actual Gb₃ levels. Interestingly, not all tumor cells within a given sample were positive for Gb₃. This heterogeneous expression pattern is in accordance with earlier findings, reporting an association of the invasive status of colorectal cancer cells with high Gb₃ expression (4). Even though Gb₃ expression may be restricted to a subset of pancreatic cancer cells, a therapeutic intervention based on vectorized STxB could still yield a clinical benefit, for example, by eliciting an antitumoral immune response caused by dying tumor cells. In the present study, an association of the proliferative status of cancer cells with Gb₃ levels was tested. However, there was no clear correlation of Gb₃ content with the proliferation marker Ki67. The Gb₃ levels found in normal tissue are most likely due to endothelial cells, immune cells, and myofibroblasts, which have been reported to express Gb₃ (24, 34, 35). We validated the endothelial expression via CD31 staining of tissue sections. Endothelial cells in normal tissue were positive for Gb₃ and intratumoral blood vessels were not strongly marked. However, we and others described a strong Gb₃ expression in tumor-associated neovascularization in human bladder cancer and colorectal carcinoma (35–38). Our findings indicate a less pronounced expression of Gb₃ in intratumoral blood vessels in pancreatic cancer, which may result from organ-specific variances concerning tumor vascularization. However, it cannot be excluded that Gb₃ may also be expressed by further cells at levels that are below the detection threshold in immunocytochemistry.

Here, we could validate an intracellular uptake along the retrograde pathway in pancreatic cancer cells, whereas no uptake of STxB was observed in DLD1 cells lacking the Gb₃ receptor. Colocalization of fluorescently labeled STxB and the Golgi marker Golgin p97 occurred within 60 minutes, and intracellular STxB was still detectable after 2 days, even though cell-specific differences were observable at late time points. Therefore, targeted chemotherapy mediated by STxB may be feasible on pancreatic cancer. As proof of principle, we used a previously established compound that consists of a topoisomerase type I inhibitor, coupled covalently to STxB (16). SN38 is the active metabolite of irinotecan, which inhibits the DNA topoisomerase I (15, 17, 39). Irinotecan is activated by hydrolysis to SN38, but only 2% to 8% of irinotecan is converted to SN38 in the liver and the tumor cells, which requires higher dosage to obtain therapeutic efficacy (40, 41). SN38 is far more efficient than irinotecan, but due to its hydrophobicity and low stability in serum, its clinical use is limited (40, 41). In vitro, the cytotoxicity was successfully shown for lung, ovarian, colorectal, and gastric cancer, and several trials investigate the clinical use of the agent (16, 17, 39, 32–43). Covalent coupling to STxB could be an effective means to stabilize SN38 (15) and has the advantage of greatly enhanced specificity. Only cells that express the Gb₃ receptor are targeted, and consequently, the clinical side effects could be expected to be greatly diminished. Endothelial cells or kidney epithelia express Gb₃ and might therefore be negatively affected by a cytotoxic STxB-based compound (44, 45). However, although Gb₃ levels and Shiga toxin (STx) binding does not change depending on patients age, the effects, in particular the hemolytic uremic syndrome, appear age dependent. Therefore, the renal damage caused by STx and conclusively in the case of therapeutic usage of STxB, the possible renal side effects cannot be estimated yet and should be evaluated (46).

The standard therapy for pancreatic cancer is gemcitabine (2, 3, 47, 48). However, irinotecan is currently analyzed as first-, second-, and third-line therapy for advanced and metastatic pancreatic cancer in gemcitabine refractory cases and showed slight but objective responses as single agent or in combination (FOLFIRI; refs. 21, 49, 50). We therefore decided to test STxB-SN38 as a novel therapeutic compound on pancreatic cancer cells. As positive control, STxB-SN38 was tested on the Gb₃-expressing colon cancer cell line HT29 (15). Proliferation assays showed cytotoxicity of STxB-SN38 on HT29 cells (not shown). To validate the receptor specificity, tests were conducted on DLD1, a Gb₃-negative colon cancer cell line. Reduced cytotoxicity was observed for STxB-SN38, whereas the cells responded to irinotecan. All pancreatic cancer cell lines tested were sensitive to treatment with irinotecan or STxB-SN38. Importantly, STxB-SN38 had a significantly stronger cytotoxic effect than irinotecan; the IC₅₀ values for STxB-SN38 were enhanced over 100-fold as compared with irinotecan, in accordance with earlier findings on colorectal cancer cells (ref. 15; Fig. 4). However, there was no positive correlation between the Gb₃ expression levels in the cell lines and sensitivity toward STxB-SN38. This may indicate that additional parameters, such as regulation of apoptotic pathways, may determine the cellular sensitivity to the cytotoxic compound. Moreover, a competition assay with a 10-fold molar excess of unlabeled STxB in competition with STxB-SN38 showed that the cytotoxicity is mainly mediated by Gb₃-triggered uptake. Taken together, STxB-mediated tumor targeting offers a promising new approach to diagnose or treat pancreatic cancer.

No potential conflicts of interest were disclosed.

The authors thank Alexandra Gnann and Widya Johannes for excellent technical assistance and Prof. Bernhard Holzmann for critical discussion.

M. Maak received funding from the Kommission für Klinische Forschung (KKF/MRI). L. Johannes was supported by Fondation InNaBioSanté (n°2009CS026) and Fondation Pierre-Gilles de Gennes pour la Recherche. L. Keller and K.-P. Janssen were funded by the Wilhelm Sander-Stiftung.

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.