The parental mouse MAb mG250 (IgG2a) and its human-mouse IgG1 chimeric version cG250 (Rencarex/Girentuximab) were provided by WILEX AG. The human CA IX PG domain-specific monoclonal antibodies M75 and IV/18, and CA domain-specific monoclonal antibodies VII/20, V/12 and V/10 were described previously (24). The mouse CA IX-specific MAb AM4 was also described earlier (45). Homosulfanilamide, a CA inhibitor with an IC50 of 0.1 mM, determined as inhibition of extracellular acidification mediated by CA IX in cell culture (7), was provided by Professor Claudiu T. Supuran (University of Florence). GST-CA (GST fused to full-length human CA IX protein) and GST-Car9 (GST fused to full-length mouse CA IX protein) were described before (24,45).

For the experiments described in this report, we used MDCK canine kidney cells permanently transfected with the full-length human carbonic anhydrase (CA9) cDNA in the pSG5C plasmid (MDCK-CA9) or with plasmids derived thereof encoding a CA domain deletion variant (MDCK-ΔCA), a PG domain deletion variant (MDCK-ΔPG) and a human alternatively spliced protein truncated in the C-terminal part of the CA domain (MDCK-hAS). Mock-transfected cells (MDCK-neo) were used as a negative control. The CGL3 tumourigenic cell line (HeLa x fibroblast hybrid) with high normoxic expression of CA IX and moderate induction by hypoxia (20) was kindly provided by Professor Eric J. Stanbridge (University of California, Irvine). Human HT-29 colorectal carcinoma cells (ATCC), which express CA IX at high endogenous level, were used for in vivo experiments. The cells were routinely cultured in DMEM with 10% FCS (BioWhittaker, Inc.). Hypoxic treatments were done in a hypoxic workstation (Ruskinn Technology, Ltd.) in a mixture of gases containing 2% O₂, 5% CO₂, 10% H₂, and 83% N₂.

Microplate wells were coated overnight at 37°C with the RIPA cell extracts diluted in PBS or 10 ng/well of GST fusion proteins. After blocking with 10% FCS in PBS, the coated wells were incubated with 10 μg/ml of mG250 or cG250, depending on the experimental setting. Peroxidase-labelled pig anti-mouse IgG or goat anti-human IgG (Sigma) were used as detectors.

Microplate wells were coated overnight with the following antigens diluted in PBS: purified CA I (200 ng/well), purified CA II (100 ng/well) and recombinant CA XII (100 ng/well), all kindly provided by Professor Seppo Parkkila (University of Tampere, Finland). Then, the coated wells were incubated with mG250 and with polyclonal sera against CA I, CA II and CA XII (all 1:1,000), respectively, as positive controls. Peroxidase-labelled pig anti-mouse IgG and pig anti-rabbit IgG diluted 1:5,000 (Sigma) were used as detectors.

An extract from MDCK-CA9 cells was adsorbed on microplate wells at a concentration corresponding to 50% of maximal binding of labelled MAbs. Coated plates were washed and saturated with 10% FCS in PBS. Serial 2-fold dilutions of purified mouse MAbs in 25 μl and a constant amount of biotinylated MAb in 25 μl were added and incubated overnight at 4°C. The plates were washed and peroxidase-labelled streptavidin (Pierce Biotechnology, Inc.) was used as a detector.

Microplate wells were coated with 50 μl/well of individual purified mouse MAbs diluted in PBS (200 μg/ml). After blocking, washing and incubation with the extract of MDCK-CA9 cells (1:50 in PBS), the set of biotinylated antibodies (5 μg/ml) was added. Binding of the detector MAbs was determined using peroxidase-conjugated streptavidin. Results were expressed as absorbance differences between the wells, in which CA IX antigen was present or absent.

CGL3 cells (with natural, hypoxia-induced expression of CA IX) and MDCK-CA9 cells (with constitutive, ectopic expression of CA XII) were plated in triplicates to wells of microplates and allowed to form a confluent monolayer overnight. Then the cells were transferred to hypoxia (2% O₂) for 24 h. The inhibitor homosulfanilamide was diluted in culture media and added to the cells in increasing amounts together with a constant amount of G250 MAb (200 μg/ml based on the saturation experiment) for the last 6 h-period of the hypoxic incubation. Medium without the inhibitor was added to the control sample. The cells were fixed with methanol (5 min at −20°C) and the amount of mG250 MAb bound to cells was determined by peroxidase-labelled anti-mouse IgG.

Tested mG250 MAb, M75 (as an anti-PG domain control) and VII/10 (as an anti-CA domain control) were bound to a 25 μl 50% suspension of Protein A Sepharose (Pharmacia) for 2 h at RT. Biotinylated extracts of MDCK-CA9, MDCK-ΔCA, MDCK-ΔPG, MDCK-hAS, and CGL3 cells (200 μl each), or media from biotinylated CGL3 cells treated with PMA (activator of shedding) and MDCK-hAS cells (1,500 μl) were pre-cleared with 20 μl of a 50% suspension of Protein A Sepharose and then added to bound MAb. Immunocomplexes collected on Protein A Sepharose were washed, boiled 5 min in Laemmli loading buffer and separated by SDS-PAGE on a 10% gel. Afterwards, the proteins were transferred onto a PVDF membrane and detected with peroxidase-conjugated streptavidin (1:1,000; Pierce Biotechnology, Inc.) followed by ECL.

HT-29 cells were incubated with 100 μg/ml mG250 antibody for 1 h at 37°C to recruit the maximum of MAb to CA IX at the cell surface. Subsequently, the cells were washed to remove unbound antibody and left in fresh medium at 37°C for the internalization of bound mG250 for different periods of time (0, 3, 24, 48 and 72 h). At the end of the internalization period, Fc portions of the antibody remaining/recycled on/to the cell surface were detected by Protein A conjugated with Alexa Fluor 488 (Invitrogen Life Technologies) diluted 1:100 for 2 h at 4°C to prevent continued internalization. Data acquisition and analysis were performed on a Guava flow cytometer using CytoSoft software.

Cells plated on sterile glass coverslips 24 h before the experiment were incubated with 10 μg/ml cG250 antibody for 30 min at 4°C to recruit the MAb to CA IX at the cell surface. Subsequently, the cells were washed to remove any unbound antibody and transferred to 37°C for various time intervals to allow for the internalization of CA IX-bound G250. Alternatively, the MAb was left in the medium throughout the experiment. At the end of the internalization period, the cells were washed and fixed in ice-cold methanol at −20°C for 5 min and sequentially treated for 1 h at 37°C with anti-human Alexa-conjugated antibody (Invitrogen Life Technologies) diluted 1:1,000 to detect internalized cG250. Finally, the cells were thoroughly washed, mounted onto slides in the Fluorescent Mounting Medium (Calbiochem), analysed with a Leica DM4500 B microscope and photographed with a Leica DFC480 camera or by a confocal laser-scanning microscope Zeiss LSM 510 Meta.

The therapeutic effect of MAb G250 was investigated in nude mice with tumour xenografts generated from HT-29 colorectal carcinoma cells. Animal handling was approved by the Slovak Veterinary Administration in accordance with EU regulations. For the immunotherapy experiment, 3 × 10⁵ HT-29 cells in 200 μl of PBS were grafted subcutaneously to the mouse back. Immediately (within 1 h) after tumour cell grafting, one group of mice received intravenous injections of 100 μg/dose of mG250 in PBS and the other two groups received PBS without the MAb. All subsequent injections were given into the tail vein twice a week throughout the tumour growth (39 days). To get closer to the situation in humans, where treatment is initiated only after detection of the growing tumour, a delayed mG250 treatment was started in the second group from day 10 after xenografting when the first tumours became palpable. Then the regimen of the treatment was the same as for the first group of animals. The third, control group received only PBS injections at the same time points as the antibody-treated animals. Tumour diameters were regularly measured and recorded. On the day of sacrifice, tumours were dissected, weighed and processed for immunohistochemistry. Serum samples were also collected from all groups to detect the CA IX ECD.

The capture antibody VII/38 (10 mg/ml, 100 ml/well) was immobilized on the surface of microplate wells overnight at 4°C. After blocking and washing, mouse serum samples were diluted 1:2 in PBS and added to the coated wells (100 ml/well) for overnight binding at 4°C. The attached antigen was then allowed to react with biotinylated M75 MAb diluted 1:5,000 (200 ng/ml) in PBS. The amount of bound detector antibody was determined after 1 h incubation with the peroxidase-conjugated streptavidin (Pierce Biotechnology, Inc.) using the peroxidase substrate ortho-phenylene diamine (Sigma).

Earlier studies showed that G250 MAb interacts with the native CA IX protein, but its domain specificity has remained unknown (22). Thus, we first analysed the binding of G250 MAb to CA IX variants with deletions of various parts of the molecule. ELISA with these CA IX-related antigens revealed that G250 MAb (both mouse and chimeric versions) bound only to those variants that contained an intact CA domain, including GST-CA and extracts from MDCK-CA9 and MDCK-ΔPG cells (Fig. 1A). Accordingly, mG250 MAb was capable of immunoprecipitating only the full-length IX CA protein and its ΔPG variant (Fig. 1B), but not the ΔCA variant lacking the catalytic domain, as was shown previously for other CA domain-specific MAbs (24). A competitive antibody-binding ELISA was then performed to analyse mutual relationships among these MAbs and derive the position of mG250 on the epitope map of the CA IX antigen. Only the CA domain-specific antibodies VII/20, V/12 and V/10 were able to compete for the binding to CA IX with labelled mG250 to an extent similar to the homologous competition by non-labelled mG250 (Fig. 1C). A capture-detection assay corroborated the relationships between the antibodies with minor differences attributable to the different arrangement of the assay (Fig. 1D).

CA IX protein exists in three naturally occurring forms, namely the full-length protein, the ECD shed to extracellular space and the alternatively spliced (AS) variant that is partially intracellular and partially secreted (20,21,46). The CA IX ECD can be detected in body fluids of cancer patients and appears to have a prognostic/monitoring value (47–53). On the other hand, the truncated AS variant is produced at a low level independently of hypoxia and tumour phenotype and for immunotherapy its interaction with CA IX antibodies is not desirable due to possible interference with the binding of the clinically relevant CA IX molecules (46). Indeed, ELISA and immunoprecipitation of CA IX variants from extracts of transfected cells or natural CA IX expressors demonstrated that mG250 MAb recognizes only the full-length CA IX protein and its extracellular domain shed to medium, but not the AS variant lacking the C-terminal part of the catalytic domain (Fig. 2A and B).

Since G250 binds to the CA domain, we analyzed whether it can also recognize related CA isoforms (particularly those that could be co-expressed with CA IX in tumour cells). We used ELISA to test the binding of G250 MAb (both mouse and chimeric versions) to purified CA I, CA II, and CA XII antigens, and found that G250 did not bind any of the CA isoforms tested (Fig. 2C). Moreover, neither mouse nor rat CA IX orthologs interacted with G250 suggesting that it specifically reacts with the human CA IX antigen (Fig. 2D). Therefore, it can be investigated in animal models with human tumour xenografts, but not with natural, induced, syngeneic tumours or mouse/rat xenografts.

The CA domain of CA IX is a globular structure containing the enzyme active site that is located in a large conical cavity with the catalytic zinc at the bottom (54). This active site can accommodate different types of carbonic anhydrase inhibitors (CAIs). Since CA IX activity plays an important role in tumour biology, these inhibitors have been investigated as clinically promising tools for detection and therapeutic targeting of hypoxia-activated CA IX (11,18,55). We previously showed that the FITC-conjugated homosulfanilamide inhibitor (FITC-CAI) binds to CA IX expressed in hypoxic cells (7). Here we examined whether interaction of cG250 MAb to the catalytic domain of CA IX is affected by FITC-CAI binding to the active site. To this end we used a competitive ELISA on monolayers of hypoxic cells, namely CGL3 cells with natural CA IX expression and MDCK-CA9 cells with ectopic CA IX expression. FITC-CAI was added in increasing amounts together with a constant amount of cG250 MAb (200 μg/ml based on the saturation experiment, data not shown). CAI did not block the internalization of cG250 either (Fig. 3C). Moreover, in both cell lines, FITC-CAI did not reduce cG250 MAb binding (Fig. 3A and B), indicating that cG250 does not bind in or close to the active site, but rather interacts with the backbone of the catalytic domain. This suggests that CA IX-expressing hypoxic tumours can be potentially subjected to a combination treatment with cG250 and CAI.

Binding of cG250 MAb to CA IX was shown to trigger receptor-mediated internalization, a process that has particular impact on the outcome of immunotherapy targeted to cancer-related antigens (56). Since we wanted to learn more about the temporal and mechanistic aspects of cG250 internalization, we followed the localization and fate of the internalized antibody for different time periods and under different conditions.

To analyze whether cG250 co-localizes with CA IX during the internalization path, the MAb was first recruited to the cell surface CA IX at 4°C, then unbound antibody was washed away, internalization was initiated at 37°C and allowed to proceed for 3, 6, 24 and 48 h. In all internalization periods, staining signals of cG250 antibody and CA IX antigen overlapped (Fig. 4A) suggesting that cG250 antibody remains associated with the antigen. Interestingly, intracellular staining of both cG250 and CA IX prevailed up till 24 h of internalization, whereas samples incubated for 48 h showed the membrane staining, apparently due to recycling of the cG250-CA IX complex back to the cell surface. Moreover, the analysis of cG250 internalization under hypoxia (2% O₂) and in acidic extracellular pH revealed that cG250 enters the intracellular space independently of these physiological conditions and confirms that it also remains associated with CA IX inside the cell (Fig. 4B). Interestingly, cG250 antibody remained associated with the cells at least for 3 days, even when the cells were split and re-plated (data not shown).

The distribution of immunotherapeutic antibodies in solid tumours reflects different physiological barriers including high cell density (57). We therefore examined internalization of cG250 in a highly packed cell monolayer under normoxia and hypoxia, respectively, using MDCK-CA9 cells with the constitutive expression of CA IX and CGL3 cells with high endogenous expression of CA IX. We found that cG250 MAb was unable to internalize into dense cells under normoxia, whereas hypoxia facilitated internalization presumably due to release of tight cell-cell contacts (Fig. 4C). This suggests that CA IX present on the surface of hypoxic and/or HIF-activated cancer cells can mediate antibody-induced internalization despite high local cell density.

We then investigated whether cG250 MAb is able to undergo consecutive internalization cycles via newly produced and/or unoccupied CA IX antigen present on the cell surface. The experimental scheme utilizing consequent green- and red-labelled G250 MAbs was set as shown in Fig. 5A. Confocal microscopic analysis revealed that both antibodies bound to CA IX and internalized irrespective of the time interval between their addition to cells (see Fig. 5). Interestingly, there were no big differences in the subcellular localization of the 1st versus 2nd antibody, and both antibodies showed considerable co-localization. A slight difference could be seen in the 1-h sample, where the red signal seemed to dominate in the intracellular space, while the green signal prevailed on the cell surface. In the samples with 24 and 72 h internalization period, both red and green signals overlapped, while a slight red intracellular staining was visible again in the sample with the 48 h internalization period. These very subtle differences might reflect partly asynchronous movement (e.g., time-shift in internalization-recycling) of the red and green MAbs in the cells.

To gain insight into the fate of internalized cG250 we used several inhibitors and markers of the internalization pathways. We could see that the pattern of cG250 internalization is very similar to transferrin (TR) a molecule known to enter cells in a clathrin-dependent manner (Fig. 6A). Consistent with this, nystatin, a cholesterol-aggregating inhibitor of clathrin-independent endocytosis, did not have any effect on cG250 internalization (data not shown). Moreover, no major differences in cG250 internalization patterns were observed following treatment with the lysosomotropic agent concanavalin A (data not shown). Indeed, cG250 MAb showed no or only a minor overlap with LysoTracker Red lysosomal marker in support of the idea that the major part of the antibody escapes lysosomal degradation and enters the recycling pathway (Fig. 6B). Finally, we performed internalization of cG250 in the presence of nocodazole, a microtubule-disrupting agent, which reduces the transit of internalized molecules to the perinuclear compartment and thereby inhibits their recycling. This led to loss of perinuclear localization of cG250, diminished recycling to the plasma membrane and to its diffuse distribution throughout the entire cytoplasm (Fig. 6C).

We further evaluated the internalized/recycled cG250 MAb with respect to the integrity of its Fc portion, which mediates ADCC response, i.e., directs cytotoxic activities of effector cells against the target cell. We first examined the capacity of cG250 to bind Protein A in immunofluorescence analysis, which showed that the antibody can bind Protein A at each point of the internalization cycle, i.e., initially on the cell surface, then inside the cells and finally on the cell surface again (Fig. 7A). Flow cytometric analysis of HT-29 cells that highly express CA IX revealed that the percentage of Protein A-positive cells (in which Protein A was bound to cG250 MAb attached to CA IX at the surface of living cells) decreased from initial 72 to 55% after 3 h internalization and to 18% after 24 h (Fig. 7B). This decline agrees with the strength of the intracellular signal observed in immunofluorescence and suggests a reduction of the antibody on the cell surface as a consequence of internalization. On the other hand, gradual elevation of the Protein A-binding cells to 31% after 48 h and 45% after 72 h suggests that the antibody recycled back to the plasma membrane, although it did not reach the initial level (Fig. 7B). This reduction could be due either to partial degradation of the internalized antibody or, more conceivably, to cell division during the long internalization periods that resulted in the relative dilution of the staining signal. Undoubtedly, these results confirm that the recycled antibody possesses an intact Fc portion.

In order to learn whether G250-based immunotherapy can be useful in tumours other than RCC, we investigated its anticancer effect against HT-29 colorectal carcinoma xenografts implanted in nude mice. This model was chosen because HT-29 cells display high expression of CA IX even under normoxia. Although hypoxic induction of CA IX in monolayer culture of HT-29 cells is relatively low, its distribution in xenografts displays a typical hypoxic pattern with high perinecrotic expression of CA IX and low or no expression around vessels as demonstrated earlier (36).

The xenografted animals were subjected to immediate as well as to delayed treatment with mG250 as described above in Materials and methods. Both groups showed a significant reduction of tumour volume and tumour weight when compared to control, placebo-treated animals suggesting that mG250 is capable of eliciting not only a protective, but also a real therapeutic anti-cancer effect in a setting similar to the treatment of cancer patients (Fig. 8A and B).

We also wanted to find out whether the presence of the HT-29 tumour is reflected in serum levels of the CA IX ECD. For this purpose, we adopted a capture-detection ELISA using a combination of the CA domain-specific VII/38 MAb and the PG domain-specific M75 MAb. The analysis revealed detectable serum levels of the CA IX ECD in tumour-bearing animals (Fig. 8C). Although levels of CA IX ECD did not fully correlate with the weights of tumours (conceivably due to differences in tumour tissue heterogeneity, extent of hypoxia and intratumoural levels of CA IX) these data principally support the monitoring potential of CA IX serum levels.