To further characterize the correlation between the up-regulation of carbonic anhydrase IX at increased cell density and the binding of the peptide CaIX-P1, fluorescence-activated cell sorting (FACS) studies were carried out. For FACS analysis HCT116, HT29 and BxPC3 cells were grown to high cellular density and incubated at the same conditions either with a rhodamine- or Alexa fluor-labeled anti-human carbonic anhydrase monoclonal antibody (anti-CAIX-mAb) or with FITC-labeled CaIX-P1 or with both. The FACS analysis demonstrated similar portions of labeled cells for the antibody and the CaIX-P1 peptide when the cells were cultivated under the same conditions. In particular, after determination of the cell autofluorescence, about 36% and 28% of HCT116 cells were found to be labeled with the anti-CAIX mAb and the CaXI-P1 peptide respectively (Figure 7).

Co-incubation of HCT116 cells at the same conditions with both molecules (rhodamine-labeled anti-human CAIX mAb and FITC-labeled CaIX-P1) revealed that about 71% of the cells labeled with anti-human CAIX mAb were co-labeled with FITC-CaIX-P1 (Figure 8). In particular, after determination of autofluorescence at 530 nm (FL-1) and 590 nm (FL-2) (Cutoff lines Figure 8), 17% of all cells were found to be labeled with rhodamine-anti-CAIX-mAb (upper boxes Figure 8). Among them only 5% was within the autofluorescence level at 530 nm (upper left box, Figure 8). The rest (12% of all cells, upper right box, Figure 8) showed a signal, which was higher than the autofluorescence signal at 530 nm, indicating that about 71% (12%/17%) of the cells labeled with anti-human CAIX mAb were co-labeled with FITC-CaIX-P1.

In addition to HCT116 cells, FACS analysis of HT29 cells incubated with FITC-CaIX-P1 or rhodamine labeled anti-CAIX mAb under conditions of increased cellular density revealed that about 33% of all cells were labeled FITC-CaIX-P1, whereas about 36% of all cells were labeled with rhodamine-anti-CAIX mAb (Figure 9).

The same experiments using the CAIX negative human pancreatic carcinoma cell line BxPC3 revealed only background level binding of both Alexa fluor-labeled anti-human CAIX mAb and FITC labeled CaIX-P1 (Figure 10).

These results indicate that the novel peptide CaIX-P1 possibly recognizes the extracellular domain of carbonic anhydrase IX in a similar manner like the monoclonal antibody and strengthens therefore the hypothesis that the dodecapeptide might be used as a lead structure for the development of specific molecules targeting CAIX in tumors.

To evaluate the binding specificity of CaIX-P1 on colorectal carcinoma HCT116 cells competition experiments were performed as previously described [14]. Co-incubation of ¹²⁵I-CaIX-P1 with the unlabeled peptide at different concentrations revealed an increasing inhibition of radioligand binding in the presence of the competitor (Figure 11A). Maximal radioligand inhibition of about 90% was reached at a competitor concentration of 10⁻⁴ M (* p < 0.05). Using different ligands, such as octreotide and the peptides HBP-1 and VA131, as negative control competitors at the same concentration (5 × 10^{− 5} M), binding of the radioligand could not be competitively abolished (Figure 11B).

The negative control peptides HBP-1 and VA131 were randomly chosen. HBP-1 (SPRGDLAVLGHKY) is a linear peptide identified by phage display technology on squamous cell carcinoma cells of the head and neck [19], whereas VA131 (PWMEY) is a derivative of the phage display identified dodecapeptide p160 [20]. The fact that binding of the radioligand was strongly inhibited by the unlabeled CaIX-P1 peptide in a concentration dependent manner but was not affected by the negative control linear peptides supports the hypothesis of a specific target binding. Similar results were previously reported for the stably CAIX expressing cell line SKRC52 [14]. However, although our results strengthen the hypothesis of a specific ligand binding, they also reveal an IC⁵⁰ value of about 1–5 μM, which is low compared to peptide-based molecules that are used in clinical targeting approaches. Prominent examples are the somatostatin analogon octreotide and the αvβ3 integrin targeting RGD sequence, which show nanomolar binding affinities towards their targets [21,22]. Therefore, a major issue of further investigation is the enhancement of the ligand’s affinity. An essential step in this direction is the identification of the amino acids in the sequence of CaIX-P1 that are important for target binding. Identification of key amino-acid residues can be achieved by various strategies, including peptide truncation from the N- or C-termini for isolation of the minimum binding sequence, or amino-acid scan, such as alanine-scan, for characterization of the relative importance of particular functional groups in the peptide sequence [23]. Recent studies applying such strategies revealed modifications in the sequence of the peptide, which might be favorable for improvement of the ligand’s target affinity and indicated a fragment (CaIX-P1-4-10) with increased binding capacity [24]. However, our study also demonstrated that such approaches require the synthesis and moreover the evaluation of a high number of derivatives, which is time and cost intensive. An attractive alternative method is the use of peptide arrays [25]. This technology is based on the principle that hundreds upon thousands of peptides are synthesized and presented on a planar surface at a time [26]. Incubation of the target with the peptide array and analysis of interactions between target and the spotted sequences can lead to identification of molecules with increasing binding affinity. In recent studies, peptide arrays were successfully applied for the identification of peptide-cell interactions [27].

Once the essential amino acids for target binding are determined, establishment of the optimal ligand conformation is required for affinity optimization. Linear peptides are characterized by increased 3D flexibility and enhanced loss of conformational entropy upon formation of the ligand-target-complex. Local or global conformational constraints can stabilize the structure, which is optimally complementary to the binding site. Targeted modifications to constrain the local conformation include for example the use of D-amino acid residues, which can stabilize a reverse β-turn structure [23]. Global conformational constraints such as cyclization of the peptidyl backbone or grafting of the binding domain into a molecular scaffold, potentially lead to an increased structural rigidity and a higher Gibbs free binding energy [28], facilitating an affinity increase.

A prerequisite for the in vivo use of an agent as targeting vehicle is a selective binding to the tissue of interest. This requirement is fulfilled when the uptake by the healthy tissues is lower than that of the target in bio-distribution experiments. The organ distribution data of ¹³¹I-labeled CaIX-P1 in mice bearing HCT116 tumors indicates that the peptide has potential in this respect, however demonstrates at the same time major challenges prior to a possible in vivo use of the molecule. In particular, the results of the organ distribution experiments reveal a higher uptake in the tumor after 30 min and 120 min circulation, compared to most healthy tissues, such as heart, spleen, liver, muscle and brain (Figure 12).

Furthermore, our studies showed that although tumor binding of CaIX-P1 decreased with time progression, a higher decrease was noticed for the healthy tissues resulting in a significant increase of the tumor-to-organ ratios over time (* p < 0.05) (Table 1).

The biodistribution experiments also indicated a slight correlation between the tumor size and radioligand uptake. In particular both at 30 min and 120 min p.i. lower tumor uptake (%ID/g tissue) and lower tumor-to-organ ratios were noticed for animals carrying subcutaneously tumors with a weight of ≤100 mg (data not shown) compared to animals with larger tumors. This result seems to be in concert with the in vitro data showing a correlation between cell density and radioligand uptake, it must be however, considered very critically and proved in further experiments, including immunohistochemistry and autoradiography studies, in detail. Main reason for that is the metabolic instability of the peptide. Previous investigation of serum stability revealed a rapid peptide degradation by serum proteases. The half-life was measured to be about 25 min [14]. Therefore, a prerequisite for safe conclusions on the in vivo uptake properties of CaIX-P1 is the improvement of the peptide’s metabolic characteristics.

A further important issue that is associated with the proteolytic degradation of CaIX-P1 is the elevated levels of radioactivity in blood. Elevated blood values represent a major drawback, since they cause increased background noise that is unfavorable for imaging applications. Metabolic instability of CaIX-P1 is found to be associated with labeling instability, which further increases the blood background. It has been shown in previous studies that the product of serum degradation is a derivative lacking the N-terminal tyrosine [14]. Since radioiodination is performed on the tyrosine residues, free radiolabeled tyrosine molecules might be produced, which then circulate in the blood stream and are responsible for increased background noise. Therefore, further studies must focus on two major aspects: the first aspect includes the improvement of the metabolic stability. This can be achieved by modifications in the amino-acid sequence, including use of unnatural amino-acids which cannot be recognized by proteases, such as D-isoforms, N-terminal acetylation or grafting of the binding sequence in a scaffold structure [29,30]. However, such modifications might induce conformational changes to the specific binding molecule and further reduce its binding affinity [31]. Furthermore, although scaffolds are successfully applied to improve the serum stability, they also lead at the same time to a significant increase of the ligand’s molecular size influencing the advantageous pharmacokinetic properties of oligopeptides, such as improved tumor penetration and rapid clearance. The second aspect refers to the improvement of radiolabeling stability. Alternatively to direct iodination, indirect labeling via a chelating moiety, covalently bound to the peptide, can be performed [32]. Macrocyclic chelators, such as DOTA or branched chelators, such as DTPA, which utilize carboxylate and amine groups to form stable complexes with radioactive metals, provide improved labeling stability [33], have however also the disadvantage that they might change the conformational structure of the specific binding ligand and negatively influence its target affinity.

Human colorectal carcinoma HCT116 cells and human pancreatic carcinoma BxPC3 cells were cultivated in RPMI 1640 cell culture medium supplemented with 10% (v/v) fetal calf serum (Invitrogen). Human colorectal adenocarcinoma HT29 cells were cultivated in DMEM supplemented with 10% FCS.

The peptides CaIX-P1 (YNTNHVPLSPKY), HBP-1 (SPRGDLAVLGHKY) and VA131 (PWMEY) were chemically synthesized on an ABI 433 A peptide synthesizer (Applied Biosystems) as described previously [14]. After HPLC based purification (Chromolith Semi Prep Column RPe18, 10 × 100 mm, Merck, Darmstadt, Germany), lyophilization and mass spectrometry analysis (MALDI-3; Kratos instruments), the peptide was radiolabeled with ¹²⁵I or ¹³¹I using the chloramine T method and the purified product was analyzed on a Chromolith Performance RP-18e 10 × 4.6 mm column (Merck). All experiments were conducted under the same radiolabeling conditions. For fluorescence microscopy studies and fluorescence-activated cell sorting, fluorescein isothiocyanate (FITC) was coupled at the N-terminus of CaIX-P1.

300,000 and 2,000,000 human colorectal carcinoma HCT116 and pancreatic carcinoma BxPC3 cells were cultivated in 6-well plates for 48 h. RPMI 1640 medium with 1% BSA was used for cell blocking. Thereafter, 1 mL RPMI 1640 medium containing about 1.0 × 10⁶ cpm ¹²⁵I-labeled peptide (specific activity ~50 GBq/μmol) was added and cell incubation was carried out for various time periods (10–120 min). To characterize the specificity of radioligand binding, co-incubation with unlabeled CaIX-P1 at various concentrations (10⁻⁴ to 10⁻⁹ mol/L) was performed. After incubation the cells were washed with ice cold PBS and lysed with 0.5 mL NaOH 0.3 mol/L. The lysate was collected and the radioactivity was measured with a γ-counter (LB951G; Berthold Technologies). Bound radioactivity, representing the amount of peptide accumulation on HCT116 cells, was calculated as percentage applied dose per 10⁶ cells. Binding specificity was further determined using octreotide, HBP-1 and VA131 as negative control competitors at the same concentration as CaIX-P1. All binding and competition experiments were performed in serum free medium in order to avoid peptide degradation by serum proteases.

300,000 and 2,000,000 HCT116, HT29 and BxPC3 cells were cultivated in 6-well plates at 37 °C for 48 h. After cell washing with 10 mL ice-cold PBS pH 7.4 and scraping with a cell scraper, the lysate was centrifuged for 3 min at 1000 rpm. The pellet was washed with 5 mL PBS and subsequently centrifuged for 3 min at 1000 rpm. 2 mL Triton X-100 1% was added to the pellet and centrifugation was performed for 10 min at 2700 rpm. After protein concentration determination, equal amounts of protein were loaded into a polyacrylamide gel, separated by electrophoresis and transferred to a nitrocellulose membrane with a Mini Trans-Blotter (100 V for 90 min). Membranes were blocked with 5% non-fat milk powder in TBST buffer for 1 h at room temperature. Incubation overnight with a primary rabbit IgG monoclonal anti-human CAIX antibody (Abcam, ab108351; 1:1000 dilution) was performed at 4 °C. Thereafter the membrane was washed with TBST and incubated with horseradish peroxidase conjugated antibody (R & D Systems, HAF008; 1:1000 dilution) at room temperature for 60 min. As control, anti-GAPDH antibody (Abcam; 1:5000 dilution) and polyclonal anti-mouse (DakoCytomation, P 0447; 1:1000 dilution) were respectively used. Antibody binding was determined using a chemoluminescence detection system.

Human colorectal carcinoma HCT116 cells were seeded in culture plates (1,000,000–2,000,000 cells/well in 3 mL medium) containing glass coverslips and were allowed to grow for 48 h. The medium was replaced by 900 μL fresh medium (without FCS) and 100 μL 10⁻⁴–10⁻⁵ M sterilized stock solution of FITC-labeled-CaIX-P1 or fluorescein at the same concentration was added to the cells. The plates were incubated for 1 h at 37 °C. After incubation the cells were washed thrice with 1 mL ice cold PBS (pH 7.4). Subsequently the cells were fixed with 2% formaldehyde in PBS (pH 7.4) at room temperature for 20 min. This procedure was performed on ice. The cells were washed with ice cold PBS (1 × 600 μL, pH 7.4) and the cover slips were mounted on a slice with DAKO fluorescent medium. Cells were photographed using a Leica DM-RA2 fluorescence microscope. Digitalized figures were processed using standard imaging software (Adobe Photoshop 6.0/Illustrator 9.0).

For fluorescence activated cell sorting 1,000,000–2,000,000 HCT116, HT29 or BxPC3 cells were seeded into 6-well plates and cultivated in 3 mL incubation medium at 37 °C for 48 h. The medium was replaced by 1 mL fresh medium (without FCS) containing FITC-CaIX-P1 at a concentration of 10⁻⁵–10⁻⁶ mol/L or 5–10 μg mouse anti-human CAIX-mAb (R & D Systems, MAB2188) or both. After overnight incubation the medium was removed and the cells were washed with 3 mL PBS. Cells incubated with anti-human CAIX-mAb were further incubated with a rhodamine conjugated anti-mouse IgG (Rockland, 610-4002) or an Alexa fluor conjugated IgG overnight. The cells were transferred in 1.5 mL Eppendorf tubes and centrifuged at 1000 rpm for 10 min. The pellet was resuspended in cell wash medium and fluorescence activated cell sorting was performed. To discriminate between labeled cells and autofluorescence of unlabeled cells, autofluorescence of HCT116, HT29 or BxPC3 cells was measured and determined through a cutoff line. Cells that showed a higher fluorescence signal than the cutoff value were considered labeled by FITC-CaIX-P1 or rhodamine/Alexa fluor labeled anti-CaIX-mAb. Flow cytometry analysis was performed in a FACS-Calibur (Becton & Dickinson) equipped with a 488 nm argon laser with filter combination for FITC (530 nm) and rhodamine (590 nm). Histogram and dotplot analysis were performed using the Cellquest software (Becton & Dickinson).

In vivo experiments were performed in Balb/c nu/nu mice, obtained from Charles River WIGA. Human colorectal carcinoma HCT116 tumors were transplanted through subcutaneous injection of a cell suspension (~5 × 10⁶ cells in OPTI-MEM) into the mouse trunk. After the tumors were grown to a size of 1.0 cm³, approximately 1 MBq of ¹³¹I-labeled CaIX-P1 was injected into the tail vein. The radioligand circulated in the mice for 15, 30 and 120 min and thereafter the animals were euthanized. Tumor, blood and selected organ tissues including heart, lung, spleen, liver, kidney, muscle and brain were removed and drained of blood. The organs were weighed and radioactivity was measured in a γ-counter. The tumor and organ uptake was calculated as percentage injected dose per gram tissue (%ID/g). All animal experiments were carried out in conformity with German and European laws for animal protection. The animal studies were approved by the Regierungspräsidium Karlsruhe, Baden-Württemberg, Germany, file reference: 35-9185.81/G-132/04.

Statistical analysis was performed using the unpaired Student t-test on the SIGMASTAT program (Jandel Scientific). A p-value of 0.05 or less was considered statistically significant.