Cancer Therapy Vol 2, 245-262, 2004

1Centre of Molecular Medicine, Institute of Virology, Slovak Academy of Sciences, Dúbravská cesta 9, 845 05 Bratislava, Slovak Republic

__________________________________________________________________________________

Key words: carbonic anhydrase IX, MN, G250, cell adhesion, pH regulation, hypoxia, cancer detection, prognosis, therapy, sulfonamides, monoclonal antibodies

Abbreviations: anion exchanger, (AE); carbonic anhydrase, (CA); cell adhesion molecule, (CAM); clear cell RCC, (ccRCC); glucose trasporter, (GLUT); hypoxia inducible factor, (HIF); hypoxia response element, (HRE); immunohistochemistry, (IHC); renal cell carcinoma, (RCC); vascular endothelial growth factor, (VEGF); von Hippel Lindau, (VHL)

Modern diagnostic approaches and therapeutic anticancer strategies stem from enormous research and technical advances, which improved our understanding of cancer hallmarks and clarified underlying molecular networks (Hanahan and Weinberg, 2000). These strategies principally focus on targeting molecules that are relevant for particular type of tumor and relate to its phenotypic characteristics. Spectrum of potential targets includes regulatory genes and proteins, whose functional modulation may affect tumor development, and cancer-associated molecules, whose expression pattern per se provides a means to recognize and destroy malignant cells. Carbonic anhydrase IX (CA IX) is one of the recently emerged targets with distribution and molecular attributes offering various opportunities for clinical utilization (Table 1).

CA IX belongs to a carbonic anhydrase family of enzymes that use a zinc-activated hydroxide mechanism to catalyze the reversible conversion of carbon dioxide to carbonic acid in the net reaction CO₂ + H₂O « HCO₃^- + H⁺ (Christianson and Cox, 1999; Supuran, 2004). Via this catalytic activity carbonic anhydrases either supply bicarbonate for biosynthetic reactions and ion transport across the membranes or consume produced/transported bicarbonate. Their proper performance is essential for various physiological processes in virtually all living organisms. Mammalian a -CAs exist in at least 15 isoforms (Figure 1), which can be divided according to various criteria to intracellular (CA I-III, VA, VB, VII, VIII, X, XI, XIII) and extracellular (CA IV, VI, IX, XII, XIV), catalytically active (CAI-VII, IX, XII-XIV) and inactive (CA VIII, X, XI), wide-spread (CA II, IV, VB, XII, XIV) and restricted to few tissues (CA I, III, VA, VI,

Table 1. SWOT analysis of CA IX for diagnostic and therapeutic usefulness

Figure 1. Mammalian CA isoforms, schematic illustration of their domain composition, enzyme activity and subcellular localization. Cytoplasmic and mitochondrial CAs consist only of the CA domain, secreted CA contains a short C-terminal extension, membrane-associated CAs in addition have a transmembrane anchor and, except for CA IV, also the cytoplasmic tail. CA IX is the only carbonic anhydrase with an N-terminal proteoglycan-like sequence, engaged in cell-to-cell adhesion.

VII). Different combinations of these properties create a diversity allowing each isoform fulfill a unique role in a specific physiological context. Active CAs are generally expressed in differentiated cells and play important roles in respiration, bone resorption, production of gastric acid and different body fluids, renal and testicular acidification and other biological processes. Loss or deregulated activity of certain isozymes has been implicated in several diseases including glaucoma, osteopetrosis, oedema from heart and renal failure, neurological and neuromuscular disorders etc. (Pastorekova et al, 2004). However, despite many efforts, no consistent link to cancer had been found before the identification of CA IX.

The discovery of CA IX was in a way unintentional and unexpected. Our initial aim was to reveal envelope proteins of hypothetical human defective retrovirus by formation of phenotypically mixed virions (pseudotypes) of vesicular stomatitis virus (VSV). Indeed, the pseudotypes with predicted properties were produced in a cell line MaTu, reportedly derived from a mammary tumor (Zavada et al, 1972). The extracts from MaTu cells, metabolically labeled with ³⁵S-methionine, contain a 58 kDa protein immunoprecipitated with various human and animal sera. This protein turned out to be transmissible via cell-to-cell contact to certain other human cells, e.g. HeLa, derived from cervical carcinoma, in spite of the fact that no morphologically distinct virions could be detected (Zavada and Zavadova, 1991).

In an effort to elucidate the identity of the enigmatic MaTu agent, we produced monoclonal antibodies (Mabs) which enabled us to distinguish between two components, MN and MX (Pastorekova et al, 1992). Cell membrane protein, designated MN, formed a "twin" band of 54 and 58 kDa on Western blots and was detected by MAb M75. This protein could be assembled into VSV virions, but it was not transmissible. The second, cytoplasmic protein MX reacted with the M67 antibody, formed a single band of 58 kDa and was transmissible, but did not integrate into the VSV pseudotype.

Importantly, the MN protein was found to be tumor-associated. Its expression was inducible by high density of cells, correlated with the tumorigenicity of HeLa x fibroblast hybrids and was expressed in several types of human carcinomas, but not in corresponding normal tissues. It was also absent from human placenta and embryos (Zavada et al, 1993). The transmissible component MX was later identified as a nucleoprotein of lymphocytic choriomengitis virus (LCMV), presumably a human strain (Reiserova et al, 1999). Further on, we concentrated on identification and analysis of MN, which after its sequencing was identified as a cellular protein and is now named as CA IX.

MN cDNA was isolated by immunoscreening of cDNA expression library derived from MX-infected HeLa cells using M75 MAb (Pastorek et al, 1994). Sequence analysis allowed for deduction of the primary structure of MN protein and prediction of its domain composition. However, due to an error in the 5' region of cDNA sequence, the N-terminal portion of MN protein was wrongly assumed to contain helix-loop-helix motif, what was corrected in the subsequent paper dealing with a genomic structure of MN (Opavsky et al, 1996). According to the correct sequence, MN cDNA codes for 459 amino acid (aa) protein with 414 aa N-terminal extracellular part linked through the 20 aa hydrophobic transmembrane region (TM) with 25 aa C-terminal intracellular tail (IC). Extracellular part is composed of 37 aa signal peptide, 59 aa region with similarity to keratan sulfate-binding domain of a large proteoglycan aggrecan (PG) and a 257 aa carbonic anhydrase domain (CA). Under non-reducing conditions, monomeric 58/54 kDa MN protein assembles into trimers reminding thus virus surface glycoproteins, what could at least partially explain its insertion into the viral envelope (Pastorekova et al, 1992). Linear illustration of the monomeric MN protein with indicated positions of important aa residues and a simplified scheme of its putative modular structure designed according to available experimental data are shown on Figure 2.

CA domain of MN protein displays a significant identity with extracellular carbonic anhydrases, particularly with the secreted CA VI (40.8%). It also contains all three conserved histidine residues that coordinate catalytic zinc and are therefore crucial for the enzyme activity. Based on the homology with the carbonic anhydrases and because at that time it was the ninth isoform identified within the a -CA family, MN protein was renamed as carbonic anhydrase IX (Hewett-Emmett and Tashian, 1996).

Exon-intron structure of the CA9 genomic region that encodes the CA domain is analogous to other mammalian carbonic anhydrase genes, further supporting its membership in a -CA family. Exons related to additional domains/regions of CA IX were probably acquired by exon shuffling. In particular the PG region is absent from the other a -CAs and represents a unique feature of CA IX. Interestingly, extracellular parts of distant CA relatives, namely two human receptor protein tyrosine phosphatases (RPTPb /g ) and a soluble rat derivative phosphacan, contain inactive carbonic anhydrase domain and are expressed in the form of proteoglycan in the nervous system (Krueger and Saito, 1992; Barnea et al, 1994). Acatalytic CA domain of RPTPb forms a deep and wide pocket and functions as a ligand-binding receptor site that mediates an interaction of RPTPb with a neuronal cell adhesion molecule contactin (Peles et al, 1995). Via this interaction, RPTPb can modulate glial-neuronal cell cross-talk and influence differentiation, adhesion and motility of neurons. This example indicates that structurally similar domain of classical CAs may potentially play a dual role of enzyme and receptor, what could be especially relevant for the transmembrane isoforms (including CA IX), which are evolutionarily older and more closely related to RPTPs than the intracellular CAs (Hewett-Emmet and Tashian, 1996).

Figure 2. Modular structure of CA IX deduced from the sequence and experimental data. Domain composition: SP—signal peptide, PG-proteoglycan-like segment, CA—carbonic anhydrase domain, TM—transmembrane anchor, IC—intracytoplasmic tail. CA IX forms homotrimers linked with disulfidic bonds mediated by cystine residues. The CA catalytic centre contains three histidine residues, which are located on the bottom of an active site cavity and bind one Zn²⁺ ion.

Several years after decoding the molecular identity of CA IX and resolution of its cDNA and genomic sequences, Grabmaier et al, (2000) revealed a full homology between CA IX and G250 antigen, which had been independently investigated for its close link to renal cell carcinomas (RCC). G250 antigen was detected on the surface of renal carcinoma cells with the monoclonal antibody G250 generated against cell homogenates from the primary RCC lesions (Oosterwijk et al, 1986). In contrast to M75 MAb, which recognizes both native as well as denatured CA IX and is useful for all immunodetection methods including immunoblotting and routine immunohistochemistry (IHC) on paraffin-embedded tissues, G250 MAb is clearly directed against a conformational epitope and thus can be used in IHC on living cells or frozen sections only and does not work in immunoblotting. This property of G250 MAb apparently complicated cloning of the corresponding cDNA and accounted for a delayed molecular characterization of G250 antigen. Nevertheless, G250 MAb has become an important imaging and targeting tool as described below.

Because G250 had been long considered as an RCC-specific antigen, it was mostly studied in relationship to RCC (Mulders et al, 2003). In between, M75 helped to uncover CA IX expression in a broad array of different tissues and to understand its regulation.

CA IX has a distinctive expression pattern: it is naturally expressed in few normal tissues, but its ectopic expression is induced in a wide spectrum of human tumors (Figure 3).

The most abundant expression of CA IX was found in the normal mucosa of the stomach and gallbladder (Pastorekova et al, 1997). Lower levels are expressed in the intestinal epithelium, where it is confined to the cryptal areas composed of cells with high proliferation capacity (Saarnio et al, 1998a). Noteworthy, amount of CA IX progressively decreases with increasing distance from the stomach toward the rectum. Other normal tissues that display weak expression of CA IX include epithelia of pancreatic ducts, male reproductive organs and lining cells of body cavity (Kivela et al, 2000; Karhumaa et al, 2001; Ivanov et al, 2001).

Almost perfectly complementary pattern can be seen when looking at the distribution of CA IX in the cancer tissues. CA IX is ectopically expressed at relatively high levels and with a high prevalence in tumors, whose normal counterparts do not contain this protein. These comprise carcinomas of the cervix uteri, esophagus, kidney, lung, breast and many other tumors (Liao et al, 1994, Turner et al, 1997, Liao et al, 1997, McKiernan et al, 1997, Vermylen et al, 1999, Bartosova et al, 2002). The opposite expression is evident also in tissues with high natural CA IX expression, such as stomach and gallbladder, which

Figure 3. Distribution of CA IX in tissues with examples of IHC staining of RCC and papilloma coli. Most of the normal tissues do not contain any CA IX. On the other hand, it is strongly expressed in the gastric and duodenal mucosa. In the tumors, its expression is mostly ectopic. Photomicrographs: Expression of CA IX in papilloma coli (top) and in renal cell carcinoma (bottom). Immunoperoxidase stainig of paraffin sections with M75 antibody and DAB.

lose or reduce CA IX upon conversion to carcinomas (Saarnio et al, 2001; Leppilampi et al, 2003). In the colonic epithelium, CA IX is present normally in the deep crypts and abnormally in the superficial adenomas and carcinomas, with the most intense staining seen in tumors with mucinous component (Saarnio et al, 1998b).

Especially striking is very high proportion of CA IX-positive specimens among the cervical, renal and lung cancers. In the cervical cancer, CA IX immunoreactivity with M75 can be observed in virtually all cervical carcinomas and the majority of cervical intraepithelial neoplasia (Liao et al, 1994). Diffuse CA IX-positive staining signal in normal cervical tissues is found only in concurrent presence of dysplasia or carcinoma and therefore it can be useful as an early diagnostic indicator of cervical neoplasia in Pap smears (Liao and Stanbridge, 1996). In the kidney cancer, CA IX protein expression is selectively linked with the most frequent carcinomas of renal clear cell type (ccRCC). High levels of CA IX are seen in primary, cystic and metastatic ccRCCs, but not in benign lesions (Liao et al, 1997). Simultaneous study using RT PCR has independently proven CA9 mRNA expression in RCC and its absence in benign renal tissue (McKiernan et al, 1997). These findings were further explored into the development of sensitive enhanced RT PCR assay for the detection of RCC cells circulating in the peripheral blood of renal cancer patients (McKiernan et al, 1999). In the lung cancer, CA IX is not found in pre-neoplastic lesions, but is readily present in malignant tumors (Vermylen et al, 1999). Some normally looking bronchial and alveolar epithelia in close vicinity to the tumors contain CA IX positive cells, whereas all other normal lung specimens sampled at a distance from the tumor are negative.

Unusual distribution of CA IX raised questions about mechanisms that control its differential expression in normal versus tumor cells and contribute to its frequent presence in many tumor types. Experimental evidence that the increased cell density can influence CA IX expression through the promoter activation redirected the attention to a transcriptional regulation of CA9 gene (Lieskovska et al, 1999). CA9 promoter analyzed under conditions of high cell density was shown to possess five regulatory regions containing several cis-acting elements (Kaluz et al, 1999). Two regions adjacent to the transcription initiation site bind AP-1 and SP transcription factors and their synergy is needed for the basic transcriptional activation of CA9 gene (Kaluzova et al, 2001). The most important regulatory element of CA9 promoter is localized on the antisense strand between the SP-1 binding site and the transcription start at position -10/-3 and consists of the nucleotide sequence 5'-TACGTGCA-3' corresponding to a hypoxia response element (HRE) (Wykoff et al, 2000). HRE element is recognized by HIF-1 transcription factor, which assembles under hypoxic conditions from constitutive b subunit and oxygen-regulated a -subunit. In normoxia, HIF-1a is modified by proline hydroxylases at two conserved proline residues in the central oxygen-dependent degradation domain and subsequently degraded via pVHL tumor suppressor protein (Epstein et al, 2001). Moreover, its transcriptional activity in blocked by factor inhibiting HIF (FIH)-mediated hydroxylation of arginine residue in C-terminal transactivation domain (Mahon et al, 2001). When oxygen level is reduced and the cell is exposed to hypoxic or anoxic conditions, HIF-1a escapes hydroxylation, accumulates and following dimerization with HIF-1b becomes transcriptionally active. Presence of the functional HRE element thus makes CA9 a transcriptional target of HIF-1 and places it among the genes regulated by hypoxia (Semenza, 2003). HRE element is utilized also in a density-induced transcription of CA9, but under these conditions it requires cooperation with juxtaposed SP-1 binding site. Activation of CA9 transcription by increased cell density, which is linked with pericellular hypoxia, involves PI3 kinase pathway and subhypoxic levels of HIF-1a (Kaluz et al, 2002).

Since the level of HIF-1a is controlled by pVHL, it is not surprising that the expression of the wild type VHL transgene can suppress the transcription of CA9 mRNA in the normoxic cells and that VHL deletion or inactivating mutation leads to release of CA9 transcription (Ivanov et al, 1998). Loss of functional pVHL is linked with a majority of clear cell renal cell carcinomas and provides explanation for frequent presence of CA IX in ccRCC. Onset of CA IX expression is an early event occurring in morphologically normal single cells within the renal tubules of patients with VHL disease and therefore provides a robust system for the identification of early foci of VHL inactivation (Mandriota et al, 2002). In addition, transcription of CA9 gene can be modulated by methylation of CpG dinucleotides within the promoter region. In RCC cell lines and in tumors, expression of CA IX is associated with hypomethylation (Cho et al, 2000, 2002). No relationship between CA9 gene expression and promoter methylation was observed in RCC/normal kidney paired specimens by Grabmaier et al, (2002), but it could be missed because the authors did not take into account VHL status and/or presence of hypoxia. Indeed, Ashida et al, (2002) showed that the expression of CA IX in cells with VHL mutation does not occur without hypomethylation of the promoter, particularly CpG sites at positions -74 and -6 with respect to the transcription start, and they proposed that VHL and methylation cooperate in the regulation of CA IX. Methylation of -74 CpG site can also influence the expression of CA IX in the carcinoma cell lines of a different origin than RCC, where it seems to represent adverse factor modifying transcriptional response to cell density (Jakubickova et al, submitted).

It is possible that CA IX expression is regulated also at higher stages of the biosynthetic trail, similarly to some other hypoxia-induced genes. There are several supportive indications, including the presence of consensus phosphorylation sites in the intracytoplasmic tail, which might affect functional performance of CA IX, and the shedding of soluble CA IX, which might control the amount of the plasma-membrane associated protein (Zavada et al, 2003). Of course, these assumptions require thorough investigations. So far, it has been clearly shown that the posttranslational stability of CA IX protein is very high (with a half life corresponding to approximately 38 h) and allows for its long persistence in reoxygenated cells (Rafajova et al, 2004). Such extended presence of CA IX on the surface of the post-hypoxic cells may have important implications for its intratumoral distribution.

The highest level of CA IX expression in vitro is achieved upon mutual aid of high cell density and hypoxia. These conditions are readily found in vivo in solid tumor tissues as consequences of physiological barriers that limit expansive growth of tumor mass. Aberrant formation and functional defects of tumor vasculature cause decreased delivery of oxygen and insufficient removal of metabolic waste exposing distant tumor cell populations to hypoxia and low pH. Depending on severity of these stresses, the affected cells respond by a number of phenotypic changes that may involve necrosis and apoptosis, cessation of the cell cycle, or adaptation enabled by the metabolic shift to anaerobic glycolysis and induction of neoangiogenesis. These changes result from transcriptional activities of HIF, which besides CA IX induces spectrum of functionally relevant genes, including VEGF (vascular endothelial growth factor) and GLUT-1 (glucose transporter) (Semenza, 2003). Hypoxia-triggered architectural and phenotypic rearrangements of tumor tissue finally convert into development of necrotic areas surrounded by the zones of surviving hypoxic cells, which adapted to stress and acquired aggressive behavior. These cells are more inclined to produce metastases and are generally refractive to conventional anticancer treatment modalities. Accordingly, the presence of the hypoxic tumors correlates with poor cancer prognosis (Wouters et al, 2002).

CA IX expression pattern in vivo clearly mirrors a distribution of hypoxic areas. The protein is localized in the perinecrotic regions of various solid tumors, including carcinomas of the breast, skin, ovary, cervix uteri, head and neck, lung, bladder (Wykoff et al, 2000, 2001; Chia et al, 2001, Giatromanolaki et al, 2001; Koukourakis et al, 2001; Loncaster et al, 2001; Olive et al, 2001). According to the measurements in head and neck carcinomas, CA IX expression starts at a distance of 40-140 m m (median 80 m m) from a blood vessel and continues toward necrosis (Beasley et al, 2001). Similar spatial relationship of CA IX to microvessels is found in bladder and lung cancer (Turner et al, 2002, Swinson et al, 2003). When compared to the distribution of HIF-1a and chemical marker of hypoxia EF5 assessed in an independent study (Vukovic et al, 2001), CA IX expression begins more distantly than HIF-1a , but more closely than EF5. This might suggest that CA IX induction requires lower oxygen levels than HIF-1a and that it occurs in a perinecrotic zone, which is larger than the zone labeled by EF5. In support of this assumption, Olive et al, (2001) found that CA IX staining extends beyond the regions binding another chemical marker pimonidazole in cervical carcinomas. Moreover, they demonstrated that CA IX-expressing cells isolated from tumor xenografts are viable, clonogenic and resistant to killing by ionizing radiation. These important findings indicate that at least a fraction of the tumor cells that express CA IX is intermediate in oxygenation and may represent potential source of metastases.

There are several studies showing positive correlation between CA IX expression and poor prognosis. In the breast tumors, CA IX is associated with necrosis and high grade of ductal carcinomas in situ (Wykoff et al, 2001), negative estrogen receptor status (Span et al, 2003), higher relapse rate and worse overall survival of patients with invasive carcinomas (Chia et al, 2001). In the head and neck cancer, CA IX expression correlates with necrosis, high microvascular density and advanced stage (Beasley et al, 2001), and with poor complete response rate to chemoradiotherapy and poor local relapse free survival (Koukourakis et al, 2001). In the non-small cell lung cancer, CA IX is a significant factor of poor prognosis independent of angiogenesis (Giatromanolaki et al, 2001) and its stromal expression is associated with advanced tumor stage (Swinson et al, 2003). In bladder cancer patients treated by radical radiotherapy with carbogen and nicotinamide, CA IX expression is significantly linked with worse cause-specific and overall survivals (Hoskin et al, 2003). In the carcinomas of the cervix, extent of CA IX expression correlates with electrode measurements of tumor oxygen and with overall survival and metastasis-free survival after radiation therapy (Loncaster et al, 2001). Additional paper describes lack of these correlations in cervical cancer (Hedley et al, 2003). The authors discuss intratumoral heterogeneity, influence by factors other than hypoxia and technical differences in staining procedures as potential causes of discrepancies and call for further studies.

Quite a different situation with respect to CA IX prognostic relationships is evident in renal cell carcinomas, where overall expression of CA IX decreases with progressing stage, grade and development of metastases, and lower CA IX staining (cutoff 85%) is independently associated with poor survival in advanced RCC (Bui et al, 2003). In addition, CA IX staining is inversely correlated with Ki-67 and combination of these two markers into a single parameter allows for stratification of patients into low, intermediate and high risk groups, significantly predicts survival and can displace histological grade (Bui et al, 2004). It is not clear, whether the expression of CA IX in RCC is predominantly a function of VHL gene mutation or whether it reflects also other factors, nevertheless, it seems to be the most significant molecular marker described in kidney cancer so far (Pantuck et al, 2003).

Ectopic expression in tumors, hypoxia-related distribution pattern and correlation with prognosis favor functional involvement of CA IX in cancer development. CA IX is a highly active enzyme with a catalytic performance and proton transfer rate similar to prototype CA II isoform (Wingo et al, 2001). Its activity can be efficiently inhibited by sulfonamides, particularly with certain derivatives that show some selectivity toward CA IX when compared to other isoenzymes (Vullo et al, 2003, Ilies et al, 2003, Abbate et al, 2004, Vullo et al, 2004, Casey et al, 2004). Similarly as other active CA isoforms, CA IX has been proposed to play a role in pH regulation. This proposal seems meaningful especially in relationship to anaerobic tumor metabolism that generates excess of acidic products, such as lactic acid and H⁺, which have to be extruded from the cell interior to maintain the neutral intracellular pH and protect the cells from death. The extrusion of metabolic waste and its poor clearance by inadequate tumor vasculature creates acidic extracellular microenvironment that is more permissive for tumor cell growth and invasion (reviewed in Stubbs et al, 2000). However, lactic acid is not the only source of acidosis and the studies of intratumoral physiological parameters indicate significant contribution of CO₂ (Helmlinger et al, 1997, Helmlinger et al, 2002). A role for CA IX in this process appears to involve catalytic conversion of CO₂ to bicarbonate and proton at the extracellular side of the plasma membrane and facilitation of the bicarbonate transport to the cell cytoplasm. In analogy to another extracellular isoenzyme CA IV, which physically interacts with bicarbonate transporters such as anion exchangers (AE) to form a transport metabolon in differentiated cells (Sterling et al, 2002), CA IX may directly cooperate with AE in tumor cells and help to neutralize their intracellular space. At the same time, the protons produced by CA IX from hydration of CO₂ may remain outside and improve acidosis of microenvironment (Figure 4). Our recent experimental data clearly fit within this concept (Svastova et al, submitted). It is well known, that the acidic extracellular milieu induces production of growth factors, increases genomic instability, perturbs cell-cell adhesion, and facilitates tumor spread and metastasis (Stubbs et al, 2000). Evidence for CA IX as a causal factor of tumor acidosis may thus support its functional involvement in these processes.

Figure 4. Proposed involvement of CA IX in the pH regulation in tumors illustrated on a model that is based on the formation of a transport metabolon composed of anion exchanger (AE) and carbonic anhydrases (in analogy to CA IV-AE-CA II metabolon described by Sterling et al., 2002). CA IX as an extracellular component of the metabolon hydrates carbon dioxide and provides bicarbonate anions to AE, which transports them to the cytoplasm in exchange for chloride anions. At the intracellular side, CA II converts bicarbonate to carbon dioxide, which diffuses out through the plasma membrane. Extracellular CA IX activity also generates protons that contribute to acidification of external pH, whereas cytoplasmic CA II activity allows for consumption of intracellular protons and contributes to neutralization of internal pH.

Besides its enzyme activity, CA IX is also a cell adhesion molecule (CAM), which can mediate attachment of cells to non-adhesive solid support (Zavada et al, 2000). This activity resides in the N-terminal end of the molecule, in the proteoglycan-like domain. The adhesion site of CA IX overlaps with the epitope for M75 monoclonal antibody — PGEEDLP, since M75 blocks adhesion of cells to the immobilized CA IX protein (Figure 5). The PG region of CA IX contains three identical repeats of the motif GEEDLP and four modified repetitions. Also an oligopeptide AITFNAQYA identified with the use of a phage display library of random heptapeptides and synthesized with addition of alanine on both ends competed both with binding of CA IX to the M75 antibody and to the cell surface receptor. The amplification of the M75 epitope in the PG region is also reflected by the capacity of CA IX molecules to bind 2-3 times more of M75 IgG molecules than of any of three monoclonal antibodies V10, V12 and VII20 with epitopes in the CA domain (Zavada et al, MS in preparation). These CA domain-specific monoclonal antibodies prepared by Zatovicova et al, (2003) represent important tools for study of structure-function relationships in CA IX molecule and allow for sensitive detection of soluble CA IX that is described below.

A remarkable feature of the PG segment of the CA IX molecule is a high content of dicarboxylic aa (24 G + E out of total 59 residues) and at the same time, a low content of basic aa (4 R + K). We observed that the acidic character of the PG is reflected by easy dissociation of CA IX from the complex formed with the M75 antibody or with the cell surface receptor already at slightly acidic pH. This property might facilitate release of cells from the tumors acidified by the products of hypoxic metabolism. The cells might then attach elsewhere in the organism where the pH is neutral or slightly basic, and start a metastatic growth (Zavada et al, MS in preparation).

Moreover, CA IX appears to play a role in intercellular adhesion. In polarized epithelial MDCK cells transfected with the human CA9 cDNA, CA IX protein co-

Figure 5. CA IX-mediated cell adhesion to a solid support. On a hydrophobic plastic, human cells (eg. CGL1) attach only on the areas coated with purified CA IX (A). Treatment of the area with CA IX-specific M75 antibody prevents cell adhesion (B), whereas M67 antibody directed to an unrelated antigen has no effect on cell adhesion (C). Reproduced from Zavada et al, 2000 with kind permission from British Journal of Cancer.

localizes with a key adhesion molecule E cadherin and destabilizes E cadherin-mediated cell-cell contacts via a mechanism that involves competitive interaction with b - catenin (Svastova et al, 2003). This capability of CA IX is reminiscent of some oncoproteins (EGFR, c-ErbB2, MUC-1) and makes it a candidate contributor to tumorinvasion that is known to require a diminished intercellular adhesion. In contrast to cadherins, which are homophilic cell adhesion molecules, CA IX is a heterophilic molecule (Zavada et al, 2000 and MS in preparation).

We propose that CA IX, being a CAM, could mediate the communication between cells and transmit signals, but there is no conclusive evidence for this so far. Certain transmembrane proteins with inactive carbonic anhydrase domains termed RPTP (receptor-type protein tyrosine phosphatase) serve as signaling molecules and are involved in differentiation of embryonic brains (Tashian et al, 2000). In accord with the proposal, CA IX expression in normal human epithelia displays polarized distribution confined to basolateral plasma membranes engaged in contacts with the neighboring cells and with the basal membrane (Pastorekova et al, 1997). In the stomach and in large bile ducts CA IX is strongly expressed in differentiated cells, whereas in the intestine its presence is restricted to the bottom of Lieberkuhn crypts, where the cells are intensely dividing and co-express the Ki-67 protein (Saarnio et al, 1998). While the cells migrate to the top of the crypts and gradually differentiate, they are losing the CA IX protein. These observations suggest that in normal tissues CA IX plays a dual role: in the stomach it supports differentiation and in the intestine it is connected with the proliferation of the cells. Consistently with the situation in man, mice with the targeted disruption of Car9 gene develop hyperplastic stomach mucosa with a reduced number of glandular pepsinogen-producing cells and increased number of surface mucus-producing cells (Gut et al, 2003). This finding demonstrates that the mouse CA IX protein is involved in the control of differentiation and proliferation of epithelial cell lineages in the stomach mucosa during the stomach morphogenesis. No remarkable phenotypic alterations were detected in the intestines of the knock-out mice.

Most of the complete CA IX is integrated in the cell membrane as a trimer composed of 54 and 58 kDa monomeric molecules linked together with disulfidic bonds. Body fluids and TC media contain a soluble form sCA IX consisting of 50 and 54 kDa polypeptides (Zavada et al, 2003). It has been proposed that sCA IX may be derived from the complete molecule by the proteolytic cleavage of the extracellular domain from TM and IC by membrane-associated proteases. While TC fluids of permanent cell lines or of primary human tumor explants contain a relatively high concentration of sCA IX (20-50 ng/ml), blood and urine of RCC patients contain extremely low levels of the antigen, of the order of 5-100 pg/ml. This is about 1000 times less than the levels of certain cancer markers, such as CEA (carcinoembryonic antigen) or PSA (prostate soluble antigen). It seems that in human body, sCA IX is rapidly cleared from the blood, but until now it has not been shown whether this is due to absorption in unknown deposits, degradation or excretion in urine. One of the tumors, for which detection of sCA IX might be feasible, is bladder carcinoma. Therein, CA IX expression is associated with the luminal surface that is in a direct contact with urine (Turner et al, 2002) and therefore the shedding of sCA IX deserves further investigation.

For several reasons explained above, CA IX appears to be suitable as a target molecule for cancer therapy. Indeed, several groups of scientists realized this fact and they are attempting to develop "magic bullets", which would be able to destroy tumor cells without causing any unwelcome side effects. Several ingenious strategies were designed and tested (Figure 6). This concentrated attack has been to a certain extent successful, as seen from more than 60 papers dealing with CA IX-targeted therapy. Most of the tested strategies indeed proved effective in in vitro experiments or in immunodeficient animals heterotransplanted with human tumors. Some of the therapies met criteria for testing in phase I and II in humans, various clinical phase II trials were completed and phase III trial in RCC patients with minimal residual disease after tumor nephrectomy is in preparation as a part of the clinical program of Wilex company.

We believe that more of basic knowledge on how CA IX operates at the molecular level is needed, including the signaling pathways used by CA IX in the tumors. Quite possibly, these pathways are not the same in different tumors. Intratumoral heterogeneity of the cells is another problem, which may require suitable complementary therapy.

A majority of clinical trials have concentrated on radioimmunotherapy of renal cell carcinoma (Mulders et al, 2003). Monoclonal antibodies labeled with ¹³¹I have the advantage that they can be used both in diagnostic and therapeutic formats. For the human use, chimeric antibodies containing the Fc fragment derived from human IgG and variable parts of Fab fragments from the mouse G250 antibody were constructed and termed cG250. Chimeric antibodies allow repeated application, without raising human anti-mouse antibodies (HAMA), which would preclude repeated use of the Mab in humans (Oosterwijk and Debruyne, 1995). The ¹³¹I-cG250 proved to be very suitable for scintigraphic imaging of RCC in the patients with a very high specificity and extremely low background, enabling detection of very small tumors (less than 1 cm).

RCC patients with primary tumors and metastases were injected with diagnostic doses of ¹³¹I-cG250 and examined by scintigraphy. Afterwards, their tumors were surgically removed, together with biopsies of normal

Figure 6. Therapeutic strategies exploiting the tumor-associated expression and plasma membrane localization of CA IX. There are several variants of "magic bullets": (a) CA IX-specific antibody serves for targeted delivery of toxins, drugs or radionuclides to tumor cells. A combination with cytokines potentiates the inhibitory effect of antibody molecules; (b) bispecific antibody molecules directed to tumor antigen and to cytotoxic cells ensure killing of the cancer cells; (c) sulfonamides (derivatives designed so as to exert maximum specificity for CA IX) will upset the pH-regulatory system of tumor cells; (d) cell-to-cell binding mediated by CA IX could be inhibited by the soluble form of CA IX receptors (yet hypothetical) or by receptor-mimicking small molecules; (e) recombinant therapeutic gene, which is under the regulatory control of CA IX promoter sequence, containing HRE. Under hypoxia, HIF binds to HRE and switches on the synthesis of the therapeutic protein which can convert an innocuous pro-drug into the toxic product, killing the tumor (Jaffar et al, 2001).

tissues, and distribution of the radioactivity was determined. Radiolabeled cG250 showed excellent tumor targeting, with only very low radioactivity in normal tissues. In vivo selectivity of the antibody was among the highest ever reported. Metastases were also strongly radiolabeled. Calculation showed that suitable doses of radioactive cG250 could achieve killing of all tumor cells, without seriously damaging any normal tissue. Scintigraphic images show no accumulation of radioactivity in the stomach, which is normally expressing high levels of the CA IX antigen (a fact not known at the time of this work). The most disappointing observation was a high heterogeneity of intratumoral distribution of ¹³¹I-cG250. This fact proved also later as a main drawback for radioimmunotherapy of RCC (simultaneously with a relatively high radioresistance of RCC). The problem of intratumoral heterogenous distribution of therapeutic antibody could not be solved even by fractionated application (Steffens et al, 1999).

The Wilex company is now advertising that their ¹³¹I-labeled product WX-G250 RIT is intended for the future treatment of biliary cancers (cholangiocarcinoma, gall bladder carcinoma), which are sensitive to radiation.

There are other interesting papers worth mentioning: a mathematical model for antibody clearance from the patients has been produced (Loh et al, 1998); to reduce the radioactive burden to the organism, a two-step treatment comprised the application of bivalent antibody with one arm derived from G250 and the other directed to a chelate binding radioactive indium. The isotope was given after the clearance of unbound antibody (Kranenborg et al, 1998). For the scintigraphic imaging, the cG250 labeled with ^99mTc (using 3 different methods of labeling) was superior to ¹³¹I-labeled antibody (Steffens et al, 1999b).

Non-radioactive MAbs were also tested. To activate antibody-dependent effector mechanisms, they were used in combination with cytokines or they were modified so as to amplify their complement-dependent cytotoxicity (CDC) or antibody-dependent cell-mediated cytotoxicity (ADCC). There are two in vitro studies on killing RCC cells with human mononuclear cells from peripheral blood of healthy donors, activated with a combination of cG250 antibody and IL-2 (Surfus et al, 1996; Liu et al, 2002). Both of these groups are suggesting that the combination immunotherapy with cG250 and cytokines such as IL-2 shows promise in the treatment of RCC. Established xenografts of human RCC in nu-nu mice were treated with G250 in combination with IF + TNF. This cocktail proved to be therapeutically more efficient than each of the components alone (Van Dijk et al, 1994).

There exist occasional reports on "spontaneous" regression in renal cell carcinoma, sometimes even resulting in disappearence of metastatic cancer. RCC is believed to be a relatively immunogenic tumor (Vissers et al, 1999). Vaccination is one of potential approaches towards immunotherapy of tumors carrying suitable antigens like CA IX. This might afford a life-long active immunity, repeatedly destroying newly emerging tumor recurrences.

Anti-idiotype vaccines represent one possible solution. Their advantages are that a large amount of the immunizing antigen can be produced repeatedly, and that the antigen is in fact not perfectly identical with the original tumor antigen. The patient's organism is probably tolerant to the antigen like CA IX, which is recognized as "self", and does not respond to it immunologically. On the other hand, an anti-idiotype vaccine is somewhat different from the original antigen and can conceivably induce formation of antibodies reacting with the immunizing antigen and also cross-reacting with the original tumor. From the mice immunized with G250, six hybridomas were obtained, producing "second" antibodies (Ab2) reactive with the G250 antibody molecules (Uemura et al. 1994a,b). These "second" antibodies mimicked the CA IX antigen and they were used for immunization of another group of mice. Their sera after immunization contained polyclonal "third antibodies" (Ab3), reactive with human RCC cells carrying the CA IX antigen. Indeed, these sera, when injected into nu-nu mice simultaneously with the live RCC cells protected them against development of the tumors. The response was quite spectacular — in certain variants it was a 100% protection. The treatment with Ab3 was highly efficient even in the mice with established human RCC xenografts (Uemura et al. 1995). Let us hope that analogous experiments in the patients will prove as effective as in the mice.

Dendritic cell vaccine is an oligopeptide capable of binding dendritic cells and so to induce a specific cytotoxic lymphocyte response directed towards RCC cells. Vissers et al, (1999) analyzed the sequence of CA IX protein for potential HLA-A2.1 binding peptides using a computer program and found 60 candidate oligopeptides. They immunized with each of them transgenic mice expressing human HLA-A2.1 and found that four of them indeed developed cytotoxic T-lymphocytes (CTLs), capable to lyse transgenic mouse cells expressing human CA IX protein as well as HLA-A2.1. One of these oligopeptides (HLSTAFARV) was able to bind human A2.1 dendritic cells in vitro. After cocultivation with autologous CD8-positive T cells and after their restimulation, peptide-specific human CTL were obtained. They possessed the capacity to destroy target cells expressing the CA IX protein. Hopefully, this is one of the ways towards immunotherapy of RCC. Interestingly, the same CD4 (+) T cells in the context of HLA-DR molecules recognized this sequence also in the naturally processed CA IX protein (Vissers et al, 2002). This finding supports the view that peptide-based vaccines indeed could raise the cytotoxic cell response in the patients.

Other vaccines derived from the sequence of CA IX were tested using mouse RCC cells transfected with human CA9 cDNA. Three nonapeptides were designed so as to be compatible both with murine H-2Kd as well as with human HLA-A24 histocompatibility antigens. One of the peptides was identified as a potential vaccine effective both in mouse and human organisms (Shimizu et al, 2003).

Ringhoffer et al, (2004) have proposed a peptide vaccine combining four antigens which are frequently expressed in renal cell carcinoma: MAGE-1, MAGE-3, CA IX and PRAME. Tumor-specific expression of at least one of these T-cell activating antigens was detected in all of 41 RCC patients examined, 80% of these patients expressed two or more tumor-associated antigens simultaneously.

An alternative method of vaccination is a culture of monocytes, derived from peripheral blood mononuclear cells (PBMC) transduced with the adenovirus-CA9 vector. This system has several advantages: PBMC are easy to obtain from the blood and they express the whole CA IX protein (Mukouyama et al, 2004).

One of the key problems of developing new cancer treatment methods is a specific and efficient delivery of therapeutic genes into tumor cells, without transfecting normal tissues. Now a technique termed antifection could provide a solution. In principle it is very simple: therapeutic genes (or vectors) are chemically conjugated with monoclonal antibody directed to a suitable tumor antigen, eg., CA IX. This antibody-vector conjugate can attach to the tumor cell, and is subsequently internalized. Finally, the antifected gene is expressed. This idea indeed works, as shown by Dürrbach et al, (1999), who conjugated CA IX-specific MAb G250 with an IL-2 vector and demonstrated its internalization and prolonged expression in mouse cells previously transfected with a human CA 9 vector. Initially, this technique was only modestly efficient, but several further refinements (Deas et al, 2002) brought about an increased delivery and expression of the reporter gene.

Another strategy of the gene therapy is unlimited supply of cytotoxic lymphocytes (CTL) specifically destroying the tumor without damaging normal tissues. Moreover, these CTL should preferably be MHC-independent because (a) tumor cells often do not express MHC; (b) cytotoxic cells should be applicable for all patients carrying the same tumor (or group of tumors) and not tailored for each patient individually with his own combination of MHC antigens. Even this goal appears to be feasible. Weijtens et al, (1996) have engineered CTL armed with a fusion protein scFv/g . The single chain antibody scFv was derived from variable parts of anti-CA IX antibody G250; it has been fused with high affinity Fc(g )RI, which can function in a CD3-independent manner. Such CTLs can specifically lyse target cells via their scFv/g chimeric receptor independent of MHC. They produce cytokines in response to binding with the target cells and like normal CTL, they recycle their scFv/g dictated lytic activity.

In these artificial CTLs, a co-regulatory function of CD2, CD3, CD11a/CD18 and of other molecules was demonstrated (Weijtens et al, 1998). The authors also examined quantitative dependences between the density of T cell chimeric receptor scFv/g and CA IX antigen density on the target cells and their effects on the CTL-mediated cytolysis and production of IL-2 and TNF (Weijtens et al, 2000).

Another ingenious and original method of cancer therapy targeted to CA IX-positive tumor cells has been designed (Shinkai et al, 2001). Submicron magnetic particles were enclosed in liposomes with covalently linked Fab fragments of the G250 antibody. Mice with tumors produced by mouse renal carcinoma cells transfected with human CA9 cDNA were injected with these immunomagnetoliposomes (or with control liposomes). After an interval, they were exposed to oscillating magnetic field, which induced vibrations of the magnetic particles. As a consequence, the temperature of the tumors increased to 43° C. This resulted in temporary growth arrest. Control tumors continued growing.

For years, virologists have been thinking of specific oncolysis caused by viruses. Initial experiments were not successful, since the cytopathogenic viruses used (Sindbis, vaccinia) were infectious for both normal and tumor cells. One of possible solutions is retargeting the virus by bispecific scFv with one arm directed against the viral surface antigen (adenovirus knob) and the other against a tumor-associated antigen (CA IX). This virus-antibody complex binds preferably CA IX-expressing tumor cells and eventually destroys them (Jongmans et al, 2003). However, the next virus generation is again a plain adenovirus with no preference for tumor cells.

There are also several theoretical possibilities to target CA IX function for the anticancer therapeutic purposes and/or utilize the mode of its regulation. These possibilities are based on a solid rationale but certainly require experimental proof of principle.

CA IX-specific sulfonamides as efficient enzyme inhibitors might potentially reduce CA IX-mediated extracellular acidification of the tumor microenvironment and neutralization of the intracellular pH, thereby decreasing invasion propensity and/or survival of tumor cells. Sulfonamide inhibitors might be also applied together with the conventional chemotherapeutic drugs to improve their uptake and efficiency. This idea is based on the experiences that intracellular accumulation of weakly electrolytic drugs and their therapeutic effects depend on the pH gradient across the plasma membrane (Gerweck, 1998; Raghunand et al, 1999; Stubbs et al, 2000). Indirect arguments in favor of the above assumptions reside in observations that in vitro invasion of tumor cells and in vivo tumor growth in xenotrasplanted animals can be diminished by a non-selective CA inhibitor acetazolamide (Teicher et al, 1993; Parkkila et al, 2000), and that different derivatives of CA inhibitors can retard tumor cell growth in culture even in nanomolar concentrations (Casini et al, 2002). However, the mechanism of sulfonamide action and possible involvement of CA IX in those studies remains unclear. . An important step toward the elucidation of the phenotypic consequences of CA IX inhibition has been recently made by the synthesis of the first CA IX-selective, membrane-impermeable sulfonamides (Casey et al, 2004; Pastorekova et al, 2004).

Function-blocking antibodies might represent alternative CA IX-targeted therapeutic tools, although at this point their potential usefulness is merely a matter of speculation based on the analogy with antibodies against some other cell surface molecules, such as EGFR and c-ErbB2. Nevertheless, series of CA IX-specific monoclonal antibodies is available (Zatovicova et al, 2003) and awaits evaluation of possible anticancer effects.

In spite of the increasing number of scientists and clinicians interested in CA IX, there are some fundamental questions which, when answered, could provide useful leads for drug development. As we see it, these are as follows:

• What are the cell surface receptors/ligands binding CA IX?
• Are the binding partners the same in normal stomach, in the intestines and in various tumors?
• What signals (if any) are transmitted between CA IX and the interacting proteins and what are the downstream targets/effectors?
• Which are additional regulatory pathways upstream of CA IX and what is their biological significance?
• What are the mechanisms underlying intratumoral and intertumoral heterogeneity of CA IX distribution?
• What are the principles determining onset of expression in the course of carcinogenesis, quantity of the protein and percentage of CA IX-positive cells, why CA IX is expressed in tumor stroma?
• What are the phenotypic consequences of CA IX expression in different tumors?

This review essentially supports the view that CA IX could be useful for targeted cancer therapy. More than sixty papers have reported a partial success, at least in vitro or in animal experiments. What is needed is the amplification of the therapeutic effects. We are showing that several entirely different approaches have been designed. The survey suggests that CA IX study is quite stimulating not only to mount the researchers' imagination, but also to prompt clinicians' interests for practical improvements of the management of cancer patients.

The research of the authors is supported by Bayer Healthcare, the Academy of Sciences of the Czech Republic (project No. K5011112), the Slovak Scientific Grant Agency (VEGA — 2/3055/23), the Science and Technology Assistance Agency (APVT—51—005802) and by the 6^th Framework Program of the European Commission (EUROXY). The authors wish to thank Dr. Juraj Kopacek (Institute of Virology, Bratislava) for the help with the graphical illustrations and to Dr. Eva Sloncova (Institute of Molecular Genetics, Prague) for the photomicrographs with IHC staining.