Carbonic Anhydrase IX Promotes Tumor Growth and Necrosis In Vivo and Inhibition Enhances Anti-VEGF Therapy

The increased metabolic production, and impaired removal, of CO₂ and lactic acid can result in a highly acidic extracellular microenvironment (1, 2). Adaptation to this is crucial for cancer progression (3). Hypoxia-inducible factor (HIF), coordinates a shift in gene expression pattern, as part of the hypoxic response (4, 5). This includes increased expression of extracellular carbonic anhydrases IX (CAIX) and XII (CAXII; ref. 6). CAIX and CAXII expression has been identified in numerous cancer types and are generally associated with hypoxia (7–11). CA9 expression is regulated by HIF-1α where the HRE/HIF-1α–binding site is located at −3 of −10 position in the promoter region of this gene (6, 12). CAIX and CAXII facilitate transmembrane removal of CO₂ (a weak acid) by hydrating extracellular CO₂ and thereby maintaining a steeper efflux gradient (13). We recently identified a major role for CAIX in regulating intracellular and extracellular pH in spheroids (14). CAIX expression maintained a more alkaline and uniform intracellular pH while producing a more acidic extracellular pH (14). We identified that the majority of cellular acid is excreted across the membrane in the form of CO₂ rather than lactic acid (15), highlighting the importance of extracellular CAs in maintaining favorable intracellular pH. The expression of CAIX is restricted in normal adult tissues (16, 17) and disruption of the mouse homolog of CAIX during development resulted in only gastric hyperplasia (18), therefore making it a good therapeutic target. CAIX is a marker for poor clinical outcome in most cancer types (10, 19–21).

There is evidence of resistance to therapy in almost all anti-VEGF clinical trials (22, 23). Resistance mechanisms may include vascular regrowth due to additional proangiogenic factors such as Delta-like ligand 4 (DLL4; ref. 24); protection of the tumor vasculature by increasing recruitment of pericytes, which densely cover vessels or proangiogenic inflammatory cells, such as proangiogenic monocytic cells; or through co-opting normal vasculature by increased invasion into local tissues as reviewed by Bergers and Hanahan (25). In addition, antiangiogenic therapies induce hypoxia (26). This is likely to promote selection or adaptation of cells able to proliferate and survive in oxygen- and nutrient-deficient environments. Hypoxic adaptation occurs through HIF-1α stabilization and increased target gene expression. This hypoxic response could provide a further mechanism of resistance enabling tumors to overcome increased hypoxic conditions. CAIX expression was associated with poorer clinical outcome in response to antiangiogenic therapy (bevacizumab) in patients with metastatic colorectal cancer (27), malignant astrocytoma (28), and with worse progression-free survival in recurrent malignant glioma (29). VEGF, the target of bevacizumab and another target gene of HIF-1α, was not associated with worse outcome in these studies (27–29). These data suggest that CAIX may contribute to this angiogenesis-independent mechanism of resistance.

Two recent investigations of the role of CAIX in xenograft growth showed that CAIX knockdown by short hairpin RNA (shRNA) reduced xenograft tumor volume in one colon cancer cell line (30), one mouse breast cancer cell line, and one human breast cancer cell line (31). One of these studies showed upregulated CAXII expression in response to CAIX knockdown (26), where dual knockdown of both CAIX and CAXII reduced xenograft growth further (30). In this study, we constitutively knocked down CAIX expression in HT29 which have high levels of hypoxia-induced CAIX, to enable analysis of CAIX function. Conversely, we overexpressed CA9 in HCT116 which have low levels of hypoxic CAIX and CAXII.

We found that the capacity of CAIX to hydrate CO₂ is negatively affected by acidity, thereby producing a self-limiting mechanism if excess acid accumulates. Overexpression of CAIX in HCT116 increased spheroid and xenograft growth rate. CAIX knockdown in HT29 cells reduced spheroid and xenograft growth rate. Surprisingly, CAIX expression increased necrosis in spheroids and xenografts (HT29 and HCT116) and increased markers for apoptosis in spheroids (HT29 and HCT116) and in vivo (HT29). CAIX expression also increased Ki-67 staining, a marker for proliferation (HT29). CAXII expression was upregulated in response to CAIX knockdown in spheroids and xenografts although not in 2-dimensional (2D) culture. We identified increased CA9 expression in response to bevacizumab treatment in U87 (glioblastoma cell line) xenografts. Furthermore, we investigated the role of CAIX in resistance to bevacizumab treatment in HT29 and U87 xenografts. The results show an enhanced effect, significantly reducing growth rate.

Cells were maintained in a humidified incubator at 5% CO₂ and 37°C. For hypoxic exposure, cells were grown at 0.1% O₂, 5% CO₂, and 37°C. HCT116 were maintained in McCoy's 5A while HT29 and U87 were maintained in Dulbecco's Modified Eagle's Medium (DMEM) both supplemented with 10% FBS. For spheroid culture, aggregation was initiated by resuspending 4 × 10⁶ cells in 250 mL of medium in 1 L spinner flasks (Techne MCS) spun at 40 rpm at 37°C, 5% CO₂ or as described previously (32).

The following constructs were used to generate stable cell lines: human CA9 cDNA (FLCA9; a gift from Dr. Jaromir Pastorek, Slovak Academy of Science, Bratislava, Slovak Republic) was cloned into pcDNA3.1(+) (Invitrogen). To knockdown CA9 in HT29, the HuSH-29 shRNA targeting CA9 (TR314250) and empty vector (R20003) were purchased from Origene. Cell lines were transfected with FuGENE 6 (Roche) according to manufacturer's instructions. To knockdown CA9 in U87 a doxycycline inducible shCA9 (V3THS_363219) and shControl (shCTL; RHS4743) vectors were purchased from Thermo Scientific and lentivirus produced by the trans-lenti shRNA packaging kit (TLP4615) and cells transduced according to manufacturers instructions (Thermo Scientific). Cells were grown under selective pressure [HCT116, 0.4 mg/mL G418 (Invitrogen); HT29, 300 ng/mL puromycin; U87, 1 μg/mL puromycin (Invitrogen)] until no mock-transfected cells remained. Individual clones were isolated with cloning cylinders (Sigma).

Cell lysates were separated on 10% SDS-PAGE and transferred to polyvinylidene difluoride membrane. Primary antibodies were used at 1:1,000 unless otherwise stated. These were mouse anti-HIF-1α, (BD Transduction Laboratories), mouse anti-CA9 (Gift from J. Pastorek, Institute of Virology, Slovak Republic), goat anti-CA12 (R&D Systems), mouse anti-Na⁺/K⁺-ATPase (1:5,000; Abcam), and mouse anti-β-tubulin (Sigma). Appropriate secondary horseradish peroxidase (HRP)-linked antibodies were used (Dako). Immunoreactivity was detected with chemiluminescence (Amersham).

RNA extraction and the quantitative PCR protocol have been described previously (33). Primer sequences: CA9 forward, CTTGGAAGAAATCGCTGAGG; CA9 reverse, TGGAAGTAGCGGCTGAAGTC; CA12 forward, GCTCTGAGCACACCGTCA; CA12 reverse, GGATAAAGGTCTGAGTTATAATGGACA; glyceraldehyde-3-phosphate dehydrogenase (GAPDH) forward, AGCCACATCGCTCAGACAC; GAPDH reverse, GCCCAATACGACCAAATCC.

Cells were detached and resuspended on ice in 100 μL of PBS, at a concentration of 2 × 10⁶ cells per mL. Ten microliters of FITC Labeled CAIX (R&D systems) or isotype IgG antibody (R&D systems) was added and incubated for 1 hour at 4°C. Cells were washed with PBS and resuspended in 500 μL of PBS. Samples were analyzed by flow cytometry using a FACSCalibur flow cytometer (BD Biosciences) and CellQuest software (BD Biosciences).

HCT116 cells were harvested to a total cell count of the order 10⁷ to 10⁸ in HEPES-buffered DMEM. Cells were lysed in 20 mmol/L HEPES, 20 mmol/L Mes, protease inhibitor (Roche), 30 mmol/L KCl, 0.5% Triton X, and a pH adjusted to the desired starting point between 6.5 and 8. The membrane fraction was separated from the cytosolic fraction by centrifugation at 100,000 rpm for 40 minutes. The pellet, containing the membrane fraction, was resuspended in fresh lysis buffer. The total CA activity was measured by a kinetic assay (as previously described; ref. 14). The pH sensitivity of CA activity was elucidated by adjusting lysis buffer pH to 6.5, 7.0, 7.5, and 8.0. CA activity was quantified relative to the spontaneous rate measured in lysis buffer only or in extracts treated with the broad spectrum CA inhibitor, acetazolamide (100 μmol/L). The CO₂ hydration rate constant was estimated by fitting experimental pH time courses with model predictions, run for different rate constants. The model consisted of 3 ordinary differential equations for pH, CO₂, and HCO₃⁻ and included pH sensitivity of the CO₂ hydration rate constant. Kinetic data were normalized to the protein content (per 1 mg/mL), measured by the Bradford assay (Biorad). The pH dependence of CA activity was fitted with a Hill equation, featuring the spontaneous hydration rate (k_spont), the maximum catalyzed activity (k_max), half-maximal pH (pK), and cooperativity (n or Hill number):

Mice were housed at Cancer Research UK Laboratories (Clare Hall) and procedures were carried out under a Home Office license. Cells were trypsinized and washed twice in serum-free medium before inoculation in mice. Six- to 7-week-old female BALB/c nu/nu mice were injected subcutaneously in the lower flank with 100 μL Matrigel (BD Bioscience) and 1 × 10⁷ cells suspended in 100 μL of serum-free medium. Tumor growth was monitored 3 times per week measuring the length (L), width (W), and height (H) of each tumor with calipers. Volumes were calculated from the formula 1/6 × π × L × W × H. When tumors reached 1.44 cm³ the mice were sacrificed by cervical dislocation. Ninety minutes before sacrifice, mice were injected intravenously with 2 mg of pimonidazole (hypoxyprobe-1; Chemicon International) as described previously (24). For bevacizumab treatment mice were injected intraperitoneally every 3 days (10 mg/kg), until sacrifice, once 2 of 5 mice from any group had xenografts which had reached 150 mm³ in size. If required, doxycycline was given through the feed at 625 mg/kg ad libitum (SDS services) from the start of the experiment.

Immunohistochemistry was carried out as previously described (24). The following primary antibodies were used at room temperature for 1 hour; CAIX (M75, 1:50); HIF-1 (BD Biosciences; 1:100); pimonidazole (hypoxyprobe-1; Chemicon International; 1:50); CD34 (MCA1825; AbD Serotec; 1:50), cleaved caspase-3 (RnD systems; 1:2,000), CAXII (1:1,000; ref. 34), and Ki-67 (M7240; Dako; 1:50). Slides were incubated with the anti-rabbit/anti-mouse secondary antibody (Dako) for 30 minutes at room temperature and washed in PBS. 3,3′-Diaminobenzidine (DAB; Dako) was applied to the sections for 7 minutes. The slides were counterstained by immersing in hematoxylin solution (Sigma-Aldrich) for 20 seconds and mounted with Aquamount (VWR). Secondary-only control staining was done routinely, these were negative. One section from each xenograft was analyzed for each stain. Slides were scored by 2 researchers blinded to groupings (as a percentage or index based on intensity and percentage positive) or analyzed quantitatively by image analysis in imageJ using color deconvolution as described previously (35). Where scores differed, sections were reviewed and a consensus result was decided.

Statistical analysis including the Student t test, one-way ANOVA, and linear regression of log transformed growth data were carried out as appropriate with GraphPad Prism 4.0b.

HCT116 cells were transfected with the CA9 coding sequence (FLCA9) or empty vector (EV). CAIX was overexpressed relative to empty vector shown by immunoblot and flow cytometry (Fig. 1A).

HT29 was transfected with an shRNA targeting CA9 (shCA9) or empty vector. CAIX knockdown in shCA9 was shown by quantitative PCR and immunoblot compared with empty vector (Fig. 1B). The expression of CA9 in shCA9 was 6% the empty vector levels at 0.1% O₂ for 72 hours (***, P < 0.001, n = 3).

U87 cells were transduced with an shCA9 or a nontargeting shRNA (shCTL). CA9 knockdown in shCA9 was shown by quantitative PCR (***, P < 0.001, n = 3; Fig. 1C) and by immunoblotting at a range of doxycycline concentrations (Fig. 1D).

HCT116 FLCA9 and empty vector membrane extracts were prepared and tested for CA activity. The subcellular fractionation was assessed by immunoblotting the membrane fraction for Na⁺/K⁺-ATPase, a membrane-specific protein (Fig. 1E). Membrane extracts from empty vector cells showed hydration at the spontaneous rate, with similar CO₂ hydration rates in the presence of acetazolamide, a CA inhibitor. In FLCA9, CAIX was catalytically active in the membrane fraction and inhibition to basal empty vector levels occurred in the presence of acetazolamide. CA activity was markedly reduced by decreasing pH, with CA activity almost completely inhibited at pH 6.0 and a pK (pH for half-maximum rate) between 6.81 and 6.86 (Fig. 1F). Best-fitting Hill curves to the pH-sensitivity data yielded Hill cooperativity numbers near 2 for full-length CAIX. As a consequence of this, the pH sensitivity of CAIX to pH is steeper than for other CA isoforms (36).

We found a significant increase in growth rate in HCT116 FLCA9 spheroids (***, P < 0.001, n = 3; Fig. 2A) and xenografts (***, P < 0.001, n = 10; Fig. 2B) compared with empty vector.

CAIX knockdown reduced the growth rate of HT29 spheroids (*, P < 0.05, n = 3; Fig. 2C) and xenografts (*, P < 0.05, n = 5; Fig. 2D) compared with empty vector.

We characterized the spheroids and xenograft tumors for several immunohistologic markers related to cell survival and angiogenesis.

In HCT116 spheroids, the percentage of CAIX-positive cells was greater in FLCA9 (95%) than empty vector (5%; ***, P < 0.001, n = 5; Fig. 3A). Similarly in HCT116 xenografts, the percentage of CAIX-positive cells was greater in FLCA9 (FL/1, 55% and FL/3, 75.5%) than empty vector (EV; EV/5, 12.5% and EV/6, 13.5%; ***, P < 0.001, n = 10; Fig. 4A).

In HT29 spheroids, CAIX expression was decreased in shCA9 (2.5%) compared with empty vector (18%; ***, P < 0.001, n = 5; Fig. 3B). The same was true in the xenografts shCA9 (6%) compared with empty vector (17%; ***, P < 0.001, n = 5; Fig. 4B).

In HCT116 spheroids, the proportion of necrosis was higher in FLCA9 (15%) than in empty vector (5%; *, P < 0.01, n = 5; Fig. 3A). The proportion of necrosis was significantly higher in FLCA9 HCT116 xenografts (FL/1, 39.5% and FL/3, 28.6%) than in empty vector (EV/5, 6.9% and EV/6, 12.1%; *, P < 0.05, n = 10; Fig. 4A).

Conversely in the HT29 spheroids and xenografts, the shCA9 had less necrosis (spheroids, 19% and xenografts, 13%) than empty vector [spheroids, 36% and xenografts, 35%; spheroids, ***, P < 0.001, n = 5 (Fig. 3B) and xenografts, *, P < 0.05, n = 5 (Fig. 4B)].

HCT116 FLCA9 spheroids showed increased apoptosis compared with empty vector as determined by immunostaining for activated caspase-3 (2.9% and 6.0%, respectively, ***, P < 0.001, n = 5; Fig. 3A) although this was not found in the xenografts (Fig. 4A).

Increased apoptosis was found in the HT29 empty vector spheroids and xenografts compared with shCA9 (spheroids, 9% and 5%; ***, P < 0.001, n = 5; Fig. 3B and xenografts; *, P < 0.05, n = 5; Fig. 4B).

In HT29 xenografts, the percentage of Ki-67–positive nuclei was higher in empty vector (89%) than in shCA9 (55%, ***, P < 0.001, n = 5; Fig. 4B). In HT29 spheroids, Ki-67 positivity was higher in hypoxic areas in empty vector (32%) than in shCA9 (15%, ***, P < 0.001, n = 5; Fig. 5A). Hypoxic areas were defined by pimonidazole staining. There was no difference in Ki-67 positivity in HCT116, FLCA9 or empty vector spheroids (Fig. 3A), or xenografts (Fig. 4A).

Reduced pimonidazole staining was associated with CAIX expression in spheroids, HCT116 FLCA9 (15%) and empty vector (20%, *, P < 0.05, n = 5; Fig. 3A), HT29 empty vector (32%), and shCA9 (55%, ***, P < 0.001, n = 5; Fig. 3B). This was due to the increase in necrosis in hypoxic areas of CAIX-positive spheroids, which reduced the amount of viable hypoxic tissue. CAIX expression did not effect the distance from the edge of the spheroids that pimonidazole staining began. There was no difference in pimonidazole staining in xenografts in HCT116 (Fig. 4A) or in HIF-1α staining in HT29 (Fig. 4B).

CAIX expression did not affect CD34 staining, a marker for blood vessels in xenografts in HCT116 (Fig. 4A) and HT29 (Fig. 4B).

CAIX increased growth rate while also increasing necrosis. Therefore, it was unclear whether CAIX expression increased the amount of viable tumor tissue at a given time point. To calculate the volume of the viable tumor, we subtracted the percentage of necrotic tumor (determined by histology, Figs. 3 and 4) from the endpoint volumes of the xenografts and spheroids expressing CAIX. These viable tumor volumes of the CAIX expressors were greater than the volumes of the spheroids and xenografts which lacked CAIX expression at the same time point. However, the latter were allowed to continue to grow to assess final growth curves.

This analysis revealed that the CAIX expressors had increased viable tumor volume in HCT116 xenografts (66% more viable tumor at day 14, ***, P < 0.01, n = 10) and spheroids (19% more viable tumor at day 13, **, P < 0.01, n = 5) and HT29 xenografts (52% more tumor tissue at day 20, P < 0.05, n = 5) and spheroids (81% more viable tumor at day 16, *, P < 0.01, n = 5; Supplementary Fig. S1).

Although we acknowledge a likely overestimation in the amount of viable tumor in the CAIX-deficient groups (HCT116 empty vector and HT29 shCA9), as necrosis was not subtracted from these volumes, this would only make the results more significant. Necrosis was not subtracted from the CAIX-deficient groups (HCT116 empty vector and HT29 shCA9) because data on the amount of necrosis was only available at the time of sacrifice of these controls, not the earlier date on which the CAIX expressors were ended.

In HT29, no effect was identified on CAXII RNA or protein levels in response to CAIX knockdown in 2D culture (Fig. 1B). In 3-dimensional (3D) culture CAIX knockdown increased CA12 expression at the RNA (by ∼2-fold, P < 0.05, n = 5; Fig. 5B) and protein level by immunohistochemistry (from 1% to 2% of cells; ***, P < 0.001, n = 5; Fig. 5D). In vivo CAIX knockdown did not increase CA12 RNA expression (Fig. 5C) but did increase CAXII protein expression in shCA9 (5%) compared with empty vector (2%; *, P < 0.05, n = 5; Fig. 5E).

We identified that CA9 expression was increased (***, P < 0.001, n = 5) in response to bevacizumab treatment with material from a previous xenograft experiment using, the glioblastoma cell line, U87 (ref. 24; Supplementary Fig. S2). HT29 empty vector and shCA9 were grown as xenografts with and without bevacizumab treatment (Fig. 6A). The shCA9 clone grew more slowly than the empty vector clone as before (75.9% the growth rate of the empty vector; *, P < 0.05, n = 5). Bevacizumab treatment also reduced xenograft growth rate (68% the rate of empty vector control; *, P < 0.05, n = 5). The HT29 shCA9 clones treated with bevacizumab grew significantly more slowly than shCA9-untreated xenografts or empty vector bevacizumab-treated xenografts (37% the growth rate of untreated empty vector; *, P < 0.05, n = 5; Fig. 6A). This was at least an additive reduction in growth rate. shCA9 xenografts had reduced CAIX expression compared with empty vector clones when treated with vehicle (PBS; *, P < 0.01, n = 5) and with bevacizumab (***, P < 0.001, n = 5; Supplementary Fig. S3). HT29 cells were also treated with acetazolamide in combination with bevacizumab. Acetazolamide or bevacizumab treatment alone did not significantly reduce growth rate however, both in combination, reduced growth significantly compared with untreated (*, P < 0.05, n = 5) or bevacizumab-treated alone (*, P < 0.05, n = 3; Fig. 6B).

U87 xenografts with doxycycline-inducible shCA9 were treated with bevacizumab. The shCA9 without doxycycline and the shCTL with or without doxycycline xenografts grew at the same rate (Fig. 6C). CA9 knockdown (33% reduction, P < 0.05, n = 5) or bevacizumab treatment alone (38% reduction, P < 0.05, n = 5) significantly reduced xenografts growth rate. CA9 knockdown and bevacizumab treatment in combination significantly reduced growth rate further (>50%), more than CA9 knockdown (*, P < 0.05, n = 5) or bevacizumab treatment (*, P < 0.05, n = 5) alone.

Bevacizumab treatment reduced the number of blood vessels in the HT29 xenografts as determined by CD34 staining [empty vector: P < 0.05, n = 5; shCA9: P < 0.05, n = 5; vehicle: *, P < 0.05, n = 5; acetazolamide (ATZ): *, P < 0.01, n = 5; Supplementary Figs. S3 and S4]. Bevacizumab treatment increased the amount of hypoxia in xenografts as determined by HIF-1α immunostaining (empty vector: *, P < 0.05, n = 5; shCA9: *, P < 0.01, n = 5). CAIX immunostaining (empty vector: ***, P < 0.001, n = 5; vehicle: *, P < 0.05, n = 5; ATZ: **, P < 0.01, n = 5), the amount of necrosis (shCA9: *, P < 0.01, n = 5) and apoptosis as determined by caspase-3 activation (empty vector: ***, P < 0.01, n = 5; shCA9: *, P < 0.05, n = 5′ Supplementary Figs. S3 and S4) was also increased with bevacizumab treatment.

We recently highlighted the importance of CAIX in regulating intracellular and extracellular pH in spheroids (14, 15, 37, 38). In addition, the proteoglycan domain of CAIX has previously been linked to negative regulation of cell adhesion through modulation of β-catenin and E-cadherin interactions (39), whereas the intracellular domain has been shown to contribute to AKT activation (40). To understand the role of CAIX in spheroids and in vivo, we knocked down CAIX in HT29 cells, which have high levels of CAIX under hypoxic stress. Several cell lines and human tumors have low levels of CAIX expression or induction by hypoxia, and the HCT116 cells provided an opportunity to investigate the role of CAIX when uniformly expressed in a tumor. Further to this, we examined the role of CAIX in resistance to severe hypoxia induced by bevacizumab treatment in HT29 and U87 cells.

We found that CAIX catalytic activity is inhibited by low pH and was half-maximal (i.e., the pK) at pH approximately 6.8. This is lower than the typical value for other CA isoforms (36), but in the range of extracellular pH measured in human tumors (pH6.5–6.8; refs. 41, 42). Consequently, H⁺ ions, produced by CO₂ hydration in the tumor milieu, can exert an important autoinhibitory effect on CAIX. The pH dependence of CAIX activity showed a cooperativity of approximately 2 [higher than reported for a truncated recombinant CAIX (ref. 43) and for other CA isoforms (ref. 36)], making CAIX activity more H⁺ sensitive as pH approaches the enzyme's pK. CAIX's relatively low pK will allow the tumor to attain a more acidic extracellular pH, whereas higher cooperativity may protect from excessive acidification by strengthening the autoinhibitory H⁺ effect below a critical pH. Although there is pH inhibition of CAIX activity, it is not autoinhibited abruptly, but gradually, and therefore significant activity is still measurable at pH 6.5. We have modeled the relationship between CA activity and extracellular pH (15).We showed that “basal” CAIX activity would have to be very mild to be completely switched off by extracellular H⁺ ions.

We show that CAIX knockdown reduced the growth rate of xenografts as has been reported in 2 recent studies (26, 31). In addition, we identify the converse, whereby overexpression of CAIX in HCT116 increased growth rate. However, we identified, for the first time, that CAIX expression was associated with increased necrosis. This was surprising as we initially expected CAIX to reduce necrosis by improving intracellular pH control. The overall effect of this was increased tumor volume with more viable tumor and more necrosis at the same time point in the CAIX expressors, of both HCT116 and HT29 cell lines. The increase in apoptosis associated with CAIX offers a mechanism to explain the increased pathologic necrosis.

We hypothesize that the necrosis may represent regions where CAIX expression has maintained a high level of proliferation (as shown in Fig. 5A) through the maintenance of a more uniformly alkaline intracellular pH. This would increase metabolic demand and deplete substrates in the hypoxic milieu. We previously identified that necrosis also induces infiltration of macrophages which may contribute to an aggressive tumor phenotype (44, 45) and necrosis is associated with poor prognosis in many cancer types including colorectal cancer (46). So the well-recognized observation of CAIX staining around necrotic areas may reflect a causative role in necrosis.

CAIX expression was associated with Ki-67 positivity in HT29, particularly in the hypoxic areas in HT29 spheroids. This is in agreement with data that CAIX maintains an intracellular environment conducive for growth under hypoxic conditions. CAIX expression increased growth in HCT116 spheroids and xenografts, however there was no difference in Ki-67 staining. While Ki-67 stains positive proliferative cells, staining does not relate to the time required for an intermitotic cycle completion (47).

A recent study identified CAXII upregulation in response to CAIX knockdown in 2D culture at the RNA and protein level (30). We have identified CAXII upregulation in response to CAIX knockdown in spheroids and in vivo, but not in 2D culture. The cause of this difference requires further study, but it may be related to differences in pH or oxygen tension between 2D and 3D culture.

We investigated CAIX knockdown in combination with antiangiogenic therapy (bevacizumab) in 2 tumor types. CAIX was recently associated with poor outcome in patients with metastatic colorectal treated with bevacizumab (27). Here, we show that CAIX knockdown and bevacizumab treatment acted additively in reducing xenograft growth rate in the colon adenocarcinoma cell line HT29 and synergistically in the glioblastoma cell line U87. In addition, the CA inhibitor acetazolamide enhanced bevacizumab treatment in HT29 xenografts. Our results provide experimental evidence that is consistent with previous immunohistologic data (27–29), showing that CAIX is part of a resistance mechanism that enables tumors to adapt to the increased hypoxia induced by bevacizumab treatment. On the basis of these data, detection of CAIX induction levels, possibly detected in blood serum or urine (48), in response to bevacizumab could help in predicting outcome to bevacizumab treatment. This highlights also the general principle of targeting the hypoxia survival mechanisms in combination with bevacizumab treatment, particularly given the chemotherapy and radiotherapy resistance of such hypoxic areas.

The effect of CAIX knockdown, reducing xenograft tumor growth alone, and at least additively in combination with antiangiogenic therapy, and the association between CAIX expression and increased necrosis, identifying a further mechanism to explain CAIX association with poorer outcome, highlight potential mechanisms by which CAIX inhibitory therapeutics, such as antibodies (49) and small-molecular inhibitors (50), may be used optimally.

No potential conflicts of interest were disclosed.

Conception and design: A. McIntyre, S. Wigfield, J.-L. Li, A.L. Harris

Development of methodology: A. McIntyre, S. Patiar, S. Wigfield, J.-L. Li, H. Turley, C. Snell, K.C. Gatter, R.D. Vaughan-Jones, A.L. Harris

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. McIntyre, S. Patiar, S. Wigfield, J.-L. Li, I. Ledaki, R. Leek, K.C. Gatter, R.D. Vaughan-Jones, P. Swietach

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. McIntyre, S. Wigfield, J.-L. Li, I. Ledaki, R. Leek, C. Snell, K.C. Gatter, P. Swietach, A.L. Harris

Writing, review, and/or revision of the manuscript: A. McIntyre, S. Patiar, S. Wigfield, J.-L. Li, R. Leek, K.C. Gatter, W.S. Sly, P. Swietach, A.L. Harris

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. McIntyre, S. Patiar, S. Wigfield, H. Turley, R. Leek, K.C. Gatter

Study supervision: A. McIntyre, K.C. Gatter, A.L. Harris

Provided reagent: W.S. Sly

This work was supported by funds from Cancer Research UK, EU Framework 7 Metoxia (to A.L. Harris), The Royal Society, Medical Research Council (to P. Swietach), and NIH grant GM034182 (to W.S. Sly).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.