Tumor-associated Carbonic Anhydrase 9 Spatially Coordinates Intracellular pH in Three-dimensional Multicellular Growths*

CA9 is a membrane-tethered, carbonic anhydrase (CA) enzyme, expressed mainly at the external surface of cells, that catalyzes reversible CO2 hydration. Expression is greatly enhanced in many tumors, particularly in aggressive carcinomas. The functional role of CA9 in tumors is not well established. Here we show that CA9, when expressed heterologously in cultured spheroids (0.5-mm diameter, ∼25,000 cells) of RT112 cells (derived from bladder carcinoma), induces a near-uniform intracellular pH (pHi) throughout the structure. Dynamic pHi changes during displacements of superfusate CO2 concentration are also spatially coincident (within 2 s). In contrast, spheroids of wild-type RT112 cells lacking CA9 exhibit an acidic core (∼0.25 pHi reduction) and significant time delays (∼9 s) for pHi changes in core versus peripheral regions. pHi non-uniformity also occurs in CA9-expressing spheroids after selective pharmacological inhibition of the enzyme. In isolated RT112 cells, pHi regulation is unaffected by CA9 expression. The influence of CA9 on pHi is thus only evident in multicellular tissue. Diffusion-reaction modeling indicates that CA9 coordinates pHi spatially by facilitating CO2 diffusion in the unstirred extracellular space of the spheroid. We suggest that pHi coordination may favor survival and growth of a tumor. By disrupting spatial pHi control, inhibition of CA9 activity may offer a novel strategy for the clinical treatment of CA9-associated tumors.

Intracellular H ϩ ions are powerful modulators of physiological and biochemical processes, including cell growth, division, and development (1)(2)(3). Many metabolic reactions within the cell generate H ϩ ions as end-products (4). Because of their high chemical reactivity, particularly with intracellular proteins, these ions must be removed if normal cell function is to continue. Tumors display a notably high rate of H ϩ ion production, partly because their growth demands a ready supply of energy, but also because much of the metabolism that supports this is anaerobic (5). Energy production via extramitochondrial routes is associated with a larger metabolic flux of acid per ATP molecule (4). Indeed, growing tumors are typically hypoxic, resulting in stabilization of hypoxia-inducible factor 1␣, which upregulates a variety of proteins, including enzymes that control glycolytic lactic acid production (6). If intracellular H ϩ ions accumulate, intracellular pH (pH i ) 3 falls and cellular activity becomes compromised. One solution is for cells to express plasmalemmal ion transport proteins that extrude excess H ϩ equivalents (7). Extrusion is achieved either by expelling H ϩ ions directly (e.g. via Na ϩ -H ϩ exchangers, H ϩ -lactate co transporters, and H ϩ -ATPases), or by importing HCO 3 Ϫ anions (e.g. via Na ϩ -dependent HCO 3 Ϫ transporters), to neutralize intracellular H ϩ ions. The pH i of tumor cells, like that of most other eukaryotic cells, appears to be relatively alkaline (pH i ϳ 7.2), indicating that they possess an efficient system for eliminating H ϩ ions (8,9).
In many cell types the enzyme carbonic anhydrase (CA) plays an important role in the regulation of pH i . The ␣-CA family expressed in eukaryotic cells comprises at least 13 catalytically active isozymes (10). These enzymes, in effect, catalyze the reversible buffering of H ϩ by HCO 3 Ϫ ions. This generates CO 2 , which is highly membrane-permeant, and therefore diffuses out of the cell. Despite significant contributions from anaerobic metabolism, many growing tumors produce large quantities of CO 2 , suggesting that chemical buffering by HCO 3 Ϫ is an important strategy for dealing with metabolic acid production (11,12). CA can further assist pH i regulation by binding to various HCO 3 Ϫ transporters forming a protein complex, a transport metabolon (13,14). CA activity may enhance the H ϩ equivalent flux of the transporter (14,15), thus increasing the efficiency of pH i regulation, although this mechanism has been disputed (16). The enzyme also facilitates H ϩ ion diffusion inside cells (17), which would otherwise be slowed considerably by the presence of intracellular fixed buffer sites, particularly on proteins. In addition, CA can facilitate CO 2 diffusion by reversibly converting the gas into diffusible HCO 3 Ϫ anions (18).
CA9 is an isoform highly expressed by many developing tumors (19 -21), including von Hippel-Lindau-defective tumors (19,22). CA9 expression correlates with hypoxia (21,23,24) and is regulated by hypoxia-inducible factor 1␣ (25). CA9 is now an established marker for several aggressive cancers such as breast cancer (26,27) and non-small cell lung cancer (28). Generic inhibitors of CA activity, like acetazolamide (ATZ), have been used experimentally to attenuate tumor growth, suggesting an important role for the catalytic activity of the enzyme (29,30). Specific targeting of CA9 by RNA interference also reduces growth and survival under hypoxic conditions in cell monolayers (31). Despite this evidence, the nature and mechanism of CA9 involvement are not well established. The enzyme is anchored principally to the surface membrane of tumor cells, with its active site oriented toward the extracellular space. It is therefore likely to catalyze the hydration of CO 2 that emerges from a cell, producing H ϩ and HCO 3 Ϫ ions. This may then facilitate the diffusion of more CO 2 from the cell. Enhanced extracellular CO 2 hydration would reduce local extracellular pH (pH e ) (19,32). Indeed, pH e in poorly vascularized tumors is notably acidic (ϳ6.9), a feature that is thought to encourage host-tissue regression, thereby assisting further tumor development (33)(34)(35). Although possible effects of CA9 on pH e have been recognized, there has been less consideration of its effect on pH i . Nevertheless, recent reports suggest that, in heterologous expression systems, CA9 forms a functional transport metabolon with HCO 3 Ϫ transporters (13), suggesting a specific role for the enzyme in pH i regulation.
In the present work we have explored the contribution of CA9 to the regulation of pH i in three-dimensional multicellular growths. We have developed a system for confocally imaging pH i spatially within cultured spheroids of RT112 cells, using the intracellular pH fluorophore, carboxy-SNARF-1. These cells, derived from bladder tumor cells, do not constitutively express CA9. Multicellular spheroids can therefore be grown that lack CA9, and these can be compared with others grown from RT112 cells stably transfected with the enzyme. We find that, although CA9 does not enhance H ϩ equivalent transport in isolated RT112 cells, its presence exerts a remarkable effect on the spatial control of pH i within a multicellular spheroid. We develop a three-dimensional computational model to explore the relationship between extracellular CA9 expression and the spatial coordination of pH i . Results of both the experiments and the model suggest a novel role for the enzyme in reducing pH i non-uniformity in multicellular structures. This role may provide insight into the functional importance of CA9 in tumor development.

EXPERIMENTAL PROCEDURES
Stable Transfection of RT112 Cells-RT112 cells were transfected with the cDNA of human CA9 (a gift from Dr. J. Pastorek, Bratislava) or empty vector using a published technique (31). To grow spheroids, cells were seeded into agarose-coated 96-well plates (ϳ20,000 cells/well) and cultured for 72 h (36).
Western Blotting-Cells were homogenized in lysis buffer (6.2 M urea, 10% glycerol, 5 mM dithiothreitol, 1% SDS, protease inhibitor), and the whole cell extract was separated on 10% SDS-PAGE and transferred to polyvinylidene difluoride mem-brane using a semidry blotter (Wep Company, UK). Primary antibodies used were mouse monoclonal M75 to CA9 (a gift from Dr. J. Pastorek), CA2 (Abcam, UK), and mouse anti-␤tubulin monoclonal antibody (Sigma). Immunoreactivity was visualized with horseradish peroxidase-linked goat anti-mouse or anti-rabbit serum at 1:1000 and detected with enhanced chemiluminescence.
Caspase Staining-To assess apoptosis, spheroids were fixed in liquid N 2 and labeled immunohistochemically with mouse caspase-3 antibodies and rabbit polyclonal anti-mouse antibodies (0.1 g/ml, R&D Systems, UK).
Measuring Functional CA Activity-This was measured in (i) intact cells suspended in CO 2 -free, isotonic buffer (130 mM NaCl, 5 mM KCl, 20 mM Hepes) or in (ii) whole cell lysates (produced by hypotonic shock in 0.1% Triton X-100, 5-min sonication), suspended in 20 mM Hepes solution. Solution pH was adjusted to 8.0 at 4°C. 1.5 ml of suspension was added to a stirred reaction vessel, and pH was monitored (Biotrode, Mettler-Toledo, UK). 0.5 ml of 100% CO 2 -saturated water was added to the cell suspension/lysate, and the rate of acidification was analyzed to obtain the CO 2 hydration constant (38). Measured rates were normalized to 1 mg of total protein per ml, measured spectrophotometrically (Bradford assay, Bio-Rad).
Superfusion and pH i Imaging-A heated superfusion chamber (37°C, 1 ml) was mounted on an inverted microscope (DM-IRBE, Leica), its base (coverslip) was pre-treated with 0.01% poly-L-lysine to aid cell adhesion. Cells/spheroids were exposed to membrane-permeant acetoxymethyl ester of carboxy-SNARF-1 (20 M) for 20 min (39). Intracellular pH was imaged confocally (TCS, Leica) with a 10ϫ lens. Carboxy-SNARF-1 was excited by the 514 nm argon laser line, and fluorescence emission was collected simultaneously at 580 Ϯ 20 nm and 640 Ϯ 20 nm. Acquisition produced 512 ϫ 512 pixel spatial maps of fluorescence every 1.8 s. Fluorescence was offset for background and averaged in regions of interest (ROIs). In experiments involving isolated cells, ROIs were defined as discrete areas of fluorescence that exceeded a threshold of 10% peak fluorescence and of area between 20 and 120 m 2 (equivalent to a cell). In experiments performed on spheroids, the outer boundary of the cluster was used to define six layers of ROIs, with concentric centers. Average ROI fluorescence was ratioed (580 to 640 nm) to eliminate differences in cellular dyeloading, light-path and specimen thickness. This is of particular importance for spheroids, in which peripheral cells tend to concentrate dye more than core cells. Fluorescence ratios were then converted into pH i (39). The pH signal was intracellular as frame-averaging showed that fluorescence was confined to round objects corresponding to cell outlines. Also, the behavior of the pH signal was consistent with predicted intracellular events.
Measuring Cell Dimensions and Buffering Capacity-Cell dimensions were estimated from the mean radius of carboxy-SNARF-1-loaded isolated cells. Intrinsic intracellular H ϩ buffering capacity (␤ int ) was measured in single cells superfused with 20 mM Hepes-buffered solution (nominally CO 2 -free). Buffering due to CO 2 /HCO 3 Ϫ (␤ CO2 ) was measured as the increase in buffering when cells were superfused with 5% CO 2 /22 mM HCO 3 Ϫ -buffered solution. Cells were first superfused with solution containing 30 mM NH 4 Cl (generating an alkalosis) and then with a sequence of solutions containing lower doses of NH 4 Cl (20, 10, 5, and 0 mM). These solution maneuvers produced stepwise decreases of pH i (40). Buffering capacity was calculated from the quotient of the estimate change in [NH 4 ϩ ] i and the measured change in pH i (back-extrapolated to the moment of solution change). Inhibitors of H ϩ equivalent transport were included to inhibit slow pH i changes that could otherwise lead to inaccurate estimates of ␤. In CO 2 /HCO 3 Ϫ buffer, allowance was made (1-2 min) for buffer equilibration (41).
Measuring H ϩ Equivalent Membrane Transport-Single cells were superfused with solution containing 30 mM NH 4 Cl for 4 -6 min ("prepulse"). Upon returning to solution without NH 4 Cl, pH i acidifies. This pH i displacement stimulates H ϩ equivalent transport. At any given pH i , its rate of recovery (dpH i /dt) multiplied by the appropriate buffering capacity gives an estimate of the transmembrane H ϩ equivalent flux (J memb H ) (39). 2 Partial Pressure-The catalytic activity of CA can be studied by changing superfusate p CO2 and [HCO 3 Ϫ ] (at a constant pH e of 7.4). Experimentally, this was done by changing superfusate buffer from Hepes to CO 2 /HCO 3 Ϫ . The slope of the initial pH i change (measured over 15 s) was measured in selected ROIs following a rise or fall in p CO2 . To obtain a measure of H ϩ equivalent flux (J CO2 H ), the slope was multiplied by ␤ int .

RESULTS
CA Expression in RT112 Cells-RT112 cells were chosen for this study, because they do not express CA9 under hypoxia (Fig.  1A). This eliminates the possibility of additional, hypoxia-induced CA9 expression at the core of spheroids, which would introduce unwanted variability into the analysis. Instead, by using a stably transfected clone, comparable levels of CA9 could be introduced uniformly across spheroids, as described below. Western blots (Fig. 1A) indicated CA9 immunoreactivity in RT112 cells transfected with the CA9 gene (CA9 expressor). Based on a comparison with ␤-tubulin expression, the level of CA9 protein expressed with transfection was of a similar order of magnitude to that induced by hypoxia in breast cancer MDA231 and MDA468 cell lines (31). In contrast, cells transfected with empty vector (EV) displayed moderate levels of CA2 (an intrinsic, intracellular CA isoform), but no CA9. Of the clones illustrated in Fig. 1A, EV and CA9 expressor clones #2 were used for subsequent experiments.
Cell surface CA9 staining was also detected in multicellular spheroids grown from CA9 expressor but not from EV cells (Fig. 1B). Fluorescence-activated cell sorting analysis indicated Western blot for CA9, CA2 and ␤-tubulin (for normalization) protein immunoreactivity in lysates of various CA9 expressor and empty vector (EV) clones. Note the inverse relationship between CA2 and CA9 expression. EV and CA9 expressor clone 2 were used for subsequent experiments. B, CA9 antibodies detect surface antigen in sections through CA9 expressor (but not in EV) spheroids. CA9 expressor spheroids showed a lower apoptotic index (caspase-3 staining). C, the rate of acidification of medium containing CA9 expressor whole cell lysate, following addition of CO 2 at 4°C, provides an estimate of CA activity (n ϭ 4/trace) in the absence (black) or presence of drugs (100 M ATZ, green; 200 nM 14v, blue). D, data (n ϭ 6 -10/bar) collected from intact (left panel) and lysed (right panel) EV and CA9 expressor cells, normalized to the spontaneous hydration rate (measured in cell-free experiments, 0.0035s Ϫ1 ) and to total protein concentration.
a normal distribution of membrane CA9 in the transfectants (not shown) and, therefore, as expected, some variability in individual cell staining. Surface CA9 was detected throughout the CA9 expressor spheroid, unlike the situation in endogenous growths where hypoxia-induced up-regulation of CA9 typically occurs in the core. Absence of CA9 immunoreactivity in EV spheroids correlated with higher levels of staining for caspase-3, a marker for apoptosis (Fig. 1B). This suggests that CA9 expression may confer a degree of protection against cell death, in agreement with previous studies performed on unclustered cells (31).
CA activity was assayed functionally by recording the time course of acidification of medium containing intact or lysed CA9 expressor or EV cells, following addition of CO 2 -saturated solution (38). An example of the acidification time course (n ϭ 4) with CA9 expressor whole cell lysates is shown in Fig. 1C. The acidification was slowed to a spontaneous (uncatalyzed) rate (Fig. 1D) by adding the CA inhibitors, 14v or ATZ, confirming the presence of catalytic activity.
When using non-lysed CA9-expressing cells in the suspension, the acidification was still considerably faster (by 66%) than the spontaneous, uncatalyzed rate, indicative of CA9 activity at the external surface of the cells (Fig. 1D). The enhanced rate was inhibited by the membrane-impermeant CA inhibitor, 14v (200 nM), whereas no effect was seen for EV cell suspensions, confirming that the active site of the transfected CA9 was indeed extracellular. The increase in CO 2 hydration rate observed with CA9-expressing RT112 cells (Fig. 1D) is comparable to that reported previously in MDA468 breast cancer cells exposed to hypoxia (31). Thus the functional activity of CA9 in transfected RT112 cells matches that induced in hypoxia-responding tumor cells. Both CA9 protein expression and the resulting functional activity in RT112 expressor cells are therefore comparable to levels in hypoxically responsive tumor cell lines.
CA9 Effects on Cell Size and Buffering Capacity-To assess effects of CA9 on pH i regulation, we first examined if its expression altered cell size or cellular H ϩ buffering capacity. Analysis of fluorescence (carboxy-SNARF-1) images of cells showed no difference between the radii of CA9 expressor cells (4.89 Ϯ 2.14 (S.D.) m) and EV cells (4.97 Ϯ 1.82 m). Intracellular buffering capacity was assessed in isolated cells from the rapid fall of pH i (recorded with carboxy-SNARF-1) in response to reductions of extracellular NH 4 Cl ( Fig. 2A) (40). Intrinsic (non-CO 2 /HCO 3 Ϫ ) intracellular buffering (␤ int ) was estimated in the presence of Hepes-buffered superfusates. Additional buffering due to CO 2 /HCO 3 Ϫ (␤ CO2 ) was estimated in CO 2 /HCO 3 Ϫ -buffered superfusates (which measures ␤ int ϩ ␤ CO2 ). Values for ␤ int and ␤ CO2 were no different in CA9 expressor and EV cells, indicating that intracellular H ϩ -buffering capacity was not influenced by CA9 (Fig. 2B).
CA Activity in Intact Cells-In an isolated cell, switching from Hepes-buffered to CO 2 /HCO 3 Ϫ -buffered superfusate caused a fall of pH i , as CO 2 entering the cell was hydrated to H ϩ FIGURE 2. No effect of CA9 expression on single cell pH i regulation. A, intracellular H ϩ buffering capacity (␤) was estimated from the size of stepwise pH i changes produced by gradual ammonium removal, as shown for a single CA9 expressor cell, superfused by a sequence of Hepes-buffered solutions containing 30 to 0 mM NH 4 Cl. Possible effects on pH i of membrane H ϩ equivalent transport were minimized by including 0.5 mM 4,4Ј-diisothiocyanatostilbene-2,2Ј-disulfonate and 50 M dimethylamiloride in superfusates. B, buffering capacity (␤) and its pH i dependence measured in the presence and absence of 5% CO 2 /22 mM HCO 3 Ϫ buffer (n Ͼ 20/bin). Experiments in nil CO 2 provided estimates of ␤ int (circles) and were best-fitted with a straight-line: 8.6 ϫ pH i Ϫ 28.1 for CA9 expressor and 3.5 ϫ pH i ϩ 6.2 for empty vector (EV). Experiments in 5% CO 2 /22 mM HCO 3 Ϫ -buffered media provided estimates of ␤ int ϩ ␤ CO2 (triangles). Dashed lines: predicted contribution from CO 2 -dependent ␤ (ϭ 2.303 ϫ [HCO 3 Ϫ ] i ). C, kinetics of intracellular CO 2 hydration, and the reverse reaction, as measured by intracellular pH in single cells. Superfusate was rapidly switched from CO 2 -free (Hepes-buffered) to 5% CO 2 /22 mM HCO 3 Ϫ , and back. 200 nM 14v had no effect on the kinetics of intracellular pH changes. The pH i time course was monitored in the presence and absence of 100 M ATZ, to quantify the spontaneous and CA-catalyzed reaction. Further analysis of the time course of CO 2 addition and removal (n Ͼ 100) is shown in D and E, respectively (n Ͼ 100). F, H ϩ equivalent membrane transport, stimulated by acid-loading RT112 cells following a 4-min, 30 mM NH 4 Cl prepulse. pH i recovery was negligible when superfusing cell with zero extracellular [Na ϩ ] (green-shaded area). On re-addition of Na o ϩ , pH i recovery was considerably faster in the presence of 5% CO 2 /HCO 3 Ϫ buffer (black trace) than in its absence (gray trace). Inclusion of 100 M ATZ (red trace) did not affect recovery. G, pH i dependence of H ϩ equivalent extrusion (n Ͼ 20/bin) in EV cells. H, pH i dependence of H ϩ equivalent extrusion (n Ͼ 20/bin) in CA9 expressor cells.

and HCO 3
Ϫ ions (Fig. 2, C and D). Switching back, after 3 min, reversed this process, causing a rapid intracellular alkalosis (Fig.  2, C and E). To assess CA catalysis, experiments were repeated in the presence of ATZ or 14v. The initial time courses in Fig. 2 (D and E) were analyzed (see supplemental material) to derive the rate constant of intracellular hydration (k f ) or dehydration (k r ). With ATZ (total CA inhibition), k f was 0.18 Ϯ 0.01 s Ϫ1 (n Ͼ 50, hence k r was 0.24 M Ϫ1 s Ϫ1 ), in agreement with previous published values for the spontaneous reaction (41). In the absence of drug, k f and k r were 2.09-fold larger in CA9 expressor cells and 2.96-fold larger in EV cells. In contrast, k f and k r were not significantly affected by 14v (catalysis of 2.05-fold and 2.93-fold in CA9 expressor and EV, respectively), indicating that, in superfused single cells, extracellular CA does not contribute catalytically to the reversible hydration of intracellular CO 2 . These data also confirm that 14v is membrane-impermeant (37,42) as, otherwise, slower pH i changes would have been measured in its presence, because of inhibition of intracellular CA isoforms. The faster rate constants measured in the absence of ATZ indicate that intrinsic intracellular CA activity was functional in both CA9 expressor and EV cells, although it was 33% lower in the former. Thus expressing extracellular CA9 in RT112 cells was associated with a coordinate reduction of intracellular CA activity, in agreement with the observed down regulation of CA2 protein expression (Fig. 1A).
CA9 Effects on H ϩ Equivalent Membrane Transport-We investigated H ϩ equivalent extrusion by prepulsing isolated cells for 4 min with extracellular NH 4 Cl (1). This maneuver induces an intracellular acid-load that stimulates plasmalemmal H ϩ equivalent extrusion and recovery of pH i to resting levels. Recovery of pH i when using Hepesbuffered superfusates was accelerated by changing the buffer to CO 2 / HCO 3 Ϫ (Fig. 2F), indicating that an HCO 3 Ϫ -dependent transporter contributes to pH i regulation. Recovery was reversibly blocked by substituting extracellular Na ϩ with N-methyl-D-glucamine (Fig. 2F), indicating that all H ϩ equivalent extrusion is Na ϩ -dependent and is probably a combination of Na ϩ /H ϩ exchange (seen in Hepes superfusate) plus a Na ϩ -HCO 3 Ϫ -dependent transporter (mediating HCO 3 Ϫ influx, active in CO 2 /HCO 3 Ϫ -superfusate). H ϩ extrusion (J memb H ) was quantified as a function of pH i (Fig. 2, G and H), by multiplying pH i recovery rate and buffering capacity for a range of pH i values. Comparable estimates for J memb H in Hepes-or CO 2 /HCO 3 Ϫ -buffered conditions were obtained in EV and CA9 expressor cells. Furthermore, inhibiting CA activity with 100 M ATZ had no effect on J memb H or on steady state pH i (ϳ7.25). Further experiments (not illustrated) have shown that 200 nM 14v, like ATZ, has no effect on H ϩ equivalent extrusion (n ϭ 15) nor on resting pH i (n ϭ 15). Thus, an operational transport metabolon involving the HCO 3 Ϫ transporter and CA9 was not apparent in isolated cells, despite the fact that the CA9 enzyme was functional. Ϫ -buffered superfusate, showing significant intracellular core-acidosis. B, similar pH i map generated from a CA9 expressor spheroid, showing a greatly reduced core acidosis. C, incubation of the CA9 expressor spheroid with 500 nM 14v results in significant core acidosis, similar to that observed in EV spheroids. D, radial pH i profiles across EV (gray) and CA9 expressor (red) spheroids. E, radial pH i profiles in the presence of 500 nM 14v. F, radial pH i profiles in the presence of 100 M ATZ. Large standing pH i gradients were recorded in EV spheroids, in the presence or absence of CA inhibitors. CA9 activity in expressor spheroids reduces standing pH i gradients 2.5-fold. However, inhibition of CA9 with 14v or inhibition of global CA activity with ATZ restores pH i gradients. Asterisks denote significant difference between EV and CA9 expressor ROI pH i (t test at 5% level). G, core to periphery pH i gradients analyzed further for CA9 expressor spheroids. 14v and ATZ increased standing pH i gradients. H, in EV spheroids, neither 14v nor ATZ had a significant effect on standing pH i gradients. CA9 Effect on Spatial pH i Uniformity-Although CA9 expression had no influence on pH i regulation in isolated cells, it had a major effect on pH i in multicellular spheroids. On average, spheroids had a principal (major axis) radius of 222 Ϯ 113 (S.D.) m (n ϭ 22) and a 3-fold shorter minor axis radius (giving an "equivalent" spherical radius of 115 m for diffusion). Fig. 3A shows a confocal scan of pH i through the middle of a spheroid of EV cells, averaged in six concentric ROIs, showing significant pH i non-uniformity when superfused with the CO 2 / HCO 3 Ϫ -buffered solution. Core regions were notably more acidic than peripheral regions. In spheroids made of CA9 expressor cells (Fig. 3B), the pH i non-uniformity was greatly reduced. Exposure of the same spheroid to 500 nM 14v for 10 min (Fig. 3C) then induced a pH i non-uniformity, similar to that seen in EV spheroids.
In Fig. 3 (D-F) we have plotted mean steady-state pH i , from the core (ROI6) to the periphery (ROI1) of spheroids. In the presence of CO 2 /HCO 3 Ϫ -buffered superfusate, steady-state pH i in EV spheroids was significantly non-uniform, the core being 0.246 Ϯ 0.01 (S.E.) units more acidic than the periphery (Fig. 3D, n ϭ 10). This spatial pH i gradient was graded with distance from the core. The gradient was not an artifact of the measurement method, because 10 M nigericin, a membrane H ϩ /K ϩionophore, collapsed pH i non-uniformity (n ϭ 8, not shown).
Large radial pH i gradients in EV spheroids were in contrast to a more uniform pH i observed in CA9-expressing spheroids. The core-periphery pH i difference was reduced to 0.098 Ϯ 0.014 (S.E.) units (Fig.  3D, n ϭ 12). This more uniform pH i , however, was disrupted by addition of ATZ (100 M), to inhibit CA globally. The spatial pH i map now resembled that seen in EV spheroids, although the radial pH i gradient was not quite as large (average core-periphery pH i difference of 0.20 Ϯ 0.014 (S.E.) units, n ϭ 12) (Fig. 3F). This may be because the mean diameter of EV spheroids (ϳ460 m) studied was larger than that of CA9 expressor spheroids (ϳ390 m), predisposing the former to larger pH i gradients. In addition, full access of ATZ to the core of spheroids may be limited by tortuosity and permeability barriers. Inhibiting only extracellular CA activity with 14v (500 nM) also greatly increased pH i non-uniformity in CA9 expressor spheroids ( Fig.  3E; note that higher doses of 14v were used to overcome potential problems of drug access). Thus both 14v and ATZ more than doubled the radial pH i gradient in CA9 expressor spheroids (Fig. 3G). In contrast, these drugs exerted no significant effect in EV spheroids (Fig. 3H). CA9 activity thus enhances the spatial control of pH i in the multicellular structure.
CA9 Effect on Temporal pH i Uniformity-We investigated the ability of CA9 to coordinate pH i dynamically by monitoring pH i in a spheroid when superfusate CO 2 partial pressure (p CO2 ) was raised from 0% to 5%. This lowers pH i as CO 2 penetrates and permeates the structure. Fig. 4A shows that, in EV spheroids, pH i fell initially in the periphery and then, with a delay of ϳ9 s, in the core (by offsetting starting pH i , only the dynamic changes are illustrated). The experiment was then run in reverse, by reducing extracellular p CO2 from 5% back to 0% (Fig.  4B). The periphery alkalinized first, as CO 2 vented out of the cells, while the core lagged behind by ϳ9 s. Thus, in the absence of CA9 expression, a temporal heterogeneity of pH i was evident FIGURE 4. CA9 coordinates pH i temporally in spheroids. Raising superfusate p CO2 from 0% to 5% (and [HCO 3 Ϫ ] from 0 to 22 mM, at a constant superfusate pH of 7.4) drives CO 2 flux into cells and reduces pH i . Reducing superfusate p CO2 from 5% to 0% (and [HCO 3 Ϫ ] from 22 to 0 mM) drives the opposite reaction. Time courses of pH i change (ϮS.E.), offset to starting pH i , in the peripheral (black traces) and core (gray traces) ROIs of EV (A and B), CA9 expressor (C and D), and 14v-treated CA9 expressor (E and F) spheroids. Time courses were deemed significantly different (denoted by an asterisk), if over a continuous period of 20 s, time-matched data points (periphery versus core) were statistically different based on a t test at 5% significance level. The time delay between core and peripheral pH i changes was abolished in the presence of CA9 activity and reintroduced by blocking CA9 pharmacologically. during dynamic events. In contrast, during an extracellular CO 2 challenge, pH i changes in the periphery and core of CA9 expressor spheroids were almost coincidental (Fig. 4, C and D). The time delay in the core now typically lagged the periphery by Ͻ2 s. The spheroid behaved dynamically more like a pH i syncytium. Inhibiting extracellular CA with 14v removed this behavior. A time delay in the core (of ϳ6 s) was reintroduced in response to the CO 2 challenge (Fig. 4, E and F). Note that the pH i changes in 14v-treated CA9 expressor spheroids were a little slower, overall, than in EV spheroids, probably because of the simultaneous down-regulation of intracellular CA2 in the former (Fig. 1). We conclude that CA9 activity helps to coordinate the temporal and spatial behavior of pH i in the multicellular structure.
In Fig. 5, we have quantified the intracellular acid-loading rate (J CO2 H ) caused by addition/removal of superfusate CO 2 . In the absence of CA activity (inhibition with ATZ), J CO2 H was slow (7-10 mM/min), and comparable in isolated cells and spheroids. Under these conditions, J CO2 H was most likely limited by the slow, uncatalyzed rate of reversible intracellular CO 2 hydration. In the presence of intracellular, but not extracellular CA activity (absence of CA9 expression or its inhibition with 14v), J CO2 H was enhanced 3-fold in isolated cells, reflecting the faster reversible hydration of intracellular CO 2 and ready access of the molecule to and from the bulk extracellular phase. J CO2 H was also enhanced in spheroids with intracellular CA but no extracellular CA9 activity, although this was less so (up to 2-fold enhancement), most likely reflecting the more complex geometry of the spheroid, restricting CO 2 movement. Enhancement of J CO2 H was, however, spatially non-uniform, being slowest in the core. CA activity confined to the intracellular domain thus accelerated J CO2 H in spheroids, but not uniformly. In the presence of both intra-and extracellular CA activity (in CA9 spheroids, without CA inhibitors), fast J CO2 H was maintained in spheroids, but it was now more spatially uniform.
Modeling Effects of CA9 on pH i -Using data gathered in the present work, we have developed a diffusion-reaction model to explore the effect of intracellular and extracellular CA activity on the spatiotemporal coordination of pH i . The model is based on a two-compartment representation of the spheroid (Fig.  6A), consisting of the intracellular space (I) and the unstirred extracellular space (E), and a well stirred superfusate (S) that is continuous with E (Fig. 6B). H ϩ , CO 2 , and HCO 3 Ϫ can therefore diffuse within E, whereas CO 2 is also permitted to permeate between E and I. Solute concentrations are clamped in S, thereby mimicking experimental protocols. In the model, CA activity is represented in both E and I (see supplemental material for model derivation and best fitting). The model was run to simulate spatiotemporal changes of pH i during reduction of superfusate p CO2 from 5% to 0% (cf. Fig. 4). The computational model successfully predicts the ability of extracellular CA at a 30-fold catalytic rate to coordinate changes of pH i in core and peripheral regions of the spheroid (Fig. 6E). The contour map (Fig. 6F) shows pH i alkalinizing at similar rates at all radial sites. In contrast, a significant core delay occurs when EV spheroids lacking extracellular CA9 are modeled (Fig. 6, C and D), or when CA9 in the expressor model is inactivated (Fig. 6, G and  H). Catalysis by extracellular CA9 is therefore predicted to minimize pH i non-uniformity during dynamic changes of p CO2 , as observed experimentally.

Effect of CA9 on Intracellular pH Is Evident at Tissue Level-
In the present work we have examined the influence on pH i regulation of expressing CA9 in RT112 cells. This enzyme appears not to affect resting pH i or its regulation in isolated cells, whereas it exerts a major influence in larger, multicellular structures. This suggests that, when it is constitutively expressed, as in various developing tumors, the functional role of CA9 may be engaged at the tissue rather than the cellular level. After transfection with the CA9 gene, enzyme expression can be readily detected at the surface of isolated cells using molecular and functional activity assays (Fig. 1). Despite this, in superfused cells, CA9 does not enhance carrier-mediated H ϩ equivalent efflux (Fig. 2, G and H) and exerts no influence on intrinsic H ϩ -buffering capacity (Fig. 2B). CA9 activity also has no effect on the kinetics of carbonic buffering within the cell and appears not to be critical for CO 2 permeation (Fig. 2, D and  E). In multicellular spheroids, however, CA9 expression reduces non-homogeneity of pH i throughout the structure. It minimizes the occurrence of an acidic core and spatially synchronizes dynamic pH i changes during fluctuations of CO 2 . These are significant effects that would not readily be predicted from experiments on isolated cells. Our results suggest that CA9 plays an integrative role in spatial pH i regulation.
Mechanism of pH i Coordination by CA9-Encoding the catalytic properties of CA9 into a computational model of H ϩ  JULY 18, 2008 • VOLUME 283 • NUMBER 29 movement predicts a pattern of pH i response very similar to that observed experimentally in spheroids (Fig. 4, compare with Fig. 6). In the model, this is a consequence of extracellular CA catalysis, facilitating the venting of intracellular CO 2 . The mathematical scheme is important, because it confirms that such facilitation, and its spatial influence on pH i , is thermodynamically and kinetically feasible.

CA9 Coordinates Tissue pH i
As mentioned in the introduction, whereas CO 2 may be generated aerobically, it can also be generated through the buffering by intracellular HCO 3 Ϫ of anaerobically produced lactic acid. Because CO 2 is readily membrane-permeant, it will pass into the extracellular space. This helps to reduce the intracellular H ϩ load imposed by metabolism. In a spheroid without a vascular supply, the extracellular space is an unstirred compartment with a degree of spatial tortuosity, which would encourage local accumulation of CO 2 . This would limit further CO 2 efflux from the cells, thus restricting the rate at which CO 2 can be vented diffusively into the superfusate. CA9 activity in the extracellular space will facilitate CO 2 diffusion, by locally catalyzing conversion of a fraction of CO 2 into HCO 3 Ϫ and H ϩ , which then diffuse in parallel with unconverted CO 2 to the bulk phase. As confirmed quantitatively by the predictions of the computational model ( Fig. 6 and supplemental material), facilitated CO 2 diffusion will thus enhance metabolite removal from the tissue. Fig. 7 illustrates schematically that some of the extracellular HCO 3 Ϫ will be captured and recycled back into the intracellular compartment via membrane HCO 3 Ϫ transport, where it will buffer more intracellular H ϩ ions. Indeed, such HCO 3 Ϫ re-uptake will be necessary if facilitated CO 2 efflux is to be sustained over significant time periods.
In the absence of CA9, the facilitated route for CO 2 efflux will be limited by the slow spontaneous rate of extracellular CO 2 hydration. Due to large diffusion distances, core regions will retain a higher p CO2 and so become more acidic. This can account for spatiotemporal non-uniformity of pH i within a spheroid (Figs. 3 and 4). In the presence of CA9, the enhanced extraction of CO 2 will ensure that pH i becomes more alkaline. When expressed in vivo, the enzyme will help to match CO 2 efflux to the local rate of metabolic activity. CA9 should therefore coordinate the rate of intracellular CO 2 excretion from different regions, which will help to synchronize pH i spatially, as well as minimizing the formation of core intracellular acidosis. details. A model was used to simulate the effect of CO 2 efflux (induced by CO 2 removal from superfusate) on spatial pH i in EV spheroids (C and D), CA9 expressor spheroids (E and F), and CA9-inactivated CA9 expressor spheroids (G and H). Data are presented as time courses of pH i in core (gray) and peripheral (black) ROIs (C, E, and G) and as contour maps (D, F, and H) showing time dependence of pH i at different radial distances from core. The model outcome supports a role for CA9 in coordinating pH i spatially during CO 2 efflux from spheroid.
Facilitated CO 2 diffusion requires that local extracellular acidosis also be regulated to some extent, so that CO 2 hydration is not ultimately impeded by H ϩ accumulation (Fig. 7). Local accumulation of extracellular H ϩ can be controlled, but not prevented entirely, by buffering and diffusive dissipation. Our computational model can simulate experimental data if extracellular H ϩ mobility is reduced relative to free H ϩ mobility in water, and if extracellular H ϩ buffering capacity is at least 4.4 mM (in units of extracellular volume, or 2.2 mM in units of total spheroid volume). Reduced extracellular H ϩ mobility can be explained by the tortuosity of the extracellular space and buffering. The simulations presented in Fig. 6 were generated assuming 7 mM extracellular buffering (in units of extracellular volume) and H ϩ mobility of 8000 m 2 /s. An interesting observation in the present experiments is the down-regulation of intracellular CA activity (partly attributable to CA2) associated with expression of CA9 (Fig. 1). In preliminary work, 4 we have observed similar behavior when CA9 is expressed in HCT116 cells (derived from human colon carcinoma). These results are in agreement with immunohistochemical staining for CA2 and CA9 in human large intestine, showing a high CA2:CA9 ratio in normal tissue but a much lower ratio in colorectal tumors (43). Furthermore, recent work on zebrafish heart has shown that manipulating intracellular CA2 expression can inversely influence the expression of extracellular CA isoforms (44). Expression of intracellular CA and extracellular CA9 may therefore be coordinately linked, although it would be premature to suggest this as a general phenomenon in all tissues. Our computational modeling (supplemental material) suggests that the venting of CO 2 would be easier to achieve if intracellular CA expression were lowered (intracellular CA can trap aerobically generated CO 2 as intracellular H ϩ ). Coordinate regulation of intra-and extracellular CA may therefore be an additional element in the mechanism for attenuating pH i heterogeneity. Complete removal of intracellular CA expression, however, would be disadvantageous, because it would slow acid removal from anaerobically respiring cells (by slowing intracellular H ϩ conversion to membranepermeant CO 2 ). For fast and coordinated acid removal, both intra-and extracellular CA must therefore be expressed, but with a dominance for the latter. In the present work, the inversely linked expression of intracellular and extracellular CA activity may explain why inhibition of CA9 with 14v did not fully reverse the enzyme's coordinating effect on pH i (Fig. 4), as part of the coordination will have been secondary to the reduction of intracellular CA activity.
We conclude that facilitation of CO 2 diffusion by the catalytic activity of extracellular CA9 is likely to be the cause of better pH i coordination in RT112 spheroids. Facilitated CO 2 diffusion has been proposed previously in tissues such as skeletal muscle, where it assists the removal of CO 2 across a poorly vascularized extracellular space and the lung surface where it assists CO 2 transfer to and from the capillaries (18,45). The present work, however, is the first to examine the consequences of such facilitation for the spatial control of pH i within a tissue.
No Apparent CA9 Effect in Isolated Cells-The proposed action of CA9 on pH i in spheroids may help to explain its lack of effect on pH i changes induced by CO 2 addition/removal in isolated cells (Fig. 5). Much of the difficulty of transmembrane CO 2 movement disappears when a small cell is adequately superfused. During CO 2 efflux, for example, because bulk solution is refreshed continuously, there will be no local accumulation of extracellular CO 2 for CA9 to dissipate, and so the role of the enzyme will not be evident.
In the present work, CA9 expression also had no effect (Fig.  2C) on carrier-mediated H ϩ equivalent extrusion (including Na ϩ -driven HCO 3 Ϫ transport). In contrast, CA9 appears to facilitate Cl Ϫ /HCO 3 Ϫ exchange when expressed heterologously in HEK293 cells (13). Once again, the lack of effect in isolated RT112 cells, caused by the rapid superfusion used in our experiments, ensures that extracellular HCO 3 Ϫ is continuously presented at equilibrium to the cell surface. If the local solution is not driven significantly out of equilibrium by HCO 3 Ϫ transport, net CA9 catalysis will not be engaged, and so the functional role of the enzyme will not be unmasked. This implies that the influence of CA activity within a HCO 3 Ϫ transport metabolon may become evident only when a significant unstirred layer exists close to the membrane. Reduced or zero perfusion of the extracellular compartment may therefore be a requirement for the participation of extracellular CA. This requirement may be satisfied in the present work on spheroids, where a significant extracellular unstirred layer exists between individual cells. Because of membrane HCO 3 Ϫ re-uptake and spatial diffusion of extracellular CO 2 , HCO 3 Ϫ and H ϩ ions (Fig. 7), the local extracellular CO 2 /HCO 3 Ϫ buffer may be driven out of equilibrium, triggering the participation of extracellular CA activity. Spatial arrangements within the spheroid are therefore more likely to favor a functional transport metabolon, although this remains to be tested experimentally. Ϫ titrates cell-generated H ϩ to produce membrane-permeant CO 2 , the form in which much acid is excreted by tumor cells. The kinetics of this reaction depends on intracellular CA activity. To prevent saturation of CO 2 removal, extracellular CO 2 must not accumulate. Extracellular CO 2 can diffuse away radially (solid black arrow) or convert to HCO 3 Ϫ and H ϩ , and diffuse in ionic form (facilitated CO 2 diffusion). Conversion is catalyzed by CA9. Some HCO 3 Ϫ is re-sequestered into cells via HCO 3 Ϫ transport (dashed gray arrow) to supply the base for further titration of intracellular H ϩ .

Spatial pH i Coordination Will Depend on CA9 Expression
Level and Tissue Size-Diffusion-reaction modeling (see supplemental material) can be used to explore the possible effect of CA9 expression levels and tissue size on the ability of the enzyme to coordinate pH i spatially. Coordination can be quantified in terms of the core-to-periphery pH i time delay during CO 2 washout (measured at a threshold pH i change of, say, 0.1 unit). Our computational model suggests that, for a typical RT112 spheroid, this time delay is likely to decrease hyperbolically as CA9 expression increases, halving in response to a 14-fold increase in extracellular CO 2 hydration rate (see supplemental Fig. S2). Extracellular CA activity in our stably transfected RT112 spheroids accelerated CO 2 hydration rate by 30-fold (see supplemental material for derivation of the measurement). This catalytic rate is predicted to facilitate CO 2 efflux near-maximally, reducing the time delay by 4-fold, thereby providing optimal conditions for the spatial control of pH i .
A key question is whether spatial pH i regulation also occurs in larger spheroids than used in the present work. We have currently examined pH i in spheroids of Յ500 m in principal diameter, to avoid the hypoxic core becoming necrotic (46), thereby failing to take up the pH dye. The spheroids comprise ϳ25,000 cells, equivalent, for example, to a typical non-vascularized tumor in early development. Our computational model suggests that a 14-fold catalytic rate of CA9 expression should still be sufficient to halve the core to periphery time delay when spheroid diameter is doubled or even tripled. Indeed, following such a size increase, a 30-fold CA9 catalytic rate as measured in RT112 spheroids, should still reduce the time delay by 3-to 4-fold. This is not to say that larger spheroids would be expected to exhibit a perfectly uniform pH i , they would not. The uncatalyzed time delay in larger spheroids would naturally increase, in line with the larger core to periphery diffusion distance. The modeling, however, predicts that optimal CA9 expression should reduce this delay by up to 75%. Spatial pH i regulation by CA9 should therefore remain prominent in larger spheroids.
In summary, the extent to which CA9 can coordinate pH i will, most likely, depend on the level of enzyme expression relative to tissue size. CA9 expression may not always succeed at unifying pH i spatially in a tissue, but it will nevertheless attenuate local pH i gradients. The predicted effect of tissue size tends to be borne out by our recent preliminary results 4 using large, spherical growths (diameter ϳ 600 m) cultured from HCT116 cells that do not normally express CA9. The larger spheroids show longer core to periphery time delays for pH i on CO 2 washout (Ͼ20 s). Both the gradients and the time delays are, however, still reduced in spheroids expressing CA9, suggesting that the enzyme's ability to coordinate pH i spatially is not restricted to RT112 spheroids.
Relevance to Normal Tissue and Growing Tumors-The potential of CA9 to coordinate pH i and reduce core acidosis in a tissue will be particularly important where vascular perfusion is low or absent. It would therefore be expected to influence pH i spatially during pathophysiological events such as ischemia. Under such conditions, an acidic area within a tissue is likely to coincide with regional hypoxia, a key trigger for CA9 expression (25). Although the present work has focused on extracel-lular CA9, other CA isoforms such as CA4 and CA12 are also expressed externally in cells such as cardiac myocytes, endothelial/epithelial cells, and neurons (10) and may contribute to spatial pH i control.
CA9 expression in many tumors has raised questions regarding its importance to cancer growth. Its potential for spatially controlling pH i furnishes an obvious survival advantage to the tumor, particularly as CA9 has one of the highest catalytic rates of all isoforms (47). Previous work has documented the tendency of tumors (33,35) and spheroids (48) to develop an acidic extracellular core, but there has been no information on spatial pH i regulation. Our data now suggest that CA9 expression will tend to unify pH i across three-dimensional tissue structures. A more uniformly alkaline tumor may be predisposed to further growth (3). The effect of CA9 on pH i may be particularly important before significant vascular perfusion of a tumor has been established. It is notable that the CA9-transfected RT112 cells used in the present work exhibit protein levels for the enzyme and extracellular CO 2 hydration rates comparable to those induced hypoxically in breast cancer cell lines (31). Furthermore, Northern blotting of various hypoxically exposed cancer cell lines, as well as a variety of tumor samples, has revealed that CA9 RNA levels (calibrated to ␤-actin RNA) are high (19). Moreover, neoplasias arising from a defective von Hippel-Lindau gene have been shown to "overexpress" CA9 (19,22). CA9 levels in our transfected RT112 cells are thus likely to be representative of those in CA9 expressing tumors.
Not all tumors express CA9, but this need not invalidate the proposal that facilitating CO 2 excretion helps to reduce intracellular acidosis. In the absence of CA9, other extracellular CA isoforms may suffice or other, as yet unknown, strategies may be implemented for spatially controlling pH i . Nevertheless, our results may not translate in simple fashion to tumors composed of many cell types, possibly with heterogeneous local expression of CA, H ϩ equivalent transporters, and H ϩ buffers, all of which will influence pH i . For example, endogenous CA9 in tumors is typically up-regulated in the hypoxic core rather than distributed more uniformly, as in the present CA9 expressor spheroids. Our three-dimensional diffusion model, however, suggests that such localization of CA9 would still be very effective at venting core CO 2 and dissipating core acidosis (see supplemental Fig. S2). This is expected, because less CA9 expression is required for removing CO 2 at the periphery of three-dimensional tissue due to the shorter diffusion distance. It is noteworthy that recent immunohistochemical staining in tumor xenografts (21) has shown that areas of CA9 expression exceed pimonidazole staining (hypoxic) areas by a factor of two. Furthermore, many tumors have "diffuse" CA9 staining (defined as Ͼ40% cell staining over 400 m) (19). For example, 68% of cervical carcinomas tested showed strong and diffuse CA9 staining (19). This pattern is similar to that detected with antibodies in CA9-expressing RT112 spheroids (Fig. 1B). The present model and results therefore point to a powerful potential role for extracellular CA9 in the integrative control of tumor pH i .
One consequence of CO 2 hydration by CA9 is the generation of extracellular H ϩ ions, and hence a local fall of pH e which drives H ϩ diffusion into the bulk extracellular phase (33,35). In preliminary unpublished work, 4 we have imaged CA9-expressing HCT116 spheroids using an extracellular, membrane-impermeant pH dye. The data show that these spheroids develop low pH e that can be minimized by inhibiting CA9 activity pharmacologically. Although extracellular buffering will help to dampen the emerging acidity, the low pH e may be beneficial to the tumor by promoting retraction of host tissue, permitting further tumor invasion. CA9 may benefit the tumor, not only by raising pH i , but also by reducing pH e . Selective suppression of CA9 activity by drugs or antibodies may therefore provide a means for controlling tumor development, by disrupting pH i coordination and pH e acidification. Although general membrane-permeant CA inhibitors have been shown to attenuate tumor growth, application of membrane impermeant CA inhibitors, such as 14v, to target extracellular CA9 selectively may offer a novel therapeutic approach.