Intracellular Carbonic Anhydrase Activity Sensitizes Cancer Cell pH Signaling to Dynamic Changes in CO2 Partial Pressure*

Background: Intracellular carbonic anhydrase (CAi) activity, commonly detected in cancer, accelerates CO2/HCO3− equilibration. Results: In cells with high CAi activity, fluctuations in pCO2 (which can arise from intermittent blood flow) evoke substantial intracellular pH oscillations that modulate downstream pathways (e.g., calcium, mTOR). Conclusions: CAi transduces the state of perfusion to intracellular signaling. Significance: Coupling between environment and cell behavior may influence cancer progression. Carbonic anhydrase (CA) enzymes catalyze the chemical equilibration among CO2, HCO3− and H+. Intracellular CA (CAi) isoforms are present in certain types of cancer, and growing evidence suggests that low levels correlate with disease severity. However, their physiological role remains unclear. Cancer cell CAi activity, measured as cytoplasmic CO2 hydration rate (kf), ranged from high in colorectal HCT116 (∼2 s−1), bladder RT112 and colorectal HT29, moderate in fibrosarcoma HT1080 to negligible (i.e. spontaneous kf = 0.18 s−1) in cervical HeLa and breast MDA-MB-468 cells. CAi activity in cells correlated with CAII immunoreactivity and enzymatic activity in membrane-free lysates, suggesting that soluble CAII is an important intracellular isoform. CAi catalysis was not obligatory for supporting acid extrusion by H+ efflux or HCO3− influx, nor for maintaining intracellular pH (pHi) uniformity. However, in the absence of CAi activity, acid loading from a highly alkaline pHi was rate-limited by HCO3− supply from spontaneous CO2 hydration. In solid tumors, time-dependence of blood flow can result in fluctuations of CO2 partial pressure (pCO2) that disturb cytoplasmic CO2-HCO3−-H+ equilibrium. In cancer cells with high CAi activity, extracellular pCO2 fluctuations evoked faster and larger pHi oscillations. Functionally, these resulted in larger pH-dependent intracellular [Ca2+] oscillations and stronger inhibition of the mTORC1 pathway reported by S6 kinase phosphorylation. In contrast, the pHi of cells with low CAi activity was less responsive to pCO2 fluctuations. Such low pass filtering would “buffer” cancer cell pHi from non-steady-state extracellular pCO2. Thus, CAi activity determines the coupling between pCO2 (a function of tumor perfusion) and pHi (a potent modulator of cancer cell physiology).

forms are present in certain types of cancer, and growing evidence suggests that low levels correlate with disease severity. However, their physiological role remains unclear. Cancer cell CA i activity, measured as cytoplasmic CO 2 hydration rate (k f ), ranged from high in colorectal HCT116 (ϳ2 s ؊1 ), bladder RT112 and colorectal HT29, moderate in fibrosarcoma HT1080 to negligible (i.e. spontaneous k f ‫؍‬ 0.18 s ؊1 ) in cervical HeLa and breast MDA-MB-468 cells. CA i activity in cells correlated with CAII immunoreactivity and enzymatic activity in membrane-free lysates, suggesting that soluble CAII is an important intracellular isoform. CA i catalysis was not obligatory for supporting acid extrusion by H ؉ efflux or HCO 3 ؊ influx, nor for maintaining intracellular pH (pH i ) uniformity. However, in the absence of CA i activity, acid loading from a highly alkaline pH i was rate-limited by HCO 3 ؊ supply from spontaneous CO 2 hydration. In solid tumors, time-dependence of blood flow can result in fluctuations of CO 2 partial pressure (pCO 2 ) that disturb cytoplasmic CO 2 -HCO 3 ؊ -H ؉ equilibrium. In cancer cells with high CA i activity, extracellular pCO 2 fluctuations evoked faster and larger pH i oscillations. Functionally, these resulted in larger pH-dependent intracellular [Ca 2؉ ] oscillations and stronger inhibition of the mTORC1 pathway reported by S6 kinase phosphorylation. In contrast, the pH i of cells with low CA i activity was less responsive to pCO 2 fluctuations. Such low pass filtering would "buffer" cancer cell pH i from non-steady-state extracellular pCO 2 . Thus, CA i activity determines the coupling between pCO 2 (a function of tumor perfusion) and pH i (a potent modulator of cancer cell physiology).
Due to rapid permeation of CO 2 gas across cell membranes (4), carbonic buffer is present in both intra-and extracellular tissue compartments. On the time scale of cellular physiology, the spontaneous equilibration of carbonic buffer is relatively slow, as exemplified by a CO 2 hydration time constant of several seconds (rate constant k f ϭ 0.18 s Ϫ1 ) (5). Consequently, biological processes that involve a change in the concentration of CO 2 , HCO 3 Ϫ or H ϩ can become rate-limited by carbonic buffer reequilibration. This limitation has presumably driven the evolution of at least a dozen mammalian carbonic anhydrase (CA) 2 isozymes that accelerate CO 2 /HCO 3 Ϫ equilibration (6, 7). The CAs are grouped as intra-(CA i ) or extracellular (CA e ) depending on the orientation of the catalytic site (5)(6)(7)(8). Activity assays and immunotechniques have identified CA i and CA e isoforms in cancer cells (9 -17). Physiologically, CA e isoforms, such as CAIX and CAXII, facilitate CO 2 and H ϩ diffusion across the continuous and tortuous interstitial space (18,19). Thus, CA e activity can improve the venting of acidic products of metabolism over the long diffusion distances found in inadequately perfused solid tumors, allowing their faster growth (20,21). The role of CA i isoforms in cancer physiology is still debated. Down-regulation of gap junctions in cancer cells prevents the intracellular compartment from becoming syncytial (22), and this restricts the spatial range over which CA i activity could facilitate CO 2 or H ϩ diffusion. Previously, it has been suggested that CA i activity facilitates the transport of HCO 3 Ϫ or H ϩ ions across membranes by reducing the extent to which cytoplasmic reactions slow the delivery or removal of the transported ion (23)(24)(25)(26). To benefit from CA activity, these transporter-evoked H ϩ or HCO 3 Ϫ fluxes would have to exceed the spontaneous chemical re-equilibration kinetics of carbonic buffer (27).
Another source of disturbance to carbonic buffer equilibrium is fluctuating CO 2 partial pressure (pCO 2 ). Cancer cells produce large quantities of CO 2 from titration of acids (e.g. lactic) with HCO 3 Ϫ and decarboxylation by mitochondria and the pentose phosphate shunt (28). Ultimately, the excess CO 2 must be removed with the blood flow. In tumors, vasomotion and hemodynamic factors can produce cycles of intermittent blood flow, commonly observed with periodicities of several minutes (29 -33). Unstable perfusion is the basis for acute hypoxia (32, 34 -40), characterized by oxygenation-reoxygenation cycles as fast as 2/min (30) and amplitudes of tens of mmHg O 2 (30,35,41,42). Episodes of inadequate blood flow produce "closed" pockets of blood, which become oxygen-depleted and accumulate CO 2 (43)(44)(45). Periodic restoration of flow returns pCO 2 and pO 2 to normal. The resulting fluctuations in pCO 2 (mirroring pO 2 ) are transmitted across the intraand extracellular compartments because CO 2 gas (unlike H ϩ or HCO 3 Ϫ ions) crosses membranes rapidly (4). Due to reversible intracellular CO 2 hydration, pH i will respond to pCO 2 fluctuations, but the coupling between these depends critically on CA i activity. Because cell behavior is strongly modulated by H ϩ ions (46 -48), CA i activity may act as an important transducer between blood flow and cell signaling.
The aim of the present study was to measure CA i activity in the native environment of intact cancer cells and to identify the physiological processes, relevant to cancer, that depend on CA i activity. We describe a novel role for CA i activity in sensitizing cancer cell pH to pCO 2 changes, thereby linking metabolism and perfusion with pH i -responsive signaling cascades.
CAII Knockdown-CA2 gene expression in HCT116 cells was silenced using one of four shRNA constructs cloned into psi-LVRU6-GFP lentiviral vector targeting the 351st, 493rd, 597th, and 695th position of CA2 mRNA (NCBI Reference Sequence NM_000067). All constructs, including scrambled-eGFP, were purchased from Genecopoiea. Stable HCT116 cell clones with silenced CA2 gene expression were selected in the presence of 2 g/ml puromycin for 2 weeks, and selected clones were pooled together for experiments.
Measuring Carbonic Anhydrase Activity in Lysates-Cells were lysed by repeated freeze-thaw cycles in buffer containing 140 mM potassium gluconate, 0.5 mM EGTA, 1 mM MgCl 2 , 15 mM Hepes, 15 mM Mes at pH 7.8 (4°C), and protease inhibitor. Membranes were removed by centrifugation (20 min at 15,000 rpm at 4°C), and the supernatant was diluted to a total protein concentration between 1 and 10 mg/ml (Bradford assay). The CA-catalyzed reaction was triggered by adding 0.33 ml CO 2saturated water to 0.67 ml of lysate in a stirred chamber at 4°C. The time course of pH (Hamilton Biotrode) was fitted with a kinetic model (19) to obtain the CO 2 hydration rate constant k f .
Fluorescence Measurements of Intracellular pH and Calcium-Cells were imaged confocally using a Zeiss LSM 700 confocal system on an Axiovert inverted microscope. To measure pH i , cells were AM-loaded for 10 min with 10 M cSNARF1 (excitation, 555 nm; emission, 580 and 640 nm). The fluorescence ratio was converted to pH i using a calibration curve obtained by the nigericin method (50). Calibration was performed twice a year for each cell line, and the stability of the curve was tested by the null point method (51). Membrane H ϩ (or H ϩ -equivalent) flux was calculated as the rate of pH i change ϫ buffering capacity (3). Intrinsic buffering capacity was measured in separate experiments using the graded ammonium removal technique (3). CO 2 /HCO 3 Ϫ -dependent buffering was estimated using the equation for an open buffer system: 2.303 ϫ [HCO 3 Ϫ ] i . Intracellular [HCO 3 Ϫ ] was calculated by best fitting a mathematical model of pH i regulation to the time course of pH i recovery from an acid or base load (see below). This approach derives instantaneous carbonic buffering capacity (rather than assuming that the buffer is distributed at equilibrium across the cell membrane at all times). To measure [Ca 2ϩ ] i , cells were AM-loaded for 20 min with 50 M Fluo3 (excitation, 488 nm; emission, Ͼ510 nm). For co-cultures, regions with distinct areas of cancer cells and fibroblasts (dis-tinguished by cell morphology and size) were selected for imaging.
Mathematical Modeling-A mathematical model of pH i control was parameterized using data for buffering capacity (␤) and membrane transport fluxes of H ϩ and HCO 3 Ϫ ions (J H and J HCO3 ) (3). The carbonic buffer equilibrium constant, K CO2 , was 10 Ϫ6.1 M, and the spontaneous CO 2 hydration rate constant (k f ) was 0.18 s Ϫ1 (the reverse rate constant, k r , was k f /K CO2 ). CA i activity was represented as a dimensionless scalar (ca). Surface area/volume ratio was 1/h where h is the monolayer height (10 m). CO 2 permeability, p, was 10 4 m/s (4). pCO 2 was a sinusoidal function, f, of amplitude A, frequency , and baseline of 5% CO 2 . The equations are as follows.

RESULTS
CA i Activity in Cancer and Fibroblast Cell Lines-Total CA i activity was measured in intact cells superfused continuously with physiological salt solution at body temperature. The CA icatalyzed reaction was triggered by switching between CO 2 / HCO 3 Ϫ -free and 5% CO 2 /22 mM HCO 3 Ϫ -buffered superfusates. The time constant of solution exchange (2.7 s; determined by fluorescently labeling one solution with 30 M fluorescein) was adequately fast to ensure that pCO 2 was manipulated rapidly. CA i -catalyzed reaction kinetics were determined from the initial rate of pH i change (reported using the fluorescent pH indicator cSNARF1 AM-loaded into cells) and intrinsic buffering capacity (52). Fig. 1A, panel i, shows an example pH i time course recorded from colorectal HCT116 cells paired with an experiment performed in the presence of the broad spectrum CA inhibitor acetazolamide (ATZ; 100 M) to determine spontaneous kinetics over a matching pH i range. CA i activity was expressed relative to spontaneous reaction kinetics (Fig. 1A, panel ii). Experiments were also performed on bladder RT112, colorectal HT29, fibrosarcoma HT1080, cervical HeLa, and breast MDA-MB-468 cells as well as dermal NHDF-Ad, colonic CCD-112-CoN fibroblasts, and intestinal InMyoFib myofibroblasts. CA i activity ranged from ϳ9-fold above the spontaneous rate in HCT116 cells to very low in HeLa, MDA-MB-468, and fibroblast/fibroblast-related cells. Thus, the CO 2 hydration capacity of cancer cell cytoplasm spanned over an order of magnitude in range and was cell line-dependent.
Soluble cytoplasmic CA (CA s ) isoforms (e.g. CAI, -II, -III, -VII, and -XIII) are plausible contributors to CA i activity. This was tested in cell lysates after removing membranes by centrifugation. Lysates were injected with CO 2 -saturated water, and the CO 2 hydration constant was estimated from the time course of medium pH at 4°C (Fig. 1B, panel i). CA s activity was quantified in terms of the CO 2 hydration rate constant relative to the measurement in the presence of ATZ (100 M). Kinetic data were normalized to total protein concentration (19). Substantial CA s catalysis was detected in lysates of cancer cells with high total CA i activity determined in intact cells. This argues that soluble isoforms underlie at least part of intracellular CO 2 hydration catalysis (Fig. 1B, panel ii). The contribution of CAII to total CA i activity was tested in HCT116 cells by knockdown using shRNA constructs 351, 495, 597 and 695 and compared with scrambled shRNA and wild-type cells. Construct 695 resulted in the most substantial decrease in CAII immunoreactivity ( Fig. 1C, panel i), and it reduced total CA i activity by ϳ90% relative to wild-type ( Fig. 1C, panel ii). To test whether the decrease in total CA i activity is attributable to disruption of CAII alone, the specificity of shRNA constructs 695 and 597 was investigated. Constructs 695 and 597 have very low sequence homology (16 -42%) with membrane-associated CAs (IV, IX, XII, and XIV) and low homology (53-68%) with soluble isoforms other than CAII (I, III, VII, and XIII). Of the soluble isoforms, HCT116 expressed only low levels of CAVII and -XIII, and immunoreactivity for CAI and -III was absent ( Fig.  1C, panel iii). Constructs 695 and 597 did not alter this expression pattern, indicating that the knockdown experiment reduced CA i activity by targeting CAII selectively. The findings indicate that CAII is a principal contributor to CA i activity in HCT116 cells.
Across the tested cancer, fibroblast, and fibroblast-related cells, CAII immunoreactivity ( Fig. 1D) correlated with in situ CA i activity (Pearson's correlation coefficient r 2 ϭ 0.83). CAII expression was notably absent in fibroblast and fibroblast-related cells, including cancer-derived Hs675.T cells. In cancer cells, CAII expression correlated strongly (r 2 ϭ 0.97) with CA s activity measured in membrane-free lysates. As expected from active site topology, the expression of membrane-tethered CAIX did not correlate with CA i activity. Hypoxia plays an important role in cancer biology by regulating gene expression through mechanisms that include hypoxia-inducible factor. To investigate the effect of long term hypoxia on CAII expression, Western blot analysis was performed on cells incubated for 48 h with the hypoxia-inducible factor-stabilizing drug dimethyl oxalylglycine (3,53). This treatment did not affect CAII expression but produced the expected induction of CAIX in RT112, HT1080, MDA-MB-468, and HeLa (54) and to a lesser degree in NHDF-Ad and CCD-112-CoN fibroblasts.
Role of CA i Activity in pH i Regulation-The supply of H ϩ or HCO 3 Ϫ ions for pH i -regulatory transport across membranes may be rate-limited by the CA i -catalyzed reaction. This was investigated in RT112, MDA-MB-468, and HCT116 cells.
RT112 cells have naturally high CA i activity and can produce large HCO 3 Ϫ fluxes at both low and high pH i (3). Acid extrusion was evoked at low pH i by means of an "ammonium prepulse" solution maneuver performed on cSNARF1-loaded cells ( Fig.  2A, panel i). An "acetate prepulse" was performed to raise pH i . Because abrupt base loading of cytoplasm upon acetate removal drives carbonic buffer out-of-equilibrium, cells were superfused briefly in Cl Ϫ -free medium to allow buffer re-equilibration (55). Subsequent re-exposure to Cl Ϫ -containing superfusates activated acid loading transporters ( Fig. 2A, panel ii).
H ϩ /H ϩ -equivalent flux was calculated as the product of the rate of pH i change and buffering capacity (the carbonic buffer component was estimated using a mathematical model to predict out-of-equilibrium behavior during dynamic pH i regulation). pH i -regulatory fluxes, shown in Fig. 2A, panel iii, were symmetrical around the resting pH i . To establish the contribution of Na ϩ /H ϩ exchange (NHE) to pH i regulation, measure-ments were repeated in the presence of 30 M 5-(N,N-dimethyl)amiloride (DMA). Acid loading was DMA-insensitive, whereas acid extrusion flux was reduced only modestly, confirming a minor role for NHE in pH i regulation in RT112 cells ( Fig. 2A, panel iii, inset). Replacing superfusate carbonic buffer with Hepes reduced flux substantially, indicating that pH i regulation relies principally on HCO 3 Ϫ transport. activity on acid/base membrane transport was inferred from the effect of 100 M ATZ (in the presence of carbonic buffer). Although ATZ targets both intra-and extracellular CAs, the latter isoforms would not be exerting a net catalytic effect because superfusion presents cells with pre-equilibrated buffer at all times (19). ATZ did not affect resting pH i or acid extrusion flux, but it limited acid-loading fluxes to no greater than ϳ14 mM/min. This flux is equal to the maximal rate of HCO 3 Ϫ deliv-ery by CO 2 hydration under spontaneous reaction kinetics (k f ϫ [CO 2 ] ϭ 14 mM/min). Thus, acid loading by means of HCO 3 Ϫ export can be rate-limited by CA i activity, but this requires large fluxes evoked by substantially raised pH i . Intracellular H ϩ ions diffuse considerably slower than expected from their low atomic weight because of extensive binding to larger buffer molecules, including immobile proteins (56,57). NHE activity could plausibly be rate-limited by a slow diffusive supply of H ϩ ions and therefore require CA i activity for delivering H ϩ ions from CO 2 hydration. As shown in noncancerous cells, CA i activity can facilitate H ϩ diffusion by improving the reaction kinetics of CO 2 /HCO 3 Ϫ , a mobile buffer (58,59). If the intrinsic H ϩ ion diffusivity in cytoplasm were sufficiently restricted, a large acid/base flux at the surface membrane would be expected to produce measurable pH i non-uniformity in the absence of carbonic buffer. This was tested in MDA-MB-468 cells, which produce very high NHE fluxes at low pH i (3). Cells were trypsinized to produce spherical cells with the smallest possible surface area-to-volume ratio, i.e. longest average surface-to-core diffusion distance. The mean radius of MDA-MB-468 cells was 7.11 Ϯ 0.12 m, which is typical of most cancer cells. NHE is functional in the presence and absence of carbonic buffer; therefore it is possible to investigate the effect of removing CO 2 /HCO 3 Ϫ on spatial H ϩ dynamics in cytoplasm. pH i uniformity during rapid acid extrusion was assessed by imaging pH i confocally near the surface membrane and at the core of the cell. The pH i time courses shown in Fig. 2B, panel i, indicate that even at maximal acid extrusion rates, pH i remained uniform in the absence of carbonic buffer. The addition of carbonic buffer did not affect the degree of pH i uniformity (Fig. 2B, panel ii). Thus, carbonic buffer and hence CA i activity are not required for ensuring adequate H ϩ diffusion in MDA-MB-468 cells. These findings are indicative of good diffusive coupling by intrinsic H ϩ buffers.
Measurements of pH i gradients in bulk cytoplasm may not identify diffusion barriers in the immediate environment of the transport protein. An example of such a barrier is slow H ϩ transfer from immobile buffers to the transporter proteins (60,61). Given that CO 2 is a highly penetrating source of H ϩ ions (26), a role for CA i activity in overcoming these localized H ϩ barriers is conceivable. To test for possible submembrane barriers, acid extrusion was measured in HCT116 cells, which have naturally high CA i activity and the capacity to produce large NHE fluxes (3). If CA i activity were important for delivering H ϩ ions to NHE protein, then removal of carbonic buffer would slow overall flux to a baseline set by the intrinsic capacity of the cytoplasm to supply the transporter with H ϩ ions. Loading cytoplasm with exogenous mobile buffers, such as imidazoles, would then be expected to rescue acid extrusion (60). Fig. 2C, panel i, shows the time course of pH i recovery from low pH i in the absence of CO 2 /HCO 3 Ϫ . This eliminates CO 2 -sourced delivery of H ϩ ions and would expose any submembrane diffusional barriers. Experiments were repeated on cells loaded with the imidazole-containing mobile buffer carnosine (6 mM for 48 h). Intrinsic buffering capacity, measured by the "graded ammonium removal" technique (62), was higher in carnosineincubated cells, consistent with the loading of cytoplasm with an exogenous buffer of pK 6.7 (similar to that of carnosine) and concentration of 12 mM (Fig. 2C, panel ii). After accounting for the additional buffering capacity, flux in carnosine-loaded cells was no different from control (Fig. 2C, panel iii). Further experiments were performed on control HCT116 cells superfused with carbonic buffer. Because HCO 3 Ϫ -dependent mechanisms (e.g. Na ϩ -HCO 3 Ϫ co-transport) are activated in the presence of carbonic buffer, NHE-mediated flux was calculated by subtracting the DMA-sensitive component from the total flux. Neither CO 2 /HCO 3 Ϫ nor carnosine increased NHE flux (Fig.  2C, panel iii), arguing that these additional mobile buffers are not required for delivering H ϩ ions to NHE. Thus, CA i catalysis is unlikely to accelerate transport by providing an additional route of H ϩ delivery. An earlier study had suggested that imidazole groups projecting from the CAII molecule deliver H ϩ ions to H ϩ transporters without involving the catalytic process (60). Because large acid extrusion fluxes could be produced by CAII-negative MDA-MB-468 cells, an acatalytic effect of CAII is unlikely to be an absolute requirement for fast NHE activity. This was explored further in CAII knockdown HCT116 cells (construct 695; Fig. 1C, panel i). Acid extrusion in carbonic buffer, which includes NHE and HCO 3 Ϫ -dependent components, was not different in CAII-negative cells compared with scrambled controls (Fig. 2C, panel iv). The HCO 3 Ϫ -dependent flux component, measured by inhibiting NHE with DMA, was also unaffected by CAII knockdown (Fig. 2C, panel iv).
Role of CA i Activity during Fluctuations in pCO 2 -To explore the effects of CA i activity during pCO 2 fluctuations associated with intermittent blood flow, superfusion of confluent monolayers (grown in slow exchange chambers) was alternated between 5% CO 2 /22 mM HCO 3 Ϫ (pH 7.4; representing normal arterial blood plasma) and 20% CO 2 /22 mM HCO 3 Ϫ (pH 6.8; representing respiratory acidosis) with a periodicity of 4 min. By labeling one solution with fluorescein (30 M) in separate experiments, the solution mixing interval was estimated to be ϳ15 s, which allows for a relatively slow transition between 5 and 20% CO 2 . The relationship between pCO 2 and [H ϩ ] i was explored further using a mathematical model parameterized using data for CA i activity and buffering capacity in HCT116 cells (3). Smooth transitions between normal and raised pCO 2 were simulated with a sinusoidal wave (Fig. 3G) over a range of periodicity and amplitude. Fig. 3H shows the predicted effect of 10-fold CA i activity on the size of [H ϩ ] i fluctuations. [H ϩ ] i fluctuations were amplified 30% for periodicities of 4 min and at least doubled for periodicities of Ͻ2 min. The relative effect of CA i activity on [H ϩ ] i dynamics was essentially independent of pCO 2 amplitude. pH i -regulatory H ϩ and HCO 3 Ϫ membrane transport is expected to curtail the extent of [H ϩ ] i changes, but mathe-matical modeling (not shown) predicts this to be a minor effect because the magnitude of membrane H ϩ /HCO 3 Ϫ transport is considerably smaller than that of CO 2 flux, particularly for short pCO 2 wave periodicities.
In summary, CA i activity sensitizes cancer cell pH to fluctuations in pCO 2 . In contrast, cytoplasm with low CA i activity behaves as a low pass filter that dampens [H ϩ ] i changes. Essentially all biological processes are pH-sensitive; therefore the effect that CA i activity has on [H ϩ ] i dynamics during pCO 2 fluctuations is potentially of major functional importance. This was explored by measuring Ca 2ϩ dynamics and kinase activity as examples of intracellular signaling.
Ca 2ϩ Signaling-The behavior of the signaling ion Ca 2ϩ was measured in Fluo3-loaded HCT116 cells during oscillations of pCO 2 between 5 and 20% (4-min periodicity). A significant increase in Fluo3 fluorescence was observed upon returning pCO 2 from 20 to 5% (Fig. 4A). Fluo3 fluorescence is modestly pH-sensitive, and to assess the magnitude of this artifact, HCT116 cells superfused in Ca 2ϩ -free and CO 2 /HCO 3 Ϫ -free media were exposed to 40 mM acetate for 4 min to evoke a pH i change comparable to that caused by the pCO 2 fluctuations. Fluo3 fluorescence did not change substantially during this protocol (Fig. 4B), arguing that the responses shown in Fig. 4A report a genuine rise of cytoplasmic [Ca 2ϩ ] ([Ca 2ϩ ] i ). The mechanism of this response was explored using inhibitor drugs and Ca 2ϩ buffers (Fig. 4C). The Ca 2ϩ response was dampened in cells AM-loaded with 1,2-bis(2-aminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetic acid (BAPTA; 100 M), adding to the evidence that Fluo3 reports [Ca 2ϩ ] i changes rather than pH i . Responses were considerably greater when intracellular Ca 2ϩ stores were depleted by pretreatment with 10 M thapsigargin (sarco/endoplasmic reticulum Ca 2ϩ -ATPase inhibitor). In the absence of extracellular Ca 2ϩ (nominally Ca 2ϩ -free; residual Ca 2ϩ buffered with 1 mM EGTA), the Fluo3 response was abolished, arguing that Ca 2ϩ influx is required. Collectively, these data suggest that cytoplasmic alkalinization evokes Ca 2ϩ influx through store-operated channels, such as Orai (64). In support of this, 2-aminoethoxydiphenyl borate (100 M) blocked the Ca 2ϩ response. The CO 2 -evoked [Ca 2ϩ ] i response was dampened in the presence of 100 M ATZ, illustrating a role for CA i activity in accentuating a cellular response to changing pCO 2 . Experiments on co-cultures of HCT116 cells with fibroblasts of low CA i activity (NHDF-Ad, CCD-112-CoN, InMyoFib; Fig. 4, D-F) confirmed that CA i activity can influence the Ca 2ϩ response to pCO 2 fluctuations in a cell-dependent manner.
Kinase Activity-Many kinase-operated signaling pathways are modulated by pH (65) and are expected to respond to changes in pCO 2 . This was tested in HCT116 cells by mea-suring the phosphorylation state of the major mitogen-activated kinases, extracellular signal-regulated kinases, c-Jun N-terminal kinases, and p38 isoforms. pCO 2 was held constant at 5 or 20% or oscillated between 5 and 20% CO 2 for six cycles (4-min periodicity) in the presence or absence of ATZ to manipulate the dynamics of the pH i response (note that at the end of pCO 2 oscillations cells were returned to 5% CO 2 for 2 min). After 26 min of superfusion, lysates were tested for kinase activity readouts to identify pH-sensitive and pH-insensitive kinases over the pH range studied. Among the kinases investigated, phosphorylation of ERK1 (Thr-202/Tyr-204) and ERK2 (Thr-185/ Tyr-187) was unaffected by pCO 2 changes (Fig. 5A). In contrast, kinases, such as JNK1 and JNK2, were dephosphorylated (Thr-183/Tyr-185) by stably raised pCO 2 . Transient exposure to 20% CO 2 (i.e. oscillating pCO 2 ) was inadequate to reduce phosphorylation, indicating that a sustained fall in pH i is nec- essary for eliciting a change in JNK1/2 activity (Fig. 5B). Stably decreased pH i also reduced the activity of mTOR complex 1 (mTORC1), reported as S6 kinase (S6K) phosphorylation at Thr-389 (65)(66)(67). Unlike the other kinases tested, mTORC1 signaling was dependent on CA i activity under fluctuating pCO 2 protocols (higher S6K phosphorylation in the presence of ATZ; Fig. 5C; note that expression of the mTOR regulator G␤L did not underlie differences in S6K phosphorylation). For the four pCO 2 protocols, S6K phosphorylation did not correlate with the time-averaged [H ϩ ] i (Fig. 5C, panel ii), arguing that the dynamics of pH i must be influencing the mTORC1 pathway. A single, transient (2-min) exposure to 20% CO 2 was sufficient for observing an effect of CA i activity on mTORC1 (higher S6K phosphorylation in the presence of ATZ; Fig. 5D). The mTOR inhibitor rapamycin (10 M) produced a substantial decrease in S6K phosphorylation (Fig. 5E), confirming the validity of using S6K as an mTORC1 readout. At constant (5%) pCO 2 , ATZ did not affect mTORC1 activity (Fig. 5E), indicating that the ATZ sensitivity of S6K phosphorylation measured under oscillating pCO 2 is not an off-target effect of the CA inhibitor. Consistent with this finding, ATZ had no effect on S6K phosphorylation in CAII knockdown HCT116 (Fig. 5F) and wild-type MDA-MB-468 (Fig. 5G), i.e. cells with low CA i activity. In summary, mTORC1 signaling responds to sharp pCO 2 changes that are attainable with high CA i activity. These proof-of-principle experiments demonstrate that CA i activity is able to modulate the coupling between extracellular pCO 2 (a function of blood flow) and intracellular signaling.

DISCUSSION
CA i Activity in Cancer Cells-In this study, we quantified CA i activity in a panel of cancer cells and compared these data with results from fibroblast and fibroblast-related cells. Activity measurements were performed under physiological conditions, in the native and undiluted cytoplasmic environment of intact cells, and in the presence of relevant regulatory cytoplasmic influences (Fig. 1A). CA i activity varied considerably among cancer cell lines from essentially absent in HeLa and MDA-MB-468 cells to high in HCT116, HT29, and RT112 cells. Low CA i activity was characteristic of fibroblasts/fibroblast-related cells, arguing that CA i catalysis in solid tumors is more likely to be associated with cancer cells rather than stromal fibroblasts. Soluble (but not secreted) and cytoplasm-facing membrane-tethered CA isoforms may contribute toward the CA i activity. Cancer cells with high total CA i activity were also positive for soluble CA activity in membrane-free cell lysates (Fig. 1B) and for CAII immunoreactivity (Fig. 1D). Genetic knockdown of CAII in HCT116 cells demonstrated that the majority of CA i activity was attributable to CAII (Fig. 1C). Additional isoforms may contribute to CA i activity in other cell lines, such as HT1080 cells that lack CAII expression and soluble CA activity, but have measurable overall CA i . Based on data from all cells investigated, CAII expression correlated strongly with soluble CA activity and was a good predictor of high total CA i activity.
CA i Activity Is Not Universally Rate-limiting for pH i Regulation-The physiological role for CA i activity in cancer is contentious although clearly distinct from that of exofacial CAs (6). Here, we explored the cellular processes that may depend on CA i activity. Previous studies have argued for a role of CA i activity in facilitating H ϩ or HCO 3 Ϫ transport across membranes (23)(24)(25)(26) and H ϩ diffusion in cytoplasm (58,59), but these interactions have not been tested robustly in cancer cells. CA i activity in the six cancer cell lines studied does not correlate with resting pH i , NHE flux, or HCO 3 Ϫ transporter flux measured previously in these cells (3). Our present data (Fig. 2) show that the H ϩ and HCO 3 Ϫ fluxes produced by cancer cells over the physiological pH i range are not of sufficient magnitude to require CA i catalysis, with the exception of acid loading by HCO 3 Ϫ export at high pH i . In the absence of CA i activity, the maximal capacity of CO 2 hydration to generate HCO 3 Ϫ ions for extrusion is ϳ14 mM/min at 5% CO 2 (i.e. k f ϫ [CO 2 ]), and this can be rate-limiting for fast HCO 3 Ϫ -dependent acid-loading transporters as measured in RT112 cells ( Fig. 2A, panel iii). However, the highly alkaline intracellular conditions that are required for producing this CA i dependence are unlikely to be typical of cancer cells. CA i -catalyzed hydration of CO 2 has recently been proposed to facilitate Na ϩ -HCO 3 Ϫ co-transport in cardiac myocytes by supplying the transport protein with H ϩ ions for titrating HCO 3 Ϫ (26). However, genetic CAII knockdown (HCT116; Fig. 2C, panel iv) or pharmacological inhibition with ATZ (RT112; Fig. 2A, panel iii) did not affect HCO 3 Ϫ -dependent acid extrusion. This difference may reflect contrasting structural and chemical properties of cardiac and cancer cytoplasm. Alternatively, CA i dependence may only be observed experimentally under high Na ϩ -HCO 3 Ϫ co-transport fluxes, which are attainable in cardiac myocytes (under hyperkalemic stimulation) but not in cancer cells. NHE activity was able to produce rapid H ϩ extrusion in MDA-MB-468 cells that naturally have very low CA i activity. This argues that CA i -independent delivery of H ϩ ions does not limit the membrane transport process (Fig. 2B) in agreement with an earlier study on ventricular myocytes (26). In HCT116 cells (high CA i activity), intrinsic cytoplasmic buffers alone supported the same magnitude of NHE flux as measured in the presence of CO 2 / HCO 3 Ϫ or after loading cells with the highly mobile buffer carnosine (Fig. 2C). Even in the absence of CO 2 /HCO 3 Ϫ (hence no CA i catalysis), activation of NHE did not evoke measurable pH i non-uniformity. This supports the case that diffusive H ϩ coupling across cancer cytoplasm is normally adequate with intrinsic (non-carbonic) buffers. Knockdown of CAII did not affect NHE activity in HCT116 cells (measured as the DMA-inhibitable component of flux; Fig. 2C, panel iv), arguing that neither catalysis nor the presence of CAII protein is necessary for high NHE fluxes. Collectively, these observations argue for the absence of rate-limiting barriers to H ϩ diffusion over the relatively small dimensions of cancer cells (mean radii, Ͻ10 m). Additionally, cancer cells typically down-regulate gap junctions (22), which in other tissues allow for greatly expanded diffusion lengths. The absence of electrical coupling explains why CA i catalysis cannot facilitate CO 2 or H ϩ diffusion appreciably in tumors as this would require HCO 3 Ϫ ions to diffuse relatively freely between cells.
CA i Activity Influences the Degree of Coupling between pCO 2 and pH i Dynamics-Most investigations of the role of CA i activity in pH i regulation have been performed under constant pCO 2 . This condition may be appropriate for well perfused tissues with stable CO 2 production and venting but is not representative of metabolically active solid tumors with intermittent blood flow. Fluctuations in tumor blood flow are the basis for time-dependent changes in pO 2 (also known as acute hypoxia) that influence tumor biology (34). During periods of inadequate capillary washout, pCO 2 rises as O 2 is depleted, and these changes reverse upon blood reperfusion. Although direct, high resolution measurements of pCO 2 fluctuations are not available because of technical limitations, the inverse correlation between time-averaged pO 2 and pCO 2 (acidity) is well established (43)(44)(45). Our standard experimental protocol of varying pCO 2 between 5 and 20% with 4-min periodicity is based on measurements of extracellular pH, blood flow, and pO 2 in solid tumors in vivo. The pH of 20% CO 2 superfusate is 6.8, which is considered to be typical of the mean extracellular pH of most solid tumors (63, 68); pH 7.4 of 5% CO 2 superfusates is normal for blood plasma. Our choice of periodicity is in the range of fluctuation frequencies of blood flow and/or pO 2 established by Fourier analysis in tumors in vivo (30,33,42).
Changes in pCO 2 will have knock-on effects on [H ϩ ] i dynamics and hence the many pH-sensitive downstream processes. The coupling between pCO 2 and pH i is strongly dependent on CA i as illustrated in Fig. 3 where the same experimental protocol of changing pCO 2 between 5 and 20% produced different Mathematical modeling demonstrates that the effect of CA i activity on [H ϩ ] i dynamics was important even with a smoother sinusoidal pCO 2 waveform (Fig. 3G), which is more representative of undulating blood flow. The ability of CA i to influence [H ϩ ] i dynamics was predicted for a wide range of pCO 2 waveform amplitudes but required a periodicity of 4 min or less for physiologically meaningful effects (Fig. 3H). This frequency dependence is expected because even at spontaneous reaction rates, [H ϩ ] i is able to track slow changes in pCO 2 . Given that fluctuations of blood flow (29 -32) and pO 2 (30, 32, 34 -41) in solid tumors have been observed with a periodicity as short a 0.5 min (i.e. two cycles/min), a range of effects on [H ϩ ] i dynamics could be achieved by regulating CA i activity: from tight temporal pCO 2 -pH i coupling at high CA i activity to low pass filtering with negligible CA i activity. Cancer cells would be able to tune their pH i responsiveness to changes in pCO 2 (blood flow) by regulating CA i activity, e.g. through CAII expression.
Our study explored the functional significance of the CA i dependence of [H ϩ ] i dynamics using Ca 2ϩ signaling and kinase-operated cascades as proof-of-principle examples (Fig.  6). In HCT116 cells, a transient pH i rise, evoked by reducing pCO 2 , increased [Ca 2ϩ ] i . Our evidence points to store-operated calcium entry as the mechanism triggering the [Ca 2ϩ ] i response (Fig. 4C). Recent work has demonstrated that storeoperated calcium entry is reduced at low pH i because of an uncoupling between the endoplasmic reticulum Ca 2ϩ sensor STIM and the surface membrane Ca 2ϩ channel Orai (64).
Because [Ca 2ϩ ] i responds dynamically to changes in the balance between influx and efflux pathways, pH i sensitivity of Ca 2ϩ entry would produce the observed [Ca 2ϩ ] i fluctuations. Under CA i inhibition with ATZ, the same pCO 2 stimulus evoked a slower and smaller [Ca 2ϩ ] i response.
Previous studies have shown that the environmental sensor mTOR, which is strongly linked with cancer (66,69,70), is inhibited at low pH i (65). The present work shows that the degree of mTOR inhibition also depends on CA i activity when pCO 2 is fluctuating. In HCT116 cells, a single cycle of pCO 2 oscillation produced a lower readout of mTORC1 signaling (S6 kinase phosphorylation) when CA i activity was intact (Fig. 5D). This inhibitory effect of CA i catalysis on mTORC1 signaling was substantiated by expanding the protocol to six cycles of pCO 2 fluctuations (Fig. 5C). mTORC1 signaling was generally higher when pCO 2 oscillations were performed on HCT116 cells with pharmacologically (ATZ) or genetically (knockdown; Fig. 5F) reduced CA i activity or on cells with naturally low CA i activity, such as MDA-MB-468 (Fig. 5G). mTORC1 activity was not a unique function of time-averaged pH i . For example, holding pCO 2 stably at 20% or fluctuating pCO 2 between 5 and 20% in the presence of ATZ produced a similar time-averaged pH i , but S6 kinase phosphorylation differed by a factor of 2 ( Fig. 5C, panel ii). We conclude that the dynamics of pH i changes must influence mTORC1. Akin to the notion that transient hypoxia (perfusion-limited) and stable hypoxia (diffusion-limited) have distinct biological consequences (71), there may be similar "frequency modulation" and "amplitude modulation" (72) aspects of H ϩ signaling in cancer. A possible explanation why some kinases (such as JNK1/2) do not respond to fluctuating pCO 2 is a slow binding of H ϩ ions to modulatory sites that acts like a low pass filter. mTOR may be able to register rapid pH i fluctuations by faster H ϩ binding kinetics. Further studies are necessary to explore these possible mechanisms.
CA i -catalyzed CO 2 hydration will produce fluctuations in [HCO 3 Ϫ ] that parallel pH i changes, and it is plausible that HCO 3 Ϫ ions interact with targets in a pH-independent manner. Removing HCO 3 Ϫ ions from cytoplasm also removes the substrate (CO 2 ) for CA i ; therefore a simple buffer substitution experiment would not be an appropriate test to distinguish the effects of H ϩ and HCO 3 Ϫ ions. However, it is generally accepted that H ϩ ions are more reactive than HCO 3 Ϫ ions. Furthermore, earlier observations of the acid response of store-operated calcium entry (64) and mTORC1 (65) were made in the absence of CO 2 /HCO 3 Ϫ , arguing for H ϩ ions as the key regulators. Growing evidence points to a negative correlation between the expression of CA i isoforms, such as CAII, and cancer disease severity (11,(13)(14)(15). Importantly for disease progression, CAII down-regulation would help to protect cytoplasmic pH from pCO 2 fluctuations and bestow cancer cell pH with a greater degree of autonomy from extracellular influences, such as those arising from intermittent blood flow (73). Conversely, rapid changes in blood flow would be registered by cells with high CA i activity. For example, mTOR-regulated metabolism and proliferation would be more responsive to fluctuating blood flow in cancer cells with higher CA i activity, particularly in regions close to aberrant blood vessels, which experience the sharpest pCO 2 fluctuations. Incidentally, these tumor regions are also important for metastasis and nutrient sensing. In solid tumors made of cancer cells with high CA i activity surrounded by a fibroblast stroma of low CA i activity, pCO 2 fluctuations would evoke distinct [H ϩ ] responses in the two cell types as illustrated by [Ca 2ϩ ] i responses (Fig. 4).
In conclusion, this study demonstrates a novel role for CA i activity in cancer as a transducer of pCO 2 fluctuations (which arise from intermittent blood flow) into a potent cytoplasmic signal (H ϩ ions). Cancer cells may alter the degree of temporal coupling between pCO 2 and pH i by tuning cytoplasmic CA i activity, e.g. through CAII expression (Fig. 6). This work highlights the importance of dynamic aspects of pH signaling as a modality distinct from responses to stable pH changes.