Proton export drives the Warburg Effect

Aggressive cancers commonly ferment glucose to lactic acid at high rates, even in the presence of oxygen. This is known as aerobic glycolysis, or the “Warburg Effect”. It is widely assumed that this is a consequence of the upregulation of glycolytic enzymes. Oncogenic drivers can increase the expression of most proteins in the glycolytic pathway, including the terminal step of exporting H+ equivalents from the cytoplasm. Proton exporters maintain an alkaline cytoplasmic pH, which can enhance all glycolytic enzyme activities, even in the absence of oncogene-related expression changes. Based on this observation, we hypothesized that increased uptake and fermentative metabolism of glucose could be driven by the expulsion of H+ equivalents from the cell. To test this hypothesis, we stably transfected lowly-glycolytic MCF-7, U2-OS, and glycolytic HEK293 cells to express proton exporting systems: either PMA1 (yeast H+-ATPase) or CAIX (carbonic anhydrase 9). The expression of either exporter in vitro enhanced aerobic glycolysis as measured by glucose consumption, lactate production, and extracellular acidification rate. This resulted in an increased intracellular pH, and metabolomic analyses indicated that this was associated with an increased flux of all glycolytic enzymes upstream of pyruvate kinase. These cells also demonstrated increased migratory and invasive phenotypes in vitro, and these were recapitulated in vivo by more aggressive behavior, whereby the acid-producing cells formed higher grade tumors with higher rates of metastases. Neutralizing tumor acidity with oral buffers reduced the metastatic burden. Therefore, cancer cells with increased H+ export increase intracellular alkalization, even without oncogenic driver mutations, and this is sufficient to alter cancer metabolism towards a Warburg phenotype.


9
In 1924, Otto Warburg and colleagues demonstrated that cancer, even in the presence of oxygen, ferments glucose to lactic acid at high rates (Warburg et al., 1924), and this was contemporaneously confirmed by the Coris (Cori and Cori, 1925). Aerobic glycolysis, commonly termed the "Warburg Effect" in cancer (DeBerardninis and Chandel, 2020; Vaupel and Multhoff, 2020), is undeniably a hallmark of primary tumors and aggressive invasive disease (Hanahan and Weinberg, 2011). This preference of tumors for aerobic glycolysis is exploited in diagnostic PET imaging of fluorodeoxyglucose ( 18 F-FDG) uptake (Kunkel et al., 2003). It is commonly believed that this increased fermentative glycolysis, and thus proton flux, is driven by oncogenes, such as RAS, MYC, HIF, and AKT (Kim et al., 2007;Wonsey et al., 2002), and that this augmented flux out-competes the ability of mitochondria to oxidize pyruvate, leading to the net production and export of lactic acid and reconversion of NADH to NAD + to maintain redox balance. Hence, the Warburg Effect could solely be an epiphenomenon of oncogene activation, which is consistent with the observation that fermentation under aerobic conditions is energetically unfavorable and does not confer any clear evolutionary benefits The current study investigates whether aerobic glycolysis can be driven by protonexport, and further investigates the impact of this on cancer aggressiveness. We demonstrate that over-expression of proton exporters is sufficient to increase aerobic glycolysis, through enhanced glucose uptake and lactate production. We further observed that proton export increased intracellular pH and increased metabolic flux at most steps in glycolysis. Finally, we observed in vivo that these proton-exporting cell lines were more aggressive, generating higher-grade tumors and increased metastases. There is a known association between acid production, aerobic glycolysis, and metastatic potential. Further, experimental metastases can be inhibited with acidneutralizing buffers. The current work adds to this literature by demonstrating that acid production per se can be sufficient to drive the Warburg Effect and promote metastasis.

Overexpression of CA-IX in cancer cells increases glycolytic metabolism 10
CA-IX hydrates extracellular CO2 to H + + HCO3 -. This facilitates CO2 diffusion away 11 from the cell reducing pH gradients across tissues. The bicarbonate generated from CO2 12 hydration, a reaction that occurs either spontaneously or is sped up by carbonic anhydrases, can 13 then reenter the cell via Na + + HCO3co-transporter (Boedtkjer, 2019;Svastova et al., 2012). 14 Numerous studies in, e.g., breast, ovarian (Choschzick et al., 2011), and astrocytoma 15 (Nordfors et al., 2013) cancers have shown that CA-IX expression correlates with poor prognosis 16 and reduced survival. Figure 1A shows the overall survival of ER+ breast cancer patients with 17 low and high (median cutoff) CA-IX expression was 143 and 69.4 months, respectively, p=9.32 18 e-7. Although the sample size was smaller, a similar pattern was seen for metastasis-free 19 survival in patients with low and high CA-IX expression, 130 vs. 50 months, respectively, 20 p=0.0012. Metastasis-free survival is also reduced in other cancers with high CA-IX expression, 21 including cervical and colorectal (van Kuijk et al., 2016). CA-IX's role in outcome makes it a 22 clinically significant target warranting further investigation. 23 To test the hypothesis that proton export can drive aerobic glycolysis, we established 24 models which over-expressed proton exporters. We transfected MCF-7 cells with a CA-IX 25 expression vector and isolated two individual clones (M1 and M6) and confirmed CA-IX protein 26 expression (Fig. 1B). MCF-7 cells do not express CA-IX under normoxic conditions, but do 27 express other carbonic anhydrases, CA-II and CA-XII (Fig. 1B). CA-IX is distinct among 28 exofacial CA's (CA-IV, CA-XII), as it contains a proteoglycan domain, which enables it to 29 maintain enzymatic activity at lower pHe (Li et al., 2011). In tissues, CA-IX can function as a 30 "pH-stat", which tumors hijack to maintain an acidic pHe (Lee et al., 2018). We also transfected 31 MCF-7 cells with an empty pcmv6 vector hereafter referred to as MOCK-2. Additionally, we 32 confirmed by ICC that CA-IX, an exofacial membrane-bound protein, was expressed on the 33 plasma membrane in both CA-IX clones (Fig. 1C). 34 To test our hypothesis that proton export can drive aerobic glycolysis, we interrogated the 35 metabolism of our CA-IX expressing clones using a Seahorse XFe96 Extracellular Flux (XF) 36 Analyzer, enzymatic, and radiochemical assays to assess both glycolytic and mitochondrial 37 metabolism. Specifically, the Seahorse glycolytic stress tests (GST) showed that both CA-IX 38 clones exhibited higher proton production rates (PPR) upon glucose stimulation compared to 39 MOCK-2 or parental clones (Fig. 1D). Using glucose, and [ 3 H]-2-deoxyglucose (2DG) uptake 40 assays ( Fig. 1E & Supplementary Fig. 1), and lactate production rate assays ( Fig. 1F & 41 Supplementary Fig. 2), we further confirmed CA-IX expressing clones had increased glycolysis 42 in normoxic conditions. Mitochondrial metabolism in these clones, as measured by Seahorse 43 mitochondrial stress test (MST), exhibited a decreased reliance on oxidative phosphorylation. In 44 the CA-IX clones, both basal oxygen consumption rates (Fig. 1G) and reduced ATP-linked 45 oxygen consumption rates (Supplementary Fig. 3) were decreased compared to MOCK-2 or 46 parental cells. CA-IX expressing clones also had hyperpolarized mitochondria (Supplementary 47 Fig. 4). Therefore, CA-IX expression did not globally upregulate all ATP turnover, but 48 upregulated aerobic glycolysis and limited reliance on oxidative phosphorylation. 49 (red). D: Glycolysis associated proton production rate (PPR) using the Seahorse extracellular flux analyzer, 59 measured post glucose injection. Data are shown as mean ± SD, N=8 biological replicates per group, statistical 60 analysis using ordinary one-way ANOVA. E: Glucose uptake of cells in each group over 24hr, measured as 61 luminescence generated using Glucose Uptake-Glo assay (Promega). N=3, statistical analysis using ordinary 62 one-way ANOVA. F: Lactate measured in extracellular media after 24hr using Sigma kit. N=3 biological 63 replicated per group, statistical analysis using ordinary one-way ANOVA. G: Basal oxygen consumption rate 64 (OCR) measured using the Seahorse extracellular flux analyzer in 5.8mM glucose concentration. Data are shown 65 as mean ± SD, N=8 biological replicates per group, statistical analysis using ordinary one-way ANOVA. 66 *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 67 68 To investigate whether these metabolic alterations were specific to the MCF-7 cells, we 69 tested other cell lines. CA-IX was over-expressed in U2-OS osteosarcoma human cells and HEK 70 293 human embryonic kidney cells. CA-IX expression upregulated glycolysis in both cell lines, 71 as seen by increased aerobic lactate production ( Supplementary Figs. 5 & 6). Even in HEK 293 72 cells, which have higher basal glycolysis than the other cell lines tested, over-expression of the 73 proton exporting CA-IX still enhanced their rate of aerobic glycolysis. 74 To delineate which steps in glycolysis were being impacted by acid export, we analyzed 75 the intracellular metabolites using mass spectrometry and a library of known metabolites (see 76 methods). Principal component analysis (PCA) showed a statistically significant separation of 77 MOCK-2 and parental MCF-7 from the CA-IX clones ( Fig. 2A). Notably, PCA also showed that 78 the parental and MOCK-2 cells had distinct metabolic profiles, but neither overlapped with the 79 CA-IX clones. Consistent with this, heatmap visualization of the top 50 most significantly 80 altered metabolites showed that the two CA-IX clones exhibited similar metabolic profiles (Fig.  81 2B) but were considerably different from the MOCK-2 and parental clones. MOCK-2 cells grow 82 more rapidly in vitro compared to the parental MCF-7 or the CA-IX clones which is likely why 83 they exhibit a different metabolic profile from the parental line (Supplementary Fig. 7). We 84 note later, however, that this difference in growth rate was not maintained in vivo when grown as 85 primary tumors (Fig.4A). Out of all metabolites assessed, the glycolytic intermediates were 86 consistently altered in the CA-IX clones compared to parental or MOCK-2. 87 The CA-IX clones exhibited increased levels of all glycolytic intermediates upstream of 88 pyruvate kinase, PK ( Fig. 2C-J), which catalyzes the penultimate step of glycolysis: the 89 conversion of phosphoenolpyruvate (PEP) + ADP → pyruvate +ATP. Thus, it appears that the 90 activities of the upstream enzymes have increased, leading PK to now become rate-limiting for 91 glycolytic flux in the CA-IX expressing cells (Fig. 2J). As the CA-IX cells have higher 92 glycolytic flux ( Fig. 1D-F), the most straightforward interpretation is that CA-IX expression de-93 inhibited all of the enzymatic steps upstream of PK. Overall, CA-IX expression enhances 94 glycolytic intermediates' flux, resulting in enhanced lactate and acid production. Our metabolomics studies suggest multiple glycolytic enzymes were impacted, we 109 therefore hypothesized that CA-IX expression might raise the intracellular pH of cells, which 110 could pleiotropically increase glycolytic enzyme rates. Most glycolytic enzymes have ionizable 111 residues that can alter their enzyme activity. Recently, these residues have been characterized for 112 all glycolytic enzymes through homology modelling, which predicted glycolytic enzyme 113 activities generally increase with pHi above neutral (Persi et al., 2018). We tested the PPR in 114 MCF-7 parental cells after altering pHi in a range of pHe media 6.6-7.4, expecting that the PPR 115 rate would increase with increasing pHi. Cells were incubated with either a chloride-containing 116 solution or an iso-osmotic low-chloride formulation which replaced chloride salts with gluconate 117 equivalents. Low-chloride solutions alter the driving force for Cl -/HCO3exchangers, effectively 118 loading cells with HCO3ions and raising pHi at the same pHe (Sasaki and Yoshiyama, 1988; 119 Wu et al., 2020). At all values of pHe from 6.6-7.2, cells in gluconate media exhibited 120 significantly higher glycolytic rates compared to those in chloride-containing media ( Fig. 3A & 121 Supplementary Fig. 8), indicating that the increased glycolytic rate is strongly dependant on 122 pHi. 123 CA-IX overexpression has been shown to increase pHi in other systems (Morgan et al., 124 2007), leading us to probe pHi in our CA-IX expressing cells. We used a pHi reporter dye, 125 cSNARF1 to measure the effects of CA-IX expression on intracellular pH (pHi). A cSNARF1 126 calibration curve was generated using nigericin/K + buffers (Supplementary Fig. 9). We then 127 loaded MCF7 parental or the CA-IX expressing clones with cSNARF1 and used fluorescence 128 imaging to measure pHi. MOCK-2 cells were unsuccessfully loaded with cSNARF1, possibly 129 due to reduction or loss of esterase activity from the integration of the MOCK-2 plasmid into the 130 MCF-7 DNA; therefore, they are not included in these analyses. We equilibrated the cells in 131 media at pHe 6.6, 6.8, 7.1, and 7.4 and co-loaded them with cSNARF1 and nuclear dye Hoechst 132 33342. The inclusion of a nuclear dye allowed post-processing to mask the nucleus, ensuring 133 measurement of cytoplasmic pH only. At the intermediate pHe values of 6.8 and 7.1, the CA-IX 134 expressing clones had significantly higher pHi compared to parental (Fig. 3C  In vitro, we utilized scratch assays and gel escape to measure migration and invasion in 160 the CA-IX clones. Scratch assays showed that clone M1 had increased migratory ability 161 ( Supplementary Fig. 10), closing the wound significantly more rapidly than controls. However, 162 this was not observed in the M6 CA-IX clone. In the gel escape assay, expansion out of the gel 163 is due to a combination of proliferation, invasion, and migration. Both CA-IX clones invaded 164 the area surrounding the gel drop substantially quicker than the parental MCF-7 clone 165 ( Supplementary Fig. 11). Compared to the MOCK-2 cells, however, only the M6 clone was 166 significantly more invasive. The apparent increased invasion rate by MOCK-2 compared to 167 parental is likely due to their increased proliferation rate as correcting for proliferation eliminates 168 the difference between MOCK-2 and parental (Supplementary Fig. 7). It is important to note 169 that these assays were carried out at neutral pH and that the invasive behavior might be further 170 enhanced by low pH, as we have shown previously for melanoma cells (Moellering et al., 2008). 171 Together, these studies suggest that CA-IX expression can enhance cell motility, increasing 172 migration and local invasion. 173 Furthermore, aggressive cancers with stem-like properties can resist anoikis, which can 174 be measured in vitro by cells' ability to form spheroids independent of attachment to a basement 175 membrane. Utilizing the hanging droplet technique, we observed that CA-IX expression enabled 176 robust, compact spheroid formation, compared to the MOCK-2 and parental MCF-7 clones, 177 which could not (Supplementary Fig. 12). This spheroid forming ability suggests CA-IX not 178 only enhances cell:cell adhesion but also suggests that increased proton export can contribute to 179 anoikis resistance when detached from the basement membrane. 180 Although the phenotype of our proton exporting CA-IX clones appeared to be more 181 aggressive, this could be an in vitro only phenomenon. We thus investigated the clones in vivo. 182 We studied the effect of CA-IX expression on primary tumor growth, as well as the ability of 183 these clones to form both spontaneous (from the mammary fat pad) and experimental (tail vein 184 injected) metastasis. For primary and spontaneous metastasis models, MOCK-2, M1 or M6 CA-185 IX MCF-7 cells were implanted in the mammary fat pads of mice, and growth was monitored by 186 caliper measurement. Although in vitro, mock cells proliferated faster (Supplementary Fig. 7), 187 this was not observed in vivo, as the tumors from CA-IX expressing clones grew significantly 188 faster. At all of the time points measured, primary tumor volume was significantly increased in 189 the CA-IX expressing clones, compared to controls ( Fig. 4A & Supplementary Fig. 13). Because formation of spontaneous metastases is a multi-step, time-consuming, and 203 complex process, we also investigated the ability of these clones to form experimental metastases 204 following tail-vein injection, which only involves the final steps of extravasation and 205 colonization. Metastases to the lung were scored blindly by a board-certified pathologist (M.B.), 206 who observed that neither of the control groups developed metastases, and that both CA-IX 207 clones had significant macrometastases (Table 1). Consistent with the spontaneous model, the 208 M1 formed fewer metastases compared to the M6, however both CA-IX clones exhibited gross 209 metastasis to the lungs (Fig. 4C), which were confirmed histologically (Fig. 4D). The resulting 210 macrometastases significantly reduced overall survival of the mice (Fig. 4E). Additional IHC 211 staining confirmed the lung metastasis expressed both CA-IX and human estrogen receptor, 212 which we used as a marker for MCF7 cells (Fig. 4F). 213 Prior studies have shown that neutralization of acidity using oral buffers inhibits 214 metastasis (Ibrahim-Hashim et al., 2017; Ibrahim-Hashim et al., 2012). As we hypothesized that 215 our M6 CA-IX clones were metastatic by virtue of increased acid production (Fig. 1D, E), we 216 asked whether buffer therapy would reduce the metastatic burden. Using the experimental 217 metastasis model, we compared untreated to buffer-treated M6 or MOCK-2 mice. As in the first 218 experimental metastasis study, all mice injected with M6 developed macrometastases, and these 219 were significantly reduced by bicarbonate ( Fig. 4G & H). Two out of ten MOCK-2 mice 220 developed very small micrometastasis, and no metastases were found in the buffer therapy 221 MOCK-2 group (Fig. 4G).

Over-expression of yeast proton pump PMA1 increases motility and metabolism 244
While we hypothesize that CA-IX is acting as a proton equivalent exporting system, there 245 are many other activities of this protein, including non-enzymatic activities, that could be 246 activating glycolysis and promoting metastasis. To test whether the observed effects on 247 glycolysis could be due to increased proton export, we utilized another model, PMA1, which 248 electrogenically pumps H + out of cells at the expense of ATP (Ferreira et al., 2001). Prior work 249 has shown ectopic expression of PMA1 in murine 3T3 fibroblasts led to tumorigenesis (Perona 250 and Serrano, 1988) and to increased aerobic glycolysis with elevated intracellular pH (Gillies, 251 1990; Martinez et al., 1994). We engineered MCF-7 cells to express PMA1 and selected two 252 clones following zeocin selection (PMA1-C1 and PMA1-C5) as well as an empty vector 253 transfected control (MOCK-1). PMA1 over-expression in C1 and C5 was confirmed, by qRT-254 PCR (Supplementary Fig. 14) and western blot (Fig. 5A). Moreover, using 255 immunocytochemistry of non-permeabilized cells (Fig. 5B & Supplementary Fig. 15) or 256 permeabilized cells (Supplementary Fig. 16 To characterize the metabolic activity of PMA1 transfectants, we again utilized the 287 glycolysis and mitochondrial stress tests (GST and MST, respectively) of the Seahorse (XFe) 288 Analyzer. The glucose-induced PPR (Fig.5C), and the glycolytic reserve (Supplementary Fig.  289 17) were significantly higher in the PMA1 clones compared to empty vector MOCK-1 or 290 parental clones, suggesting functional activities of the transfected pump. In contrast to the CA-IX 291 transfectants, there were no significant differences in oxygen consumption rates (OCR) between 292 PMA1 clones and controls (Fig. 5D). This could be due to increased energy demand from the 293 ATPase proton pump. We further confirmed these metabolic alterations by measuring glycolytic 294 flux. PMA1 clones had significantly higher glucose consumption rates (Fig. 5E) and lactate 295 production rates (Fig. 5F) compared to MOCK-1 or parental MCF-7 clones. These data, together 296 with the CA-IX results, indicate that acid export can drive cells to exhibit a Warburg phenotype. 297 As with the CA-IX transfectants, we also measured invasion and migration using gel 298 escape and circular wound healing assays, respectively. Compared to the MOCK-1 cells, both 299 PMA1 clones expanded significantly more out of the gel drop ( Fig. 5G & Supplementary Fig.  300 18). In the circular "wound-healing" assay, we monitored the migration of cells into a cell-free 301 area. Again, compared to MOCK-1, both PMA1 clones had increased migration rates (Fig. 5H & 302 Supplementary Fig. 19). Together, these results indicate that cellular invasion and migration 303 were significantly enhanced by PMA1 expression and acid production. To investigate if proton export enhanced aggressiveness, as seen in the CA-IX model, we 309 measured the PMA1 cells' metastatic ability in vivo in both spontaneous and experimental 310 metastasis models. In our in vitro studies, the proliferation rates of PMA1-C1 and the empty 311 vector MOCK-1 clones were similar, whereas the growth rate of PMA-C5 was significantly 312 slower (Supplementary Fig. 20). Hence, we omitted PMA-C5 in our in vivo studies to reduce 313 the possibility of proliferation being a confounding variable. In both the spontaneous and tail 314 vein metastases models, only 1 of 10 mice in each MOCK-1 group developed lung metastases. In 315 contrast, 7 of 12 PMA-C1 mice developed lung metastases following tail vein injection and 4 of 316 9 formed spontaneous metastases (Supplementary Table 2.). 317 318 Lung metastases of PMA1 cells, visualized by H&E staining (Fig. 6A), were further 319 validated by immunohistochemistry (IHC) of PMA1 (Fig. 6A) and RNA analysis of FFPE lung 320 tissue for PMA1 gene expression (Supplementary Fig. 21). Notably, primary tumors revealed 321 no significant growth differences between PMA1 and MOCK-1 tumors (Fig. 6B) or final tumor 322 volume (Fig. 6C). However, blind grading (1 to 4+) of H&E-stained tumor sections by a board-323 certified pathologist (A.L.) indicated that PMA1 primary tumors were of significantly higher 324 grade compared to MOCK-1 controls (p=0.016). The average grade was 2.7 ± 0.52 for MOCK-1 325 compared with 3.4 ±0.67 for the PMA1 primary tumors (Fig. 6D and 6E). Maintenance of 326 PMA1 expression in vivo was confirmed with quantitative IHC of the resected primary tumors, 327 demonstrating a significant difference in PMA1 protein expression between the MOCK-1 (19%± 328 3.0, n = 10) and PMA-C1 (90%± 2.5, n = 9) tumors ( Fig. 6F & 6I). 329 330 331 332 Additional IHC of PMA1 tumors showed that they had significantly lower CA-IX 359 expression than MOCK-1 (Fig. 6H &6K). As CA-IX plays a vital role in regulating tumor pH 360  (Fig. 6G &6J), which has been associated with 363 increased aggressiveness in breast cancer (Doherty et al., 2014). Notably, other proteins, such as 364 glucose transporter 1, GLUT1 (Supplementary Fig. 22), the sodium hydrogen exchanger 1, 365 NHE1 (Supplementary Fig. 23), and MCT4 (Supplementary Fig. 24) showed no differences 366 between the PMA1 and MOCK-1 groups. These data suggest that the increased glycolytic flux, 367 which requires glucose uptake by GLUT1, can be accommodated by native protein levels of 368 GLUT1 (i.e., it is not rate-limiting). 369

DISCUSSION
The primary goal of this study was to determine if acid export per se could drive aerobic glycolysis, the "Warburg Effect", in cancer. Aerobic glycolysis, a cancer hallmark, is often associated with more aggressive tumors. Numerous studies have attempted to determine why tumors favor fermentative glycolysis, even in the presence of sufficient oxygen (Gatenby and  Thompson, 2012). However, none of these theories clearly demonstrate why so many cancers favor aerobic glycolysis. We hypothesize that acid export per se and extracellular acidification provides a distinct selective advantage, and that it is enabled by increased glucose fermentation. This theory was first proposed as the "acid To test the hypothesis that acid export increases glycolysis in cancer, we over-expressed a proton exporter, CA-IX, in an OXPHOS dominant cell line, MCF-7. In MCF-7 cells, CA-IX is not expressed under normoxic conditions, and over-expression resulted in cells that more rapidly exported acid and up-regulated glycolysis. Both glucose consumption and lactate production rates increased. Other studies have shown that CA-IX over-expression can increase lactate production, albeit under hypoxic conditions (Jamali et al., 2015). As glycolysis was specifically increased, we investigated the mechanisms whereby acid export could be causing this shift in metabolism. We broadly interrogated cellular metabolism using a targeted metabolomics panel and found that the most significantly altered metabolites in the CA-IX expressing clones were glycolytic intermediates. Specifically, all glycolytic intermediates upstream of pyruvate kinase were increased. Enzyme activity can be affected by pH, including those in the glycolytic pathway, and are most active at the pH of their subcellular compartment from acidic lysosomes to alkaline mitochondria (Persi et al., 2018). For glycolytic enzymes, pH optima are slightly on the alkaline side of neutral (7.2-7.4), meaning that raising pH above neutrality will globally increase activities of glycolytic enzymes. Because the data suggested pleiotropic increases in enzyme activity, we measured intracellular pH using fluorescence ratio imaging in our CA-IX expressing clones. These clones, had a higher pHi in biologically relevant extracellular pH conditions, compared to the parental. Specifically, at pHe 6.8 and 7.2, at which CA-IX can function as a proton equivalent exporter, the CA-IX clones had increased pHi compared to parental. One caveat of this experiment was that the MOCK cells were unquantifiable as they did not accumulate SNARF-1, possibly due to increased activity of multidrug resistance transporters.
In addition, the experiments in low-chloride directly implicate increased pHi in regulating aerobic glycolytic flux. These data indicate loading cells with HCO3ions raises the intracellular pH sufficiently to enhance enzyme activity and result in increased glycolytic flux.
We tested our hypothesis using two more models and another acid exporting protein to minimize cell line and protein-specific effects. Similar to the MCF-7 results, over-expression of CA-IX in U2-OS and HEK293 cells increased glycolysis compared to controls. Although parental HEK293 cells are more glycolytic than the other cell lines chosen, over-expression of CA-IX still enhanced glycolysis. We tested another acid exporting protein, PMA-1, which has an unequivocal activity of exporting protons at the expense of ATP. PMA-1 over-expression in MCF-7 cells similarly resulted in increased glycolysis, as measured by increased glucose uptake and lactate production. These findings, together with our CA-IX results, suggest that expression of proton exporting activity can up-regulate aerobic glycolysis, likely through a global increase of intracellular pH. Notably, while the CA-IX transfectants had reduced oxygen consumption, this was not observed in the PMA1 cells. The cause of this difference is not known and may reflect differences in the bioenergetic requirements for the two transporting systems.
We hypothesize that acid export-driven glycolysis would make them more aggressive, as 371 measured in vitro with motility and invasion assays, and in vivo with experimental and 372 spontaneous metastases studies. Glycolysis and acidity have been correlated with poor prognosis 373 and metastasis (Walenta et al., 2000;Webb et al., 2011a). Our focus on CA-IX as an acid 374 exporter was due to its clinical relevance in many cancer types, such as breast, ovarian, and 375 astrocytoma, where CA-IX over-expression correlates with poor prognosis, reduced survival, and 376 reduced metastasis-free survival. This suggests CA-IX specifically and perhaps acid export 377 generally enhances cancer aggressiveness and subsequently metastasis. In our models, in vitro Our encouraging in vitro results led us to take these models, CA-IX and PMA-1, in vivo. 386 We found that aggression and metastasis were higher in both PMA-1 and CA-IX transfectants. 387 Primary tumor growth was enhanced in the CA-IX model compared to controls. In the PMA-1 388 model, a pathologist blindly scored the primary tumors a higher grade compared to control 389 tumors. However, spontaneous metastasis after primary tumor resection was not significantly 390 increased in PMA-1 or CA-IX transfectants compared to controls ( Table 1 & Supplementary  391   Table 2). It is notable, however, that in our spontaneous models 10/45 mice with PMA-1 or CA-392 IX clones had metastases, compared to 2/20 control mice with parental or MOCK-transfections. 393 In contrast, our experimental metastasis model, which skips the intravasation step, showed 394 enhanced experimental metastasis in both models, with 17/25 mice in PMA-1 or CA-IX 395 transfectants developing metastasis, compared to 1/21 metastases in control mice (Table 1 &  396   Supplementary Table 2). We did not quantify the number or size of metastatic lesions, because 397 the important metric is binary: i.e. whether or not these clones were able to metastasize at all. A 398 related study in 4T1 breast cancer showed that inhibiting CA-IX reduced tumor growth and 399 experimental metastasis (Lou et al., 2011). However, inhibition is different than induction, and 400 4T1 are highly glycolytic and acidic to begin with. However, this study does indicate the 401 importance of acid export and its role in enhancing tumor growth. Due to the robust enhanced 402 metastasis formation in the experimental metastasis studies, it indicates that acid export can 403 facilitate tumor cell extravasation out of the blood vessels and colonization of metastatic sites. 404 In our CA-IX model buffer therapy significantly reduced tumor burden in the lungs 405 compared to their untreated counterparts. Although this did not completely prevent metastasis, 406 combinations of buffer therapy with specific acid exporter inhibitors may be necessary. CA-IX is 407 minimally expressed in normal tissue and could be a viable therapeutic target (Silvia Pastorekova 408 et al., 1997) and other proton pump inhibitors are currently in clinical trials (Fais, 2015). Many 409 studies have hinted at the importance of acidity, and many have proposed reasons as to why 410 cancer cells favor aerobic glycolysis, but few have proposed that acidity is the driver. This study 411 represents the first to test whether acid export can increase aerobic glycolysis and enhance cancer 412 aggressiveness, rather than acid merely being a by-product. 413

Construction of stable cell lines 415
Plasmids. Yeast plasma membrane ATPase 1 (PMA1) cDNA construct was designed based on 416 the sequence (Accession Number: NM_001180873; Saccharomyces cerevisiae S288c PMA1). 417 The codons were optimized for the suitable expression in mammalian cells and restriction 418 enzyme sequences Hind III and Xho I were inserted at the 5' and 3' ends of the full-length 419 sequence, respectively. The fully designed DNA sequence was commercially synthesized (Blue 420 Heron clones were verified using western blotting. Cell lines were tested for mycoplasma using 445 MycoAlert assay (Lonza). 446

Oxygen consumption and proton production rate measurements (OCR and PPR) 499
Real-time oxygen consumption (OCR) and proton production rate (PPR) were measured 500 by using the Seahorse Extracellular Flux (XFe-96) Analyzer (Seahorse Bioscience, Chicopee, 501 MA). The cells were seeded in an XFe-96 microplate (Seahorse, V3-PET, 101104-004) in 502 normal growth media overnight. The growth media were replaced with DMEM powder base 503 media ( Sigma D5030) supplemented with 1.85g/L sodium chloride and 1mM glutamine, and the 504 cells were incubated in the media in the absence of glucose, when testing glycolysis, in a non-505 CO2 incubator for one hour prior to the measurement. PPR and OCR were measured in the 506 absence of glucose associated with the non-glycolytic activity, followed by two sequential 507 injections of D-glucose (6mM) and oligomycin (1µM) in real-time, which are associated with 508 glycolytic activity and glycolytic capacity (reserve). The mitochondrial stress test was also used 509 where cells were incubated in glucose (5.5mM), and glutamine (1mM) containing media and 510 basal OCR and PPR measured, prior to sequential injection of Oligomycin (1 µM), associated 511 with ATP linked OCR, FCCP(1µM) associated with mitochondrial reserve capacity and 512 Rotenone/Antimycin A (1µM). Following the measurements, protein concentrations were 513 determined in situ for each well using a standard BCA protein assay (Thermo Scientific Pierce). 514 The OCR and PPR values were normalized to µg protein. Results were also normalized using 515 Celigo High Throughput Micro-Well Imaging Cytometer (Nexcelom Bioscience) by bright-field 516 direct cell counting and normalized per 10K cells prior to assay. 517 Glucose consumption and lactate production assays 518 reached 90% confluence, the growth media were removed, and the cells were washed twice in 530 PBS and incubated in serum-free and phenol-red free media for 24 h. The media were collected 531 from 24 h incubation for both glucose consumption and lactate production assays. The cells were 532 trypsinized and the cell densities were determined. Glucose quantification was conducted using 533 glucose colorimetric/fluorometric assay kit (BioVision, K606-100) as described per 534 manufacturer instruction. The lactate assay kit II (BioVision, K627-100) was used to measure L 535 (+)-Lactate in the culture media according to the manufacturer's instructions. Data were 536 normalized by cell density per well and were reported as lactate production and glucose 537 consumption as pmol per cell. 538 Glucose uptake radioactive assay 539 Cells were seeded in 24 well plates to 80% confluence. Cells were incubated for 1hr with 540 1 µCi of Deoxy-D-glucose, positive pixels. Fluorescence at 580 and 640 nm was averaged, background offset and ratioed for 581 each particle representing a cell. 582

Cell invasion and migration assay in vitro 583
In vitro cell motility and invasiveness were measured by methods as previously reported 584 with some modifications 30. The motility change was measured by the circular wound healing 585 assay using OrisTM Cell Migration Assay Kit (Platypus, CMAU101). Cells were plated on a 96-586 well plate at 1 x 10 6 cells/ml while a cell seeding stopper masker the circular area at the center of 587 each well. The cell seeding stoppers were removed 24 hours after the plating and cells were 588 cultured a further 30 hours to monitor the closing of the cell-free area (wound area Bioscience, MA). After 7 days of culture, the cell expansion from the droplets was quantified by 605 Celigo single colony verification algorithm or Image J after fixing cells in 3.7 % formaldehyde 606 and staining in crystal violet solution. The larger area occupied by cells represents the higher 607 invasion potential. 608 Normalization of invasion and migration assays. The results of invasion and migration assays 609 were normalized by the proliferation rates of the cells. Proliferation rates were calculated by a 610 linear fit of cell growth, 48 hours after seeding the cells (See Figure S16)  For the spontaneous metastasis studies, approximately 10×10 6 cells (either MCF-641 7/MOCK-1 and MCF-7/PMA1-C1) in 100 µl of PBS +100 µl of Matrigel were injected into the 642 mammary fat pads of mice. Once tumors reached approximately 400 mm 3 , or 6 weeks post-cell 643 injection, the tumors were resected, fixed, and stained with H&E, or PMA1 antibody. Three 644 months after resection, the mice were sacrificed, and lung sections were examined for lung 645 metastases. The tumors were measured twice every week throughout the study with a digital 646 caliper and volume values were calculated with the formula V= (Length×Width 2 )/2. The body 647 weights were monitored twice a week throughout the study. 648 presence of stromal over-growth and mitosis, necrosis, spindle cells differential and chromatin 670 activity. Tumor burden in the lung was measured by Aperio ImageScope (Leica Biosystems, IL) 671 and calculated as % of total lung tissue and compared between groups. Tumors were drawn 672 around by hand using the Aperio software and confirmed by a pathologist (M.M.B), % area of 673 tumor in lungs was then calculated by comparing the area of total lung tissue to the area of the 674 tumor within lungs. For all lung metastases IHC, three different sections were taken from each 675 lung and analyzed, with 5-6 sections taken between each analyzed slice to ensure entirety of 676 lungs, and metastatic burden was analyzed. 677 678

RNA analyses in formalin-fixed, paraffin-embedded (FFPE) tissue 693
FFPE tissue samples were cut in 10µm-thick sections on a microtome, and 694 deparaffinized by deparaffinization solution (Qiagen, 19093). Total RNA was extracted from 695 deparaffinized FFPE sections with the miRNeasy FFPE kit (Qiagen, 217504) following the 696 manufacturer's protocol. Real-time qRT-PCR analyses for PMA1 mRNA were described above 697 in the qRT-PCR section. Experimental Ct values from PMA1 amplification were normalized with 698 GAPDH Ct values and were expressed relative to MCF-7/MOCK-1 control Ct values. 699

Statistical analyses 700
A two-tailed unpaired student T-test or Welch's T-test was employed to determine 701 statistical significance. Ordinary one-way ANOVA with Geisser Greenhouse correction and 702 Tukey's Multiple Comparison test, with a single pooled variance. A p-value of less than 0.05 was 703 considered statistically significant or otherwise indicated. Kaplan Meier Curve was used to 704 analyze overall survival in mouse models with Log-Rank Test curve comparison. 705

Data and code availability 706
The authors declare that the data supporting the findings of this study are available within 707 the paper and its Supplementary information files. The code is available in the Supplementary 708 data. Figures with raw data associated include Fig. 5H, associated raw data are found in 709 Supplementary figures S19. 710