Identification of distinct slow mode of reversible adaptation of pancreatic ductal adenocarcinoma to the prolonged acidic pH microenvironment

List of antibodies was provided in Supplementary Table S1. Unless otherwise specified, all reagents and chemicals were purchased from either Thermo Fisher Scientific (Carlsbad, CA, USA) or Sigma-Aldrich by Merck (Darmstadt, Germany).

Human pancreatic ductal adenocarcinoma SUIT-2 and BxPC-3 cell lines were freshly acquired from JCRB (Japanese Collection of Research Bioresources, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka, Japan) and BCRC (Bioresource Collection and Research Center, Hsinchu, Taiwan) cell banks, respectively. The STR (short tandem repeat) profiles of both cell lines were validated by FIRDI (Food Industry Research and Development Institute, Hsinchu, Taiwan), and the certificates of authentication can be provided upon request. Cells were cultured as monolayers in RPMI-1640 medium (Cat. #31800-022, Gibco by Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 23.8 mM NaHCO₃ and 10% (v/v) FBS at 37 °C in a humidified atmosphere of 5% CO₂ in air, and were routinely examined for mycoplasma contamination using the EZ-PCR mycoplasma detection kit (Biological Industries, Kibbutz Beit HaEmek, Israel). To assess the impact of acidic pHe on PDAC cells, the bicarbonated RPMI-1640 medium was further supplemented with 25 mM HEPES/PIPES (bioWORLD by GeneLinx, Dublin, OH, USA) followed by the addition of appropriate amount of 1 N HCl to either pHe 7.4 for use as the buffer control medium, or to pHe 6.7 for use as the acidic culture medium as per the guidelines recommended by Michl et al. [29]. Each cell sample was adapted to the intended pH by adding the corresponding pH-adjusted medium to freshly split cells, followed by intensive daily monitoring (e.g., before and after each experiment or when cultivation took place in the humidified CO₂ incubator) to ensure the desired medium pH level to be accurate within 0.1 units throughout experiments. To achieve statistical power, all cell samples subjected to different periods of acidic exposures were prepared in 3 separate experiments.

Cell viability was determined by trypan blue (Invitrogen by Thermo Fisher Scientific, Eugene, OR, USA) dye exclusion assay and expressed as the percentage of living cells counted. To measure changes in cellular proliferation, the MTT (thiazolyl blue tetrazolium bromide) assay was performed. The blue formazan crystals were solubilized with DMSO, and the colored solution was read at 570 nm using a SpectraMax 250 microplate reader (Marshall Scientific, Hampton, NH, USA).

The oxygen consumption rate (OCR) and the extracellular acidification rate (ECAR) of test cell samples were assessed using a Seahorse XFe24 Extracellular Flux Analyzer (Agilent, Santa Clara, CA, USA) as per the guideline provided by the manufacture. Briefly, cells were seeded in the XFe24 microplates and cultured in conditioned medium for 16 h (37 °C under 5% CO₂ and 95% humidity) to allow attachment and growth. Prior to the start of Seahorse assays, the optimal FCCP (carbonyl cyanide-4(trifluoromethoxy)phenylhydrazone) concentrations were determined by titration studies. For the OCR measurements, cells were washed in pre-warmed medium (2% FBS in RPMI-1640 without NaHCO₃, pH adjusted to 7.4) and equilibrated at 37 °C for 1 h before the assay. For the ECAR measurements, cells were washed in pre-warmed glucose-free RPMI-1640 medium (Cat. R1383, Sigma-Aldrich by Merck) supplemented with 2% (v/v) FBS without NaHCO₃, pH adjusted to 7.4 and equilibrated at 37 °C for 30 min before the experiments. Data were normalized by cell number, and graphs were calculated and plotted using Agilent Seahorse Wave Desktop software.

To visualize cellular morphological changes, all cells cultured in their corresponding pHe medium were collected and plated in 35-mm diameter μ-dishes (Ibidi, Martinsried, Germany). A minimum of 10 randomly chosen microscopic fields (~ 70 cells per field) were imaged using an AF6000 LX microscope (Leica Microsystems, Singapore) equipped with Zyla sCMOS camera (Andor by Oxford Instruments, Belfast, UK) and objective lens (HCX PL Fluotar L 20x/0.40 CORR PH1) with 1.8x digital zoom. To observe mitochondrial morphology and cristae ultrastructure, the TEM (transmission electron microscopy) images of each cell group were taken on an H-7650 electron microscope (Hitachi High-Technologies, Tokyo, Japan) equipped with a CCD (charge-coupled device) camera (Advanced Microscopy Techniques, Danvers, MA, USA). The density of cristae was calculated based on 45 randomly selected mitochondria per cell group with ImageJ software (https://​imagej.​nih.​gov/​ij).

To detect the changes of autophagy, approximately 2 × 10⁴ cells obtained from each sample were incubated with 200 nM LysoTracker Red DND-99 for lysosome visualization, followed by labeling with 1x CYTO-ID Green dye (Enzo Life Sciences, Farmingdale, NY, USA) for tracking autophagic vacuoles, and with 1 μg/mL Hoechst 33342 for nucleus staining. After incubation, the stained cells were subjected to image analysis by a TCS SP5 confocal microscope (Leica Microsystems). For ROS measurement, cells were stained with 5 μM CellROX Green dye for oxidative stress detection and then counterstained with 1 μg/mL Hoechst 33342 for confocal imaging. To determine mitochondrial superoxide production and membrane potential (∆ψ mt), cells were respectively stained with 5 μM MitoSOX Red and 200 nM TMRE (tetramethylrhodamine ethyl ester) as per the manufacturer’s instructions. To observe mitochondrial dynamics, cells were incubated with 200 nM MitoTracker Orange CMTMRos, followed by labeling with 2 drops/mL ActinGreen 488 reagent for tracking F-actin and mounted with ProLong Diamond mounting medium containing DAPI for confocal imaging. The fluorescence intensity profiles were plotted using LAS X software (Leica Microsystems). The super-resolution imaging of mitochondria was carried out using a Nikon N-SIM system equipped with an Apochromat 60×/1.27 numerical aperture water-immersion objective len (Nikon Instruments, Tokyo, Japan). Raw SIM images were obtained and reconstructed using Nikon Elements software. The lengths of mitochondria were analyzed by Imaris 8.0 software (Bitplane, Badenerstrasse, Zurich, Switzerland). For the assessment of filopodia assembly, cells were immunostained with anti-VASP antibody followed by Alexa Fluor 488-conjugated secondary antibody. Actin filaments were revealed by Alexa Fluor 594 phalloidin (Invitrogen) staining. The images were taken with the TCS SP5 confocal microscope, followed by fluorescence signal quantification with MetaMorph software (Molecular Devices, San Jose, CA, USA).

To characterize cell cycle phase, the collected cells were stained with propidium iodide solution and subjected to a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA, USA). The proportions of cells in different phases of the cell cycle were calculated using CellQuest software (BD Biosciences). The levels of cellular ROS, mitochondrial superoxide, and ∆ψ mt were determined by FACSCalibur and CellQuest using the CellROX Green, MitoSOX Red and TMRE staining, respectively.

Cell migration was assessed using time-lapse microscopy as previously reported [30] with slight modifications. In brief, approximately 2 × 10⁴ cells in each study group were seeded in 6-well plates in their corresponding pHe medium. On the next day, cells were placed on a Leica AF6000 LX microscope and their movement was recorded for 3 h at 20 min intervals at 37 °C. Individual cells were tracked for motility analysis with Image-Pro Plus software (Media Cybernetics, Rockville, MD, USA). To determine cell invasive capability, cells were stained with 1 μM CellTracker Deep Red dye (Invitrogen), and then seeded in 35-mm glass bottom culture dishes (MatTek, Ashland, MA, USA) precoated with Matrigel Matrix (Corning, Tewksbury, MA, USA) supplemented with 25 μg/mL DQ-collagen IV (Invitrogen). The dishes were left for 16-18 h before nuclear staining with 1 μg/mL Hoechst 33342. Images of live-cell invasion were photographed by the TCS SP5 confocal microscope followed by quantitative analysis using Imaris 8.0 software.

To generate HA-tagged constructs, the HA (YPYDVPDYA) epitope-tag was cloned into the pGW1-CMV expression vector (Addgene, Watertown, MA, USA) as a positive control to assess transfection efficiency as previously described [31]. The construction of two sub-clones, pGW1-HA-Drp1^S637A and pGW1-HA-Drp1^S637D, was engineered as detailed elsewhere [32]. Cells were transfected with each of the recombinant plasmids using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s guidelines. For gene-silencing studies, non-targeting control siRNA and sequence-specific SMARTpool siRNA against Mfn2 (see Supplementary Table S2) were purchased from Dharmacon by Horizon Discovery (Cambridge, UK). Transfections were performed using Lipofectamine RNAiMAX transfection reagent (Invitrogen) as per manufacturer’s protocol.

For whole-cell protein extraction, cells were harvested and lysed in 1x RIPA lysis buffer (Merck Millipore) containing PhosSTOP phosphatase inhibitors and cOmplete protease inhibitor cocktail (Sigma-Aldrich by Merck). Protein concentrations were determined by the Pierce BCA assay kit (Thermo Fisher Scientific) using BSA as standard. Equal protein quantities of cell lysates (20-50 μg) were electrophoresed on a range of 10-15% (w/v) SDS-PAGE depending on the size of the target proteins, and then processed for Western blotting with corresponding primary antibodies followed by incubation with HRP-conjugated secondary antibodies. Chemiluminescence signals were visualized using an ECL Western Detection System (Merck Millipore) as recommended by the manufacturer. Unless otherwise specified, β-actin served as loading control in all Western blot experiments.

To investigate changes in OPA1 oligomerization correlated with altered cristae ultrastructure, cells were crosslinked with 1 mM bismaleimidohexane for 20 min at 37 °C. After crosslinking, cells were quenched and washed with PBS supplemented with 0.1% β-mercaptoethanol, and were then lysed in RIPA buffer for gradient gel electrophoresis using EVOgel (GeneDireX, Taoyuan, Taiwan). OPA1 oligomerization was analyzed by Western blotting using an anti-OPA1 antibody.

Total RNA was extracted using RNeasy mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s guidelines. The quality of RNA was ascertained both by gel electrophoresis and Agilent 2100 Bioanalyzer. Microarray analysis was conducted on the high-resolution Human Transcriptome Array 2.0 platform (Affymetrix, Santa Clara, CA, USA) by the Core Instrument Center at National Health Research Institutes, Taiwan. Raw intensities were normalized using SST-RMA (signal space transformation-robust multi-chip analysis) algorithm. Fold-change values were analyzed by Partek Genomics Suite statistical software (Partek, St. Louis, MO, USA). Gene expression profiles were subjected to GSEA (Gene Set Enrichment Analysis, https://​www.​gsea-msigdb.​org) and IPA (Ingenuity Pathway Analysis, Qiagen) functional enrichment analysis for a variety of gene sets including the pre-defined gene sets, the hallmark gene sets and the representative MSigDB (Molecular Signatures Database) signatures (see Supplementary Table S3). The relative gene expressions of leading edge subsets were viewed as heatmaps by MeV software (https://​webmev.​tm4.​org). The analyzed gene sets were limited to those that contained between 15 and 500 genes. The permutation was conducted 1000 times with random combinations according to default-weighted enrichment statistics, along with the use of a signal-to-noise metric to rank genes based on their differential expression levels across different cell groups.

Gene expression data and relevant clinical information were obtained from The Cancer Genome Atlas pancreatic cancer dataset (TCGA_PAAD cohort) downloaded from GDC data portal (https://​portal.​gdc.​cancer.​gov). Primary pancreatic tumor samples with available overall survival data (n = 176) were selected for analysis. The Kaplan-Meier curve was used to estimate the effects of gene on the overall survival of patients. The patients were classified into low and high expression groups based on the median-centered gene expression values. The difference between low or high expression groups was assessed by log-rank test.

Data are presented as the means ± SD from at least three individual experiments, unless specified otherwise. Results were analyzed by GraphPad Prism statistical software (GraphPad Software, San Diego, CA, USA) using unpaired Student’s t-test. A p value less than 0.05 was interpreted as statistically significant for all comparisons.

To investigate the short- and long-term effects of extracellular acidification on pancreatic tumor cells, two representative PDAC cell lines (SUIT-2 & BxPC-3) were subjected to different periods of acidic pHe stress. Here, SUIT-2 was utilized as the primary cell line for the majority of studies whereas BxPC-3 served as the second cell line to confirm the key findings and to rule out cell line-specific effects. As illustrated in Fig. 1a, five groups in three replicates per cell line were established including: 1) control group (Ctrl) in which PDAC cells were grown in control medium at physiological pHe 7.4; 2) buffered group (Buff) in which PDAC cells were incubated in control medium supplemented with non-volatile HEPES/PIPES buffer at pHe 7.4; and 3-5) three acid-treated groups in which PDAC cells were cultured in control medium buffered with HEPES/PIPES at pHe 6.7 for a short period of 2-3 weeks to reflect the initial stage of extracellular acidification (denoted hereafter as S.A. for short-term acidic exposure); a medium period of 4-5 months to mirror the reported median TVDT in pancreatic cancer patients [22, 23] (denoted hereafter as M.A. for mid-term acidic exposure); or a long period of 10-12 months to simulate the TVDT in patients with early primary lesions [24‐26] (denoted hereafter as L.A. for long-term acidic exposure). The pHe 6.7 represents the modest level of extracellular acidity frequently detected in different types of solid tumors including PDAC [12, 16, 33‐37]. Each culture medium was carefully prepared as recommended by Michl et al. [29] with the desired pHe value stably maintained within ±0.1 units to ensure quality control and data reliability. Consistent with previous findings [38], we found no substantial effect of salinity on test cells from salt ions present in the mildly acidic medium (data not shown).

Upon initial exposure to acidotic pHe stress, the proliferation of SUIT-2 PDAC cells was severely impaired as compared to those grown at pHe 7.4 (see S.A. vs Ctrl or Buff, Fig. 1b). The diminished cell proliferation was slightly restored after some generations of acidic exposure (see M.A. in Fig. 1b). When cultivated for a longer period up to ~ 1 year, SUIT-2 tumor cells gradually recovered and proliferated at a rate closer, but not yet equal, to those maintained at pHe 7.4 (see L.A. vs Ctrl or Buff, Fig. 1b). No apparent reduction in cell viability was detected during the course of the experiment (Fig. 1c). Cell cycle analysis revealed a disproportionate increased percentage of G1 phase and a decreased percentage of S phase in PDAC cells under initial acidic stress (see S.A. vs Ctrl or Buff, Fig. 1d), which indicate a cell-cycle arrest at G1 phase likely induced by acute extracellular acidosis for subsequent cell growth inhibition. This notion was further corroborated by Western blotting: the protein levels of G1-phase regulators (e.g., CDK4, CDK6, Cyclin D1, & Cyclin E) were down-regulated and some CDK inhibitors such as p21^Kip1 were up-regulated under early acidic stress condition (see S.A. in Fig. 1e). After a prolonged acidic exposure, most of the altered expressions in G1/S cell cycle transition regulators were reversed to normal levels, and the high proportion of G1-phase cells was reduced to near, but not equal, to those of the controls at pHe 7.4 (see L.A. vs Ctrl or Buff, Fig. 1d-e). Overall, these results point to a slow recovery of the proliferation rate of viable PDAC cells to external acidotic stress stimuli, most likely by progressively overcoming G1 arrest signals to promote cell cycle progression in a time-dependent manner.

To further assess the differences between short-term response and long-term adaptation of solid tumor cells to extracellular acidification, we examined the phenotypic changes of SUIT-2 tumor cells under various periods of acidic pHe challenge. Upon initial acidotic stress exposure, SUIT-2 cells promptly lost their cobblestone appearance with cell-cell contacts and instead induced a spindle phenotype with a concomitant accumulation of autophagic vacuoles (see S.A. vs Ctrl or Buff, Fig. 1f-g). This early phenotypic stress response is similar to previous findings on other types of solid tumor cells [17, 18] that low pHe induced an EMT-like process associated with an activation of autophagy to sustain cell viability and growth. Quantification of protein expression revealed an overexpression of LC3B II, Beclin 1, and ATG7 markers with a lack of p62 accumulation (see S.A. in Fig. 1h), which confirm the presence of autophagy induction with an enhanced autolysosomal degradation upon acute acidic stress stimuli. Interestingly, SUIT-2 cells reverted to their former phenotype when gradually adapted to the prolonged extracellular acidity (see L.A. in Fig. 1f), and the early acidosis-induced autophagic vacuoles and autolysosomes were nearly completely abrogated along with decreased LC3B II and Beclin 1 expression (Fig. 1g-h). This gradual decline in autophagy over time is different from previous reports claiming that the acidosis-induced cell autophagy was constantly active and not reversed [12, 13]. Further western blot analyses showed that, similar to other tumor types studied [39‐42], both NF-κB signaling and c-IAP2 expression were significantly higher in the S.A. state than in the control states, thus exerting an apoptotic resistance especially under acute acidic stress (Fig. 1i). Since low environmental pH can be viewed as a metabolic stressor and that metabolic changes are often intertwined with the dynamic transition between epithelial and mesenchymal phenotypes [11, 43], our data suggest that tumor cells under acute acidotic pressure rapidly activated EMT and autophagy as early pro-survival responses to overcome metabolic stress and to resist apoptosis. They later developed alternative metabolic strategies to enable their adaptations to a constant and prolonged acidic pHe stress.

Cellular energy status is an important factor for cell cycle progression, which may be compromised by environmental or metabolic stress that either interfere with ATP production or accelerate ATP consumption [44]. To evaluate the energy metabolic status of SUIT-2 cells under various periods of acidic pHe stress, we performed Seahorse extracellular flux assays to determine and compare the oxygen consumption rate (OCR) and the extracellular acidification rate (ECAR) between five different PDAC cell states. Upon early exposure to extracellular acidosis, PDAC cells displayed a substantial decrease in basal respiration, ATP turnover, mitochondrial maximal respiration, and spare respiratory capacity (see S.A. vs Ctrl or Buff, Fig. 1j-k). When SUIT-2 cells were acclimatized to the persistent acidic stress conditions, the initially reduced metabolic parameters caused by acute acidosis were shown to be progressively restored (see L.A. in Fig. 1j-k). Real-time ECAR measurements further revealed a similar recovering trend of basal glycolysis and maximal glycolytic capacity after SUIT-2 cells chronically adapted to the prolonged acidic stress stimuli (Fig. 1l-m). Together these results suggest that early acidotic pHe stress caused severe mitochondrial damage and ATP shortage. Nevertheless, after being cultivated for a sufficiently long period of time, PDAC cells progressively evolved the ability to reinvigorate mitochondrial activity and to repair dysfunctional metabolism to generate the energy needed for survival and proliferation under an acidic pHe microenvironment.

Mitochondria are highly dynamic organelles that can sense stress signals and rapidly remodel their structure and network [45‐47]. Here, we present evidence of an unusual mode of mitochondrial network dynamics highlighted by an early rapid response and a later slow adaption to constant acidotic pHe stress. Super resolution 3D N-SIM microscopy showed the mitochondria of control SUIT-2 tumor cells at pHe 7.4 as speckled and spherical nucleus-surrounding clusters (see Ctrl and Buff in Fig. 2a-b & Videos S1-S2). When exposed to acute extracellular acidosis, however, the mitochondria of SUIT-2 cells rapidly fused into distinct filamentous reticular networks termed as SIMH (stress-induced mitochondrial hyperfusion, [48]) and spread extensively throughout the cytoplasm (see S.A. in Fig. 2a-b & Video S3).

After a prolonged period of acidic pHe exposure, the hyperfused mitochondrial network underwent a slow progressive ultrastructural transformation from the interconnected reticular SIMH to the punctate fragmented disconnected morphology scattered over the cytoplasm (see M.A. and L.A. in Fig. 2a-b & Videos S4-S5). This characteristic reorganization of mitochondrial architecture was further confirmed by TEM. The prominent elongated mitochondria with abnormal cristae were promptly induced by acute acidotic stress (see S.A. in Fig. 2c), but were slowly restructured into a nearly normal configuration after SUIT-2 cells steadily habituated to the acidic pHe microenvironment (see L.A. in Fig. 2c). The acid-induced cristae modulation was illustrated by ImageJ software analysis showing a significant decrease in normal cristae and a large increase in distorted cristae per surface area of mitochondria (Fig. 2d-f). Under extended acidic conditions, the distorted cristae ultrastructures were progressively reversed to a near normal status as those seen at pHe 7.4 (Fig. 2d-f). The full-range mitochondrial length distributions of five SUIT-2 cell states were analyzed by Imaris software. The data showed a mitochondrial network hyperfusion with a wide range of mitochondrial lengths as an early responsive machinery (see S.A. in Fig. 2g). This was followed by a mitochondrial network fragmentation with shorter mitochondrial lengths as a late adaptative mechanism to the prolonged acidic pHe pressure (see L.A. in Fig. 2g).

The dynamism and health of the mitochondrial network is delicately regulated by fusion/fission events [45‐47]. Figure 2h shows that there is a significantly elevated expression of inactive form p-DRP1^Ser637 detected exclusively in the S.A. cell state in addition to an increase of DRP1 total protein in three acid-treated cell states. The increased phosphorylation at serine 637 of fission protein DRP1 suppressed its translocation to the mitochondria [46, 47], which enhanced mitochondrial fusion and prevented mitochondria from autophagic degradation (Fig. 2a-b & Video S3). Notably, after an extended exposure to the same acidotic pHe stress, the expression of p-DRP1^Ser637 was reversed to a level similar as control cells at pHe 7.4 (Fig. 2h), thus leading to an activation of mitochondrial fission to trigger mitochondrial fragmentation (Fig. 2a-b & Video S5). A simultaneous up-regulation of mitochondrial pro-fusion protein MFN2 was also seen in the S.A. cell state, which enables more fusion activity by tethering adjacent mitochondria together under early acidic pHe pressure (Fig. 2h). Furthermore, the tightness of mitochondrial cristae junctions has been reported to be regulated by the oligomerization of long and short forms of cristae-remodeling protein OPA1 [46, 47]. Hence, the decrease of OPA1 oligomerization detected in the S.A. cell state points to a marked loss of normal cristae structures; in the L.A. cell group many distorted cristae were steadily restored to normal and functional configurations (Fig. 2c-f & i).

Oxidative stress refers to elevated intracellular levels of reactive oxygen species (ROS). It is a common issue for cancer cells under external stimuli because of its ability to initiate cell death [46]. Considering that we did not observe any substantial reduction in cell viability during the mild acidic stress exposure (Fig. 1c), we attempted to assess and compare the status of oxidative stress among five different SUIT-2 cell states. Herein, oxidative stress was measured based on the levels of cellular ROS, mitochondrial ROS, and mitochondrial membrane potential using CellROX Green, MitoSOX Red, and TMRE fluorescent probes, respectively. Figure 2j & m shows that no apparent difference was found for the production of overall cellular ROS in SUIT-2 cells under normal or acidic pHe conditions. Compared to control cells at pHe 7.4, however, there was a statistically significant inhibition of both the MitoSOX-measured superoxide and the TMRE-measured membrane potential in PDAC cells under acute acidotic stress (see S.A. in Fig. 2k, l, n, & o). These data strongly suggest that PDAC cells rapidly reduced oxidative phosphorylation and lowered mitochondrial ROS production in response to acute extracellular acidosis.

Increased ROS produced by damaged mitochondria can trigger mitophagy—a crucial mechanism for regulating mitochondrial quality and cellular homeostasis especially under stress [45]. SUIT-2 PDAC cells could induce massive SIMH to prevent mitochondria from mitophagic degradation by steric obstruction (Fig. 2a-h & Video S3). Thus, we manipulated DRP1 fission and MFN2 fusion activities to disintegrate mitochondrial hyperfusion and promote selective mitophagy that ultimately leads to the disruption of tumor cell adaptation to extracellular acidosis. As presented in Fig. 3a, we individually overexpressed the control HA-tagged protein, the phosphodeficient DRP1^S637A, or the phosphomimetic DRP1^S637D variant [31] in three different cell states (Ctrl, S.A., & L.A.). Despite the modest cytotoxicity to the infected cells, we found that overexpression of the DRP1^S637D variant had no statistical difference in terms of MTT proliferation assay versus those overexpressing the HA-tagged protein regardless of culture in physiological pHe 7.4 or mild acidic pHe 6.7 (Fig. 3a). In contrast, overexpression of the DRP1^S637A mutant in the S.A. cell group resulted in more severe cell death than the Ctrl and L.A. cell groups (Fig. 3a). A lower amount of phosphorylated DRP1^S637 was also detected in the S.A. cell group overexpressing phosphodeficient DRP1^S637A. This synergistically enhanced DRP1 fission activity fragmented the hyperfused mitochondria followed by selective mitophagy (e.g., differential expressions of LC3B II, BNIP3, & BNIP3L/NIX) and apoptotic induction (e.g., elevated levels of BAK1, BAX, & Cleaved Caspase 3) (see S.A. vs Ctrl in Fig. 3a). Notably, overexpression of the same DRP1^S637A variant in the L.A. cell state did not induce mitophagy or apoptosis probably due to the presence of restored fragmented and relatively healthy mitochondria in long-term acid-adapted PDAC tumor cells.

To obtain more convincing evidence on the impact of perturbing dynamic mitochondrial networks on tumor cell responsiveness to acidic pHe stress, we used a more efficient and less cytotoxic approach to transfect the same three SUIT-2 cell states (Ctrl, S.A., & L.A.) with either negative control non-targeting siRNAs (siRNA-N.C.) or siRNAs against MFN2 (siRNA-MFN2). Unlike the control siRNA-treated cells, knockdown of MFN2 expression with a parallel decrease of phosphorylated DRP1^Ser637 in the S.A. cell group led to a reduced mitochondrial fusion and an enhanced mitochondrial fission that collectively enabled selective mitophagy (e.g., up-regulation of LC3B II & BNIP3) followed by cell apoptosis (e.g., increased levels of BAK1, BAX, & Cleaved Caspase 3) (see S.A. in Fig. 3b). In contrast, there was no significant inhibitory effect of siRNA-MFN2 on cell proliferation nor was there impaired viability in the L.A. cell group (see L.A. in Fig. 3b). These data highlight the critical role of mitochondrial network dynamics for the systematic response and adaptation of PDAC cells to the external acidic stimuli. A disruption in this dynamic process severely hampered the capability of pancreatic tumor cells to deal with the acidified microenvironment. This subsequently caused an inadequate response or intolerance to the stress of extracellular acidosis especially under acute exposure situation. However, PDAC cells develop a near normal active metabolic state via a slow mode of mitochondrial network reorganization after a constant and prolonged acidic exposure. At this long-term acid-adapted L.A. cell state, the same strategy of manipulating fission/fusion proteins to fragment the already fragmented mitochondria may not be as effective as it was to the hyperfused mitochondria present in the early acidity-responsive S.A. cell state.

Accumulating evidence links mitochondrial subcellular distribution to tumor cell morphology, proliferation, and metastatic potential [49]. Figure 4a shows that many mitochondria in SUIT-2 PDAC cells under acidic pHe stress moved from near the nucleus to the cell periphery via anchoring to actin filaments. The elevated fluorescence intensity of MitoTracker (red) along the arrows (white) indicated an anterograde movement of mitochondria toward the plasma membrane (Fig. 4b). Statistical analysis of the quantification of the mitochondrial moving distances also validated such phenomenon (Fig. 4c-e). Interestingly, when SUIT-2 cells acclimated to the constant stress of extracellular acidity, many mitochondria were fragmented from reticular SIMH (Fig. 2a-b) but still remained at the leading edge of the cell periphery (Fig, 4a-e). This scene was most likely to support ATP-consuming activities for cell membrane dynamics, migration, and invasion.

Tumor cell migration and invasion were recently proposed to be regulated by mitochondrial trafficking [49]. By performing time-lapse imaging to track the migration of individual SUIT-2 cells exposed to different periods of extracellular acidity, we found that solid tumor cells under acute acidic pHe are more prone to adopt to faster migration than those at control pHe 7.4 (see S.A. vs Ctrl or Buff, Fig. 4f). This migration ability remained robust when PDAC cells further adapted to the prolonged acidic pHe stress (see M.A. & L.A. in Fig. 4f). Statistical analysis of the migration distance and velocity confirmed that SUIT-2 cells under pHe 6.7 had a higher motility than those under pHe 7.4 (Fig. 4g-h). Confocal images of endogenous VASP—a vasodilator-stimulated phosphoprotein essential for filopodia formation—revealed that it was more localized in a line along the tips of the protruding membrane. This presentation suggested an augmented filopodia formation induced by acidotic stress (see Fig. 4i-j). This increased filopodia formation was supported by high levels of actin nucleation and polymerization markers including WAVE2, Profilin 1, phosphorylated VASP^Ser157, and RAC1/CDC42^Ser71 (see L.A. in Fig. 4k). Remarkably, E-cadherin was almost completely abolished during acidic exposure whereas a drastically enhanced N-cadherin expression was detected when SUIT-2 cells progressively adapted to the constant acidic pHe conditions (Fig. 4l). Other transcription factors and mesenchymal markers like SNAIL, Fibronectin 1, and Vimentin also had a differential expression pattern similar to that of N-cadherin, indicating a persistent EMT with epithelial cobblestone phenotype (Fig. 1f) likely associated with cellular adaptation to the prolonged acidic pHe exposure (Fig. 4l). Extracellular matrix glycoproteins such as Laminin-γ2 and proteolytic enzymes such as MMP1, CTSS & CTSB were radically up-regulated under acute acidotic stress, but all become less expressed when tumor cells steadily adapted to extended extracellular acidity (Fig. 4l). Some metastasis-related factors including ARHGAP29, DSG2, and ST3GAL1 were up-regulated after an extended exposure to acidic pHe (Fig. 4l). Further quantification of protein expression of autophagy, pro-survival, mitochondrial dynamics, and metastasis-related markers showed that many of these critical regulators were differentially expressed within the first few days of acute exposures to extracellular acidosis (Fig. S1). Together these findings demonstrate that PDAC cells rapidly increased mitochondrial motility under acidic stress. They had a more aggressive potential even though their proliferation rate was strongly reduced and most mitochondria were in a hyperfused state associated with ATP insufficiency. By further incubating SUIT-2 cells in 3D overlay culture, we observed an extensive pericellular degradation of DQ-collagen IV under acidic pHe conditions especially at the L.A. cell state (Fig. 4m-n). These data indicate that PDAC tumor cells became more malignant with increased migratory and invasive potentials when adapted to continuous long-term extracellular acidity.

BxPC-3 cells were used as the second PDAC cell line to validate the aforementioned observations of the short-term response and long-term adaptation of SUIT-2 cells to the acidic pHe microenvironment. As shown in the S.A. state of Fig. S2, upon initial exposure to acidotic stress, BxPC-3 cells also exhibited impaired proliferation (Fig. S2a), increased G1 phase arrest (Fig. S2b), induced EMT phenotype (Fig. S2c), activated cytoprotective autophagy (Fig. S2d), diminished metabolic capacity (Fig. S2e-S2h), stimulated mitochondrial hyperfusion (Fig. S2i-S2j) and enhanced invasive property (Fig. S2k) that are similar to those observed in the S.A. state of SUIT-2 cells. Consistent with the primary SUIT-2 cell line, after BxPC-3 cells progressively habituated to the prolonged acidic pHe exposure, many of the altered phenomena were reversed to close to those of the controls at pHe 7.4 except with more aggressive potentials with increased invasiveness (see L.A. in Fig. S2a-S2k). In should however be noted that, while the majority of differential protein expression patterns were similar between the two cell lines, the autophagy-related markers in BxPC-3 cells appear to be less susceptible to induction by acute extracellular acidity (see S.A. vs Ctrl or Buff, Fig. S2l). These observations are possibly due to the higher basal levels of endogenous autophagy in parental cells (see Ctrl in Fig. S2l), so that when exposed to acidic pHe the degree of induced activation of protective autophagy to mitigate acidotic stress to sustain cell survival is less drastic in BxPC-3 cells than that in SUIT-2 cells.

The maintenance of stable medium pH is critical for data reproducibility and reliability. However, as recently stated by Michl and co-workers [29], there are still no consensus guidelines for best practices in controlling pH in live-cell culture systems. In this study, we prepared and constantly monitored incubation media pH as per recommendations by Michl et al. [29] and carried out principal component analysis (PCA) for both the untreated and acid-treated PDAC cell groups to determine the key factors involved in adaptation to extracellular acidosis as well as the potential buffer effects from acidic pHe cultures. Figure S3a shows that the PCA plot displayed a high degree of concordance between the sample duplicates with two relatively distinctive subpopulations of data points representing a substantial difference in gene expression between PDAC cells at physiological pHe 7.4 and mild acidic pHe 6.7. Further statistical analysis on the Ctrl and Buff cell groups revealed that out of 67,528 gene probes analyzed, 99.48 and 99.06% of probes remained unchanged in expression in PDAC cells under short- or long-term buffering conditions, respectively (Fig. S3b). Even after nearly 1 full year of continuous cell culture passage, only 0.39% of probes were up-regulated (fold change ≥2); 0.55% were down-regulated (fold change ≤0.5) to statistical significance. These results—together with the very slight differences in cell phenotype, proliferation, cell cycle phase, and energy metabolic status in between the control and buffer-treated PDAC cell populations (see Ctrl vs Buff, Figs. 1b, d, j-m, 2a-g, & j-o)—indicate that the diverse short-term responses and long-term adaptations of tumor cells to external stimuli do not result from the effect of the buffering regime but are rather due to the acidic pHe stress. Hence, our work strongly supports the guidelines of Michl et al. [29] for proper management of medium pH, which would be advantageous for researchers in the field to minimize untoward pH fluctuations and erroneous inferences for assessing the impact of extracellular acidification on test cells regardless of the short- or long-term culture.

We next compared the global gene expression profiles with public expression databases and determined several critical gene signatures linked to PDAC cell adaptation to the acidic pHe environment. There were four functional hallmarks (invasion_of_tumor_cell_lines; advanced_malignant_tumor; cell_movement_of_tumor_cell_lines; and epithelial_mesenchymal_transition) and two canonical pathways (adherens_junctions_interactions and LKB1_pathway) identified from the comparisons between each of Buff, S.A., and L.A. cell populations (Fig. 5a-b). This analysis suggested that the activation states of the four functional classes are selectively enhanced along with extended periods of tumor cell exposure to acidic conditions whereas the LKB1_pathway is primarily involved in metabolism and growth control [50] during early extracellular acidosis. The adherens_junctions_interactions pathway is predominantly suppressed when PDAC cells adapt to the long-term acidic pressure. These hallmark signatures agree with those shown in Figs. 1 & 2 in that the proliferation and energetic metabolism of PDAC cells were largely restricted when first exposed to extracellular acidosis, but became progressively more aggressive and malignant after long-term adaptation to the acidic pHe stress (see Fig. 4). Further heatmap analysis revealed that the top differentially expressed genes belonged to the leading-edge subsets of the indicated functional categories (Fig. 5c-h), suggesting these genes may be involved in the early response or late adaptation of PDAC cells to extracellular acidosis. Kaplan-Meier survival curves on these leading-edge subsets disclosed a total of thirteen acid-adaptation up-regulated genes (F3, COL12A, RARG, LMO7, VANGL1, ITGB1, MYC, NOTCH2, SERPINE1, EREG, DPYD, PLOD2, & CALU) and six down-regulated genes (STRADA, TSC2, SIK3, CADM1, RPTOR, & MLST8) whose expression levels are respectively negatively and positively correlated with the overall survival of pancreatic cancer patients in the TCGA-PAAD cohort (Fig. 5i).

To highlight the molecular targets that are specifically involved in the long-term adaptation of PDAC tumor cells to the acidic pHe microenvironment, we analyzed and compared mRNA differential expression profiles between L.A. and S.A. cell states. Two distinct gene sets were created and annotated for functional inference: the chronic_acidosis_adaptation up-regulated signature (see Table S4) and the chronic_acidosis_adaptation down-regulated signature (see Table S5). Their expression levels are differentially impacted by long-term cell adaptation to extracellular acidosis. The differential expression patterns of the defined gene sets were visualized as a heatmap in Fig. 6a with a total of fourteen up-regulated genes (CLDN16, CDK14, CTTNBP2NL, LOX, PRRG4, ZPLD1, ZBTB38, NRAS, LINC00707, PRR16, FOSL2, OSMR, ST3GAL1, & MIR4668). One down-regulated gene (STMN3) was further identified and significantly associated with the overall survival of pancreatic cancer patients (Fig. 6b). These fifteen genes, together with those nineteen genes identified from GSEA analysis in Fig. 5, may represent the central molecular targets to help clarify how solid tumor cells such as PDAC respond and adapt to the enduring acidotic pHe stress and then proceed to more aggressive and metastatic stages.