Connexin-43 channels are a pathway for discharging lactate from glycolytic pancreatic ductal adenocarcinoma cells

Abstract

Glycolytic cancer cells produce large quantities of lactate that must be removed to sustain metabolism in the absence of oxidative phosphorylation. The only venting mechanism described to do this at an adequate rate is H⁺-coupled lactate efflux on monocarboxylate transporters (MCTs). Outward MCT activity is, however, thermodynamically inhibited by extracellular acidity, a hallmark of solid tumours. This inhibition would feedback unfavourably on metabolism and growth, raising the possibility that other venting mechanisms become important in under-perfused tumours. We investigated connexin-assembled gap junctions as an alternative route for discharging lactate from pancreatic ductal adenocarcinoma (PDAC) cells. Diffusive coupling (calcein transmission) in vitro was strong between Colo357 cells, weaker yet hypoxia-inducible between BxPC3 cells, and very low between MiaPaCa2 cells. Coupling correlated with levels of connexin-43 (Cx43), a protein previously linked to late-stage disease. Evoked lactate dynamics, imaged in Colo357 spheroids using cytoplasmic pH as a read-out, indicated that lactate anions permeate gap junctions faster than highly-buffered H⁺ ions. At steady-state, junctional transmission of lactate (a chemical base) from the spheroid core had an alkalinizing effect on the rim, producing therein a milieu conducive for growth. Metabolite assays demonstrated that Cx43 knockdown increased cytoplasmic lactate retention in Colo357 spheroids (diameter ~150 μm). MiaPaCa2 cells, which are Cx43 negative in monolayer culture, showed markedly increased Cx43 immunoreactivity at areas of invasion in orthotopic xenograft mouse models. These tissue areas were associated with chronic extracellular acidosis (as indicated by the marker LAMP2 near/at the plasmalemma), which can explain the advantage of junctional transmission over MCT in vivo. We propose that Cx43 channels are important conduits for dissipating lactate anions from glycolytic PDAC cells. Furthermore, lactate entry into the better-perfused recipient cells has a favourable alkalinizing effect and supplies substrate for oxidative phosphorylation. Cx43 is thus a novel target for influencing metabolite handling in junctionally-coupled tumours.

Figure 1

Gap junctional connectivity in PDAC cells. (a) Western blot (one of three repeats) for MCT1 and MCT4; CAIX induction is used as a marker of hypoxic signalling. (b(i)) Microarray data (BioGPS dataset E-GEOD-21654) for expression of connexin-coding genes in PDAC cell lines (see Supplementary Information for list; note that Cx23, Cx25 and Cx30.2 were not determined). (ii) Expression re-plotted on logarithmic scale, highlighting data for BxPC3, Colo357 and MiaPaCa2 cells. (c,i) Western blot (one of four repeats) for Cx43, Cx45 and Cx26 under normoxic conditions, after treatment with 1 mm dimethyloxalylglycine (DMOG), and incubation in 2% O₂ (hypoxia). (ii) Fluorescence recovery after photobleaching (FRAP) protocol for measuring junctional calcein permeability. Specimen trace for Colo357 monolayer. Carbenoxolone (CBX; 100 μm) inhibited coupling. Mean±s.e.m. of 15 Colo357, 15 BxPC3 and 10 MiaPaCa2 cell clusters. Unpaired t-test *P<0.01 (control vs CBX), ^#P<0.01 (DMOG vs normoxia). (d(i)) shRNA knockdown of Cx43 (GJA1 gene) in Colo357 (lentiviral delivery) reduces Cx43 expression; two constructs (of four tested) with best knockdown efficiency are shown. Knockdown efficiency (% KD) determined densitometrically from the change in Cx43/actin ratio from three blots (mean±s.e.m.). (ii) Cx43 knockdown reduces cell-to-cell coupling assayed by FRAP. Note that enhanced green fluorescent protein (eGFP) signal associated with lentivirally-infected cells, is negligible (<10%) compared with calcein fluorescence and does not contribute to fluorescence recovery. Specimen time courses shown; histogram shows mean±s.e.m. (n=11, 8, 9); unpaired t-test *P<0.01.

Figure 2

Junctional transmission of lactate and H⁺ ions. (a(i)) A source of H⁺ ions drives passive exchange of protonated and unprotonated buffers. Cytoplasmic H⁺ diffusivity probes buffer mobility in a cell. (ii) Cell-to-cell H⁺ permeability probes buffer transmission through gap junctions. (b(i)) Measuring H⁺ diffusion in a Colo357 cell, superfused in Hepes-buffered solution containing 1 mm NVA, the caged H⁺-compound. H⁺-uncaging (shaded region) produces a H⁺ microdomain that spreads diffusively in cytoplasm. Time courses of [H⁺] in three regions of interest (indicated by arrows) best-fitted with diffusion equation. (ii) Mean±s.e.m. (N=12, 15, 10). (c(i)) Measuring junctional H⁺ permeability between Colo357 cells. H⁺ ions were uncaged from 0.5 mm NVA in a central cell and time courses of [H⁺] in the central and neighboring cells derive junctional H⁺ permeability. (ii) Mean±s.e.m. (N=10, 11, 12, 10, 14, 12, 18, 10, 12, 14, 10). Experiments repeated in presence of carbenoxolone (CBX; 100 μm) and on GJA1 knockdown cells. Unpaired t-test *P<0.01 (effect of CBX), ^#P<0.02 (effect of DMOG). (d(i-a)) Regional exposure of a monolayer to 40 mm lactate-containing Hepes-buffered solution using a double-barrelled micropipette. For illustrative purposes, one microstream was labelled with 30 μm fluorescein; (b) fluorescence of cSNARF1-loaded monolayer; (c) pH_i before and (d) during regional exposure to lactate. (ii) Rapid junctional lactate transmission dissipates syncytial [lactate]_i gradient and permits further MCT-dependent H⁺-lactate entry into cells under the lactate-containing microstream (shaded green). (iii) Steady-state intracellular pH across monolayer of wild-type Colo357 cells measured after 2 min of regional exposure to lactate, minus intracellular pH under resting conditions. Experiment repeated in presence of CBX (100 μm). Mean±s.e.m. (N=7 WT, 10 CBX). (iv) Measurements on Colo357 cells transduced with shRNA #1 (knockdown) or scrambled construct. Mean±s.e.m. (N=5 WT, 5 CBX). ANOVA *P<0.05 for effect of CBX and for the effect of knockdown.

Figure 3

Intracellular lactate diffuses ahead of H⁺ ions in Colo357 spheroids. (a) pH_i response in cSNARF1-loaded Colo357 monolayer to 40 mm lactate-containing solution (probes MCT activity) followed by 30 mm ammonium-containing solution (tests speed of solution change). Inset: lactate-evoked pH_i changes are not rate-limited by solution change. (b) Membrane lactate permeability (P_mct,lac) calculated from MCT activity. Gap junctional blocker CBX (100 μm) had partial inhibitory effect on P_mct,lac. Mean±s.e.m. (N=20, 19, 20, 19). (c,i) pH_i map through equatorial plane of cSNARF1-loaded Colo357 spheroid (radius 110 μm). (ii) pH_i dynamics in peripheral layer (11 μm-thick rim; black time course) and core (radius 11 μm; red time course). (iii) Two routes of lactate entry into cells at the spheroid core. (d) pH_i response, in presence of CO₂/HCO₃⁻, to (i) adding and (ii) removing lactate. Biphasic response at core is blocked by CBX (100 μm; paired experiment). Mean±s.e.m. (N=8). (e) pH_i response, in absence of CO₂/HCO₃⁻, to (i) adding and (ii) removing lactate; extracellular buffering was provided by 20 mm or 1 mm Hepes (paired experiments). Mean±s.e.m (N=7). MCT activity at the core is rate-limited by slow H⁺ supply attained under a low (1 mM) [Hepes] buffering regime; this exposes a higher junctional component of lactate handling at the spheroid core. (f) Measurements in 1 mm Hepes performed on size matched spheroids grown from Colo357 cells transduced with shRNA #1 or scrambled construct. Unpaired measurements. Radius: 125.4±2.3 μm, 127.7±2.4 μm. Mean±s.e.m (N=5, 4).

Figure 4

Junctional venting of lactate from the hypoxic core alkalinizes cells at the rim of Colo357 spheroids. (a) Steady-state pH_i across the equatorial plane of a Colo357 spheroid. (b) Radial pH_i gradient in Colo357 spheroids showing alkalinization at rim. Replacing glucose with galactose reduces degree of peripheral alkalinization. Unpaired t-test with Bonferroni correction, **P<0.01. Mean±s.e.m. (N=12 glucose, 8 galactose). (c) Peripheral alkalinization is smaller in Cx43-knockdown Colo357 spheroids (shRNA #1; Figure 1d) compared to scrambled controls. Mean±s.e.m. (N=14 shRNA #1, 12 scrambled). (d) CBX reduces peripheral alkalinization; the modest acidification at the core likely relates to the partial inhibitory off-target effect on MCT. Mean±s.e.m. (N=11 control, 8 CBX). (e) Mechanism by which junctional lactate transport alkalinizes recipient cells at the spheroid rim. Lactate and H⁺ ions are co-produced by hypoxic metabolism. HCO₃⁻ (for example, taken-up by a bicarbonate transporter, BT) titrates cytoplasmic H⁺ ions in hypoxic cells.

Figure 5

Intracellular lactate retention measured by biochemical assays. (a) Collection protocol: Colo357 cells were grown as monolayers for 96 h or as spheroids for 36–120 h. Between 30 and 100 spheroids were collected for measuring radius. Aliquots of medium and of cells were collected for biochemical assays; cells were lysed in 400 μl water and centrifuged to remove debris. (b) Lysates from monolayers grown in glucose- or galactose-containing medium were assayed for [lactate]/protein ratio. Lactate is produced only in the presence of glucose, and to comparable levels in wild-type and shRNA-transduced Colo357 cells (time-matched incubations). Mean±s.e.m. (six lysates each). (c) [lactate]_i/[lactate]_e ratio (a measure of intracellular lactate retention) in wild-type monolayers (six lysates) and spheroids as a function of radius (four batches each containing >1000 spheroids). Mean±s.e.m. Unpaired t-test: * denotes significant increase in lactate retention relative to monolayer (P<0.05). (d) Upper panel: intracellular lactate retention in Colo357 cells infected with scrambled construct or shRNA #1, grown as monolayers (six lysates each) or as spheroids for 3 or 5 days. Mean±s.e.m. (four batches each containing >1000 spheroids). Unpaired t-test: * denotes significant increase in lactate retention relative to monolayer (P<0.05). ^# denotes significant difference between scrambled and shRNA #1 for duration-matched experiments (P<0.01). Lower panel: radius of spheroid, Mean±s.d. (e) Schematic diagram illustrating the routes for venting lactate. (i) In the absence of junctional coupling, all lactate traffic is through MCTs. Efflux in hypoxic regions with poor diffusive coupling is curtailed by an unfavourable transmembrane gradient; this leads to intracellular lactate retention. (ii) With junctional coupling, lactate can dissipate between cells and access normoxic regions that have a more favourable transmembrane gradient for MCT-facilitated off-loading; this reduces intracellular lactate retention.

Figure 6

Spheroid co-cultures of Colo357 cells with different levels of Cx43. (a) Colo357 spheroid co-culture (1:1) of wild-type (non-fluorescent) cells mixed with eGFP-fluorescent GJA1-knockdown cells (shRNA construct #1). Upper panels: transmission image superimposed with fluorescence measured confocally across equatorial plane (pinhole=3 μm). Lower panels: fluorescence maps normalized to mean signal (blue pixels indicate below-average fluorescence; red pixels indicate above-average fluorescence), showing segregation by level of Cx43. Cx43-positive wild-type cells are more abundant in the peripheral rim. (b) Experiment repeated with the same imaging settings with shRNA construct #2, confirming the result with construct #1. (c) Experiment repeated with the same imaging settings with scrambled construct. Scrambled and wild-type cells mixed randomly, without clear radial segregation. (d) Analysis of eGFP fluorescence, normalised to spheroid-wide signal, as a function of radial depth: cells with high Cx43 levels segregate preferentially in the proliferating rim. Mean±s.e.m. (N=11 scrambled, 13 shRNA #1, 22 shRNA #2). Table shows results of ANOVA.

Figure 7

Cx43 immunoreactivity in MiaPaCa2 cells studied in orthotopic xenograft mouse models. Cx43 staining (left column) in sections of the primary tumour, invasion into adjacent tissues and sites of metastasis. Staining for human EGFR (middle column) identifies MiaPaCa2 cells. Overlay of Cx43 and EGFR immunoreactivity analysed by image-pair registration and colour filtering to identify staining (right column). (a) Cancer cells in the primary tumour remained Cx43 negative (overlay is blue, indicating EGFR signal only). (b) Example of a lung metastasis with absent Cx43 staining and (c) with more prominent Cx43 staining that co-localizes with an EGFR-positive cluster. (d) Clusters of Cx43-positive MiaPaCa2 cells (white arrows) within the host pancreas that co-localize with EGFR staining. MiaPaCa2 cell invasion into (e) mucosa of duodenum and (f) stomach muscularis, showing stronger and more uniform Cx43 staining (white arrows), but also a number of Cx43-negative clusters (black arrows). Clusters that are positive for both Cx43 and EGFR are indicated by purple colour on the overlay.

Figure 8

Distribution of LAMP2 in primary and invading MiaPaCa2 cells in vivo. (a) Diffuse LAMP2 staining in MiaPaCa2 cytoplasm but not at cell margins in sections of primary tumour. (b) In a subpopulation of MiaPaCa2 cells invading (i) duodenal mucosa and (ii) stomach muscularis, LAMP2 staining appeared more concentrated at cell margins, a marker of chronic extracellular acidosis.