Collective invasion induced by an autocrine purinergic loop through connexin-43 hemichannels

Abstract

Progression of epithelial cancers predominantly proceeds by collective invasion of cell groups with coordinated cell-cell junctions and multicellular cytoskeletal activity. Collectively invading breast cancer cells express the gap junction protein connexin-43 (Cx43), yet whether Cx43 regulates collective invasion remains unclear. We here show that Cx43 mediates gap-junctional coupling between collectively invading breast cancer cells and, via hemichannels, adenosine nucleotide/nucleoside release into the extracellular space. Using molecular interference and rescue strategies, we identify that Cx43 hemichannel function, but not intercellular communication, induces leader cell activity and collective migration through the engagement of the adenosine receptor 1 (ADORA1) and AKT signaling. Accordingly, pharmacological inhibition of ADORA1 or AKT signaling caused leader cell collapse and halted collective invasion. ADORA1 inhibition further reduced local invasion of orthotopic mammary tumors in vivo, and joint up-regulation of Cx43 and ADORA1 in breast cancer patients correlated with decreased relapse-free survival. This identifies autocrine purinergic signaling, through Cx43 hemichannels, as a critical pathway in leader cell function and collective invasion.

Graphical abstract

Figure S1.

Expression of Cx43 in breast cancer cells during collective cancer cell invasion. (a) Original colored fluorescence overview of breast cancer lesions with zoomed insets on normal ducts and tumor invasive margins. (b) Trainable cell phenotyping function of the Inform software to identify different cell types within a multicellular tissue based on spectral fluorescence fingerprinting. Images display the results of the per-cell phenotyping analysis in normal ducts (luminal epithelial cells [LE]) and collective invasion (CI) patterns in fibrous and in adipose tissue. In each region, different cell types are marked by colored dots, as indicated. For all regions devoid of bilayered ducts, cells of epithelial origin, based on pan-cytokeratin, were considered as invasive breast cancer cells. (c) Unmixed composite images of Cx43 intensity in epithelial cancer cells within the tumor margins of representative samples with negative (upper panel) and mid-level (lower panel) Cx43 expression. (d) Heterogeneous expression of Cx43 within cancer cells of the collective invasion patterns in different tumor regions and samples. Values represent the percentage of Cx43 positive cancer cells. (e) Summary of Kaplan–Meier survival analysis correlating RFS with Cx43 expression in breast cancer patients categorized based on grade or molecular subtype. Bold and underlined values indicate positive (black) or inverse (green) association with Cx43 expression. P values, log-rank test. (f) Quantitative PCR for nine connexin subtypes relevant for the mammary gland (Cx26, Cx30, Cx30.1, Cx32, Cx36, Cx37, Cx40, Cx43, and Cx45). Connexin mRNA expression in 4T1 (black) and MMT (gray) cells from 3D collagen invasion cultures. Values represent mean normalized mRNA levels and SEM from three independent experiments. (g) Cx45 localization in collectively invading strands in 3D collagen. Confocal images represent maximum intensity projections from 3D confocal stacks. Arrowheads, cytoplasmic protrusions. Scale bars: 100 µm (c), 50 µm (c, inset), 20 µm (g).

Figure 1.

Increased expression of Cx43 in breast cancer cells during collective cancer cell invasion in vivo and in vitro. (a) Multispectral microscopy of breast cancer tissue sections stained for Cx43 (green), pan-cytokeratin (red; epithelial cells), vimentin (white; stromal cells, cancer cells after epithelial-to-mesenchymal transition). The three panels represent example slices for normal ducts and the invasion zones in the collagen-rich fibrous or adipose tissue. Detection of Cx43 in myoepithelial cells surrounding the luminal epithelium and epithelial cancer cells along cell–cell junctions. Arrows and arrowheads depict Cx43 at cell–cell contacts with puncta or linear patterns, respectively. (b) Colocalization analysis of Cx43 versus pan-cytokeratin expression and classified myoepithelial cells, luminal epithelial cells, and invading cancer cells in fibrous or adipose tissue. Cx43 and pan-cytokeratin levels were acquired using supervised automated tissue segmentation software (Inform). (c) Cx43 levels in luminal epithelium cells (LEC) and cancer cells within the fibrous or adipose tissue from 13 clinical samples (see Table 1 for sample details). Dotted line represents the threshold level for Cx43 negativity, determined by receiving operator characteristic (ROC) analysis with intensity values obtained from the luminal epithelial cells as control. (d) Kaplan–Meier survival plot predicting RFS for high versus low Cx43 expression in basal-type breast cancer patients (Györffy et al., 2010). P values, log-rank test. (e and f) Distribution of Cx43 along cell–cell junctions (white arrowheads) and at the polar extensions (red arrowheads) of leader cells during collective invasion of 4T1 (e) and MMT (f) spheroids in 3D collagen. (g) Intensity distribution of Cx43 and F-actin along the circumference of a leader cell (LC; e, white dashed border). Values show the pixel intensity with background subtraction. Confocal images are displayed as maximum-intensity projections from a 3D confocal stack. (h) Cx43 expression and distribution during collective invasion of 4T1 cells in the stroma of the mouse mammary gland, 4 d after implantation and monitored by confocal microscopy. Scale bars: 100 µm (a, e, and f, overview), 50 µm (a, inset), 20 µm (e and f, spheroid 2; h), 10 µm (e, f, and h, inset). Cyt. Ext., cytoplasmic extension. Norm, normal.

Figure S2.

Collectively invading cancer cells maintain Cx43-mediated GJIC and extracellular nucleotide release. (a) β-Catenin localization along cell–cell junctions between both leader cells and follower cells in collective strands of 4T1 and MMT cells invading 3D fibrillar collagen. Arrowheads, cell–cell junctions. (b) Intensity distribution of β-catenin and F-actin along the circumference of a leader cell (white dashed border). Values show the pixel intensity with background subtraction. (c) Multicellular 3D gap FRAP. Single confocal slices of calcein-labeled invasive 4T1 strands before and after photobleaching. Contours indicate regions of interest for bleaching, including follower regions (red: R1; blue: R2), an invasive front (green: R3), and a single detached cell (SC; purple), in addition to the unbleached neighboring regions D1 and D2. (d) Normalized mean calcein fluorescence intensity of the selected regions. (e) Western blot for Cx43 and β-tubulin as loading control. Whole-cell lysates extracted from 4T1 and MMT cells with stable expression of shNT or shCx43; ±SD. (f and g) Fluorescence recovery of calcein-labeled MMT leader cells expressing shNT or shCx43 during invasion in 3D collagen (f) and physically connected cells in 2D culture (g). (h) Values represent the means and SEM of at least 19 bleached cells from 2D culture per condition from three independent experiments. P values, two-tailed unpaired Mann–Whitney test. (i) Relative dye transfer in parachute assay of 4T1 with nontargeting shRNA (NT) and Cx43 shRNA (Cx43) after 6 or 12 h of cell–cell (donor–receiver) encounter. Values represent the means and SEM from two independent experiments. (j) Summary of the effect of Cx43 down-regulation and/or CBX treatment on the reduction of individual nucleotide and nucleosides concentration detected in the media of 4T1 and MMT collagen cultures by HPLC (related to Fig. 2 k). Concentrations from CBX-treated and Cx43 shRNA cultures were compared with control cells (Ctrl, untreated cells with nontargeting shRNA). P values, ANOVA with Dunnett’s multiple comparison test (ns, not significant). Scale bars: 20 µm (a, c, f, and g), 10 µm (a, inset).

Figure 2.

Cx43 mediates GJIC and extracellular release of nucleotides/nucleosides during collective invasion. (a) Workflow of time-resolved 3D gap FRAP, including calcein labeling of multicellular spheroids in 3D collagen cultures and 3D gap-FRAP procedure of leader and follower cells within a 3D invasion strand by asymmetric region of interest selection and photobleaching, followed by recording the change in fluorescence intensity over time. (b) Single confocal slices of calcein-labeled invasive 4T1 strands before and after photobleaching. Dashed contours represent the bleached leader cell (LC; red), FC (blue), and SC (purple) in control media and LC in the presence of CBX (white). (c) Normalized calcein fluorescence intensity in bleached LCs, FCs, and SCs in the presence or absence of CBX. Values represent normalized mean fluorescence intensities with SEM; 11–16 cells per treatment condition from four independent experiments. (d) Effect of CBX on percentage fluorescence recovery after photobleaching of LCs and FCs. Values are represented as the means and SEM of three independent experiments. P values, two-tailed unpaired Mann–Whitney test. (e) Average fluorescence recovery in cells stably expressing control vector (shNT) or Cx43 shRNA (shCx43) during invasion in 3D collagen. Values represent the normalized mean intensities and SEM of four leader cells for shNT and pooled two leader cells and two follower cells for shCx43 condition. (f) Inhibition of fluorescence recovery of MMT cells after Cx43 down-regulation (cell groups in Petri dish culture). Normalized mean intensity and SEM of 19–22 cells from three independent experiments. P value, two-tailed unpaired Mann–Whitney test. (g and h) Parachute assay to measure de novo junction formation (g) followed by dye transfer into nonlabeled cells (h) and example histograms obtained by flow cytometry showing the time-dependent alterations of calcein label in donor and recipient cells. (i) Intensity of calcein in 4T1 receiver cells after 12 h of incubation with 4T1 donor cells in control conditions and during inhibition with CBX or after Cx43 down-regulation. Median (red line) intensities from four independent experiments. P values, Kruskal-Wallis test with Dunn’s multiple comparison test. (j and k) HPLC analysis of purines released from 4T1 and MMT spheroids into the supernatant after 24 h of invasion in 3D collagen. Values represent average concentrations of each purine (j) and the relative change of the total purine concentrations after treatment with CBX and/or Cx43 down-regulation in 4T1 and MMT cells (k). Mean values and SEM from four (j) or SD from three (k) independent experiments. P values, ANOVA Dunnett’s multiple comparison test. Scale bar: 20 µm (b). Neg ctrl, negative control; Norm., normalized; FC, follower cell; LC, leader cell; SC, single detached cell.

Figure S3.

Cx43 dependence of leader cell function. (a and b) Resazurin cytotoxicity assay for 4T1 and MMT cells in 3D collagen for 24 h in the presence of vehicle control and escalating doses of CBX (a) and 18aGA (b). Values represent the mean relative fluorescence unit (RFU) with SEM of two or three independent experiments. P values, one-way ANOVA with Bonferroni multiple comparison test. (c) Bright-field images after 24 h of MMT cell invasion in 3D collagen in the presence of CBX or an inactive structurally homologous control compound (GLZ). Data represent the medians (black line), 25th/75th percentiles (box) and maximum/minimum (whiskers) of 8–11 MMT spheroids per condition pooled from three independent experiments. P value, two-tailed unpaired Mann–Whitney test. (d) Efficacy of collective invasion in 4T1 spheroids in the presence of escalating doses of 18aGA. Bright-field images after 24 h of culture. Arrowheads (c and d), tips of invasion strands.(e) 4T1 leader cell (LC) initiation in the presence of escalating doses of CBX. Data represent mean values and SEM from 23 to 40 spheroids per condition pooled from at least three independent experiments. P value, ANOVA with Bonferroni multiple comparison test. (f) Speed of MMT leader cells during collective invasion in 3D collagen. Values represent median (black line), 25th/75th percentiles (box) and maximum/minimum (whiskers) of seven or eight leader cells from one experiment. P value, two-tailed unpaired Mann–Whitney test. (g) Cumulative collective invasion of MMT cells. Data represent the medians (black line), 25th/75th percentiles (box) and maximum/minimum (whiskers) of 8–11 MMT spheroids per condition pooled from three independent experiments. P value, two-tailed unpaired Mann–Whitney test. (h) Cumulative collective invasion of 4T1 and MMT spheroids in the presence of escalating doses of 18aGA. Data represent mean values and SEM from 12 spheroids per condition pooled from three independent experiments. P value, ANOVA with Bonferroni multiple comparison test. Scale bars: 100 µm (c and d).

Figure 3.

Cx43 dependence of leader cell function. (a) Bright-field images of 4T1 spheroids after 24 h of invasion in 3D collagen in the presence of increasing concentration of CBX. Arrowheads, tips of invasion strands. (b) Mean cumulative length of collective invasion strands in response to CBX. Mean values and SEM from three independent experiments each comprising seven spheroids/condition. P values, Kruskal–Wallis test with Dunn’s multiple comparison test. (c) Maximum intensity projection from a confocal 3D stack, showing F-actin and DAPI in control and CBX-treated spheroids (100 µM). (d–g) Cx43-dependent collective invasion. Bright-field images (d), cumulative collective invasion (e and f), and leader cell initiation (g) of spheroids composed of 4T1 or MMT cells stably expressing Cx43 or nontargeting shRNA (NT). Values represent the medians (black line), 25th/75th percentiles (box) and maximum/minimum (whiskers) from three or four independent experiments with three to eight spheroids each. P values, two-tailed unpaired Mann–Whitney test. (h) Bright-field images of 4T1 spheroids in response to transient Cx43 down-regulation using two different Cx43 RNAi probes. (i) Number of leader cells (LCs) per spheroid. P values, Kruskal–Wallis test with Dunn’s multiple comparison test. (j) Maximum intensity projection from a confocal 3D stack showing mosaic spheroids composed of fluorescence-coded MMT cells stably expressing NT or Cx43 shRNA as indicated (1:1 ratio); arrowheads depict red or green leader cells. (k) Frequency of green or red leader cells. Bars represent the mean values and SEM pooled from one (j1) and three (j2) independent experiments. (l) Effect of CBX on the kinetics of established invasion strands. Mean length per invasion strand after addition of CBX at 0 h (intervention) and washout at 24 h (rescue). Data show the mean values and SEM from six to eight invasion strands from one spheroid per condition. Scale bars: 100 µm (a), 50 µm (c, d, and h), 25 µm (c, inset). LC, leader cell.

Figure 4.

Cx43 hemichannel function in leader cell induction by purine nucleotide release. (a–c) Rescue of invasion in 4T1 and MMT spheroids in 3D collagen in the presence of CBX and escalating doses of ADO; for comparison with untreated baseline levels refer to Figs. 3 b and S3 (c and e). Representative bright-field images (a), cumulative collective invasion (b), and number of leader cells (LC) per spheroid (c). Mean values and SEM from four independent experiments, each comprising 5–11 spheroids/condition. P values, Kruskal–Wallis test with Dunn’s multiple comparison test. (d and e) Rescue of leader cell deficiency after Cx43 down-regulation by exogenous ADO. (d) Maximum intensity projection from a 3D confocal stack showing mosaic spheroids with MMT cells stably expressing NT or Cx43 shRNA (1:1 ratio) in the absence or presence of ADO and (e) frequency of LC with green color expressing Cx43 shRNA. Bars represent the means and SEM from three independent experiments, each comprising three to six spheroids. P values, two-tailed unpaired Mann–Whitney test. Scale bars: 50 µm (a), 100 µm (d).

Figure S4.

Cx43-mediated nucleotide release in LC initiation. (a) Leader cell (LC) initiation in MMT spheroids expressing shCx43 after 24 h of invasion in collagen in the presence and absence of exogenous ADO. Values display median (black line), 25th/75th percentiles (boxes) and maximum/minimum values (whiskers) of 30–35 spheroids per condition pooled from three independent experiments. P value, two-tailed unpaired Mann–Whitney test. (b) Purinergic receptor mRNA expression in 4T1 (black) and MMT (gray) cells during collective invasion in 3D collagen assessed by quantitative PCR. Values represent the mean normalized mRNA levels and SEM from three independent experiments. (c) MMT leader cell initiation in the presence of vehicle and escalating doses of PSB36. Values represent mean values and SEM of at least 20 spheroids per condition pooled from three independent experiments. P values, Kruskal–Wallis test with Dunn’s multiple comparison test. (d) Leader cell initiation in response to DCPX treatment 12–15 h of invasion of 4T1cells in collagen. Values display medians (black lines) 25th/75th percentiles (boxes) and maximum/minimum values (whiskers), 23–27 spheroids from three independent experiments. P values, two-tailed unpaired Mann–Whitney test. (e) Mitotic figures in 4T1 spheroid after 24 h of invasion in collagen. Values represent average number and SEM of mitotic figures per spheroids, three spheroids (≥1,000 cell/spheroid) per condition. P value, two-tailed unpaired Mann–Whitney test. (f) Leader cell initiation of 4T1 and MMT spheroids after 12–15 h of invasion in collagen in the presence of ADORA2b inhibitor PSB1115. Values display medians (black lines) 25th/75th percentiles (boxes) and maximum/minimum values (whiskers), 4–12 spheroids from one experiment. P values, Kruskal–Wallis test with Dunn’s multiple comparison test. (g) Expression and distribution of ADORA1 in 4T1 spheroids treated with vehicle control or CBX for 24 h. Images represent single confocal slices. (h) ADORA1 expression of the outer cellular layer. Data represent the background-corrected mean fluorescence and SD of four (control) or six (CBX) spheroids from one experiment. P value, two-tailed unpaired Mann–Whitney test. Scale bars: 50 µm (g), 25 µm (g, inset).

Figure 5.

Autocrine purinergic receptor signaling maintains leader cell functions. (a) Bright-field images of 4T1 spheroids cultured in 3D collagen for 24 h in the presence of escalating concentrations of PSB36. (b) Median numbers of leader cells (LC) per spheroid (red line) from three independent experiments. P values, Kruskal–Wallis test with Dunn’s multiple comparison test. (c) Maximum intensity projection from a 3D confocal stack of 4T1 spheroid in the presence or absence of PSB36. Arrowheads, leading extensions of individual LCs. (d) Maximum intensity projections of 3D MMT spheroids with different treatments and (e) resulting number of leader cells per spheroid. Data represent the medians (black line), 25th/75th percentiles (boxes) and maximum/minimum values (whiskers) from two independent experiments each comprising 10 to 16 spheroids/condition. P values, ANOVA with Bonferroni multiple comparison test. Scale bars: 50 µm (a and d), 20 µm (c), 25 µm (d, inset), 10 µm (c, inset).

Figure S5.

Hemichannel-ADORA1 loop in other collective but not single-cell invasion models and downstream AKT activation. (a) Distribution of Cx43, β-catenin, and F-actin in NMuMG spheroids during collective invasion in 3D collagen after 68 h. Images represent maximum intensity projections from a 3D confocal stacks. Arrowheads, focal Cx43 hemichannel localization in leader cells. (b) Bright-field images of NMuMG spheroids embedded in 3D collagen for 68 h in the presence or absence of CBX. Arrowheads, protrusive leader cells. (c) Cumulative median length of invasion per spheroid (black line), 25th/75th percentiles (box) and maximum/minimum (whiskers) from 42 to 75 stands from six (control) and eight (CBX) spheroids (one experiment). P values, two-tailed unpaired Mann–Whitney test. (d) Bright-field images of NMuMG spheroids in 3D collagen after 48 h culture in the presence of PSB36 (5 or 20 µM). (e) Mean number of leader cells per spheroid and SEM from seven to nine spheroids per condition (one experiment). P values, Kruskal–Wallis test with Dunn’s multiple comparison test. (f) HPLC analysis of purines released from individualized 4T1 cells into the supernatant after 24 h of invasion in 3D collagen. Values represent concentrations of each purine from one experiment. (g) Speed of 4T1 leader vs individual cells in 3D collagen. Values represent the medians (black line), 25th/75th percentiles (box) and maximum/minimum (whiskers) of six leader cells (one experiment) and 19 individual cells tracked over 24 h (two independent experiments). P value, two-tailed unpaired Mann–Whitney test. (h and i) Percentage of individualized 4T1 cells with elongated morphology after 12, 24, and 40 h in collagen and treatment with CBX (h) and PSB36 (i). Data show the means and SEM from 17–33 cells per condition from two (h) or one (i) experiments. DPCPX, 8-cyclopentyl-1,3-dipropylxanthine. (j) Distribution of AKT, phospho-AKT473, and F-actin in MMT spheroid invasion culture. 3D confocal maximum intensity projections after 24 h of culture. Dashed line indicates the region of analysis in k. (k) Distribution of AKT and phospho-AKT473 and F-actin along extending protrusions of leader cells. Data represent the mean intensities and SEM of F-actin, AKT, and phospho-AKT473 from at least 12 leader cells from one experiment. (l) Phospho-AKT473 levels in response to 24 h of PSB36 treatment during 4T1 spheroid culture in 3D collagen. (m) Phospho-AKT473 levels in of 4T1 and MMT 2D cultures after ADO treatment for 0.5, 1, 3, 6, and 12 h. Numbers in (l, m) represent the ratio of phospho-AKT473 over total AKT mean intensities from one experiment. Scale bars: 50 µm (a, b, d, and j), 25 µm (a and j, insert).

Figure 6.

AKT regulation and requirement for leader cell function and invasion. (a) Maximum intensity projection of phospho-AKT473 and F-actin staining in MMT spheroids treated with vehicle control or CBX. Maximum intensity projection of 3D confocal stack. (b) Distribution of phospho-AKT473 along protrusions of leader cells in both conditions with dot plot comparing phospho-AKT473 levels in the last 10 µm of leader cell protrusions in vehicle- and CBX-treated condition. Line graphs and dot plot represent intensities of phospho-AKT473 from 11 (control) or 21 (CBX) leader cells from three or four spheroids per respective condition. P values, two-tailed unpaired Mann–Whitney test. (c) Western blot for phospho-AKT473, AKT and β-tubulin as loading control. Whole-cell lysates extracted from MMT spheroids after 24 h of invasion into 3D collagen in the presence of 0, 75, and 100 µM of CBX. (d) Values represent normalized intensities of phospho-AKT473/total AKT with median values (red line) from two (75 µM) or three independent experiments (0 and 100 µM). P value, Kruskal–Wallis test with Dunn’s multiple comparison test. (e) Bright-field images of 4T1 and MMT spheroids embedded in 3D collagen for 15 h in the presence of DMSO or AKT inhibitor IV. (f) Number of leader cells (LC) per spheroid; red line represents the median value from 18 to 25 spheroids (4T1) and 16–19 spheroids (MMT) pooled from two independent experiments. P values, two-tailed unpaired Mann–Whitney test. Scale bars: 50 µm (a), 25 µm (a, inset), 100 µm (e).

Figure 7.

Pharmacological interference with ADORA1 inhibits collective invasion in vivo. (a) Invasion of 4T1 dual-color cells (H2B/mCherry; LifeAct/YFP) from microtumors implanted into the mouse mammary fat pad and imaged by multiphoton microscopy through a mammary window at days 0, 2, and 4 after implantation. Mice were treated with DMSO (vehicle) or 30 mg/kg PSB36 via intraperitoneal injection. Arrowhead, tip of collective invasion strand. (b and c) Analysis of invasion as the average area covered by the lesion normalized to the number of nuclei (b) and the tumor growth expressed as number of cells per lesion (c). Data in b and c represent the medians (black line), 25th/75th percentiles (boxes), and maximum/minimum values (whiskers) from at least nine implanted tumors from four independent experiments. P values, two-tailed unpaired Mann–Whitney test. (d) Detection of ADORA1 in fibrous or adipose tissue invasion zones of breast cancer samples using multispectral microscopy. (e) Relative ADORA1 levels in epithelial cancer cells quantified from eight independent lesions (see Table 1 for sample characteristics). Dotted line represents the threshold level for ADORA1 negativity, determined by ROC analysis with intensity values obtained from the luminal epithelial cells as control. Scale bars: 100 µm (a and d), 50 µm (a and d, inset).

Figure S6.

Role of ADORA1 in preclinical spontaneous metastasis and prognostic relevance in clinical samples. (a) Detection of ADORA1 in normal ducts within clinical samples using multispectral microscopy. (b) Frequency of ADORA1-positive cancer cells with collective invasion patterns in breast lesions. Data represent the fibrous and adipose tissue from individual patients. (c) Overviews of lung sections isolated from DMSO and PSB36 (30 mg/kg) treated mice, stained for cytokeratin-8. Insets and arrowheads depict macro- and micrometastases. (d) Average number of micro- and macrometastases normalized per lung section. Data represent a total of 135–145 lung sections from three independent mice per group. (e) Relative frequency of micro- or macrometastases per treatment group: DMSO: 39 micrometastases, 59 macrometastases, 98 total; PSB36: 39 micrometastases, 91 macrometastases, 130 in total. Data show the mean percentages and SD from three mice. P value, two-tailed unpaired Mann–Whitney test. (f and g) Kaplan–Meier survival analysis predicting RFS in (f) basal-type breast cancer patients. (g) Correlation of ADORA1 or combined ADORA1 and Cx43 expression in clinical samples. Classification was based on tumor grade and molecular subtype (Györffy et al., 2010). Bold and underlined values show the positive (black) or inverse (green) association with ADORA1/Cx43 expression. (f and g) P values, log-rank test. Scale bars: 100 µm (a), 50 µm (a and c, inset), 1,000 µm (c, overview).