Arachidonic acid-induced Ca²⁺ entry and migration in a neuroendocrine cancer cell line

Cyclopiazonic acid (CPA) was purchased from Calbiochem. Arachidonic acid was obtained from MP Biomedicals. Thapsigargin and SK&F 96365 (SKF) were purchased from Tocris Bioscience. Ketoprofen, ethylene glycol tetraacetic acid (EGTA) and 2-aminoethoxydiphenyl borate (2-APB) were obtained from Sigma. Leukotriene C₄ (LTC4) and Nordihydroguaiaretic acid (NDGA) were purchased from Cayman chemical.

BON cells were cultured in flasks with a 1:1 solution of DMEM and F12K supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin and maintained at 37 °C in a humidified incubator with 5% CO₂. All cell culture reagents were obtained from Life Technologies, unless specifically indicated. For cell transfection, 5 × 10⁶ BON cells were electroporated with plasmid vectors containing shRNAs against Orai1 or Orai3 (OriGene Technologies) using the Amaxa nucleofector II device (Lonza) as per manufacturer’s instruction. All available shRNA constructs supplied by the manufacturer resulted in comparable knockdowns for each subunit, and the data from these experiments were pooled. Transfection efficiency was estimated to be approximately 90% for both constructs. Control cells were transfected with an identical plasmid that contained a scrambled message. Moreover, mock-transfected and untransfected cells were used as additional controls. All plasmids expressed either GFP or RFP and a gene for puromycin resistance. Transfected cells were maintained 2 days in culture medium containing 0.2 μg/ml puromycin dihydrochloride. In addition to western blotting (see below), knockdown of Orai1 and 3 were functionally verified using live cell Ca²⁺ imaging at different time points and the maximum effects were observed at 48 h after transfection.

Lysates were prepared from BON cells at 48 h post-transfection by extraction of cellular proteins in RIPA buffer (containing 25 mM Tris–HCl, pH 7.6, 150 nM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS). The cell lysates were concentrated and desalted using the YM-100 Microcon centrifuge filter (Sigma Aldrich). Approximately, 50 μg of concentrated cell lysates were run on 8% Bis–Tris gels under denaturing conditions. Following separation, the protein bands were transferred onto polyvinylidene fluoride (PVDF) membranes for 90 min using the Criterion blotter (Biorad) under semi-wet conditions. The membranes were then incubated for 2 h in a solution of 5% non-fat dry milk in TBS-T buffer (containing in mM, 20 Tris/HCl, 150 NaCl and 0.1% Tween-20) to block non-specific interactions. The membranes were then cut in two and incubated overnight at 4°C with rabbit polyclonal antisera against Orai1 or Orai3 (Prosci). The membranes were washed with TBS-T buffer and probed for 2 h using an HRP-conjugated anti-rabbit secondary antibody (Millipore). Finally, the blots were washed with TBS-T buffer and treated for chemiluminescence visualization using the ECL Western Blotting substrate (Pierce). Detection of protein bands was performed using a LAS 3000 imaging system (Fujifilm). After matching the position of the bands of interests with respect to the molecular mass reference ladder (Invitrogen), the membranes were stripped using Restore solution, (Pierce) and re-probed with HRP-conjugated mouse monoclonal antibody against β-actin (Abcam) and visualized by chemiluminescence. Quantification of band density was achieved using NIH ImageJ software. The intensity of bands for Orai1 and Orai3 were comparable in untransfected, mock transfected and scrambled shRNA-transfected BON cells.

Changes in cytosolic Ca²⁺ levels in live cells were performed using live-cell fluorescent imaging as described previously [30]. Briefly, BON cells cultured on glass coverslips were loaded with a 2 μM fura-2 AM solution prepared in a physiological saline (containing in mM 140 NaCl, 5 KCl, 2.2 CaCl₂, 1 MgCl₂, 10 HEPES, and 5 glucose, pH 7.4) for 30–40 min at room temperature. Changes in intracellular Ca²⁺ were represented as the ratio of fura-2 fluorescence at 510 nm evoked by sequential excitation at 340 and 380 nm at a frequency of 1 Hz. Typically, for live cell imaging experiments changes in Ca²⁺ for 20–40 cells were independently monitored and analyzed. For each experiment, 2–3 coverslips were analyzed and averaged results from at least three independent experiments were used perform statistical analysis.

Rate of Ca²⁺ entry independent of Ca²⁺ release or buffering was estimated by monitoring the rate of quenching of the fura-2 fluorescence signal excited at 360 nm in response to Mn²⁺ entry as previously described [30]. The rate of Mn²⁺ quench of the fura-2 fluorescence in response to pharmacological stimulation was compared with the basal rate under the unstimulated condition. The fold change in the rates of quenching from basal values was determined.

BON cell migration was assessed as described previously using a modified Boyden chamber assay [31]. Briefly, 5.0 × 10⁴ BON cells were re-suspended in serum free media (SFM) and were seeded into the upper chamber, the bottom surfaces of which were coated with 50 μg/ml Type-1 rat-tail collagen (Corning Incorporated). The lower chamber was loaded with SFM containing AA with or without pharmacological inhibitors. In some experiments, the cells were pre-treated with 1 μM thapsigargin dissolved in Ca²⁺-free SFM for 15 min. In another experiment, the concentration of free Ca²⁺ was reduced to approximately 0.7 mM by adding EGTA to the media. The cells were allowed to migrate from the upper to lower chamber for 8–12 h. Following this time-period the number of migrated cells were counted and expressed as percentage of the total number seeded.

Immunofluorescence was used to identify various NMT markers in BON cells prior to or following arachidonic acid treatments. BON cells (5 × 10⁴ cells) were seeded onto glass coverslips or Boyden chamber inserts placed in a 6-well dish containing growth media. Following treatments, cells were fixed using a 4% paraformaldehyde solution and washed with phosphate buffered saline (PBS) (Life tech). Cells were then permeabilized using a 0.5% Triton X-100 solution (Sigma) for 5 min at and a blocking buffer containing 5% non-fat dry milk added for 1 h. Following PBS washes, samples were incubated overnight at 4 °C with the appropriate primary antibody in a humidified chamber. After PBS washes, glass coverslips were incubated with secondary antibodies conjugated to Alexa Fluor 488 or Alexa Fluor 546 for 1 h and mounted with Vectashield antifade medium (Vector Laboratories). F-actin was labeled using phallotoxin conjugated to Alexa Fluor 546 or 633. Images were obtained using a Leica SP5 Confocal microscope.

The results from the live cell imaging and Mn²⁺-quench, were analyzed using one-way analysis of variance (ANOVA) and Dunnette’s tests were performed to assess significance among treatment groups. BON cell migration experiments using genetic and pharmacological manipulations were assessed by Tukey’s multiple comparison tests. The statistical significance of the western blot data was analyzed by 2-tailed t tests.

An initial set of experiments was performed to pharmacologically characterize store-operated and arachidonic acid (AA)-activated forms of Ca²⁺ entry in BON cells by selective blockade using a broadly selective Ca²⁺ channel blocker SK&F 96365 (SKF) or 2-aminoethoxydiphenyl borate (2-APB) that has been shown to be less effective at blocking Orai3-containing Ca²⁺ channels. Store-operated Ca²⁺ entry (SOCE) was assessed using a standard protocol to deplete the endoplasmic reticulum (ER) stores of Ca²⁺ by treatment with 30 μM cyclopiazonic acid (CPA), a reversible inhibitor of the sarco-endoplasmic reticulum Ca²⁺-ATPase (SERCA). The CPA was dissolved in a saline containing nominal Ca²⁺ (0.5 mM EGTA with no added Ca²⁺). As shown in Fig. 1a, following CPA treatment there was a transient increase in concentration of cytoplasmic Ca²⁺ ([Ca²⁺] _cyt ) that indicated depletion of Ca²⁺ from the ER ([Ca²⁺] _ER ). Following restoration of 2.2 mM Ca²⁺ to the bath solution, a second, larger transient elevation in [Ca²⁺] _cyt was observed indicative of SOCE. As shown in Fig. 1a, c, activation of SOCE using standard protocol resulted in a change of 0.38 ± 0.05 ratio units (n = 3), whereas, treatment with 30 μM SKF (n = 3) or 50 μM 2-APB (n = 3) greatly attenuated the SOCE responses. The mean peak amplitudes of the response in SKF- or 2-APB-treated cells were 0.04 ± 0.00 and 0.14 ± 0.05 ratio units, respectively (n = 3). As shown in Fig. 1b, treatment with SKF or 2-APB did not have a significant effect on the store-content, as reflected by the CPA-mediated Ca²⁺ release. While, the average amplitude of the CPA-mediated release in control cells was 0.13 ± 0.02 ratio units, treatment with SKF and 2-APB resulted in Ca²⁺ release with the average amplitudes of 0.14 ± 0.02 and 0.13 ± 0.01 ratio units, respectively.

We next tested whether addition of AA activated a distinct Ca²⁺ entry pathway. Application of exogenous 1–30 µM AA induced elevations in [Ca²⁺] _cyt in a concentration-dependent manner (Additional file 1: Figure S1). A sub-maximal dose of 6 μM AA was used to treat cells for all the experiments described in the current study. As shown in Fig. 1d, application of 6 μM AA in nominal Ca²⁺-containing bath solution induced an initial transient rise in [Ca²⁺] _cyt that was followed by robust sustained elevation in [Ca²⁺] _cyt following return of Ca²⁺ in the bath solution. Typically, the secondary responses were complex waveforms and we used the peak amplitude measured within the first 300 s post-application as an index of the magnitude of the AA-induced Ca²⁺ entry. On average in about 60% of the cells, treatment with AA resulted in Ca²⁺ response with bimodal kinetics, while the other 40% of cells responded with a gradual rise and sustained increase in [Ca²⁺] _cyt . These responses were in contrast to those evoked by SOCE that showed only a transient rise in cytosolic calcium. It was unclear as to what underlies the difference between these populations. Examples of these AA-induced responses are shown in Additional file 1: Figure S1. The average change in amplitude for all the experiment was 1.48 ± 0.29 ratio units (n = 3). To further distinguish AA-induced Ca²⁺ entry from SOCE we assessed the effects of known pharmacological inhibitors of SOCE. As shown in Fig. 1d, e, the AA-induced rise in [Ca²⁺] _cyt was substantially diminished by 30 μM SKF with average change in amplitude of 0.36 ± 0.06 ratio units (n = 4). In contrast, the AA-induced rise in [Ca²⁺] _cyt was not diminished by treatment with 50 μM 2-APB and was 2.21 ± 0.12 ratio units (n = 3), demonstrating that the SOCE and the AA-induced Ca²⁺ entry pathways were pharmacologically distinguishable.

We next performed a series of manganese quench assays to measure the rates of quenching and follow divalent ion entry independent of potential contribution by Ca²⁺ buffering, clearance or release from internal stores. The rates of the Mn²⁺-induced quench of the fluorescence response following activation or inhibition of SOCE and the AA-induced responses was determined and compared against the rate of quenching in unstimulated cells. As shown in Fig. 2a, b, CPA-induced store depletion resulted in a nearly sixfold faster rate of quenching than that measured in unstimulated cells. The normalized rate of quenching for unstimulated cells was 1.00 ± 0.10 fluorescent unit (FU)/s (n = 4), whereas that after CPA treatment was 5.56 ± 0.87 FU/s (n = 4). Treatment with both SKF, as well as, 2-APB diminished the rate of quenching compared to the unstimulated cells to 0.94 ± 0.50 FU/s (n = 3) and 0.58 ± 0.37 FU/s (n = 4), respectively. Treatment with 30 µM SKF or 50 µM 2-APB alone had no effect on the unstimulated rate of entry (data not shown). Consistent with the activation of a Ca²⁺ entry channel, treatment with AA resulted in rapid quenching of the fura-2 fluorescence signal. In this case the rate of quenching was 14.36 ± 0.59 FU/s (n = 4), indicating that the rate of entry was approximately 2.5 fold faster than the CPA-evoked rate of entry, and 14 times faster than the unstimulated rate of entry. In contrast to the CPA-treated cells, 2-APB treatment had no significant effect on the rate of AA-induced entry, and the rate of quenching was 24.62 ± 5.26 FU/s (n = 4). Treatment with SKF however, resulted in significant diminishment of entry. The normalized rate of AA-induced quenching was reduced to 1.45 ± 0.12 FU/s (n = 4) by treatment with SKF.

As shown in Fig. 1d, treatment with AA induced a rise in [Ca²⁺] _cyt in the absence of extracellular Ca²⁺. Therefore, we asked if the rise in [Ca²⁺] _cyt resulted from ER Ca²⁺ release and whether this release was sufficient to depleted the [Ca²⁺] _ER and trigger SOCE. To determine whether AA-induced Ca²⁺ release from the ER, we pre-treated the cells in nominal Ca²⁺ containing saline with thapsigargin (TSG), an irreversible inhibitor of the SERCA pump, and then applied AA. As shown in Fig. 3a (upper panel), 1 μM TSG induced robust, transient release of Ca²⁺ from ER stores. Subsequent application of AA did not induce significant additional release. The mean amplitude of this signal was only 0.007 ± 0.001 ratio units (n = 4). In contrast, in time-matched control experiments (lower panel), where TSG treatment was omitted, AA evoked Ca²⁺ release with mean amplitude of 0.24 ± 0.05 ratio units (n = 4), a value similar in magnitude to TSG-evoked release. These results indicated that AA mobilized Ca²⁺ release from a TSG- or CPA-sensitive intracellular store of Ca²⁺. Thus, to test whether AA treatment depleted [Ca²⁺] _ER , a signal that is necessary to trigger canonical SOCE, we reversed the sequence of treatment as shown in Fig. 3b (upper panel). When AA was applied first, the [Ca²⁺] _cyt gradually increased to a new steady state about 200 s. Subsequent addition of TSG resulted in an additional release of Ca²⁺ with mean amplitude of 0.08 ± 0.01 ratio units (n = 3). However, this Ca²⁺ release signal was significantly smaller than that evoked by TSG alone (on average 0.19 ± 0.03 ratio units (n = 3) as shown in a time-matched control experiment (Fig. 3b: lower panel). This demonstrated that AA mobilized [Ca²⁺] _ER , but did not deplete the ER in our experimental paradigm. Moreover, mobilization of [Ca²⁺] _ER by treatment with AA was not sufficient to trigger SOCE (Fig. 3d, upper panel). As shown in Fig. 3d, e, following restoration of Ca²⁺ in the bath, the mean change in amplitude for cells that were pre-treated with AA in nominal Ca²⁺-containing solution was 0.32 ± 0.20 ratio units (n = 3) (Fig. 3d, upper panel). If pre-treatment with AA was omitted, the mean change in the signal amplitude was 0.19 ± 0.06 ratio units (n = 3). These slight changes in ratio values reflected the change in steady state between Ca²⁺-containing and Ca²⁺-free media and indicated that (1) AA-induced mobilization of [Ca²⁺] _ER did not activate Ca²⁺ entry and (2) that continuous presence of AA was needed to sustain AA-induced Ca²⁺ entry.

Because SOCE and the AA-induced Ca²⁺ entry pathways could be distinguished functionally in BON cells, we investigated the requirements of Orai1 and Orai3 channel subunits for these modes of Ca²⁺ entry. Specific channel subunits in BON cells were selectively knocked down by transfecting with plasmids that expressed short hairpin RNAs (shRNAs) for Orai1, Orai3 and scrambled sequences for either genes of interest as controls. Since, previous work indicated that BON cells express low levels of Orai2 mRNA, the role of this homolog in mediating Ca²⁺ entry was not assessed in the current study [30]. Western blotting was performed 48 h post-transfection to validate the effectiveness of knockdown of protein expression. As shown in Fig. 4, Orai protein band densities were normalized to the β-actin loading control and expressed as a ratio. The Orai1 expression level in cells that were transfected with scrambled shRNA was determined to be 1.21 ± 0.21 ratio units (n = 4). In contrast, when cells were transfected with Orai1 shRNA the ratio value was 0.26 ± 0.07 ratio units (n = 4) and represented approximately 78% knockdown of Orai1. Expression of Orai3 was knocked down by 82% when compared to control cells that were transfected with scrambled shRNA. (The average ratio values of the Orai3 bands in cells transfected with scrambled and Orai3 shRNAs were 1.95 ± 0.29 (n = 4) and 0.35 ± 0.10 ratio units (n = 4), respectively). In addition, cell lysates from knocked down cells were probed for the expression of both Orai1 and Orai3 homologs to test for changes in the expression of the non-targeted homolog. In these experiments, the average Orai1 ratio value of expression in cells transfected with Orai3 shRNA was 1.07 ratio units (n = 2) and the ratio value of Orai3 expression in cells transfected with Orai1 shRNAs was 1.82 ratio units (n = 2). These values were comparable to control values and indicated there was no significant effect on the expression of the non-targeted homolog. An example of this control is shown in Additional file 2: Figure S2.

The functional consequences of knockdowns of Orai homologs on Ca²⁺ entry were investigated using live cell Ca²⁺ imaging methods. As shown in Fig. 5a, c, knockdown of Orai1 significantly attenuated the amplitude of SOCE in comparison with cells transfected with scrambled shRNA. While the mean SOCE amplitude for cells where Orai1 was knocked down was 0.16 ± 0.02 ratio units (n = 5), the amplitude for cells transfected with scrambled shRNA was 0.38 ± 0.04 ratio units (n = 5). However, knockdown of Orai3 did not significantly diminish the mean SOCE amplitude. In these cells, the peak amplitude of SOCE was on average 0.27 ± 0.01 ratio units (n = 5). As shown in Fig. 5b, the genetic manipulation had no significant effect on the CPA-evoked Ca²⁺ release signal. The average amplitude of the CPA-evoked release signal in cells that were transfected with scrambled shRNA was 0.16 ± 0.02 ratio units and that for cells transfected with Orai1 and Orai3 shRNAs were 0.15 ± 0.01 and 0.12 ± 0.01 ratio units, respectively.

In contrast to the SOCE responses, knockdown of either Orai1 or Orai3 significantly decreased the amplitude of the AA-induced responses compared to cells transfected with scrambled shRNA. While the mean amplitude of the AA-induced Ca²⁺ entry in cells transfected with scrambled shRNA was 1.68 ± 0.32 ratio units (n = 5), cells that were transfected with Orai1 or Orai3 shRNAs had significantly reduced peak amplitudes of 0.29 ± 0.09 (n = 6) and 0.29 ± 0.10 (n = 6) ratio units, respectively. These results indicated that both Orai1 as well as Orai3 were required for mediating Ca²⁺ entry in response to AA and that channels that contained predominantly Orai1 were required for SOCE.

We next selectively activated store-operated or AA-induced Ca²⁺ entry to assess their relative effectiveness to induce migration in BON cells. We first tested whether the induced cell migration required extracellular Ca²⁺. As shown in Fig. 6a, reduction of extracellular Ca²⁺ had minimal effect on the basal migration compared to SFM alone. For example, the percentage of cells migrated in SFM was 3.89 ± 0.40% (n = 4) while the percentage of cells migrated in a reduced Ca²⁺-containing SFM was 1.96 ± 0.40% (n = 3). Moreover, direct activation of SOCE by treatment of BON cells with either 1 μM TSG or 30 µM CPA failed to enhance cell migration significantly above basal levels with only 3.54 ± 0.47% cells migrated (n = 3). In contrast, migration of BON cells was enhanced approximately threefold in response to treatment with 6 μM AA where 14.43 ± 1.61% of cells migrated (n = 4). Reducing extracellular Ca²⁺ abrogated the AA-induced enhancement of cell migration (2.15 ± 0.31%; n = 3) consistent with a requirement for Ca²⁺ entry. To further assess the relative contributions of store-operated- and AA-induced Ca²⁺ entry in BON cell migration, the pharmacological inhibitors of SOCE, SKF and 2-APB, were tested on AA-induced BON cell migration. Treatment with SKF significantly reduced the AA-induced cell migration. In these experiments only 2.82 ± 0.58% cells on average migrated (n = 4). In contrast, 2-APB had no significant effect on the AA-induced enhancement of cell migration when compared to AA alone. In these experiments 12.76 ± 0.53% cells on average migrated (n = 4). It was interesting to note that in contrast to the acute effects of AA-treatment where Ca²⁺ entry was enhanced in the presence of 2-APB, inclusion of 2-APB did not enhance AA-induced cell migration. This most likely resulted from a difference in the time courses of treatment for Ca²⁺ imaging and Boyden chamber experiments. Taken together, activation of AA-induced Ca²⁺ entry, but not SOCE, could mobilize a sub-population of BON cells to migrate in a Boyden chamber.

However, because AA can be readily metabolized into other bioactive classes of compounds such as leukotrienes and prostaglandins, we assessed whether AA-metabolites contributed in BON cell migration. Moreover reports by Mohamed Trebak’s group demonstrated that leukotriene-C4 could mediate a store-independent Ca²⁺ entry via Orai1/3 channels [14, 24]. To test whether AA-metabolites were involved in BON cell migration, we treated the cells with pharmacological inhibitors of AA-metabolism. Ketoprofen, a non-selective inhibitor of cyclooxygenase I and II, and nordihydroguaiaretic acid (NDGA), a pan-lipoxygenase inhibitor were included along with AA in the migration assays. As shown in Fig. 6b, neither of these compounds caused a significant reduction in AA-induced cell migration. In presence of ketoprofen or NDGA, 13.7 ± 0.76% (n = 3) or 13.12 ± 0.59% (n = 3) cells migrated, respectively. These data were consistent with the idea that AA itself, and not its metabolite caused the enhancement of cell migration.

Because AA-induced Ca²⁺ entry is thought to be mediated by a channel that contains both Orai1 and Orai3, we investigated the role of these homologs in AA-induced cell migration. As shown in Fig. 7c, d, knockdown of either Orai1 or Orai3 in BON cells resulted in a dramatic reduction in AA-induced migration compared to cells transfected with scrambled shRNA. While 13.45 ± 0.77% (n = 3) of the cells transfected with scrambled shRNA migrated when, only 2.69 ± 0.32% (n = 3) of cells transfected with Orai1 shRNA and 2.68 ± 0.15% (n = 3) of cells transfected with Orai3 shRNA migrated in response to application of 6 μM AA. Since knocking down Orai3 had a minimal effect on SOCE, the result of this experiment was consistent with our pharmacological data that suggested that AA-induced Ca²⁺ entry is the dominant Ca²⁺ entry pathway regulating BON cell migration in vitro.

Transformed cells can exhibit a migratory phenotype that correlates with increased metastatic potential. In BON cells that express both epithelial and neuroendocrine features, this can be achieved via neuroendocrine-to-mesenchymal transition (NMT). This transition is characterized by a reduction in neuroendocrine markers and altered expression of mesenchymal proteins [32‐34]. Thus, we assessed whether AA treatment could induce morphological changes and alter expression of markers of NMT including chromogranin A (CGA) E-cadherin, α-smooth muscle actin (α-SMA) and snail family transcriptional repressor 1 (Snail1).

In these experiments, cells were plated onto coverslips or applied to transwell inserts, serum starved for 4 h, incubated overnight or for up to 48 h in growth medium supplemented with 6 µM arachidonic acid or 0.1% DMSO as vehicle control and subsequently assessed by immunofluorescence. Morphological changes induced by AA treatment were visualized by F-actin staining using fluorescently labeled phalloidin. The majority of control cells appeared clustered and were rounded or oblong in shape (~ 97%), whereas the majority of AA treated cells (~ 93%) exhibited an elongated, flattened shape that was often decorated with multiple cellular processes. Additional images of these morphologies are shown in Additional file 3: Figure S3. As indicated above, migration assays revealed that AA treatment induced migration in a subset of BON cells. Therefore, we removed and imaged chamber inserts following AA treatment and assessed the morphology of non-migrated and migrated cells using confocal microscopy. As shown in Fig. 7a, b, staining for F-actin revealed distinctive morphologies between these two populations. The unmigrated cells on the surface of the transwell insert appeared to be in clusters and exhibited rounded shapes whereas the migrated cells on the bottom of the insert typically exhibited flattened irregular shapes with processes. These respective morphologies were similar to those observed for cells cultured on glass coverslips that were treated with vehicle or AA.

One hallmark of NMT is a reduction in the neuroendocrine secretory granule protein CgA. Under our experimental conditions, approximately 34% of cells in control groups exhibited perinuclear and subplasmamembrane distribution of CgA signal. Following overnight AA treatment, less than 15% of cells showed labeling. Figure 7c, d indicated that following AA treatment there was about a 50% reduction in the number of cells expressing CgA. Moreover, the treated cells appeared much more mesenchymal in morphology and some cells showed persistent expression and ubiquitous distribution of the CgA label through out the cell including the filipodial extensions. Longer treatments (24–48 h) showed a further reduction in the number of CgA expressing cells.

Furthermore, as shown in Fig. 7e, f, there was a clear reduction in the intensity of E-cadherin staining following AA treatment. In comparing images of control and AA-treated cells, there was typically greater than 60% reduction in the mean pixel intensity of the E-cadherin (green) signal. The findings were consistent with NMT phenotype. Of note, the E-cadherin signal in BON cells had an atypical nuclear distribution, similar to that previously shown to correlate with invasiveness in pancreatic NETs [34].

In contrast, we did not observe changes in expression levels of the mesenchymal proteins α-SMA and Snail1 following AA treatment (data not shown). The lack of an effect may be in part be explained the surprising expression of these markers in BON cells prior to AA treatment.