Background
Tumor cells within a tissue are constantly under the selective pressure of the surrounding environment, characterized by heterogeneity in terms of cellular components, availability of nutrients, fluctuating oxygen levels, poor tissue perfusion and low pH [
1,
2]. Such a complex environment is poorly reproduced by standard in vitro culture conditions, which provide excessive levels of nutrients and growth factors, an atmospheric oxygen pressure never reached in human tissues (21%) and the physiological pH conditions of healthy tissues (pH 7.4). Drug-resistance is one of the most important problems in clinical oncology [
3], yet anticancer drugs are normally identified and tested using screening procedures that do not consider the biochemical and physical complexity of solid tumors [
4]. This classical approach usually results in the identification of substances that inhibit cell proliferation, but not necessarily affecting the survival of non-proliferating cancer cells. One potential approach to overcome multicellular resistance is the use of multicellular spheroids (MCS), since this 3D culture system better reproduces the oxygen, nutrients and pH gradients observed in human tumor tissues [
5,
6]. Results from different studies have indicated that such model systems reduce the chances of selecting anticancer compounds with insufficient pharmacological properties.
One of the simplest biochemical properties affecting drugs’ activity, distribution and intracellular accumulation is their charge status in different pH conditions [
7,
8]. Cells within solid tumors are known to be exposed to chronic and/or intermittent acidic conditions as a consequence of high metabolic rates, producing acid metabolites which are not sufficiently removed because of scarce vascular perfusion [
1,
9]. Acidification of the extracellular environment causes cell cycle arrest, decreased glycolysis and protein synthesis, autophagy and increased angiogenesis and metastases [
8,
10]. Besides promoting malignant progression of tumors, tumor acidosis represents an important mechanism of drug resistance, likely associated with poor drug efficacy [
3,
8,
11,
12]. In an attempt to overcome acidosis-mediated drug resistance, we developed a drug screening assay to target the colon cancer HCT116 cell line chronically adapted to acidosis. The screening of the Prestwick library compounds led to the identification of the photosensitizer Verteporfin (VP) as an effective compound with preferential activity towards cells in acidic conditions. We also discovered several aspects of the Verteporfin mechanism of action that raise several cautions with regards to interpretation of experimental results with this drug.
Materials and methods
Chemical and antibodies
RPMI-1640 (SH30255.01), trypsin (SH40003.12), phosphate-buffered saline (PBS, SH40003.12) and Fetal Bovine Serum (FBS) (SV30160.03) were purchased from HyClone. Sodium Bicarbonate (25080) and RPMI-1640 without NaHCO3 (51800) were purchased from Gibco. Bafilomycin A1 (BafA1, B1793), protease cocktail tablets EDTA-free, phosphatase inhibitors (P5726, P0044) and Verteporfin were purchased from Sigma. Protein assay dye reagent concentrate (500-0006), protein assay standard (5000-0007) and dry milk (170-6404) were purchased from Bio-Rad. The following antibodies were used: LC3B (Cell Signaling Technology, 2775), SQTM1/p62 (BD Biosciences, 610833) and β-actin (Sigma, A5441). HRP-conjugated anti-rabbit (NA934V) and anti-mouse (NXA931) antibodies, ECL system (RPN2106) and PVDF membranes (RPN303F) were purchased from GE Healthcare. The fluorogenic substrate Suc-LLVY-AMC (S-280) was purchased from Boston Biochem (USA).
Cell culture
The colon carcinoma cell line HCT116 was cultured in RPMI-1640 medium supplemented with 10% FBS, antibiotics and 2 g/l NaHCO
3. The low pH adapted HCT-116 cell line (AA-HCT116) was obtained by growing the parental cells in RPMI-1640 medium buffered at pH 6.8 (0.15 g/l NaHCO
3) for 3 months as previously described [
13,
14]. In the settings used, media pH was stable over 48 h and dropped by about 0.1–0.2 units after 3 days in culture for cells kept at pH 6.8 and by 0.3–0.4 units for parental cells. Since parental cells have a higher proliferation rate, so was medium pH dropping faster and cells were passaged every 2–3 days. AA-HCT116 cells were passaged every 4–5 days. The h-TERT RPE1 cell line was cultured in DMEM F12 medium supplemented with 10% FBS and antibiotics. The MelJuSo UbG76V-YFP cell line was cultured in DMEM medium completed with 10% FBS and antibiotics.
HCT116 cells were plated at 60,000 cells/well in 100 µl medium in XF24-well cell plates. Subsequently medium was replaced with 500 µl Seahorse assay medium (1 mM pyruvate, 25 mM glucose and 2 mM glutamine). Oligomycin, FCCP, rotenone and antimycin A were injected according to the Seahorse XF cell mito stress kit. The mitochondrial function was assessed detecting the ECAR and OCR values by using the Seahorse XF analyzer.
Drug screening assay
AA-HCT116 cells were plated at 15,000 cells/well in flat-bottom 96-wells plate, left under the laminar hood for 1 h and then moved to the incubator. On the next day, all plates were treated simultaneously with 10 µM of each of the compounds from the Prestwick chemical library (one compound per well). For each plate, quadruplicate wells were used as untreated controls and quadruplicate wells used for the positive control (VLX600 10 µM) [
15]. Plates were then moved back to the incubator and 48 h later cell viability was assessed by the acid phosphatase assay [
16]. Hits were identified using a threshold of 70% inhibition of cell viability. Hit confirmation was performed testing each compound in a dose range 0–15 µM in duplicate wells.
Cell viability and clonogenic assay in monolayer and multicellular spheroids cultures
Cells were plated in 96-well plates and the next day treated with different concentrations of the specific compounds. Forty-eight hours after treatment cell viability was evaluated by using the acid phosphatase assay.
For analysis of clonogenic cell survival, HCT116 and AA-HCT116 cells were plated at 300 cells/well in 6-well plates and treated with VP for 8 or 48 h, following culture for 6–8 days in drug-free media. Plates were then washed in PBS and colonies were stained with Giemsa.
Clonogenic cell survival was also assessed on HCT116 MCS treated with different compounds. Briefly, MCS were obtained by a modification of the hanging-drop method applied to 96-well plates as previously described [
17]. After drug treatment for 3 days, the MCS were collected, washed in PBS, and a single cell suspension was obtained after incubation with Accutase (Biolegend, 423201). Equal volumes of cells for each treatment condition were plated in triplicate in 6-well plates for 6–8 days, followed by staining the colonies with Giemsa. The software ImageJ was used to perform semi-automated colony counting.
Analysis of intracellular VP accumulation
Cells were plated in black 96-wells plates (Costar) and allowed to adhere overnight.
On the next day, VP was added in a range 0–15 μM for 4 h. After removal of drug containing medium, cells were extensively washed with PBS. VP fluorescence was recorded with excitation wavelength of 450 nm and an emission wavelength of 695 nm [
18] using a fluorescence microplate reader Tecan Infinite M1000 (Tecan, Männedorf, Switzerland). For each condition quadruplicate wells were used and VP fluorescence was normalized to protein content for each well.
Qualitative analysis of intracellular VP was performed by laser-scanning fluorescence microscopy with a Leica confocal microscope.
Studies on autophagic flux
Analysis of the autophagic flux by Western blot was done on different cell lines. Experiments were performed by plating cells in their appropriate medium and allowing them to adhere overnight. On the next day, VP was added at specified concentrations for 4, 8 or 24 h and BafA1 (100 nM) was added during the last 2 h of incubation. Cells were then collected for Western blotting. Analysis of the autophagic flux was performed using the ratio of normalised levels of LC3-II in the presence and absence of BafA1. All experiments were performed in darkness conditions.
Western blotting
Cells were washed with PBS on ice and collected by scraping in cold PBS. The cell pellet was lysed in RIPA buffer (150 mM NaCl, 50 mM Tris pH 7.4, 1% Nonidet P-40, 0.1% SDS and 0.5% sodium deoxycholate) in presence of protease and phosphatase inhibitors. The protein concentration was determined using the Biorad Protein Assay (Biorad Laboratories) and equal amount of proteins (20 µg) was loaded on pre-casted acrylamide gels (4–12% SDS-PAGE, NuPage). The proteins were transferred from the gel to PVDF membrane for 2 h at 4 °C. Red Ponceau staining of the membranes verified the proper loading and transfer. Membranes were blocked in 5% blotting grade dry milk in TBS with 0.1% Tween (TBS-T) for 1 h at room temperature and then incubated with primary antibodies diluted in 5% BSA in TBS-T overnight at 4 °C. The next day membranes were washed and incubated for 1 h at room temperature with the appropriate HRP-conjugated secondary antibody and the binding was revealed by the ECL system.
Analysis of proteasomal activity
The fluorogenic substrate Suc-LLVY-AMC was used to measure the chymotrypsin-like activity of the proteasome on cellular lysates. Briefly, cells were treated with VP or BZ for 4 h, lysed in presence of 0.05% NP40, 1 µM DTT and 2 µM ATP, followed by the addition of Suc-LLVY-AMC (10 μM). Fluorescence intensity was recorded using a Tecan Infinite M1000 microplate reader with excitation wavelength 380 nm and emission wavelength 460 nm.
The cell line MelJuSo UbG76 V-YFP was used as reporter of proteasome inhibition [
19]. Cells were plated overnight and then exposed to VP or BZ. Detection of accumulated Ub-YFP was performed by Western blot analysis using an anti-GFP antibody.
RNA sequencing was performed in biological triplicate using Strand-specific TruSeq library preparation and Ribo-Zero ribosomal depletion. Tophat2, HTSeq, and DESeq 2 were utilized for alignment, quantification, and differential expression analysis, respectively, with the hg19 genome and Ensembl v73 [
20‐
22]. All bioinformatics were preformed in the R programming language [
23] and scripts related to the differential expression analysis, gene ontology analysis, and community detection analysis are publically available (
https://github.com/GranderLab/acidAdaptedRNAseq) as an R package facilitating reproduction of the analysis and associated figures. Principal component analysis and hierarchical clustering was performed using the prcomp and heatmap.2 functions, respectively, within the R statistical environment. Data was centered and scaled (mean = 0, sd = 1) for each gene previous to clustering and regularize log transformed previous to PCA.
Gene ontology term enrichment
Significant (alpha < 0.05) genes in the differential expression analysis were used for biological processes GO term enrichment analysis with the topGO software [
24]. All quantified genes, defined as counts per million > 1 in at least 3 samples, were utilized as the gene universe. Terms with less than 5 annotated genes were not included in the significance testing procedure. Significance testing was performed using the classic Fisher method and the top 1000 terms with the lowest p-value were included in downstream analysis unless otherwise specified.
To understand the relationship between the significant GO terms, the GO graph was retrieved for all significant (alpha < 0.05) GO terms using the GOSim Bioconductor package [
25,
26]. The GO graph was then utilized for community detection via the spin-glass algorithm [
27] from the igraph package [
28] using a maximum of 200 possible spin states. The authority score was subsequently calculated for all nodes in the GO graph to determine the node with the highest in-degree for each community [
29]. The GO graph was then collapsed on the nodes with the highest authority scores (community node) by merging all nodes into said node and simplifying the graphs edges. In cases when several nodes had identical authority scores equaling the max authority score for that community’s nodes, one was chosen at random to represent the community. All nodes not presently contained in the collapsed graph, were merged into the graph and edges were re-drawn between the merged nodes and the community node. Results were visualized with the graph package [
30].
Statistical analysis
All data were obtained from at least three experimental replicates and presented as mean ± SEM, if not otherwise indicated. Data analysis was performed using Graphpad Prism 6.0 (GraphPad Software, CA). Differences between groups were analysed with parametric or non-parametric tests according to the distribution of the values. The specific test used is indicated in figure legends. The significance level was set as P < 0.05.
Discussion
Tumor acidosis has been acknowledged as an important feature of solid tumors that can dramatically alter pathobiology of the disease [
8,
10]. Although it has been known that acidosis also negatively modulates drug sensitivity, few studies have aimed to identify agents with preferential activity towards cancer cells cultured in acidic conditions. Standard 2D cell culture has long been applied for the identification of compounds with activities on either proliferation or cell survival in preclinical settings. Conventional culture conditions imply the use of media with high glucose content, non-physiological oxygen pressure and alkaline pH set at 7.4. Overall, such conditions are poorly representative of the complex physiological environment that cancer cells experience
in vivo and hence drug screening performed in these simplified conditions might overestimate the potential clinical efficacy of identified compounds. With the aim to partially overcome such limitations, we developed and characterized HCT116 cells chronically adapted to low pH culture conditions. To our knowledge, the characterization of the transcriptome of acid adapted cells has thus far been limited. RNAseq analysis was applied to define major differences between parental and adapted cell lines, as summarized in Fig.
3. Although the full characterization and validation of these data was not in the scope of this study, the analysis suggests distinct biological properties of the low pH adapted AA-HCT116 cells, including a slower proliferation rate and an increased expression of several markers implicated in autophagy and cell differentiation. In line with previous reports on other cell types [
34,
43], AA-HCT116 cells show upregulation of several autophagy genes. AA-HCT116 cells also upregulate the expression of CD274 (PDL-1), a molecule whose expression on tumor cells mediates immunosuppression by inhibiting the activation of CTL and NK cells. Interestingly, acidosis is known to have a strong immunosuppressive activity, thus modulation of tumor acidosis has been suggested as a mechanism to improve immune therapy [
44‐
46].
An initial pilot screen with clinically used anticancer drugs showed that nearly all drugs had a lower cytotoxicity on AA-HCT116 as compared to parental HCT116 cells, confirming that cells chronically adapted to acidosis are resistant to many chemotherapeutics. Interestingly, Everolimus was recently reported to selectively target melanoma cells adapted to acidic pH [
47] and in our study the activity of Everolimus and Sorafenib was similar in the two cell lines.
These findings prompted us to use systematic screening to identify compounds with cytotoxic activity also against acid-adapted colon carcinoma cells, leading to the identification of VP as an interesting hit, with equipotent activity also on the adapted cell line. VP has been safely used as a photosensitizer in photodynamic therapy (PDT) of acute macular degeneration and it is under investigation in PDT of pancreatic cancer [
48]. Recently, it was suggested that VP specifically targets Yes-associated protein 1 (YAP1), a transcriptional coactivator in the Hippo signaling pathway [
37]. Based on this report, VP has been used as a pharmacological tool to study the protumorigenic activity of YAP1 in several cancer models [
49‐
51]. Moreover, some studies have also suggested that VP inhibits YAP activity in the absence of light activation [
39,
52]. However, a very recent study has shown that VP demonstrate activity against cancer cell lines that is independent of YAP1 and instead dependent on the ability to induce high MW protein oligomers, in particular STAT3 and SQSTM1 [
36]. In fact, VP was recently shown to induce cross-linking of SQSTM1 [
41]. In line with these studies, we observed that treatment with VP induces the accumulation of protein oligomers, including SQSTM1, GRP78 and mTOR.
Here, we extend these studies to demonstrate that the activity of VP towards cancer cells is pH-dependent and light-dependent and associated with protein cross-linking and alterations of polyubiquitinated proteins. The exact mechanism by which VP induces such a phenomenon is still unclear. However, we could exclude any involvement of the proteasome since we did not observe any inhibition of proteasomal function upon VP treatment using three different methods. Importantly, we observed that treatment of cells and protein extraction in complete darkness totally abolished the effects of VP on the observed accumulation of poly-ubiquitinated proteins, suggesting that extreme care should be taken in experimental studies addressing the mechanism of action of this drug, as recently pointed out by Kostantinou and colleagues [
53]. In fact, contrary to many studies where exposure to ambient light is not specifically addressed in experimental procedures, these authors present strong experimental evidence that cross-linked oligomers and formation of high MW complexes is mostly a light dependent mechanism. In line with this, we observed that the effects of VP on cell viability was also light-dependent since exposure of VP to ambient light for 30 min before treatment greatly increased VP effects in both HCT116 and AA-HCT116 cells.
Although the scope of this study was not to unravel the molecular mechanism by which VP affects cancer cell viability, we observed that VP has a preferential activity towards colon cancer cells grown in acidic conditions. This might be simply related to the higher intracellular availability of VP in both AA-HCT116 and parental HCT116 cells transiently exposed to acidic medium. This behavior may be predicted by the acidic pKa of the compound, which likely facilitates the uptake and intracellular availability.
In conclusion, our study describes the phenotypic and transcriptomic profile of acid adapted colon cancer cells. We used these cells to screen a library of FDA-approved compounds and identified Verteporfin as a hit compound preferentially targeting acidic cancer cells. Given that VP (Visudyne®) is already available for therapeutic applications, its use as anticancer agent might be easily translated into clinical oncology for specific cancer types.
Authors’ contributions
PP, JTS, NB, MM, IK, DY performed experiments and data analysis; TL, MH, PD, SL designed the study and analysed data, ADM designed the study and led the project. PP, JTS, PD, TL, MH, SL and ADM wrote and reviewed the manuscript. All authors read and approved the final manuscript.