Background
Glioblastoma Multiforme (GBM) is the deadliest tumor of the central nervous system and, due to its location, aggressive biological behavior and diffuse infiltrative growth, presents disproportionately high morbidity and mortality. The cornerstone of therapy consists of maximal well-tolerated surgical resection followed by radiotherapy plus concurrent and adjuvant chemotherapy with temozolomide (TMZ) [
1]. Despite this optimized treatment schedule, GBM is characterized by high rates of recurrences, therefore the median patient survival ranges from 12 to 15 months, where less than 5% of patients survive for more than 5 years after diagnosis [
1]. Moreover, since GBM is characterized by diffusely infiltrative growth and unusual ability in repairing the therapy-induced damages, a complete eradication can be exploited quite rarely [
2], which ultimately leads to a high rate of recurrences characterized by increased aggressiveness. Systemic therapy with TMZ, though with its limitations, remains the standard of care for GBM, but the frequent onset of chemoresistance in relapsed GBM generates a compelling need for novel therapeutic strategies. Of note, there is presently no valid alternative to the aforementioned single-drug approach using TMZ in concomitance of radiotherapy or alone in the adjuvant treatment.
SI113, a small molecule identified by virtual screening of a molecular library with respect to SGK1 crystal structures [
3], has proven to delay the cell cycle progression with cell accumulation in G0-G1 and block cancer growth in preclinical settings, both in vitro and in vivo [
4,
5]. SGK1, a structural and functional analogue of AKT [
6], is a key regulator in a number of patho-physiological cell functions [
7‐
9], thus the possibility to modulate its activity can be functional in several diseases [
10]. Indeed, SI113 inhibits epithelial-to-mesenchymal transition and subverts cytoskeletal organization in human cancer cells [
11] and, specifically in GBM, potentiates the cytotoxic effects of radiotherapy [
12] as well as those of mitotic spindle poisons [
13]. In the attempt to delve into the mechanism of action of this compound and assess the status of diverse signal transduction pathways in GBM cell lines [
14], we initially employed a Reverse-Phase Protein Arrays (RPPA) platform [
15‐
17], a technology designed for multiplexed, antibody-based relative quantification of specific cellular proteins along with their post-translational modifications. RPPA results outlined distinct molecular profiles between anchorage-dependent established cell lines and patient-derived neurospheres as far as pivotal cellular pathways governing cell growth and metabolism are concerned, either at the baseline or under the effect of SI113. Furthermore, SI113 triggered an autophagic response in GBM cells, ultimately leading to cytoprotective autophagy in neurospheres, thus suggesting that its administration concomitant with an inhibitor of the autophagic process could effectively hinder GBM growth.
Methods
GBM cell lines
ADF human GBM cells [
18] were a gift from Dr. W. Malorni (Istituto Superiore di Sanità, Rome, Italy). U373MG and T98G GBM cells were provided by Dr. C. Leonetti (IRCCS - Regina Elena National Cancer Institute, Rome, Italy). Cell line authentication was performed by short tandem repeat (STR) profiling, which resulted in > = 70% match for 8 loci as per interrogation of the ATCC STR profiling database. Similar to previous reports in the literature [
19], our U-373 MG matched ATCC’s U-251 MG cell line. Cells were cultured as previously reported [
13], were Mycoplasma-free and used for a maximum of 20 passages.
GBM3-Luc anchorage-independent neurospheres are a primary cell clone, growing in suspension, derived from a specimen of a GBM patient operated in the IRCCS - Regina Elena National Cancer Institute, Rome, Italy, collected according to the current institutional ethical guidelines. GBM3-Luc have been characterized as follows: a) derived from a GBM removed in 2012, negative for MGMT promoter methylation and wild-type for IDH1 gene; b) grow as neurospheres; c) tumorigenic in mice (orthotopic growth); d) CD133: negative; e) CD44: 40% positive; f) CD56: 97% positive; g) engineered to stably express a luciferase reporter gene. GBM-I is another primary neurosphere cell line, gift from Dr. A. Eramo (Istituto Superiore di Sanità, Rome, Italy) [
20,
21]. Both neurospheres were cultured in DMEM/F12 stem cell medium as described [
20,
21].
Drugs
SI113 was synthesized as previously reported [
3] and diluted at a 50 mM concentration in DMSO. Quinacrine (QC) (Sigma-Aldrich, St. Louis, MO) was diluted at a 10 mM concentration in phosphate-buffered saline (PBS).
RPPA
For RPPA analysis, 3.5 × 103 GBM3-Luc, ADF and U373MG cells were seeded onto 6-well microtiter plates; then cells from three different passages were used for biological replicates of individual experimental conditions. In details, the RPPA experimental design included two time points, i.e. 2 and 8 h, and two drug doses, namely the concentration resulting in 30 and 50% residual viability as measured at 48 h (see below for details on the cell viability assay used), i.e. IC30 and IC50, respectively. Control samples, i.e. cells treated with a concentration of drug vehicle identical to that of drug-treated samples, were included at each analyzed time point.
Protein extracts as well as RPPA lysate printing and immunostaining, were performed according protocols established in our laboratory [
17,
22]. Image analysis for spot recognition, quantification and normalization was carried out using ‘MicroVigene’ v5.2 software (
http://www.vigenetech.com/MicroVigene.htm) (VigeneTech Inc). Data analysis was performed on averaged technical and biological replicate values for each individual sample condition by means of ‘R’ v3.5.0 (
https://www.r-project.org/) (R Foundation for Statistical Computing) and ‘RStudio’ v1.1.414
https://www.rstudio.com/ (RStudio) using the following installed packages: base, plyr, tidyverse, ggsignif, FactoMineR, factoextra, RColorBrewer, Bioconductor and shiny.
Immunoblot analysis
In order to validate the RPPA results, cell lysates were processed for western blot as described [
23] and filters were probed using the antibodies listed in Additional file
1: Table S1. Further antibodies used throughout the study: mouse anti-p62 (SQSTM1) MoAb GT1478 (1:1000, BD Thermo Fisher Scientific, San José, CA); rabbit anti-LC3 PoAb (1:1000, MBL International, Woburn, MA); mouse anti-β-actin MoAb (1:10000, MP Biomedicals, Aurora, OH); Rabbit anti-Nucleolin PoAb (1:1000, Abcam, London, UK); mouse anti-GAPDH (1:24000, Sigma Aldrich, Saint Louis, MO).
Cell viability assay
Cells were seeded in 96-well plates at a concentration of 5 × 103 cells/well and then treated with drugs at the given concentrations. CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI) was employed to determine the relative number of viable cells, after 48 h of treatment, by means of a GLOMAX 96 Microplate Luminometer (Promega). Control samples were treated with the same final concentration of drug solvent(s) (DMSO and/or PBS).
Anchorage-dependent GBM cells were plated at a concentration of 1–2 × 102 cells/well in 6-well plates. After 24 h, vehicle(s), SI113, QC or a combination of both (as indicated) was added, and the culture was incubated for 48 h. Cells were then washed, cultured for additional 12 d and subsequently stained using a 5% crystal violet solution in order to assess the colony number.
GBM3-Luc and GBM-I cells were plated at 2.5 × 105 cells per well in stem medium with 10 ng/ml bFGF and 20 ng/ml EGF in a 6-well plate and treated with vehicle(s), SI113, QC or a combination of both compounds as indicated. Following 48 h treatment, GBM3-Luc cells were dissociated into single cell suspension by means of TrypLE Express (Gibco, Life Technologies), counted, diluted at the appropriate concentration and re-seeded in triplicate into new 6-well plates (1 × 102 cells per well). At d 26 cells were examined by means of an inverted microscope and neurospheres were counted on averaged triplicates of 9 fields/well using a 4 x objective. Neurosphere counting was blind and independently performed by two investigators.
GBM-I cells were dissociated, diluted and reseeded into new 6-well plates at the concentration of 5 × 102 cells/well. Since GBM-I cells grow more slowly and form smaller neurospheres when compared with GBM3-Luc cells, the former were all pelleted 26 d after treatment, fixed in 2% PFA, stained with 2% crystal violet and cytocentrifuged on a slide.
Cytofluorimetric assays
Evaluation of autophagy was performed by staining cells with Cyto-ID Autophagy Detection Kit (Enzo Life Sciences, Farmingdale, NY) optimized for detection of autophagy in live cells by flow cytometry [
24]. Samples were analyzed with a dual-laser FACScalibur flow cytometer (BD Biosciences Franklin Lakes, NJ).
Statistical and data analysis
Antibodies were sub-selected based on the concordance between the two analyzed time points. In details, data from all drug treatment conditions were log10-transformed and the Pearson’s correlation indices (‘r’) were calculated, stratified by antibody and cell line, between the 2 and 8 h time points. Endpoints included in subsequent data analyses comprised the roughly 25% displaying r values > 0.64 (third quartile) in at least one cell line out of the three analyzed (Additional file
1: Table S1). AMPK-α pS485 and AMPK-β pS108, but not AMPK-α pT172, scored r values above the threshold criteria. Nonetheless, we included AMPK-α pT172 since we found correlation (r = 0.72) in GBM3-Luc cells between AMPK-α pT172 and its functional substrate ACAC p79. Along similar lines, we utilized 4E-BP1 pT70 and ERK1–2 pT202-pY204, although not reaching the threshold criteria, since they are functionally related to key pathway targets.
Unless otherwise specified i) all experimental conditions were tested in technical triplicates and experiments performed at least three times, ii) all RPPA point-and-line plots include individual values as well as mean ± standard deviation (SD) and iii) all other results are expressed as a mean ± standard error (SE). Data from in vitro experiments were analyzed by One-way ANOVA test followed by Tukey’s Multiple Comparisons Test (GraphPad Prism v5). RPPA results were analyzed by a custom ‘R’ algorithm designed to i) test the normality assumption via Lilliefors or Shapiro-Wilk tests, ii) perform ANOVA and Bonferroni post-hoc tests if the data are normally distributed or iii) perform Kruskal-Wallis followed by Wilcoxon rank sum or signed rank tests as well as by FDR p value adjustment, in case of non-normal data. Statistical significance is reported on plots using the following notation: *P < 0.05, **P < 0.01, ***P < 0.001.
The assessment of synergy between two drugs was done using the algorithm described by Fransson et al. [
25], where synergy is characterized by a Combination Index (CI) < 0.8, while a CI between 0.8 and 1.2 indicates an additive effect and values > 1.2 indicate an antagonistic effect.
As far as RPPA data are concerned, we opted for n = 9 (3 technical × 3 biological replicates) for each individual experimental condition (cell line, drug concentration and time point) in order to achieve a power of 0.8 to detect a statistically significant (FDR 5%) difference in at least 20% of the endpoints, given > 75% non-overlapping populations. We generated custom R code based on ‘R Shiny’ and useful for interactive visualization of RPPA data by plots, dendrograms and heatmaps. The aforementioned code can be made available upon request.
For in vitro data (normally distributed) we tested the homogeneity of variance by Levene’s test and we found out that most of the conditions evaluated were homoscedastic. Accordingly, we performed ANOVA.
Discussion
The small molecule SI113 was originally identified as an inhibitor of the SGK1 kinase activity and the role of the drug in this context has been documented satisfactorily [
4,
5,
12,
13]. SGK1 plays a key functional role in the PI3K/mTOR pathway and is able to sustain AKT-independent mTORC1 activation [
6].
The strong impact of SI113 on cancer cell survival has been highlighted by several previous studies. Nonetheless, here we aimed at exploring in depth the pharmacological capability of SI113 in interfering with major signal transduction pathways in GBM cells. Indeed, SI113 was effective in hindering proliferation and cloning efficiency in anchorage-dependent GBM cell lines and neurospheres.
RPPA revealed a peculiar role of SI113 in GBM3-Luc neurospheres, where it provoked a rapid (already at 2 h and IC30) reduction of cell cycle related endpoints, i.e. Cyclin B1 and phospho-ERKs, and promoted a survival-oriented autophagic process. Indeed, upon treatment with SI113, a series of key pathway players, including mTOR pS2448 and its downstream targets 4E-BP1 pT37–46 and S6 pS235–236, as well as AMPK-α pT172 and ACACA pS79, displayed a striking correlation selectively in neurospheres. Concordantly, treatment with SI113 caused a decrease in cleaved PARP D214 in GBM3-Luc cells, thus excluding the involvement of an apoptotic process in these cells. In line with these data, analysis of the expression of LC3 and p62 showed that both GBM-3 Luc and GBM-I neurospheres underwent autophagy when exposed to SI113.
It should be remarked that GBM is characterized by the concomitance of diverse pathway alterations including, but not limited to, RTKs, PI3K/mTOR, Ras/MAPKs and cell cycle [
54]. In this regard, it is widely accepted that mTOR inhibition leads to ERK activation in a PI3K-dependent manner [
55] and that receptor tyrosine kinase-driven signals are rewired accordingly to selective blockade of either PI3K/mTOR or Ras/MAPK axes [
56,
57]. These acquired notions pose the rationale for designing combination therapies involving concomitant targeting of PI3K and MAPKs [
58]. Interestingly, SI113 was capable of targeting neurospheres by inhibition not only of PI3K/mTOR pathway but also of the MAPKs cascade, as represented by the noticeable drop in the levels of ERK1–2 phosphorylation in SI113-exposed GBM3-Luc cells (Additional file
4: Figure S2). Such a scenario implies that, for a minimally effective inhibition of GBM cell growth, both the aforementioned pathways should be targeted [
59]. Once again, SI113 is a very interesting candidate compound for GBM therapy, being capable of hindering multiple pathways and thus exposing GBM3-Luc neurospheres vulnerabilities.
Additionally, the autophagy inhibitor QC, which, as expected [
50‐
52], displayed considerable toxicity towards all the GBM cell lines tested (see Table
1) showing also an unquestionable ability to cooperate with SI113 in hindering GBM cells growth capabilities. Indeed, we found a strong synergy between SI113 and QC in contrasting key features of both anchorage-dependent GBM cells and neurospheres. Since the latter ones are usually resistant to standard anticancer treatments and display cancer stem cells behaviors, our results hold a substantial clinical significance.
Recently, ferroptosis, a peculiar form of cell death occurring through Fe (II)-dependent lipid peroxidation, has been shown to play a major role in cancer cell apoptotic processes [
60]. In addition, it has been reported that the blockade of autophagy by means of QC can enhance sensitivity to TMZ in GBM neurospheres by igniting ferroptosis [
61].
In summary, autophagy allows GBM cells to survive in a hostile environment and is regarded as a cytoprotective adaptive reaction, particularly in cancer stem cells [
62‐
64]. Herein, exposure to SI113 forced neurosphere cells towards a pro-survival autophagic response that was efficiently constrained by the autophagy inhibitor QC, thus provoking the suppression of a survival pathway that these cells likely depend on. Therefore, we envisage a combination therapy approach that, differently from standard “one-hit” targeted therapies, could be based on synthetic lethality, inhibiting primitive cancer cell survival pathways, being thus less prone to the emergence of resistant clones and providing strong benefits in a foreseen clinical setting. Indeed, GBM due to its heterogeneity cannot be considered as a “single pathway disease” and doesn’t represent the best environment to assay the effects of a targeted therapy. Conversely, in this disease there is a window of opportunity for the so-called dirty drugs that, far from being specific, can take advantage of selective vulnerabilities of GBM cells, e.g. dependence from aberrant energy supply pathways or pro-survival autophagic response to hostile environment. In such a scenario, our results show the importance to have a complete signal transduction portrait in order to address the most efficient therapeutic strategies.
Conclusions
All so far published in vitro and in vivo preclinical studies based on SI113 hold great promises for its use in a combinatorial therapeutic regimen in cancer, including GBM [
4,
12]. In this respect, a novel therapeutic strategy consisting in TMZ plus SI113 and QC, may provide an effective drug cocktail against GBM [
14]. Notably, SI113 has been demonstrated to potentiate the effects of radiotherapy in GBM cells [
12], thus advocating its clinical use in GBM. Furthermore, QC or other quinolone derivatives, e.g. the widely used chloroquine, display well-established levels of toxicity and dosage in humans, making them ideal candidates for drug repositioning [
14].
Therefore, all the aforementioned elements converge in supporting a Phase 1 clinical trial for safety and dosage assessment of SI113 and QC or its analogs, in order to evaluate the effects of their co-administration with TMZ (and, when still possible, radiation therapy) in GBM patients who have exhausted all available treatment options.
Acknowledgements
The authors wish to thank Dr. Walter Malorni (Istituto Superiore di Sanità, Rome, Italy) for providing the ADF cell line; Dr. Carlo Leonetti (IRCCS Regina Elena National Cancer Institute, Rome, Italy) for providing the U373MG and T98G cell lines; Dr. Iole Cordone (IRCCS Regina Elena National Cancer Institute, Rome, Italy) for her help in characterizing the GBM3-Luc cell line; Dr. Giulia Carpinelli (Istituto Superiore di Sanità, Rome, Italy) for the analysis of drug-related toxicity data.