2016 Fasting Boosts Sensitivity of Human Skin Melanoma To Cisplatin - Induced Cell Death
2016 Fasting Boosts Sensitivity of Human Skin Melanoma To Cisplatin - Induced Cell Death
2016 Fasting Boosts Sensitivity of Human Skin Melanoma To Cisplatin - Induced Cell Death
a r t i c l e i n f o a b s t r a c t
Article history: Melanoma is one of leading cause of tumor death worldwide. Anti-cancer strategy includes combination
Received 23 September 2016 of different chemo-therapeutic agents as well as radiation; however these treatments have limited ef-
Accepted 28 September 2016 ficacy and induce significant toxic effects on healthy cells. One of most promising novel therapeutic
Available online xxx
approach to cancer therapy is the combination of anti-cancer drugs with calorie restriction. Here we
investigated the effect Cisplatin (CDDP), one of the most potent chemotherapeutic agent used to treat
Keywords:
tumors, in association with fasting in wild type and mutated BRAFV600E melanoma cell lines. Here we
Melanoma
show that nutrient deprivation can consistently enhance the sensitivity of tumor cells to cell death in-
CDDP
Starvation
duction by CDDP, also of those malignancies particularly resistant to any treatment, such as oncogenic
BRAF BRAF melanomas. Mechanistic studies revealed that the combined therapy induced cell death is char-
Apoptosis acterized by ROS accumulation and ATF4 in the absence of ER-stress. In addition, we show that auto-
phagy is not involved in the enhanced sensitivity of melanoma cells to combined CDDP/EBSS-induced
apoptosis. While, the exposure to 2-DG further enhanced the apoptotic rate observed in SK Mel 28 cells
upon treatment with both CDDP and EBSS.
© 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
http://dx.doi.org/10.1016/j.bbrc.2016.09.149
0006-291X/© 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: F. Antunes, et al., Fasting boosts sensitivity of human skin melanoma to cisplatin-induced cell death,
Biochemical and Biophysical Research Communications (2016), http://dx.doi.org/10.1016/j.bbrc.2016.09.149
2 F. Antunes et al. / Biochemical and Biophysical Research Communications xxx (2016) 1e7
interventions also up-regulates the AMPK signalling, an important 2.5. Real time PCR analysis
pathway with different targets for cancer treatment [15]. This
pathway is also involved in autophagy regulation [16], a possible Total RNA was extracted by using Trizol reagent (Invitrogen) as
mechanism involved in the action of fasting since during the stress recommended by the supplier. cDNA synthesis was performed
induced by nutritional deprivation autophagy is up-regulated as a using a reverse transcription kit (Promega) according to the man-
survival mechanism [17]. Though, it is known that aggressive tu- ufacturer's recommendations. Quantitative PCR reactions were
mors that harbor Ras mutations are energetically compromised performed by using a Rotor-Gene 6000 (Corbett Research Ltd)
under starvation conditions since this mutation limits a higher thermocycler. Maxima SYBR Green/ROX qPCR Master Mix (Thermo
induction of the already increased basal autophagy. In this way Scientific) was used to produce fluorescently labeled PCR products.
there is not enough energy supply, leading tumor cells to cell death Primer sets for all amplicons were designed using Primer-Express
[18,19]. 1.0 software (Roche).
Cisplatin (CDDP) is one of the most potent chemotherapeutic ATF4 forward: 50 -GTGGCCAAGCACTTCAAACC-30
agent used to treat a wide range of malignancy as lung cancers, ATF4 reverse: 50 -CCCGGAGAAGGCATCCTC-30
head and neck tumors, ovarian carcinoma, among others [20]. ATF6 forward: 50 -TTTGCTGTCTCAGCCTACTGTGGT-30
Before the development of targeted therapies, cisplatin was one of ATF6 reverse: 50 -TCCATTCACTGGGCTATTCGCTGA-30
the first choice to treat melanoma, however with a poor response Xbp-I forward: 50 -GAATGAAGTGAGGCCAGTG-30
rate [1]. CDDP resistance may occur at several levels as the cellular Xbp-I reverse: 50 -GAGTCAATACCGCCAGAATC-30
extrusion of the drug [21], mechanism of DNA repair activation [22] L34 forward: 50 -GTCCCGAACCCCTGGTAATAGA-30
and autophagy induction [23]. Nevertheless, identifying new uses L34 reverse: 50 -GGCCCTGCTGACATGTTTCTT-30
for existing drugs might be an effective therapeutic approach to L34 mRNA level was used as an internal control and results were
overcome drug resistance in diverse malignancy and sustain pa- expressed as previously described [24].
tient's survival [29]. Therefore, we aimed to evaluate the opportu-
nity of using fasting cycles combined with cisplatin treatment as a 2.6. Western blotting analysis
novel valuable clinical approach to treat human skin melanoma
independently of its mutational status. Cells were lysed in Cell Lytic buffer (Sigma-Aldrich) supple-
mented with protease and phosphatase inhibitors (protease in-
hibitor cocktail plus 5 mM sodium fluoride, 0.5 mM sodium
2. Materials and methods
orthovanadate, 1 mMsodium molybdate, 50 mM 2-
chloroacetamide, 2 mM 1,10-phenanthroline monohydrate, and
2.1. Materials
0,5 mM PMSF; Sigma-Aldrich). Lysates were incubated at 4 C for
30 min in ice. After centrifugation at 4 C for 10 min at 13,000 rpm
Cisplatin (Sigma Aldrich), Bafilomycin A1 (Sigma Aldrich), Z-
to remove insoluble debris, protein concentration was determined
VAD-fmk (Santa Cruz Biotechnology), Necrostatin-1 (Sigma
using a Bradford assay (Biorad). 20 mg of total proteins were
Aldrich), N-acetyl-cysteine (Sigma Aldrich), 2-deoxy-glucose
resolved by using NuPAGEBis-Tris gels (Life Technologies) and
(Sigma Aldrich), EBSS (Sigma Aldrich).
electroblotted onto nitrocellulose (Protran, Schleicher&Schuell) or
PVDF (Millipore) membranes. Blots were incubated with primary
2.2. Cell culture antibodies resuspended in 5% non-fat dry milk in PBS plus 0.1%
Tween-20 overnight at 4 C. Detection was achieved using a
CHL-1 and SK Mel 28 metastatic melanoma cells were cultured horseradish peroxidase-conjugate secondary antibody (1:5000 in
in DMEM high glucose (Sigma-Aldrich) supplemented with 10% FBS 5% non-fat dry milk in PBS plus 0.1% Tween-20, Jackson Immuno
(Life Technologies), 2 mM L-glutamine, and 1% penicillin/strepto- Research Laboratories), visualised with ECL (GE Healthcare) and
mycin solution, at 37 C under 5% CO2. images were recorded by using a Chemidoc apparatus (BioRad).
Primary antibodies were: anti-PARP cleaved (1:1000, Cell Signal-
ling), anti-p62 (1:2000 MBL), anti-LC3B (1:1000, Cell Signalling),
2.3. Apoptosis evaluation
anti-Ambra1 (1:1000, Cell Signalling), anti-Beclin-1 (1: 500, Santa
Cruz Biotechnology); anti-Atg5 (1:500, Santa Cruz Biotechnology);
Briefly, 1,5 105 CHL-1 or SK Mel 28 cells were treatment as
anti-Actin (1:5000, Santa Cruz Biotechnology) and anti-GAPDH
indicated, for 24 h. Then, cells were fixed with methanol-acetone,
(1:20000, Calbiochem).
pelleted and resuspended in RNAse (50 mg/ml in PBS) and incu-
bated at 37 C for 15 min followed by propidium iodide (25 mg/ml in
2.7. Statistics
PBS) staining. The cell cycle distribution was evaluated by flow
cytometry. 10.000 events were acquired using a FACS Calibur cy-
All values are represented as the mean ± SD. Significance was
tometer (Becton-Dickinson, Mountain View,CA, USA) and the sub-
evaluated by ANOVA one-way followed by Tukey test for multiple
G1 phase analyzed using FlowJo software [24].
comparisons among control and treatments. ANOVA two-way fol-
lowed by Bonferroni post-test was used for group analysis. Differ-
2.4. ROS and DJm assessment ences were considered significant with p 0.05. At least three
independent experiments were conducted to warrant that the re-
Briefly, 1,5 105 CHL-1 or SK Mel 28 cells were treatment as sults were representative.
indicated and cells harvested at indicate time points. Then, cells
were pelleted and resuspended in H2-DCFDA (10 mM in PBS) or in 3. Results and discussion
TMRM (50 nM in PBS), to detect ROS or to evaluate the DJm,
respectively. Cells were incubated at 37 C for 15 min in the dark, 3.1. Starvation enhances melanoma cells sensitivities to cisplatin-
and 10.000 events were acquired by using a FACS Calibur cytometer induced cell death
(Becton-Dickinson, Mountain View,CA, USA). Data analysis was
performed using FlowJo software. Previous results show severe nutrient restriction (starvation)
Please cite this article in press as: F. Antunes, et al., Fasting boosts sensitivity of human skin melanoma to cisplatin-induced cell death,
Biochemical and Biophysical Research Communications (2016), http://dx.doi.org/10.1016/j.bbrc.2016.09.149
F. Antunes et al. / Biochemical and Biophysical Research Communications xxx (2016) 1e7 3
associated to chemo-therapeutic drugs regimen results in compared to individual treatments. Moreover, no PARP cleavage
enhanced cell death induction in many cancer, ameliorating clinical was observed in SK Mel 28 cells, indicating that cell death induction
outcome of many therapy resistant malignancies [12e14]. Due to could be delayed by constitutive active pro-survival pathways
the notorious resistance of human melanoma to clinical treatment characterizing the oncogenic BRAF melanoma cells [7,25].
and the absence of a definitive therapy, we decided to explore this Since the observed cleavage of PARP and positive staining with
opportunity in human skin melanoma. To this aim, we used a wild PI are indicative of an involvement of the apoptotic pathway, we
type (CHL-1) and a BRAFV600E cell line (SK Mel 28) as a model. Both decided to unveil the molecular pathway activated by the com-
cell lines were treated with CDDP (50 mM) or cultured in EBSS bined CDDP/EBSS treatment in both cell lines. Therefore, both cell
(nutrient deprivation) alone or in combination for 24 h, and cell lines were untreated or treated with CDDPþEBSS for 24 h in
death induction was evaluated by measuring the sub-G1 distribu- presence or absence of the pan-caspase inhibitor Z-VAD-FMK
tion of cells, by flow cytometric analysis of propidium iodide- (ZVAD, 50 mM), or the necroptosis inhibitor Necrostatin-1 (NEC,
stained (PI) samples. Data reported in Fig. 1A indicate CHL-1 cells 50 mM), or with the ROS (reactive oxygen species) scavenger NAC
are sensitive to both CDDP or EBSS treatment, while SK Mel 28 (N-acetyl-cysteine, 10 mM), and cell death induction was evaluated
resulting highly resistant as expected, due to the presence of by flow cytometric analysis of PI-stained samples. Interestingly, the
oncogenic BRAF [25]. However, combined exposure to both CDDP inhibition of caspases resulted in almost complete abrogation of
and EBSS resulted in consistently enhanced cell death induction in CDDPþEBSS-induced cell death, confirming apoptosis as the main
both wild-type and oncogenic BRAF melanoma cells, compared to pathway activated by combined treatment in both cell lines
individual treatments (70% and 35% respectively). To better char- (Fig. 1C). However, NEC was also able to partially inhibit the cell
acterize the synergy of combined treatment, CDDPþEBSS, on cell death induction in both cell line, indicating the involvement of
death induction we performed a time-course experiment in which other cell death pathways (Fig. 1C). Importantly, NAC displayed a
both cell lines were untreated or treated for 6 or 8 h with CDDP, significant cell death inhibition similar to ZVAD administration,
EBSS or CDDPþEBSS, and cell death induction was evaluated by thus indicating a key role of ROS in the apoptotic pathway stimu-
western blotting analysis of cleaved PARP, a well-characterized lated by the combined administration of CDDP and EBSS (Fig. 1C), in
apoptotic marker. As showed in Fig. 1B, combined treatment of both cell lines.
CHL-1 with CDDPþEBSS resulted in earlier cell death induction, Collectively these data indicate that nutrient deprivation can
Fig. 1. Starvation potentiates CDDP-induced apoptosis in melanoma cells. (A) Flow cytometric analysis of cell death (sub-G1 phase) in CHL-1 and SK Mel 28 cells treated 24 h
with EBSS (STV ¼ nutrient shortage), CDDP (50 mM) or CDDPþSTV, and stained with propidium iodide. (B) Representative immunoblots of western blotting analysis of cleaved PARP
levels in CHL-1 and SK Mel 28 cells treated as indicated. Actin was used as loading control. (C) Quantitative analyzes of cell death (sub-G1 phase) in CHL-1 and SK Mel 28 cells
treated for 24 h with STV (EBSS), CDDP (50 mM), CDDP, Z-VAD-fmk (50 mM), Necrostatin-1 (NEC, 50 mM), N-acetyl-cysteine (NAC, 10 mM), as indicated. (mean ± SD; n ¼ 3)
*p < 0.0001 compared with control cells. #p < 0.0001 compared to cisplatin. p < 0.0001 compared to CDDP plus STV. One-way ANOVA, pos-test Tukey.
Please cite this article in press as: F. Antunes, et al., Fasting boosts sensitivity of human skin melanoma to cisplatin-induced cell death,
Biochemical and Biophysical Research Communications (2016), http://dx.doi.org/10.1016/j.bbrc.2016.09.149
4 F. Antunes et al. / Biochemical and Biophysical Research Communications xxx (2016) 1e7
consistently enhance the sensitivity of tumor cells to cell death induced cell death in melanoma cells under nutrient deprivation.
induction by chemotherapeutic drugs, also of those malignancies To this aim, we evaluated the autophagic flux in cells subjected to
particularly resistant to any treatment, such as oncogenic BRAF nutrient deprivation, CDDP treatment or a combination of the two
melanomas. stimuli, evaluating the expression of the two well-known auto-
phagic markers p62 and LC3, in presence or absence of bafilomycin
3.2. Role of mitochondria and ER stress in nutrient deprivation- A1 (BAF, 5 nM). Therefore, both CHL-1 and SK Mel 28 cells were
enhanced melanoma cell death induction subjected to starvation or exposed to CDDP for 2, 4 and 6 h and p62
protein levels or LC3 conversion (LC3-II) were monitored by
Results reported in Fig. 1C clearly indicate the involvement of western blotting analysis. Data reported in Fig. 3A and B show CHL-
ROS in the apoptotic process stimulated by combined CDDP and 1 are sensitive to both CDDP or EBSS administration, while SK Mel
EBSS exposure of melanoma cells. To confirm this data, the pro- 28 were sensitive to CDDP-induced while resistant to STV-induced
duction of ROS was evaluated in both melanoma cell lines 2 h after autophagy. The latter is in line with previous results showing
CDDP, EBSS or CDDPþEBSS administration, by analyzing the fluo- intrinsic basal autophagy deregulation in oncogenic BRAF mela-
rescence of the ROS specific probe DCFDA (20 ,70 -dichlorofluorescein noma cells [7]. Interestingly, combined treatment of both mela-
diacetate) by flow cytometry. Data reported in Fig. 2A confirmed a noma cell lines with CDDP and EBSS showed no significant
significant increase in ROS production in cells exposed to combined modulation of either p62 protein levels nor LC3-II accumulation,
CDDP and EBSS. However, it is important to note the major thus indicating that autophagy is not involved in enhanced sensi-
contribution of nutrient deprivation (STV ¼ EBSS) to ROS produc- tivity of melanoma cells to combined CDDP/EBSS-induced
tion in both cell lines treated with CDDPþEBSS (compare 2nd and apoptosis (Fig. 3C). This conclusion are supported by results re-
4th histograms of each panel in Fig. 2A, respectively). ported in Fig. 3D showing no accumulation of key autophagic
ROS production during apoptosis induction/execution is often proteins Ambra1, Beclin-1 or Atg5, in the same experimental
the result of imbalanced mitochondrial homeostasis. Therefore, to conditions.
evaluate the status of this organelle, both CHL-1 and SK Mel 28 cells
were untreated or treated with CDDP, EBSS or CDDPþEBSS for 2, 4 3.4. Glucose metabolism is involved in melanoma cells survival
or 8 h, and the mitochondrial membrane potential (MMP) was
monitored by flow cytometric analysis of TMRM-stained cells. Our It is well-known that cancer cells are characterized by an altered
analysis, reported in Fig. 2B, revealed an early (2 h) pronounced metabolism, favoring glycolysis even in aerobic conditions, an
alteration of MMP in EBSS or CDDPþEBBS treated oncogenic BRAF event knows as “Warburg effect” [8]. Since cancer cells are addicted
cells (SK Mel 28, Fig. 2B right panel), possibly accounting for the to use glucose as primary source of energy and building blocks
early ROS production observed in this cell line (Fig. 2A, right panel). supplier, we asked whether nutrient deprivation (also resulting in
However, a minor effect was observed in CHL-1 cells at any time glucose shortage) might be responsible for cancer cell deregulation
point, with a major effect observed after 8 h of treatment. These of metabolism resulting in enhanced sensitivity to pro-apoptotic
data parallel the different amount of ROS produced by the two cell stimuli. To this end, we exposed our cell lines to CDDP in absence
lines and confirm nutrient deprivation (EBSS) as the main cause of or presence of 2-deoxy-glucose (2-DG, 10 mM), an irreversible in-
cellular stress. hibitor of glycolysis, supplemented to normal cell culture (high
Nutrient shortage typically results in protein synthesis attenu- glucose) or EBSS medium, and evaluated the apoptotic rates after
ation through eIF2a (eukaryotic initiation of translation) phos- 24 h. Our data indicate that inhibition of glycolysis enhanced cell
phorylation, resulting in the inhibition of cap-dependent sensitivity to pro-apoptotic stimuli (CDDP þ 2-DG) similarly to
translation while favoring the translation of IRES-containing general nutrient deprivation (CDDP þ EBSS) in BRAF wild-type cells
mRNAs, such as ATF4, and thus resulting in downstream up- (CHL-1, Fig. 4, left panel). On the other hand, oncogenic BRAF cells
regulation of this factor. The analysis of ATF4 mRNA levels in both seems to be more sensitive to combined EBSS þ 2-DG compared to
melanoma cell lines upon CDDP or EBSS treatment indeed revealed CDDP þ 2-DG treatment, possibly indicating that the glucose
an enhanced expression of this factor under nutrient deprivation metabolism is fundamental for the survival or these cancer cells.
(STV ¼ EBSS) alone or in combination with CDDP (Fig. 2C). Since Importantly, the exposure to 2-DG further enhanced the apoptotic
CDDP alone resulted in a slight up-regulation of ATF4, that may also rate observed in SK Mel 28 cells upon treatment with both CDDP
indicate the induction of an ER stress response, we asked if this and EBSS, indicating that a ‘complete’ nutrient deprivation might
process was involved in the apoptotic pathway stimulated by consistently boost pro-apoptotic induction in the notoriously
combined treatment of melanoma cells with CDDP and EBSS. To resistant oncogenic BRAF melanoma cells (Fig. 4, right panel,
answer this question we evaluated the expression of two other ER compare the two right-most histograms).
stress marker, ATF6 and XBP1, in the same experimental conditions. Several studies in animal models have suggested that calorie
Results reported in Fig. 2C (right panels) evidenced no significant restriction is able to significantly increase the life span and decrease
modulation of ATF6 while a slight up-regulation of XBP1 in SK Mel the risk of cancer development [28]. The data presented in this
28 in response to CDDP. Collectively these data indicate a minor or study confirm this hypothesis indicating that nutrient deprivation
no involvement of ER stress in our model. increases the efficacy of CDDP treatment in melanoma irre-
spectively of the BRAF status. Particularly, interesting is the effect of
3.3. Autophagy does not contribute to nutrient deprivation the combined treatment with 2-DG that is able to induce similar
enhanced cell death sensitivity of melanoma cells to CDDP cell death levels both wild type and mutated BRAF cells. However,
further in vivo experiments are required to better define the
Autophagy is a catabolic process promptly activated by cells regulation/deregulation of glucose metabolism in wild-type and
under several stress conditions such as hypoxia, infection, nutrient oncogenic BRAF melanoma cells, and its role in their survival under
deprivation and in presence of cytotoxic compounds, such as CDDP stress conditions [29e30].
[26]. Although it represents primarily a pro-survival process, pro-
longed or deregulated autophagy may result or contribute to cell Conflict of interests
death induction/execution [27]. Therefore, we asked whether
autophagy was involved in the enhanced cell sensitivity to CDDP- None.
Please cite this article in press as: F. Antunes, et al., Fasting boosts sensitivity of human skin melanoma to cisplatin-induced cell death,
Biochemical and Biophysical Research Communications (2016), http://dx.doi.org/10.1016/j.bbrc.2016.09.149
F. Antunes et al. / Biochemical and Biophysical Research Communications xxx (2016) 1e7 5
Fig. 2. Involvement of mitochondria and ER stress on CDDP/STV induced cell death of melanoma cells. (A) Quantitative analyzes of DCFDA fluorescence by flow cytometry in
CHL-1 and SK Mel 28 cells treated for 2 h with STV, CDDP (50 mM) and CDDP plus STV (B) Measuring of DJm alteration with TMRM by flow cytometry in CHL-1 and SK Mel 28 cells
treated for 2e4 and 8 h with STV, CDDP (50 mM) and CDDP plus STV. CCCP (50 mM) was added as a positive control to loss of DJm. (C) Gene expression analyzes by qPCR of ATF4,
ATF6 and XBP-1 in CHL-1 and SK Mel 28 cells treated for 6e8 h with STV, CDDP (50 mM) and CDDP plus STV. (mean ± SD; n ¼ 3) *p < 0.005 compared with control cells. One-way
ANOVA, pos-test Tukey.
Please cite this article in press as: F. Antunes, et al., Fasting boosts sensitivity of human skin melanoma to cisplatin-induced cell death,
Biochemical and Biophysical Research Communications (2016), http://dx.doi.org/10.1016/j.bbrc.2016.09.149
6 F. Antunes et al. / Biochemical and Biophysical Research Communications xxx (2016) 1e7
Fig. 3. Analysis of autophagy induced by EBSS or CDDP in melanoma cells. p62 and LC3-II western blotting analysis of CHL-1 and SK Mel 28 cells untreated or treated 2, 4 or 6 h
with EBSS (STV) (A) or CDDP (50 mM) (B), (BAF, 5 nM). (C) p62 and LC3-II western blotting analysis of CHL-1 and SK Mel 28 cells treatment for 6 h with STV (EBSS), CDDP (50 mM),
CDDP plus STV, in presence or absence of bafilomycin A1. (D) Western blotting analysis of the autophagic proteins Ambra1, Beclin-1 and Atg5 in CHL-1 and SK Mel 28 cells treated
for 6e8 h with STV, CDDP (50 mM) or a combination. GAPDH was used as loading control.
Fig. 4. Glucose metabolism is involved in melanoma cells survival. Quantitative analysis of apoptosis (sub-G1 phase) in CHL-1 and SK Mel 28 cells treated with STV, or CDDP
(50 mM) as indicated, in presence or absence of 2-deoxy-glucose (2-DG e 10 mM) for 24 h. Cells were stained with propidium iodide and analyzed by flow cytometry (mean ± SD;
n ¼ 3) *p < 0.0001 compared with control cells. One-way ANOVA, pos-test Tukey. #p < 0.0001 compared to cisplatin. Two-way ANOVA, post-test Bonferroni.
Please cite this article in press as: F. Antunes, et al., Fasting boosts sensitivity of human skin melanoma to cisplatin-induced cell death,
Biochemical and Biophysical Research Communications (2016), http://dx.doi.org/10.1016/j.bbrc.2016.09.149
F. Antunes et al. / Biochemical and Biophysical Research Communications xxx (2016) 1e7 7
Please cite this article in press as: F. Antunes, et al., Fasting boosts sensitivity of human skin melanoma to cisplatin-induced cell death,
Biochemical and Biophysical Research Communications (2016), http://dx.doi.org/10.1016/j.bbrc.2016.09.149