In vitro prediction of the efficacy of molecularly targeted cancer therapy by Raman spectral imaging
- First Online:
Mutational acquired resistance is a major challenge in cancer therapy. Somatic tumours harbouring some oncogenic mutations are characterised by a high mortality rate. Surprisingly, preclinical evaluation methods do not show clearly resistance of mutated cancers to some drugs. Here, we implemented Raman spectral imaging to investigate the oncogenic mutation resistance to epidermal growth factor receptor targeting therapy. Colon cancer cells with and without oncogenic mutations such as KRAS and BRAF mutations were treated with erlotinib, an inhibitor of epidermal growth factor receptor, in order to detect the impact of these mutations on Raman spectra of the cells. Clinical studies suggested that oncogenic KRAS and BRAF mutations inhibit the response to erlotinib therapy in patients, but this effect is not observed in vitro. The Raman results indicate that erlotinib induces large spectral changes in SW-48 cells that harbour wild-type KRAS and BRAF. These spectral changes can be used as a marker of response to therapy. HT-29 cells (BRAF mutated) and SW-480 cells (KRAS mutated) display a smaller and no significant response, respectively. However, the erlotinib effect on these cells is not observed when phosphorylation of extracellular-signal-regulated kinase and AKT is monitored by Western blot, where this phosphorylation is the conventional in vitro test. Lipid droplets show a large response to erlotinib only in the case of cells harbouring wild-type KRAS and BRAF, as indicated by Raman difference spectra. This study shows the great potential of Raman spectral imaging as an in vitro tool for detecting mutational drug resistance.
KeywordsRaman imaging Erlotinib Oncogenic mutation BRAF KRAS Colon cancer
Cancer is one of the leading causes of death worldwide. Statistics from the World Health Organization indicate that 8.2 million cases of cancer death were registered in 2012 . This high number of cancer deaths is due to the failure of many therapeutic strategies to suppress tumour progression. Some oncogenic mutations render tumours aggressive, which is accompanied by poor prognosis and a decreased response to therapy [2, 3, 4]. Therefore, researchers in drug development are competing to find new potent anticancer drugs that can overcome the mutational drug resistance. Hundreds of new anticancer drugs appear each year. To evaluate their potencies, they are studied in a series of in vitro and in vivo preclinical tests. The promising candidates are then proceeded to clinical trials, to be tested on human cancer patients . However, preclinical methods do not predict the resistance of some oncogenic mutations to therapy, which is later detected in many clinical studies. For instance, inhibitors of epidermal growth factor receptor (EGFR) showed great therapeutic potential in solid tumour suppression . However, some of them failed in patients with KRAS-mutated cancers .
RAS is a membrane-anchored guanosine triphosphatase protein . It controls multiple downstream signalling pathways and acts as an on/off signalling switch (Fig. 1). Activated EGFR sets RAS to on mode, which in turn activates multiple effector proteins such as rapidly accelerated fibrosarcoma (RAF), phosphatidylinositiol 3-kinase (PI3K), TIAM, phospholipase Cε (PCLε), and RAL for cell proliferation and survival [17, 18, 19]. In the case of EGFR inhibition, RAS remains in off mode, and the cell cycle is arrested . However, mutated RAS is chronically switched to on mode, and it becomes independent of EGFR activation. Therefore, RAS-mutated tumours are aggressive, invasive, and metastatic. Moreover, it is accompanied by a decreased response to EGFR-inhibition therapy [4,7,17,18,21, 22, 23, 24, 25]. KRAS mutation accounts for 86 % of all RAS mutations . It is commonly found in the most deadly cancer types: lung (33 %), colon (40 %), and pancreatic (90 %) cancers [27,28]. KRAS mutation also is reported to be predictive for poor prognosis and low survival rate in cancer [22,25,29].
RAF is one of the well-identified RAS effector proteins, with serine/threonine kinase activity . RAF is activated by binding with the active form of RAS, which subsequently stimulates the mitogen-activated protein kinase–extracellular-signal-regulated kinase (ERK) pathway proteins through a cascade of autophosphorylation events towards cell proliferation (Fig. 1) . BRAF mutation occurred in up to 80 % of skin cancers and 5-10 % of colon cancers . In addition, BRAF mutation is accompanied by an elevated kinase activity, which increases ERK phosphorylation [32,33]. Clinical studies showed that cancer patients with BRAF mutation have a relatively poor prognosis [25,34].
Oncogenic mutations are commonly detected by DNA sequencing and methods based on polymerase chain reaction [35,36]. In vitro assessment of drug effects is done separately by cytotoxicity assays . Although, some in vivo methods such as genetically engineered cancer models revealed promising results in detecting drug resistance to mutations [37,38], they still have some drawbacks, such as requiring a long time and unpredictability of tumour formation . To the best of our knowledge, no in vitro method has reported so far the impact of oncogenic mutations on response to EGFR molecularly targeted therapy.
Raman micro-spectroscopic imaging is an emerging technique in biomedical research. Raman spectroscopy can measure biological samples in an aqueous physiological environment. It is a label-free, non-invasive technique with high spectral/lateral resolution and great reproducibility [40, 41, 42, 43]. Raman spectral imaging can classify cancerous human tissues [44,45]. It can be used for imaging of single cells and characterisation of subcellular components [46, 47, 48, 49, 50]. Furthermore, Raman imaging can be conducted to monitor drug uptake and its effect on single cells [51, 52, 53, 54, 55, 56, 57]. In our previous work we investigated the distribution and metabolism of erlotinib in SW-480 colon cancer cells using its unique –C ≡ C– band at 2100 cm−1, which is used as a marker band for erlotinib localisation .
Here, we implemented Raman imaging coupled with hierarchical cluster analysis (HCA) to monitor the response of colon cancer cells to erlotinib therapy. We report in vitro evidence that detects the effect of oncogenic KRAS and BRAF mutations on the cellular response to erlotinib. The Raman results show that colon cancer cells experience a large spectral response to erlotinib, but colon cancer cells expressing oncogenic BRAF or KRAS mutations experience small or no relevant effects, respectively. Furthermore, the largest effect is observed in lipid droplets of cancer cells harbouring wild-type BRAF and KRAS that were treated with erlotinib.
Material and methods
The colon cancer cell lines SW-48, HT-29, and SW-480 were purchased from American Type Culture Collection. Cells were cultured in Dulbecco’s modified Eagle’s medium (Life Technologies, Darmstadt, Germany) supplemented with 10 % fetal bovine serum (Life Technologies, Darmstadt, Germany), 2 mM l-glutamine, and 5 % penicillin–streptomycin, and were incubated at 37 °C in a 10 % CO2 atmosphere. Cells were subcultured to 80 % confluence, detached by trypsin–EDTA (0.25 %) (Gibco trypsin solution, Life Technologies, Darmstadt, Germany), centrifuged at 1500 rpm for 3 min and diluted to 10 %, then seeded again in culture medium. Raman measurements were performed on cells grown on CaF2 windows (Korth Kristalle, Kiel, Germany) to avoid Raman scattering from regular glass slides. Cells were incubated with erlotinib (Tarceva; Roche, Switzerland) at 10 μg/ml at 37 °C in 10 % CO2 for 12 h. Subsequently, cells were fixed in 4 % paraformaldehyde (VWR International, Darmstadt, Germany) and then submerged in phosphate-buffered saline (Life Technologies, Darmstadt, Germany).
Confocal Raman microscopy
Cancer cells were measured with an alpha300 AR confocal Raman microscope (WITec, Ulm, Germany) as described previously [45,50,53]. Briefly, the setup excitation source was a frequency-doubled Nd:YAG laser of 532 nm (CrystaLaser, Reno, NV, USA) with an output power of around 40 mW. The laser radiation was coupled into a Zeiss microscope through a wavelength-specific single-mode optical fibre. An achromatic lens used as a collimator for the laser beam, and a holographic band-pass filter focused the beam on the sample through a Nikon NIR APO (×60/1.00 numerical aperture) water immersion objective. The sample was located on a piezoelectrically driven microscope scanning stage, which has an x,y resolution of 3 nm and a z resolution of 0.3 nm. Raman back-scattered light was collected through a microscopic objective and passed through a holographic edge filter into a multimode fibre (50-μm diameter) and into a 300-mm focal length monochromator, which incorporated a 600/mm grating blazed at 500 nm. A back-illuminated deep-depletion charge-coupled device camera operating at −60 °C was used to detect Raman spectra. We acquired Raman imaging measurements by raster scanning the laser beam over cells and collecting a full Raman spectrum at speed of 0.5 s per pixel, with a pixel resolution of 500 nm.
HCA and principal component analysis (PCA) were performed on Raman spectra of cells. Raman hyperspectral results were exported to MATLAB 8.2 (The MathWorks, Natick, MA, USA). Scripts written in-house were used to perform data preprocessing and multivariate analyses. Raman spectra without a C–H band at 2850–3000 cm−1 were treated as the background and were deleted. An impulse noise filter was used to remove cosmic spikes, and the Raman spectra were interpolated to a reference wavenumber scale. All spectra were then baseline corrected with a third-order polynomial and were also vector normalised. HCA was performed on the regions from 700 to 1800 cm−1 and 2800 to 3050 cm−1 with use of Ward’s clustering in combination with the Pearson correlation distance. All Raman mean spectra obtained from HCA were normalised with the phenylalanine band near 1008 cm−1. We calculated Raman difference spectra of cells by subtracting the mean spectrum of erlotinib-treated cells from that of control cells for each component identified by HCA. PCA was performed on the mean spectra for all subcellular components identified by HCA of both control and erlotinib-treated cells, and the first three PCs generated denote the higher variances within the data set . PC1 loadings of respective subcellular components of control and erlotinib-treated cells were paired for comparison.
Phosphatase inhibitor/buffer mixture II (Sigma-Aldrich, Munich, Germany) and protease inhibitor cocktail (Roche) were used to harvest the cells. They were resolved by sodium dodecyl sulfate–10 % polyacrylamide gel electrophoresis, and transferred to Immobilon-P membranes (Millipore, Schwalbach am Taunus, Germany). The membranes were then incubated with antibodies to activate phosphorylated ERK1 and ERK2 (p-ERK; Cell Signaling Technology, Danvers, MA, USA), total ERK1 and ERK2 (Cell Signaling Technology, Danvers, MA, USA), phosphorylated AKT (p-AKT; Cell Signaling Technology, Danvers, USA), and total AKT (Cell Signaling Technology, Danvers, USA). Antibodies were detected with the appropriate anti-mouse horseradish peroxidase conjugated secondary antibody enhanced by chemiluminescence (Pierce, Life Technologies, Darmstadt, Germany), and images were captured with a Versa Doc 5000 imaging system (Bio-Rad, Munich, Germany).
Results and discussion
Downstream signalling and oncogenic mutations by Western blot
Furthermore, no significant difference in p-ERK level was observed for erlotinib-treated SW-48, HT-29 and SW-480 cells compared with the control. This is perhaps due to the activating mutations of BRAF in HT-29 cells and KRAS in SW-480 cells as well as the EGFR mutation in SW-48 cells. Similar results were observed for the p-AKT level for erlotinib-treated SW-48 and SW-480 cells. On the other hand, the p-AKT level was reduced in erlotinib-treated HT-29 cells. This means that the Western blotting results did not significantly discriminate between the KRAS and BRAF status in these cells. These results are also in agreement with a study conducted to detect the effect of erlotinib on p-ERK and p-AKT in a series of pancreatic cancer cell lines with wild-type and mutated KRAS .
Effect of erlotinib on SW-48 cells by Raman spectral imaging
Erlotinib is a reversible competitive inhibitor of the TK domain of EGFR that binds to its adenosine 5′-triphosphate binding site. To detect the effect of erlotinib binding to the TK domain, several Raman measurements of SW-48 cells were performed. Cells were incubated with erlotinib at 10 μg/ml for 12 h. The concentration of erlotinib used in this study is similar to that detected in the plasma of cancer patients taking this drug. The oral dosage of erlotinib in clinics is 150 mg/day, and the erlotinib concentration in the plasma was found to be 2.5–29 μg/ml . No contribution from erlotinib to the Raman spectra itself was expected because of its low concentration, which cannot be detected under our experimental conditions.
Next, HCA was performed in the spectral ranges from 700 to 1800 cm−1 and 2800 to 3100 cm−1 in order to construct a cluster map based on the Raman intensity image (Fig. 4, panel a); the result is shown in Fig. 4, panel b. The clusters were then merged to four major subcellular components: plasma membrane, cytoplasm, nucleus, and lipid droplets. This merging process was based on use of the spectral marker bands of each cell component. Afterwards, mean spectra of the four clusters were calculated (Fig. 4, spectra c–f). The outermost cluster (Fig. 4, spectrum c) is considered to contain the plasma membrane. However, the plasma membrane is smaller than the laser spot (less than 1 μm); therefore, the membrane cluster might contain contributions from the buffer and cytoplasm (Fig. 4, spectrum d). The mean spectrum of the nucleus (Fig. 4, spectrum e) contains characteristic Raman bands such as the stretching vibration of DNA phosphodiester at 790 cm−1. The spectrum of the cytoplasm (Fig. 4, spectrum d) and that of the nucleus (Fig. 4, spectrum e) both contain the Raman spectral band of the ring-breathing mode of phenylalanine at 1008 cm−1, which represents their protein constituent. However, cytoplasm spectra do not contain the DNA marker band at 790 cm−1 [55, 56, 57,70]. The Raman mean spectrum of lipid droplets (Fig. 4, spectrum f) shows characteristic bands of the C = O stretching vibration of the ester form of fatty acids at 1750 cm−1 and the = C―H stretching vibration of unsaturated fatty acids at 3018 cm−1. The spectra of the nucleus and lipid droplets are consistent with those reported previously [42,43,49,50].
Resistance of oncogenic mutations to erlotinib therapy by Raman spectral imaging
It is worth mentioning that BRAF mutation affects mainly one signalling pathway for cell proliferation, whereas KRAS mutation affects multiple pathways (Fig. 1); therefore, the impact of BRAF mutation is less than that of KRAS mutation. In agreement with the present Raman results, clinical studies on human patients harbouring BRAF- or KRAS-mutated tumours have shown a limited response to TK inhibitors [22,25,29,34,78]. On the other hand, the Western blot results could not discriminate between erlotinib-treated cells harbouring wild-type KRAS and BRAF and those with KRAS and BRAF mutations through monitoring the ERK phosphorylation. In addition, the AKT phosphorylation is inhibited only in cells with BRAF mutations. In Western blot, antibodies are used to mark single proteins although cellular proteins are highly coupled in protein networks, and monitoring of a single protein may not reflect the in vivo cellular response to drugs such as erlotinib. This might explain the differences between the findings of in vitro drug studies and the drug effect observed in cancer patients. On the other hand, Raman microscopy reflects the overall chemical composition of the cell, and Raman difference spectra can detect the changes in the whole cell contents, which would be an efficient way to exhibit the overall cellular and subcellular component changes upon drug treatment.
Developing a new preclinical method that would predict the resistance of mutated cancers to therapy can reduce the cost, time, and effort spent in the new drug discovery process. It is one step further to increase the accuracy of in vitro drug evaluation, before these drugs proceed to clinical trials to be tested on human patients. In fact, clinical trials suffer from many ethical and economical limitations . Consequently, increasing the accuracy of in vitro evaluation tests would have a significant impact on accelerating the process of new anticancer drug evaluation and final approval.
Raman microscopy coupled with clustering analysis such as HCA can be used to detect the subcellular changes induced by drug treatment. Raman difference spectra can also emphasise the oncogenic mutational drug resistance in EGFR-targeted therapy. Erlotinib resistance in BRAF- and KRAS-mutated colon cancers is clearly reflected in their Raman difference spectra. Cancer cells with wild-type KRAS and BRAF show clear changes upon erlotinib treatment, whereas BRAF- and KRAS-mutated cancer cells display a limited change or no significant changes, respectively. Consistently, clinical studies revealed that human cancer patients with BRAF- or KRAS-mutated tumours have the worst therapy prognosis. Accordingly, in vitro Raman imaging is a valuable addition that can enhance the sensitivity of the preclinical evaluation methods for new anticancer agents.
We thank Carsten Kötting for helpful discussions. We also thank Dennis Petersen for his assistance in the PCA. This study was supported by the Protein Research Unit Ruhr Within Europe (PURE), the Ministry of Innovation, Science and Research (MIWF) of North Rhine–Westphalia, Germany, and the Center for Vibrational Microscopy, European Regional Development Fund, the European Union, and North Rhine–Westphalia, Germany, and the Deutsche Forschungsgemeinschaft under grant number: SFB 642. This study was also supported by a fellowship to H.K.Y. from the Konrad Adenauer Foundation.
Conflict of interest
The authors declare that they have no conflict of interest.
- 1.International Agency for Research on Cancer, World Health Organization (2014) World cancer report 2014. International Agency for Research on Cancer, LyonGoogle Scholar
- 5.Zips D, Thames HD, Baumann M (2005) New anticancer agents: in vitro and in vivo evaluation. In Vivo 19:1–7Google Scholar
- 6.Goel S, Hidalgo M, Perez-Soler R (2007) EGFR inhibitor-mediated apoptosis in solid tumors. J Exp Ther Oncol 6:305–320Google Scholar
- 13.Marchetti A, Milella M, Felicioni L, Cappuzzo F, Irtelli L, Del Grammastro M, Sciarrotta M, Malatesta S, Nuzzo C, Finocchiaro G, Perrucci B, Carlone D, Gelibter AJ, Ceribelli A, Mezzetti A, Iacobelli S, Cognetti F, Buttitta F (2009) Clinical implications of KRAS mutations in lung cancer patients treated with tyrosine kinase inhibitors: an important role for mutations in minor clones. Neoplasia 11:1084–1092CrossRefGoogle Scholar
- 14.Brugger W, Triller N, Blasinska-Morawiec M, Curescu S, Sakalauskas R, Manikhas GM, Mazieres J, Whittom R, Ward C, Mayne K, Trunzer K, Cappuzzo F (2011) Prospective molecular marker analyses of EGFR and KRAS from a randomized, placebo-controlled study of erlotinib maintenance therapy in advanced non-small-cell lung cancer. J Clin Oncol 29:4113–4120. doi:10.1200/JCO.2010.31.8162 CrossRefGoogle Scholar
- 18.Bos JL (1989) ras oncogenes in human cancer: a review. Cancer Res 49:4682–4689Google Scholar
- 21.Amado RG, Wolf M, Peeters M, Van Cutsem E, Siena S, Freeman DJ, Juan T, Sikorski R, Suggs S, Radinsky R, Patterson SD, Chang DD (2008) Wild-type KRAS is required for panitumumab efficacy in patients with metastatic colorectal cancer. J Clin Oncol 26:1626–1634. doi:10.1200/JCO.2007.14.7116 CrossRefGoogle Scholar
- 23.Benvenuti S, Sartore-Bianchi A, Di Nicolantonio F, Zanon C, Moroni M, Veronese S, Siena S, Bardelli A (2007) Oncogenic activation of the RAS/RAF signaling pathway impairs the response of metastatic colorectal cancers to anti-epidermal growth factor receptor antibody therapies. Cancer Res 67:2643–2648. doi:10.1158/0008-5472.CAN-06-4158 CrossRefGoogle Scholar
- 31.Borràs E, Jurado I, Hernan I, Gamundi M, Dias M, Martí I, Mañé B, Arcusa À, Agúndez JA, Blanca M, Carballo M (2011) Clinical pharmacogenomic testing of KRAS, BRAF and EGFR mutations by high resolution melting analysis and ultra-deep pyrosequencing. BMC Cancer 11:406. doi:10.1186/1471-2407-11-406 CrossRefGoogle Scholar
- 35.Kan Z, Jaiswal BS, Stinson J, Janakiraman V, Bhatt D, Stern HM, Yue P, Haverty PM, Bourgon R, Zheng J, Moorhead M, Chaudhuri S, Tomsho LP, Peters BA, Pujara K, Cordes S, Davis DP, Carlton VEH, Yuan W, Li L, Wang W, Eigenbrot C, Kaminker JS, Eberhard DA, Waring P, Schuster SC, Modrusan Z, Zhang Z, Stokoe D, de Sauvage FJ, Faham M, Seshagiri S (2010) Diverse somatic mutation patterns and pathway alterations in human cancers. Nature 466:869–873. doi:10.1038/nature09208 CrossRefGoogle Scholar
- 36.Oxnard GR, Paweletz CP, Kuang Y, Mach SL, O’Connell A, Messineo MM, Luke JJ, Butaney M, Kirschmeier P, Jackman DM, Janne PA (2014) Noninvasive detection of response and resistance in EGFR-mutant lung cancer using quantitative next-generation genotyping of cell-free ülasma DNA. Clin Cancer Res 20:1698–1705. doi:10.1158/1078-0432.CCR-13-2482 CrossRefGoogle Scholar
- 41.Popp J, Tuchin VV, Chiou A, Heinemann S (eds) (2011) Handbook of biophotonics. Volume 2: photonics for health care. Wiley-VCH, WeinheimGoogle Scholar
- 45.Mavarani L, Petersen D, El-Mashtoly SF, Mosig A, Tannapfel A, Kötting C, Gerwert K (2013) Spectral histopathology of colon cancer tissue sections by Raman imaging with 532 nm excitation provides label free annotation of lymphocytes, erythrocytes and proliferating nuclei of cancer cells. Analyst 138:4035–4039. doi:10.1039/c3an00370a CrossRefGoogle Scholar
- 62.Vartanian S, Bentley C, Brauer MJ, Li L, Shirasawa S, Sasazuki T, Kim J-S, Haverty P, Stawiski E, Modrusan Z, Waldman T, Stokoe D (2013) Identification of mutant K-Ras-dependent phenotypes using a panel of isogenic cell lines. J Biol Chem 288:2403–2413. doi:10.1074/jbc.M112.394130 CrossRefGoogle Scholar
- 65.Salzer R, Siesler HW (eds) (2009) Infrared and Raman spectroscopic imaging. Wiley, WeinheimGoogle Scholar
- 66.Diem M, Chalmers JM, Griffiths PR (eds) (2008) Vibrational spectroscopy for medical diagnosis. Wiley, ChichesterGoogle Scholar
- 68.Ali S, Banerjee S, Ahmad A, El-Rayes BF, Philip PA, Sarkar FH (2008) Apoptosis-inducing effect of erlotinib is potentiated by 3,3’-diindolylmethane in vitro and in vivo using an orthotopic model of pancreatic cancer. Mol Cancer Ther 7:1708–1719. doi:10.1158/1535-7163.MCT-08-0354 CrossRefGoogle Scholar
- 69.Ling Y-H, Lin R, Perez-Soler R (2008) Erlotinib induces mitochondrial-mediated apoptosis in human H3255 non-small-cell lung cancer cells with epidermal growth factor receptorL858R mutation through mitochondrial oxidative phosphorylation-dependent activation of BAX and BAK. Mol Pharmacol 74:793–806. doi:10.1124/mol.107.044396 CrossRefGoogle Scholar
- 73.Tirinato L, Liberale C, Di Franco S, Candeloro P, Benfante A, La Rocca R, Potze L, Marotta R, Ruffilli R, Rajamanickam VP, Malerba M, De Angelis F, Falqui A, Carbone E, Todaro M, Medema JP, Stassi G, Di Fabrizio E (2014) Lipid droplets: a new player in colorectal cancer stem cells unveiled by spectroscopic imaging. Stem Cells. doi:10.1002/stem.1837 Google Scholar
- 75.Doherty KR, Wappel RL, Talbert DR, Trusk PB, Moran DM, Kramer JW, Brown AM, Shell SA, Bacus S (2013) Multi-parameter in vitro toxicity testing of crizotinib, sunitinib, erlotinib, and nilotinib in human cardiomyocytes. Toxicol Appl Pharmacol 272:245–255. doi:10.1016/j.taap.2013.04.027 CrossRefGoogle Scholar
- 76.Hülsmann HJ, Rolff J, Bender C, Jarahian M, Korf U, Herwig R, Fröhlich H, Thomas M, Merk J, Fichtner I, Sültmann H, Kuner R (2014) Activation of AMP-activated protein kinase sensitizes lung cancer cells and H1299 xenografts to erlotinib. Lung Cancer 86:151–157. doi:10.1016/j.lungcan.2014.09.001 CrossRefGoogle Scholar
- 78.Ulivi P, Delmonte A, Chiadini E, Calistri D, Papi M, Mariotti M, Verlicchi A, Ragazzini A, Capelli L, Gamboni A, Puccetti M, Dubini A, Burgio M, Casanova C, Crinò L, Amadori D, Dazzi C (2014) Gene mutation analysis in EGFR wild type NSCLC responsive to erlotinib: are there features to guide patient selection? Int J Mol Sci 16:747–757. doi:10.3390/ijms16010747 CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.