MALDI-MS Protein Profiling of Chemoresistance in Extracellular

Letter Cite This: Anal. Chem. XXXX, XXX, XXX−XXX

pubs.acs.org/ac

MALDI-MS Protein Profiling of Chemoresistance in Extracellular Vesicles of Cancer Cells Gerald Stübiger,*,†,‡,∇ Michael D. Nairn,§,∇ Tom K. Abban,§ Matthew E. Openshaw,§ Luis Mancera,∥ Barbara Herzig,† Michael Wuczkowski,† Daniel Senfter,⊥ and Robert M. Mader#,‡

Anal. Chem. Downloaded from pubs.acs.org by UNIV OF SUNDERLAND on 11/02/18. For personal use only.

†

Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria ‡ Comprehensive Cancer Center of the Medical University of Vienna, Spitalgasse 23, 1090 Vienna, Austria § Shimadzu/Kratos Analytical, Trafford Wharf Road, M17 1GP Manchester, United Kingdom ∥ Clover Bioanalytical Software SL, Avda De La Innovacion 1, 18016 Granada, Spain ⊥ Department of Paediatrics, Molecular Neuro-Oncology Research Unit, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria # Department of Medicine I, Clinical Division of Oncology, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria S Supporting Information *

ABSTRACT: Cancer cells communicate with the whole organism via extracellular vesicles (EVs), which propagate molecular information in support of the malignant phenotype. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) was employed for protein profiling of EVs derived from CCL-228 as the primary colon tumor, the lymph node metastasis CCL-227, and subclones resistant to 5, 25, and 125 μM 5-fluorouracil (FU). EVs were harvested from cell culture supernatant by ultracentrifugation to serve as a model for circulating cancer cell-derived biomarker carriers from body fluids (i.e., liquid biopsy). Protein mass spectra were recorded using standard MALDI matrixes (e.g., CHCA, sinapinic acid) in the range m/z 2000−20000 on different MALDI-TOF-MS systems and subjected to multivariate data analysis . By using hierarchical clustering, PCA and PLS-DA, discriminatory protein patterns of the EVs from the different cell populations were obtained. Peaks in the range m/z 2000−6500 and m/z 5500−15500 were found to be unique to EVs and the cells, respectively. This clearly demonstrates the differential expression of proteins in EVs as the result of an increasing chemoresistance of their parent cells. The sensitivity of the MALDI-MS based assay was in the low μg/mL (≈1.2−5 × 1010 particles/mL) range. Consequently, our MALDI-MS protein profiling approach shows the potential to serve as novel tool for minimally invasive cancer diagnostics and chemotherapy monitoring in the future, e.g., for early detection of therapy resistance without biopsy.

M

In recent years, an increasing number of cancer researchers have turned their attention to the analysis of biomarkers present in biological fluids. This highly dynamic field of research is called “liquid biopsy”. Thereby, a specific focus is circulating tumor cells (CTCs), circulating tumor DNA (ctDNA), and extracellular vesicles (EVs). The latter representing cellular organelles in the nanometer range derived from the plasma membrane or the endosome. A specific type of EV is exosomes representing small vesicles (30−200 nm) that are known to contain DNA, mRNA, miRNA, proteins, lipids, and viruses.3,4 By transferring this molecular information

atrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) represents a very robust, sensitive, and high-throughput technology that allows the simultaneous detection of hundreds of molecules over a broad mass range (100 kDa). These features made MALDI-MS the method of choice for the rapid identification of microorganisms based on recording of characteristic protein profiles (i.e., biotyping). This initiated a paradigm shift in clinical microbiology during the past decade.1 So far, MALDIMS based profiling is almost exclusively used in this field of clinical application. However, its favorable features make it a very interesting platform for a broader range of applications.2 Thus, MALDI-MS is expected to revolutionize other areas of clinical application like cancer research and diagnostics in the near future. © XXXX American Chemical Society

Received: August 18, 2018 Accepted: October 26, 2018

A

DOI: 10.1021/acs.analchem.8b03756 Anal. Chem. XXXX, XXX, XXX−XXX

Letter

Analytical Chemistry

Figure 1. MALDI mass spectra of extracellular vesicles (EVs) from CCL-228 cells recorded in the range m/z 2000−14000 using CHCA matrix (10 mg/mL) with (A) the MALDI-8020 and (B) the AXIMA-Performance platform. In (C) the MALDI mass spectra of proteins from the EVs recorded in the range 1.5−400 μg/mL and in (D) the mean signal-to-noise (S/N) ratios of the eight most abundant peaks in the range m/z 3000− 16000 (indicated by asterisks) at each of the measured protein concentrations are shown. Six of the eight peaks could be detected at 6 μg/mL with an average S/N ratio of ∼3. Values represent mean ± standard deviation (SD) of three independent measurements (n = 3). Particle numbers from μg protein were calculated according to the theoretical mass of exosomes according to literature.7

wide molecular repertoire is taking place during chemoresistance development.10,11 More recently, it could be demonstrated that exosomes were able to transfer regulatory RNAs (e.g., miR-200) between CCL-227 and adjacent cells.12 A proteomics study on EVs from breast cancer cells identified a panel of proteins related to the generation of therapeutic induced senescence (TIS), which allows tumor cells to successfully resist chemotherapy.13 This clearly demonstrates the importance of EVs as carriers of molecular information from cancer cells and their involvement in cell signaling and modulation of gene expression during the development of anticancer drug resistance. The aim of our present study was to evaluate the potential of MALDI-MS protein profiling to monitor chemoresistance directly at the level of cells and EVs, which are readily available in clinical samples (e.g., exosomes). The same in vitro cancer model based on a primary colon cancer (CCL-228) and its lymph node metastasis cell line (CCL-227) together with subclones resistant to 5, 25, and 125 μM FU was used as cited above.14 For quality control, the final EV-containing ultracentrifugation fraction was measured by using nanoparticle tracking analysis (NTA) and dynamic light scattering (DLS), respectively (see Experimental details in the Supporting

between cells, exosomes are able to reprogram other cells in the organism to support malignant progression.5 They are secreted by all cell types in culture and are found in vivo in many body fluids (blood, urine, saliva, cerebrospinal and epididymal fluid, amniotic liquid, breast milk, etc.) containing a multitude of potential biomarker molecules related to their cellular origin.3,6 The concentration of circulating exosomes in human plasma can vary from 108 to 1010 particles/mL, which can increase up to 1014 particles/mL in plasma samples of cancer patients.7 Thus, they represent very promising sources of novel biomarkers, attractive targets for noninvasive diagnostics and vectors of therapy.8 One of the most important obstacles to cancer chemotherapy is the development of resistance to anticancer drugs. 5Fluorouracil (FU) is one of the most widely used agents, particularly in the treatment of colorectal cancer. One crucial question is how resistance emerges when initially sensitive cells develop increasingly chemoresistant clones. In a previous study, FU-resistant subclones of colon cancer cells (CCL-227) were generated by continuous exposure to increasing concentrations of FU over a time period of >2 years.9 With this cellular model, it could be shown by gene expression arrays (e.g., GeneChip), real-time PCR, and proteomic analysis that a B


Letter


Figure 2. Representative 3-D plots of the processed protein mass spectra of (A) cells and (B) EVs from the different cell populations recorded with standard MALDI-TOF/RTOF platform. In (C) dendrograms of a hierarchical clustering analysis using Clover MS Software from cells and EVs of the cell lines CCL-228 and CCL-227 and the corresponding 5, 25, and 125 μM 5-fluorouracil (FU) resistant subclones are shown. Samples of three independent cell culture experiments (#1−3) were used for analyis (see Figure S-2C,D).

Taking the eight most abundant peaks (m/z 3453 ± 3, 3776 ± 4, 4605 ± 5, 5663 ± 6, 6104 ± 6, 6905 ± 7, 11320 ± 11, 13799 ± 14) into account, informative mass spectra could be recorded down to the lower μg/mL range with a limit of detection (LOD) above 6 μg/mL (≈1.2 × 1010 particles/mL) (Figure 1C,D). This corresponds to the amount of EVs found in human plasma under normal and pathophysiological conditions (e.g., in cancer patients).16,17,7 Protein extracts of the cells and EVs from three biological replicates (i.e., independent cell cultures) were measured in order to determine the quality and reproducibility of the assay (Supplemental Figure S-2C,D). With two different sample/ matrix preparation methods, the mean interassay variability was found in the range 39.5−40.5% for EVs and 30.8−40.8% for cells, respectively (Supplemental Figure S-3). This demonstrates the robustness and reproducibility of the MALDI-MS approach independent of the sample types (i.e., cells and EVs) and the sample preparation method used. Signals with an individual coefficient-of-variation (CV) less than 30−40% between the cell cultures were selected as reasonable peak biomarkers for subsequent statistical data analysis. Peaks with a higher interassay variation were excluded

Information). With both techniques, a sample related size distribution of 100−400 nm with a major peak between 100 and 200 nm was obtained. The particle concentration was determined in the range (0.2−1.4) × 107 particles/mL (Supplemental Figure S-1). The results are in agreement with the size of EVs and exosomes known from literature.3,15 Although a detailed characterization was not performed, this indicates the enrichment of these types of particles from the cell culture medium. During method development, different MALDI matrixes (e.g., CHCA, sinapinic acid, etc.) were tested to detect protein related peaks of the cells and EVs in different mass ranges. The most informative mass spectra were obtained by using CHCA in the mass range m/z 2000−20000 while sinapinic acid showed a much lower signal-to-noise (S/N) ratio. This especially concerns the mass range below m/z 10000 where CHCA shows cleaner spectra with a higher number of peaks (Supplemental Figure S-2A,B). No relevant signals where detected above 20 kDa (data not shown). Thereby, mass spectra recorded in the mass range above m/z 4000 on both instrument platforms were found to be qualitatively almost identical in the number and relative distribution of the peaks (Figure 1A,B). To test the sensitivity, serial dilutions of the cell culture samples were measured. C


Letter


Figure 3. MALDI mass spectra of (A) EVs from CCL-227 colon cancer cells with varying degrees of 5-fluorouracil (FU) resistance (red, green, orange, and pink) and the corresponding score plot (upper trace) recorded with benchtop MALDI-TOF-MS. In (B) the results of a partial leastsquares data analysis (PLS-DA) of the same cell populations and their EVs using Mass++ DA software are shown. The software settings used for data analysis are displayed. Nine independent measurements were made from each EV sample. Peak lists were normalized to total ion count (TIC) for statistical analysis.

ing and intercellular communication between cancer cells in order to promote the malignant phenotype.20 In the next step, data sets from the linear benchtop MALDITOF-MS platform were used for analysis using Mass++ DA software. By overlay of the MALDI-MS data from analysis of the cells, no obvious differences between the cell populations were observed (Supplemental Figure S-4). In contrast, the mass spectra of the EVs reveal patterns of more differentially expressed protein peaks. In order to study the differences between the groups in detail, partial least-squares data analysis (PLS-DA), which looks at more subtle features of the peaks, was employed. Figure 3 shows the corresponding score plots and results of PLS-DA of the EVs from the different CCL-227 cell populations compared to their parent cells. Despite the visual similarity of the spectra (Figure 3A), the different cell populations with their varying degrees of FU resistance could be separated on the basis of statistical analysis of the individual samples (Figure 3B). Thereby, by examination of the data collected from the EVs, an even greater distinction between the naive cells and FU-resistant subclones could be observed. This is in part due to the greater number of differentiating peaks (“peak shifts”) present in the EV samples compared to those of the cells. On the basis of ANOVA, more than 60 peaks in the range m/z 2000−15500 that show significant differences between the samples were found (Supplemental Table S-1). From these, >40 peaks in the range m/z 2000−6500 were found to be unique to EVs and >10 between m/z 5500 and 15500 were unique for the cells. Some peaks were found conserved across both groups, while most of them occurred differentially expressed in the resistant and nonresistant groups. These peaks served as differentiating markers (i.e., unique features) for the classification of the samples using PLS-DA.

because of reduced biological reproducibility and consequently lesser reliability as peak biomarkers. For initial statistical exploration the data, sets of cells and EVs recorded with our standard matrix-assisted laser desorption/ionization time-of-flight/reflectron time-of-flight (MALDI-TOF/RTOF) instrument were processed (e.g., peak alignment, normalization, etc.) using Clover MS software (Figure 2A,B) and subjected to principal component analysis (PCA) and hierarchical clustering. Thereby, a distinct separation of the cells and EVs into two main clusters was obtained (Figure 2C). This clearly indicates that both populations contain discriminatory protein profiles, which is in agreement with proteomics analyses of exosomes of different cellular origins.18 Moreover, a subclustering according to the degree of FU resistance was obtained. Interestingly, the subclustering was less accurate when data from cells were used for analysis. This indicates that protein profiles of EVs seem to better reflect the changes taking place during development of chemoresistance. The present colon cancer model is characterized by an orchestrated molecular response involving different sets of genes during the development of chemoresistance to FU.10 This covers cytoskeleton rearrangements, cell adhesion molecules, insulin signaling, cell cycle control (checkpoint suppressor), and classic drug resistance mechanisms such as ABCC6, which might be reflected in the molecular content of EVs. Recently, proteomics analyses revealed that among the proteins carried by exosomes, several factors of transcriptional and translational regulation with putative influence on cancer-related pathways were found.19 This seems to be related to their inherent role as vehicles of specific molecular information involved in intracellular signalD



■

An interesting feature of this analysis was the linear grouping of the data derived from the FU-resistant vesicles, whereby the samples from naive CCL-227 cells grouped together in a far distant cluster from the FU-resistant cell populations (Figure 3B). Moreover, CCL-227 subclones with resistance to 5 μM FU showed a distinct separation from those to 25 and 125 μM FU, respectively. Remarkably, the number of unique peaks of EVs decreased with increasing FU resistance in the CCL-227 subclones compared to those of naive cells. In contrast, the number of unique peaks of the cells remained almost unchanged but showed a considerable signal intensity increase in the highly resistant subclones (Supplemental Figure S-5 and Table S-1). This corroborates the results from previous RNA expression studies where more significant changes in the expression of a higher number of genes was observed during early or intermediate stages than with a high degree of FU resistance.10,11 Thereby, downregulation of genes was the most prominent feature in the low- and intermediate-resistance phenotypes, while upregulation becomes prominent in late stages only. Obviously, these transciptional differences at the mRNA level directly account for changes in the protein profiles amenable to MALDI-MS analysis. The similarities in the mass range where the peaks appear and the sample preparation features (e.g., extraction solvents and matrix properties) indicate that the peak biomarkers of cancer cells and EVs may resemble those found also in classical biotyping approaches (i.e., basic proteins of intracellular origin). In summary, this work demonstrates the proof-of-concept that MALDI-MS protein profiling can be used to characterize EVs (e.g., exosomes) from different cancer cell populations and also between those expressing different levels of susceptibility to chemotherapy. To our knowledge this was the first time that MALDI-MS was used for the analysis of chemoresistance in cancer cells and to identify the potential of EVs to differentiate between increasing stages of anticancer drug resistance on the molecular level. Thereby, unique features (i.e., peak biomarkers) of EVs appeared to exhibit a better differentiation potential than those of the parent cells themselves. Translated to in vivo conditions, this implies a strong argument for the advantages of liquid biopsy for diagnostic purposes. Moreover, our results show a high degree of interinstrumental (i.e., standard vs benchtop MALDI-TOF-MS platforms) and interassay reproducibility. This makes the method very promising for application and validation in clinical settings. It is anticipated that applying this technique to minimally invasive diagnostics from body fluids (e.g., human plasma) will essentially contribute to improve cancer therapy monitoring in the future.

■

Letter

AUTHOR INFORMATION

Corresponding Author

*G. Stübiger. E-mail: [email protected]; g_ [email protected]. ORCID

Gerald Stübiger: 0000-0002-7128-8158 Author Contributions ∇

These authors contributed equally to this work

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank Maria Kalipciyan for performing cell culture and isolation of extracellular vesicles and Gema Méndez Cervantes for help with the preparation of the dendrograms. The work was in part supported by the Austrian Federal Government within the COMET K1 Centre Program, Land Steiermark and Land Wien.

■

REFERENCES

(1) Bizzini, A.; Greub, G. Clin. Microbiol. Infect. 2010, 16, 1614− 1619. (2) Fuh, M.; Heikaus, L.; Schlüter, H. Int. J. Mass Spectrom. 2017, 416, 96−109. (3) Vlassov, A. V.; Magdaleno, S.; Setterquist, R.; Conrad, R. Biochim. Biophys. Acta, Gen. Subj. 2012, 1820, 940−948. (4) Kastelowitz, N.; Yin, H. ChemBioChem 2014, 15, 923−928. (5) Fang, T.; Lv, H.; Lv, G.; Li, T.; Wang, C.; Han, Q.; Yu, L.; Su, B.; Guo, L.; Huang, S.; Cao, D.; Tang, L.; Tang, S.; Wu, M.; Yang, W.; Wang, H. Nat. Commun. 2018, 9, 191. (6) Gupta, A.; Pulliam, L. J. Neuroinflammation 2014, 11, 68. (7) Zarovni, N.; Corrado, A.; Guazzi, P.; Zocco, D.; Lari, E.; Radano, G.; Muhhina, J.; Fondelli, C.; Gavrilova, J.; Chiesi, A. Methods 2015, 87, 46−58. (8) Kourembanas, S. Annu. Rev. Physiol. 2015, 77, 13−27. (9) Mader, R. M.; Sieder, A. E.; Braun, J.; Rizovski, B.; Kalipciyan, M.; Mueller, M. W.; Jakesz, R.; Rainer, H.; Steger, G. G. Biochem. Pharmacol. 1997, 54, 1233−1242. (10) Schmidt, W. M.; Kalipciyan, M.; Dornstauder, E.; Rizovski, B.; Steger, G. G.; Sedivy, R.; Mueller, M. W.; Mader, R. M. Int. J. Cancer 2004, 112, 200−212. (11) Mader, R. M.; Schmidt, W. M.; Mitterer, G.; Gerner, C. Int. J. Clin. Pharmacol. Ther. 2005, 43, 599−600. (12) Senfter, D.; Holzner, S.; Kalipciyan, M.; Staribacher, A.; Walzl, A.; Huttary, N.; Krieger, S.; Brenner, S.; Jäger, W.; Krupitza, G.; Dolznig, H.; Mader, R. M. Hum. Mol. Genet. 2015, 24, 3689−3698. (13) Kavanagh, E. L.; Lindsay, S.; Halasz, M.; Gubbins, L. C.; Weiner-Gorzel, K.; Guang, M. H. Z.; McGoldrick, A.; Collins, E.; Henry, M.; Blanco-Fernández, A.; Gorman, P. O.; Fitzpatrick, P.; Higgins, M. J.; Dowling, P.; McCann, A. Oncogenesis 2017, 6, e388. (14) Tentes, I. K.; Schmidt, W. M.; Krupitza, G.; Steger, G. G.; Mikulits, W.; Kortsaris, A.; Mader, R. M. Exp. Cell Res. 2010, 316, 3172−3181. (15) Peterson, M. F.; Otoc, N.; Sethi, J. K.; Gupta, A.; Antes, T. J. Methods 2015, 87, 31−45. (16) Gercel-Taylor, C.; Atay, S.; Tullis, R. H.; Kesimer, M.; Taylor, D. D. Anal. Biochem. 2012, 428, 44−53. (17) Cheng, H. H.; Yi, H. S.; Kim, Y.; Kroh, E. M.; Chien, J. W.; Eaton, K. D.; Goodman, M. T.; Tait, J. F.; Tewari, M.; Pritchard, C. C. PLoS One 2013, 8, e64795. (18) Dozio, V.; Sanchez, J-C. J. J. Extracell. Vesicles 2017, 6, 1302705. (19) Ung, T. H.; Madsen, H. J.; Hellwinkel, J. E.; Lencioni, A. M.; Graner, M. W. Cancer Sci. 2014, 105, 1384−1392. (20) Lee, T. H.; D’Asti, E.; Magnus, N.; Al-Nedawi, K.; Meehan, B.; Rak, J. Semin. Immunopathol. 2011, 33, 455−467.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.8b03756. Exerimental section, table of peak biomarker significance and abundance, and figures of particle size measurements of EVs, MALDI mass spectra and profiles, and Venn diagrams displaying the number of unique peaks (PDF) E


MALDI-MS Protein Profiling of Chemoresistance in Extracellular

Recommend Documents