Identification of Cell Surface Proteins for Antibody-Based Selection of Human Embryonic Stem Cell-Derived Cardiomyocytes Dennis Van Hoof,†,#,‡,| Wilma Dormeyer,‡,| Stefan R. Braam,†,§ Robert Passier,†,§ Jantine Monshouwer-Kloots,†,§ Dorien Ward-van Oostwaard,†,§ Albert J. R. Heck,‡ Jeroen Krijgsveld,‡,¶,⊥ and Christine L. Mummery*,†,§,⊥ Hubrecht Institute, Developmental Biology and Stem Cell Research, Uppsalalaan 8, 3584 CT Utrecht, The Netherlands, Biomolecular Mass Spectrometry and Proteomics Group, Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands, and Leiden University Medical Center, Department of Anatomy and Embryology, Einthovenweg 20, 2333 ZC Leiden, The Netherlands Received December 10, 2009
The absence of identified cell surface proteins and corresponding antibodies to most differentiated derivatives of human embryonic stem cells (hESCs) has largely limited selection of specific cell types from mixed cell populations to genetic approaches. Here, we describe the use of mass spectrometry (MS)-based proteomics on cell membrane proteins isolated from hESCs that were differentiated into cardiomyocytes to identify candidate proteins for this particular lineage. Quantitative MS distinguished cardiomyocyte-specific plasma membrane proteins that were highly enriched or detected only in cardiomyocytes derived from hESCs and human fetal hearts compared with a heterogeneous pool of hESC-derived differentiated cells. For several candidates, cardiomyocyte-specific expression and cell surface localization were verified by conventional antibody-based methodologies. Using an antibody against elastin microfibril interfacer 2 (EMILIN2), we demonstrate that cardiomyocytes can be sorted from live cell populations. Besides showing that MS-based membrane proteomics is a powerful tool to identify candidate proteins that allow purification of specific cell lineages from heterogeneous populations, this approach generated a plasma membrane proteome profile suggesting signaling pathways that control cell behavior. Keywords: Human Embryonic Stem Cell • Differentiation • Cardiomyocyte • SILAC • Cell Surface Protein • EMILIN2
Introduction Many predicted applications of embryonic stem cells (ESCs) derive from their ability to self-renew indefinitely and form all cell types present in the body at birth.1 Potential uses range from human cells in a dish for testing therapeutic reagents, the creation of disease models and therapies based on replacing cells lost or malfunctioning as a result of damage and disease.2-5 For all of these purposes, pure populations of selected cell types will be a likely prerequisite. While several cell surface markers are available for selecting undifferentiated cells [e.g., stagespecific embryonic antigen (SSEA)-3,6 SSEA-47 and TRA-1-60 and TRA-1-818 and others9] selecting differentiated derivatives * To whom correspondence should be addressed. Phone: +31 71 526 9307. E-mail:
[email protected]. † Hubrecht Institute, Developmental Biology and Stem Cell Research. # Present address: Diabetes Center, University of California San Francisco, 513 Parnassus Avenue, San Francisco, California 94143, USA. ‡ Utrecht University. | These authors contributed equally to this work. § Leiden University Medical Center. ¶ Present address: EMBL, Genome Biology Unit, Meyerhofstrasse 1, 69117, Heidelberg, Germany. ⊥ These authors share equal contribution as senior authors.
1610 Journal of Proteome Research 2010, 9, 1610–1618 Published on Web 01/20/2010
of most lineages, with the exception of endothelial cells,10 has had to rely on genetic methods.11,12 The reason for this is that only few specific plasma membrane proteins for antibody targeting have been identified, and most cell type markers are transcription factors and structural proteins expressed intracellularly. Here, we have addressed this problem through a proteome analysis of plasma membrane proteins isolated from human ESC (hESC) cultures induced to differentiate to cardiomyocytes (hESC-CMs).13-15 In previous studies, we have shown that during cardiomyocyte formation, hESCs recapitulate gene expression patterns that occur during normal heart development,16 and successfully survive engraftment into mouse hearts that have undergone experimental myocardial infarction.5,17-19 Even though injection of mixed hESC derivatives containing ∼25% hESC-CMs results in selective survival of cardiomyocytes and progenitors over time, grafts containing human cardiovascular derivatives (cardiomyocytes, vascular endothelial cells and smooth muscle cells) exclusively will eventually be required should these procedures reach clinical application.5,17-19 The only method described to date which partially purifies hESC-CMs from mixed cell populations uses Percoll gradient centrifugation.20,21 10.1021/pr901138a
2010 American Chemical Society
Cell Surface Proteins of hESC-Derived Cardiomyocytes
research articles
This has been used successfully to enrich for cardiomyocytes from H7 cells, but it is presently not clear whether this is a generic method applicable to multiple lines. Mapping the cell surface proteome could lead not only to the identification of cell surface markers for antibody-based lineage selection, but also to the identification of growth factor receptors, signaling pathways and transporters. These could be used to enhance differentiation to mesoderm derivatives, such as hESC-CMs, in vitro or to any other lineage, as required. Recent developments in sensitivity and resolution of modern mass spectrometry (MS) methodologies as well as improved multidimensional separation techniques for proteins and peptides have promoted proteomic research tremendously. Stable isotope labeling with amino acids in cell culture (SILAC) provides the possibility of comparing protein expression levels quantitatively in different samples.22,23 Combined, these advancements have set the stage for MS-based strategies to profile the stem cell proteome in a generic manner,24-30 or focusing on specific subclasses of proteins, such as membrane proteins.31 Here, we present the application of SILAC-based quantitative MS in combination with enrichment of membrane proteins to identify hESC-CM-associated cell surface markers. Verification of surface expression using antibodies showed that elastin microfibril interfacer 2 (EMILIN2) could be used successfully for fluorescence-activated cell sorting (FACS)-based selection of hESC-CMs without genetic modification.
Results Strategy for the Comparative Analysis of Plasma Membrane Proteins. Our main objective was to identify cell surface proteins that are specific for cardiomyocytes derived in vitro from hESCs. To this end, SILAC was used to identify and quantify plasma membrane proteins differentially expressed between differentiated hESC (hEDCs), enriched populations of hESC-derived cardiomyocytes (hESC-CMs), and primary human fetal cardiomyocytes (hFCMs) (Figure 1). The rationale behind this comparative approach was that proteins identified in both hESC-CMs and hFCMs, but not in hEDCs, are specific for cardiomyocytes. Thus, only hESC-CMs were labeled metabolically, after which the membrane proteins from all the different cell types were isolated and digested (Figure 1), using the strategy optimized previously for hESCs.31 The labeled peptides from hESC-CMs were mixed with those from hEDCs (SILAC mix 1) to distinguish differences, and mixed with those from hFCMs (SILAC mix 2) to find similarities in protein expression patterns (Figure 1). Identification of Plasma Membrane Proteins. Detailed information on the recorded MS/MS spectra and the Mascot database search results including protein and peptide scores are available at https://bioinformatics.chem.uu.nl/supplementary/DvanHoof_CM. In total, 2244 proteins were identified in SILAC mix 1 and 1402 proteins in SILAC mix 2 (Supplemental Table 1). The false-positive rates of the identifications in SILAC mix 1 and 2 were 1.7% and 1.4%, respectively, as determined by reversed database searches. To predict cellular localization, the proteins identified were classified according to their GO annotation (Supplemental Table 1). In SILAC mix 1, 2003 of the identified proteins had known unique GO identifiers. The cellular localization of 1437 of these annotated proteins was known, 365 of which originated from a component that might harbor cell surface markers (i.e., from the plasma membrane, cell surface, extracellular region or matrix). In SILAC mix 2, 1348 of the identified proteins had a known unique GO identifier.
Figure 1. Schematic illustration of the SILAC approach used for the relative quantitation of plasma membrane proteins of hEDCs, hESC-CMs and hFCMs. The undirected differentiation of hESCs into a heterogeneous pool of hEDCs and the cardiomyocyte isolation of hFCMs from a human fetal heart were performed in media containing unlabeled “light” amino acids, while the differentiation of hESCs into hESC-CMs was performed in medium containing stable isotope-labeled “heavy” lysine (13C615 N2-K) and arginine (13C6-15N4-R). The microsomal fraction of the light hEDCs, heavy hESC-CMs and light hFCMs was prepared, and the proteins were digested by sequential triple cleavage. The heavy peptides of the hESC-CMs were mixed in a 1:1 ratio with the light peptides of the hEDCs (SILAC mix 1) and the hFCMs (SILAC mix 2), respectively. The resulting complex peptide mixtures were fractionated by strong cation exchange chromatography (SCX), and analyzed by nanoLC-MS/MS. The proteins were identified by searching the IPI human database using the Mascot algorithm, and quantified by calculation of the intensity ratios of the heavy/light peptide pairs using the MSQuant algorithm.
The cellular origin was known for 1043 of these annotated proteins; 339 originated from cell components potentially harboring cell surface markers. Quantitation of Plasma Membrane Proteins. All proteins identified in SILAC mix 1 and 2 were quantified in a relative manner by integration of the extracted peptide ion chromatograms and subsequent calculation of the intensity ratios of the heavy/light peptide pairs. The relative quantitation represents the average heavy/light intensity ratio of the peptide pairs detected for each protein, thereby reflecting the abundance ratio of the protein in each cell type [i.e., the protein ratio hESCCM/hEDC in SILAC mix 1 (Figure 2A) and hESC-CM/hFCM in SILAC mix 2 (Figure 2B)]. Proteins with a ratio g2 (log2 ) 1) were denoted as being more abundant in hESC-CMs than in hEDCs or hFCMs, whereas proteins with a ratio of e0.5 (log2 ) -1) were considered as significantly less abundant in hESCCMs than in hEDCs or hFCMs. Journal of Proteome Research • Vol. 9, No. 3, 2010 1611
research articles
Figure 2. Overview of the relative quantitation of the proteins identified in SILAC mix 1 and SILAC mix 2. Plotted are the log2 values of the average “heavy/light” intensity ratio of the peptide pairs detected for each protein identified in SILAC mix 1 (A) and 2 (B). The standard deviation bars were calculated as [(loge × standard deviation of the ratio)/ratio]. Proteins with intensity ratios outside the gray shaded area, i.e., proteins with an average intensity ratio g2 (log2 ) 1) or e0.5 (log2 ) -1) were considered as significantly higher or lower abundant in hESC-CMs than in hEDCs and hFCMs, respectively.
Classification of Quantified Plasma Membrane Proteins. To extract the most promising candidate hESC-CM cell surface markers, proteins identified in both experiments were combined and classified based on the observation that proteins increased (log ratio >1), decreased (log ratio
