Profiling signaling polarity in chemotactic cells

Abstract

Cell movement requires morphological polarization characterized by formation of a leading pseudopodium (PD) at the front and a trailing rear at the back. However, little is known about how protein networks are spatially integrated to regulate this process at the system level. Here, we apply global proteome profiling in combination with newly developed quantitative phosphoproteomics approaches for comparative analysis of the cell body (CB) and PD proteome of chemotactic cells. The spatial relationship of 3,509 proteins and 228 distinct sites of phosphorylation were mapped revealing networks of signaling proteins that partition to the PD and/or the CB compartments. The major network represented in the PD includes integrin signaling, actin regulatory, and axon guidance proteins, whereas the CB consists of DNA/RNA metabolism, cell cycle regulation, and structural maintenance. Our findings provide insight into the spatial organization of signaling networks that control cell movement and provide a comprehensive system-wide profile of proteins and phosphorylation sites that control cell polarization.

Directed cell migration or chemotaxis in response to a chemokine gradient is involved in development, immune function, angiogenesis, as well as pathological processes associated with inflammation and cancer cell metastasis (1). Cell locomotion is a highly polarized process characterized by protrusion of a leading pseudopodium (PD) (lamellipodium) at the front and establishment of a trailing rear compartment or tail region at the back (1, 2). This process is regulated through organization of the actin-myosin cytoskeleton in response to signal transduction processes that operate downstream of chemokine and integrin adhesion receptors. These receptors serve as antennae to sense changes in chemokine and extracellular matrix gradients, which provide navigational cues to direct cell movement. However, the molecular signaling mechanisms that control cell polarity and chemotaxis are not fully understood.

Emerging evidence indicates that cell behavior is regulated by the complex organization of multiple signaling networks in time and space. The specific propagation of biological information in a complex biological system is achieved through the spatiotemporal assembly of multiprotein scaffolds, protein activation cascades, protein turnover, and specific posttranslational modifications of proteins, including the phosphorylation and dephosphorylation of specific amino acid residues. Although significant progress has been made in identification of specific signaling proteins and regulatory pathways that mediate cell migration, little is known about how these signals are integrated into spatial networks to achieve morphological polarity and directional movement at a system level.

To understand the complexity of signal organization in migrating cells, our laboratory developed a system that facilitates the differential purification of the PD and cell body (CB) compartments of chemotaxing cells for proteomic analysis using microporous filters (3, 4). This model system recapitulates physiological events associated with chemotaxis, including gradient sensing and pseudopodial protrusion through small openings in the vessel wall during cancer cell metastasis (5). This model has revealed important regulatory pathways and novel proteins that mediate cell movement (3, 6–8).

Recent advances in mass spectrometry allow the identification of thousands of unmodified and posttranslational modified peptides from complex biological samples, including specific sites of phosphorylation (9, 10). This approach combined with bioinformatics analysis is able to map thousands of protein changes in a given biological sample and to determine their interrelationships (11). In the current study, we applied global proteome profiling in combination with newly developed quantitative phosphoproteomics approaches to map and compare the CB and PD subproteomes. Our analyses revealed networks of actin regulatory proteins and kinase-substrate cascades as well as signaling and metabolic pathways that partition to different poles of the migrating cell. Our findings provide insight into the protein networks that control cell polarization and chemotaxis and could lead to a systems-wide understanding of how chemotactic cells organize signaling polarity to achieve directional movement.

Results

Purification of Pseudopodia and Analysis of Signaling Polarity in Migrating Cells.

To identify proteins associated with cell polarization and migration, we used 3.0-μm microporous filters, which facilitate the exposure of large numbers of cells to a gradient of chemoattractant l-α-lysophosphatidic acid (LPA) and pseudopodia purification for proteomic analysis (3, 7, 8) (Fig. 1A). In this system, cells sense the directional cue and respond biochemically by amplification of migratory signals as illustrated by the robust ERK activation in response to a gradient, but not a uniform concentration of the chemoattractant LPA (Fig. 1B Left). The amplified signaling response then drives the actin-myosin machinery to produce pseudopodial extensions through the small openings in the filter toward the chemoattractant in the lower chamber (3, 7). Under these conditions, cells achieve both morphological and biochemical polarity as illustrated by confocal imaging and the localized ERK activation at the front and the desensitization of chemokine-directed ERK activation at the rear of the cell (Fig. 1B Right). Polarized pseudopodia and cell bodies can then be manually separated from either side of the filter into the appropriate buffer for biochemical analysis (3, 4).

Fig. 1.

Microporous filter system for purification and mass spectrometry analysis of PD and CB fractions isolated from polarized cells. (A) Three-dimensional reconstruction of a Cos-7 cell expressing GFP and protruding PD through 3.0-μm pores of the filter (dashed red line) to the lower surface. The PD- and CB-localized proteins are differentially harvested, digested with trypsin, and then analyzed by LC-MS/MS (1) or dual-step ¹⁸O/¹⁶O labeling (2), followed by IMAC to isolate phosphorylated peptides as described in Experimental Procedures. The number of peptides sequenced and proteins identified is shown. (B) (Left) Cos-7 cells on microporous filters were either not treated (NT) or stimulated with a uniform (U) or gradient (G) concentration of LPA (100 ng/ml) for the indicated times and then Western blotted for activated ERK (ERK-P) or total ERK protein. (Right) Cos-7 cells were stimulated (+) or not stimulated (−) with an LPA gradient for 30 min to allow cells to polarize by protruding PD through 3-μm porous filters. CBs on the tops of the filters were then stimulated with LPA for the indicated times before being collected and Western blotted for activated and total ERK. Note that the CB is not responsive to LPA-induced ERK activation in polarized cells.

Global Characterization of the PD and CB Proteomes.

Significant evidence indicates that cell polarization requires the partitioning of proteins to the front and back (3, 7). A change in protein concentration at the different poles is believed to facilitate spatial signaling events that drive gradient-sensing mechanisms and cytoskeletal reorganization leading to morphological polarity (2). To identify relative differences in protein abundance in the PD and CB compartments, we used 2D liquid chromatography-coupled tandem MS (LC-MS/MS) to identify and detect these protein abundance changes. Specifically, by calculating the ratio of peptide spectrum counts (PD/CB) assigned to the peptide for each protein in each fraction, it is possible to quantify the relative abundance for a given protein between the two samples. To note, this spectrum count-based method may be affected by parameters such as sample processing and instrument analysis. However, this relative quantitative approach has been shown to be successful when it is based on comparing the same protein across multiple, but similar, samples (12, 13) provided with proper controls in experimental design and analysis. Using this approach, we recorded 494,775 and 493,875 MS/MS spectra from the CB and PD fractions to which we assigned sequences to 26,119 and 26,325 distinct peptides, respectively. In total, 5,266 proteins were identified [supporting information (SI) Table 1; see also SI Tables 2–6]. Of these, only proteins identified by two or more independent peptides (3,509) were included for further analysis.

A protein was considered to be enriched in a particular fraction if its PD/CB ratio fell outside 1.5 and 0.67. Also, proteins that were detected in only one fraction (unique) were considered to be enriched and were included for analysis. All other proteins were considered to be equally abundant in the two fractions. Using these criteria, 1,073 proteins were enriched in the PD, 1,088 proteins were enriched in the CB, and 1,348 proteins were equally distributed (SI Table 1). SDS/PAGE analysis and a plot of the distribution pattern of the quantified proteins in each fraction are shown in Fig. 2.

Fig. 2.

Functional comparison of the PD and CB proteomes. (A) The bars represent the quantitative distribution of the number of PD proteins identified by LC-MS/MS relative to corresponding proteins identified in the CB fraction, as determined in Experimental Procedures. (B) The relative abundance can also be indicated from the intensity of bands on Coomassie-blue-stained SDS/PAGE gel for PD and CB proteins (Left) or by quantitative Western blotting of the indicated proteins (Right). R_B indicates the PD/CB ratio as measured by Western blot and densitometry. R_m ratio was measured by LC-MS/MS as in A. (C) Comparison of the pathways that are relevant to the CB- and PD-enriched proteins (P < 0.05; the lower P value indicates the pathway is more relevant to the analyzed proteins). The pathways are extracted from 1,088 CB-enriched proteins (PD/CB <0.67) and 1,073 PD-enriched proteins (PD/CB >1.5) using Ingenuity Pathways software.

We next used quantitative Western blotting and densitometry to confirm the relative changes in protein abundance measured by LC-MS/MS. There was an excellent correlation of protein abundance ratios detected by the two different methods for several protein classes, including cytoskeletal-associated proteins (caldesmon, α-actinin, drebrin, and dynein), cytoplasmic-signaling proteins (Akt and ERK2), and membrane proteins (EphA2 and EGFR) (Fig. 2B Right). Importantly, LC-MS/MS analysis also revealed the expected changes in abundance of several proteins previously shown to concentrate in the PD (Lasp-1 and cortactin) (7, 14) or CB (histones, lamin B2, and serine/threonine kinase VRK1) (15, 16) of polarized cells (SI Table 1). However, the possibility remained that our protein profile reflects changes in membrane to cytoplasmic ratios in PD and CB, rather than those related to cell polarization. To directly measure the membrane/cytoplasmic ratio, PDs and CBs were harvested from polarized cells expressing membrane-localized GFP (CAAX-GFP) and examined for relative differences in GFP levels by Western blotting. There was only a 1.4-fold increase in the level of GFP in the PD compared with the CB, indicating that the membrane/cytoplasmic ratio has not been significantly altered (Fig. 2B). Additionally, subcellular component categorization based on gene ontology annotation (www.geneontology.org) revealed an essentially equal distribution of all subcellular compartments between the CB and PD fractions except for nuclear (1.34-fold increase for CB) and cytoskeleton (1.41-fold increase for PD) (SI Table 2). This distribution represents a real, but limited, bias when comparing across samples, but the combined ratio of the amount of the proteins in the two subcellular compartments in each fraction is almost equal, and thereby will not affect the quantitative comparison of the proteins in other subcellular compartments, i.e., membrane proteins (see SI Text and SI Figs. 5–9 for a more in-depth discussion). Together these findings demonstrate the feasibility of determining relative quantitative differences in protein abundance in the CB and PD proteomes of polarized cells.

Analysis of Signaling Pathways and Networks of Proteins Spatially Localized to the CB and PD Compartments.

We next sought to find the functional interrelationship of the proteins identified in the PD and CB proteomes and determine whether they are compartmentalized into distinct pathways in the front and back of polarized cells. We first examined proteins that are enriched in the PD and CB fractions because these proteins are more likely to play an important role in cell polarization. For functional annotation, we used the Ingenuity Pathways Analysis program and the Ingenuity Pathways Knowledge Base (IPKB), which is a system-wide database of biological pathways created from millions of relationships of proteins, genes, and diseases (11). The Ingenuity Pathways Analysis program can analyze a large genomic or proteomic data set to find the most significant canonical pathways and networks relevant to the data set based on the calculated probability score by searching the IPKB (12). The canonical pathway analysis revealed that 17 distinct pathways were represented by PD-enriched proteins, whereas 14 canonical pathways signify the CB-enriched protein fraction (Fig. 2C) (P < 0.05). Interestingly, a comparison of these pathways revealed a high degree of spatial separation of signaling pathways between the front and back of polarized cells. For example, many pathways that regulate the actin cytoskeleton, integrin signaling, and axon guidance are more highly represented in the PD, whereas pathways that regulate estrogen receptor signaling, cell cycle, and purine metabolism are more dominant in the CB. It is also notable that the ERK and phosphatidylinositol-signaling pathways are highly represented in the PD relative to the CB (Fig. 2C). These pathways have been previously shown to regulate cell polarity, pseudopodial dynamics, and cell movement (17–20). In contrast, analysis of proteins that are equally abundant in the front and back revealed a more diverse distribution of pathways in metabolism, biosynthesis, posttranslational modifications, and cellular signaling (Fig. 2C and SI Fig. 10). However, some of these proteins and pathways may still play a role in establishing cell polarity, which is not related to concentration differences at the front and back of the cell. For example, some components associated with cytoskeleton (GIT1, profilin 1, and CRKL), growth factor signaling (MAP2K4, MAP2K1, STAT1, ERK1, and ERK2), and metabolism are equally distributed in the PD and CB fractions (SI Table 1). This class of proteins may provide specific functions at the front and back that are regulated by specific posttranslational modifications and protein–protein interactions. This response is best illustrated by ERK, which is highly phosphorylated and activated in the PD relative to the CB, but is equally abundant in the front and back of the cell (Figs. 1B and 2B).

We reconstructed a functional interactome of the enriched proteins present in PD and CB proteomes using all known direct and indirect protein interactions (e.g., protein A activates protein B, protein B activates protein C, and then proteins A and C have indirect interactions) in the IPKB database. The functional integration of proteins in the different proteomes revealed distinct interactomes that operate in the front and back of the cell (SI Fig. 11). The majority of proteins in the PD interactome are involved in cell assembly, cell organization, cell movement, cell morphology, and cell–cell signaling, including actin cytoskeleton, integrin signaling, ephrin receptor signaling, and other signaling proteins (SI Fig. 11A and SI Table 4). The CB interactome regulates DNA replication, recombination and repair, gene expression, RNA posttranscriptional modification, cell cycle, small molecular biochemistry, cellular assembly, and cellular growth and proliferation (SI Fig. 11B and SI Table 5). In contrast, the networks represented by the equally abundant proteins have more diverse functions in biosynthesis, molecular transport, metabolism, cell assembly and morphology, and cell death (SI Fig. 12).

Quantitative Comparison of Protein Phosphorylation in the PD and CB.

To further our understanding of the phosphorylation-based signaling pathways involved in PD control, a newly developed quantitative phosphoproteomic approach was applied for comparative analysis of CB and PD that combines immobilized metal affinity chromatography (IMAC) for specific phosphopeptide enrichment with a dual stable isotopic labeling (trypsin-catalyzed ¹⁸O/¹⁶O and ¹⁸O/¹⁶O-methanol esterification) approach, and LTQ-FT with MS² and MS³ function for CB- (¹⁶O) and PD- (¹⁸O) specific fractions (SI Fig. 13). The CB and PD were lysed in buffer containing sodium orthovanadate and calyculin to inhibit the tyrosine and serine/threonine phosphatases, respectively, before IMAC and LC-MS analysis. In total, we recorded 20,446 MS/MS (MS²) fragmentation events with neutral loss-triggered acquisition of 4,082 MS/MS/MS (MS³) spectra. From the 1,399 phosphopeptides identified, many peptides were sequenced multiple times due to the ¹⁸O/¹⁶O pairs, MS²/MS³ redundancy, and different charge states, which collapsed into 228 unique phosphopeptides corresponding to 197 proteins (SI Table 3). Among these unique phosphopeptides, 179 are singly phosphorylated (78.5%) and 49 are doubly phosphorylated (21.5%). The sites of phosphorylation for 198 peptides (86.8%) were confidently assigned based on manual inspection of the corresponding fragmentation mass spectra.

Conservative analysis of phosphorylation sites was performed by using an in-house program, Blastpro, which allows the direct comparison of large proteomic data sets based on amino acid sequence homology within and across species (8).The results revealed that 89% and 80% of the phosphorylation sites identified in humans are conserved in mouse and rat, respectively, whereas 44% are conserved in lower vertebrates (zebrafish) (SI Table 3).

Of the 228 phosphopeptides identified and quantified, 77 (33.7%) showed a 1.5-fold or more enrichment in the PD, whereas 96 phosphopeptides were enriched in the CB by 1.5-fold or higher (42.1%). The remaining peptides did not show significant changes falling within the 1.49 to 0.66 range. Notable proteins that showed significant increases in phosphorylation in the PD fraction include the serine/threonine kinase ERK2 (34.1) (SI Fig. 14), tyrosine-protein kinase receptor UFO precursor (Axl) (15.83), EphA2 (14.82), and phosphoinositide 3-kinase (PI3K) (8.22). CB proteins that showed a significant increase in phosphorylation included phospholipase C, ε-1 (0.001), protein kinase N2, and PIK3 γ-adaptor protein (0.09). It is notable that the 34.1-fold increase in ERK2 phosphorylation is in the activation domain (T*EY*). This observation supports our previous work that ERK is highly activated in the PD, which is necessary for proper cell movement (17, 18). Although the function of the novel phosphosites of Axl, EphA2, and PI3K are not known, these proteins were previously shown to play critical roles in cell migration (21–23).

The observed differences in phosphoprotein abundance within polarized cells may be due to subcellular compartmentalization, phosphorylation/dephosphorylation, and/or protein synthesis/degradation. Because the true stoichiometric change in phosphorylation of a given protein depends on its absolute abundance, we compared the quantitative ¹⁸O/¹⁶O ratio of the detected phosphoprotein with its corresponding total protein level detected in each fraction by LC-MS/MS. Of the 197 phosphoproteins identified by IMAC, 135 (69%) were identified and relatively quantified in the global proteome profiling experiment (SI Tables 1 and 3). For these proteins, the ratio of the phosphorylated form of the protein to its corresponding total protein level was calculated and is shown in SI Table 3. Several important observations were made from this correlation. First, some phosphoproteins (e.g., Eph A2) showed a high level of phosphorylation that correlated with increased protein levels in either the CB or PD compartments. Second, a few phosphoproteins showed increased abundances in the different compartments, but displayed inverse levels of phosphorylation (e.g., serine/threonine-protein kinase pctaire-2 and PLEKHA7 protein). Third, although some proteins were present at similar levels in the CB and PD, their level of phosphorylation was significantly different, either increased or decreased in abundance (e.g., ERK) (SI Table 3). Fourth, a few phosphoproteins were equally distributed between the CB and PD, but displayed unique phosphorylation sites depending on their cellular location. For example, the singly phosphorylated peptide sequence YHGHS*MSDPGVSYR derived from pyruvate dehydrogenase E1 component α-subunit was significantly enriched in the CB (0.34), whereas its doubly phosphorylated form YHGHS*MSDPGVS*YR was highly up-regulated in PDs (3.04), suggesting that it is differentially regulated depending on its subcellular location (SI Table 3 and SI Fig. 15). These findings highlight the significance of this approach for revealing spatially regulated phosphoproteins and their phosphorylation sites in polarized cells.

Functional Comparison of Phosphoproteins in the CB and PD.

Interestingly, when comparing the identified 135 phosphoproteins, which are also found within the global protein data set, the majority (79%) were not equally distributed in the PD and CB, but rather showed a marked increase or decrease in the front and rear of the cell (SI Fig. 16). This result is in contrast to the entire protein complement, which shows the majority of proteins to have approximately equal distributions between the CB and PD (Fig. 2A). This finding suggests that polarized cells selectively partition phosphoproteins with specialized functions to the front and rear of the cell. Indeed, Ingenuity Pathway Analysis revealed that the PD phosphoproteome has a greater percentage of proteins involved in signal transduction processes and regulation of actin cytoskeleton, whereas the CB proteome has a greater percentage of proteins involved in estrogen receptor signaling, nucleotide excision repair pathway, and cell cycle regulation (Fig. 3A). Notable phosphorylation sites known to be involved in cell movement and enriched in the PD include ERK, α-parvin, and cortactin (SI Table 3) (24–26). These findings are consistent with a significant body of evidence showing that phosphorylation is a critical component that regulates cytoskeletal and focal adhesion-signaling dynamics during PD formation (3, 17). However, at the rear of the cell, a higher percentage of the proteome was represented by housekeeping proteins, DNA, and RNA regulatory proteins, including RNA polymerase largest subunit, CDK7, and lamin. As expected, functional analysis of the equally distributed phosphoproteins such as cAMP-dependent protein kinase, 6-phosphofructokinase, and elongation factor 1-δ revealed more diverse functions in signaling and metabolic pathways (data not shown). Together these findings indicate that directional movement requires spatial localization of distinct phosphoproteins within the front and back of the cell.

Fig. 3.

Functional comparison of the PD and CB phosphoproteomes. (A) Comparison of the pathways that are relevant to the CB- and PD-enriched phosphoproteins (P < 0.05), as in Fig. 2A. (B) The distribution pattern of kinase phosphorylation motifs and associated kinase classes of identified phosphopeptides. The percentage indicates the ratio of the number of proteins containing a particular motif to the total number of proteins in each fraction.

Assignment of Identified Phosphorylation Sites to Known Kinases.

Phosphorylation motif analysis by the software Scansite was used to assign a putative kinase to the identified phosphorylation site (27). We queried our data set with 10 distinct kinase-recognition motifs (Fig. 3B). Of these, proline-directed kinases represented the largest group (44%), followed by the basophilic kinases (36%). We then determined the number of each motif present in the PD and CB fractions. Interestingly, comparison of these findings revealed specific compartmentalization of several kinase classes (Fig. 3B). ERK1/2-, GSK-3β-, and CK2-recognition motifs were significantly increased in the PD compared with the CB, whereas basophilic 2 and CDK1/2/5 kinases were significantly increased in the CB compartment. ERK1/2 and GSK-3β have been shown to localize to the PD and regulate cell migration, whereas CDK1/2/5 enzymes are necessary for cell cycle regulation (17, 28).

Characterization of the Ras/ERK-Signaling Pathway in the PD by Combined Proteome and Phosphoproteome Profiling.

ERK activity is strongly amplified in the PD induced by a gradient of LPA, and this activation is necessary for PD extension and chemotaxis (SI Table 3 and SI Fig. 14) (17, 18). Also Western blot analysis shows that the upstream kinase of ERK, i.e., MEK, is significantly activated in the PD compared with the CB (Fig. 4A). Using the Ingenuity software, we mapped the signaling networks in PD that could activate ERK in the PD (Fig. 4B). The signaling network not only revealed the canonical RAF/MEK/ERK pathway, but also revealed at least three possible upstream pathways that could activate ERK. The first pathway involves the transactivation of epidermal growth factor receptor (EGFR) by LPA receptors. This response is mediated by metalloproteinase ADAM 10, which releases membrane-bound HBEGF binding to the EGFR (29). The activated EGFR then recruits Shc and SOS to activate the RAF/ERK pathway. The second pathway occurs through the activation of RAF by activated Src, which is highly enriched in PD (PD/CB = 3.4) (29). Finally, the activity of ERK could also be enhanced by Integrin-signaling components, which are highly enriched in the PD (Fig. 4B) (30).

Fig. 4.

Signaling pathways that regulate ERK in response to LPA-induced PD formation. (A) Western blot shows the total and phosphorylation levels of MEK and ERK in PD and CB compartments. (B) Three possible signaling pathways that may activate ERK in response to LPA-induced PD extension. ERK activity can also be negatively regulated by EphA2. (C) Known ERK substrates and positive/negative regulators. The relative abundance of each protein is included in parenthesis. PDU indicates that the protein is unique in PD.

ERK localization and/or activity can also be positively or negatively regulated by many other proteins directly or indirectly. Therefore, we also used the Ingenuity Pathway Analysis to determine all known ERK regulators that localize to the PD (Fig. 4C). Important ERK-interacting proteins identified include PEA15 and KSR, which are enriched 2.7 and 4 times, respectively. PEA15 can sequester activated ERK in the cytoplasm, preventing its translocation to the nucleus (31), and KSR provides a scaffold to assemble the Raf/MEK/ERK-signaling module in the membrane, leading to an increase in signal amplitude and duration (32). These proteins may play an important role in sequestering activated ERK and its regulatory components in the PD. On the contrary, the dual-specificity phosphatase 3 (DUSP3) can dephosphorylate and hence deactivate ERK (Fig. 4C) (33). In addition, activated EphA2 was also previously shown as a negative regulator of the RAF/MEK/ERK pathway (Fig. 4B) (34).

Finally, we sought to identify all possible ERK substrates that reside in the PD. The signaling network revealed eight known ERK substrates that are enriched in PD, including α-parvin, myosin light chain kinase, and cortactin (18, 26, 33) (Fig. 4C). These substrates are known to be involved in cell migration. It is also notable that the ERK phosphorylation site identified in cortactin (Ser 405; SI Table 3) has been previously shown to play an important role in cell motility (24, 25). Further analysis of the data set for proteins that contain both an ERK kinase-recognition sequence (P-X-[S T]-P) (35) and an ERK docking site ([R K]-X_2–6-ψ-X-X, where ψ indicates hydrophobic residues Leu, Ile, and Val) or a binding motif (F-X-F) (33) revealed 12 other possible ERK substrates, including the uncharacterized proteins KIAA0731 and FLJ4447 (SI Table 6). Although these putative ERK substrates have not been previously linked to cell motility, they may play an important role in this process. Together these findings illustrate the spatial organization of the up- and downstream protein networks that regulate the Ras/ERK signaling pathway in polarized chemotactic cells and reveal possible new proteins that may contribute to this process.

Discussion

We investigated the spatial organization of the proteome and phosphoproteome in the front and rear of polarized cells by combining global proteome profiling with the newly developed phosphoproteomics approach. Bioinformatics analyses of these results provided insight into the system-wide organization of a large number of signal transduction pathways and protein and phosphoprotein networks. We focused our bioinformatics primarily on the pool of enriched proteins because these proteins are more likely to play a role in establishing cell polarity. The functional annotation and interactome of these proteins largely reflected the spatial regulation of actin dynamics and focal adhesion dynamics at the leading front of the cell and housekeeping and nuclear functions at the rear of the cell. However, many proteins were equally abundant in the front and back of cells, including numerous cytoskeletal and metabolic proteins. Some of these proteins are also likely to contribute to the process of cell movement.

A fascinating finding was that the pathways involved in regulating cell polarity during axon guidance were highly represented in the PD proteome. These findings suggest that the fundamental ability of cells to sense directional cues and polarize toward various stimuli involve conserved regulatory elements shared across diverse cell types. Also homology analysis revealed a high degree of amino acid sequence conservation among identified phosphorylation sites in human, mouse, and zebrafish (SI Table 3). In the end, these evolutionarily conserved signals must impinge on the actin cytoskeleton and focal adhesion structures in a temporal and spatial manner to mediate adhesiveness and cell shape changes necessary for morphological polarization.

It has been predicted that upward of one-third of the total cell proteome is phosphorylated at any given time (9). In contrast, our overall phosphoproteome coverage (228 sites) only represents a small fraction of these total events in a cell. Regardless, we were able to target a substantial number of key phosphorylation events of proteins and kinases specific to our cell model system and known to regulate migration such as the canonical Ras/ERK pathway. We also uncovered phosphoproteins and sites of phosphorylation (e.g., EphA2) not previously linked to cell polarization and chemotaxis. These proteins and their specific sites of phosphorylation provide important new targets to investigate for a role in cell migration. We anticipate that the implementation of extensive protein- and peptide-level fractionations before enrichment and MS analysis will likely provide a significant increase in the detection of phosphorylation sites, which we are currently pursuing.

In summary, by integrating contemporary quantitative proteomics with methods for the fraction of polarized cells, we have created a comprehensive and valuable resource database for investigating signal polarity in chemotactic cells. This work, along with the rapid technological advances in the proteomic and bioinformatics fields, will allow the identification of even more protein signatures and their integration into signaling networks at the systems level. This approach will provide a detailed understanding of cell polarization and chemotaxis and generate computational models and testable hypotheses. Future work to integrate the PD proteome and phosphoproteomes with emerging disease-perturbed genomic and proteomic databases will provide a novel approach for the development of diagnostics related to cancer progression and immune dysfunction. This approach was recently illustrated by computational mining of large-scale genomic and proteomic databases to discover the metastatic and PD-associated protein α-actinin (8). In the end, the application of proteomics and bioinformatics to cell chemotaxis will provide a better understanding of how signaling polarity and cell movement integrate into the whole organism in health and disease.

Experimental Procedures

Sample Preparation.

Low-passage Cos-7 (ATCC, Rockville, MD) cells were grown in DMEM (GIBCO, Invitrogen, Carlsbad, CA) supplemented with 10% FBS (Gemini Bio-Products, Woodland, CA) to 70–80% confluence. The growth of PD was induced by LPA as previously described (3, 4). The PD and CB were harvested into a lysis buffer [7 M urea, 2 M thiourea, 40 mM Tris (pH 8.4), protease inhibitor mixture, 2 mM vanadate, and 50 nM calyculin]. The samples were then digested with trypsin, and the resulting peptides were desalted and quantified.

Dual-Step ¹⁸O/¹⁶O Labeling and IMAC Enrichment of Phosphopeptides.

Five hundred micrograms of desalted tryptic peptides from PD and CB fractions was labeled with ¹⁸O/¹⁶O, respectively, by using the previously described procedures (36). Phosphopeptides were enriched by IMAC as previously described with some modification (37). Phosphopeptides were eluted by 150 μl of 50 mM Na₂HPO₄ (pH 9.0) and were acidified with acetic acid immediately after elution.

LC-MS/MS and MS³ Analysis.

The peptides from whole CB and PD lysates were analyzed by 2D LC-MS/MS on a linear ion trap (LTQ) mass spectrometer (ThermoFinnigan, San Jose, CA), and the phosphopeptides enriched from ≈100 μg combined CB and PD peptides by IMAC were analyzed by LC-MS²/MS³ on an LTQ-Fourier Transform (LTQ-FT) mass spectrometer (ThermoFinnigan). The LTQ-FT acquiring method was described previously with some modification (38). All MS/MS and MS³ spectra were searched against the human International Protein Index database using the SEQUEST program. Initial filtering criteria were based on a reversed database searching results to achieve a >95% confidence (39). Additionally, the confidence of the peptide identifications was also evaluated by independently calculating a probability-based confidence score by using the program Peptideprophet, which resulted in a composite data set score of 98% confidence (40). The phosphorylation sites were assigned and manually confirmed (see SI Text and SI Fig. 17 for details).

Bioinformatics.

The pathways and signaling networks were analyzed by using the Ingenuity software (Ingenuity Systems, www.ingenuity.com). Briefly, we queried all PD- and CB-enriched and equally distributed proteins against the IPKB, and the significance values (P value) associated with pathways were calculated by using the right-tailed Fisher's exact test to measure how likely it is that proteins from the data set file are relevant to the pathway. The test was performed by comparing the number of user-specified proteins that participate in a given pathway, relative to the total number of occurrences of these proteins in all pathway annotations stored in the IPKB database.

The kinase substrate motifs were predicted by Scansite using a quick matrix method (www.scansite.mit.edu/quickmatrix.html).

Abbreviations

CB

cell body

EGFR

epidermal growth factor receptor

IMAC

immobilized metal affinity chromatography

IPKB

Ingenuity Pathways Knowledge Base

LC-MS/MS

liquid chromatography-coupled tandem MS

LPA

l-α-lysophosphatidic acid

LTQ

linear ion trap

PD

pseudopodium.