Highly multiplexed quantitative proteomic and phosphoproteomic analyses in vascular smooth muscle cell dedifferentiation

Abstract

Restenosis, re-narrowing of arterial lumen following intervention for cardiovascular disease, remains a major issue limiting the long-term therapeutic efficacy of treatment. The signaling molecules, TGFβ (transforming growth factor-beta) and Smad3, play important roles in vascular restenosis, but very little is yet known about the down-stream dynamics in global protein expression and phosphorylation. Here, we develop a highly multiplexed quantitative proteomic and phosphoproteomic strategy employing 12-plex N,N-dimethyl leucine (DiLeu) isobaric tags and DiLeu Tool software to globally assess protein expression and phosphorylation changes in smooth muscle cells (SMCs) treated with TGFβ/Smad3 and/or SDF-1α (stromal cell-derived factor). A total of 4086 proteins were quantified in the combined dataset of proteome and phosphoproteome across 12-plex DiLeu-labeled SMC samples. 2317 localized phosphorylation sites were quantified, corresponding to 1193 phosphoproteins. TGFβ/Smad3 induced up-regulation of 40 phosphosites and down-regulation of 50 phosphosites, and TGFβ/Smad3-specific SDF-1α exclusively facilitated up-regulation of 27 phosphosites and down-regulation of 47 phosphosites. TGFβ/Smad3 inhibited the expression of contractile-associated proteins including smooth muscle myosin heavy chain, calponin, cardiac muscle alpha-actin, and smooth muscle protein 22α. Gene ontology and pathway enrichment analysis revealed that elevated TGFβ/Smad3 activates cell proliferation and TGFβ signaling pathway, sequentially stimulating phosphorylation of CXCR4 (C-X-C chemokine receptor 4). SDF-1α/CXCR4 activated extracellular signal-regulating kinase signaling pathway and facilitated the expression of synthetic marker, osteopontin, which was validated through targeted analysis. These findings provide new insights into the mechanisms of TGFβ regulated SMC dedifferentiation, as well as new avenues for designing effective therapeutics for vascular disease.

Graphical abstract

1. Introduction

Over one million angioplasty-stent procedures are performed each year to treat cardiovascular disease. This treatment is initially effective; however, 20–30% of patients eventually develop intimal hyperplasia and constrictive remodeling that re-narrowing blood vessel [1]. Previous studies have implicated transforming growth factor-beta (TGFβ) signaling involved in this process [2–5]. TGFβ binds to cell surface receptor and induces downstream signaling protein Smad3 phosphorylation. Phosphorylated Smad3 forms a complex with Smad2 and Smad4, which subsequently translocates to the nucleus and regulate gene expression. TGFβ regulates gene expression in a context-dependent manner based on Smad-interacting proteins, the activity of signal pathways, and the epigenetic landscape [6]. Activated TGFβ signaling and other signaling pathways induce smooth muscle cell (SMC) dedifferentiation and transform SMC from quiescent, contractile to synthetic, proliferative phenotype [4]. Elevated levels of TGFβ and Smad3 also up-regulate stem cell-related genes, most notably chemokine receptor type 4 (CXCR4) [7], a receptor for stromal cell-derived factor 1 (SDF-1α). Binding of SDF-1α to CXCR4 induced protein phosphorylation and activate MAPK kinase in the injured arterial wall [7,8]. TGFβ and SDF-1α triggered expression and phosphorylation of downstream signaling molecules are primary driving factors in vascular restenosis by accumulating proliferative SMCs and modulating arterial diameter by remodeling of arterial wall after vascular injury. However, very little is known about the downstream protein expression and phosphorylation in an environment similar to vascular injury.

To decipher the molecular mechanism of vascular restenosis, we employed highly multiplexed quantitative proteomics and phosphoproteomics (collectively referred to as (phospho)proteomics) in SMCs sequentially treated with Smad3/TGFβ and SDF-1α. Mass spectrometry (MS) has emerged as an effective technique to study large-scale phosphoprotein regulation in many biological systems [9–15]. In-house developed 12-plex N,N-dimethyl leucine (DiLeu) isobaric tags were utilized for global (phospho)proteome profiling, greatly increasing the analytical throughput (Fig. S1). The DiLeu tags consist of an N,N-dimethyl leucine reporter group, a balance group, and a triazine ester amine-reactive group to selectively label peptide N-termini and lysine side chains [16]. This set of tags barcodes up to 12 samples, enabling analysis of multiplexed samples in a single assay to facilitate high-throughput quantification and reduce run-to-run variability. To ensure accurate quantification, we developed DiLeu tool to specifically correct isotopic interference derived from the multiple isobaric tags.

Parallel quantification of the proteome and phosphoproteome enables distinguishing whether regulation of certain phosphorylation sites (referred to as phosphosites) was a product of changes in phosphorylation site stoichiometry or that of changes in protein expression [17]. Owing to the improved selectivity to reliably discriminate the targeted analytes from complex background interferences, parallel reaction monitoring (PRM)-based targeted quantification approach was further applied to validate specific de-differentiation marker in SMCs [18]. This study revealed novel expression patterns and identified new phosphoproteins or phosphosites in response to TGFβ or SDF-1α treatment, which are important to decipher molecular changes after vascular injury.

2. Experimental

2.1. Cell culture

Rat smooth muscle cells (SMCs) were isolated from the thoracoabdominal aorta of male Sprague-Dawley rats. SMCs at passages 4 to 5 were used for all experiments and maintained in DMEM supplemented with 10% fetal bovine solution (FBS) at 37 °C with 5% CO₂. Cell viability was >95% as assayed using the Trypan Blue exclusion method. Adenoviral vectors expressing Smad3 (AdSmad3) and control green fluorescent protein (AdGFP) were constructed as previously described [4]. SMCs were infected for 4h with AdSmad3 (or AdGFP) (3×10⁴ particles/cell) in DMEM containing 2% FBS, allowed to recover for 20 h with 10% FBS, and then starved with 0.5% FBS for 24 h followed by treatment with human recombinant TGFβ (5 ng/mL, R&D Systems, Minneapolis, MN) or equivalent amount of solvent (final 4 μM HCl and 1μg/mL bovine serum albumin) for 24 h. Half of the SMCs infected with AdSmad3 or AdGFP were treated with 100 ng/ml SDF-1 for 10 min at 37 °C. The dishes were washed three times with cold PBS media, and the SMCs pellet was collected. In total, four groups of SMCs were prepared based on different overexpression of protein (AdSmad3 or AdGFP) and ligand treatments (TGFβ and SDF-1α). Each group contained three biological replicates.

2.2. Cell lysis and protein digestion

The digestion buffer was prepared to the following specifications: 8 M urea, 50 mM Tris•HCl, 5 mM CaCl₂, 30 mM NaCl, 1x protease inhibitor tablet (Roche, Penzberg, Germany), and 1x phosphatase inhibitor tablet (Roche) at pH 8. Cell pellets were reconstituted in digestion buffer and lysed with a probe sonicator at 4 °C. The lysates were analyzed for total protein content via a bicinchoninic acid protein assay kit (Thermo Scientific Pierce, Rockford, IL). Equal amounts of total protein (250 μg) from each sample were reduced and alkylated by dithiothreitol and iodoacetamide, respectively. Trypsin (Promega, Madison, WI) was added to each sample at a 50:1 w/w protein:enzyme ratio and incubated at 37 °C for 18 hours. The digestion was quenched by acidification with 10% trifluoroacetic acid (TFA) to pH 3. Resultant peptides were purified via C18 Sep-Pak cartridges and dried down via SpeedVac.

2.3. DiLeu labeling

The complete set of 12-plex DiLeu tags was synthesized as previously described [16]. Each inactive label was suspended to 40 μg/μL in anhydrous N,N-dimethylformamide and combined with 0.7× limiting molar ratios of 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium tetra-fluoroborate and N-methylmorpholine. These solutions were then vortexed for one hour to yield activated amine-reactive DiLeu tags. The mixture was centrifuged at 8,000g for 1 min. The resulting supernatant was transferred immediately for peptide labeling. Peptide was resuspended in 0.5 M triethylammonium bicarbonate and labeled by adding the activated DiLeu at a 5x w/w tag excess. The reaction solutions were then vortexed for 2 hours and quenched by adding hydroxylamine to a final concentration of 0.25%. The tag scheme was as follows: AdGFP group-115a, 117a, 118b, AdSmad3/TGFβ group-115b, 116c, 118a, AdGFP/SDF-1α group-116a, 117c, 118c, AdSmad3/TGFβ/SDF-1α group-116b, 117b, 118d.

2.4. Phosphopeptide enrichment

Labeled samples were combined at a 1:1:1:1:1:1:1:1:1:1:1:1 ratio, purified with C18 Sep-Pak cartridges, and resuspended in 80% acetonitrile (ACN) 0.1% TFA. The IMAC phosphopeptide enrichment protocol was adapted from methods reported before [19]. Magnetic Ni-NTA beads (Qiagen, Hilden, Germany) were prepared with the following steps: wash with water (3x), 40 mM EDTA (pH 8) incubation with shaking for 45 minutes, wash with water (5x), 100 mM FeCl₃ incubation with shaking for 45 minutes, and a final wash with 80% ACN 0.1% TFA (3x). The sample was then added to the beads and vortexed for 60 minutes. The initial supernatant and subsequent three washes with 80% ACN 0.1% TFA were saved as the “phosphopeptide-depleted” sample. After two additional 80% ACN 0.1% TFA washes, phosphopeptides were collected by two elution with 50% ACN 0.7% NH₄OH and then neutralized with 4% formic acid (FA).

2.5. Sample fractionation

Prior to MS analysis, high-pH (HpH) reversed-phase (RP) high-performance liquid chromatography (HPLC) separation was performed on a Waters Alliance (e2695 separation module, 2489 UV/vis detector monitored at 215 nm). Elutions from the phosphopeptide-enriched sample were combined and resuspended in HpH solvent A (water, 10 mM NH₄HCO₂, pH 10) and separated on a 150 mm × 2.1 mm, 5 μm, 100 Å, C18 column (Phenomenex). After 3 minutes of column-loading in 100% HpH solvent A at 0.2 mL/min, HpH solvent B (90% ACN, 10 mM NH₄HCO₂, pH 10) was linearly ramped from 0% to 35% over 50 minutes. Initial fractions were collected every 1.5 minutes and recombined into a total of 20 final fractions based on the UV chromatogram trace.

Peptides not retained on IMAC beads from phosphopeptide-depleted sample were dried down via SpeedVac and resuspended in strong cation exchange (SCX) chromatography solvent A (10 mM NH₄HCO₂, pH 3). The sample was loaded on a 200 mm × 2.1 mm, 5 μm, 300 Å, poly(2-sulfoethyl aspartamide) column (PolyLC) with 100% SCX solvent A at 0.2 mL/min for 20 minutes. SCX solvent B (500 mM NH₄HCO₂, pH 6.8) was ramped with a linear gradient from 0% to 50% over 50 minutes. The excess DiLeu tags were eluted in the first 20 minutes. Fractions were then collected when SCX solvent B was initiated. All fractions were combined and resuspended in HpH solvent A (water, 10 mM NH₄HCO₂, pH 10) to perform HpH-RP HPLC fractionation as described above.

2.6. Mass spectrometry analysis

Online nano-LC separation was performed on a nanoAcquity UPLC (Waters Corporation). Phosphopeptide and peptide fractions were dried down, resuspended in 15 μL of 0.1% FA, and injected onto a self-fabricated capillary column (16 cm length, 75 μm i.d.) packed with reversed-phase C18 material (1.7 μm, 130 Å, Waters Corporation). Samples were loaded onto the column in 100% solvent A (water, 5% dimethylsulfoxide, 0.2% FA, pH 2–3) at a flow rate of 0.3 μL/min. Separation occurred during the following non-linear gradient of solvent B (ACN, 5% dimethylsulfoxide, 0.2% FA, pH 2–3): 0–36 min from 4% to 12%, 36–68 min from 12% to 22%, 68–80 min from 22% to 30%. Although the solid-phase chemistry of the HPLC and nano-LC columns were very similar, reversed-phase peptide separation at pH 10 is considered to be highly orthogonal to separation at pH 2–3 due to a substantial difference in net charges on peptides at different pH [20].

Phosphopeptides were acquired on an Orbitrap Elite mass spectrometer, and peptides were acquired on a Q Exactive mass spectrometer (Thermo Fisher Scientific). MS and tandem MS (MS/MS) spectra were collected via a top 15 data-dependent acquisition method. For precursor MS scans, the AGC, maximum injection time, and resolution (at m/z 400) were set to 1×10⁶, 100 ms, and 30000, respectively. The mass range for precursor scan was from m/z 350 to 1500. MS/MS fragmentation was performed via HCD with 28% normalized collision energy. The first mass was set to m/z 110. For tandem MS scans, AGC, maximum injection time, and resolution (at m/z 400) were set to 1×10⁵, 110 ms, and 60000, respectively. Dynamic exclusion time was set to 30 seconds. The threshold for minimum precursor intensity was 1000. The inclusion list for targeted PRM analysis contained the most abundant precursor ions of protein osteopontin, calponin (CNN), smooth muscle protein 22α (SM22α), and skeletal muscle actin (ACTA1).

2.7. Data analysis

All peptide and protein identification was performed using the COMPASS (v1.4) software suite [21], and the Open Mass Spectrometry Search Algorithm (OMSSA) [22]. Each file was searched against tryptic peptides from a target-decoy protein database for Rattus norvegicus (UP000002494, UniProt, canonical and isoforms, downloaded on 12/03/2015), allowing a maximum of two miscleavages. Carbamidomethyl cysteine, DiLeu N-terminus, and DiLeu lysine were selected as fixed modifications. Oxidation of methionine was set to variable modification. Additional variable modifications for the phosphoproteome were phosphorylation of serine with neutral loss, phosphorylation of threonine with neutral loss, and phosphorylation of tyrosine. The protein-level FDR and minimum peptides-per-group were set to 1% and 1, respectively.

After Protein Hoarder’s 1% FDR curation, remaining PSMs mapped to protein groups were subjected to PTM localization analysis by the PhosphoRS algorithm [23]. Class I (>75% localization probability) phosphopeptides were filtered out for further quantification. To facilitate accurate protein and phosphoprotein quantification via DiLeu, we created a custom software suite called The DiLeu Tool. Briefly, The DiLeu Tool used the raw data .mgf files, the Protein Hoarder output, and the PhosphoRS output to quantify expression levels of localized and unlocalized phosphoproteins. After PSMs were assigned to phosphopeptide by PhosphoRS, each PSM’s DiLeu tag intensities were retrieved from the .mgf file. After intensities were purity-corrected and normalized for channel-bias, a mean intensity for each tag was calculated at the phosphopeptide-level. For protein quantification, the DiLeu tool only used the raw data .mgf files and the ProteinHoarder output. Mean tag intensities were then assigned to their corresponding sample groups to calculate sample group-level (phospho)peptide intensities. Heteroscedastic Student’s t-test was performed for sample group-level binary comparisons. ANOVA-based multiple sample t-test was performed for quaternary comparisons. Full data analysis is provided in Supplemental Methods.

3. Results and discussion

3.1. Schematic illustration of highly multiplexed quantification of protein expression and PTMs in large datasets.

Isobaric labeling of proteins from vascular SMCs with 12-plex DiLeu tags was employed for quantitative (phospho)proteomic analyses. SMCs overexpressing Smad3 with additional TGFβ treatment (AdSmad3/TGFβ) or further stimulated with SDF-1α (AdSmad3/TGFβ/SDF-1α) were used to mimic molecular environment after vascular injury; in parallel, SMCs overexpressing green fluorescent protein (GFP) were used as control (Fig. 1). AdSmad3/TGFβ and AdSmad3/TGFβ/SDF-1α treated SMCs were the primary experimental groups, while groups of AdGFP and AdGFP/SDF-1α served as their respective controls. The 12-plex DiLeu tags were appropriately distributed into four groups of differentially treated SMCs to avoid quantification bias resulting from deuterium shift on retention time. DiLeu-labeled tryptic digests underwent phosphopeptide enrichment via immobilized iron (III) affinity chromatography (IMAC) [19], after which the workflow was split into two parallel experiments for the phosphopeptide-enriched portion and phosphopeptide-depleted portion. The enriched portion was fractionated via off-line high-pH reversed-phase high-performance liquid chromatography (HPLC) and analyzed by LC-MS/MS for identification and relative quantification of phosphoproteome. The depleted portion was subjected to strong cation exchange (SCX) chromatography to remove excess DiLeu components, fractionated by high-pH reversed-phase HPLC, and analyzed by LC-MS/MS for identification and relative quantification of the proteome.

Figure 1.

Schematic illustration for 12-plex DiLeu labeled quantitative proteomics and phosphoproteomics in SMCs. SMCs were infected with adenovirus expressing Smad3 with additional TGFβ treatment (AdSmad3/TGFβ) or control green fluorescent protein (AdGFP), or further stimulated with SDF-1α (AdGFP/SDF-1α, AdSmad3/TGFβ/SDF-1α) in triplicates. SMC lysates were digested with trypsin and strategically labeled with 12-plex DiLeu tags as indicated. Phosphopeptide enrichment was accomplished with immobilized iron affinity chromatography (IMAC). Both proteome and phosphoproteome samples were fractionated through high-pH reversed-phase (HpH-RP) chromatography prior to LC-MS/MS analysis.

We also developed a DiLeu-Tool software suite to facilitate quantification of protein expression and PTMs in large datasets. The main function of DiLeu-Tool is to correct the interferences of isotopic peaks to neighboring primary reporter ion peaks through a series of equations in the form of 12×12 matrix as detailed in Supplementary Methods. The DiLeu-Tool has been designed for use after database searching and phosphosite localization by COMPASS and PhosphoRS, respectively [21,23]. Peptide-spectrum matches (PSMs) were assigned to phosphopeptides by PhosphoRS, and each PSM’s DiLeu tag intensities were retrieved from the raw data. After intensities were purity-corrected and normalized for channel-bias, a mean intensity for each tag was calculated at the phosphopeptide-level. Mean tag intensities were then assigned to their corresponding sample groups to calculate sample group-level (phospho)peptide intensities. (Phospho)peptide abundances provided a complementary view of large-scale dynamics across the four sample groups, and thereby facilitated the determination of true alterations in phosphorylation from changes in protein expression by normalizing phosphopeptide ratios to the corresponding protein changes [24]. A detailed description of the DiLeu-Tool software can be found in Supplemental Methods.

3.2. Proteomic and phosphoproteomic profiling and quantification in SMCs.

Following the curation of Open Mass Spectrometry Search Algorithm (OMSSA) search results to a 1% protein-level FDR [22], a total of 4345 unique phosphopeptides corresponding to 1542 distinguishable phosphoprotein groups were identified. Phosphoserine, phosphothreonine, or phosphotyrosine modifications were identified on 11736 phosphopeptide spectral matches (PSM), corresponding to a PSM-level phosphopeptide enrichment efficiency of 81.9%. These PSMs were mapped to 2812 distinct phosphopeptides, 2317 of which were localized at the residue-level by PhosphoRS with a probability of 75% or greater (Class I). The relative frequencies of phosphosites with localized phosphoserine, phosphothreonine, and phosphotyrosine were 87%, 11%, and 1.0%, respectively (Fig. 2A). In total, 4086 proteins were quantified in the combined dataset of proteome and phosphoproteome across all 12-plex DiLeu-labeled SMC samples. Among those, 2893 proteins were unique to the proteome, 629 proteins were measured only in the phosphoproteome, and 564 proteins were quantified in both datasets (Fig. 2B).

Figure 2.

Quantitative proteomics and phosphoproteomics in SMCs. (A) Distribution of phosphorylation on serine, threonine, and tyrosine among the Class I phosphorylation sites. (B) Venn diagram showing the overlap of proteins and phosphoproteins quantified in the 12-plex DiLeu labeled SMCs. Histogram displaying all log₂-transformed ratios of proteins (C) and phosphosites (E) in the binary comparison of AdSmad3/TGFβ and AdGFP infected SMCs, as well as that of proteins (D) and phosphosites (F) upon AdSmad3/TGFβ- specific SDF-1α treatment via quaternary comparison of the four groups of differently treated SMCs. (G) Biological function enrichment analysis of regulated proteome and phosphoproteome between AdSmad3/TGFβ and AdGFP treated SMCs and changes in these biological functions upon AdSmad3/TGFβ-specific SDF-1α stimulation. A dotted line shows the significance level (p-value < 0.05).

To evaluate the regulation induced by elevated TGFβ/Smad3, we performed a binary comparison between the AdGFP and AdSmad3/TGFβ groups (student’s t-test). An ANOVA-based, multiple-sample t-test was performed for quaternary comparisons across all the groups to monitor the effects resulting from SDF-1α in the presence of AdSmad3/TGFβ, referred to as TGFβ/Smad3-specific SDF-1α (Fig. S2A). 790 proteins were significantly regulated after TGFβ stimulation with elevated Smad3 in the combined dataset of proteome and phosphoproteome (p-value < 0.05). Of those, 53 proteins were found in both datasets (Fig. S2B). TGFβ/Smad3-specific SDF-1α treatment stimulated 1235 proteins to be differentially regulated, of which 85 were quantified in both the proteome and phosphoproteome (Fig. S2C). Normalization of phosphosites was performed to the abundance-altered proteins with p-value < 0.05; 164 and 282 phosphosites were normalized in the study of TGFβ/Smad3 and TGFβ/Smad3-specific SDF-1α, respectively.

Our 12-plex DiLeu quantification strategy revealed significant changes in protein expression and phosphorylation across four different sample conditions. Using a threshold of two standard deviations to consider protein and phosphosites significantly up- or down-regulated (after ratio normalization by The DiLeu Tool), 42 proteins were significantly up-regulated, and 66 proteins were down-regulated in the proteome of SMCs treated with TGFβ/Smad3 (Fig. 2C and Table S1), whereas 21 proteins were up-regulated and 15 were down-regulated upon SDF-1α treatment in the presence of AdSmad3/TGFβ (Fig. 2D and Table S2). With regard to the degree of phosphorylation, TGFβ/Smad3 induced up-regulation of 40 phosphosites and down-regulation of 50 phosphosites (Fig. 2E and Table S3), and TGFβ/Smad3-specific SDF-1α facilitated up-regulation of 27 phosphosites and down-regulation of 47 phosphosites (Fig. 2F and Table S4). Although both histograms appear to be symmetrical and Gaussian-like, the TGFβ/Smad3 condition displays a wider distribution compared to additional SDF-1α stimulation for ratios within the significance thresholds.

To gain insights into the biological function over-represented among the dynamic proteins and phosphoproteins, we performed gene ontology (GO) enrichment analysis through Ingenuity System Pathways Analysis (IPA) software. Only proteins and phosphoproteins displaying differentiated changes (p-value < 0.05) were analyzed in IPA. The biological functions are organized based on the -log (p-value) of the top 15 ranked GO-terms in the phosphoproteome of TGFβ/Smad3-specific SDF-1α (Fig. 2G). As the intrinsic character of SMCs, cellular assembly and organization was the most over-represented biological function in the phosphoproteome of TGFβ/Smad3, with or without SDF-1α treatment. Cellular movement was significantly enriched in the proteome of TGFβ/Smad3 in the absence or presence of SDF-1α. As expected, cellular growth and proliferation were over-represented in both the proteome and phosphoproteome of elevated TGFβ/Smad3, and these biological processes were accelerated even further in SMCs following SDF-1α treatment, revealing that SDF-1α can facilitate cell proliferation and phenotype exchange of SMCs.

3.3. Hierarchical clustering of altered phosphoproteins across differentially treated SMCs.

Hierarchical clustering was conducted to globally characterize groups of phosphosites that show similar or opposite regulation trends among the differentially treated groups. 795 differential phosphosites remained after an ANOVA-based multiple-sample test (p-value < 0.05). Column-wise clustering of triplicate measurements of the four differently treated SMCs revealed that the intragroup experimental variation was smaller than intergroup biological variation. Row-wise clustering of phosphosites highlighted three patterns of protein phosphorylation that showed a similar or opposite trend among the four groups (Fig. 3A). In general, the two control groups (AdGFP and AdGFP/SDF-1α) kept a consistent expression trend, indicating that SDF-1α did not trigger large-scale activation or inhibition of protein phosphorylation in the absence of TGFβ/Smad3. Pattern 1 revealed proteins that were dephosphorylated after TGFβ/Smad3 treatment relative to matched controls, suggesting TGFβ/Smad3 may increase phosphatase or decrease kinase activity specific to these proteins or phosphosites. Pattern 2 consisted of protein phosphorylation levels that were elevated in SMCs treated with AdSmad3/TGFβ; however, SDF-1α treatment abrogated the effects of TGFβ/Smad3 and returned phosphorylation to the basal level. For example, elevated TGFβ/Smad3 led to increased phosphorylation of Ser2691 in AHNAK (log₂[ratio] = 1.33, p-value = 9 × 10⁻⁵), but TGFβ/Smad3-specific SDF-1α treatment significantly down-regulated phosphorylation of Ser2691 (log₂ [ratio] = −2.19, p-value = 1.25 × 10⁻⁸). A total of 24 phosphorylation sites for AHNAK were identified (Fig. S3). Pattern 3 included phosphosites that were up-regulated in TGFβ/Smad3 treated SMCs and increased even further upon SDF-1α incubation.

Figure 3.

Phosphorylation clusters in differently treated SMCs. (A) Hierarchical clustering of 12-plex DiLeu-labeled phosphopeptides among the differently treated SMCs. The signal intensity was converted into heat map colors using the minimum (green) and maximum (red) values in the entire data set. (B) Proteins that displayed changes in phosphorylation upon SDF-1α treatment were used to build an interaction network using the STRING database and selecting high stringency (score > 0.7) interactions inferred from experimental data. The size of node is proportional to the number of interacting partners involved in the network. (C) Five molecular complexes were clustered in the large protein-protein interaction network via MCODE. The biological function of each complex was annotated in DAVID.

To determine the relationships among these differential phosphoproteins, we performed protein-protein interaction in String [25,26] and GO enrichment analysis in DAVID [27] (Fig. 3B). Cell differentiation was the most significantly enriched biological process from Bonferroni statistical test (p-value = 6.45 × 10⁻⁸). Cardiac muscle alpha-actin (Actc1), involved in various types of cell motility, was the key node of this biological process with multiple functional interactions with other proteins. Cytoskeleton organization was the second most over-represented biological process (p-value = 3.63 × 10⁻⁷). Molecular complex detection (MCODE) clustering algorithm was further used to detect the densely connected subnetwork within the large protein-protein interaction network, which can be considered molecular complexes [28]. Five most prominent molecular complexes were predicted and demonstrated to be involved in the TGFβ signaling pathway, muscle contraction, relation of cardiac muscle contraction, mRNA processing, and mitotic spindle organization (Fig. 3C).

3.4. TGFβ/Smad3 inhibits the expression of contractile proteins in SMCs.

Next, we investigated the specific effect resulting from the elevated TGFβ/Smad3 in SMCs. Contractile-associated proteins and phosphoproteins, such as smooth muscle myosin heavy chain (MHC), calponin (CNN), smooth muscle protein 22α (SM22α), and cardiac muscle alpha-actin (ACTC1) were significantly down-regulated in TGFβ/Smad3 elevated group (Fig. 4). Calponin is an actin filament associated protein, which regulates smooth muscle contraction via interaction with F-actin and inhibition of actomyosin magnesium-ATPase. As reported, phosphorylation of Ser175 leads to the loss of both properties [29]. Ser175 and Thr180 were located in the second actin-binding site, which was phosphorylated by protein kinase C (PKC) and Ca²⁺/calmodulin-dependent protein kinase II [30]. This suggests that Thr180 could be another functional site involved in the formation of tensional forces. More interestingly, phosphorylation of CXCR4 on serine 316 was significantly up-regulated by 1.86 fold in AdSmad3/TGFβ-treated group compared to control (p-value < 0.05). Canonical pathway enrichment analysis demonstrated that TGFβ signaling pathway was over-represented in AdSmad3/TGFβ-treated SMCs with regarding the control AdGFP (p-value = 0.017). (Phospho)proteins associated with TGFβ signaling pathway are shown in Fig. S4. As expected, Smad3 and TGFβ were up-regulated in the proteome of AdSmad3/TGFβ-treated SMCs. The TGFβ treatment caused altered phosphorylation at three sites on Smad3, Thr132, Ser416, and Ser418. Phosphorylation of Ser418 was significantly down-regulated (-log₂[ratio] = −1.41) after normalization of protein Smad3 expression. Phosphorylation at Ser418 can promote ubiquitin-directed degradation [31], suggesting that TGFβ suppressed the phosphorylation of Ser418 to avoid proteasomal degradation of activated Smad3. Dual specificity mitogen-activated protein kinase kinase 1 (MAP2k1) is an important component of MAP kinase signal transduction pathway. MAP2k1 was significantly up-regulated (p-value < 0.05) in TGFβ/Smad3 treated group compared to control, indicating possible activation of the ERK signaling pathway.

Figure 4.

Changes in expression and phosphorylation of contractile proteins upon TGFβ/Smad3 treatment. Decreased expression and phosphorylation of SMC-specific contractile-associated proteins, including smooth muscle myosin heavy chain, calponin, SM22, and actin, and increased phosphorylation of CXCR4. An asterisk (*) indicates p-value < 0.05 based on two-sample student’s t-test. The error bars represent standard deviations.

3.5. Activation of ERK signaling pathway upon TGFβ/Smad3-specific SDF-1α treatment.

To further obtain insights into the function of CXCR4, SMCs treated with AdSmad3/TGFβ were incubated with SDF-1α, which bound to CXCR4 and activated downstream signaling. IPA canonical pathway analysis results showed that the ERK/MAPK signaling pathway was activated with SDF-1α treatment (p-value = 0.026). Thirty-four (phospho)proteins involved in ERK pathway were identified and quantified (Fig. 5). Phosphorylated serum response factor (SRF) was the most significantly up-regulated phosphoprotein. SRF functions as an important transcriptional regulator of both proliferation and differentiation genes. It regulates SMC differentiation marker gene expression by binding to a conserved CArG (CC(A/T₆GG) motif, which is present in almost all of the SMC-specific promoters [32]. This DNA-binding domain of SRF spans a 90-amino-acid peptide (aa 133–222), termed MADS box [33]. Phosphorylation of Ser162 in the MADS box has been reported to attenuate SRF-DNA binding via phosphate-phosphate repulsion and steric hindrance, which consequently inhibits expression of myogenic differentiation genes but activates immediate-early proliferation genes [34]. Phosphorylation of Thr159 by cAMP-dependent kinase (PKA) was also reported to repress SMC-specific transcription via attenuation of SRF-dependent smooth muscle α-actin promoter activity [35]. Along with Ser162 and Thr159, Ser220 was located in the MADS box and significantly up-regulated upon the treatment of SDF-1α (log₂ [ratio] = 0.53, p-value = 0.02), indicating that phosphorylation of Ser220 may be a function site to promote gene expression into proliferation status.

Figure 5.

Expression and phosphorylation changes in ERK signaling pathway upon TGFβ/Smad3-specific SDF-1α treatment. (A) Quantification data for proteins and phosphosites are color-coded: orange, protein up-regulated in proteome with p-value < 0.05; blue, protein down-regulated in proteome with p-value < 0.05; white, protein not quantified in proteome; red phosphate group, phosphosite up-regulated with p-value < 0.05; green phosphate group, phosphosite down-regulated with p-value < 0.05; gray, protein or phosphosite quantified but not significantly regulated. (B) Normalized fold change of phosphosites in A (see equation in Fig. S2A). An asterisk (*) indicates p-value < 0.05 from ANOVA-based multi-sample test across all four groups of differently treated SMCs. The error bars represent standard deviations.

PKC family members regulate the function of other proteins or themselves by phosphorylation of serine and threonine. The PKC family consists of three subfamilies, conventional PKCs (α, β, γ), novel PKCs (δ, ε, η, θ, μ), and atypical PKCs. βI and ε isoenzymes of PKC have been discovered to involve in the activation of ERK1/2 and cell proliferation of vascular SMCs [36,37], while PKCδ has dual roles in regulating cell proliferation and migration [38]. Structural and biochemical analyses of PKC confirmed that PKC-βII activation was regulated by three phosphorylation residues: T500 in the activation loop, and T641 and S660 in the carboxyl terminus [39]. The three residues are conserved among all the PKC family members. Protein expression of PKCδ showed no difference with the treatment of SDF-1α, but phosphorylation of S662 significantly increased by 1.32 fold (p-value = 0.04), indicating that phosphorylation of S662 located in the carboxyl terminus may also work as a positive regulator of cell proliferation.

Furthermore, IPA-based upstream regulator analysis identified versican as the most significant factor that might regulate protein change in response to TGFβ/Smad3 before SDF-1α treatment (p-value = 9.66 × 10⁻¹⁰) and following treatment (p-value = 4.15 × 10⁻¹¹). Versican is a large extracellular matrix proteoglycan, contributing to the expansion of pericellular ECM, which is required for cell proliferation and migration [40]. The regulation of versican was indicated to be inhibited by TGFβ/Smad3, but additional SDF-1α activated this whole regulatory system (Fig. S5).

3.6. SDF-1α stimulates the expression of protein osteopontin in SMCs.

Osteopontin is a phosphorylated acidic glycoprotein containing an Arg-Gly-Asp (RGD) motif which can interact with CD44 and several integrins to initiate cell migration [41]. Osteopontin protein and mRNA expression increased via α_vβ₃ integrin signaling in response to vascular injury in rat arterial SMC [42]. Our results showed up-regulation of osteopontin after AdSmad3/TGFβ treatment; the expression was further increased upon additional SDF-1α stimulation. Phosphorylation of osteopontin is important to inhibit human vascular smooth muscle cell calcification, but there is little information about the location of phosphorylation sites [43]. Our phosphoproteome data showed that the phosphorylation sites located at both N-terminal (Ser26, Ser27) and C-terminal (Ser312, Ser313) were down-regulated in AdSmad3/TGFβ with or without SDF-1α, suggesting that dephosphorylation of osteopontin may contribute to cell proliferation.

To confirm the expression of osteopontin across the four different treatment groups, parallel reaction monitoring (PRM) was employed for targeted analysis. Protein database search results from a discovery proteomics experiment were used to build a spectral library, and the chromatograms of predicted transitions were extracted using Skyline. Fig. 6A shows the tandem mass spectrum and top 8 PRM transitions of 12-plex DiLeu-labeled osteopontin peptide SDAIASQASSK, where a wealth of b- and y-ions are mapped for confident peptide identification. A dot-product score (dotp) of 0.96 for this peptide indicates the strong correlation between peak intensities of the transitions acquired from PRM and the library spectrum. Targeted quantification results displayed good reproducibility with coefficient of variation <10% across biological triplicates for each treated SMC group (Fig. 6B). It also confirmed that AdSmad3/TGFβ can stimulate up-regulation of osteopontin, and this effect was further strengthened by SDF-1α with a p-value of 0.004. Whereas, the contractile proteins including calponin (CNN), smooth muscle protein 22α (SM22α), and skeletal muscle actin (ACTA1) were all downregulated through the targeted analysis in AdSmad3/TGFβ-treated SMCs and the statistical significance was more prominent upon TGFβ/Smad3-specific SDF-1α stimulation (Fig. S6).

Figure 6.

Targeted quantification of osteopontin in SMCs. (A) Example tandem mass spectrum and PRM transitions of a 12-plex DiLeu-labeled osteopontin peptide SDAIDSQASSK. (B) PRM-based targeted quantification of 12-plex DiLeu-labeled osteopontin peptides in SMCs. The error bars represent standard deviations of three biological replicates.

4. Conclusion

Here, we present global quantitative proteomic and phosphoproteomic analyses of SMCs using high-resolution enabled 12-plex DiLeu isobaric tags, allowing simultaneous triplicate comparison of four sample conditions. The integrated proteomic and phosphoproteomic analyses provide a comprehensive view of system-wide signal transduction in SMCs. This investigation firstly demonstrated the feasibility of the 12-plex DiLeu tagging strategy for large-scale protein and post-translational modification (PTM) analyses to reveal the underlying molecular mechanism in a biological system. Additionally, we developed and employed a new software package, DiLeu Tool, to facilitate quantitative analyses of 12-plex DiLeu-labeled proteins and PTMs at large scale.

Our results demonstrate that TGFβ/Smad3 negatively regulates the expression of contractile proteins, such as smooth muscle myosin heavy chain, SM22α, calponin, and actin, which is in agreement with previous transcriptome result that TGFβ promotes SMCs dedifferentiation and loss of signature contractile characteristics. More interestingly, TGFβ/Smad3 activated phosphorylation of CXCR4 at the site of Ser316, which is located in CXCR4 45-amino acid serine/threonine-rich C-terminal as a potential PKC phosphorylation site [44]. Binding of SDF-1α to its receptor CXCR4 is crucial to mediate SMC mobilization and recruitment to injury sites after arterial injury. Pathway enrichment analysis demonstrated 34 proteins associated with ERK signaling pathway were activated upon TGFβ/Smad3-specific SDF-1α stimulation. Our current study revealed SDF-1α/CXCR4 axis could help drive the specific events that convert SMCs to the hyperplasia phenotype through activation of signaling pathways, such as ERK signaling. The significantly increased expression of osteopontin, a protein involved in neointima formation, further confirmed the phenotype switch. These data provide crucial hints to explore novel therapeutic targets for restenosis.

Highlights:

• The 12-plex DiLeu tagging strategy was implemented for large-scale protein and phosphorylation quantification in parallel for the first time.

• The DiLeu Tool software was developed to facilitate highly multiplexed quantification.

• The integrated proteomic and phosphoproteomic profiling provided a comprehensive view of SMC dedifferentiation.

• TGFβ/Smad3 regulated SMC dedifferentiation and SDF-1α/CXCR4 induced SMC migration through activation of ERK signaling.