IFATS Collection: Combinatorial Peptides Identify α5β1 Integrin as a Receptor for the Matricellular Protein SPARC on Adipose Stromal Cells

Abstract

The biological features of adipose stromal (stem) cells (ASC), which serve as progenitors for differentiated cells of white adipose tissue (WAT), are still largely undefined. In an initiative to identify functional ASC surface receptors, we screened a combinatorial library for peptide ligands binding to patient-derived ASC. We demonstrate that both primary and cultured human and mouse stromal cells express a conserved receptor targeted by peptides found to mimic SPARC, a matricellular protein that is required for normal WAT development. A signaling receptor for SPARC has not as yet been determined. By using the SPARC-mimicking peptides CMLAGWIPC (termed hPep) and CWLGEWLGC (termed mPep), isolated by panning on human and mouse cells, respectively, we identified the α5β1 integrin complex as a candidate receptor for SPARC. On the basis of these results, we evaluated ASC responses to SPARC or SPARC-mimicking peptide exposure. Our results suggest that extracellular SPARC binds to α5β1 integrin at sites of focal adhesions, an interaction disrupting firm attachment of ASC to extracellular matrix. We propose that SPARC-mediated mobilization of ASC through its effect on α5β1 integrin complex provides a functional basis for the regulation of WAT body composition by SPARC. We also show that α5β1 integrin is a potential target for ASC-selective intracellular delivery of bioactive peptides and gene therapy vectors directed by the SPARC-mimicking peptides.

INTRODUCTION

Mesenchymal stromal cells, also referred to as mesenchymal stem cells (MSC), are multipotent progenitors capable of differentiating into tissues of mesodermal origin, such as bone, cartilage, and adipose tissue [1, 2]. MSC were originally defined as fibroblastic colony-forming units (CFU-F) derived from mixed bone marrow cells plated on uncoated plastic [2, 3]. Subsequently, MSC have been found within organs other than bone marrow, one of which is white adipose tissue (WAT) [4].

Human WAT can be conveniently obtained in high abundance as a byproduct of minimally invasive procedures such as lipoaspiration or abdominoplasty. The stromal-vascular fraction (SVF) of WAT can serve for isolation of CFU-F with properties similar to those of CFU-F from bone marrow [5–8]. Depending on the species and the type of WAT, the SVF contains 20%–60% fibroblastic cells, which associate with microvasculature in vivo [8]. These adipose stromal cells, initially described as preadipocytes and now commonly also referred to as adipose stromal (stem) cells (ASC), are multipotent, as demonstrated by their in vivo differentiation into a number of tissues of mesenchymal lineage [6]. Notably, ASC may present a viable alternative to bone marrow MSC as a source of mesenchymal cells for autologous cell transplantation [6].

While the therapeutic potential of ASC is being extensively explored and shows great promise, characterization of molecules that mediate ASC functions, such as interaction with other cellular components of WAT, has remained limited. Identification of functional cell surface proteins could be applied for ASC prospective isolation, targeting, and potentially endogenous mobilization. As a step toward this goal, we initiated a systematic characterization of the ASC surface proteome. In this study, we used phage display in an approach that has previously yielded peptide ligands for receptors on various cell types [9–12] and allowed identification of the corresponding receptors [12–18] for cell- and tissue-directed delivery of drugs, imaging agents, and transgenes in vitro and in vivo [13, 19–21].

Selection of a combinatorial random peptide library against patient-derived ASC identified peptides mimicking SPARC (secreted protein, acidic and rich in cysteine). SPARC, also known as osteonectin, is a prototypical matricellular protein abundantly secreted by adipocytes [22–24]. SPARC has reported functions in adipose tissue development, as well as in tumor cell mobilization and metastasis [25, 26]. The potential role of SPARC in stem cell physiology has not been systematically explored, and a signaling cell surface receptor for SPARC has remained elusive. By using the SPARC-mimicking peptides as “bait,” we purified an ASC surface receptor for SPARC and identified it as a protein complex containing α5 (CD49e) and β1 (CD29) integrins, also known as the fibronectin receptor subunits, previously reported to mark MSC in various organs, including WAT [6, 27–31]. Our results strongly suggest that SPARC binds to the α5β1 integrin complex at focal adhesions. This interaction is consistent with the dependence of adipocyte differentiation on α5β1 interaction with extracellular matrix (ECM) and establishes a possible functional basis for ASC adhesion in the regulation of WAT growth. The SPARC-mimicking peptide ligands allowed directed intracellular delivery of a proapoptotic peptide and of a chimeric adeno-associated virus phage (AAVP) vector expressing an antilipogenic transgene. Together, these results provide insight into the role of SPARC-α5β1 interaction in WAT homeostasis and establish a foundation for prospective approaches to target ASC in vivo.

MATERIALS AND METHODS

Adipose Cell Isolation and Culture

Human subcutaneous WAT sample was obtained by a lipoaspiration procedure under informed consent from a deidentified Caucasian donor (patient 1: female, age 43, body mass index [BMI] 33) and used as a source of ASC used in the experiments shown in Figures 1–7. To address the possibility of interindividual variability in the ASC receptor expression, the critical binding and migration experiments were reproduced (supplemental online Fig. 2) with ASC derived from four more donors (patient 2: male, age 31, BMI 27.5; patient 3: female, age 48, BMI 24; patient 4: female, age 31, BMI 20; patient 5: female, age 46, BMI 38.5). As source of mouse ASC, visceral WAT isolates from C57Bl/6 mice fed regular chow or prefed a high-fat diet to induce obesity [16] have been used with similar results. ASC were isolated as previously described [5–8]. Briefly, minced WAT was digested in dispase (BD Biosciences, San Diego, http://www.bdbiosciences.com)/collagenase type I (Worthington Biochemical, Lakewood, NJ, http://www.worthingtonbiochem.com) solution under gentle agitation for 1 hour at 37°C, filtered on 500- and 250-µm Nitex filters (BD Biosciences), centrifuged at 200g for 5 minutes to separate the SVF pellet from adipocytes, and extensively washed. The SVF was plated in EGM-2MV (Cambrex, Walkersville, MD, http://www.cambrex.com) or Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum (FCS) on uncoated plastic, and after overnight attachment of ASC, the majority of contaminating hematopoietic cells were washed off. For library selection, passage 1 human ASC were used, whereas for the remainder of the experiments, mouse and human ASC from passages 4–10 grown in DMEM/FCS or EBM-2/FCS (Cambrex) were used. Depletion of hematopoietic endothelial cells was confirmed by flow cytometry with CD45 and CD31 antibodies, respectively, using LSR II (BD Biosciences). Differentiation of 3T3-L1 postconfluent preadipocytes was induced by treatment with insulin, dexamethasone, and isobutylmethylxanthine in DMEM/FCS [32]. For the cell migration assay, mouse ASC were detached with 2.5 mM EDTA and resuspended in DMEM/FCS containing the indicated concentrations of recombinant human SPARC (rhSPARC), hPep, or a control peptide, and 5 × 10⁴ cells were applied to the upper chamber of a 6.5-mm membrane well (polycarbonate; pore size, 5 µm) in Transwell polystyrene plates (Corning Costar, Acton, MA, http://www.corning.com/lifesciences). Migrated cells were counted after attachment in the lower chamber 1 day later by washing with phosphate-buffered saline (PBS) and staining with Crystal Violet (Fisher Scientific International, Hampton, NH, http://www.fisherscientific.com).

Figure 1.

Adipose stem cell (ASC)-binding Peps identify potential domains of SPARC-cell interaction. (A): Binding of phage to human ASC in individual rounds of a random Pep library selection. Relative phage recovery in subsequent rounds of selection reveals a progressive increase in phage binding. For the insertless phage (fd-tet), used as a negative control for each round of selection, mean binding is shown. (B): Human ASC binding of hPep-phage and mPep-phage. In (A, B), the ratio of phage TU bound over TU applied to cells is plotted. Negative controls: fd-tet and phage displaying an unrelated (control) Pep sequence. Error bars: SEM (%) from the three rounds (A) or from two independent experiments (B). (C): Amino acid sequences of Pep selected on human ASC (green underline) and mouse 3T3–L1 cells (orange underline) matched to the sequence of human SPARC (signal Pep sequence not included). Previously characterized domains of SPARC are indicated; pep4.2 and pep2.1 sequences are highlighted in yellow. Pep similarity criteria: at least three amino acids identical (red) and at least one similar (blue) to the correspondingly positioned SPARC amino acids. Abbreviations: E-C, extracellular Ca(2+)-binding domain; Pep, peptide.

Figure 7.

SPARC and ASC migration. (A): Activation of ASC motility by recombinant human SPARC and hPep, compared with untreated cells (mock) and cont peptide, assayed by quantification of cell transmigration through 5-µm-pore membranes. Shown are numbers (±SEM) of migrated cells within 24 hours from quadruplicate wells. *, statistically significant (p < .03) increase in cell migration over levels observed for cont peptide. (B): A model for SPARC function in white adipose tissue (WAT). ASC reside in the perivascular niche via firm adhesion to the ECM/BM through α5β1 integrin. SPARC (SPARC+) maintains balance among ASC quiescence, proliferation, and differentiation: nutrients supplied through systemic circulation result in ASC lipogenesis and their differentiation into adipocytes that secrete SPARC, leading to deadhesion and proliferation/migration of other uncommitted ASC. Thus, upon continuous nutrient input, adipogenesis is accompanied by concomitant proliferation and migration of ASC for their incorporation into WAT neovasculature, resulting in tissue growth. In the absence of SPARC (SPARC−), ASC differentiate into adipocytes that do not secrete SPARC, which promotes ECM/BM attachment of ASC, hence shifting the balance toward extended lipogenesis, resulting in adipocyte hypertrophy. Arrows, cell fates; red line, point at which SPARC interferes with the interaction between α5β1 and ECM. Abbreviations: ASC, adipose stem cells; BM, basement membrane; cont, control; ECM, extracellular matrix.

Isolation and Quantification of Phage-Peptide Binding to Cells

Random peptide library or isolated clones of M13-derived bacteriophage based on the vector fUSE5 displaying the insert CX₇C on the pIII protein were selected on cells by the biopanning and rapid analysis of selective interactive ligands (BRASIL) method [11]. Cells were detached with 2.5 mM EDTA and were resuspended in DMEM containing 1% bovine serum albumin (BSA); 10⁵ cells were incubated with 10⁹ transforming units (TU) of phage. The cell mixture was passed by centrifugation through the organic phase, and bound phage were recovered, quantified, and, when necessary, processed for sequencing of the peptide-coding DNA [33].

Peptide Sequence Analysis

Sequence similarity search for proteins mimicked by ASC-binding peptides was performed with BLAST (http://www.ncbi.nlm.nih.gov/BLAST). For identification of peptide motifs mimicking SPARC, Peptide Match software was codified in Perl 5.8.1 on the basis of RELIC [34]. The program searches for similarity between a seven-mer peptide sequence and a protein sequence by matching every hexapeptide comprising the seven-mer (in both orientations) with every hexapeptide comprising the protein from the amino (N) to the carboxy (C) terminus in single-residue shifts. The peptide-protein similarity scores for each residue were calculated on the basis of a BLOSUM62 amino acid substitution matrix modified to adjust for rare amino acid representation. Mapping of sequences along the protein was performed with web-based software codified in Perl 5.8.1 and Common Gateway Interface, and sequences were graphed with R-Project, version 2.2.1.

SPARC Receptor Purification Form ASC Protein Extract

Peptides hPep, mPep, and the control peptide CARAC [21] were chemically synthesized, cyclized via their N- and C-terminal cysteines, and purified to at least 95% purity (Polypeptide Laboratories, Torrance, CA, http://www.polypeptide.com). For protein extraction, a culture of cells isolated from a patient used for library selection (Fig. 1) was expanded in EBM-2/FCS. Cells from 10 passages (10⁸ cells per passage), detached with 2.5 mM EDTA and washed in 4°C PBS containing 0.2 mM phenylmethylsulfonyl fluoride (PMSF)/Complete-11697498 Protein Inhibitor cocktail (Roche, Mannheim, Germany, http://www.roche.com) PI cocktail (Complete-11697498), were pelleted and frozen in aliquots at −80°C. All the aliquots were thawed and pooled for extraction by solubilization in column buffer (PBS containing 1 mM CaCl₂ 1 mM MgCl₂, 50 mM n-octyl-β-D-glucopyranoside, 0.2 mM PMSF/Roche protein inhibitor cocktail), with subsequent incubation overnight at 4°C. High-speed centrifugation at 4°C was performed to remove insoluble debris. For affinity chromatography, 1 mg of peptide was coupled (via its C terminus) onto 1-ethyl-3-(3-demethyaminopropyl)carbodiimide-HCI Sepharose (Pierce, Rockford, IL, http://www.piercenet.com), and the column was equilibrated with column buffer containing 1% Triton X-100. Peptide-coupled resin was incubated with 20 mg of ASC extract, washed extensively with column buffer by gravity flow, and depleted of minimally bound proteins by a control elution with 2 mM untargeted peptide (CARAC) in column buffer. Bound proteins were eluted with 2 ml of 2 mM hPep or mPep peptide, followed by 0.1 M glycine (pH 2.8); protein in the eluate was monitored by absorbance at 280 nm. Aliquots of eluted 0.5-ml fractions normalized for protein concentration were subjected to SDS-polyacrylamide gel electrophoresis (PAGE) and either stained with Gelcode (Pierce) or processed for immunoblotting. Proteins transferred to Trans-Blot nitrocellulose membranes (Bio-Rad, Hercules, CA, http://www.bio-rad.com) were blocked, and they were probed with mouse anti-human β1 (1 µg/ml; R&D Systems Inc., Minneapolis, http://www.rndsystems.com), goat anti-human α5 (0.1 µg/ml; R&D Systems), or mouse anti-human α6 (1:2,000; Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com) antibodies in PBST (PBS/0.1% Triton X-100); signal was detected with secondary alkaline phosphatase IgG conjugates (1:5,000; Bio-Rad) and the nitroblue tetrazolium (NBT)/5-bromo-4-chloro-3-indolyl phosphate (BCIP) substrate system (Roche).

Immunofluorescence

Cells grown in chamber slides (Lab-Tek II; Nalge Nunc International, Rochester, NY, http://www.nalgenunc.com) were washed and subsequently incubated with 0.1 µg/ml hPep, 0.1 µg/ml mPep, or the indicated concentration of rhSPARC [35] prior to fixation with 4% paraformaldehyde. Fixed cells were rendered permeable with 0.2% Triton X-100, blocked in 5% serum, and incubated for 12–16 hours in PBST containing the following primary antibodies: rabbit anti-SPARC peptide 4.2 IgG [36] (1:1,000), rabbit anti-rhSPARC IgG [35] (10 µg/ml), mouse anti-human β1 IgG (10 µg/ml; R&D Systems), rat anti-mouse β1 IgG (20 µg/ml; Chemicon, Temecula, CA, http://www.chemicon.com), mouse anti-paxillin IgG1 (BD Biosciences; 1:1,000), or antisera against cyclized KLH-coupled peptides produced in rabbits (1:100; Genemed Synthesis Inc., San Francisco, http://www.genemedsyn.com). Phage internalization was measured as described [37] by incubation of blocked cells for 1 hour in medium containing 30% FCS and subsequently for 12 hours with 5 × 10⁹ TU of phage in medium containing 2% FCS. Unbound phage particles were washed off with PBS containing 10% BSA followed by 50 mM glycine/150 mM NaCl buffer (pH 2.8) and subsequent incubation with rabbit anti-fd phage antibody (1:500; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com). After washes in PBST, cells were incubated with the following donkey secondary antibodies (Jackson Immunoresearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com): anti-mouse fluorescein isothiocyanate-conjugated IgG (1:300) and anti-rabbit Cy3-conjugated IgG (1:400). DNA was labeled with Hoechst 33258 as necessary, and immunofluorescence images were acquired with a Leica DMR upright fluorescence microscope (Leica, Heerbrugg, Switzerland, http://www.leica.com) or an Olympus IX70 inverted fluorescence microscope (Olympus, Tokyo, http://www.olympus-global.com).

Induction and Quantification of Apoptosis

3T3-L1 cells 5 days postinduction for adipocyte differentiation [32] were treated in triplicates in 96-well plates for 24 hours at 37°C with increasing concentrations of mPep-_D(KLAKLAK)₂ or control peptides: CGRRAGGSC-GG-_D(KLAKLAK)₂ or _D(KLAKLAK)₂ peptide alone (AnaSpec, Inc, San Jose, CA, http://www.anaspec.com). Cell survival was evaluated with the WST-1 assay (Roche) and subsequent absorbance at 440 nm, as well as by phase-contrast microscopy.

Design of the Targeted AAVP for Cell Transduction

Plasmids for the production of targeted AAVP particles were made according to established protocols [19]. The recombinant AAV carrying the mouse uncoupling protein 1 (Ucp1) expression cassette was inserted into the intergenomic region of a phage fUSE5-MCS vector [38] encoding the pIII-mPep fusion protein. The resulting construct was confirmed for Ucp1 expression by transfection of mammalian cells with AAVP-Ucp1 DNA. The AAVP particles were produced in K91Kan Escherichia coli. For transduction, 10⁶ TU per cell of AAVP was applied in insulin-containing differentiation medium to 3T3-L1 cells that had been preinduced (2 days) for adipocyte differentiation. Ucp1 expression was confirmed by reverse transcription-polymerase chain reaction (RT-PCR) with Ucp1-specific primers and by immunofluorescence with anti-Ucp1 polyclonal antibodies (1:1,000; RDI, Concord, MA, http://www.researchd.com).

RESULTS

ASC-Binding Peptides Map Functional Domains of SPARC

We isolated SVF cells from patient-derived (liposuction) WAT and enriched them for ASC through depletion of endothelial and hematopoietic cells on the basis of established protocols [5, 7, 8]. To isolate ligands binding to ASC surface receptors, we selected a phage-displayed random cyclic CX₇C (C, cysteine; X, any residue) peptide library [33] on human ASC by the BRASIL method [11]. To attempt to identify sequences with conservation of both sequence and function between human and mouse preadipocyte cell surface proteomes, we performed an identical selection on mouse 3T3-L1 cells, a preadipocyte cell line commonly used for the study of adipogenesis [32]. In each screen, we observed progressively increasing phage recovery over the course of three rounds of biopanning (Fig. 1A), indicative of binding peptide selection for both human and mouse cells. For each screen, we determined sequences of peptides in 96 clones of phage selected in the third round of selection. The most frequent phage clones displaying peptides CMLAGWIPC (human cell screen) and CWLGEWLGC (mouse cell screen), termed hPep and mPep, respectively, were compared individually with insertless fd-tet phage and phage displaying an unrelated peptide sequence: there was ~10-fold more efficient binding of these two peptides to human ASC over the background observed for the controls (Fig. 1B; supplemental online Fig. 1). Analysis with Student’s t test adjusted for multiple comparison by the Benjamini-Hochberg method indicated a statistically significant difference in the TU recovery between hPep and insertless phage (p = .03), as well as between mPep and insertless phage (p = .03), whereas the difference in the TU recovery between control and insertless phage was not found significant (p = .52).

Comparison of hPep and mPep sequences revealed identities for two corresponding amino acid positions and class conservation for another two positions (Fig. 1C). This suggested that hPep and mPep bind to a receptor system conserved between human and mouse cell surface proteomes, consistent with comparable binding efficiency of the two peptides to ASC (Fig. 1B; supplemental online Fig. 2A). It is well recognized that selections of combinatorial peptide libraries on cells yield ligand peptides that bind to cell surface receptors via structural mimicry of the corresponding native biological ligands of these receptors. Often, such mimicry can occur at the primary protein structure level [11, 12, 15, 17, 18]. Accordingly, we used the sequence of ASC-binding peptides for identification of ligand-receptor interactions operating in the context of the ASC cell surface. Through automated searching of protein sequence databases [12, 17, 18], we identified the motif present in hPep and mPep peptides to be similar to a sequence within the matricellular protein SPARC (Fig. 1C).

SPARC is a prototypical matricellular protein that modulates interactions between cell surface and ECM molecules, thus controlling cell adhesion, differentiation, and angiogenesis [24–26, 39]. Automated computer-assisted alignment [12, 17] of all sequences isolated in the library selection against the SPARC amino acid sequence revealed that 50 of 192 (26%) of the selected peptides had similarity to SPARC. Moreover, peptides appeared to cluster within several “hot-spots” (Fig. 1C), indicating that the matches were not random. Notably, the primary hotspots correspond to the peptides 2.1 and 4.2 (pep4.2), previously identified as key sequences that mediate binding of SPARC to cells and inhibit cell attachment [36]. Consistent with their proposed binding to an interspecies conserved receptor system, the ASC-binding mPep and hPep were found to correspond to the cell-binding pep4.2, which is 100% conserved between human and mouse SPARC. To demonstrate the significance of the similarities, we performed a control computer-assisted analysis of a matching number of peptide sequences from unselected library using the same algorithm. Fisher’s exact test indicated that peptides selected on both human ASC and 3T3-L1 cells have a cumulative similarity match to SPARC pep4.2, being statistically significant compared with unselected library (p = .002 and .004, respectively).

To confirm that the ASC-binding mPep and hPep mimic SPARC pep4.2, we performed immunoblotting and immunofluorescence (Fig. 2) experiments on human ASC exposed to hPep. Antisera raised against hPep (Fig. 2B) and against mPep (Fig. 2C), but not preimmune sera, produced a pattern highly similar to that obtained with anti-SPARC pep4.2 antibodies (Fig. 2A). This staining pattern is reminiscent of the distribution of focal adhesions revealed with antibodies against markers such as paxillin [40]. Consistent with cell surface secretion of endogenous SPARC by cultured ASC [25, 39, 41], these data suggest that both anti-SPARC and anti-peptide antibodies recognize SPARC on the cell surface. Addition of soluble hPep (Fig. 2D–2F), mPep (not shown), or rhSPARC (Fig. 4D) to the ASC in serum-depleted medium prior to fixation resulted in an enhancement of the observed pattern, suggesting that these SPARC-mimicking peptides can bind to a SPARC receptor on the ASC surface. These data confirm that these ASC-binding peptides indeed map the domain of SPARC that interacts with its cell surface receptor.

Figure 2.

Adipose stem cell (ASC)-binding mPep and hPep mimic SPARC pep4.2 domain. Cultured human ASC were incubated in serum-free medium without (A–C) or with (D–F) synthetic hPep. Immunofluorescence on fixed cells was performed (without permeabilization) with antibodies against SPARC pep4.2 (A, D), against hPep SPARC-mimicking peptide (B, E), and against mPep SPARC-mimicking peptide (C, F). Signal (red) was revealed with secondary Cy3-conjugated antibodies; nuclei are blue. Note focal signal pattern identical for anti-SPARC pep4.2, anti-hPep, and anti-mPep antibodies (corresponding to native extracellular SPARC), which was enhanced by the addition of hPep to the cells. Scale bar % 5 µm.

Figure 4.

SPARCβ1 interaction at focal adhesions. Human adipose stem cells (ASC) (A, B), or mouse ASC (C–H) were incubated without (A, C) or with rhSPARC in serum-free medium at 40 µg/ml (B, D) or 5 µg/ml (E–H). Immunofluorescence on fixed cells was performed with (A) or without (B–H) permeabilization with anti-SPARC peptide 4.2 antibodies (A–D, F, H) and anti-β1 antibodies (E, F) or anti-paxillin antibodies (G, H). Signal was detected by Cy-3 conjugated (red) and fluorescein isothiocyanate-conjugated (green) secondary antibodies. Focal colocalization of SPARC and integrin β1 (arrowheads) is indicated (E, F), and the identity of these foci was confirmed by colocalization with anti-paxillin antibodies (G, H). Note change in cell shape in response to focal adhesion disassembly by 40 µg/ml rhSPARC (D). Scale bar % 5 µm (A–D) and 1 µm (E–H). Abbreviation: rhSPARC, recombinant human SPARC.

Biochemical Purification of SPARC Receptor from ASC

To identify the ASC surface receptor of SPARC, we used the SPARC-mimicking peptides for biochemical isolation of their membrane protein targets. Proteins extracted from the ASC used for library selection (Fig. 1) were used for affinity chromatography on carboxy-coupled hPep. Specifically bound proteins, eluted with hPep and mPep, appeared in discrete fractions. Testing of the eluates coated onto plastic for phage-peptide binding confirmed that several fractions contained proteins binding to hPep- and mPep-displaying phage but not to control (insertless) phage. These fractions were concentrated and resolved by denaturing gradient PAGE. Four bands eluted with hPep migrated at ~250, 125, 70, and 55 kDa (Fig. 3A). Mass spectrometric analysis of these bands revealed that the 125-kDa band contained integrin chains α5 (CD49e) and β1 (CD29), whereas the 250-kDa band contained talin-1, and the 70-/55-kDa bands contained fragments of β-actin and the chaperone GRP-58 (also termed PDIA-3 and ER-60). We also performed an independent purification of the SPARC receptor with mPep as bait, which yielded an identical protein profile (data not shown). Immunoblotting of elution fractions with antibodies specific to individual integrin chains confirmed isolation of α5 and β1 integrins with hPep and mPep, whereas control antibodies did not detect other integrins, such as α6 (Fig. 3B). The transmembrane α5β1 integrin heterodimer has previously been shown to execute cell signaling via interaction with intracellular talin-1 and actin filaments [28]. Therefore, our results indicate that both mouse and human SPARC-mimicking peptides interact with an intact multiprotein complex, with the transmembrane integrin component [27] corresponding to the apparent cell surface target of SPARC.

Figure 3.

Purification of SPARC receptor from adipose stem cells (ASC). SPARC-mimicking peptide hPep (1 mg) immobilized on resin was incubated with 30 mg of protein extract from human ASC. SPARC-binding ASC proteins were purified by washing, mock elution with an unrelated peptide CARAC (cont. elution), and subsequent elution with hPep. Proteins eluted with the peptide (elution fr. 1–4) and with 0.1 M glycine, pH 2.5 (postelution), were collected (0.5 ml each) and resolved by gradient 4%–15% SDS-polyacrylamide gel electrophoresis. (A): Gelcode staining of a gel resolving the collected fractions. Proteins corresponding to the four bands eluted with hPep (arrowheads) identified by mass spectrometry are indicated on the right. (B): Immunoblotting of the collected fractions with antibodies against integrins α5, β1, and α6 confirmed the specific isolation of α5 and β1 integrins (arrows) with hPep. Abbreviations: cont., control; fr., fraction.

SPARC-Integrin Functional Interaction at Focal Adhesions

To confirm binding of SPARC to the α5β1 integrin complex on the cell surface, we performed a series of experiments with human and mouse ASC (Fig. 4). Anti-SPARC immunofluorescence on detergent-treated human ASC confirmed intracellular distribution of SPARC (Fig. 4A), consistent with the reported expression pattern [41]. In agreement with the established role of secreted SPARC in effecting a state of intermediate adhesion compatible with cell motility [24–26, 39], immunofluorescence on detergent-untreated human ASC revealed an association of SPARC with sites on the cell membrane that were reminiscent of focal adhesions (Fig. 4B). For both human and mouse ASC, the focal cell surface SPARC signal was essentially identical to the pattern observed upon incubation of ASC with the SPARC-mimicking peptides (Fig. 2D–2F). Confirming the reported function of SPARC in focal adhesion disassembly, addition of 1 #M rhSPARC to firmly attached and fully spread human (Fig. 4A) and mouse (Fig. 4C) ASC resulted in their contraction and change in shape (Fig. 4B, 4D). Consistent with α5β1 integrin as the receptor for SPARC (Fig. 3), anti-integrin immunofluorescence revealed a high degree of β1 colocalization with the cell surface foci staining for SPARC, whereas a control integrin (α6) was not detected (Fig. 4E, 4F; data not shown). Finally, we used antibodies against paxillin (Fig. 4G), a focal adhesion marker, which colocalized with β1 integrin [42, 43]. Our results demonstrate that the cell surface foci of SPARC overlap with the sites of β1/paxillin coexpression (Fig. 4H) and thus strongly suggest that cell surface SPARC binds to α5β1 integrin at sites of focal adhesions.

α5β1 Integrin Complex Internalizes SPARC-Mimicking Peptides

Integrins on the cell surface have been shown to internalize their corresponding ligands, a property useful for cell targeting [21, 38, 44]. To test whether α5β1 integrin could be used as a portal for experimental intracellular delivery of therapeutic compounds, we performed a phage internalization assay [37] with hPep and mPep. By anti-phage immunofluorescence of cultured cells preincubated with phage displaying hPep and mPep, we demonstrated that phage particles were internalized by human (Fig. 5A, 5B) and mouse (Fig. 5E) ASC. In cultures of mouse ASC, phage particles were internalized predominantly by the smaller, actively proliferating cells with compact nuclei, which formed rapidly expanding colonies, but not by the larger, spread cells with diffuse nuclei typical of quiescence (Fig. 5E). The cytoplasmic accumulation of internalized phage was not observed for the insertless fd-tet control (Fig. 5C, 5F). Control non-MSC cells (human PC-3 prostate carcinoma) did not internalize hPep-phage (Fig. 5D) or mPep-phage (data not shown), indicating ASC target selectivity.

Figure 5.

α5β1 integrin internalizes SPARC-mimicking peptides. Culture of human adipose stem cells (ASC) (A–C), human prostate PC3 carcinoma cells (D), and mouse ASC (E, F) subjected to internalization of 10⁹ TU phage displaying hPep (A, D, E), mPep (B), or control insertless phage (C, F). Washed cells were fixed and analyzed for intracellular phage detected with Cy-3 immunofluorescence (red). Peptide uptake was particularly high in clones of actively dividing small ASC populations with compact nuclei (arrows) but not in quiescent cells (arrowheads), as revealed by DNA fluorescence (blue [E, F]). Scale bar = 5 µm.

Proapoptotic Cell Targeting with SPARC-Mimicking Peptides

To confirm that α5β1 integrin can serve as a viable ASC target in a functional assay, we used an amphipathic peptide sequence (KLAKLAK)₂ that disrupts mitochondrial membranes and causes apoptosis upon receptor-mediated cell internalization [16, 37, 45]. A synthetic peptide composed of the ASC-targeting mPep peptide in tandem with the proteolysis-resistant D-enantiomer of (KLAKLAK)₂ linked by a glycinylglycine bridge, termed mPep-_D(KLAKLAK)₂, was incubated with a culture of mouse 3T3-L1 preadipocytes induced toward adipocyte differentiation. According to a quantitative cell survival assay [37], mPep-_D(KLAKLAK)₂ induced rapid apoptosis in cultured preadipocytes, whereas nontargeted _D(KLAKLAK)₂ or _D(KLAKLAK)₂ fusion with an unrelated peptide targeting interleukin-11 receptor [18, 37] had no effect (Fig. 6A). Analysis with Student’s t test adjusted for multiple comparison by Benjamini-Hochberg method indicated a statistically significant increase in the peptide-induced cell death by 0.1 mM mPep (p = .003), whereas 0.1 mM control peptide-KLAKLAK₂ did not significantly induce cell death (p = .36). Microscopic analysis of the cell culture confirmed a characteristic apoptotic phenotype of cells treated with 0.1 mM mPep-_D(KLAKLAK)₂ and no effect after exposure to the control _D(KLAKLAK)₂ peptide fusions at the same molar concentration (Fig. 6B, 6C).

Figure 6.

Targeted intracellular peptide and gene delivery with SPARC-mimicking peptides. (A): Dose-response effect of mPep-GG-_D(KLAKLAK)₂ on mouse 3T3–L1 preadipocytes induced toward adipocyte differentiation. Cells were treated with increasing concentrations of mPep-GG-_D(KLAKLAK)₂ or negative controls (_D(KLAKLAK)₂ untargeted or fused with an unrelated peptide) for 24 hours. Apoptosis was assessed by the WST-1 assay. Error bars: SEM from triplicate wells. (B, C): Micrographs of 3T3–L1 cells in (A) after 24 hours of treatment: 0.1 mM mPep-GG-_D(KLAKLAK)₂ induced apoptosis (B), whereas untargeted _D(KLAKLAK)₂ had no effect on 3T3–L1 differentiation (C). (D–G): Delivery of an antilipogenic transgene (Ucp1 expressed from AAVP) with SPARC-mimicking peptides into 3T3–L1 cells induced toward adipocyte differentiation. AAVP-Ucp1 displaying mPep efficiently transduced preadipocytes, which resulted in Ucp-1 expression detected (arrow) with anti-Ucp1 antibody (red immunofluorescence) 5 days post-transduction (D), whereas untargeted AAVP-Ucp1 did not transduce preadipocytes (E). Micrographs taken 7 days post-transduction demonstrate that expression of Ucp-1 upon mPep-targeted AAVP-Ucp1 delivery prevented adipocyte differentiation of 3T3–L1 cells (F), whereas the untargeted AAVP-Ucp1 vector did not prevent their differentiation into adipocytes (G). Scale bar % 5 µm. Abbreviations: AAVP, adeno-associated virus phage; OD, optical density.

Cell Delivery of Gene Expression Vectors with SPARC-Mimicking Peptides

We also asked whether the SPARC-α5β1 ligand-receptor system could be used for experimental gene delivery. Recently, a chimeric prokaryotic-eukaryotic vector, termed AAVP, was developed that can display a peptide of interest on phage particles and deliver a gene expression cassette into mammalian cells upon transduction [19]. As a proof of principle, we chose to direct AAVP to mouse cells for targeted expression of Ucp1. Expression of Ucp1 modulates the metabolic state of adipocytes, thus resulting in the inhibition of lipogenesis [46]. We constructed an AAVP-Ucp1 vector displaying mPep on the particle surface and used it to transduce mouse 3T3-L1 preadipocytes induced to differentiate into adipocytes. Five days after transduction, Ucp1 was detected by RT-PCR and by immunofluorescence with anti-Ucp1 antibodies in ~30% of preadipocytes (Fig. 6D). In contrast, an equivalent aliquot of nontargeted AAVP-Ucp1 vector failed to deliver the Ucp1 cassette into cells, as assessed by both RT-PCR and immunofluorescence (Fig. 6E). Consistent with the published kinetics of the transgene expression in the AAVP system [38, 47], 7 days of expression of the SPARC mimetic-directed Ucp1 transgene resulted in inhibition of adipocyte differentiation (Fig. 6F). In contrast, 7 days post-transduction, cells exposed to nontargeted AAVP-Ucp1 vector displayed the anticipated phenotype of lipid-accumulating adipocytes (Fig. 6G). These data indicate that internalization of the SPARC-mimicking peptides by the receptor allows for specific and efficient intracellular uptake of the hybrid viral particles. Our results therefore show that α5β1 integrin, like other integrins [44], could serve as a selective target of peptide-directed interventions.

SPARC-α5β1 Integrin Interaction and ECM Attachment of ASC

Finally, on the basis of our results indicating SPARC binding to α5β1 integrin at sites of focal adhesions, we hypothesized that the corresponding function of SPARC is to disrupt integrin-mediated cell attachment to fibronectin in the ECM. To test this possibility, we assayed the capacity of SPARC to interfere with ASC attachment by measuring cell motility in the presence of rhSPARC, hPep, or a control peptide (Fig. 7A; supplemental online Fig. 2B). By quantifying transmigration of ASC through 5-µm pores, we detected a concentration-dependent elevation in motility of patient-derived ASC. According to analysis with Student’s t test adjusted for multiple comparison by Benjamini-Hochberg method, a 4 µg/ml concentration of rhSPARC (p = .015) or hPep (p = .038) was sufficient to induce a statistically significant increase in cell migration, compared with 4 µg/ml control peptide (p = .235). On the basis of our collective results, we propose a model for the function of SPARC in WAT (Fig. 7B).

DISCUSSION

Stromal organ compartments are a convenient source of progenitor cells for mesenchymal tissues [2]. Much of the information on the biological properties of stromal cells is available from studies on bone marrow-derived MSC [1–3]. It has been shown that MSC can be mobilized from their niches and migrate to other tissues in response to inflammatory signals [2, 48]. The exact cell surface interactions that lead to mobilization of MSC from their niches, as well as chemokine gradients that result in their homing to secondary organs, are poorly understood.

In this study, we reasoned that a combinatorial approach could identify ligands of functional ASC receptors. Indeed, we show that certain peptides selected on the basis of their binding to ASC resembled previously characterized sequences within the SPARC protein. Peptides CMLAGWIPC (hPep) and CWLGEWLGC (mPep) were found to mimic the region of SPARC implicated in the inhibition of cell adhesion [36]. Our protein binding and cell internalization studies indicate that α5β1 integrin is the receptor of SPARC, hPep, and mPep. We show that this ligand-receptor system is conserved on both primary human ASC and bone marrow MSC with little or no donor-specific variation (Fig. 1; supplemental online Fig. 2), as well as on mouse ASC and 3T3-L1 cells (Fig. 1). This provided the well-characterized 3T3-L1 cell line as a convenient model to study the α5β1 complex as a potential target for intracellular delivery of bioactive peptides and viruses. Our immunolocalization experiments and cell motility assays show that SPARC-α5β1 binding occurs at focal adhesions and leads to cell detachment. This function of SPARC and the SPARC-mimicking peptides apparently accounts for the difference between the immunofluorescence patterns for β1/paxillin and for SPARC upon addition of SPARC to ASC. Whereas β1 and paxillin are immunolocalized on the intact focal adhesions that largely colocalize with endogenous SPARC expressed at low levels (Fig. 4E–4H), exogenous SPARC at a higher concentration detaches cells, thus leading to the pseudopodia contraction and apparent β1 membrane clustering reported for this and other integrins [27, 28].

Although intracellular proteins that bind to SPARC have been identified, the cell surface receptor for extracellular SPARC has remained elusive [26]. Stabilin-1 [49] and integrin-linked kinase [41] have been identified as proteins mediating SPARC-cell surface signaling. However, the highly restricted expression of the former and the intracellular localization of the latter have indicated that another, cell surface-exposed complex must be involved. The known function of SPARC in focal adhesion disassembly and cell detachment [24, 26, 39] has pointed to focal adhesions, which are known to depend on integrins, as locations of the SPARC cell surface target. The correlation of SPARC localization with integrin-linked kinase activity [41] has supported the possibility that SPARC might bind to integrins. Moreover, the dependence of an integrin β1/talin/paxillin association on SPARC within focal adhesions of lens epithelial cells has been shown [40]. Although the α5β1 complex had not been directly isolated as a SPARC receptor, coimmunoprecipitation experiments indicate complex formation between SPARC and β1 (M. Weaver, G. Workman, E.H. Sage, manuscript submitted for publication). Recently, SPARC was shown to inhibit integrin β1-mediated adhesion and growth factor-dependent survival signaling of ovarian cancer cells, and the possibility of direct binding of SPARC to integrins has been proposed [50]. Therefore, our results are consistent with published reports and for the first time directly identify the β1 integrin, in complex with α5 integrin, intercellular talin-1, and actin, as a target of SPARC on the ASC surface. Our data indicate that SPARC-mediated disruption of the interaction between α5β1 integrin and its established ECM ligand, fibronectin [27], at least partially accounts for the cell detachment function of SPARC.

Although increased expression of α5 and β1 on undifferentiated 3T3-L1 preadipocytes and ASC has been reported previously [6, 51], these integrins are also upregulated in the stroma of organs other than WAT [30]. Indeed, the β1 integrin subunit (CD29) has been acknowledged as a marker useful for prospective bone marrow MSC isolation, with high interspecies reproducibility compared with other antigens [30, 31]. Our results are consistent with previous reports on selective upregulation of the α5β1 complex in nondifferentiated MSC and 3T3-L1 preadipocytes [29, 51]. The β1 integrin subunit has been implicated as a key molecule in stem cell-extracellular matrix interaction [6, 31]. The expression and function of β1 in MSC migration and engraftment have been reported [52]. Consistent with our data, it has been shown that the physiological role of this integrin is mediated through β1/talin-1 signaling on actin-mediated focal adhesions [27, 28].

On the basis of the established role of SPARC in WAT physiology [24, 25, 39] and our preliminary data indicating that SPARC suppresses adipogenesis, in part through regulation of ASC adhesion (J. Nie and E.H. Sage, unpublished results), we propose that SPARC regulates adipocyte differentiation in vivo. Previous comparison of adipocytes and vascular stroma demonstrated SPARC expression in vivo to be mainly associated with the adipocyte fraction [22], consistent with SPARC upregulation in differentiated 3T3-L1 adipocytes [23]. According to our model (Fig. 7B), extracellular SPARC secreted by adipocytes represses WAT growth by its interference with ASC adhesion to the extracellular matrix. If so, this functional interaction creates an autoregulatory loop that maintains the balance between ASC proliferation and differentiation into adipocytes. This working model is consistent with the α5 integrin downregulation in adipocytes [51] and low expression of SPARC in ASC (in comparison with that of adipocytes), as well as the increased adiposity characteristic of the SPARC-null mice [22, 23, 25]. It is currently unclear whether residual SPARC expression by ASC also plays a role in tissue homeostasis.

In addition to local tissue effects, SPARC could be involved in systemic mobilization of cells, a role consistent with its function in tumor cell metastasis [26]. Integrin complexes have been revealed as key cell-matrix attachment anchors, signaling through which can mobilize stem cells [31, 53]. The role of the focal adhesion complex involving integrin β1 and paxillin in leukocyte trafficking has been demonstrated [42, 43]. Furthermore, binding of SPARC to vascular cell adhesion molecule-1 (VCAM-1) causes the cytoskeletal rearrangement that is important for leukocyte transmigration [54]. Our experiments suggest that SPARC could execute a similar effect on ASC anchored by integrins in the extracellular matrix. Tantalizingly, SPARC has actually been implicated recently in the biology of stem cells, in agreement with our observations [55, 56]. Patient-derived human ASC promote angiogenesis by secretion of growth factors concomitant with their engraftment into skeletal muscle [7], and it remains to be tested whether ASC migration is mediated by SPARC in vivo. Reports on transient expression of SPARC as a regulator of fibroblastic cell migration in heart disease [23, 57] also support our hypothesis that SPARC-integrin interactions mediate MSC mobilization and subsequent homing. The relative importance of ASC, versus MSC from bone marrow and other tissues, as a dynamic source of mesenchymal progenitor cells in disease and development remains to be determined in future experimental work.

CONCLUSION

Our study reinforces combinatorial peptide libraries as an effective resource for isolation of cell-targeting ligands [9, 10, 13, 16, 17, 19–21] and adds SPARC to the list of α5β1 integrin ligands [27]. It also has several novel translational implications. Perhaps most significantly, we hypothesize that the SPARC-mimicking peptides identified here could be useful as potential stromal cell mobilization agents. We also demonstrate that an α5β1 integrin-dependent internalizing receptor system can be used to direct cytotoxic compounds and transgenes into ASC. Because the SPARC protein is highly conserved, the β1 integrin-targeting peptides described here are equally functional as delivery vehicles for both mouse and human ASC.