Research ArticleCell Biology

Proteomic Analysis of Integrin-Associated Complexes Identifies RCC2 as a Dual Regulator of Rac1 and Arf6

See allHide authors and affiliations

Science Signaling  08 Sep 2009:
Vol. 2, Issue 87, pp. ra51
DOI: 10.1126/scisignal.2000396


The binding of integrin adhesion receptors to their extracellular matrix ligands controls cell morphology, movement, survival, and differentiation in various developmental, homeostatic, and disease processes. Here, we report a methodology to isolate complexes associated with integrin adhesion receptors, which, like other receptor-associated signaling complexes, have been refractory to proteomic analysis. Quantitative, comparative analyses of the proteomes of two receptor-ligand pairs, α4β1–vascular cell adhesion molecule–1 and α5β1–fibronectin, defined both core and receptor-specific components. Regulator of chromosome condensation–2 (RCC2) was detected in the α5β1–fibronectin signaling network at an intersection between the Rac1 and adenosine 5′-diphosphate ribosylation factor 6 (Arf6) subnetworks. RCC2 knockdown enhanced fibronectin-induced activation of both Rac1 and Arf6 and accelerated cell spreading, suggesting that RCC2 limits the signaling required for membrane protrusion and delivery. Dysregulation of Rac1 and Arf6 function by RCC2 knockdown also abolished persistent migration along fibronectin fibers, indicating a functional role for RCC2 in directional cell movement. This proteomics workflow now opens the way to further dissection and systems-level analyses of adhesion signaling.


Multicellular existence relies on the engagement of integrin cell surface receptors with the extracellular matrix (ECM) and counter-receptors on adjacent cells. Integrins are a family of 24 αβ heterodimers that result from selective noncovalent associations of 18 α and 8 β subunits (1, 2), which comprise large extracellular and typically small cytoplasmic domains. The particular combination of extracellular domains determines ligand-binding specificity, the structural basis of which enables the family to be subdivided into four main classes (3). Arg-Gly-Asp (RGD)–binding integrins, including all five αV integrins, α5β1, α8β1, and αIIbβ3, share the ability to recognize ligands such as fibronectin (FN), vitronectin, and fibrinogen, which contain the RGD tripeptide motif. Leu-Asp-Val (LDV)–binding integrins include α4β1, α4β7, α9β1, and the four members of the β2 subfamily and recognize active sites containing LDV-related sequences in ligands such as vascular cell adhesion molecule–1 (VCAM-1), mucosal addressin cell adhesion molecule–1 (MAdCAM-1), and intercellular cell adhesion molecule–1 (ICAM-1). A domain–containing integrins (α1β1, α2β1, α10β1, and α11β1) bind laminin and collagen, whereas the non–αA domain–containing laminin-binding integrins (α3β1, α6β1, α7β1, and α6β4) are highly selective laminin receptors.

Integrin-ligand binding controls a diverse range of cellular functions in both health and disease and is orchestrated by a combination of membrane-associated signaling complexes and mechanosensitive connections to the actin cytoskeleton (46). Approximately 150 components of integrin signaling complexes have been described, and on the basis of their reported pairwise interactions, a hypothetical integrin “adhesome” has been created (7). However, the networks that translate ligand binding into functional readouts, the extent of receptor-specific and spatiotemporal variation in composition, the stoichiometry of complex components, and the molecular mechanisms that coordinate different pathways within the signaling network are unknown. To date, systematic global analyses of the integrin signaling nexus have been hampered by the inherent lability and inaccessibility of plasma membrane–associated complexes. Here, we report the development of a quantitative proteomic methodology suitable for the analysis of transmembrane receptor-ligand complexes and the application of the methodology to two integrin-ligand pairs: α4β1–VCAM-1 and α5β1–FN; these structurally related integrins were selected because they exhibit distinct functional activities with respect to cell migration and exertion of cellular contractile forces (8, 9). We reasoned that their proteomes would contain both shared and receptor-specific interacting proteins.


A quantitative proteomic methodology for the analysis of integrin adhesion complexes

Integrin-associated complexes were isolated from α5+4+ K562 cells (K562 cells stably transfected with α4 integrin) by a ligand affinity purification approach (Fig. 1A). α5+4+ K562 cells display a similar degree of attachment to VCAM-1 compared to that mediated by endogenous α5β1 to FN and show similar activation-state modulation of α4β1 compared to that of the endogenous receptor (10). The restricted integrin profile of α5+4+ K562 cells, unlike most other cells, thus enabled heterodimer-specific adhesion and signaling events to be pinpointed.

Fig. 1

A quantitative proteomic pipeline for the analysis of integrin adhesion complexes. (A) Schematic workflow of the ligand affinity isolation strategy in conjunction with MS data acquisition, validation, and interpretation. LC-MS/MS, liquid chromatography–tandem mass spectrometry; PPI, protein-protein interaction. (B and C) The specific nature of the isolated integrin-ligand complexes was shown by Western blotting, as compared with control pull-down assays (α-TfR, transferrin receptor antibody; Control, non–integrin-binding VCAM-1 D40A mutant). α5β1–FN cytoskeletal complexes were isolated with the cell membrane–permeable cross-linker DTBP but not the cell membrane–impermeable cross-linker DTSSP (B), and α4β1–VCAM-1 cytoskeletal complexes were isolated with DTBP (C). Bead-coated proteins used for each affinity purification (AP) and proteins probed for by immunoblotting (IB) are indicated. Mouse IgG fragments are indicated by a white arrowhead.

Paramagnetic beads were coated with integrin ligands (FN for α5β1 or VCAM-1 for α4β1) and mixed with cells in suspension to induce the formation of adhesion complexes. After sonication and detergent extraction, bead-associated material was probed by Western blotting for the presence of integrin subunits and the adhesion complex components talin and paxillin. Consistent with previous attempts to identify integrin-associated proteins by immunoprecipitation, low amounts of talin and paxillin were recovered (Fig. 1B). However, the addition of the membrane-permeable cross-linker dimethyl-3,3′-dithiobispropionimidate (DTBP) before cell lysis increased the recovery of adhesion complex components compared to that of the controls, which were an antibody against transferrin receptor (TfR) or a VCAM-1 Asp40→Ala (D40A) mutant that does not bind to integrins (Fig. 1, B and C). In contrast, the non–membrane-permeable cross-linker 3,3′-dithiobis(sulfosuccinimidylpropionate) (DTSSP) increased the recovery of the primary receptor-ligand complex, but failed to retain talin and paxillin (Fig. 1B). Disruption of the actin cytoskeleton with cytochalasin D did not affect the recovery of talin and paxillin, showing that the integrin-associated complexes were not simply trapped in an extended actin network (fig. S1). Together, these data show that specific and efficient isolation of integrin-associated complexes is possible, but that it requires use of a membrane-permeable cross-linker before cell lysis.

Tandem mass spectrometry (MS) identified with ≥99% confidence 620, 418, and 319 proteins in FN, VCAM-1, and non–integrin-binding control samples, respectively. Of the characterized FN receptors (3), α5 was the predominant α integrin subunit detected in FN-induced complexes, whereas β3 integrin was not detected. In addition, α4β1 was the only integrin receptor detected in VCAM-1–induced complexes. These data show that the complexes isolated by FN and VCAM-1 are primarily recruited by α5β1 and α4β1 integrins, respectively.

Relative abundance of identified proteins was measured by spectral counting (11, 12). Hierarchical clustering was used to identify patterns of recruitment of proteins between samples and to aid the objective assimilation of multiple, large proteomic data sets (13) (Fig. 2 and table S1). Unbiased Pearson correlation identified five major groups of proteins on the basis of their relative abundance in the different samples, which corresponded to clusters of proteins enriched to VCAM-1 (correlation = 0.66), FN (correlation = 0.72), both FN and VCAM-1 (correlation = 0.71), the negative control (correlation = 0.64), or all three conditions (correlation = 0.65) (displayed from top to bottom, respectively, in Fig. 2A). We reasoned that the groups identified by hierarchical clustering represented proteins that were specifically enriched either in α5β1–FN or in α4β1–VCAM-1 complexes, were core components of both types of complex, or were nonspecifically recruited. As an important test of the unbiased clustering, the α4 and α5 integrin subunits were found to reside, as expected, in clusters enriched in VCAM-1 and FN samples, respectively (Fig. 2, B and C). Known adhesion complex proteins, including myosin IIA, vinculin, VASP (vasodilator-stimulated phosphoprotein), filamin, ACF7 (actin cross-linking family protein 7), IQGAP1 [IQ motif–containing guanosine triphosphatase (GTPase)–activating protein 1], Rap1b, and protein phosphatase 2, were found in a dual FN- and VCAM-1–enriched cluster, characterized by a higher abundance of the proteins in the FN samples compared to that in VCAM-1 samples (Fig. 2D). The shared β1 integrin subunit resided in a separate dual FN- and VCAM-1–enriched cluster (Fig. 2E), along with several well-characterized adhesion complex proteins, including kindlin-3, talin-1, zyxin, fascin, α-actinin–4, moesin, 14-3-3 protein β/α, myosin IIB, and actin-related protein Arp2/3 complex subunit 3 (Fig. 2E and tables S1 and S2). The proteins identified in the dual FN- and VCAM-1–enriched clusters can therefore be considered core components of the two complexes.

Fig. 2

Hierarchical clustering of proteins identified by MS analysis of FN, VCAM-1, and control affinity purifications. (A) Complete output of unsupervised hierarchical clustering analysis of identified proteins. Quantitative heat maps display mean spectral counts as a percentage of the total number of spectra identified in each analysis. Associated dendrograms display hierarchical clustering on the basis of uncentered Pearson correlation using complete linkage. Correlations at selected nodes are indicated. (B to E) Selected ligand-specific and core β1 integrin–enriched proteins are displayed, indicated by blue bars in (A).

Analysis of the clusters specifically enriched to either integrin ligand revealed a much larger number of proteins enriched in FN samples compared to that in VCAM-1, suggesting that FN has a greater agonistic potential to recruit the adhesion machinery than VCAM-1 (Fig. 2, B and C, and tables S1 and S2). For FN, this cluster again contained several well-studied adhesion complex proteins including CD98hc (cluster of differentiation 98 heavy chain), ILK (integrin-linked kinase), LASP1 [LIM (Lin-11, Isl-1, and Mec-3) and SH3 (Src homology 3) protein 1], disabled-2, Arp2/3 complex components, and 14-3-3 proteins, but it also contained many candidate adhesion components. The smaller VCAM-1 cluster contained proteins such as Arf1 (adenosine 5′-diphosphate ribosylation factor 1), SHP2 [SH2 (Src homology 2) domain–containing protein tyrosine phosphatase 2], and casein kinase 1 that were detected exclusively in VCAM-1 samples, as well as a group of proteins (including galectin-1, Rab5C, and Rac1) that were detected in both integrin ligands but enriched in VCAM-1 compared to FN samples. Intriguingly, some proteins that have been documented as components of mature integrin-containing adhesion complexes, such as tensin, FAK (focal adhesion kinase), p130Cas, and Src, were not identified in our analyses. A distinguishing feature of these proteins is that they are responsive to the application of force (2, 4, 1416), and their absence is consistent with the fact that bead-bound complexes were isolated without the specific application of tension. A similar pattern of recruitment was observed for small signaling GTPases. Rac1, which drives adhesion initiation, was detected, but RhoA, which is activated by force and drives the maturation of adhesion complexes (17), was not. Together, these findings suggest that the complex components listed in Fig. 2 and tables S1 and S2 represent early nascent adhesions. In the future, it will be important to adapt the workflow to isolate complexes from adherent cells or from cells in suspension to which force has been applied.

FN- and VCAM-1–induced integrin interaction networks

To interrogate the molecular organization of the isolated adhesion complexes, we used protein interaction network analysis. To convert the lists of integrin-associated complex components into interaction networks, a subtractive proteomic strategy was used. Only proteins enriched at least twofold over the amount in non–integrin-binding control samples (406 for FN and 185 for VCAM-1; table S2) were mapped onto an interactome constructed from a union of human protein interaction data sets (1820) and the literature-curated interactome of integrin adhesion complexes (7). The integrin-associated interaction networks were unlikely to be generated by chance as they were distinctly nonrandom, with significantly higher numbers of interactions and clustering coefficients than those observed for control simulations (table S3). For proteins that directly interact with β1 integrin (the 1-hop neighborhood), an increased number mapped onto the α5β1–FN network compared to the α4β1–VCAM-1 network (16 compared to 9 proteins), consistent with the conclusions drawn from Fig. 2. As a result, the 2-hop neighborhood of α5β1 was far more expansive than that of α4β1 (34 compared to 12 proteins; Fig. 3). That fewer α4β1–VCAM-1–specific proteins were identified is consistent with previous studies that suggest α4β1 forms weaker cytoskeletal interactions, resulting in diminished adhesion complex formation, and is therefore specialized for highly motile cell types (8, 21).

Fig. 3

Protein-protein interaction network models for α5β1–FN and α4β1–VCAM-1 complexes. (A and B) Proteins identified in α5β1–FN (A) and α4β1–VCAM-1 (B) complexes after a subtractive proteomics strategy were mapped onto a human interactome to generate interaction network models. Proteins identified within two path lengths (the 2-hop neighborhood) of β1 integrin are shown, with the 1- and 2-hop neighborhoods forming the inner and outer circles, respectively. Protein identifications (nodes) are colored according to their relative enrichment to α5β1–FN (red) or α4β1–VCAM-1 (blue) and are ordered in a clockwise fashion relating to their relative enrichment to α5β1–FN. Proteins identified in both α5β1–FN and α4β1–VCAM-1 data sets are indicated by a black node border; absence of a black border indicates unique identification in that integrin-ligand data set.

Additional normalization of quantitative MS data (to the β1 integrin spectral count) and coloring proteins (nodes) by fold enrichment to α5β1–FN (red) or α4β1–VCAM-1 (blue) allowed changes in relative abundance of protein recruitment between α5β1 and α4β1 to be assessed. The α5β1 network is predominantly composed of proteins specifically enriched to FN over VCAM-1, whereas the α4β1 network consists of proteins more evenly balanced in this respect (Fig. 3). In addition, core β1 integrin interactions identified by hierarchical clustering (Fig. 2) are common to both ligand-occupied α5β1 and α4β1, although the relative abundance of the core interactors within these adhesion complexes can differ. Furthermore, some well-characterized adhesion complex components, including talin, kindlin-3, α-actinin, and zyxin, are recruited to the same extent to α4β1–VCAM-1 and α5β1–FN (white, pale blue, or pale red nodes; Figs. 3 and 4). Thus, the reduced number of proteins detected in α4β1–VCAM-1 complexes is due to the lack of recruitment of specific proteins, rather than an overall reduction in the amount of the components or affinity differences between the integrins and their ligands. The differences in “flavors” of integrin complexes observed here reveal surprisingly diverse heterodimer-specific β1 integrin signaling events and provide a powerful tool for the future analysis of context-specific integrin adhesion signaling.

Fig. 4

Recruitment of RCC2 to FN-bound complexes modulates the activation of Rac1 and Arf6. (A) A focused view of the α5β1–FN network, highlighting the intersection of the β1 integrin, Rac1, and Arf6 1-hop neighborhoods. Protein identifications are colored as described in Fig. 3. CacyBP-SIP, calcyclin-binding protein–Siah-1–interacting protein; GART, trifunctional purine biosynthetic protein adenosine-3; GDI, GDP dissociation inhibitor; IMPDH2, inosine monophosphate dehydrogenase 2; PFK, phosphofructokinase-1; SF3b155, splicing factor 3b, subunit 1, 155 kD; SRP14, signal recognition particle 14 kD. (B) Verification of the specific recruitment of RCC2 to α5β1–FN complexes by Western blotting. (C to F) The activities of Rac1 (C and D) and Arf6 (E and F) were measured by effector pull-down assays in combination with quantitative Western blotting with fluorophore-conjugated antibodies. Activities were compared between control (Ctrl) and RCC2 knockdown (KD) cells in suspension (C and E) or during spreading on FN (D and F). Equivalent loading between experiments was confirmed by blotting crude lysates for mitochondrial Hsp70 (mtHsp70). Axes are in arbitrary units assigned according to the relative activity of cells in suspension. Each panel is representative of at least four separate experiments. Error bars indicate SEM; P values were calculated for direct comparisons of peak time points with a Student’s t test, as indicated.

Role of regulator of chromosome condensation–2 in FN-dependent adhesion complex formation, cell spreading, and migration

Examination of the integrin-ligand interaction networks revealed proteins situated at nodal points that connected identifiable signaling subnetworks. Among these proteins, regulator of chromosome condensation–2 (RCC2), a molecule that has not previously been linked to integrin signaling, occupied a key position between the β1 integrin, Rac1, and Arf6 1-hop neighborhoods of the α5β1–FN network (Fig. 4A). Rac1 was detected in both α5β1–FN and α4β1–VCAM-1 samples but enriched in α4β1–VCAM-1 (blue node with black border; Fig. 4A), whereas Arf6 was only detected in α5β1–FN samples (red node with no border; Fig. 4A). RCC2 was detected by MS in α5β1–FN, but not α4β1–VCAM-1 samples, a finding confirmed by Western blotting (Fig. 4B). FN-dependent Rac1 regulation controls membrane protrusion (22, 23), and Arf6 has been implicated in integrin-dependent membrane trafficking (24). Although little is known about RCC2, it has been reported to bind directly to the nucleotide-free form of Rac1 and speculated to act as a Rac1 guanine nucleotide exchange factor (GEF) (25). In addition, a large-scale affinity-based study mapping protein-protein interactions identified the association of RCC2 with Arf6 (26).

To test the contribution of RCC2 to FN-dependent Rac1 regulation (23, 27), RCC2 was targeted by RNA interference (RNAi). A reduction in RCC2 abundance to less than 20% had no effect on the basal Rac1 activity of cells in suspension (Fig. 4C), but, surprisingly, it augmented FN-dependent activation of Rac1 (Fig. 4D). Arf6 activation assays were also performed, with RCC2 knockdown having no effect on the basal activity of Arf6 (Fig. 4E). When suspended cells adhered onto a FN substrate, a wave of Arf6 activation was induced (Fig. 4F), consistent with a previous study (24) and similar to that observed for Rac1 (Fig. 4D). Moreover, RCC2 knockdown accelerated the FN-induced wave of Arf6 activation during cell spreading (Fig. 4F). These data suggest that, rather than acting as a GEF, RCC2 limits activation of both Rac1 and Arf6.

We hypothesized that altered Rac1 and Arf6 signaling would affect cell behavior, in particular the ability to form adhesion complexes, to spread, and to migrate. Therefore, the effect of RCC2 on adhesion complex formation during spreading on FN was assessed (Fig. 5, A and B, and fig. S2). The total area of adhesion complexes of wild-type mouse embryonic fibroblasts (MEFs) and cells in which RCC2 had been silenced was measured over time. Adhesion complex formation was accelerated after knockdown of RCC2, with adhesion area being significantly higher than that of control cells at 30 to 90 min (P < 0.05; Fig. 5, A and B). The period of augmented adhesion complex area coincided with the duration of enhanced Rac1 and Arf6 activities, which would be expected to augment both adhesion complex nucleation and trafficking of β1 integrin to the membrane.

Fig. 5

RCC2 modulates FN-dependent adhesion complex formation, cell spreading, and directional migration. (A and B) Adhesion complex formation during spreading of control and RCC2 knockdown (KD) MEF cells on FN. Cells were fixed at the indicated times and stained for vinculin to permit measurement of focal adhesion area with ImageJ software (A). NS, not significant. Forty-five cells were analyzed for each cell type at each time point. Data are representative of two independent experiments. Images shown are representative of cells that were allowed to spread for 60 min (B). (C and D) Spreading of cells transfected with control or RCC2 siRNAs. The percentage of MEF cells on FN (C) or B16-F10 cells on FN or VCAM-1 (D) counted as spread was plotted as a function of time. The shifts in the measure of half-maximal cell spreading (control minus RCC2 knockdown; mean ± SD) were as follows: MEF cells spread on FN, 3.21 ± 0.68 min (14.9 ± 2.5% acceleration compared to control); B16 cells spread on FN, 8.34 ± 2.44 min (14.1 ± 2.9% acceleration compared to control); B16 cells spread on VCAM-1, −0.40 ± 1.54 min (−0.5 ± 3.3% acceleration compared to control). (E to G) Control and RCC2 knockdown MEF cell migration on cell-derived matrix. Cells were allowed to spread on cell-derived matrix for 3 hours before imaging at 10-min intervals for 12 hours. Migration paths of 90 cells for each cell type were tracked with ImageJ software (E). The tracks of cells from three different fields of view from triplicate wells have been combined into each panel. Data are representative of three independent experiments. Persistence of migration (F) was determined by dividing the linear displacement of a cell by the total distance moved [accumulated distance (G)]. Gray blocks represent the experimentally determined threshold for the random migration of cells on FN. Error bars indicate SEM; P values were determined with a Z-test, as indicated.

To assess the functional outcome of altered GTPase signaling upon RCC2 depletion, the spreading of MEF cells on FN, and B16-F10 mouse melanoma cells on FN or VCAM-1, was examined. Consistent with the enhanced kinetics of Rac1 and Arf6 activation, depletion of RCC2 enhanced the rate of spreading of both cell types on FN (Fig. 5, C and D, and fig. S3). B16-F10 cell spreading on VCAM-1, however, was unaffected by RCC2 knockdown (Fig. 5D and fig. S3), consistent with the recruitment of RCC2 to α5β1–FN but not to α4β1–VCAM-1 complexes.

A key functional readout of altered GTPase signaling upon knockdown of RCC2 was obtained by examining the migration of MEF cells on preformed, fibrous cell-derived matrices. Cells plated on such matrices normally migrate with a high degree of directional persistence (23, 28). In comparison to control cells, RNAi directed against RCC2 reduced persistent migration by 50% (Fig. 5, E and F). RCC2 knockdown did not, however, affect total distance moved (Fig. 5G). Together, these data show that RCC2 is a functionally relevant component of the machinery required to maintain persistent migration in response to an ECM cue and, intriguingly, that this role is mediated by limiting the activity of Rac1 and Arf6.


Using an integrin-targeted proteomic strategy, we have identified RCC2 as a component of FN-induced signaling pathways that regulate directional cell migration. RCC2 has been previously characterized as a passenger protein, localizing to the centromere during specific stages of the cell cycle. RCC2 is required for alignment of chromosomes on the spindle (chromosome congression), centromeric targeting of the chromosome passenger complex (CPC), and activation of Aurora-B kinase, the catalytic subunit of the CPC (25, 29). The CPC functions to regulate chromosomal and cytoskeletal events during cell division, and CPC proteins have well-defined roles in cell cleavage (30). It is therefore likely that RCC2 is also essential for cell cleavage. RCC2 shares structural homology with other RCC1 family GEFs that act on Ran, Arf1, and Rab proteins, but it is unique among this family with respect to its ability to bind Rac1 (25, 31). Inactivation of Rac is essential for cytokinesis (32), and it is becoming increasingly apparent that integrins help determine the orientation of cytokinesis and subsequent positioning of daughter cells (33, 34). How integrin signaling and the activity of RCC2 are spatiotemporally regulated during cell adhesion, spreading, and mitosis remains to be resolved, and it is intriguing that RCC2 colocalizes at the midbody with talin, an integrin-binding cytoskeletal protein, during cytokinesis (35).

As well as identifying a large number of known adhesion-related components, our analyses have highlighted numerous putative integrin-associated proteins. For example, proteins with functions relating to protein translation were found to be specifically enriched to FN, VCAM-1, or both integrin ligands. In support of this finding, the translational machinery can reposition to sites of adhesion (36), and β1 integrin is required to stimulate FN-mediated translation in fibroblasts (37). Moreover, we detected LIM domain–containing proteins, such as the integrin-associated proteins zyxin and paxillin, that can bind nucleic acids and thus may serve a scaffolding role in translational processes (36). Furthermore, a quantitative proteomics approach has identified ribonucleoprotein complexes, likely to be involved in protein translation, which appear during early stages of cell spreading at locations close to sites of adhesion (38).

In summary, this first non–candidate-based analysis of integrin ligand–induced signaling has identified core and receptor-specific components of α5β1 and α4β1 integrin signaling complexes and pinpointed RCC2 as a coordinator of FN-dependent adhesion signaling pathways, a dual activity limiter of Rac1 and Arf6, and a regulator of adhesion complex formation, cell spreading, and directional migration. The abundance of active Rac1 governs the persistence of cell migration through the formation of peripheral lamellae (28). We speculate that RCC2 acts to restrict activation of Rac1 and Arf6 to a degree that enables effective membrane protrusion and integrin trafficking, thereby permitting cells to both maintain directional migration and respond to the topographic features of their environment. The development of this proteomics pipeline now allows the molecular composition of various adhesion complexes to be measured directly and opens the way to systems-level analyses of adhesion.

Materials and Methods


Human plasma FN was purchased from Sigma-Aldrich (Poole, UK). Fc-tagged recombinant soluble VCAM-1, comprising immunoglobulin (Ig) domains 1 and 2, and mutant VCAM-1(D40A), containing a D40A point mutation in the integrin-binding site and termed control throughout this study, were expressed from the pIG1 vector and purified as described (39).

Monoclonal antibodies used were mouse antibodies against human β1 integrin (JB1A; provided by J. A. Wilkins, University of Manitoba, Winnipeg, MB, Canada), against human paxillin (clone 349, BD Biosciences, Oxford, UK), against human vinculin (hVIN-1, Sigma-Aldrich), against human TfR (H68.4, Invitrogen, Paisley, UK, and OKT9, European Collection of Cell Cultures, Porton Down, UK), against Rac1 (BD Biosciences), against Arf6 (Sigma-Aldrich), and against human mitochondrial heat shock protein 70 kD (Hsp70) (JG1, ABR-Affinity BioReagents, Golden, CO). Polyclonal antibodies used were goat antibody against human α4 integrin (C-20, Santa Cruz Biotechnology, Santa Cruz, CA), rabbit antibody against human α5 integrin (H-104, Santa Cruz Biotechnology), goat antibody against human β1 integrin (N-20, Santa Cruz Biotechnology), goat antibody against human talin (C-20, Santa Cruz Biotechnology), and rabbit antibody against RCC2 (Bethyl Laboratories, Montgomery, TX). Other antibodies used were donkey antibody against goat IgG or against mouse IgG conjugated to Alexa Fluor 680 (Invitrogen) and donkey antibody against mouse IgG or against rabbit IgG conjugated to IRDye 800 (Rockland Immunochemicals, Gilbertsville, PA).

Cell culture

K562 human chronic myelogenous leukemia cells stably transfected with α4 integrin (α5+4+ K562 cells) were a gift from M. E. Hemler (Dana-Farber Cancer Institute, Boston, MA) and were cultured in RPMI 1640 medium supplemented with 10% (v/v) fetal calf serum (FCS; Lonza Bioscience, Wokingham, UK), 2 mM l-glutamine, and geneticin (1 mg/ml; Invitrogen). The generation of immortalized MEF cells has been described previously (23). To allow expression of the large T antigen, MEF cells were cultured at 33°C in Dulbecco’s modified Eagle’s medium (DMEM; Sigma-Aldrich) supplemented with 10% (v/v) FCS, 2 mM l-glutamine, and interferon-γ (20 U/ml; Sigma-Aldrich). B16-F10 mouse melanoma cells were cultured in DMEM supplemented with 10% (v/v) FCS and 2 mM l-glutamine. All cell lines were maintained at 37°C in a humidified 5% (v/v) CO2 atmosphere unless otherwise stated.

Ligand affinity purification assay

The isolation of plasma membrane fractions enriched for integrin membrane complexes was based on the protocol of Plopper and Ingber (40), modified here to permit the isolation of integrin-associated protein complexes for proteomic analysis. Ligand or antibody was coupled to tosyl-activated paramagnetic Dynabeads M-450 (4.5-μm diameter; Invitrogen) as described in the manufacturer’s protocol. For FN-coated beads, beads were incubated with 0.125 mM bis(sulfosuccinimidyl) suberate cross-linker (Thermo Fisher Scientific, Waltham, MA) for 15 min at 25°C to increase covalent bond formation between FN and beads. Coated beads were incubated with α5+4+ K562 cells in DMEM containing 25 mM Hepes, glucose (4.5 mg/ml), and l-glutamine (DMEM-Hepes) (Invitrogen) supplemented with 0.2% (w/v) bovine serum albumin and 0.4 mM MnCl2 at 70 rpm for 30 min at 37°C. DTBP or DTSSP cross-linker (Thermo Fisher Scientific) in DMEM-Hepes was added to bead-bound cells to a final concentration of 3 mM, where applicable, and samples were incubated for a further 30 min at 37°C (70 rpm). Cross-linker was quenched with 20 mM tris-HCl before isolation of bead-bound cells on a magnet. Samples were washed with CSK buffer [10 mM piperazine-N,N′-bis(2-ethanesulfonic acid) (pH 6.8), 50 mM NaCl, 150 mM sucrose, 3 mM MgCl2, and 1 mM MnCl2] supplemented with 20 mM tris-HCl (pH 8.5) and 2 mM Na3VO4. Cells were lysed in CSK+ buffer [CSK buffer supplemented with 0.5% (w/v) Triton X-100, leupeptin (10 μg/ml), aprotinin (10 μg/ml), 0.5 mM 4-(2-aminoethyl)benzenesulfonylfluoride hydrochloride (AEBSF), and 2 mM Na3VO4] for 30 min on ice with sonication (VibraCell VCX 500, Sonics & Materials, Newtown, CT). After cell lysis, beads were washed with CSK+ buffer, and affinity-purified proteins were eluted from beads with reducing sample buffer [50 mM tris-HCl (pH 6.8), 10% (w/v) glycerol, 4% (w/v) SDS, 0.004% (w/v) bromophenol blue, and 8% (v/v) β-mercaptoethanol]. Protein samples were separated from beads with a magnet, resolved by SDS–polyacrylamide gel electrophoresis (PAGE), and analyzed by Coomassie staining or Western blotting (see Supplementary Materials).

Mass spectrometry and data analysis

Details of the MS and data analysis are provided in the Supplementary Materials.

RNAi of RCC2

Small interfering RNA (siRNA) duplexes targeting mouse RCC2, sequence (sense) CCAACGUGGUGGUUCGAGA (with ON TARGET modification for enhanced specificity), and nontargeting control duplex were purchased from Dharmacon (Thermo Fisher Scientific). Oligo (0.8 nmol) was transfected into a 90% confluent 75-cm2 flask of MEF or B16-F10 mouse melanoma cells with Lipofectamine 2000 reagent (Invitrogen). After 24 hours, the cells were passaged and retransfected to achieve a reduction in protein abundance to below 20%. Cells were passaged and then cultured for a further 24 hours before use. For each experiment, RCC2 abundance was tested by Western blotting.

GTPase activity, adhesion area measurement, cell spreading, and cell migration assays

Details of the Rac1 and Arf6 activity assays, adhesion area quantification, and cell spreading and migration assays are provided in the Supplementary Materials.


We thank S. Warwood (University of Manchester) for MS data acquisition, A. G. Garrow (Unilever, Sharnbrook, UK) for providing the human protein interaction data sets, J. N. Selley (University of Manchester) for bioinformatic support, E.-J. Keevil (University of Manchester) for assistance with MS sample preparation, J. Kott (University of Manchester) for assistance with microscopy, J. A. Askari (University of Manchester) for discussions, and A. D. Sharrocks, C. Ballestrem, C. H. Streuli, R. A. North, and N. Papalopulu (all University of Manchester) for critical reading of the manuscript. This work was supported by grants 045225 and 074941 from the Wellcome Trust (to M.J.H.) and by a Biotechnology and Biological Sciences Research Council (BBSRC) Collaborative Awards in Science and Engineering Ph.D. studentship, sponsored by GlaxoSmithKline (A.B.). The Biomolecular Analysis Facility mass spectrometer and Bioimaging Facility microscopes used in this study were purchased with grants from the BBSRC, Wellcome Trust, and the University of Manchester Strategic Fund.

Supplementary Materials


Fig. S1. Disruption of the actin cytoskeleton does not affect recruitment of talin and paxillin to isolated α5β1–FN complexes.

Fig. S2. Knockdown of RCC2 accelerates adhesion formation.

Fig. S3. Knockdown of RCC2 augments FN-mediated cell spreading.

Table S1. Hierarchical clustering analysis of proteins identified in FN, VCAM-1, and VCAM-1(D40A) affinity purifications (full data sets).

Table S2. Specifically enriched proteins identified in FN and VCAM-1 affinity purifications.

Table S3. Protein-protein interaction network statistics.


File 1. Table S1 as an Excel file.

File 2. Table S2 as an Excel file.

File 3. Interaction network model for the α5β1–FN complex (XGMML format).

File 4. Interaction network model for the α4β1–VCAM-1 complex (XGMML format).

References and Notes

View Abstract

Stay Connected to Science Signaling

Navigate This Article