Microtubules regulate GEF-H1 in response to extracellular matrix stiffness

Endogenous GEF-H1 activity is regulated by extracellular matrix stiffness

Activated Rho GEFs have a significantly higher affinity for the nucleotide-free state of Rho than for either GTP- or GDP-bound Rho (Garcia-Mata et al., 2006). This aspect can be exploited to isolate activated Rho GEFs, which can be pulled down using glutathione S-transferase (GST)–RhoA (G17A), a point mutation that mimics the Rho nucleotide-free state and was previously used to bind activated Rho GEFs (Garcia-Mata et al., 2006; Dubash et al., 2007).

To determine whether GEF-H1 exchange activity for RhoA is regulated by 3D matrix stiffness, we cultured T47D and normal murine mammary gland (NMuMG) cells in compliant and stiff collagen gels for 2 h. We previously demonstrated that increasing the concentration of collagen in 3D gels increases the elastic modulus, creating gels that are stiffer (Provenzano et al., 2009). We found that different cells experience elastic modulus differently, and thus the concentration that is “stiff” or “compliant” needs to be determined empirically. Thus, for these studies, we defined a stiff matrix (high density) experimentally for each individual cell line as the concentration of collagen that cannot be contracted >10%, whereas a compliant matrix (low density) is the concentration of collagen that a particular cell line can contract at least 40%. For the highly contractile NMuMG cells, we used 2 mg/ml collagen as compliant and 4 mg/ml collagen as stiff, whereas for the less contractile T47D human breast carcinoma cell line, we used 1.3 mg/ml collagen as compliant and 2 mg/ml collagen as stiff. Using the GST-Rho(17A) pull-down assay, we determined that the amount of activated GEF-H1 was significantly increased in stiff compared with compliant collagen gels in both T47D and NMuMG cells (Figure 1, A and B), demonstrating that GEF-H1 exchange activity for RhoA is regulated by 3D collagen matrix stiffness.

To determine whether GEF-H1 regulates RhoA downstream of matrix stiffness, stable and transient GEF-H1 knockdown cells were generated. NMuMGs were infected with vector control–pGIPz or GEF-H1 short hairpin RNA (shRNA) #1 or stably transfected with vector control-pLK0.1 or GEF-H1 shRNA #2. In addition, T47D cells were stably transfected with anti–human GEF-H1 shRNA or control nontargeting shRNA. Knockdown was determined by Western blot (Figure 2A and Supplemental Figure S1). Although GEF-H1 is implicated in mitosis (Krendel et al., 2002; Bakal et al., 2005; Birkenfeld et al., 2007, 2008), there was no observable mitotic effect (unpublished data), nor was there any difference in cell proliferation on cells with stable GEF-H1 knockdown when cultured in collagen gels (Supplemental Figure S2). However, the stable cell lines did retain some GEF-H1 expression, which may allow them to continue through mitosis.

Consistent with previous data (Wozniak et al., 2003), in shRNA control cells, the amount of active, GTP-bound RhoA was increased in stiff compared with compliant collagen gels (Figure 2B and Supplemental Figure S1). However, in cells in which GEF-H1 was knocked down, this typical increase in RhoA activity in response to increased matrix stiffness was lost (Figure 2B and Supplemental Figure S1). Of interest, with NMuMG cells that were stably transfected, we noted a slight increase in RhoA activity under compliant matrix conditions when GEF-H1 was knocked down compared with control cells. This suggests that there may be additional regulatory mechanisms that are opposed by GEF-H1 in the absence of a stimulus, a finding that is consistent with results in which GEF-H1 knockdown affects the resting state of RhoA activity in studies of growth factor stimulation (Tsapara et al., 2010). Moreover, GEF-H1–knockdown cells contracted a compliant 3D collagen matrix more than twofold less compared with shRNA control cells (Figure 2, C and D). The inability of cells expressing GEF-H1 shRNA to contract the matrix is interesting, as RhoA activity was slightly elevated in compliant matrices upon GEF-H1 knockdown. However, these data are consistent with the results of Tsapara et al. (2010) in which expression of a dominant-negative GEF-H1 significantly reduced contractility on compliant 2D surfaces.

As a converse approach, green fluorescent protein (GFP)–tagged GEF-H1 was overexpressed in NMuMG cells (Figure 2E). Overexpression of GEF-H1 resulted in an overall increase in the amount of active, GTP-bound RhoA in both compliant and stiff matrices (Figure 2F). Furthermore, GFP-GEF-H1–overexpressing cells contracted a compliant 3D collagen matrix significantly more than GFP control cells. To validate that the effect of increased GEF-H1 was through Rho, cells were treated with cell-permeable exoenzyme C3, a nonreversible inhibitor of Rho. Collagen gel contractility in control GFP and GFP-GEF-H1–overexpressing cells was inhibited by treatment with exoenzyme C3 (Figure 2G). Combined, these data demonstrate that GEF-H1 regulates matrix contractility through regulation of RhoA.

Microtubule stability regulates GEF-H1 in response to 3D matrix stiffness

It has been demonstrated that GEF-H1 is regulated by microtubules, such that it is held inactive by binding to polymerized microtubules and is released and is free to regulate Rho under conditions of microtubule instability and subsequent depolymerization (Birkenfeld et al., 2008). To explore the possibility that microtubule stability is responsive to changes in 3D matrix stiffness as a potential mechanism for GEF-H1 activation, we cultured NMuMG cells in compliant and stiff collagen gels and treated them with vehicle control, paclitaxel to stabilize microtubules, or nocodazole to depolymerize microtubules. Changes in microtubules and actin organization, as well as endogenous GEF-H1 localization (unpublished data), in response to matrix stiffness were visualized by immunofluorescence. Using acetylated tubulin as a marker for microtubule stability, we observed in both compliant and stiff collagen gels that paclitaxel increased the amount of acetylated microtubules compared with vehicle control, whereas nocodazole decreased the presence of acetylated microtubules (Figure 3A). These treatments thus served as benchmarks for increased and decreased microtubule stability. Of importance, there was a striking difference in acetylated tubulin levels in NMuMG cells that were cultured in stiff compared with compliant collagen gels: in compliant 3D matrix conditions, the acetylated microtubules appeared more linear and organized, whereas in stiff matrices acetylated microtubules were more punctate and disorganized (Figure 3, A and B). Quantification of the structural differences in acetylated microtubules of cells cultured in stiff and compliant 3D collagen gels was done using a curvelet-based analysis program, CurveAlign (www.loci.wisc.edu/software/curvelet-based-alignment-analysis), which outputs a normalized measure of anisotropy based on the distribution of curvelet coefficients (Candès et al., 2006) generated from each image. This analysis demonstrated that the punctate and disorganized structures found in the cells grown in stiff collagen gels produced a more isotropic distribution of curvelet coefficients, whereas the linear and organized structures found in cells cultured in compliant collagen gels resulted in a distribution with a more distinct directionality or anisotropy (Figure 3B). Western blot analysis of acetylated- tubulin levels confirmed an increase in acetylated tubulin in compliant compared with stiff collagen gels (Figure 3C). This suggests that the changes in microtubule stability may be at least part of the mechanism for GEF-H1 regulation in response to changes in 3D matrix stiffness. Of interest, there was no obvious change in the overall structure of cortical actin organization between stiff and compliant 3D collagen matrices with or without any treatment. However, this assessment may have missed subtler differences and did not assess the actin in protrusive structures, which have been shown to differ between stiff and compliant collagen matrices (Wozniak et al., 2003).

When cells cultured in compliant and stiff 3D collagen gels were treated with nocodazole, which induces microtubule depolymerization, the regulated activation of GEF-H1 in response to matrix stiffness was lost, and there was a significant increase in GEF-H1 activity in compliant matrices, nearly matching GEF-H1 activation in stiff 3D collagen gels (Figure 3D). The fact that GEF-H1 activity did not further increase in stiff matrices in response to nocodazole treatment may indicate that the activation of GEF-H1 is already maximal under these conditions, perhaps because of the minimal acetylated microtubules that are present when cells are in stiff matrices (Figure 3A). In contrast, when cells cultured in compliant and stiff matrices were treated with paclitaxel to inhibit microtubule disassembly and thus enhance microtubule stability, there was significantly less GEF-H1 activation in stiff 3D collagen gels compared with control treatment (Figure 3D). These data suggest that microtubule disassembly and dynamic instability are a critical component of GEF-H1 activation and its regulation in response to matrix stiffness.

Extracellular signal-regulated kinase 1/2 is not required for GEF-H1 activation in response to 3D matrix stiffness

In addition to microtubules, GEF-H1 activity can be regulated downstream of extracellular signal-regulated kinase (ERK) 1/2, which directly phosphorylates GEF-H1, increasing its exchange activity for RhoA in 2D culture (Fujishiro et al., 2008). NMuMG cells were cultured in compliant and stiff collagen gels treated with vehicle control or the mitogen-activated protein kinase (MAPK) kinase (MEK) inhibitor UO126 to disrupt downstream signaling through ERK1/2, and then GEF-H1 activation was measured in each condition. GEF-H1 activation was increased in stiff compared with compliant 3D collagen gels after control treatment, and this was unchanged when UO126 was used to block MEK/ERK signaling (Figure 4A). In addition, there was no change in the activation of RhoA itself with increased matrix stiffness when comparing control treatment and treatment with MEK inhibitor (Figure 4B). These data were unexpected because we had previously linked matrix stiffness to ERK activation (Provenzano et al., 2009), but they demonstrate that ERK1/2 regulation of GEF-H1 is not required for GEF-H1 activation or downstream RhoA activity in response to 3D matrix stiffness in epithelial cells.

GEF-H1 is required for 3D cell migration and invasion in vitro

GEF-H1 is increased in migratory cells and has been shown to regulate 2D cell migration (Nalbant et al., 2009; Heasman et al., 2010; Tsapara et al., 2010). NMuMG cells expressing control and GEF-H1 shRNA were cultured in compliant and stiff collagen gels in a Transwell invasion assay. Knockdown of GEF-H1 expression disrupted the ability of cells to migrate through and invade the collagen gel in compliant gels compared with vector control cells (Figure 5A). We found that both control and GEF-H1 knockdown cells migrated significantly less in stiff compared with compliant collagen matrices, which may be due to an increase in the physical resistance encountered by the cell or an increase in the randomness of the migration at the expense of crossing the Transwell. We validated that this difference was not due to an effect on cell proliferation between NMuMG shRNA control and GEF-H1 knockdown cells when cultured in compliant or stiff 3D matrices (Supplemental Figure S2). To determine whether GEF-H1 plays a role in a carcinoma cell, we stably transfected the more invasive MDA-MB-231 human breast carcinoma cell line with either shRNA control or GEF-H1 specific shRNA #5. The knockdown determined by Western blot was comparable to T47D and NMuMG cells, and no effect on cell proliferation was noted (Supplemental Figure S2). Highly invasive MDA-MB-231 breast carcinoma cells also required GEF-H1 and RhoA for cell migration and invasion (Figure 5B). Again, invasion through a dense gel was less than through a compliant gel, but knockdown of GEF-H1 did significantly inhibit invasion of MDA-MB-231 cells through stiff gels.

Reagents and antibodies

Type I collagen was purchased from BD Biosciences (Bedford, MA), and all culture media were purchased from Life Technologies (Grand Island, NY). Nocodazole and paclitaxel (Sigma-Aldrich, St. Louis, MO) and the MEK inhibitor UO126 (Promega, Madison, WI) were dissolved in dimethyl sulfoxide (DMSO). Antibodies used were as follows: GEF-H1 (55B6) and RhoA (Cell Signaling, Danvers, MA), α-tubulin (clone B-5-1-2) and acetylated α-tubulin (Sigma-Aldrich), Rho (A-, B-, C-, clone 55; BD Biosciences, San Jose, CA), ACTIVE MAPK and ERK1/2 (Promega, Madison, WI), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH: Santa Cruz Biotechnology, Santa Cruz, CA). Horseradish peroxidase (HRP)–conjugated secondary antibodies were purchased from Jackson ImmunoResearch Laboratories (Westgrove, PA). Bisbenzimide and tetramethylrhodamine isothiocyanate (TRITC)–conjugated phalloidin was purchased from Sigma-Aldrich, and Alexa Fluor secondary antibodies were purchased from Invitrogen (Carlsbad, CA).

shRNA and DNA constructs

Lentiviral nontargeting pLKO.1 vector control, pGIPz vector control, and nine mouse (sequences in pLK0.1: GCAGGAGATTTACAACCGAAT, CGACGCTTTATACTTGAGCTT, CCCTCATTTGTCCTAC­ATGTA, TCAGTGACATTCACACACGTT, and sequences in pGIPz: CCGTGAGAACCTTGAAGATTAT, ACCTGTCCGTGGAATCCCTTATT, CTCCCTTATTGATGAAGG-TGTA, AAAGAAGCAAGATGTCATCTAT, ACAGGACATTCCTGAAGAGACA) and five human-specific (pLK0.1 sequences: GCGGCGAATTAAGATGGAGTT, CGATGCCCTGTACTTGAGTTT, CGCTCTGTCCATCGAAACTTT, CGTAGGCAATTAGAGATCGAA, CCCAACCTGCAATGTGACTAT) GEF-H1 shRNAs were purchased from Open Biosystems (Huntsville, AL). Mature sense sequences used to generate GEF-H1–knockdown cell lines were CCGTGAGAACCTTGAAGATTAT (mouse, pGIPz #1), GCAGGAGATTTACAACCGAAT (mouse, pLK0.1 #2), and CCCAACCTGCAATGTGACTAT (human pLK0.1 #5). Human GEF-H1 cDNA was purchased from Open Biosystems and was cloned into pBABE lentiviral GFP expression vector. The resulting construct was confirmed by sequence and expression analysis. GST-tagged Rho(G17A) in the bacterial expression vector pGEX-4T1 was a kind gift from Keith Burridge (University of North Carolina, Chapel Hill, NC).

Cell culture and transfection

NMuMG cells were maintained in high-glucose DMEM containing 0.2 U/ml bovine insulin and 10% fetal bovine serum (FBS). T47D human carcinoma cells were maintained in RPMI 1640 containing 0.2 U/ml bovine insulin and 10% FBS. MDA-MB-231 cells were maintained in low-glucose DMEM containing 10% FBS. All cell lines were incubated at 37°C with 5% CO₂. Cells were stably transfected using FuGENE 6 (Roche Bioscience, Indianapolis, IN) with species-specific GEF-H1 shRNA or pLKO.1 vector control. Phoenix (HEK293) viral packaging cells were transfected with pGIPz-GEF-H1 shRNA, GIPz alone, GFP-tagged GEF-H1, or GFP alone along with 5:1 ratio of pCMV 8.9.1 and VSV-G. At 24 h posttransfection, cells were moved to 30% serum for 48 h. Virus-containing media were collected, mixed with 1:1000 dilution of 1 μg/ml polybrene (Millipore, Billerica, MA), and used to infect target cells (RNAi Consortium, Broad Institute of MIT and Harvard, Cambridge, MA). For overexpressing cell lines, NMuMG cells were infected with lentiviral-packaged, GFP-tagged human GEF-H1 or GFP alone. At 48 h postinfection, cells were selected in 5 μg/ml puromycin, and expression was confirmed by Western blot. For knockdown, NMuMg cells were infected with lentiviral-packaged pGIPz GEF-H1 shRNA #2 or pGIPz alone. At 48–72 h postinfection, cells were seeded into collagen gels for RhoA activity measurements.

Cells were cultured in type I collagen gels as described previously (Wozniak and Keely, 2005). Briefly, gels containing cells were poured with serum-free media plus 5 μg/ml bovine serum albumin (BSA) to control for serum effects into six-well plates and allowed to polymerize for 1 h at 37°C. A 2-ml amount of complete or BSA media was added, and gels were detached from the sides and bottom of the dish (floating). The time at which the gels were rendered floating was considered day and time 0. For drug treatment experiments, gels were pretreated for 30 min with 1 ml of BSA media containing DMSO or 10 μM nocodazole, 10 μM paclitaxel, or 10 μM UO126 and then floated in an additional 1 ml of drug containing BSA media for 2–4 h.

Compliant (low-density) and stiff (high-density) matrix conditions were achieved by altering the concentration of type I collagen in the collagen gel. The concentration of collagen that was considered compliant or stiff was determined experimentally for each cell type based on the ability of the cells to undergo morphological changes and/or their ability to physically contract the collagen gel (Wozniak and Keely, 2005). For T47D cells, 1.3 mg/ml type I collagen was used for compliant and 2.0 mg/ml was used for stiff matrix conditions. For NMuMG and MDA-MB-231 cells, 2.0 mg/ml type I collagen was used for compliant and 4.0 mg/ml was used for stiff matrix conditions (Wozniak et al., 2003; Gehler et al., 2009).

Western blotting and immunofluorescence

Protein expression was assessed by Western blot. Briefly, cells were lysed in Laemmli buffer and boiled for 3 min, and proteins were separated by SDS–PAGE. Proteins were transferred to poly­vinylidene fluoride membrane, and membranes were blocked using 3% BSA plus 0.3% Tween-20 in TBS (TBST). Membranes were probed with 1:1000 anti–GEF-H1, 1:1000 anti-Rho, 1:5000 anti-GAPDH, 1:1000 anti–acetylated α-tubulin, 1:1000 anti–α-tubulin, 1:2500 active MAPK, or 1:2500 ERK1/2, followed by incubation with 1:5000 HRP-conjugated secondary antibodies. Proteins were visualized using ECL reagents (Pierce, Rockford, IL).

Localization of GEF-H1, as well as of microtubule and actin organization, in collagen matrices was determined by immunofluorescence similar to that previously described (Wozniak et al., 2003). Briefly, cells in gels were treated with DMSO, 10 μM nocodazole, or 10 μM paclitaxel as described earlier. Gels were fixed in 4% paraformaldehyde, permeabilized in 0.5% Triton X-100, and blocked in 3% BSA. GEF-H1 and acetylated α-tubulin primary antibodies were used at 1:500 and incubated overnight at 4°C. Alexa 488 and 594 secondary antibodies were used at 1:500 and incubated 1 h at room temperature. TRITC-conjugated phalloidin was used at 1:1000 and incubated 1 h at room temperature. Cells were counterstained with bisbenzamide at 1:15,000 for 5 min at room temperature. Fluorescence microscopy was carried out using an Olympus FluoView FV1000 confocal microscope (Olympus, Center Valley, PA) with a 60×/1.42 numerical aperture oil immersion objective. FIJI ImageJ (FIJI Consortium, National Institutes of Health, Bethesda, MD) was used for linear adjustments of brightness and contrast. Actin morphology was determined by observer assessment of randomized photographs for nine cells in each treatment group. Three independent observers assessed the structure and density of cortical actin fibers, and since there were no obvious differences, further quantification was not pursued. Cell proliferation was assessed in control and GEF-H1–knockdown cells by counting 4′,6-diamidino-2-phenylindole–stained nuclei in four fields of view in two gels per condition for each cell line. Data are expressed as an average number of nuclei in each cell line and collagen density.

Microtubule structural quantification

Quantitative analysis of microtubule perturbation was performed using CurveAlign, a specialized Matlab program (MathWorks, Natick, MA) based on the curvelet transform (Candès and Donoho, 1999; Candès et al., 2006). The curvelet transform is an overcomplete nonadaptive representation of an image using a series of energy measurements that range across scale, orientation, and position; the m-term approximation using a curvelet representation outperforms both the Fourier and wavelet representations and is nearly as good as the best adaptive techniques (Candès and Donoho, 1999). CurveAlign outputs a normalized measure of anisot­ropy based on the distribution of curvelet coefficients generated from each image and was used to quantify the structural differences between the microtubules of cells grown in stiff collagen gels and those of cells grown in compliant collagen gels. CurveAlign can be downloaded at http://loci.wisc.edu/software/curvelet-based-alignment-analysis.

Collagen matrix contraction measurements and morphogenesis

Gel contraction measurements were performed as described previously (Keely et al., 2007). Gels were poured and released in complete media as described. For Rho inhibition, the gels were released into media containing 50% glycerol (vehicle control) or 1 μg/ml cell-permeable exoenzyme C3 transferase. The time at which the gels were rendered floating was considered day 0. Gel diameter was measured every day for 10 d. For matrix contraction, day 10 gels were lysed in 2× RIPA buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 2 mM EDTA, 2 mM NaF, 2 mM NaVO₄, 2% NP-40, 0.5% sodium deoxycholate) as described previously for total protein concentration determined by amido black staining (Sheffield et al., 1987). Data were presented as total contraction (in millimeters) over time or as day 10 contraction normalized to total protein. For morphogenesis, gels that had been incubated for 10 d were fixed with 4% paraformaldehyde. Phase contrast microscopy was carried out using a Nikon TE300 inverted microscope (Nikon, Melville, NY) equipped with a CoolSNAP fx CCD camera (Tucson, AZ). Images were acquired using SlideBook, version 4.2 (Intelligent Imaging Innovations, Denver, CO), and processed using Adobe Photoshop CS2 (Adobe Systems, San Jose, CA).

RhoA activity assays

RhoA activity was measured using a Rho-GLISA kit (Cytoskeleton, Denver, CO) and normalized to total RhoA determined by Western blot analysis as described previously (Keely et al., 2007). Briefly, 1 × 10⁶ cells were cultured in gels floated in BSA media for 2 h. Gels were lysed on ice in 400 μl of lysis buffer provided. After centrifugation, 75 μl of supernatant was assayed in a plate format following the G-LISA protocol, and 60 μl of supernatant was loaded on an SDS–PAGE gel for Western blot analysis of total RhoA levels. Active RhoA detected by chemiluminescence was normalized to total RhoA determined by densitometry using Image J (National Institutes of Health, Bethesda, MD).

Affinity precipitation of activated GEFs

Active GEF-H1 was affinity precipitated from cell lysates using the Rho(G17A) mutant, which cannot bind nucleotide and therefore has high affinity for activated RhoA GEFs (Garcia-Mata et al., 2006). A total of 5 × 10⁶ cells were cultured in gels poured with BSA media and allowed to polymerize for 1 h. Gels were released and floated in BSA media for an additional 2 h and then lysed on ice in 2× RIPA buffer. After centrifugation, cell lysates were incubated with GST or GST-Rho(G17A) in the presence of glutathione-S Sepharose beads (GE Healthcare, Piscataway, NJ) for 2 h at 4°C. Activated GEF-H1 in the precipitates and total GEF-H1 in the cell lysates was detected by Western blot as described. Active and total GEF-H1 were quantified by densitometry as described, and the amount of active GEF-H1 was normalized to total GEF-H1 for each sample.

3D cell migration and invasion in vitro and in vivo tumorigenesis

To assess cell migration and invasion in 3D cultures, 2 × 10⁵ NMuMG or 3 × 10⁵ MDA-MB-231 cells were cultured in compliant and stiff collagen matrices poured with BSA media in a Transwell insert with 8- or 12-μm pore-size filter membrane, respectively (Corning, Corning, NY) and allowed to polymerize for 1 h. Once polymerized, 650 μl of full serum media or BSA media as a serum control was filled in the lower chamber, and cells were allowed 24 h to migrate and transverse the Transwell filter. Cell migration and invasion for each condition were determined according to manufacturer's recommendations. Briefly, gels and media were removed from the tops and bottoms of the insert and the wells were rinsed with phosphate-buffered saline. Cell dissociation solution (Trevigen, Helgerman, CT) containing 2 μg/ml of Calcein AM live-cell stain (Molecular Probes, Eugene, OR) was added to each well so that the bottom of the Transwell insert was contacting the liquid but not submerged and incubated at 37°C for 60 min with occasional gentle agitation. After incubation, the bottom of the Transwell filter was gently rinsed, and the number of migrated cells was determined by fluorescence with 485-nm excitation and 520-nm emission. Data are presented as the percentage of migrated cells for each condition.

Statistical analysis

All statistical analysis was completed using the program GraphPad Prism 4 (GraphPad Software, La Jolla, CA). For direct comparisons of two conditions, a Student's t test was preformed. For comparisons of multiple parameters an analysis of variance was completed, followed by a Bonferroni post hoc test. Error bars are represented as SEM.