Role of the Focal Adhesion Protein Kindlin-1 in Breast Cancer Growth and Lung Metastasis

Metastasis, which is responsible for most cancer mortality, occurs through a multistep process, including invasion of tumor cells into the adjacent tissues, intravasation, survival in the circulation, extravasation from blood vessels, initiation and maintenance of micrometastases at distant sites, and finally, growth of cancer cells to produce secondary tumors (1–5). At each step, metastatic cancer cells face multiple obstacles that are overcome with molecular alterations that modify the expression and function of specific metastasis-related genes (1–3,6,7). Since 2000, the application of genomic profiling methods to the analysis of human breast tumors has generated expression profiles that are predictive of metastasis (8–10). However, although such analyses are very powerful for identifying prognostic markers, it has been difficult to definitively attribute the development of metastasis to the specific contributions of these genes because of the lack of experimental verification. Furthermore, comparative expression profiling of cancer cells with different metastatic potentials in animal models has led to the identification of metastasis promoting genes (8,11–14), although much work remains to be done to validate their clinical relevance. Thus, the characterization of both clinically relevant and functionally important metastasis genes is a crucial step toward the identification of potential therapeutic targets for prevention and treatment. In previous transcriptomic studies of human metastatic samples (15,16), we identified Fermitin family member 1 (FERMT1), also called Kindlin-1, in a six-gene signature that discriminates primary breast tumors with higher propensity to metastasize to the lungs.

Metastasis, which is responsible for most cancer mortality, occurs through a multistep process, including invasion of tumor cells into the adjacent tissues, intravasation, survival in the circulation, extravasation from blood vessels, initiation and maintenance of micrometastases at distant sites, and finally, growth of cancer cells to produce secondary tumors (1)(2)(3)(4)(5).At each step, metastatic cancer cells face multiple obstacles that are overcome with molecular alterations that modify the expression and function of specific metastasis-related genes (1)(2)(3)6,7).
Since 2000, the application of genomic profiling methods to the analysis of human breast tumors has generated expression profiles that are predictive of metastasis (8)(9)(10).However, although such analyses are very powerful for identifying prognostic markers, it has been difficult to definitively attribute the development of metastasis to the specific contributions of these genes because of the lack of experimental verification.Furthermore, comparative expression profiling of cancer cells with different metastatic potentials in animal models has led to the identification of metastasis promoting genes (8,(11)(12)(13)(14), although much work remains to be done to validate their clinical relevance.Thus, the characterization of both clinically relevant and functionally important metastasis genes is a crucial step toward the identification of potential therapeutic targets for prevention and treatment.
In previous transcriptomic studies of human metastatic samples (15,16), we identified Fermitin family member 1 (FERMT1), also called Kindlin-1, in a six-gene signature that discriminates primary breast tumors with higher propensity to metastasize to the lungs.Although this signature was suggested to be a surrogate of the basal-like subtype of breast cancers, we further demonstrated that it predicts lung metastases independent of the molecular subtypes of breast tumors (16,17).
Most kindlin-related studies focus on kindlin relevance in skin disease.To date, the biological implications of kindlin-1 have not been shown in cancer.Here, we conducted a study to investigate the role of kindlin-1 at the clinical, cellular, and molecular levels.First, we investigated whether Kindlin-1 is associated with lung metastasis in human breast cancer and poor prognosis in lung cancer.We then addressed the role of kindlin-1 in the aggressive phenotypes of tumor cells and assessed its involvement in breast cancer progression using an orthotopic mouse model.

Patients and Cell Lines
Patients.Lung, liver, brain, and bone metastases from breast cancer patients were obtained from the Curie Institute/René Huguenin Hospital (Saint-Cloud, France), the University of L'Aquila (L'Aquila, Italy), and IDIBELL (Barcelona, Spain).In addition, the microarray expression profiles of breast cancer metastases from an independent cohort of patients were analyzed (27).
A total of 516 primary tumors were obtained from the Curie Institute/René Huguenin Hospital.Of those, 502 [66 grade 1, 261 grade 2, and 175 grade 3 tumors, as scored according to Bloom and Richardson histoprognostic grading (28); see Supplementary Table 1, available online] were selected to encompass the various stages of breast cancer progression because grading information was available on those patients.The mean age of the patients was 60.7 years (range = 25-91 years), and the median follow-up was 120.5 months (range = 13-347 months).Seventy-six patients developed lung metastases.
Paraffin-embedded sections of matching breast tumors and metastases (n = 22) were obtained from the Curie Institute/René Huguenin Hospital and the University of Liège (Belgium).Adjacent normal host tissues were also observed in the same sections.
Written informed consent was obtained from each patient according to the recommendations of the local ethics committee.This study was approved by the review boards and ethics committees of the respective institutions.

Prior knowledge
Kindlin-1 encodes one of the three kindlin proteins implicated in several human genetic disorders, including Kindler syndrome, a genetic skin pathology characterized by blistering, atrophy, and photosensitivity.Kindlin-1 may also play a role in several cancers, but the nature of that role is unknown.

Study design
The effects of kindlin-1 overexpression and silencing on cell signaling and metastasis-like functions were assessed in human breast cancer cell lines and in breast tumor growth and lung metastasis in mice.The association between kindlin-1 expression and metastasis-free survival was analyzed in tumor tissue from 516 breast cancer patients.

Contribution
Kindlin-1 expression in human breast tumors was statistically significantly associated with lung metastasis and lung metastasis-free survival.Kindlin-1-expressing cells displayed characteristics that are hallmarks of metastasis, whereas Kindlin-1-silencing prevented tumor growth and lung metastasis in mice.

Implications
Kindlin-1 expression may be useful in identifying breast cancer patients with higher risk of developing lung metastasis.In addition, targeting kindlin-1 function may be an effective strategy for blocking metastasis in breast and other cancers.

Limitations
Kindlin-1 expression in other cancers besides breast cancer, which metastasize to the lungs, was not assessed.Although kindlin-1 expression was associated with other genes important in metastasis initiation, the molecular dissection of this association was not addressed.

From the Editors
actually be a melanoma line (36).The phenotype of HMEC cells was verified by Lonza by immunostaining for specific markers, and a certificate of analysis confirmed the authentication of cells.The phenotypes of the other cells were verified by ATCC by immunostaining for specific markers.The protocols for verifying the cell lines are available on the ATCC Web site (http://www.atcc.org/Portals/1/Pdf/CellBiologyStandards.pdf).The cell lines 67NR, 168 FARN, and 4T07 were used before passage 20, and the 4T1 cell line was used before passage 13.

Constructs and Transfections
Expression Constructs.The human Kindlin-1 cDNA, kindly provided by Dr Beckerle (Huntsman Cancer Institute, Salt Lake City, UT), was subcloned into the phCMV2-HA and pIREShyg3 vectors.Transfections were performed using Lipofectamine Reagent (Invitrogen) following the manufacturer's instructions.Stable transfectants were grown in the presence of 1 mg/mL geneticin G418 or 50 µg/mL hygromycin (Sigma-Aldrich).At least two clones of each cell line were used for in vitro experiments to exclude clonal variation.

Quantitative Real-Time Polymerase Chain Reaction(qRT-PCR).
Total RNA extraction, cDNA synthesis, PCR conditions, and normalization methods were described in detail elsewhere (40).Briefly, total RNA was isolated using a standard acid-phenol guanidium method.RNA concentration and purity were measured on the Nanodrop ND-1000 (Nanodrop Technologies, Wilmington, DE).First-strand cDNA was synthesized using a SuperScriptII Reverse Transcriptase kit (Invitrogen) according to the manufacturer's guidelines.All PCR reactions were performed with an ABI Prism 7700 Sequence Detection System (Applied Biosystems, Carlsbad, CA) and the SYBR Green PCR Core Reagents kit (Applied Biosystems).TATA box-binding protein (TBP) transcripts were used as an endogenous RNA control, and each sample was normalized on the basis of its TBP content (40).Kindlin-1-specific primers were forward: 5′AAG GAA CTT GAA CAA GGA GAA CCA CT 3′ and reverse: 5′ GGC ACA ACT TCG CAG CCT CTA 3′.For each primer pairs a standard curve was performed using a serial dilution of a pool of cDNAs consisting of several normal tissues (placenta, fetal brain, testis, and breast).The linearity of standard curves was verified, with all coefficients of variation between 0.96 and 0.99.Antibody Production.Polyclonal antibodies against kindlin-1 were generated by inoculating rabbits with two distinct human kindlin-1 peptides corresponding to amino acids 652-666 (FLSTRSKDQNETLDE) and amino acids 663-677 (TLDEDLFHKLTGGQD) (Eurogentec, Seraing, Belgium).Antibodies were affinity purified on a sepharose matrix.
Clonogenicity Assay.MCF7 transfected cells (1 × 10 2 ) were seeded in the presence of 2 × 10 5 parental cells in 60-mm diameter Petri dishes and maintained for 3 weeks under geneticin selection.Colonies were fixed with methanol and stained with 0.5% crystal violet.Results are representative of three experiments performed in triplicate.
Transwell Migration Assay.Migration assays were performed in triplicate using cell culture inserts with 8.0-µm pore size membranes (Becton Dickinson, Franklin Lakes, NJ) according to the manufacturer's protocol.MCF7 cells (5 × 10 4 ) were plated in the top chamber in DMEM.In the bottom chamber, culture medium containing 10% FBS was used as a chemoattractant.For invasion assays, the top chamber was pre-coated with 4 µg/cm 2 of Matrigel (BD Biosciences).Twenty four hours later, cells were fixed, stained using crystal violet, and counted to estimate total membrane area.
Collagen I Invasion Assay.The following precooled components were gently combined and defined as type I collagen solution: four volumes of type I collagen (stock is 3.49 mg/mL), five volumes of calcium-and magnesium-free Hank's balanced salt solution, one volume of MEM (10×), one volume of 0.25M NaHCO 3 , 2.65 volumes of culture medium, and 0.3 volumes of 1 M NaOH.For each test condition, 1.25 mL of type I collagen solution was added to one well of a six-well plate, homogeneously spread, and solidified on a flat surface in a humidified atmosphere of 10% CO 2 in air at 37°C for at least 1 hour.Control or Kindlin-1 transfected MCF7 single cells (2 × 10 5 ) suspended in 1 mL culture medium were seeded on top of the type I collagen gel and incubated on a flat surface in a humidified atmosphere of 10% CO 2 in air at 37°C.Cell morphology was studied and invasion was scored after 24 hours (41).The number of invasive and noninvasive cells was counted in ten randomly selected microscopic fields of objective 20× using an inverted phase contrast microscope (DMI 3000B; Leica, Wetzlar, Germany).The invasion index was calculated as the ratio of the number of cells that invaded the gel divided by the number of noninvasive cells counted in each field.
Wound Healing Assay.MDA-MB-435S cells were cultured to confluence in 24-well plates pre-coated with 5 µg/mL laminin-111.Monolayers were scratched with a pipette tip and washed with PBS twice to remove debris.Cells were maintained in DMEM with 10% FBS and antibiotics and imaged by phase contrast microscopy (Eclipse TS100, Nikon).Images were captured at the time and 24 hours after scratching.

Mouse Studies
All mouse experiments were conducted according to French veterinary guidelines, the Curie Institute guidelines on animal care, and those formulated by the council of Europe for experimental animal use (L358-86/609EEC).Transduced 4T1 cells (5 × 10 5 ) were injected subcutaneously into the fourth mammary gland of 12-week-old female BALB/c mice (n = 10 controls and n = 6 Kindlin-1knockdown mice).Twenty-four days later, all mice were killed by cervical dislocation, and primary tumors and lungs were removed, measured with calipers, and photographed.Tumor volumes were calculated as: volume (cm 3 ) = a × b 2 /2 (a and b are the two registered perpendicular diameters, with a > b).Serial lung sections (every 300 µm) were stained with Mayer's hematoxylin.The number of metastatic foci was determined in a blinded manner by two independent anatomopathologists.

Statistical Analysis
All statistical calculations were performed using PASW Statistics (version 18.0; SPSS Inc, Chicago, IL).Comparisons were performed using a two-sided unpaired Student t test.All experimental data presented are representative of at least two independent experiments performed in triplicate.To determine whether Kindlin-1 expression levels were able to define low-and high-lung metastasis risk populations, the optimal cutoff point to categorize patients into the two distinct risk groups was evaluated by use of the receiver operating characteristic method in the training set (n = 516 and 62 for breast and lung cancer, respectively) and then applied to the validation sets (MSKCC, EMC, NKI, or combined cohort for breast cancer and Memorial Sloan-Kettering Cancer Center, Dana-Farber Cancer Institute, or combined cohort for lung cancer).Survival distributions were estimated by the Kaplan-Meier method, and the significance of differences between survival rates was ascertained using the log-rank test.Multivariable analysis using Cox proportional hazards model was used to assess the independent contribution of each variable to lung metastasis-free survival.All variables included in the Cox model were categorical: estrogen receptor status was considered negative or positive according to the anatomopathological examination; lymph node status was referred to as positive when at least one lymph node was affected; the basal-like subtype was evaluated according to the intrinsic gene signature (42), and the "Lung Metastasis Signature" as reported in Minn et al. (12).The proportional hazards assumption was tested using tests and graphs based on the Schoenfeld residuals.
Oncomine database queries (https://www.oncomine.org)were performed using the same criteria for all tested cancer types.Briefly, we searched all of the gene expression datasets for the Kindlin-1 gene in the "cancer versus normal analysis" category.To avoid discrepancies among the results, we used the same Affymetrix probe (218796_at) Affymetrix, Santa Clara, CA.All studies are presented in the results section, except two out of four datasets on prostate cancer because they all gave similar results.
All statistical tests were two-sided.P values less than .05were considered to be statistically significant, and where appropriate, the difference in means and the 95% confidence interval (95% CI) are indicated.

Evaluation of Kindlin-1 as a Breast Cancer Lung Metastasis-Associated Gene
We previously identified (15) Kindlin-1 as a highly differentially expressed gene in lung metastases from primary breast tumors compared with metastases in other organs (6.9-fold increase, P < .001,Student t test).These microarray results have been confirmed both at the mRNA and protein levels in additional samples of metastases (Supplementary Figure 1, A and B, available online).Kindlin-1 expression was visible in lung metastases but not in adjacent host parenchyma (Supplementary Figure 1, B, available online), indicating that kindlin-1 is specifically expressed by cancer cells.Non-lung metastases showed low to no kindlin-1 expression (Supplementary Figure 1, B, available online).
Moreover, we analyzed whether kindlin-1 is differentially expressed in breast tumors according to the metastatic site.Kindlin-1 expression was tested by qRT-PCR on an additional sample of 516 primary breast tumors with a well-documented clinical follow-up (Supplementary Table 1, available online).Higher expression was observed in tumors metastasizing to the lung than in tumors metastasizing to bone (mean normalized mRNA levels: lung-metastasizing tumors vs bone-metastasizing tumors, 1.58 vs 0.69, difference = 0.89, 95% CI = 0.33 to 1.46, P = .002,Figure 1, B). Immunohistochemical analysis of primary tumors and subsequent metastases (n = 22) confirmed that patients metastasizing to the lungs had strong immunoreactivity in their primary tumors and their lung metastases, but not in their non-lung metastases (Figure 1, C).

Kindlin-1 as a Prognostic Factor for Breast Cancer Lung Metastasis
To test the prognostic value of Kindlin-1 expression, we performed a univariate analysis on the same sample of tumors from 516 patients (training set).We defined the stratification cutoff to classify patients into high-and low-Kindlin-1 groups in the training set and generated Kaplan-Meier plots to illustrate the survival differences among these groups for both the training and validation (n = 721) datasets (Figure 1, D and E and Supplementary Figure 2, A  and B).Survival analyses revealed that the risk of lung metastasis was statistically significantly higher for patients with increased Kindlin-1 mRNA levels (lung metastasis-free survival: training set, hazard ratio [HR] of lung metastasis = 1.85, 95% CI = 1.00 to 3.43, P = .03;validation set, HR of lung metastasis = 2.55, 95% CI = 1.39 to 4.69, P = .001,log-rank test).In contrast, we found no association between Kindlin-1 expression and the outcome with regard to overall distant metastasis or specific bone metastasis (Supplementary Figure 2, C-E, available online).
To determine whether Kindlin-1 is associated with a poor prognosis subtype of breast cancer, we analyzed Kindlin-1 expression in the NKI cohort that was categorized into five molecular subgroups as defined by the "intrinsic signature" (42).Indeed, the basal subgroup with the worst survival rates had the highest levels of Kindlin-1 transcripts (P < .001,Student t test, Supplementary Figure 2, F, available online) indicating that kindlin-1 is a marker of breast tumor aggressiveness and is potentially linked to the basal-like phenotype.
Finally, we performed a multivariable Cox proportional hazards analysis on the validation dataset (n = 721) including the variables Kindlin-1 expression, estrogen receptor status, lymph node status, basal-like subtype, and the Lung Metastasis Signature derived from a MDA-MB-231 xenograft mouse model (12).Kindlin-1 expression was the only parameter statistically significantly associated with the risk of breast tumors metastasizing to the lungs (Supplementary Table 2, available online).Kindlin-1 as a prognostic marker for breast cancer lung metastasis.A) Kindlin-1 mRNA expression (relative units) in primary tumors (grades 1-2 or 3; Gene Expression Omnibus accession no.GSE7390) vs lung metastases (GSE11078 and GSE14020) as determined by microarray gene expression.Grades 1-2 vs 3, P = .01,grade 3 vs lung metastases, P = .02,and grades 1-2 vs lung metastases, P < .001,Student t test.B) Kindlin-1 expression in tumors metastasizing exclusively to lungs vs to bones, as determined by qRT-PCR on an additional series of 516 primary breast tumors with a well-documented clinical follow-up (Supplementary Table

Effect of Kindlin-1 on the Metastatic Capacities of Breast Cancer Cell Lines
Kindlin-1 expression was higher in invasive and metastatic human tumor cell lines (MDA-MB-231 and MDA-MB-468) than in poorly invasive cell lines (MCF7 and SKBR3) (Figure 3, A).Moreover, kindlin-1 was only expressed in the highly metastatic 4T1 cells but not in the three nonmetastatic cell lines (67NR, 168FARN, 4T07), all derived from a single mouse mammary tumor (36).These observations suggest that kindlin-1 increases the metastatic capacities of breast cancer cells.To further investigate this hypothesis, we assessed the effect of stable Kindlin-1 expression in MCF7 cells on proliferation and clonogenicity.According to the proliferation assay, the number of Kindlin-1-cells was 2.7-fold higher than the control cells at 96 hours (Proliferation as measured by optical density at 490 nm for control cells vs Kindlin-1-cells, 0.25 vs 0.68, difference = 0.42, 95% CI = 0.34 to 0.51, P < .001, Figure 3, B).Moreover, the Kindlin-1-cells were more clonogenic than controls.Notably, control cells formed tighter and more compact colonies, whereas the Kindlin-1 colonies had a dispersed pattern, which may reflect the effects of kindlin-1 on cell spreading, migration, and invasion (Figure 3, C).Indeed, control cells were minimally motile, whereas Kindlin-1 cells showed statistically significantly more cell migration (number of migrating cells, Kindlin-1-cells vs control, 164.66 vs 19.00, difference = 145.6,95% CI = 79.1 to 212.2, P = .004,Figure 3, D).This result was confirmed using a wound healing assay with MDA-MB-435S cancer cells.Kindlin-1 cells showed a more extensive wound closure area with individual random migratory cell behavior compared with control cells (Figure 3, E).Kindlin-1-MCF7 cells were much more invasive through a native collagentype I matrix (invasion rate, Kindlin-1 cells vs control = 9.65% vs 1.92%: difference = 7.73%, 95% CI = 4.75% to 10.70%, P < .001, Figure 3, F).In addition, Kindlin-1-cells showed increased local spreading with formation of cell extensions, whereas control cells had a typical epithelial morphology (Figure 3, G).Moreover, two distinct shRNA showing a similar extinction of Kindlin-1 expression reduced the migratory and invasive capacities of 4T1 cells (Supplementary Figure 4, A-C).

Implication of Kindlin-1 in the Epithelial-Mesenchymal Transition
We next explored whether Kindlin-1 overexpression in MCF7 and MDA-MB-435S cells induces morphological changes and cytoskeleton reorganization.Indeed, Kindlin-1 cells exhibited a disruption in cell contacts, a spindle-shaped and round morphology, and the formation of actin stress fibers and lamellipodia, which are hallmarks of the epithelial-mesenchymal transition (EMT) and cell motility (Figure 4, A).Moreover, Kindlin-1 cells had increased levels of mesenchymal markers and decreased levels of epithelial markers, consistent with a Kindlin-1 mediated EMT (Figure 4, B  and C).In addition, Kindlin-1-MCF7 cells had more diffuse staining of E-cadherin, in contrast to the membrane-localized staining in the control lines (Figure 4, D), suggesting that kindlin-1 overexpression also leads to a redistribution of E-cadherin from the membrane to the cytoplasm.
We used a luciferase reporter assay to examine whether kindlin-1 represses transcription of the E-cadherin gene (CDH1).Cells overexpressing Kindlin-1 had lower CDH1 promoter activity than control cells (50% decrease in luciferase activity relative to control, difference in fold change: 0.52, 95% CI = 0.43 to 0.60, P = .001,Figure 4, E).Mutation of the E-boxes, which are critical for transcriptional repression of CDH1 (39), abrogated the effect of kindlin-1 on the CDH1 promoter (Figure 4, E).Moreover, we found that kindlin-1 stimulates the expression of zinc finger protein SNAI2 and Twist-related protein 1, two major CDH1 repressors (8.5-and 5.4-fold change in relative mRNA levels, respectively), whereas the expression of the zinc finger protein SNAI1 repressor was unaffected (Figure 4, F and data not shown).
Finally, we used the specific TGFbR1 kinase inhibitor SB431542 to ascertain whether Kindlin-1-induced EMT would require signaling through the TGFb receptor.This inhibitor prevented the phosphorylation of Smad2 (Figure 5, D) and the TGFb-induced EMT in Kindlin-1 cells (Figure 5, E).Together, these data suggest that Kindlin-1-induced EMT is TGFb dependent.
Notably, after exposure to TGFb, Kindlin-1-siRNA-treated HMECs maintained cell-cell adhesion and partially retained their original cobblestone-like epithelial morphology (Figure 6, A, B).In contrast, control cells treated with TGFb lost their cell-cell contacts and displayed a fibroblast-like morphotype (Figure 6, B).Consistently, the depletion of the E-cadherin and b-catenin epithelial markers induced by TGFb was suppressed in Kindlin-1silenced cells.Similarly, the induction by TGFb of the mesenchymal markers vimentin, fibronectin, and a-actin was repressed in Kindlin-1-silenced cells (Figure 6, C).These results indicate that  kindlin-1 suppression restricted the TGFb-dependent EMT phenotype.

Effect of Kindlin-1 Knockdown on Primary Breast Tumor Growth and Lung Metastasis
To further investigate whether Kindlin-1 may regulate invasive tumor growth and lung metastasis, we used the highly metastatic 4T1 syngeneic mouse model in which kindlin-1 expression was depleted.The orthotopic injection of Kindlin-1 shRNA-4T1, or control shRNA-4T1 cells (n = 6 and 10, respectively) into the mammary fat pad of syngeneic BALB/c mice led to the formation of primary mammary tumors in all mice.However, the mean tumor size in Kindlin-1-knockdown mice was statistically significantly lower (85% decrease) than in control mice (mean tumor size, Kindlin-1-depleted cells vs control cells, 0.36 cm 3 vs 2.46 cm 3 , difference = 2.10 cm 3 , 95% CI = 1.21 to 2.98, P < .001, Figure 7, A and B).
Notably, immunohistochemistry of tumors derived from our mouse model showed that the Kindlin-1-silenced tumors exhibited higher E-cadherin levels with more membranous staining than control tumors (Figure 7, C).Taken together, these data suggest that properties of the EMT are manifest in Kindlin-1expressing cells in vivo and are weakened when Kindlin-1 is silenced.
Finally, we assessed the effect of kindlin-1 depletion on the lung metastatic potential of breast cancer cells (Figure 7, D and E).Although macroscopic metastases were found in lung sections of control mice, no lung micrometastases were detected in Kindlin-1-knockdown mice (mean number of metastatic loci in the lungs of Kindlin-1knockdown mice vs control mice, 0 vs 4.60, difference = 4.60, 95% CI = 1.79 to 7.41, P = .003,Figure 7, D).These results show that Kindlin-1 silencing inhibited lung metastasis of 4T1 breast cancer cells.

Discussion
To our knowledge, this is the first study to show that kindlin-1 expression in breast tumors is associated with lung metastasis and lung metastasis-free survival.We provide evidence that kindlin-1 is potentially a clinically relevant mediator of lung metastasis in breast cancer and possibly other carcinomas metastasizing to the lung such as colon cancer, for which overexpression of kindlin-1 was previously reported in a small cohort of patients (seven of 10 patients) (55).In addition, we show that kindlin-1 is overexpressed in lung primary tumors, in which 60-fold increased expression of the gene was previously described (six of 10 lung tumors) (55).Importantly, Kindlin-1 expression is strongly associated with metastasis-free survival of patients with lung adenocarcinomas.
In addition, our in vitro studies showed that kindlin-1-expressing cells displayed increased proliferation, clonogenicity, and invasion, which are hallmarks of the initiation and progression of primary cancers to metastasis (3,7).Our data are consistent with previous reports on Kindler syndrome keratinocytes and Kindlin-1-deficient mice showing decreased proliferation, impaired cell adhesion, and delayed cell spreading (26,56,57).Although a higher propensity to develop cancer has been suggested for Kindler syndrome patients, there is a clear distinction between this inherited disease in which Kindlin-1 is inactivated and the sporadic cancers reported here in which Kindlin-1 is wild type and overexpressed.The mechanisms underlying skin cancers in Kindler Syndrome are likely to be different from the Kindlin-1 alterations described here.
Consistent with the role of kindlin-1 in the progression of human cancers, we have shown that Kindlin-1 silencing prevented tumor growth and lung metastasis in mice using an orthotopic and syngeneic breast tumor model.Although metastasis-associated genes are usually not involved in the growth of primary tumors, kindlin-1, like some other mediators of breast cancer metastasis to the lungs (12,30), may facilitate both breast tumorigenicity and lung metastasis.However, we did not find any association between kindlin-1 expression and breast tumor size in clinical human breast cancer patients.Indeed, a multivariable analysis assessing Kindlin-1 expression and tumor size in our validation breast tumor dataset (n = 721) showed that Kindlin-1 was an independent predictor of lung metastasis (not shown).
Kindlin-1 has been shown to be an essential regulator of integrin signaling and cellular adhesion to the ECM proteins such as fibronectin and laminin (22).A recent study (58) demonstrated that b1 integrin signaling is a critical regulator of the proliferation of micrometastatic breast cancer cells in lungs.Furthermore, this process was dependent on the activation of focal adhesion kinase (FAK), which is, like b-integrins, a binding partner of kindlin-1 (59).Thus, the possible implication of the integrin-FAK axis in kindlin-1-mediated lung metastasis is not excluded.Because integrins and adhesion to ECM components may regulate cadherindependent cell adhesion mechanisms, kindlin-1 might function at the intersection of multiple signaling pathways and cellular functions during neoplasia.
We have provided evidence that kindlin-1 initiates TGFbdependent EMT operating through reduced expression and relocalization of E-cadherin coupled with the induction of several mesenchymal markers such as N-cadherin and fibronectin.Both TGFb and EMT are potent components of signaling and cellular alterations involved in cancer initiation, invasive growth, and metastasis (60).Most notably, transduced cancer cells expressing kindlin-1 exhibit a constitutive TGFb-dependent activation of the canonical TGFb signaling elements Smad-2 and Smad-3 coupled with the transcription of several TGFb and Smad-inducible genes such as CTGF, EDN1, and MMP9, which are known to be involved in breast cancer progression.In line with our findings, TGFb has been shown to enhance lung metastasis formation in several transgenic mouse models (61,62).More recently, a TGFb response gene signature was shown to predict lung metastasis in human breast cancer (53).The authors demonstrated that TGFb signaling in the breast tumor microenvironment primes cancer cells for metastasis to the lungs (53).
This study has several potential limitations.First, we were not able to assess the prognostic value of Kindlin-1 expression in other cancers metastasizing to the lung because of the lack of documented clinical follow-up.In future investigations, it will be important to determine the involvement of Kindlin-1 in large and well-documented cohorts of colon and bladder cancer patients.Second, our study did not explain in detail the reciprocal crosstalk between kindlin-1 and TGFb in EMT.Clearly, kindlin-1 may provide a new link between TGFb, the ECM, and cell-cell adhesion systems in the tumor microenvironment.Although, our results suggest interdependent functions of kindlin-1 and TGFb in the fine-tuning of EMT and neoplasia, the molecular dissection of this reciprocal crosstalk needs further investigations.
In conclusion, our study suggests that the role of kindlin-1 in human cancers should be assessed further and its possible value for guiding diagnostics and therapeutics considered.First, Kindlin-1 analysis can be used to identify breast cancer patients with higher risk of developing lung metastasis.Second, targeting kindlin-1 function may be an effective strategy for blocking the EMT and metastasis.Given the importance of the EMT and lung metastasis in the progression of other tumor types, Kindlin-1 may also have much broader clinical utility relevant to other cancer types.

Figure 1 .
Figure 1.Kindlin-1 as a prognostic marker for breast cancer lung metastasis.A) Kindlin-1 mRNA expression (relative units) in primary tumors (grades 1-2 or 3; Gene Expression Omnibus accession no.GSE7390) vs lung metastases (GSE11078 and GSE14020) as determined by microarray gene expression.Grades 1-2 vs 3, P = .01,grade 3 vs lung metastases, P = .02,and grades 1-2 vs lung metastases, P < .001,Student t test.B) Kindlin-1 expression in tumors metastasizing exclusively to lungs vs to bones, as determined by qRT-PCR on an additional series of 516 primary breast tumors with a well-documented clinical follow-up (Supplementary Table1, available online).Means and upper 95% CIs are shown for one experiment.P = .002,Student t test.C) Representative immunohistochemical staining of kindlin-1 in matched pairs of primary tumors and metastases from two individuals with breast cancer from a larger sample (n = 22).Non-lung

1
Figure 1.Kindlin-1 as a prognostic marker for breast cancer lung metastasis.A) Kindlin-1 mRNA expression (relative units) in primary tumors (grades 1-2 or 3; Gene Expression Omnibus accession no.GSE7390) vs lung metastases (GSE11078 and GSE14020) as determined by microarray gene expression.Grades 1-2 vs 3, P = .01,grade 3 vs lung metastases, P = .02,and grades 1-2 vs lung metastases, P < .001,Student t test.B) Kindlin-1 expression in tumors metastasizing exclusively to lungs vs to bones, as determined by qRT-PCR on an additional series of 516 primary breast tumors with a well-documented clinical follow-up (Supplementary Table1, available online).Means and upper 95% CIs are shown for one experiment.P = .002,Student t test.C) Representative immunohistochemical staining of kindlin-1 in matched pairs of primary tumors and metastases from two individuals with breast cancer from a larger sample (n = 22).Non-lung Figure 1.Kindlin-1 as a prognostic marker for breast cancer lung metastasis.A) Kindlin-1 mRNA expression (relative units) in primary tumors (grades 1-2 or 3; Gene Expression Omnibus accession no.GSE7390) vs lung metastases (GSE11078 and GSE14020) as determined by microarray gene expression.Grades 1-2 vs 3, P = .01,grade 3 vs lung metastases, P = .02,and grades 1-2 vs lung metastases, P < .001,Student t test.B) Kindlin-1 expression in tumors metastasizing exclusively to lungs vs to bones, as determined by qRT-PCR on an additional series of 516 primary breast tumors with a well-documented clinical follow-up (Supplementary Table 1, available online).Means and upper 95% CIs are shown for one experiment.P = .002,Student t test.C) Representative immunohistochemical staining of kindlin-1 in matched pairs of primary tumors and metastases from two individuals with breast cancer from a larger sample (n = 22).Non-lung metastases were harvested from the bone (top) and uterus (bottom) of patient 1, and from the liver (top) and spleen (bottom) of patient 2. Original magnification ×400, scale bar = 50 µm.D) Kaplan-Meier curves showing the lung metastasis-free survival of patients with tumors expressing high vs low levels of Kindlin-1 in the training set of 516 breast cancer patients analyzed by qRT-PCR, HR of lung metastasis = 1.85, 95% CI = 1.00 to 3.43, P = .03.E) Kaplan-Meier curves showing the lung metastasis-free survival of patients with tumors expressing high vs low levels of Kindlin-1 in the combined cohort of 721 breast cancer patients corresponding to three independent microarray datasets (29-32), HR of lung metastasis = 2.55, 95% CI = 1.39 to 4.69; P = .001,log-rank test.All statistical tests were two-sided.CI = confidence interval; HR = hazard ratio; qRT-PCR = quantitative realtime polymerase chain reaction.

Figure 3 .
Figure 3. Kindlin-1 involvement in aggressive cancer phenotypes in vitro.A) Expression of kindlin-1 protein in human and mouse metastatic and nonmetastatic cell lines.B) Effect of ectopic expression of Kindlin-1 in MCF7 cells on cell proliferation as monitored by the MTS assay.The absorbance at 490 nm was measured at a different time points.Results are representative of three independent experiments performed in triplicate.Means and 95% confidence intervals are shown for one experiment; P < .001,Student t test.C) Clonogenic assays were performed with MCF7 control (left panel) and Kindlin-1 (right panel) transfectants.The higher magnifications in the insets highlight the morphological differences between colonies of control and Kindlin-1 cells.D) The total number of migrating MCF7 cells was assessed using transwells.Results are representative of two independent experiments performed in triplicate.Means and upper 95% confidence intervals of triplicates are shown for one experiment; P = .004,Student t test.E) Effect of ectopic expression of Kindlin-1 in MDA-MB-435S cells (right) on a wound healing assay.Microscopic observations were recorded after scratching the cell surface following the indicated periods.Scale bar = 200 µm.F) Invasion of MCF7 transfectants was assessed in type I collagen matrix.Results are representative of three independent experiments performed in triplicate.The invasion index was calculated by counting the number of invading and non-invading cells in 10 fields.The means and upper 95% confidence intervals represent one experiment performed in triplicate; P < .001,Student t test.G) Morphology of MCF7 cells in the collagen matrix by phase contrast microscopy.Arrows indicate cell extensions.Scale bar = 20 µm.All statistical tests were two-sided.

Figure 4 .
Figure 4. Role of Kindlin-1 in EMT in breast cancer cells.A) Morphological changes between control and Kindlin-1 cells growing on laminin substrate (5 µg/mL) as observed by immunofluorescence microscopy in MCF7 and MDA-MB-435S cells.The cytoskeletal reorganization was visualized by F-actin labeling.The arrow shows a lamellipodia-like structure.Scale bar = 20 µm.B) Expression of EMT-related proteins (E-cadherin, fibronectin, g-catenin, vimentin, N-cadherin) in control and Kindlin-1 cells.C) Immunostaining of E-cadherin, ZO-1, N-cadherin, and vimentin in the MCF7 transfectants.Cells were seeded on laminin substrate (5 µg/mL).Cells were then stained with PBS containing 1 µg/mL DAPI (diamidino-2-phenylindole) to visualize the nuclei.Epithelial and mesenchymal markers were labeled with the same antibodies as used for western blots, except for mouse monoclonal anti-ZO-1 antibody (1/ZO-1 clone, 1:5000, BD Biosciences).Scale bar = 20 µm.D) Fluorescence staining of F-actin and E-cadherin in the MCF7 transfectants at higher magnification.F-actin was localized using fluorescent-labeled phalloidin conjugated to tetramethylrhodamine isothiocyanate (Sigma-Aldrich) at a dilution of 1:100 of a methanolic stock (200 U/mL) solution.The arrow indicates a partial cytoplasmic relocalization of E-cadherin.Scale bar = 20 µm.E) MCF7 cells were transfected with the wild-type or mutant pGL3-E-cad reporter construct in the presence or absence of Kindlin-1.The luciferase activity was determined and normalized to an internal control.Data are representative of two independent experiments performed in triplicate.Means and upper 95% confidence intervals are shown for triplicates of one experiment.Difference in fold change for wild type CDH1 promoter activity: P = .001,for mutant CDH1 promoter activity: P = .2,two-sided Student t test.F) Expression of EMT-related transcriptional factors zinc finger protein SNAI1, zinc finger protein SNAI2, and twistrelated protein TWIST1 in stable MCF7 transfectants was analyzed by immunoblotting.CDH1 = E-cadherin; EMT = epithelial-mesenchymal transition; GAPDH = glyceraldehyde 3-phosphate dehydrogenase; SNAI1 = snail homolog 1; SNAI2 = snail homolog 2; TWIST1 = twist homolog 1.

Figure 5 .
Figure 5.Effect of Kindlin-1 on TGFb signaling and TGFb-induced EMT in human mammary cell lines.A) MDA-MB-231 cells transfected with the p3TP-lux reporter construct in the presence or absence of Kindlin-1 were treated for 16 hours with TGFb1 (2 ng/mL).Luciferase activity was determined and normalized to an internal control.Data are representative of two independent experiments performed in triplicate.Means and upper 95% confidence intervals are shown for triplicates of one experiment.B) Expression of total and phosphorylated Smad2 and Smad 3 proteins in control-MCF7 and two clones of Kindlin-1-MCF7 cells (Kindlin-1-Cl1 and Kindlin-1-Cl2).C) Expression of TGFb target genes in Kindlin-1-expressing MCF7 cells or Kindlin-1-depleted HMECs was evaluated by quantitative real-time polymerase chain reaction.Each bar represents expression of a target gene as fold change relative to the controls.Results are representative of one experiment performed for several clones.Differences in fold change are statistically significant: for CTGF, P = .02;for EDN1: P < .001;for EGR1: P = .10;for TGFb2: P = .004,for MMP9: P = .003,Student t test.D) Expression of phosphorylated Smad2 in MCF7 cells transiently expressing kindlin-1 after treatment for 24 hours with the TGFbR1 inhibitor SB431542 (10 µM).E) Morphology of Kindlin-1-expressing MCF7 cells untreated or treated with SB431542 was revealed by phase contrast microscopy.Scale bar = 200 µm.CTGF = connective tissue growth factor; EDN1 = endothelin 1; EGR1 = early growth response 1; EMT = epithelial-mesenchymal transition; GAPDH = glyceraldehyde 3-phosphate dehydrogenase; HMEC = human mammary epithelial cells; MMP9 = matrix metallopeptidase 9; TGFb1 = transforming growth factor b1; Smad2 = SMAD family homolog 2.

Figure 7 .
Figure 7. Effect of Kindlin-1 knockdown on tumorigenic and metastatic potential to lungs in highly invasive 4T1 breast cancer cells.A) 4T1 cells transduced with Kindlin-1 shRNA-2 or control shRNA were injected into the mammary fat pad of syngeneic BALB/c mice (n = 6 and 10, respectively).At 24 days after injection, mean tumor volume was determined in each group of mice; bars represent the means and upper 95% confidence intervals; P < .001,Student t test.B) Two representative images of primary tumors from control shRNA and Kindlin-1 shRNA mice, scale bar = 1 cm.C) Primary tumor sections from the Kindlin-1-depleted cells were stained for kindlin-1 and E-cadherin.Original magnification ×400, scale bar = 50 µm.D) Mean number of lung nodules in control shRNA and Kindlin-1 shRNA mice (n = 6 and 10, respectively).Bars represent the means and upper 95% confidence intervals; P = .003,Student t test.E) Two representative images of lungs harvested from control and Kindlin-1 shRNA mice (top) and of hematoxylin-stained sections of these lungs (bottom).The black arrows indicate metastatic foci.Original magnification ×100, scale bar = 50 µm.shRNA = short hairpin RNA.