Research ArticleCancer

SILAC identifies LAD1 as a filamin-binding regulator of actin dynamics in response to EGF and a marker of aggressive breast tumors

See allHide authors and affiliations

Sci. Signal.  30 Jan 2018:
Vol. 11, Issue 515, eaan0949
DOI: 10.1126/scisignal.aan0949

LAD1 marks aggressive breast cancer

The actin cytoskeleton is a framework of filaments that gives cells shape. Dynamic regulation of the cytoskeleton coordinates complex cell behaviors, such as cell-cell communication, migration and invasion, and cell division, the regulation of which is critical to development, tissue homeostasis, and disease. Roth et al. showed that signaling by the epidermal growth factor receptor (EGFR) promoted cell migration–associated actin dynamics through the phosphorylation of the protein LAD-1 (see also the Focus by Chiasson-MacKenzie and McClatchey). LAD1 was also a marker of aggressive subtypes or cases of breast tumors in patient samples; thus, it might be a useful biomarker to inform clinical decisions.

Abstract

Mutations mimicking growth factor–induced proliferation and motility characterize aggressive subtypes of mammary tumors. To unravel currently unknown players in these processes, we performed phosphoproteomic analysis on untransformed mammary epithelial cells (MCF10A) that were stimulated in culture with epidermal growth factor (EGF). We identified ladinin-1 (LAD1), a largely uncharacterized protein to date, as a phosphorylation-regulated mediator of the EGF-to-ERK pathway. Further experiments revealed that LAD1 mediated the proliferation and migration of mammary cells. LAD1 was transcriptionally induced, phosphorylated, and partly colocalized with actin stress fibers in response to EGF. Yeast two-hybrid, proximity ligation, and coimmunoprecipitation assays revealed that LAD1 bound to actin–cross-linking proteins called filamins. Cosedimentation analyses indicated that LAD1 played a role in actin dynamics, probably in collaboration with the scaffold protein 14-3-3σ (also called SFN). Depletion of LAD1 decreased the expression of transcripts associated with cell survival and inhibited the growth of mammary xenografts in an animal model. Furthermore, LAD1 predicts poor patient prognosis and is highly expressed in aggressive subtypes of breast cancer characterized as integrative clusters 5 and 10, which partly correspond to triple-negative and HER2-positive tumors. Thus, these findings reveal a cytoskeletal component that is critically involved in cell migration and the acquisition of oncogenic attributes in human mammary tumors.

INTRODUCTION

The erythroblastosis B/human EGF receptor (ERBB/HER) family of receptor tyrosine kinases (RTKs) consists of four members, including the epidermal growth factor receptor (EGFR), which binds with EGF and six other growth factors, and ERBB2/HER2, which binds with no known growth factor (1, 2). This receptor family plays important roles in the progression of breast and other cancers. For example, about 15% of human mammary tumors overexpress HER2, and about one-third of another subtype of breast cancer, the basal-like, displays EGFR overexpression (3). In addition to tumors highly expressing EGFR or HER2, several other tumor types often secrete growth factor ligands of EGFR, or they present mutant forms of either the receptors or the respective downstream signaling effectors, such as RAS proteins and mutants of phosphatidylinositol 3-kinase (PI3K) (4). On growth factor binding, EGFR undergoes homodimerization, or heterodimerization with HER2, thereby initiating branched kinase cascades responsible for concurrent stimulation of diverse cellular processes, including cell proliferation and migration (5). An especially useful method enabling high-granularity mapping of the downstream branched kinase cascades is stable isotope labeling by amino acids in tissue culture (SILAC), which entails metabolic labeling of proteins and permits high-throughput identification of phosphorylation substrates using mass spectrometry (MS) (6). For example, using SILAC and EGF-stimulated cells, it was shown that many substrates contain several phosphorylation sites, which display different kinetics of modification, suggesting that they integrate input from diverse signaling pathways (7).

Yet another MS analysis uncovered ligand-induced network modules that are specific to HER2-overexpressing cells (8). The cellular outcome of the EGFR-HER2 network often entails cell cycle progression and cell migration. Rapid motility of EGF-stimulated cells is propelled by the actin cytoskeleton. Actin-related protein (ARP2/3)–mediated nucleation and branching of actin filaments (F-actin), along with cytoskeleton remodeling by RHO family small guanosine triphosphatases (GTPases) (9, 10), underlay cell migration. In addition, the actin cytoskeleton is regulated by a large group of actin-binding proteins, both monomer binders (for example, profilin) and polymer binders, such as the tensins and filamins (11). Filamin A (FLNA) is a 280-kDa dimeric protein, which supports oncogenic phenotypes (12). Each monomer comprises an actin-binding domain (ABD) and a rod segment consisting of 24 highly homologous immunoglobulin-like repeats. Through dimer formation, FLNA anchors actin fibers to each other, thereby controlling cell structure and rigidity. It also anchors transmembrane proteins, such as integrins (13, 14).

To better resolve the effects of growth factors on target cells, we are applying transcriptome and proteome analyses on untransformed human mammary cells, which migrate or proliferate in response to EGF (15, 16). By using SILAC and focusing on events distal to tyrosine phosphorylation, we identify herein ladinin-1 (LAD1), a relatively uncharacterized 59-kDa protein (17), as a cytoskeletal phosphorylation substrate of the EGFR pathway. We further report two physical partners of LAD1, namely, FLNA and 14-3-3σ (also called SFN), and implicate LAD1 in actin fiber remodeling, proliferation, and motility of mammary cells. In line with these observations, we found that LAD1 differentially fractionates with monomers and polymers of actin, and its depletion inhibits tumorigenesis in an animal model of basal-like breast cancer. Furthermore, increased abundance of LAD1 is typical to high-grade breast tumors, characterizes both basal-like and HER2-positive human mammary tumors, and predicts poor prognosis of breast cancer patients. Together, these observations uncover a new regulator of the cytoskeleton and attribute to this protein, LAD1, specific roles in cell proliferation and migration, which explain the rather strong prognostic impact on human breast cancer.

RESULTS

SILAC-based proteomic analyses identify candidate serine and threonine phosphorylation substrates mediating the biological effects of EGF on mammary cells

Our previous studies using untransformed human mammary cells indicated that exposure to EGF stimulated the extracellular signal–regulated kinase (ERK) mitogen-activated protein kinase pathway and consequently increased cellular motility (15) and cell cycle progression (16). EGF induced migratory and metabolic and/or survival activities (respectively) of MCF10A-immortalized human mammary cells, which were incubated in regular animal serum or in serum that underwent previous dialysis [dialyzed serum (D. serum); Fig. 1, A and B]. Because of our interest in identifying yet-undiscovered phosphorylation substrates of the EGFR pathway, especially late-induced serine and threonine phosphorylation substrates, we applied SILAC. To enable metabolic labeling, we cultured MCF10A cells in SILAC solutions containing either medium or heavy isotopes of lysine and arginine. Thereafter, cells were stimulated with either EGF (medium isotopes) or serum growth factors (heavy isotopes) and harvested after specific time intervals (Fig. 1C). At each time point, the corresponding extracts were mixed and processed for MS. Analysis of phosphopeptides identified several serine phosphorylation sites of EGFR but no phosphotyrosine-containing peptides. Together, about 93% of the phosphate groups corresponded to phosphoserines, 7% corresponded to phosphothreonine, and only 0.3% was due to phosphotyrosines (Fig. 1D and table S1). Next, the phosphopeptide data were clustered according to the peak time of phosphorylation events (Fig. 1E). Comparison of phosphorylation and dephosphorylation events after EGF or serum stimulation uncovered extensive similarities. These observations are consistent with the mitogenic activity shared by both EGF and serum, but they raise the possibility that the previously reported ability of EGF, but not serum, to induce mammary cell migration (18) is due to subtle differences in kinetic and other phenomena rather than the result of phosphoproteins specifically induced by each stimulant.

Fig. 1 SILAC-based phosphoproteome analysis of EGF-stimulated mammary cells.

(A) MCF10A cells (0.5 × 105) were seeded in Transwells without or with epidermal growth factor (EGF; 10 ng/ml) in the presence of either untreated or dialyzed serum (D. serum). Sixteen hours later, cells were fixed in paraformaldehyde (3%), washed, and stained using crystal violet. Histograms are means ± SE of three experiments. ***P < 0.001. Scale bars, 1 mm. n.s., nonsignificant. (B) MTT assay of MCF10A cells (2 × 103) seeded in a 96-well plate and grown for 4 days with or without EGF (10 ng/ml). Data are means ± SE of quadruplicate experiments (***P < 0.001). (C) A schematic representation of the experimental layout of stable isotope labeling by amino acids in tissue culture (SILAC)–based analysis. Arrowheads mark time points of cell lysate preparation. (D) Relative abundances of phosphorylated residues in the proteomic data set (PXD008100 in ProteomeXchange). (E) Heatmap representation of the abundance of specific phosphopeptides in MCF10A cells stimulated with either EGF or serum for the indicated time intervals. Signals were normalized to the unstimulated state and sorted according to the timing of their peak (see the color bar for reference; missing values are shown in white).

Biochemical analyses identify LAD1 as a downstream phosphoserine and phosphothreonine effector of EGFR

To identify phosphorylation events that are specifically induced by EGF, we focused on phosphopeptides that changed upon EGF stimulation but were differently modified in response to serum treatment (tables S2 and S3). This analysis led to the identification of three candidate proteins: desmoplakin, a component of desmosomes; transmembrane protein 131, a poorly characterized transmembrane protein; and LAD1 (Fig. 2A). Closer analysis of LAD1 revealed seven distinct phosphorylation events (Fig. 2B). Each of these exhibited slightly different kinetics, with phosphorylation of Ser38 occurring early after either stimulus and modification of Ser64 occurring upon EGF but not upon serum stimulation.

Fig. 2 Proteomic and biochemical data identifies LAD1 as a downstream target of active EGFRs.

(A) Listed are proteins exhibiting phosphorylation changes in response to EGF or serum treatment. The relevant phosphorylated amino acids are indicated. The color bar represents signal intensities. TMEM131, transmembrane protein 131. (B) A heatmap of all ladinin-1 (LAD1) phosphopeptides identified using SILAC. The relevant phosphorylated amino acid, either serines or threonines, is indicated. (C) Serum-starved MCF10A cells were stimulated with EGF or serum for different time intervals, and extracts were immunoblotted (IB) with the indicated antibodies (AB). The blot is representative of three replicates. (D) MCF10A cells overexpressing LAD1-V5 were serum-starved and then preincubated with the indicated inhibitors before EGF stimulation. Wort, Wortmannin; GAPDH; glyceraldehyde 3-phosphate dehydrogenase. Cell extracts were as indicated. Blot is representative of three replicates. (E) Serum-starved MCF10A cells were stimulated for 60 min with EGF. Extracts were subjected to immunoprecipitation (IP) with an antibody to LAD1 and incubated without or with calf intestinal alkaline phosphatase (CIP) before immunoblotting as indicated. Data are from three experiments. (F) MDA-MB-231 cells ectopically expressing the indicated mutants of V5-tagged LAD1 were serum-starved and preincubated with U0126 before stimulation with EGF. Cell extracts were analyzed as in (D).

According to a previous report, the murine form of LAD1 is a secreted protein functioning as a component of the basement membrane (17). To address an extracellular location, we subjected MCF10A cells to surface biotinylation and captured the resulting biotinylated proteins using avidin. The results confirmed the reported membrane localization of the hemidesmosomal integrin α-6, but most LAD1 molecules were not captured (fig. S1A), in line with an intracellular location. Immunoblotting of the endogenous LAD1 revealed a time-dependent gel mobility shift that occurred (albeit differently) in response to stimulation with either EGF or serum (Fig. 2C). This shift was more prominent when we analyzed the human protein tagged with a V5 peptide (Fig. 2D). Aiming at identifying the signaling pathway that covalently modifies LAD1, we treated MCF10A cells with inhibitors targeting specific signaling pathways (Fig. 2D and fig. S1B). Specifically, blocking the kinase activity of EGFR or mitogen-activated protein kinase kinase (MEK) using specific kinase inhibitors abolished the gel mobility shift of LAD1. By contrast, cells treated with an upstream inhibitor of the AKT pathway, namely, wortmannin, retained the shift. To verify that the mobility shift corresponded to protein phosphorylation, we applied a calf intestinal alkaline phosphatase. As expected, treatment with the phosphatase abolished the LAD1 mobility shift (Fig. 2E). As an additional control, we mutated the EGF-specific site of phosphorylation, Ser64, into either an alanine, which cannot undergo phosphorylation, or to an aspartate, which mimics phosphorylation by introducing an extra negative charge. Consistent with our model, whereas the aspartate mutant (S64D) displayed a constitutive gel shift, the serine-to-alanine mutant (S64A) underwent no mobility shift in response to EGF (Fig. 2F). In conclusion, LAD1 undergoes EGF-induced phosphorylation on several sites by a kinase activated by the MEK pathway, and this modification can be detected by a gel mobility shift.

Genomic, proteomic, and transcriptional reveal characteristics of LAD1

Note that the name LAD1 was given also to two proteins distinct from the 59-kDa protein identified by our MS analysis: Linear immunoglobulin A bullous dermatosis autoantigen (LAD-1) is a secreted 120-kDa protein, which is localized to the basement membrane of skin (19), and lymphocyte-specific protein tyrosine kinase–associated adapter/Rlk- and Itk-binding proteins, which received the name LAD (or RLK/TXK- and ITK-binding adaptor; RIBP) (20). Because only very few studies previously addressed p59-LAD1, we examined inducibility by EGF and analyzed both the protein and the respective gene. The human LAD1 gene is localized to chromosome 1, and its exonic regions appear to be well conserved among mammals (fig. S2A). However, we found that LAD1 is less conserved in other vertebrate species, and it does not exist in invertebrates, indicating a late evolutionary origin. LAD1 was first cloned from a murine mouse skin cDNA library (17). The encoded protein harbors no recognizable structural domains, other than an arginine-rich N-terminal stretch and six serine-glutamate-lysine (SEK) tripeptide motifs, the function of which remains unknown. mRNA inducibility was studied by using four independent methods, namely, pausing (and traveling) of RNA polymerase II (RNAPII) at the promoter and gene body; triacetylation of histone 3’s lysine 27 (H3K27Ac), a marker of active transcription; RNA sequencing (RNA-seq); and reverse-transcriptase polymerase chain reaction (PCR) of the respective mRNA (fig. S2, A and B). All four approaches detected a delayed and moderate EGF-induced transcription, which was associated with increased recruitment of RNAPII to the respective promoter and enhancer. Several structure prediction servers (FoldIndex and SPINE-D) identified no distinct structures, but using protein sequence data sets, we found a weak sequence similarity between LAD1 and caldesmon, an actin- and calmodulin-binding protein (homologous segments are highlighted in fig. S2C). Analyses performed using GeneCards and cBioPortal (www.cbioportal.org/index.do) showed that LAD1 is transcribed in many types of carcinoma (fig. S2D), and the gene is expressed in several epithelial organs, such as the kidney and lung (fig. S2E). Moreover, very rare missense point mutations within LAD1 were reported in breast cancer patients (fig. S2C). Together, these observations portrayed LAD1 as an unstructured, evolutionarily “young” protein, which is widely expressed and undergoes transcriptional induction by EGF.

LAD1 partly colocalizes with the actin cytoskeleton

Immunostaining of MCF10A cells showed that endogenous LAD1 was organized not only along filaments decorating the cell edges but also within the cell body, largely resembling the distribution of actin (Fig. 3A). Co-staining with phalloidin revealed that LAD1 partially colocalized with fibrillar actin of both the stress fibers and the cellular cortex (Fig. 3A). To corroborate these observations, we isolated subcellular fractions corresponding to the cytoplasm, plasma membrane, and the cytoskeleton of MCF10A cells. Separately immunoblotting each fraction localized LAD1 to the cytoskeleton and, to some extent, also to the cytoplasm (Fig. 3B).

Fig. 3 LAD1 is partly colocalized with the actin cytoskeleton and its C-terminal region associates with F-actin.

(A) MCF10A cells were stained for LAD1 (red), actin (green), and nuclei (blue). LAD1 and actin colocalization revealed in the merged images (yellow). Scale bar, 10 μm. (B) Serum-starved MCF10A cells ectopically expressing V5-LAD1 were treated with EGF for the indicated time intervals, fractionated, and immunoblotted as indicated. The black and gray arrowheads mark the endogenous and ectopic LAD1-V5, respectively. Data are representative of three replicates. (C) MCF10A cells were visualized using the indicated antibodies, phalloidin, LifeAct-cherry (red), or LAD1-GFP (green). Scale bar, 10 μm. (D) Serum-starved MCF10A cells were treated with EGF, fixed, and probed for LAD1 (green), F-actin (red), and 4′,6-diamidino-2-phenylindole (DAPI). Scale bar, 30 μm. (E) Extracts from MCF10A and HCC70 cells (2 × 104) were preincubated (10 min) in F-actin–stabilizing buffer, in the absence or presence of phalloidin. Thereafter, soluble and filamentous actin forms were separated using an ultracentrifuge, and the fractions were probed with the indicated antibodies. Sup, supernatant. (F) MDA-MB-231 expressing V5-tagged forms of wild-type (WT) or deletion mutants of LAD1 [N terminal (N-term.) amino acids 1 to 276 and C-terminal (C-term.) amino acids 277 to 517] were treated and processed as in (E). Data are representative of two experiments. Cont., control.

Further analyses of LAD1 confirmed colocalization with the actin rather than the tubulin cytoskeleton (Fig. 3C). Next, we constructed a plasmid encoding a green fluorescent protein (GFP) fused to LAD1 and stably expressed it in MCF10A cells already expressing the F-actin dynamic marker mCherry-Lifeact. Along with validating colocalization of LAD1 with actin stress fibers (Fig. 3C, arrows), the images we obtained showed that some actin bundles, especially those confined to lamellipodia, were devoid of LAD1 (Fig. 3C, arrowheads). Taking advantage of EGF-induced migration of MCF10A cells (21), we focused on lamellipodia of stimulated cells. However, despite EGF-induced lamellipodia formation and scattering of cell clusters, we observed no visible changes in LAD1 localization (Fig. 3D, arrowheads mark lamellipodia).

The static distribution of LAD1 and constitutive association with F-actin raised the possibility that phosphorylated LAD1 molecules stably and specifically associated with F-actin. For fractionation of soluble and filamentous actin in live cells, we applied the G-actin/F-actin In Vivo Assay Kit on both MCF10A cells and on HCC70 mammary tumor cells. Cells were harvested in the presence or absence of phalloidin, which binds and stabilizes F-actin. The lysates were centrifuged at 100,000g. Supernatants containing G-actin and the pelleted F-actin were analyzed using gel electrophoresis and immunoblotting. As expected, the abundance of G-actin exceeded that of the filamentous form, which was enhanced by phalloidin (Fig. 3E). Surprisingly, however, in both cell lines and in three experiments, the phosphorylated form of LAD1 specifically associated with the F-actin–containing fraction, whereas the free form of actin cofractionated with the unmodified form of LAD1 (Fig. 3E). Next, we constructed two V5-tagged deletion mutants of LAD1, an N- and a C-terminal fragment (amino acids 1 to 276 and 277 to 517, respectively) and repeated the fractionation assay (Fig. 3F). This assay indicated that the C-terminal portion of LAD1, unlike the N-terminal portion, coprecipitated with F-actin. In summary, our observations indicate that two states of LAD1 exist in living cells: a soluble form and a modified form, which coprecipitated with actin filaments. Notably, EGF-induced phosphorylation of LAD1 might allow the C-terminal portion of the molecule to interact, either directly or indirectly, with the actin cytoskeleton.

Forced depletion of LAD1 suppresses cell migration and matrix invasion

The association of LAD1 with the actin cytoskeleton raised the possibility that this phosphorylation effector of EGFR has a role to play in cellular motility. To test this prediction, we stably reduced the abundance of LAD1 in MCF10A cells (by 30 to 40%; Fig. 4A) using short hairpin RNAs (shRNAs). Next, we used a two-chamber cell culture format to assay cell migration. This assay revealed that LAD1 depletion resulted in slower migration rates of mammary cells (Fig. 4B). Furthermore, coating the intervening filter with a collagen-rich matrix, called Matrigel, confirmed that two independently established LAD1-depleted cell lines partly lost their invasion capability (Fig. 4B). In line with these observations, when cells were briefly plated on fibronectin and shortly later washed away, most LAD1-depleted cells were removed, but most control cells remained adherent (Fig. 4C). The ability of cells to first adhere to and then invade through matrix barriers depends, in part, on formation of actin-filled and proteolytically active protrusions of the plasma membrane, called invadopodia (22). As expected, assays that quantified invadopodia formation after plating cells on fluorescent gelatin in EGF-containing full medium confirmed that LAD1 depletion impaired invadopodia-forming activity (Fig. 4D): Less subnuclear, gelatin-free characteristic spots were detectable when LAD1 expression was diminished. Similar effects were observed when cells were preincubated in the absence or presence of EGF (fig. S3). In conclusion, LAD1 supports the ability of cultured mammary cells to migrate, adhere to, and invade through extracellular barriers.

Fig. 4 LAD1 regulates the migratory and invasive potential of mammary cells.

(A) Extracts from MCF10A cells stably expressing shRNAs specific to LAD1 (or a scrambled shRNA) were immunoblotted using the indicated antibodies. Quantification shows normalized LAD1 abundance. Data are means ± SE of three independent replicates. (B) MCF10A cells expressing shLAD1 were incubated for 24 hours in Transwell chambers or in Matrigel-coated chambers. Migration and invasion were quantified after staining with crystal violet. The experiments were repeated thrice. Scale bar, 1 mm. *P < 0.05, **P < 0.01. (C) The indicated cells were plated on fibronectin, and 60 min later, cells that did not adhere to the substrate were removed. The numbers of adherent cells were determined after staining (in triplicates). *P < 0.05. (D) The indicated derivatives of MCF10A cells (1 × 105) were incubated in EGF-containing medium on plates precoated with fluorescein isothiocyanate (FITC)–labeled gelatin. After 12 hours of incubation, the cells were fixed and counterstained with rhodamine-phalloidin and DAPI. The broken line insets were magnified (×2) and presented (upper right corners). Quenched spots of gelatin were quantified in at least eight nonoverlapping fields. *P < 0.05, **P < 0.01 [one-way analysis of variance (ANOVA) with Bonferroni’s correction]. Scale bars, 10 μm. Data are means ± SEM. Each experiment was performed in duplicates.

Depletion of LAD1 is associated with slower rates of cell proliferation

Upon long-term culturing, we noted that LAD1-depleted MCF10A cells displayed relatively slower rates of growth. To quantify the difference, we examined two independently established stable lines and confirmed that knockdown associated with moderately slower cell proliferation (Fig. 5A). Immunoblotting revealed that LAD1 of MCF10A cells displayed reduced electrophoretic mobility when compared to LAD1 of the triple-negative HCC70 breast cancer cells, but no LAD1 was detected in MDA-MB-231 breast cancer cells (Fig. 5B). Next, we determined the fraction of MCF10A cells undergoing DNA replication by measuring bromodeoxyuridine (BrdU) incorporation. Whereas stimulation of control cells with EGF increased the fraction of DNA-replicating cells by almost eightfold, the corresponding fractions displayed by the two lines of LAD1-delpeted cells were significantly smaller (Fig. 5C, left panel). In line with this observation, when testing MDA-MB-231 cells, in which we overexpressed a V5 peptide–tagged LAD1, we observed moderately stronger BrdU incorporation signals, in comparison to control cells (Fig. 5C, right panel).

Fig. 5 Mammary cell proliferation is inhibited when LAD1 is depleted.

(A) MTT analysis of viability in MCF10A cells expressing shLAD1 after 4 days in culture. Data are means ± SEM of six replicates. *P < 0.05, ***P < 0.001 (one-way ANOVA with Bonferroni’s correction). (B) Extracts of the indicated breast cancer lines were probed as indicated. (C) Serum-starved MDA-MB-231 cells expressing LAD1-V5 and shLAD1 MCF10A cells were stimulated with EGF (12 hours). Bromodeoxyuridine (BrdU) was added, and 60 min later, cells were stained. DAPI- and BrdU-stained nuclei were counted (in >5 fields). Data are means ± SEM of three experiments. *P < 0.05, ***P < 0.001. (D) MCF10A cells (5 × 103) stably expressing shLAD1 (or shScramble) were grown in EGF-containing Matrigel (2 weeks), acini were photographed, and ImageJ was used to determine volumes. ***P < 0.0001. Scale bar, 100 μm. Data are from three experiments. (E) MCF10A cells were incubated with RO-3306 (10 nM; 15 hours), washed, cultured (<90 min), and stained for DAPI, LAD1 (green), and acetylated tubulin (red). (F) Cells pretreated as in (D) were fixed and probed at metaphase. Scale bar, 10 μm. (G) Cells ectopically expressing LAD1-GFP and LifeAct-cherry were pretreated as in (D). Frames taken from live cell movies are shown. Scale bar, 10 μm.

Our next set of experiments, which used three-dimensional (3D) cultures of MCF10A cells, further supported the possibility that LAD1 plays positive roles in mammary cell proliferation: When cultured in Matrigel in the presence of EGF, MCF10A cells normally form spheres that gradually lose their luminal (internal) cells (23). LAD1-depleted cells, which were cultured for 2 weeks in Matrigel, formed significantly smaller acini, relative to control shScramble cells (Fig. 5D), in line with observations made in 2D cultures.

Because LAD1 is involved in cell proliferation and colocalizes with and likely regulates actin fibers, we examined the subcellular distribution of LAD1 during mitosis, the main phase of the cell cycle regulated by filamentous actin. At the same time, as the microtubule network is remodeled to generate a bipolar spindle, cells disassemble stress fibers and replace their interphase actin cytoskeleton with a contractile mitotic actomyosin cortex that later supports cytokinesis by localizing into the cleavage furrow (24). To enrich for cells undergoing mitosis, we synchronized MCF10A cells using RO-3306, an inhibitor of the cyclin-dependent protein kinase 1, which arrests cells at G2. After removal of the inhibitor, cells were visualized at various intervals, and mitotic stages were captured. An antibody specific to acetylated tubulin marked mitotic spindles, and a fluorescent phalloidin was used to visualize actin fibers (Fig. 5E). As expected, mitotic microtubules underwent stage-specific reorganizations, including disassembly in anaphase and bundling into arrays in telophase (25). In comparison, during metaphase, LAD1 displayed cortical distribution that overlapped that of F-actin (Fig. 5F), but in anaphase, it assumed a more diffused pattern of distribution. However, during the final step of mitosis, in telophase, LAD1 exhibited a more cortical distribution, including filaments and small puncta that decorated the cleavage furrow. Movies taken while live cells underwent mitosis confirmed co-distribution of LifeAct-cherry and LAD1-GFP, as well as their cortical distribution at telophase (Fig. 5G). In conclusion, the growth-stimulatory effects of LAD1 in 2D and 3D cultures, along with the observed stage-specific reorganization, propose that this phosphoprotein has functional roles to play in cell proliferation.

LAD1 phosphorylation at specific sites regulates mammary cell proliferation and migration

Because our MS analysis of LAD1 identified multiple sites of phosphorylation and both depletion and overexpression of LAD1 altered rates of migration and proliferation of human mammary cells, we addressed potential roles for specific sites. To this end, we prepared sets of serine-to-alanine and serine-to-aspartate replacement mutants of LAD1 and stably expressed them in MDA-MB-231 cells, which naturally express very low amounts of LAD1. In addition to single serine mutations, we constructed mutants of the vicinal serines at 355 and 356 of the C-terminal region. Proliferation and survival assays, which were performed for 2, 4, and 6 days in the presence of EGF, as well as cell migration assays, which are presented in fig. S4, A and B, revealed no reciprocal relations when individual serine-to-alanine mutations were compared to the respective serine-to-aspartate forms, with the potential exception of the vicinal serines at 355 and 356. Nevertheless, replacement of two N-terminal serine residues, namely, Ser38 and Ser57, to either alanines or aspartates, moderately reduced cell growth. By contrast, the same mutants, in addition to mutations affecting position 64, enhanced rates of cell migration by 50 to 70%, similar to the corresponding aspartate mutants. The only exception was the vicinal double mutant at 355 and 356; the aspartate form stimulated migration, whereas the alanine form slightly inhibited cell migration. In conclusion, mutagenesis of specific phosphor-sites of LAD1, which were identified using MS, raised the possibility that phosphorylation of LAD1 at specific serine residues can affect cellular functions.

Proteomic and yeast two-hybrid screens identify filamins, actin–cross-linking cytoskeletal proteins, as potential ligands of LAD1

The observations we described so far indicated that LAD1 decorates fibrillar structures that partly colocalize with actin, its phosphorylated form cosediments with F-actin rather than with G-actin, and like actin, it supports both proliferation and cell migration. Because LAD1 harbors no recognizable ABD, we assumed that an adaptor protein mediates the interactions with F-actin. To test this prediction, we performed yeast two-hybrid (Y2H) screens. Full-length human LAD1 cDNA was cloned into a bait construct as an N-terminal fusion to LexA. The construct was used to screen a human lung cancer cDNA library, as previously described (26), and positive clones were selected. The respective prey fragments were amplified and sequenced. The resulting sequences were used to identify the corresponding interacting proteins, and a confidence score was assigned to each interaction (table S4). According to the data we obtained, two forms of the actin-binding proteins called filamins (FLNA and FLNB) specifically bind with LAD1, and they share immunoglobulin-like repeat domains 18 and 19 as LAD1 interaction sites.

To identify additional proteins that interact with LAD1, we applied MS. This approach used a V5 peptide–tagged version of LAD1, which was transiently expressed in HeLa cells. Whole-cell extracts were subjected to immunoprecipitation using an antibody to the V5 peptide. Proteins specifically bound to LAD1 were eluted using a V5 peptide, resolved by gel electrophoresis, and gel slices were applied to MS. This analysis identified several proteins, including nucleolin, nucleophosmin, LAD1 itself, and both FLNA and FLNB (table S4). Notably, these two forms of filamin received the highest scores when the peak areas of the MS data were calculated. Hence, we concluded that filamins, actin cross-linkers originally identified as cancer-promoting proteins (12), serve as binding partners of LAD1.

The C-terminal part of LAD1 physically interacts with FLNA in mammary cells

FLNA links actin fibers to each other, as well as to transmembrane proteins (such as specific integrins), thereby affecting cell structure and rigidity (13, 14). In line with the Y2H and the MS results, co-staining analyses revealed almost identical patterns of FLNA and LAD1 subcellular distribution. Both filamentous and punctate localizations were observed (Fig. 6A): Whereas the filamentous pattern likely reflects co-distribution of LAD1 and FLNA along actin stress fibers, the puncta may represent intermediate junctions (also called adherens junctions), whose cytoplasmic face is linked to the actin cytoskeleton through stress fibers. To verify colocalization, we applied the proximity ligation assay (PLA), which uses oligonucleotides attached to antibodies against proteins of interest, to form circular DNA strands when bound in close proximity (27). Accordingly, we applied antibodies to the endogenous forms of LAD1 and FLNA and observed strong protein-protein interactions (Fig. 6B). As expected, the amplification reactions excluded nuclei, and we observed no bright spots when using single antibodies. In line with strong interactions, the anti-V5 immunoprecipitates that we prepared from LAD1-V5–expressing cells contained readily detectable FLNA, but no signal was observed in control immunoprecipitates prepared from cells stably expressing the empty vector (Fig. 6C).

Fig. 6 LAD1 physically interacts and colocalizes with FLNA.

(A) Images of MCF10A cells probed for LAD1 (red), filamin A (FLNA; green), and nuclei (DAPI, blue). Arrowheads mark colocalization. Scale bar, 10 μm. (B) MCF10A cells were probed with antibodies recognizing LAD1 and FLNA and processed for proximity ligation assay (PLA) with a tetramethylrhodamine-5-isothiocyanate (TRITC) probe (red). Counterstaining used DAPI (blue) and phalloidin-FITC (green). Single-antibody control experiments are shown. Scale bar, 10 μm. (C) Extracts of MDA-MB-231 cells overexpressing the indicated forms of LAD1 were subjected to coimmunoprecipitation using beads conjugated to an anti-V5 antibody. The eluates were blotted as indicated, and bands were quantified in biological duplicates. Both short (short exp.; 10 s) and longer film exposure (long exp.) are shown. (D) MCF10A cells underwent transfection with siFLNA oligonucleotides. Serum-starved cells were stimulated with EGF for the indicated time intervals, and extracts were immunoblotted as indicated. The experiment was repeated thrice. (E) A model depicting the putative signaling events leading to association of LAD1 with actin treadmilling. LAD1 undergoes phosphorylation at multiple sites downstream to the MEK pathway. This augments binding of LAD1 with 14-3-3σ, a cytoskeletal solubility cofactor. Another ligand of LAD1 is FLNA, which cross-links actin filaments. By receiving extracellular signals, LAD1 might regulate actin polymerization, thereby controlling cell migration and proliferation during cancer progression.

The ABD of FLNA is located at the N terminus of the rod-like molecule characterized by 24 immunoglobulin-like repeats, of which the C-terminal repeat serves as a dimerization domain and repeats 17 to 20 apparently bind with LAD1. To map the corresponding interaction site within LAD1, we applied the above-described V5-tagged N- and C-terminal portions of LAD1 (Fig. 3F). Coimmunoprecipitation analyses indicated that the C-terminal half of LAD1, which cosediments with F-actin, is the domain that binds with FLNA (Fig. 6C). To examine potential effects of LAD1 phosphorylation, we tested a set of serine site mutations but found that no serine-to-alanine/aspartate markedly altered the amount of coprecipitated FLNA (fig. S4, C and D). In summary, our observations support existence of FLNA-LAD1 physical complexes in living cells. These are held by repeats 17 to 20 of FLNA and by the C-terminal portion of LAD1, which enables recruitment of LAD1 to cross-linked filaments of actin. Notably, analysis of point mutants of LAD1 imply that the binary LAD1-FLNA complex, and by inference, the ternary LAD1-FLNA-actin filaments, is constitutive rather than inducible by extracellular cues like EGF.

LAD1 affects signaling and transcription networks

Because it has been reported that cells treated with the hepatocyte growth factor exhibit poor invasiveness if their FLNA is depleted (28), we assumed that the uncovered FLNA-LAD1 interaction plays functional roles beyond the cytoskeleton. To examine this, we used small interfering RNAs (siRNAs) to deplete FLNA in MCF10A cells, which were later starved for 16 hours and thereafter stimulated with EGF (Fig. 6D). Although FLNA depletion minimally influenced EGF-induced degradation and phosphorylation of EGFR and AKT (Ser473), we observed an interesting effect on ERK1/2 activation: While ERK stimulation after a short treatment with EGF (10 min) was only slightly affected, a time-dependent diminution of phosphor-ERK was clearly evident in FLNA-depleted cells. To further establish a potential link between LAD1, the MEK-to-ERK (rather than the PI3K-to-AKT) pathway, and FLNA, we stably knocked down LAD1 in HCC70 breast cancer cells using two nonoverlapping shRNAs (fig. S5A). As expected, stimulation of control cells was followed by activation of ERK and AKT, along with robust phosphorylation of EGFR and modification of LAD1. However, in both lines of shLAD1-treated cells, we observed relatively weak (AKT) and strong (ERK) enhancement of activation. Paradoxically, EGFR phosphorylation signals were lower (and shorter) in LAD1-depleted cells, which raised the possibility that long-term selection of shLAD1-expressing clones resulted in compensatory hypersensitization of downstream signals.

Because the ERK pathway culminates in transcription regulation, the observed effects on ERK activation predicted an association with transcriptional program. This scenario was tested by contrasting gene expression programs instigated upon EGF-induced stimulation of HCC70 cells, before and after LAD1 depletion. RNA-seq, followed by data normalization, identified 460 genes that were differentially decreased and 363 genes that underwent increased abundance in LAD1-depleted cells (table S5). PCR was used to confirm altered abundance of several transcripts (see fig. S6). To determine which pathways were enriched, we used the gene ontology (GO) annotation analysis and used the GOrilla server while setting the false discovery rate (FDR) at 0.05 (table S6). In line with our other findings, the most affected annotations identified in LAD1-depleted cells related to cell cycle control, epithelial cell proliferation, and regulation of cell motility.

LAD1 binds with SFN

How could LAD1, along with its partner, FLNA, possibly regulate the upstream MEK to ERK pathway? One potential link has been uncovered by a previous targeted proteomic study, which identified both basal and EGF-induced physical complexes of LAD1 containing the SFN adaptor (29). SFN is induced by p53 (30), acts as a cytoskeletal solubility cofactor (31), and binds primarily to phospho-proteins, and its overexpression associates with regulation of the MEK-ERK pathway. As expected, coimmunoprecipitation analyses confirmed formation of physical complexes containing LAD1 and SFN in MDA-MB-231 cells (fig. S5B), but due to abundant and diffuse localization of SFN, we were unable to unequivocally determine colocalization in MCF10A cells, even after stimulation with EGF (fig. S5C). To circumvent this, we applied PLA and observed strong interactions between LAD1 and SFN (fig. S5D). When actin filaments were disrupted with latrunculin A, we observed no effect on PLA signals (fig. S5D). Nevertheless, treatment with phalloidin, which stabilizes F-actin, significantly reduced the PLA signal, implying that actin polymerization dissociates the LAD1-SFN complex. In conclusion, the uncovered interactions between FLNA and LAD1 may be critical not only for the regulation of G- to F-actin ratio but also for sustaining EGF-induced activation of ERK. Physical recruitment of SFN to LAD1-FLNA complexes may disrupt their interactions with F-actin. Figure 6E schematically summarizes our mechanistic studies: While ligand-induced activation of EGFR stimulates both the PI3K-AKT and the MEK-ERK pathway, the latter is involved in phosphorylation of LAD1 on multiple residues. These events may enhance preexisting complexes comprising LAD1 and SFN. As an outcome, the LAD1-SFN complex may differentially associate with cross-linked actin filaments. Notably, a complex comprising SFN, soluble actin, and cytokeratin has been reported in basal-like breast cancer (31), but functional relations between the two complexes of actin remain unknown.

LAD1 depletion reduces tumorigenicity of breast cancer cells in an animal model and high abundance of LAD1 associates with poor prognosis of breast cancer patients

Previous studies established strong associations between FLNA and cancer. For example, overexpression of FLNA in breast cancer was linked to lymph node metastasis and vascular invasion (32). In the same vein, targeting FLNA inhibited RAS-driven lung adenocarcinomas in a murine model (33). These observations prompted us to examine potential oncogenic functions of the FLNA partner, LAD1. To this end, we used the aforementioned LAD1-depleted HCC70 cells, which are basal type A human breast cancer cells harboring mutations in p53 and phosphatase and tensin homolog (PTEN) (34). The strong effects of two shRNAs on LAD1 abundance, which we verified using PCR and immunoblotting (Fig. 7A), permitted further studies. As expected, the two LAD1-depleted lines displayed moderately slower rates of DNA synthesis, as determined by a BrdU incorporation assay (Fig. 7B). We then implanted cells under the skin of immunocompromised mice and monitored tumor growth at the site of implantation. LAD1-depleted tumors had significantly slower rates of growth (Fig. 7C) and greater abundance of protein markers of apoptosis and growth arrest (Fig. 7D), consistent with several in vitro lines of studies, all supporting the ability of LAD1 to support cell proliferation.

Fig. 7 LAD1 depletion retards basal-like breast tumors in animals, and high LAD1 associates with poor patient prognosis.

(A) HCC70 cells stably expressing shLAD1 were analyzed using immunoblotting and polymerase chain reaction (PCR). PCR signals were quantified relative to β2-microglobulin. Blots are representative of three experiments. (B) HCC70 cells were incubated in the presence of BrdU (60 min). BrdU- and DAPI-stained nuclei were counted in several fields. Shown are histograms. Data are means ± SEM from three experiments. *P < 0.05, ***P < 0.001 by one-way ANOVA with Tukey’s correction. (C) HCC70 cells (3.5 × 106) expressing shLAD1 were implanted in six CD1/nude mice, and tumorigenic growth was monitored. Data are mean tumor volumes (±SEM). Growth rate effects of shLAD1.1 versus shLAD1.2 compared to shScramble were P < 0.05 and P < 0.01, respectively, by two-way ANOVA of repeated-measures with Dunnett’s test. (D) Extracts prepared from the xenografts presented in (C) were analyzed using immunoblotting as indicated. Note that two tumors were analyzed per group. (E) Analysis of LAD1 in the 10 integrative subtypes of breast cancer (n = 1980). (F) Patients were divided into three groups: 15% upper and lower LAD1 mRNA abundance and the rest. Breast cancer–related deaths are shown (brackets), along with survival curves (P < 0.001).

Next, we used a large clinical data set, called METABRIC (3538), which includes data from ~2000 breast cancer patients, who were followed for >20 years from initial diagnosis. Similar to the abundance of FLNA and SFN, high expression of LAD1 transcripts was associated with higher tumor grade; however, we detected no association of LAD1 transcript abundance and the respective abundance of its physical partners (fig. S7, A and B). In addition, because METABRIC contains only a few advanced stage patients, we could not examine relations between LAD1 transcript abundance and tumor stage. Further analysis indicated that LAD1 expression was especially high in relatively aggressive molecular subtypes of breast cancer, namely, basal-like and HER2-positive tumors (Fig. 7E): Surveying the 10 integrative clusters, which are based on combined genomic and transcriptomic landscapes (39), revealed that, in similarity to that of FLNA and SFN, LAD1 transcript abundance was highest in integrative cluster 10, which incorporates mostly triple-negative tumors (fig. S7C). Likewise, LAD1 transcript abundance was relatively high in integrative cluster 5, which identifies almost all cases with ERBB2/HER2 amplification. The full cohort of patients was then divided into three groups according to LAD1 transcript abundance: upper (15%), lower (15%), and intermediate expression (70%). Survival analyses using Cox multivariable regression and Kaplan-Meier (KM) curves revealed that LAD1 overexpression is independently and significantly associated with shorter disease-free survival [Cox model: hazard ratio, 1.69350 (95% confidence interval, 1.29140, 2.22080); log-rank P < 0.001; Fig. 7F]. Similarly, analyses of FLNA and SFN attributed poor prognosis to high abundance of the latter, but FLNA amounts did not reach statistical significance (Fig. 7F and fig. S7D).

In summary, the in vitro studies, animal model analysis, and the clinical data we presented are in line with LAD1 acting as a phosphorylation substrate downstream of EGFR. Further, this newly discovered partner of FLNA and F-actin appears to be involved in proliferation and migration of mammary cells. These in vitro functions are reflected in poor prognosis of specific subtypes of breast cancer. Conceivably, alternative high-order complexes containing LAD1, FLNA, fibrillar actin, and SFN regulate cytoskeletal rearrangements, as well as signaling pathways and transcription networks needed for robust proliferation and migration of breast cancer cells.

DISCUSSION

Growth factors acting as mitogens, survival factors, motility inducers, or angiogenesis promoters play critical roles in cancer progression (40, 41). Specifically, by activating the RTKs of the HER/ERBB group, growth factors of the EGF family instigate cascades of phosphorylation events. The relevance of these cascades to tumor initiation and progression is exemplified by the ability of oncogenic mutations to generate persistent growth factor signals by mimicking specific steps of the downstream signaling pathway (42). Here, we aimed at uncovering yet unknown phosphorylation effectors, which are distal to EGFR activation. For this, we applied the SILAC protocol (43) on human mammary cells and compared two stimuli: EGF, which induces proliferation and migration of mammary cells, and serum, which instigates proliferation but no migration. The observed patterns of phosphorylation and dephosphorylation were remarkably similar. Moreover, in line with a previous report (7), we found that a few phosphorylation and dephosphorylation targets contained several sites, which displayed different kinetics of phosphorylation. These observations raise the intriguing possibility that a biological outcome, such as cell migration or proliferation, might be specifically encoded by the collective action of protein modifications subtly differing in temporal or quantitative terms.

LAD1 emerges from our study as an interesting example of the subtle differences model: Both EGF and serum factors modified (with different kinetics) similar sets of serine and threonine residues. Only limited information was available on LAD1 before our investigation, and no protein partners of LAD1 were previously reported. Motoki et al. (17), who cloned the murine LAD1 gene, proposed that it acts as a secreted component of the basement membranes. However, our analyses and a recent report (44) indicated an intracellular function. In keeping with the observations we made in breast cancer, it was reported that LAD1 is one of the most highly overexpressed genes in ovarian serous papillary carcinomas, compared to the normal ovarian epithelium (45).

How exactly LAD1 phosphorylation is translated to regulation of cell migration and proliferation likely relates to the complex control of actin polymerization and cross-linking. Thus, our PLA assays proposed that SFN, which binds with phosphorylated LAD1, favors depolymerization of F-actin (31), thereby engaging LAD1 in actin treadmilling (model in Fig. 6E and fig. S5D). Both MS analyses and Y2H screens identified another important interaction, which entails physical binding with filamin family proteins. FLNA acts as a large scaffolding protein that uses the N terminus to bind with actin and the C-terminal part to bind with >80 known protein ligands, such as integrin β and the small GTPase RalA (46). Perhaps the best characterized function of FLNA is to cross-link actin filaments. Hence, filamin-mediated recruitment of LAD1 to specific bundles of actin might underlay localization of LAD1 to cortical regions, adherens junctions, and stress fibers, as well as explain the effects of LAD1 on mammary cell proliferation, adhesion, and migration. Because our extensive mutagenesis of LAD1’s serine phosphorylation sites failed identifying an essential site, we assume that the LAD1-FLNA complex is constitutive rather than EGF-inducible. By contrast, although our results and a previous report (29) indicated that SFN binds with LAD1 before EGF stimulation, it appeared that complex formation is enhanced after stimulation. Whether LAD1 phosphorylation, downstream to MEK activation, controls association of the presumably constitutive FLNA--LAD1 complex with actin bundles remains unknown.

Several lines of new evidence associate LAD1 with oncogenic phenotypes of human mammary cells. First, as we demonstrated herein, LAD1 acts downstream to a signaling pathway often engaged in cancer, namely, the EGFR-to-ERK cascade of protein kinases. Second, high LAD1’s transcript abundance correlates with shorter survival of breast cancer patients and is characteristic to two relatively aggressive molecular subtypes of this disease, HER2-positive and basal-like tumors. Third, our animal studies attributed a tumorigenic activity to LAD1, and the in vitro assays implied that LAD1 confers to mammary cells accelerated proliferation and ability to migrate across tissue barriers—attributes shared by metastatic breast tumors. Previous lines of evidence associating LAD1 with cancer are consistent with our results. Thus, it was previously proposed that LAD1 is included in a short list of markers of basal-like tumors (47). Similarly, a recent study that explored gene expression, copy number, and DNA methylation in triple-negative breast cancer identified a group of hyperactivated genes, including LAD1 (48). Another interesting study addressed thyroid carcinoma and identified seven BRAF-induced genes, including LAD1, that are specific to tumors expressing the BRAF V600E mutation (49). Yet, another study, which addressed laryngeal squamous cell carcinoma, observed up-regulation of LAD1, along with several actin-binding proteins (50). Together, LAD1 emerges from our study and from previous reports as a poor prognosis marker and a potential supporter of transformed phenotypes. Better understanding of the interactions among LAD1, filamins, and the actin cytoskeleton will likely shed light on yet-unknown modes of cancer cell regulation. An especially challenging endeavor will be the identification of functional domains within LAD1, which is unstructured and contains no recognizable functional domains.

MATERIALS AND METHODS

Materials and cell lines

The following antibodies were used: antibodies to LAD1 (HPA028732) from Sigma-Aldrich or from Thermo Fisher Scientific (PA5-2244); ERK2, AKT2, and RAS-GAP from Santa Cruz Biotechnology; FLNA and histone H3 from Abcam; and EGFR, PDCD4, RNAPII, and phosphorylated AKT from Cell Signaling Technology. DAPI (4′,6-diamidino-2-phenylindole) and conjugated phalloidin were from Invitrogen. Stable isotope-labeled arginine and lysine were from Cambridge Isotope Laboratories. Sequences of all primers and shRNAs are listed in table S7. Cell lines were from the American Type Culture Collection. All cell lines were tested for mycoplasma. MCF10A cells were cultured as previously described (51). Where indicated, cells were starved overnight for serum and thereafter stimulated with EGF (10 ng/ml) or with horse serum (5%). The following inhibitors were administered 30 min before cell stimulation: AG1478 (10 μM), PD184352 (10 μM), U0126 (5 μM), wortmannin (200 nM), and RO-3306 (10 nM). For 3D cultures, acini were left to grow for 14 days in a medium containing Matrigel (3%), fixed in methanol-acetone, and processed for microscopy.

Transfection of plasmids and siRNA oligonucleotides

siRNA transfections used ON-Target SMART oligonucleotides (Dharmacon). pENTER221-LAD1 and pENTER221-SFN were a gift from S. Wiemann [Deutsches Krebsforschungszentrum (DKFZ)]. LAD1 domains were cloned into D-TOPO (Invitrogen). The corresponding V5 peptide–tagged forms of LAD1 were cloned into pLenti6 (Invitrogen), and a GFP-tagged form was cloned into pLVU (Addgene clone #24177). Point mutations were generated using PCR and megaprimers (table S7) (52) and verified by sequencing. V5 peptide–tagged SFN was cloned into pLEX307 (Addgene clone #41392). Viral particles encoding LAD1, SFN, pTk93 Lifeact-mCherry (a gift from I. Cheeseman, Whitehead Institute, Cambridge; Addgene plasmid #46357), and shRNAs targeting LAD1 were produced in human embryonic kidney (HEK)–293FT cells.

Purification and quantification of mRNA

Isolation of mRNA was carried out using the PerfectPure RNA Cultured Cells kit (from 5-Prime). Real-time quantitative PCR analysis was performed using Fast SYBR Green Master Mix (from Applied Biosystems).

Cell fractionation, BrdU incorporation assays, and immunoblotting

Extraction of cells, immunoblotting, BrdU incorporation, and MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assays were performed as described (51). For MTT, cells (2 × 103) were seeded in a 96-well plate, in triplicates, and grown for 4 days in full medium, unless indicated otherwise. For BrdU assays, serum-starved cells were incubated with EGF for 12 hours, followed by a 60-min pulse of BrdU, fixation in PFA (paraformaldehyde) (3%), and processing for imaging. For subcellular fractionation, we used the ProteoExtract Subcellular Proteome Extraction kit (Calbiochem). Membrane proteins were isolated using the Cell Surface Isolation kit (Pierce, Thermo Fisher Scientific). For fractionation of soluble and filamentous actin in live cells, we applied the G-Actin/F-actin In Vivo Assay Biochem Kit (Cytoskeleton Inc.).

Mass spectrometry

The EGFR pathway initially engages well-characterized tyrosine phosphorylated substrates, which are later replaced by less understood sets of serine- and threonine-modified substrates, which were the targets of our MS analyses. MCF10A cells were grown in labeling media for >10 cell divisions. Serum was predialyzed and later mixed with a lysine- and an arginine-free base medium. Labeled lysine and arginine were added to form the complete labeling medium. Peptide isolation after cellular extraction was performed using equal amounts of extracts from heavy-, medium-, and light-isotope cultures, which were mixed in lysis buffer [8 M urea, 75 mM NaCl, 50 mM tris (pH 8.2), a protease inhibitor mixture, 1 mM NaF, 1 mM β-glycerophosphate, 1 mM sodium orthovandate, 10 mM sodium pyrophosphate, and 1 mM phenylmethylsulfonyl fluoride]. The samples were sonicated, and proteins were reduced for 30 min at 60°C with 2.8 mM dithiothreitol. Thereafter, lysates were treated for 30 min with 8.8 mM iodoacetamide in 100 mM ammonium bicarbonate and digested overnight at 37°C (in 2 M urea and 25 mM ammonium bicarbonate) using a modified trypsin from Promega. A second digestion was performed for 4 hours. For peptide desalting, we used C18 tips. The peptides were dried and resuspended in 0.1% formic acid and then separated using a strong cation exchange (SCX) column. Metal affinity was used to enrich fractions for phosphopeptides, and this was followed by a chromatography step that used a titanium dioxide (TiO2) column. The phospho-enriched peptides were resolved by means of reverse-phase chromatography on 0.075-mm × 200-mm fused silica capillaries packed with a Reprosil reverse-phase resin (Dr. Maisch HPLC GmbH). The peptides were eluted using linear acetonitrile gradients (7 to 40%; 214 min) and 95% acetonitrile with 0.1% formic acid in water (8 min) at a flow rate of 0.25 μl/min. MS used an ion-trap mass spectrometer (Orbitrap, Thermo Fisher Scientific) in the positive mode, followed by collision-induced dissociation of the most dominant ions selected from the first MS scan. The MS data were analyzed using MaxQuant (53) and Perseus (54), along with the human International Protein Index (IPI) database.

Analyses of protein-protein interactions

Proteins were immunoprecipitated from cleared cell lysates using beads conjugated to an antibody against the V5 peptide. After 2 hours of incubation at 4°C, complexes were washed three times and incubated for 10 min in an equal volume of a V5 peptide–containing solution (1 mg/ml). Eluates were subjected to electrophoresis and immunoblotting. For MS analysis, lanes of acrylamide gels corresponding to LAD1 coimmunoprecipitation were dissected, and trapped proteins were digested using trypsin and later analyzed using the Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific).

Y2H screens

Y2H screens were outsourced to Hybrigenics Services. The coding sequence for human full-length LAD1 was PCR-amplified and cloned into pB29 as an N-terminal fusion to LexA. The construct was used as a bait to screen a random-primed human lung cancer cDNA library. Overall, 91.1 million interactions were screened as previously described (26). One hundred twenty-two His+ colonies were selected in a medium lacking tryptophan, leucine, and histidine. The prey fragments of the positive clones were amplified and sequenced at their 5′ and 3′ junctions. The resulting sequences were used to identify the corresponding interacting proteins in the GenBank database (National Center for Biotechnology Information). A confidence score was attributed to each interaction, as described previously (55).

Analyses of histone acetylation and RNAPII pausing and traveling

Cell fixation, chromatin immunoprecipitation (ChIP), and nucleotide sequencing were performed as described previously (56). ChIP, using antibodies to acetylated histone 3 and RNAPII, was followed by DNA sequencing (ChIP-seq). Reads were mapped to the human genome assembly using Bowtie2 (57). Histone acetylation (H3K27Ac) and RNAPII signals for the genome browser views were calculated by approximation of ChIP fragment density at each position in the genome using “makeUCSCfile” of HOMER (http://homer.ucsd.edu/homer/ngs/ucsc.html).

RNA-seq analysis

RNA libraries of biological duplicates were prepared according to a standard protocol and subjected to nucleotide sequencing. Sixty–base pair–long, single-end, RNA-seq reads were aligned to the human genome with Spliced Transcripts Alignment to a Reference (STAR) (58). Reads per gene were counted using the summarizeOverlaps function from the GenomicAlignments package in R with the RefSeq transcriptome as reference. Reads per gene were normalized across samples using the DESeq2 package (table S5). GOrilla was used for functional clustering of genes that exhibited a 1.5-fold change (and P < 0.05) in shLAD1-treated cells, relative to the control, selecting clusters with FDR < 0.05 and fold change greater than 0.5 (on a log2 scale) (59).

Immunofluorescence and PLAs

Cells were grown on fibronectin-coated glass coverslips and fixed in PFA (3%) and Triton X-100 (0.05%) or in methanol (FLNA only). All primary antibodies were incubated with cells for 2 hours. After three washes with saline, cells were incubated for 60 min at room temperature with DAPI, phalloidin, and secondary antibodies. For PLA, cells were hybridized with primary antibodies followed by secondary antibodies against Rabbit PLUS (DUO92002) and against Mouse MINUS (DUO92002) and processed using the Duolink In Situ Detection Kit (red) containing a tetramethylrhodamine-5-isothiocyanate probe (Sigma-Aldrich). Thereafter, coverslips were washed and placed, cells face down, onto drops of an anti-fade reagent (from Dako). Samples were examined using a spinning disk confocal microscope (Zeiss).

Cell migration, invasion, adhesion, and invadopodia formation assays

Migration and invasion assays, as well as invadopodia formation assays, were performed as we previously described (21, 51).

Animal studies

All animal experiments were approved by the Weizmann Institute’s Animal Care and Use Committee. HCC70 cells (3.5 × 106) stably expressing shRNAs were inoculated subcutaneously in the right leg of nine female nude mice (6 weeks old). Tumor growth was monitored once every 3 days with a caliper to measure width (W) and length (L), and tumor volume was calculated according to the formula: 3.14 × (W2 × L)/6.

Analyses of clinical data sets

LAD1 expression was analyzed using the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) data set (1980 breast cancer patients) (3638). We performed KM estimates for three groups of patients split according to the 15 and 85% percentiles of expression and tested the null hypothesis of having the same survival function with the log-rank test. To check that this effect was not produced by a correlation of expression values with well-known prognostic clinical variables, we fitted a multiple Cox model stratified by number of positive lymph nodes, tumor size, and tumor grade (modeled linearly). LAD1 expression was included as a continuous variable, although we summarized the hazard ratio estimate as the risk ratio between the 85 and 15% quartiles of expression. Survival models were fitted using the survival and the rms packages in R.

Statistical analyses

Analysis of in vitro assays, which were performed in triplicates, used one-way analysis of variance (ANOVA) with Tukey’s corrections. Analysis of xenograft results used repeated-measures two-way ANOVA with Bonferroni’s correction and two-sided Dunnett’s test to compare the repeated measures to the control.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/515/eaan0949/DC1

Fig. S1. LAD1 is an intracellular protein that undergoes MEK-dependent phosphorylation inresponse to EGF.

Fig. S2. Genomic, proteomic, and transcriptional characteristics of LAD1.

Fig. S3. Knockdown of LAD1 impairs invadopodia formation by mammary cells.

Fig. S4. Phosphorylation of LAD1 on specific serine sites moderately regulates cell migration and viability, without markedly affecting FLNA binding.

Fig. S5. LAD1 increases EGF-inducible ERK phosphorylation and physically interacts with14-3-3σ (SFN).

Fig. S6. Effects of LAD1 on specific transcripts of breast cancer cells.

Fig. S7. Like LAD1, SFN expression correlates with poor prognosis in breast cancer patients;FLNA expression displays distinct pathological patterns.

Table S1. SILAC data.

Table S2. EGF-induced phosphorylation changes.

Table S3. Serum-induced phosphorylation changes.

Table S4. Lists of putative partners of LAD1 revealed by using either Y2H screens or proteomicanalyses of coimmunoprecipitataed proteins.

Table S5. RNA-seq data.

Table S6. GOrilla analysis.

Table S7. Primers.

REFERENCES AND NOTES

Acknowledgments: We thank A. Bershasdsky, B. Geiger (Weizmann Institute), and S. Wiemann (DKFZ, Heidelberg) for helpful advice, as well as R. Rotkopf (Life Sciences Core Facilities, Weizmann Institute), M. Gomaa (Department of Pathology, Hadassah Medical Center), and all the members of our laboratory. pENTER221-LAD1 and pENTER221-SFN were gifts from S. Wiemann (DKFZ, Heidelberg). pTk93 Lifeact-mCherry was a gift from I. Cheeseman (Whitehead Institute, Cambridge; Addgene plasmid #46357). pLVU/GFP was a gift from L. Ittner (University of New South Wales, Sydney; Addgene plasmid #24177). pLEX307 was a gift from D. Root (Broad Institute, Cambridge; Addgene plasmid #41392). Funding: Our research was supported by the Israel Science Foundation, the European Research Council, the German-Israeli Project Cooperation (DIP; Deutsche Israelische projektkooperation), the Israel Cancer Research Fund, and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. Y.Y. is the incumbent of the Harold and Zelda Goldenberg Professorial Chair. Author contributions: Conception and design of the study, as well as writing the paper: L.R., G.T., G.P., and Y.Y. Acquisition of data: L.R., S.S., M. Lauriola, G.T., G.P., D.G., O.H., E.P., M. Lindzen, M.M., Y.E., S.L., A.N., N.N., and T.Z. Analysis and interpretation of data: L.R., A.S.-C., E.P., M.S., O.M.R., A.A., C.C., and Y.Y. Supervision of research and analyses: Y.Y., A.A., and C.C. Competing interests: The authors declare that they have no competing interests. Data and materials availability: RNA sequence data were submitted to Gene Expression Omnibus (submission GSE103531) and are currently available at: www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=uxidoquknlejbyd&acc=GSE103531). The proteomic (SILAC) data are available through ProteomeXchange with the following identifier: PXD008100.
View Abstract

Navigate This Article