Research ArticleCancer

Memo Is a Copper-Dependent Redox Protein with an Essential Role in Migration and Metastasis

See allHide authors and affiliations

Science Signaling  10 Jun 2014:
Vol. 7, Issue 329, pp. ra56
DOI: 10.1126/scisignal.2004870

This article has a correction. Please see:


Memo is an evolutionarily conserved protein with a critical role in cell motility. We found that Memo was required for migration and invasion of breast cancer cells in vitro and spontaneous lung metastasis from breast cancer cell xenografts in vivo. Biochemical assays revealed that Memo is a copper-dependent redox enzyme that promoted a more oxidized intracellular milieu and stimulated the production of reactive oxygen species (ROS) in cellular structures involved in migration. Memo was also required for the sustained production of the ROS O2 by NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase 1 (NOX1) in breast cancer cells. Memo abundance was increased in >40% of the primary breast tumors tested, was correlated with clinical parameters of aggressive disease, and was an independent prognostic factor of early distant metastasis.


Most breast cancer patients die from metastases, making it crucial to identify proteins and signaling pathways involved in tumor cell dissemination. Our group has previously identified Memo as being essential for breast cancer cell motility in response to several receptor tyrosine kinases. Tumor cells with Memo knockdown fail to migrate in response to epidermal growth factor (EGF), heregulin (HRG), fibroblast growth factor, or serum (14). Memo also interacts with insulin receptor substrate 1 and insulin-like growth factor receptor 1 (5, 6). Thus, Memo links extracellular signals to intracellular responses through multiple receptors.

Memo is evolutionarily conserved, with homologs found in all branches of life (7, 8). The 2.1-Å crystal structure of Memo shows that it is structurally homologous with iron- and zinc-binding enzymes, but has not provided evidence for metal binding (7). Here, we present evidence that Memo is a metal-binding enzyme that required Cu(II) for its oxidase activity. Reactive oxygen species (ROS) function as intracellular second messengers that regulate physiological processes by controlling the activity of redox-regulated proteins including transcription factors, kinases, and phosphatases (912). We showed that Memo affected localized ROS production, cell migration, and in vivo metastasis. NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidases (NOXs) are major ROS-generating complexes in cells. We present results showing that Memo was required for sustained O2 production in response to NOX activation. Moreover, we showed that increased Memo abundance was present in >40% of primary human breast tumors and correlated with poor prognostic factors and distant recurrence-free survival.


Memo is required for cell motility and invasion

Memo knockdown decreases directed motility of breast cancer cell lines in Transwell and Dunn chamber assays (1, 2, 13). We examined the effect of Memo knockdown using wound closure and invasion assays. We established MDA-MB-231 breast cancer cell lines that expressed a control short hairpin RNA (shRNA) line (shLZ) or one of two independent Memo shRNAs (sh2 or sh5); both shRNAs efficiently decreased Memo abundance (Fig. 1A, lower panel). Loss of Memo significantly impaired the cellular migration of both of the knockdown cell lines in wound closure assays (Fig. 1A) and also in SKBR3 and T47D lines expressing sh5 (fig. S1A). The ability of MDA-MB-231 tumor cells to invade through Matrigel was also significantly reduced by Memo knockdown by sh5 (Fig. 1B), which reduced Memo to a greater extent than sh2 (Fig. 1B, lower panel). Restoration of Memo abundance by expressing a nontargetable (NT) Myc-tagged Memo in sh5 cells restored invasiveness to control amounts, confirming the specificity of the shRNA knockdown for Memo (Fig. 1B). We further analyzed the invasion phenotype using MCF10A-ErbB2/ErbB3 cells, which have increased abundance of the EGF receptors ErbB2 and ErbB3 and which form hyperproliferative, invasive structures and display loss of polarity when grown in three-dimensional (3D) culture (14). We generated MCF10A-ErbB2/ErbB3 cell lines in which Memo knockdown was induced by the addition of doxycycline. Doxycycline-induced knockdown of Memo in these cells did not affect proliferation (Ki67 staining, Fig. 1C), but led to a significant reduction in the number of invasive structures and a concomitant increase in the number of polarized structures, as defined by the basal marker laminin V and the apical Golgi marker GM130 staining, respectively (Fig. 1C and fig. S1B). Together, these data confirmed that Memo is important for breast tumor cell migration and invasion. Additionally, increased Memo abundance alone did not appear to be transforming because Memo overexpression did not change the invasion, proliferation, or luminal filling of MCF10A or MCF10A-ErbB2 cells grown in 3D culture (fig. S1C). Memo overexpression in MCF10A and human embryonic kidney (HEK) 293 cells, however, increased migration in wound closure assays (fig. S1D).

Fig. 1 Memo is needed for migration and invasion in vitro.

(A) Relative wound closure of MDA-MB-231 control shLZ and Memo knockdown (KD) sh2 and sh5 cells, grown in full medium. Data represent means ± SD of six fields. Western blot analysis of Memo abundance is shown below. (B) Invasion of MDA-MB-231 shLZ control, Memo KD sh2 and sh5 cells, and sh5 nontargeting Memo-reconstituted cells (sh5-NT), through Matrigel-coated Boyden chambers. Data represent means ± SD (n = 4 to 8 chambers per condition). Western blot analysis of Memo abundance is shown below. (C) Confocal images of MCF10A-ErbB2/ErbB3 cells with inducible expression of shLZ and sh5 grown in 3D culture in the presence of doxycycline (dox). Images are representative of three independent experiments. The percentages of Ki67-positive cells, invasive structures, and polarized structures were quantified. Data represent average percentages ± SEM (n = 5 wells per condition). Western blot analysis of Memo abundance is shown below.

Memo is required for breast cancer metastasis

To test the in vivo role of Memo in tumor progression, we used mice that received orthotopic injections of MDA-MB-231 cells, a model that represents the poor-outcome, triple-negative (TN) subtype and is one of the few xenograft models that spontaneously metastasize in mice (15). The parental line, a control shRNA (shLZ) line, and the two lines expressing independent Memo shRNAs (sh2 and sh5) were injected into mice. Control and Memo knockdown tumors grew with similar kinetics to similar volumes (Fig. 2A), Memo knockdown was retained in the tumors (fig. S2A), and immunohistochemistry did not reveal significant differences in angiogenesis or proliferation between the control and sh5 groups (fig. S2B). Our in vitro analyses with other Memo knockdown and reconstituted tumor cell lines also did not reveal changes in cellular proliferation (fig. S2C).

Fig. 2 Memo is required for invasion and tumor cell migration in vivo.

(A) Orthotopic tumor growth of MDA-MB-231 control (parental and shLZ) and Memo KD (sh2 and sh5) cells in mice. Curves are the mean tumor volumes ± SEM (n = 8 mice per group). Inset: Western blot analysis of Memo abundance in injected cells. (B) Relative number of spontaneous lung metastases per unit lung area, normalized to final tumor mass (metastatic index), in the animals described in (A). Data are the means ± SEM (n = 8 mice per group). For statistical analyses, each shMemo group was compared to the combined parental and shLZ control groups. (C) Average number of tail vein–derived lung surface metastases ± SEM 4 weeks after injection (n = 6 to 10 mice per group). (D) Intravital time-lapse imaging of orthotopic tumors from GFP- and shLZ- or sh5-expressing MDA-MB-231 cells. The average numbers of cells extending protrusions and motile cells were determined per z-stack. Data are the means ± SEM of seven to eight separate z-stacks (n = 3 mice per group). Inset: Western blot analysis of Memo abundance in injected cells. Lower panels: One motile shLZ-expressing tumor cell is outlined, and frames are 4 min apart; one protruding sh5 tumor cell is outlined, and frames are at 0, 8, 22, and 28 min.

To quantify metastases, we stained lung sections of tumor-bearing animals for human vimentin. Lesion numbers were lower in mice with tumors formed from cells with sh5-mediated Memo knockdown than in those with tumors formed from the parental cell line or cells expressing shLZ (Fig. 2B). To examine whether Memo knockdown affected the proliferation of lung metastases, we measured the area of each individual lung lesion. The median lung lesion size was smaller in sh5 knockdown mice, but not sh2 knockdown mice, compared to control groups (fig. S2D). Memo abundance was lower in sh5 knockdown cells than in sh2 cells (fig. S2A), suggesting that reduced Memo abundance can attenuate the proliferative ability of the tumor cells in the lung environment.

To monitor the ability of the breast cancer cells to extravasate, we injected the control shLZ and two Memo knockdown cell lines into tail veins. The number of experimental lung metastases was significantly decreased for both Memo knockdown cell lines (Fig. 2C). To analyze the role of Memo in the early stages of metastasis, we used intravital imaging to look at the periphery of green fluorescent protein (GFP)–expressing shLZ and Memo knockdown sh5 primary tumors. Compared to controls, Memo knockdown tumors had more cells extending protrusions, but fewer motile cells (Fig. 2D and movies S1 and S2). Together, these data suggest that the decreased number of spontaneous lung metastases in mice with Memo knockdown tumors is due at least partially to a defect in intravasation from the primary site, as well as a decreased potential to extravasate into the lungs. The results show that Memo has an essential role in metastatic dissemination.

Reconstitution of Memo knockdown cells with NT Memo rescues the in vivo phenotype

We next tested the in vivo effect of reconstituting MDA-MB-231 cells with sh5-mediated Memo knockdown with Myc-tagged NT Memo rescue vector (sh5-NT). Tumors formed from GFP-expressing shLZ control, sh5, and sh5-NT cell lines (Fig. 3A) grew with similar kinetics to similar volumes (fig. S3A), and Memo knockdown or NT Memo abundance was retained in the tumors (fig. S3B). Intravital imaging of the periphery of sh5 primary tumors indicated that these tumors had more cells extending protrusions and fewer motile cells (as observed in Figs. 2D and 3B). The reconstituted sh5-NT tumors had similar numbers of protruding and motile cells as the control shLZ tumors (Fig. 3B and movie S3). Likewise, reconstitution of Memo in sh5 tumors caused a significant increase in lung metastases (Fig. 3C). The three groups had similar numbers of circulating tumor cells, likely due to large intragroup variation (fig. S3C). Nonetheless, these results suggest that defects in metastatic dissemination to the lungs were specific to Memo loss because tumors with reexpression of Memo shared many of the characteristics of control tumors.

Fig. 3 Reconstitution of Memo expression rescues tumor cell migration and invasion in vivo.

(A) Western blot analysis of Memo abundance in control shLZ, Memo sh5 KD, and sh5 nontargeting Memo reconstituted (sh5-NT) MDA-MB-231 cells. (B) Intravital time-lapse imaging of orthotopic tumors from GFP-expressing MDA-MB-231 shLZ, sh5, and sh5-NT cells. The average numbers of cells extending protrusions (left graph) and motile cells (right graph) were determined per z-stack. Data are the means ± SEM of seven to eight separate z-stacks (n = 3 to 5 mice per group). (C) Relative number of spontaneous lung metastases per unit lung area, normalized to final tumor mass (metastatic index), from mice with orthotopic tumors derived from shLZ-, sh5-, or sh5-NT–expressing MDA-MB-231 cells. Data are the means ± SEM (n = 6 to 15 mice per group).

Memo is a copper-dependent redox enzyme

Although the role of Memo in cell migration is well documented (1, 2, 4, 13), nothing is known about its biochemical function. Memo is structurally homologous to a class of bacterial nonheme, iron-dependent dioxygenases (7), but these enzymes [for example, protocatechuate dioxygenase (PD)] hydrolyze catechols, and Memo lacks this activity (fig. S4A). However, Memo has a putative metal-binding pocket consisting of three His, one Asp, and one Cys that resembles those of metal-dependent redox enzymes (Fig. 4A). To test whether Memo has redox activity, we measured its ability to generate O2, an intermediate product of many metal-catalyzed redox reactions. Of eight metal ions tested, Memo generated O2 only in the presence of Cu(II) (fig. S4B). Indeed, Memo has oxygen-dependent, copper-reducing activity (Fig. 4B), which the dioxygenase PD lacks (fig. S4C).

Fig. 4 Memo has copper-dependent redox activity.

(A) Structural model of Memo showing the putative active-site pocket and associated residues (7). Histidine residues presumed to be required for copper binding are shown in red. (B) Cu(II) reduction assay, measured as micromolar Cu(II)-reducing equivalents (CRE) per nanomole purified protein, in normoxic and hypoxic conditions. (C) Cu(II) reduction assay with Memo mutants, measured as CRE normalized to wild type (WT) (normoxic conditions). (D) O2 production by WT or H192A Memo in the presence or absence of 50 μM CuCl2 and 4 U of SOD, measured as relative light units (RLU). (E) O2 production by WT or H192A Memo, preloaded with the indicated metal or assayed in the presence of Cu(II). (F) Enzymatic properties of WT and H192A Memo during the generation of O2 using CuCl2 as a substrate (Km, substrate affinity; Kcat, substrate turnover; Kcat/Km, enzymatic efficiency). Kcat was determined from fig. S4G and is the maximum number of Cu(II) molecules converted per 50 nmol protein per 10 s. (G) Quantitative MS analysis of the oxidation state of four cysteines in the indicated RhoA peptides. Data represent means ± SD of four independent experiments.

We analyzed the Cu(II)-reducing ability of forms of Memo with mutations in the amino acid residues in the putative active site (Fig. 4A). The three histidine mutants had the lowest activities both as recombinant proteins (Fig. 4C) and when expressed in HEK293T cells as Myc-tagged proteins (fig. S4, D and E). The H192A mutant produced significantly less O2 in the presence of Cu(II) than wild-type Memo; the production of O2 by both proteins was abolished upon addition of superoxide dismutase (SOD) (Fig. 4D).

We next asked whether Cu(II) was stably or transiently present in the active site by measuring O2 production after preloading Memo with Cu(II) and using dialysis to remove free Cu(II). Cu(II) preloading of Memo resulted in about 20-fold higher activity than no preloading or to preloading with Fe(II) (Fig. 4E), suggesting that Memo retained Cu(II) once it was bound. Cu(II)-preloaded H192A Memo produced significantly less O2 than wild-type Memo (Fig. 4E).

Next, we determined the enzymatic properties of wild-type and H192A Memo by measuring O2 production, using Cu(II) as a substrate (fig. S4, F and G). Wild-type Memo had a lower affinity (Km) for Cu(II) than H192A Memo, but a fourfold higher catalytic turnover (Kcat) of Cu(II) (Fig. 4F). These results suggest that Cu(II) binds the mutant more readily than wild-type, but is either not efficiently reduced or rapidly dissociates, the end result being that wild-type Memo reduces Cu(II) more efficiently than the mutant (Kcat/Km). During this process, the active-site histidines could act as bases to donate an electron to Cu(II), generating a Cu(I) intermediate that can reduce and activate molecular oxygen to produce O2 (fig. S4H). For Memo to remain active in vivo, it would need to be regenerated, namely, reduced, either actively (for example, by reductases) or passively (for example, by reduced glutathione), possibilities that will be analyzed in the future.

Finally, we examined the effect of Memo on RhoA, a protein that interacts with Memo (16), is redox-sensitive (17, 18), and is involved in migration. Quantitative mass spectrometric (MS) analysis of the net oxidation state of RhoA cysteines in recombinant wild-type and H192A Memo proteins showed that Cu(II) preloading was required for Memo to affect the oxidation of RhoA (fig. S4I), as well as the oxidation of individual cysteines in RhoA (Fig. 4G). The increase in RhoA oxidation status was also significant, albeit to a lesser extent, with H192A Memo (Fig. 4G and fig. S4I), which is in line with its lower catalytic activity.

Memo influences the cellular redox status

Having established that Memo is a copper-binding redox enzyme, we then explored Memo’s cellular activity. First, we tested invasion in the presence of the copper chelator tetrathiomolybdate (TM). Treatment of MDA-MB-231 shLZ cells with TM resulted in a 40% decrease in invasion in vitro (Fig. 5A), showing that removal of copper affects signaling pathways required for this process. Cells with sh5-mediated Memo knockdown, which exhibited impaired invasion capacity, were not affected by TM treatment (Fig. 5A), suggesting that Memo might be the major copper-binding target involved in invasion in vitro by MDA-MB-231 cells.

Fig. 5 Memo promotes a more oxidized intracellular environment.

(A) Invasion of MDA-MB-231 control shLZ and Memo KD sh5 cells treated with 1 nM TM, through Matrigel-coated Boyden chambers. Data represent means ± SD (n = 4 chambers per condition). (B) Western blot analysis of the abundance of reduced and total actin, Rac1, and RhoA in lysates incubated with AMS. Quantification of reduced RhoA, relative to total and normalized to shLZ, is shown below. Data represent means ± SD of four independent experiments. (C) Relative abundance of mRNAs encoding catalase, GCLC, and GPx1 in the indicated MDA-MB-231 cell lines. Data represent means ± SD of six independent experiments.

Next, we asked whether the presence or absence of Memo affected the cellular redox status, first by examining specific proteins, including RhoA, Shc, and actin, which interact with Memo (1, 2, 16), as well as Rac1, another protein involved in migration. Nonreducing SDS–polyacrylamide gel electrophoresis (SDS-PAGE) analysis of lysates treated with the heavy thiol-binding probe AMS (4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid) revealed that reduced RhoA was present only in MDA-MB-231 cells expressing sh5, but not in control cells expressing shLZ or sh5-expressing cells reconstituted with Memo (Fig. 5B). These results support the MS results obtained with recombinant proteins. In the same lysates, the redox status of Rac1 and actin was not altered (Fig. 5B). In addition, SKBR3 cells with Memo knockdown had increased amounts of reduced RhoA and the p52 isoform of Shc (fig. S5A).

Finally, we examined the abundance of mRNAs encoding key redox homeostasis enzymes, namely, catalase, glutamate-cysteine ligase (GCLC), and glutathione peroxidase (GPx1). The abundance of the mRNA encoding catalase, which converts H2O2 to H2O, was decreased in sh5-expressing MDA-MB-231 (Fig. 5C) and SKBR3 cells (fig. S5B). In addition, the mRNA for GCLC, a rate-limiting enzyme for glutathione synthesis, was also lower in sh5-expressing MDA-MB-231 cells (Fig. 5C). Conversely, the abundance of the mRNA for GPx1, which oxidizes GSH [glutathione (reduced form)] to GSSG (oxidized glutathione), was increased in both MDA-MB-231 and SKBR3 cells with sh5-mediated knockdown of Memo (Fig. 5C and fig. S5B). These changes were reversed in the sh5-NT reconstituted MDA-MB-231 cells (Fig. 5C). Together, these results suggest that the cellular environment is more reduced when cells have low Memo abundance, as indicated by the redox status of individual proteins and the abundance of mRNAs encoding key redox enzymes.

Memo increases ROS concentrations in protrusions and promotes O2 production

Because Memo is a redox enzyme, we measured ROS concentrations in Memo knockdown cells. There are several major ROS sources within cells including the mitochondrial transport chain, peroxisomes, and redoxisomes (19). Measurement of the concentrations of intracellular ROS, using carboxy-H2DCF-DA (DCF-DA) followed by fluorescence-activated cell sorting (FACS), revealed no significant differences between MDA-MB-231 cells expressing control shLZ and with sh5-mediated Memo knockdown (Fig. 6A). However, by visualization and quantification of ROS using the CellROX Deep Red reagent, we determined that ROS concentrations in cellular protrusions, specifically the lamellipodium and lamella (Fig. 6B), were significantly lower in sh5-expressing cells (Fig. 6C). In control cells, 3.5% of total cellular ROS was found in these protrusions, and in Memo knockdown cells, the amount was significantly decreased to 1% (Fig. 6C). Reconstitution of Memo in sh5-expressing MDA-MB-231 cells restored ROS concentrations to the values above those in the control cells (Fig. 6C). Similar results were obtained with SKBR3 and T47D cells (fig. S6, A and B). GFP-tagged Memo colocalized with ROS in these protrusions in SKBR3 cells (fig. S6C). Thus, Memo is required for production of a small percentage of cellular ROS, which can be visualized because of its localized nature, but is not measurable by FACS.

Fig. 6 Memo stimulates the generation of ROS in cellular protrusions and NOX-mediated O2 production.

(A) Intracellular ROS concentrations in MDA-MB-231 shLZ and sh5 Memo KD cells under normal growth conditions as determined by carboxy-H2DCF-DA treatment followed by FACS. Data represent means ± SD of three biological replicates from two independent experiments. (B) MDA-MB-231 cell stained to visualize the nucleus (blue), microtubules (green), and actin (red). Protrusions (lamellae) are circled. The lamellipodia, actin-rich structures at the leading edge of the protrusion, are indicated by the white arrows. (C) CellROX staining of MDA-MB-231 shLZ control, sh5 Memo KD, and sh5-NT Memo reconstituted cells after starvation and serum stimulation. Outlines indicate cell boundaries, and arrows indicate protrusions (for magnifications, see upper images). Images are representative of three independent experiments. The percentage of ROS present in protrusions relative to total intracellular ROS concentrations was quantified (graph, right). Data represent means ± SD (n = 16 images per group). (D) Representative graph of long-term NOX activity, as measured by O2 production (RLU), upon PMA stimulation of MDA-MB-231 shLZ, sh5, and sh5-NT cells (left). The area under the curve was quantified (middle). Western blot analysis of Memo abundance in control shLZ, Memo sh5 KD, and sh5 nontargeting Memo reconstituted (sh5-NT) MDA-MB-231 cells is shown (right). Data represent means ± SD of three independent experiments.

The plasma membrane is a site of localized ROS signaling (20, 21). The ROS-producing NOXs promote migration (11, 12, 22), localize to cellular membranes, and are concentrated in the lamellae of motile cells (23). We examined whether Memo might influence the activity of a specific pool of NOX. NOX1, which is activated by protein kinase C (PKC) (19, 22), was detected at the mRNA level in MDA-MB-231 cells (fig. S6D). We thus measured O2 production by live MDA-MB-231 cells after stimulation with the PKC activator phorbol 12-myristate 13-acetate (PMA). After PMA addition to control shLZ-expressing cells, there was a rapid peak of O2 production at 60 s (fig. S6E), then a second broader wave between 1.5 and 200 min that peaked at 60 min (Fig. 6D). All cells produced O2 at similar amounts 60 s after PMA treatment (fig. S6E), showing that PKC was activated in all lines (20, 22). However, cells with Memo knockdown did not produce the second wave of O2 during the 200-min time course, a response that was restored in sh5-expressing cells reconstituted with Memo (Fig. 6D). The higher amounts of O2 produced by the sh5-NT cells may reflect the greater abundance of Memo in the reconstituted cells compared to the control cells (Fig. 6D, right panel). The general NOX inhibitor diphenyleneiodonium chloride (DPI) blocked both basal and PMA-induced O2 production (fig. S6F), indicating that the O2 generated in response to PMA was NOX-mediated. In summary, the results suggest that Memo promotes a more oxidized environment in cellular protrusions formed upon stimulation with growth factors and that Memo is required for sustained O2 production in response to NOX activation.

Increased Memo abundance in breast cancer correlates with aggressive disease parameters and early events

The abundance of Memo has been reported to be increased in a human pancreatic adenocarcinoma cell line (24), and Memo is a target of miR-125b, a microRNA that is down-regulated early in breast cancer development, at the preinvasive stage (25, 26). To investigate whether Memo plays a role in human breast disease, we performed immunohistochemistry on a breast cancer cohort arrayed on tissue microarrays (27, 28) (table S1). We determined that Memo abundance was low in normal breast tissue and significantly higher in >40% of the tumors (Fig. 7, A and B).

Fig. 7 Memo abundance is increased in breast cancer and is correlated with aggressive disease.

(A) Immunohistochemical quantification of Memo in human normal and breast cancer samples. The number and percentage of samples with low and high Memo abundance are shown. Statistical analysis was performed by contingency table analysis with Fisher’s exact test (JMP 10). (B) Example of Memo abundance in nonneoplastic (N) and invasive tumor (T) structures within the same core. (C) Distribution of the number of cases displaying low and high cytosolic Memo (C-Memo) and nuclear Memo (N-Memo). C- and N-Memo abundance are significantly correlated (P < 0.0001). (D) Examples of C- and N-Memo staining. (E) Correlation between high total Memo (cytoplasm and/or nucleus), C-Memo, and N-Memo with patients’ clinical and histopathological parameters at surgery. £: The odds ratio (OR) and P value of the univariate logistic regression model, high total Memo. *: The OR and P value of the bivariate logistic regression model, high C-Memo adjusted for N-Memo. $: The OR and P value of the bivariate logistic regression model, high N-Memo adjusted for C-Memo. (F) Multivariate logistic regression model of the association between patients’ characteristics at surgery and the probability of positive Total, C-, or N-Memo. (G) Multivariate analysis of high C- or N-Memo with distant metastasis and OS. The hazard ratios (HRs) comparing C- or N-Memo high compared to low were estimated with a multivariable Cox proportional hazards model, controlled for clinical and histopathological parameters at surgery within 5 years (≤5 years) and after 5 years (>5 years) from surgery. Significant P values are indicated in red.

Memo is present in the cytoplasm and nucleus in normal cells and in breast tumor cell lines (8). To examine whether the subcellular localization of Memo correlates with clinical or histopathological parameters, we categorized the breast tumor samples according to Memo abundance and localization. Memo was found at low abundance in both the cytoplasm and the nucleus of >90% of normal breast samples and displayed high nuclear abundance in the remaining cases (Fig. 7A, normal breast, high). In the tumor samples, Memo was found at low abundance in both the nuclear and cytoplasmic compartments in 56.3%, high in both of these compartments in 23.3%, high exclusively in the cytoplasm in 6.6%, and high exclusively in the nucleus in 13.8% (Fig. 7, C and D, and fig. S7A). We next performed a correlation analysis of Memo status with the different molecular subtypes and various clinical and histopathological parameters (Fig. 7E and table S2). Because the function of Memo could vary between cellular compartments, a correction for the possible confounding effect of the contribution of cytosolic Memo to the nuclear Memo analysis, and vice versa, was made. Predominantly high cytosolic Memo significantly correlated with many parameters indicative of aggressive disease (such as high grade, estrogen receptor/progesterone receptor–negative, Ki67+, ErbB2+, and p53+) (Fig. 7E, cytosolic Memo, and table S2), as well as with the poor-prognosis, TN, and luminal B (Lum B) molecular subtypes (29, 30). Predominantly high nuclear Memo inversely associated with poor prognostic factors (high tumor grade, estrogen receptor/progesterone receptor–negative, and TN tumors) (Fig. 7E, nuclear Memo, and table S2). These dichotomous results explain why none of the examined parameters, except p53+ status, significantly correlated with total Memo staining (Fig. 7E and table S2, total Memo) and suggest that the cellular localization of Memo can determine its effect on tumor biology. Additionally, in a multivariate logistic regression model, increased cytosolic Memo retained its correlation with estrogen receptor/progesterone receptor–negative and p53+ status (Fig. 7F, cytosolic Memo).

Finally, in a time-dependent survival analysis, cytosolic Memo was strongly prognostic of distant metastasis and overall survival within 5 years (Fig. 7G and tables S3 and S4), outperforming lymph node status, the single most important indicator of disease-free survival and overall survival in breast cancer (31) (table S4). Given the limited size of the breast subtype groups, we could not evaluate the independent prognostic value of cytosolic Memo with regard to the different molecular subtypes. Nonetheless, these results provide evidence that cytosolic Memo is an independent prognostic factor for more aggressive tumors, with higher risk of early distant recurrence and death.


Metastasis is the major cause of breast cancer–related death. Accordingly, studies aimed at understanding the molecular underpinnings of tumor cell spread, as well as their survival and proliferation in metastatic sites, are of utmost importance to identify new anticancer approaches. Herein, we demonstrated that Memo was crucial for breast tumor cell migration and in vivo metastasis and was prognostic of poor patient outcome. In addition, we showed that Memo was an oxidase that was required for sustained NOX-mediated O2 production and increased localized ROS abundance. Furthermore, loss of Memo altered the redox status of several proteins involved in important biological processes, including migration.

We propose that the role of Memo in cancer is to promote tumor cell motility. In our in vitro analyses, including chemotactic Transwell migration (1, 2, 4), invasion in 3D matrices, and wound closure assays (this paper), we have consistently observed that Memo is required for cell motility. Moreover, we showed that tumors with Memo knockdown had significantly fewer motile cells and lung metastatic lesions. How does Memo influence motility and metastasis? As discussed below, Memo has a role in many pathways that influence migration. Related to the new results in this paper, the role of Memo in promoting localized ROS production in the lamellae may be important. The NOXs, which promote migration (11, 12, 22), are concentrated in the lamellae of motile cells (21, 23). NOX activation contributes to spatially controlled increases in ROS concentrations, thereby modulating the function of proteins important for cell motility (10, 32). NOXs produce ROS and, in turn, are activated by ROS, suggesting that a positive feed-forward mechanism might amplify O2 production. We propose a model (Fig. 8) in which Memo, which is also present at the cell cortex (8), influences activity of redox-sensitive proteins that are required for sustained activation of specific NOX complexes. Memo is also required for membrane localization of several proteins involved in migration (2, 16). It is not known if the catalytic activity of Memo is required for this function, or if Memo can act as a scaffold protein independent of its enzymatic activity. Generation of a specific Memo inhibitor, or a truly catalytic-dead Memo mutant, would be required to address these questions.

Fig. 8 Model for Memo’s function in intracellular ROS production and migration.

(A) Extracellular stimuli can activate signaling pathways, such as PKC, leading to NOX1 activation. (B) NOX-dependent redox signaling, occurring at cellular membranes, leads to spatially restricted increases in ROS concentrations, thereby modulating the function of proteins important for cell motility. (C) In addition, NOX complexes themselves are activated by ROS, suggesting that a positive feed-forward mechanism might amplify O2 production. Memo, which is also recruited to the cell cortex upon growth factor stimulation, is necessary for sustained activation of NOX signaling, perhaps by influencing the activity of redox-sensitive proteins. (D) Memo also controls the cellular redox status of RhoA, Shc, and, potentially, other proteins. (E) Memo also influences pathways promoting cell motility and, as we show herein, invasion and metastasis.

In ErbB2-induced migration, the process in which we discovered Memo (1), Memo has been linked to RhoA, the RhoA effector and formin mDia1, microtubule capture, and chemotaxis (33). Phospholipase C–γ, which is stimulated by ErbB2, also converges on Memo and participates in microtubule capture (13). Moreover, we have shown that Memo increases the actin-depolymerizing and actin-severing activity of cofilin (2). Together, these results suggest that Memo regulates various pathways that have roles in migration (Fig. 8).

Our results also showed that in Memo knockdown cells, there are general changes in the redox environment that are reflected in alterations in the abundance of mRNAs for enzymes that control redox homeostasis. These changes were consistent with a more reduced cellular environment in the absence of Memo, and were further reflected in the redox status of RhoA and p52 Shc, both of which showed an increase in their reduced forms in Memo knockdown cells. On the other hand, actin and Rac1, both of which are subject to redox control (34, 35), were not altered in Memo knockdown cells. The redox status of a protein is controlled by multiple inputs including ROS concentration, cellular localization, and upstream signaling pathways (9, 19, 21). The different outcome of Memo loss on the examined proteins likely reflects the complex nature of ROS control. Because RhoA activation is ROS-dependent (17) and downstream of NOX (36), it is possible that the redox changes in RhoA induced by the absence of Memo reflect alterations in NOX activity. Alternatively, RhoA might be a direct Memo substrate because Memo oxidizes RhoA in vitro. Further analyses will be necessary to determine the pathway by which Memo controls the redox status of cellular proteins.

Our study of 407 human breast tumors indicated that Memo abundance was prognostic of poor patient outcome. This is noteworthy because the breast cancer patients in the studied cohort generally had a good prognosis. Indeed, 65% did not have lymph node involvement; >90% were estrogen receptor/progesterone receptor–positive and had low-grade tumors, and/or had no distant recurrence at 12.1 years. Increased abundance of cytosolic Memo, which occurred in ~7% of the tumors, significantly predicted early metastasis and death, and outperformed nodal status for both outcomes. In this analysis, high cytoplasmic Memo significantly correlated with the proliferative marker Ki67. Memo loss does not affect proliferation in normal (4) or cancer cell lines (this work). It is possible that cultured cell lines cannot respond to changes in Memo abundance. However, although MDA-MB-231 xenografts lacking Memo grew at the same rate as control xenografts, lung metastatic lesions from Memo knockdown tumors were significantly smaller. These results suggest that in the lung environment, Memo abundance might affect survival or proliferation of cancer cells.

The data from the tissue microarray also suggested that Memo has distinct roles in different cellular compartments and that increased nuclear Memo, which was present in about 14% of the tumors, correlated with good prognostic parameters. We have not explored the role of Memo in the nucleus, although Memo interacts with the estrogen receptor and increases its transcriptional activity (6). Moreover, Memo can regulate some transcriptional pathways. Indeed, overexpressing Memo in MCF10A cells triggers epithelial-to-mesenchymal transition by increasing the abundance of the transcription factors Snail and Zeb1/2 (5). We did not specifically examine epithelial-to-mesenchymal transition in our analyses, but considering its importance in migration, the results are in line with our studies. It is also noteworthy that Memo conditional knockout mice have a reduced life span, which is accompanied by other characteristics of premature aging (3), which could be related to the redox activity of Memo.

Memo is structurally homologous with metal-binding enzymes (7), and we showed that the active-site pocket of Memo most likely coordinates a copper ion. Other copper-dependent proteins such as SOD1 (37), or indirectly hypoxia-inducible factor 1α (HIF-1α) (38, 39), are important for tumor growth and angiogenesis (4042). We showed that the copper chelator TM reduced the invasive ability of control MDA-MB-231 cells, but not those with Memo knockdown, suggesting that Memo might be the major copper-bound enzyme that promotes invasion in this model. Early clinical trials with copper chelators are ongoing ( (43), and Memo inhibition might contribute to the antitumor effects of these compounds. To conclude, our data suggest that further analyses of the mechanisms by which Memo is activated, its downstream targets and localized functions, are warranted, based on its integral role in human metastatic disease and its function as a modulator of cellular ROS signaling.


Ethics statement

A tissue microarray from 407 female breast cancer patients was analyzed retrospectively under protocols approved by the Institutional Ethical Committee of the European Institute of Oncology. All animal experiments were carried out according to Swiss guidelines governing animal experimentation, approved by the Swiss veterinary authorities (licenses 2286 and 1412).

Reagents and antibodies

β1-HRG was from R&D Systems; PMA, CuCl2, CuSO4, (NH4)2Fe(SO4)2, ZnCl2, FeCl3, CoCl2, Na2MoO4, MnCl2, CaCl2, SOD, doxycycline, and ammonium TM were from Sigma. DPI was from Millipore. Carboxy-H2DCF-DA, CellROX Deep Red reagent, and AMS were from Invitrogen. TO-PRO-3 was from Molecular Probes Inc. Growth factor–reduced Matrigel was from BD Biosciences. Doxycycline chow was from Bio-Serv. The monoclonal mouse anti-human Memo antibody was purified in-house (2). Other antibodies are as follows: β-actin MAB1501, Rac1 23A8, and laminin V MAB19562 (Merck Millipore Corp.); vimentin NCL-VIM-V9 (Novocastra Leica Biosystems); CD31 550274 and GM130 610822 (BD Biosciences); phospho–histone H3 9701 (Cell Signaling Technology Inc.) and α-tubulin MS-581-P1 (Thermo Fisher Scientific); Ki67 18-0191 and GFP A-11122 (Invitrogen Corp.); ErbB2 21N (1); and c-Myc 9E10 sc-40 and RhoA sc-179 (Santa Cruz Biotechnology).

Cell culturing and proliferation assays

MDA-MB-231, T47D, SKBR3, HEK293, and HEK293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum (FCS) (GIBCO Invitrogen) at 37°C, 5% CO2. MCF10A-derived lines were grown in 2D and 3D culture as described (14). For all experiments performed with doxycycline-inducible lines, cells were pretreated with doxycycline for a minimum of 3 days, and treatment was maintained for the duration of the experiment. To measure proliferation, cells were seeded in full-growth medium, and three wells were counted each day up to 4 days after plating, using trypan blue exclusion, in a Beckman Coulter Vi-CELL counter (Beckman Coulter).

Vectors and cloning

For a list of constructs and primers, see table S5. Lentiviral constructs were purchased from Sigma or cloned in-house. Synonymous mutations were introduced into the sh5 recognition site of the N-terminally Myc-tagged Memo construct in pcDNA3.1 (7). This construct was further subcloned to produce Myc-Memo-sh5-NT-pLHCX (referred to as sh5-NT).

Lentiviral and retroviral infections

Cells were infected overnight at 37°C with lentiviral or retroviral particles at a multiplicity of infection of 2, in the presence of polybrene (5 μg/ml) (Sigma). The medium was then changed, and the cells were incubated for 24 hours at 37°C. Successfully infected cells were selected using puromycin and hygromycin (both from Sigma) or FACS-sorted for GFP fluorescence.

Wound healing and invasion assays

Cells were grown to confluency and either maintained in full medium or starved overnight in DMEM containing 0.1% FCS. Monolayers were scratched, and wound healing was monitored using a Widefield TILL5x, Axiovert 200M (5% CO2 and 37°C chamber). Time-lapse images were digitally captured every 20 min with a CCD (charge-coupled device) camera for 8 hours. The area recovered by the migrating cells was calculated using ImageJ. Matrigel invasion assays were performed with BD BioCoat Growth Factor Reduced Matrigel Invasion Chambers as per the manufacturer’s instructions (Becton Dickinson). Briefly, 2.5 × 104 MDA-MB-231 cells were seeded in the upper chamber in DMEM containing 0.1% bovine serum albumin (BSA). The lower chamber was filled with DMEM containing 10% FCS. TM (1 nM top and bottom chambers) was added at the time of seeding. Cells were left to invade for 24 hours, at which point the upper chambers were cleaned using a cotton swab. Cells on the lower surface of the membrane were fixed with 4% paraformaldehyde, stained with crystal violet, imaged, and quantified with ImageJ. MCF10A-ErbB2/ErbB3–derived lines were cultured in 3D in the presence of doxycycline and 1.2 nM HRG. After 7 days, the cells were fixed, stained for immunofluorescence, and imaged as previously described (14).

Animal experiments

All experiments were performed with 6-week-old female nonobese diabetic/severe combined immunodeficient mice (Charles River Laboratories); 5 × 105 cells were injected into the second mammary fat pad or into the tail vein. Mice were sacrificed 28 days after injection. Lungs and tumors were processed as previously described (44). All immunohistochemical stainings were done using the Ventana Discovery XT biomarker platform. The right lobe of the lung was paraffin-embedded, and sections were taken every 50 μm through the entire block and were stained using antibodies against human vimentin or GFP. Tumor sections were stained using antibodies against phospho–histone H3 and CD31. Digital slide image data were generated using a slide scanner (Zeiss MIRAX, version 1.12) at a final magnification comparable to ×400. Whole-slide automated quantitative assessment of stained tissue areas was performed with Definiens XD software (version XD 2.0, Definiens AG) and an in-house developed image analysis algorithm. Blood burdens were assessed as described (45). Single tumor cells from blood were grown in DMEM/20% fetal bovine serum and counted 10 days after plating.

Intravital imaging by multiphoton microscopy

For intravital imaging by multiphoton microscopy (IVI-MP) experiments, stable or doxycycline-inducible shRNA GFP-expressing cells were injected into the second or fourth mammary fat pad of mice. Inducible lines were cultured in doxycycline-containing medium for 3 days before injection. Recipient mice were fed doxycycline-containing chow 2 days before injection and for the remainder of the experiment. Orthotopic primary tumors were imaged 5 weeks after injection. The animals were imaged, as described previously in (45), on a custom-built microscope (46).

Cell and tumor lysates

Cells were lysed in Triton X-100 buffer as described (2). Tumors were homogenized in NP-40 lysis buffer (44) using the Precellys 24 tissue homogenizer in 2-ml tubes with ceramic beads (20 s at 5500 rpm at 4°C). The extracts were clarified using a tabletop centrifuge at maximum speed for 20 min at 4°C, and protein concentrations were determined.

Expression and purification of recombinant proteins

Human Memo cDNA (complementary DNA) was cloned into pOPINF (47) and expressed in BL21(DE3) Escherichia coli as described (48). Cell pellets were freeze-thawed, resuspended in lysis buffer [50 mM tris (pH 7.5), 500 mM NaCl, 20 mM imidazole, 0.2% Tween, Complete EDTA-free protease inhibitor cocktail (Roche), and Benzonase (3 U/ml) (Sigma)], and homogenized with the EmulsiFlex-C3 high-pressure homogenizer (Avestin GmbH). Cleared lysates were batch-purified on Ni-NTA beads, and bound protein was cleaved by incubating with His-tagged 3C enzyme for 2 hours at +4°C in 50 mM tris (pH 7.5), 500 mM NaCl, and 20 mM imidazole. The eluted protein was monitored on ÄKTA Purifier (GE Life Sciences), concentrated, and purified to >96% using Sephadex 75 gel filtration in 20 mM tris (pH 7.5), 200 mM NaCl, and 0.02% NaN3. The ultrapure, eluted Memo was concentrated with an Amicon Ultra filtration unit. Glutathione S-transferase (GST)–RhoA in the pxRB096 vector (provided by X. Bustelo, University of Salamanca, Spain) was induced in BL21(DE3) E. coli using 0.3 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 4 hours at 30°C. Cell pellets were freeze-thawed and incubated with BugBuster lysis buffer (Millipore) supplemented with 2 mg of lysozyme, 60 U of Benzonase, and Complete EDTA-free protease inhibitor cocktail for 20 min at room temperature (RT). Cleared lysates were batch-purified on glutathione Sepharose 4B resin (GE Healthcare), and the bound GST-RhoA was incubated in thrombin buffer [20 mM tris-HCl (pH 8.4), 150 mM NaCl, 2.5 mM CaCl2, 15 U of biotinylated thrombin (Novagen)]. To remove free thrombin, beads were sedimented, and the supernatant was incubated with 50 μl of NeutrAvidin beads (Thermo Fisher Scientific) that were then sedimented, and the concentration of cleaved RhoA in the supernatant was measured.

Quantitative real-time PCR analyses of selected targets

Total RNA was collected from cells using the QIAshredder and RNeasy Mini Kit (Qiagen). cDNA synthesis and amplification were performed with GE Healthcare Ready-To-Go You-Prime First-Strand Beads according to the manufacturer’s instructions. Quantitative real-time polymerase chain reaction (qPCR) was performed with the ABsolute qPCR Master Mix (Thermo Fisher Scientific) and SetpOnePlus Real-Time PCR (Applied Biosystems). Data were normalized to h18S abundance. See table S5 for primer sequences.

Dioxygenase activity measurements

Intradiol or extradiol dioxygenase activity was measured as described (49). In brief, 400 μM protocatechuic acid was added to each well of a 96-well plate containing either 0.1 nmol Memo or 0.1 nmol PD in 10 mM tris-HCl (pH 8.5). The cleavage of protocatechuic acid was monitored spectrophotometrically by the decrease in absorbance at 290 nm. The absorbance of the blank sample was subtracted from each reading.

Cu(II) reduction assays

The Cu(II) reduction capacity of 0.1 nmol recombinant Memo was assayed as described (Cell Biolabs). The spectrophotometric absorbance at 490 nm was read in a SpectraMax 190 absorbance plate reader (Molecular Devices) and correlated to a uric acid standard curve, yielding copper-reducing equivalent values (μM CRE). The activity of protein-free samples was subtracted. To measure Memo activity in cells, pcDNA Myc-Memo and Myc-Memo mutants were transiently expressed in HEK293 cells for 72 hours. Cells were lysed in radioimmunoprecipitation assay buffer, and 500 μg of lysate was loaded onto 50 μl of 50% EZview Anti-Myc Sepharose slurry. Beads with bound Myc-Memo were resuspended in 35 μl of 10 mM tris-HCl (pH 8.5); 5 μl was used for Western blot quantification, and 30 μl for copper reduction activity measurements. Assays under hypoxic conditions were performed as described above, except all buffers were flushed with N2 for 2 hours before the assay.

Superoxide detection assays

O2 production was determined by using a luminol oxidation assay (Sigma) with slight modifications. In brief, recombinant protein or 1 × 106 cells were used per reaction; in some assays, 4 U of SOD was added. To stimulate NOX activity, PMA (1 μg/ml) was added to cultures; 10 μM DPI was added to block NOX. Values from blank samples, without Memo or PMA, were subtracted from each reading. For short-term (200 s) and long-term (220 min) assays, luminol oxidation was measured as RLU for 10 s every 10 s or 30 min, respectively, using a Berthold Centro XS3 LB 960 plate reader (Berthold Technologies GmbH). For assays done with metal-preloaded recombinant Memo, 50 pmol Memo (5 μM final) was incubated with 50 μM CuCl2 or (NH4)2Fe(SO4)2 in 50 mM tris-HCl (pH 6.8) for 30 min at 4°C. Excess metal ions were removed by dialysis against the same buffer for a minimum of 3 hours at 4°C. For the assays in fig. S4 (F and G), the indicated amounts of Memo or Cu(II) were used.

Redox status of recombinant RhoA and other proteins

The oxidation status of cysteine residues in RhoA was determined by quantitative MS. Recombinant RhoA was incubated in the presence or absence of recombinant, Cu(II)-preloaded or nonpreloaded wild-type Memo or H192A Memo, in 50 mM tris-HCl (pH 6.8), for 30 min at RT. Then, the samples were incubated for 30 min at RT in the dark with 35 mM heavy iodoacetamide-13C2, 2-d2 [heavy indole-3-acetic acid (IAA)] (Sigma). IAA was removed by trichloroacetic acid/acetone precipitation. The remaining oxidized cysteines were reduced with 20 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP) for 30 min at RT, then alkylated by 35 mM IAA-12C2 (light IAA). Samples were then digested for 6 hours at 37°C with endoproteinase Lys-C, followed by overnight digestion in trypsin at 37°C. Diluted samples were applied on an EASY-nLC 1000 liquid chromatography (LC) system, coupled to a LTQ Orbitrap Velos MS (both from Thermo Scientific). Quantitative analysis of the ratio light/heavy IAA incorporation was done with Progenesis LC (Nonlinear Dynamics).

For intracellular redox status of different proteins, cells were harvested in phosphate-buffered saline (PBS) with inhibitors and then sonicated, and supernatants were collected. Equal amounts of protein extracts were treated for 30 min at RT with 15 mM AMS. Samples were separated by 15% nonreducing and reducing SDS-PAGE.

Detection of intracellular ROS concentrations

For quantification of ROS concentrations by FACS, cells were washed twice with PBS and incubated with 20 μM carboxy-H2DCF-DA in PBS for 10 min at 37°C. Cells were then washed twice, trypsinized, resuspended in PBS with 10% FCS, spun down for 3.5 min, and resuspended in PBS with 0.1% BSA. Samples were then analyzed using a BD LSRII SORP (BD Biosciences). For quantification of ROS concentrations by imaging, growing or serum-starved cells, stimulated as indicated, were cultured in 5 μM CellROX Deep Red reagent. Cells were then incubated at 37°C for 30 min, washed with growth medium, and imaged at excitation/emission maxima 640/665 nm in a temperature- and CO2-controlled chamber using an Olympus IX81 spinning disc microscope driven by MetaMorph software (Molecular Devices). CellROX intensities in cell protrusions were measured and quantified relative to the total ROS concentrations in the cell, using ImageJ.

Analyses of breast cancer tissue microarrays

Clinical and pathological characteristics of 407 female breast cancers (27, 28, 50) are shown in tables S1 and S2. Staining for Memo was performed on 2-μm sections, using the EnVision Plus/horseradish peroxidase (HRP) system (Dako). The specificity of the Memo antibody was verified by staining breast tumor cell lines with Memo overexpression or Memo knockdown (fig. S7B). For scoring, a semiquantitative approach with scores ranging from 0 to 3 was used. For the correlation analyses of Memo abundance with clinical parameters, three categories were defined: total Memo, considering Memo abundance in the nucleus and/or the cytoplasm; cytosolic Memo, considering only cytoplasmic Memo; and nuclear Memo, considering only nuclear Memo. Memo scores in normal tissue (n = 32) ranged between 0 and 1.0; tumor samples with immunohistochemistry scores >1.0 were considered high, and those with ≤1.0 were considered low.

Association between clinical parameters at surgery and probability of high Total Memo was evaluated by univariate analysis. To control for possible confounding effects of the contribution of cytosolic Memo to nuclear Memo, and vice versa, the association of clinical variables with the probability of high cytosolic Memo and high nuclear Memo was evaluated by bivariate logistic regression models, adjusted for nuclear Memo or cytosolic Memo, respectively. The impact of patients’ characteristics on the probability of high cytosolic Memo or nuclear Memo (adjusted for nuclear Memo or cytosolic Memo, respectively) was also assessed by multivariable logistic regression models adjusted for estrogen receptor/progesterone receptor, Ki67, ErbB2, p53, grade, and histology, and is expressed as ORs with 95% confidence interval. A time-dependent multivariable Cox model was used to estimate the HRs comparing cytosolic Memo high to low within 5 years and after 5 years from surgery, controlled for age at diagnosis, tumor size, lymph node status, grade, and status of estrogen receptor/progesterone receptor, ErbB2, p53, and Ki67, as well as nuclear Memo, when evaluating cytosolic Memo, and cytosolic Memo, when evaluating nuclear Memo. SAS statistical software was used for all analyses (SAS Institute Inc.).

Statistical methods

Unless stated otherwise, statistical significance of data following a normal distribution, as determined by the Shapiro-Wilk normality test, was determined using an unpaired, two-tailed Student’s t test, assuming unequal variances or a one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for multiple comparisons. Data that failed the normality test were analyzed using the Wilcoxon rank sum test (one- and two-condition comparisons), or the Kruskal-Wallis test followed by Dunn’s post hoc test for multiple comparisons. Differences were considered significant if the P value was ≤0.05 (*P ≤ 0.05, **P < 0.01, ***P < 0.001, for all tests).


Fig. S1. Memo is required for cellular migration and invasion; Memo overexpression is not transforming.

Fig. S2. Memo down-regulation does not decrease in vitro cellular proliferation or primary tumor outgrowth, but may decrease the survival or proliferation of disseminated tumor cells in the lungs.

Fig. S3. Reconstitution of Memo restores invasion and tumor cell migration in vivo.

Fig. S4. Analyses of Memo activity.

Fig. S5. Memo influences the intracellular redox status.

Fig. S6. Analysis of ROS and selected mRNAs in Memo knockdown cell lines.

Fig. S7. Memo immunohistochemistry.

Table S1. Clinicopathological characteristics of the breast cancer cohort analyzed for Memo abundance.

Table S2. Correlations between Memo status and clinical and pathological parameters in the breast cancer cohort.

Table S3. Observed events in the breast cancer cohort.

Table S4. A time-dependent multivariable Cox model to estimate HRs.

Table S5. Primers, oligos, and vectors used.

Movie S1. Motility of MDA-MB-231 tumor cells expressing shLZ.

Movie S2. Cell protrusions in MDA-MB-231 tumor cells with sh5-mediated Memo knockdown.

Movie S3. Motility of sh5-expressing MDA-MB-231 tumor cells reconstituted with Memo


Acknowledgments: We thank O. Pertz, S. Kondo, and all members of the Hynes Lab for fruitful discussions. We thank C. Luise for excellent technical assistance; P. Maisonneuve, S. Confalonieri, E. Dama, and G. d’Ario for advice on the statistical analysis; and G. Viale for human tissue samples. Funding: N.E.H. was supported by grants from Susan G. Komen for the Cure SAC110041, SNF 310030A-121574 and 310030B-138674 and KFS 02528-02-2010, and the Novartis Research Foundation. P.P.D.F. was supported by grants from the Associazione Italiana per la Ricerca sul Cancro, the Italian Ministry of Education, University and Research (MIUR) and of Health, the Monzino Foundation, and the European Research Council. A.B. was supported by Fondation Arc and INCa-DGOS-Inserm 6038. Author contributions: G.M., I.N., M.V., and N.E.H. planned and wrote the paper; G.M., I.N., T.S., N.A., A.D., S.L., J.W., D.H., J.S., J.J.K., H.G., A.F., and D.S. planned and performed experiments; D.D. performed the statistical analysis of the breast tumor tissue microarray; M.V., G.M., and P.P.D.F. created and performed the immunohistochemical analysis of the breast tumor tissue microarray; and M.B.-A. and A.B. provided material, expertise, and discussions over the course of the experiments and the preparation of the paper. Competing interests: The authors declare that they have no competing interests.

Stay Connected to Science Signaling

Navigate This Article