The Cargo Protein MAP17 (PDZK1IP1) Regulates the Cancer Stem Cell Pool Activating the Notch Pathway by Abducting NUMB

MAP17 (DD96, PDZKIP1) is a small (17 kDa), nonglycosylated membrane-associated protein located on the plasma membrane and the Golgi apparatus (1–3) of the proximal tubule cells of the kidney. It contains two transmembrane regions and a hydrophobic C-terminus that encodes a PDZ-binding domain, that allow its interaction with several PDZ domain–containing proteins, like PDZK1 (4–7).

MAP17 overexpression enhances tumorigenic properties of tumor cells, with increased proliferation, reduced apoptosis, increased colony formation in soft agar, and increased growth ratios in tumors in nude mice (8, 9). MAP17 is overexpressed in a great variety of human carcinomas (10, 11). In many tumors, including glioblastomas, lymphomas, and cervical, breast, prostate, and ovarian carcinomas, MAP17 overexpression is strongly correlated with tumoral progression (7, 11, 12). While adenomas and benign tumors, as well as normal tissues, rarely express MAP17, a high proportion (50%–90%) of late-stage or metastatic tumors show high levels of MAP17, correlating with a more dedifferentiated phenotype (7, 9, 12). These findings highlight the relevance of this gene in tumorigenic process and tumor development.

Increasing evidence has shown that Notch signaling regulates aspects of asymmetric division and stemness (13). Aberrant Notch signaling is found in most types of cancers, either carcinomas or those of mesenchymal origin (14, 15). Notch activation depends on the interaction of ligand receptors between neighboring cells, being followed by γ-secretase cleavage, releasing the active form of the Notch intracellular domain (NICD). Upon its release, NICD translocates to the nucleus, forming a ternary complex with CSL and Mastermind-Like 1 (MAML1) to transactivate its target genes, including HES and HEY gene families (13, 16). NUMB has an antagonistic influence on Notch pathway, inhibiting Notch signaling by binding directly to the NICD domain, thus preventing its access to the nucleus (17–20). NUMB is also able to directly inhibit Notch by recruiting ITCH to polyubiquitinate Notch (21).

The mechanism responsible for the increased tumorigenicity of cells expressing MAP17 is not known yet. To precisely identify the molecular mechanism induced by MAP17 that increases cell tumorigenic properties, we performed a search for MAP17 partners, finding that NUMB physically interacts with MAP17. This physical interaction leads to a mislocalization of NUMB, which increases nuclear NICD, and consequent Notch pathway activation. As a consequence, MAP17-expressing cells increase the presence of stem-related transcription factors along with the stemness of tumor cells.

All methods were performed in accordance with the relevant guidelines and regulations of the Institute for Biomedical Research of Seville (IBIS, Seville, Spain) and University Hospital Virgen del Rocio (HUVR, Seville, Spain).

MAP17, previously cloned into pBabe-puro (12), was cloned into plasmid pGBKT7 using EcoRI-BamHI sites. To test MAP17 C-terminal deletions, an opal STOP codon was introduced by mutagenic PCR in DNA sequence corresponding to amino acids 71, 102, and 110, obtaining the pGBKT7-MAP17_FL, pGBKT7-tMAP17, pGBKT7-MAP17₇₀, and pBKT7-MAP17₁₁₀ vectors. Truncated MAP17 (tMAP17) was also obtained by mutagenic PCR in pBabe-puro plasmids. NUMB was amplified from pCMV6-XL5-NUMB (SC301023, Origene) and cloned into pGADT7 vector using EcoRI-BamHI sites. As a result, MAP17 was fused to the C-terminus of GAL4-binding domain and NUMB to the C-terminus of GAL4 activation domain. All vectors were amplified using E. coli DH5α strain. Y187 and Y2HGold yeast strains were used for yeast two-hybrid assays.

To downregulate NUMB expression, short hairpin RNA (shRNA) against NUMB or scrambled sequence control in pB-RS vectors were obtained from Origene (TR311064). Cells were transfected with shRNA plasmids and selected with 2 μg mL⁻¹ of hygromycin. After selection, two of the four shRNAs against NUMB were selected, shNb2 (TCAGCAGACAGGCATACAGAGGTTCCTAC) and shNb4 (shN in the text; ATCATTCCGTGTCACAACAGCCACTGAAC), after Western blot analysis and qPCR analysis. MAP17 shRNA was previously described (22).

Y2HGold strain was transformed with all MAP17-expressing fusion vectors, while Y187 strain was transformed with NUMB-expressing fusion vector. Briefly, yeast cells were diluted in 150 μL of 35% PEG 5000, 0.2 mol/L lithium acetate, and 0.1 mol/L dithiothreitol (DTT), transformed with 1 μg of each vector by incubating the cells for 45 minutes at 45°C and centrifuged at 550 rpm. Y2HGold cells were plated in synthetic defined (SD) medium without Trp, while Y187 cells were plated in SD medium without Leu. Colonies grew for 5 days at 30°C. For mating, Y2HGold-MAP17 cells were mixed with Y187-NUMB cells in YPD and cultured for 3 days at 30°C. Subsequently, cells were resuspended in sterile water, after measuring OD_600nm. This allowed seeding a similar number of cells in SD plates in four selective conditions: without Trp, Leu nor His; without Trp, Leu, His nor Ade, and the same plates plus X-α-Galactose. Cells grew for other 5 days at 30°C.

T47D, HeLa, Calu3, and 293T cells were obtained from the European Collection of Authenticated Cell Cultures (ECACC) commercial repository at the beginning of this study. No further authentication was performed in these cell lines. Commercial cells were maintained in DMEM (Sigma) while AA, AX, and AW cells, derived from sarcoma patients, were maintained in F10 medium (Sigma). All cultures were supplemented with 10% FBS (Life Technologies), penicillin, streptomycin, and fungizone (23). Clonogenicity assays, Holo- and paraclone analysis (24), colonies in soft agar, and tumorspheres analysis were performed as described previously (25, 26).

Tumor tissues were obtained via surgical resection of sarcomas performed at Virgen del Rocio Hospital (Seville, Spain) after the patient provided written informed consent according to a protocol approved by the local ethics committee (CEI 2013/PI002). The experiments were performed according to the European guidelines for laboratory animal care. This study was approved by the IBIS Institutional Animal Care and Use Committee.

Sarcoma patient-derived xenografts (PDX) were processed as indicated in ref. 27. Upon reaching a size of 1,500 mm³, the mice were euthanized, and the tumors were used to obtain cells to perform tumorsphere experiment. Briefly, tumors were cut in small pieces and cultured in 6-cm² plates with 1 mL of F10 for 24 hours, to allow the attachment of cells to the plate. After that, another 2 mL of F10 medium was added to the plate, and the cells grew for a week before performing the tumorsphere experiment, as described before, using 10,000 cells per well. The investigator was blinded to the data from MAP17 levels and therefore to the outcome.

Protein extracts and nuclei purification for Western Blot analysis were obtained as described previously (25).

For coimmunoprecipitation assays and Western blot detection, antibodies against NUMB (ab4147, Abcam) were used at 1 μg/mL. For Western blot detection, we used MAP17 (MABC522, Merck, 1:500 dilution), Notch (Santa Cruz Biotechnology, s-6014-R, 1:200 dilution), and hnRNP C1/C2 (4F4; Santa Cruz Biotechnology, sc-32308, 1:400 dilution) antibodies. α-Tubulin (T9026, Sigma) was used as a control. Horseradish peroxidase–labeled rabbit anti-mouse (ab97046, Abcam, diluted 1:5,000) and goat anti-rabbit (ab97051, Abcam, diluted 1:5,000) secondary antibodies were used.

To detect MAP17–NUMB interaction, anti-NUMB antibodies were incubated with protein G-Sepharose beads and incubated at 4°C for 3 hours. HeLa or T47D cell extract (2 mg; either overexpressing MAP17 or not) were added to the beads in a final volume of 1 mL and incubated at 4°C for 16 hours. The beads were washed twice with RIPA buffer and once with cold PBS. Western blot analyses were performed as described previously (28, 29).

Cells were cultured onto glass coverslips in 6-well plates, for 36 hours, washed with PBS, and fixed with PBS + 4% paraformaldehyde for 20 minutes, and washed twice with PBS. Cells were permeabilized with PBS + 0.5% Triton X-100 for 5 minutes, washed twice with PBS, and blocked with PBS + 0.1% Triton X-100 + 3% BSA for 30 minutes at room temperature. Then, anti-NUMB (ab14140, Abcam) or anti-NICD (ab8925, Abcam) were added to cells, in 1 mL of PBS + 0.1% Triton X-100 + 3% BSA at a final concentration of 1 μg/mL and incubated at 4°C for 6 hours with gentle stirring. After that, samples incubated with anti-NUMB were washed three times with PBS + 1% Triton X-100 for 5 minutes each, and anti-MAP17 (9, 10), was added to cells in 1 mL of PBS + 1% Triton X-100 at a final concentration of 2 μg/mL, being incubated overnight at 4°C with gentle stirring. After that, coverslips were washed four times with PBS + 1% Triton X-100 for 5 minutes each time, with a final wash with PBS for 5 minutes.

For NICD quantification immunofluorescence assays, Alexa Fluor goat anti-rabbit IgG (A-11008, Life Technologies) was added at a 1:250 dilution to the cells in 1 mL of PBS + 0.1% Triton X-100 + 3% BSA and incubated in dark at room temperature for 2 hours with gentle stirring. Cells were then washed three times (5 minutes each) with PBS + 0.1% Triton X-100. Finally, coverslips were mounted on a slide with a drop of mounting solution (Prolong Gold Antifade, Life Technologies) and dried. NICD images were acquired with a Leica TCS-SP2-AOBS-UV confocal microscope by sequential scanning of the emission channels. Nuclear NICD was quantified using ImageJ software and expressed as the fluorescence in nuclei relative to the normalized cytoplasm fluorescence.

For proximity ligation assay (PLA) labeling, we followed manufacturer's instructions (DUO92101, Sigma). Images were acquired with a Leica TCS-SP2-AOBS confocal microscope. The number of interactions between NUMB and MAP17 was quantified using ImageJ software.

Total RNA was purified as described previously (25). To detect changes in gene expression, we used the following probes, all from Life Technologies: MAP17 (Hs00906696_m1), HES1 (Hs00172878_m1), HES5 (Hs01387463_g1), KLF7 (Hs00748636_s1), ID2 (Hs04187239_m1), GLI1 (Hs01110766_m1), KLF4 (Hs00358836_m1), SOX9 (Hs01001343_g1), NANOG (Hs04260366_g1), OCT4 (Hs00999632_g1), NUMB (Hs01105433_m1), and GAPDH (Hs03929097_g1). Quantitative PCR was performed as described previously (25).

Labeling of HeLA and T47D cells with antibodies against CD44, CD24, and CD133, and its detection by FACS was performed as described previously (25).

To detect correlations between MAP17 and genes related to Notch pathway or stem cell genes, a total of 28 databases of different tumors (breast, lung, colon, and cervix; Supplementary Table S1) were analyzed using R2 software (Genomics Analysis and Visualization Platform, http://r2.amc.nl). All datasets are freely available in R2 webpage. To perform these correlations, MAP17 (Pdzk1ip1, 219630_at) was used to find correlations with all the genes that appear in each database, fixing a P value lower than 0.05 to find statistically significant correlations. Probe 219630_at, which corresponds to MAP17, was used for all Affymetrix datasets, with the exception of TCGA and Budinska datasets, where we used the unique probe for MAP17 gene.

Correlations for MAP17 and associated genes were determined through analysis of the GSE34053, GSE20916, GSE39395, GSE14773, Lung Adenocarcinoma TCGA 515 dataset, and Tumor Breast Carcinoma TCGA 1097 public dataset available at Gene Expression Omnibus. The public array analyses were performed using the R package Bioconductor (http://bioconductor.org). Our analysis tools incorporated functionalities from several other R packages as follows: Geo Query, AFFY, AFFYPLM, GENEFILTER, and LIMMA. For preprocessing, we used AFFY tool applying rma methods for background correction, quantile method for data normalization, PMonly for PM–MM correction and median-polish for aggregation. Differential gene expression was calculated using Bayesian modeling provided by limma package tool. Adenocarcinoma TCGA 515 dataset and Tumor Breast Carcinoma TCGA 1097 were analyzed by automatized tools provided at Oncomine (Compendia Biosciences, www.oncomine.org), R2: Genomics Analysis and Visualization Platform (http://r2.amc.nl/).

To precisely identify the molecular mechanism induced by MAP17 that increases the tumorigenic properties of cells, we performed a yeast two-hybrid screening for MAP17 partners, identifying NUMB as MAP17 interactor. To confirm this physical interaction in cells, we overexpressed full-length MAP17 and immunoprecipitated it with NUMB antibodies, finding that MAP17 also binds to NUMB in human cells (Fig. 1A).

To characterize MAP17–NUMB interaction, we generated several mutants of MAP17 by cutting different regions of the C-terminus, which contain the PDZ-binding domain (Fig. 1B). Two of these mutants were specific for the PDZ-binding domain. MAP17(1–110) lacked only the last 4 amino acids, and MAP17(1–101) lacked the last 13 amino acids. Furthermore, we generated a MAP17(1–70) mutant lacking the entire intracellular region of MAP17. We then explored their interaction with NUMB in using yeast two-hybrid system interactions (Fig. 1C). We found that MAP17–NUMB interaction was disrupted only for MAP17(1–101) and MAP17(1–70) mutants, MAP17(1–101) being the mutant lacking the smallest region unable to bind NUMB, so we referred to it as tMAP17 (truncated MAP17) in the following sections.

To confirm the MAP17–NUMB interaction in cell, we used the in situ PLA which allows direct detection of in vivo protein interactions with high specificity and sensitivity. Protein interaction can be readily detected and localized with single molecule resolution and objectively quantified in vivo (30). To this end, we overexpressed the vector only (EV), MAP17, or tMAP17 in HeLa and T47D cells (Fig. 1D; see also Supplementary Fig. S1). After selection, we performed an in situ PLA and found a clear interaction of MAP17 with NUMB; however, this interaction was not detected with tMAP17 (Fig. 1E and F).

Next, we examined the functional relevance of this interaction. NUMB antagonizes Notch signaling activities by the ubiquitination of the membrane-bound Notch receptor and the subsequent degradation of NICD following receptor activation. Therefore, if MAP17 has a functional role involving NUMB, its overexpression should alter Notch pathway. To determine this, we measured NICD nuclear levels by immunofluorescence in cells overexpressing MAP17, EV, or tMAP17 using NICD antibodies (Fig. 2A). We observed an increased intensity of NICD in the nucleus of cells overexpressing MAP17, showing in EV or tMAP17 cells a similar lower intensity of nuclear NICD, as was confirmed after nuclear signal quantification by ImageJ software (Fig. 2B).

To confirm this finding, we performed nuclear fractionation, and levels of nuclear NICD were determined, detecting that nuclei of MAP17-expressing cells contained more nuclear NICD than control cells (Fig. 2C).

In addition to the known HES1 gene as target of Notch pathway, other target genes also include HES5, KLF7, and ID2 (31). Furthermore, the Notch target HES1 is a repressive transcription factor that binds the first intron of GLI1 and inhibits its expression (32). We measured the levels of these Notch-dependent transcripts in EV, MAP17, or tMAP17 cells, finding increased transcripts levels of HES1 and HES5 in MAP17-expressing cells compared with expression in EV and tMAP17 cells (Fig. 2D). Furthermore, genes that should be repressed under Notch activation, such as KLF7 and GLI1, showed lower levels of transcription in MAP17-expressing cells (Fig. 2D). These results were identical in both cell lines, confirming that MAP17 overexpression activates Notch pathway. Only ID2 showed a different behavior, showing the expected decrease in transcription only for HeLa cells and an increment in T47D cells, maybe due to cell line–specific reasons.

Notch target genes have been connected with the maintenance of the cell's potential for self-renewal, suggesting that Notch pathway activation triggers the production of stem transcription factors (33). Cancer stem cells (CSC), or cancer-initiating cells, show an increase in stem cell factors, such as OCT4, NANOG, SOX9, and KLF4 (34, 35). Therefore, we analyzed the effect of MAP17 overexpression on these CSC markers, observing that cells overexpressing MAP17 contained significantly increased mRNA levels of all these stem cell transcripts (Fig. 3A).

To confirm the cancer stem–like phenotype of MAP17-expressing cells, we measured cellular subpopulations showing CSC surface markers. T47D is a mammary tumor cell, its CSC subpopulation described as CD44⁺/CD24⁻ (28, 36), while HeLa CSC subpopulation has been described as CD133⁺. Therefore, we measured these subpopulations in EV, MAP17, or tMAP17 cells. Both in HeLa- and T47D-overexpressing MAP17 cells, we found an increase in these surface markers compared with EV or tMAP17 cells (Fig. 3B and C).

To analyze the possible role of increased MAP17 levels on the CSC phenotype, we also measured the formation of tumorspheres and clonal growth, commonly associated to the cancer-initiating cell phenotype and the ability to generate new colonies. The last experiment allows distinguishing between holoclones, cells considered derived from stem cells; paraclones, differentiated cells incapable to reconstitute a culture; and meroclones, with intermediate properties between holo- and paraclones (24, 26, 37). Human tumor cell populations can be maintained in serum-free suspension cultures, growing as clusters of cells called “tumorspheres” (38). These tumorspheres display some self-renewal ability upon disaggregation, being enriched with multi-potent epithelial progenitors (34) and an increase in the expression of CSC markers (35). Therefore, we measured whether MAP17 overexpression alters the ability of tumor cancer cell lines to form tumorspheres.

EV, MAP17, or tMAP17 cells were subjected to disaggregation. The resulting single-cell suspension was seeded in complete media, allowing to grow during 5 days to form tumorspheres (Fig. 3D). In both cell lines, the number of tumorspheres was significantly increased only in MAP17-overexpressing cells (Fig. 3E).

Cell lines generated from carcinomas consistently produce in vitro colony patterns similar to those produced by the stem cells of normal epithelia. On the basis of the different types of colony morphologies formed, it is possible to predict some stem cell characteristics (24). Several findings now suggest that malignant cell lines may in fact retain patterns of stem cell behavior (37, 39–41). For example, malignant cell lines derived from holoclones contain subpopulations of cells with putative stem cell characteristics (24, 37). Therefore, we measured whether MAP17 overexpression alters the ability of both cancer cell lines to form holoclones. Trypsinized cells were seeded at low density and grew for 14–20 days. After that, we analyzed the phenotypic characteristics of the individual colonies and the holoclones counted, observing a significant increase holoclone percentage due to MAP17 overexpression (Fig. 3F and G).

The increase in CSC properties in cells with high MAP17 should correlate with a higher number of colony-initiating cells. Therefore, we measured the number of colonies formed after seeding of cells at a very low density (Supplementary Fig. S2). As expected, based on previous results, MAP17-expressing cells formed more colonies than EV or tMAP17. This was also true for colony formation in soft agar (Supplementary Fig. S3).

In summary, these results show that, in human tumor cells, MAP17 overexpression induces the expression of CSC phenotype, which could be responsible for an increase in the stem cell–like properties.

To further evaluate the role of MAP17 expression, we selected Calu3, a tumor cell line with endogenous high levels of MAP17, and downregulated its expression by shRNA ectopic expression (Fig. 3H). We measured Notch pathway activity and the physiologic behavior associated with MAP17 decrease in Calu3 cells, observing changes in levels of Notch-dependent transcripts (Fig. 3I), confirming the relevance of MAP17 expression in Notch pathway activation. In addition, MAP17 downregulation induced a clear reduction in all CSC transcription markers, confirming the functional relationship between MAP17 and stemness (Fig. 3J).

Finally, when MAP17 was downregulated in Calu3 cells, we observed an obvious reduction in the associated stemness phenotype, with a significant decrease both in holoclone and tumorsphere numbers (Fig. 3K and L). These results also confirm that the maintenance of high MAP17 levels is important in maintaining the CSC phenotype.

If MAP17 overexpression is functionally connected to NUMB sequestration, then NUMB downregulation should mimic MAP17 overexpression. We tested the effect of NUMB downregulation both in Notch pathway activation and stem cell–like behavior of tumor cells. To do this, we constitutively overexpressed shRNA against NUMB in the same tumor cell lines, showing a decrease in both mRNA and protein levels (Fig. 4A and B). To avoid off-target effects, we analyzed two different shRNA against NUMB, with similar results (Supplementary Fig. S4). Like in MAP17-overexpressing cells, we observed an increase in Notch pathway activation (Fig. 4C). NUMB downregulation also increased mRNA levels of stem cell factors, correlated with increased colony formation (Fig. 4D and E). In addition, we also observed an increase in holoclone and tumorsphere numbers (Fig. 4F and G).

Finally, to confirm the CSC-like phenotype with downregulated NUMB cells, we measured the cellular subpopulations showing CSC surface markers, as was done with the MAP17-expressing cells. We found that HeLa and T47D cell expressing NUMB shRNA showed a higher percentage of CD133⁺ and CD44⁺/CD24⁻ cells, respectively, confirming that NUMB downregulation is equivalent to MAP17 overexpression in its ability to induce a CSC-like phenotype (Fig. 4H).

Taken together, these data indicate that NUMB sequestration is sufficient to activate Notch pathway and increase the stem cell–like properties of tumor cells, pointing a possible role of NUMB as tumor suppressor by reducing Notch pathway activation and the stemness properties of tumor cells.

To approach our findings in a more in vivo situation, we took low passaged sarcoma cell lines generated in our laboratory directly from sarcoma tumors. The lines have been propagated no longer than 20 passages, so the number of genetic alterations should be lower compared with lines such as HeLa, highly propagated during years.

AA and AW lines, with very low levels of MAP17, and AX line, with low levels of MAP17, were transfected to force its MAP17 constitutive overexpression. In addition, we overexpressed MAP17 shRNA in AX cell line, with endogenous MAP17 expression. After selection, 5,000 cells were seeded to measure the number and size of tumorspheres. We observed that, in all cases, MAP17 overexpression produced a higher number and increased size of tumorspheres in these low-passage sarcoma cell lines (Fig. 5A).

Like for T47D and HeLa cells, mRNA levels from either Notch pathway and stem cell genes were highly activated in cells with high levels of MAP17 (Fig. 5B, black and dark gray bars).

Finally, we took 5 models of PDXs of sarcoma growing in mice. After tissue disaggregation, we directly measured MAP17 levels and seeded 10,000 cells to measure the number of tumorspheres (Supplementary Table S2). When plotted to establish one-to-one correlation, we found a strong direct correlation of tumorsphere numbers and MAP17 levels for each tumor (Fig. 5C, R² =0.8206). In these tumorspheres from primary sarcomas, we also measured the activation of the Notch pathway and the Stem cell–related factors, correlating their levels to the MAP17 levels. We observed a clear correlation between the MAP17 levels and the activation of the Notch pathway and the activation of the Stem cell–related factors (Fig. 5D).

All these data strongly suggest that MAP17 drives Notch pathway activation, tumorsphere formation, and activation of the stem cell transcription machinery.

To correlate MAP17 with Notch pathway and stem cell gene transcription, we performed meta-analysis of 28 different public databases of transcriptome from tumors, 6 from lung tumors, 2 from cervix, 10 from breast, and 10 from colon (Supplementary Table S1).

From all these databases, we analyzed correlation of MAP17 expression with Notch pathway genes and stem cell genes. Only genes with P < 0.05 were considered (Supplementary Figs. S5 and S6), and from then, only those genes with homogeneous behavior among all databases were taken further (Fig. 5E). We observed a clear set of genes positively (red) or negatively (blue) correlated with MAP17 expression either among the Notch pathway or the Stem cell gene machinery (Fig. 5E). Among these genes are the genes measured previously as end points or markers of the routes, clearly and robustly supporting the relation of MAP17 with the Notch pathway activation in vivo.

Finally, GSE34053 database, from colon tumors, have transcription of sorted CD133⁺ cells. We can observe that CD133⁺ cells carry higher MAP17 expression along with more HES1, OCT4, and SOX9 (Fig. 6A). We also observed a clear correlation between the expression levels of MAP17 and HES1 (r = 0.560), OCT4 (r = 0.828), SOX9 (r = 0.858), and CD133⁺ expression (Fig. 6B and C). Similar correlations were observed for lung adenocarcinoma TCGA database of (Fig. 6D and E), and breast adenocarcinoma TCGA database (Fig. 6F and G).

Because MAP17 is overexpressed in a high percentage of tumors, Notch pathway is activated in these cells increasing the percentage of cancer-initiating cells. This defines a new mechanism of Notch pathway activation and Stem cell pool increase that may be active in a large proportion of tumors.

Notch signaling pathway, a critical pathway governing embryonic development, is involved in the maintenance of tumor stemness and cancer metastasis. Increased activity of the Notch pathway has been reported in a variety of tumor cell lines and in tumors of different origin, including lung, colon, breast, and prostate tumors and sarcomas, melanomas, leukemias, and lymphomas (42–47). In these studies, Notch activity also appeared to participate in cancer metastasis by modulating the epithelial–mesenchymal transition (EMT), the tumor angiogenesis processes, and the anoikis resistance of tumor cells (48–50).

MAP17 is repressed in most tissues but is activated in tumors as a consequence of progressive demethylation (51) and/or oncogenic activation of the promoter (10). Therefore, MAP17 levels accumulate as a tumor grows. By sequestering NUMB, MAP17 allows nuclear localization of NICD and its transcriptional activation, increasing the stemness of cancer cells. By modulating Notch, MAP17 may also increase metastasis, contributing to the malignancy of these tumors. Our data clearly show that upon MAP17 expression, the Notch pathway is activated and, therefore, also the stemness of tumor cells. Because MAP17 is prevalent in approximately 50% of tumor types (7), our data suggest that half of tumors might have an activated Notch pathway, this activation being independent of the canonical protease activation of the receptor. Furthermore, our own data (Supplementary Fig. S7) suggest that these cells with activated MAP17 are more dependent on the Notch pathway activation, as these cells with high levels of MAP17 are more sensitive to the Notch inhibitor DAPT (Supplementary Fig. S7). Because of its small size and structure, MAP17 is unlikely a good target, but we have previously observed that MAP17-expressing cells are more sensitive to the proteasome inhibitor bortezomib (27, 52). It will be interesting to test proteasome inhibitors to study possible synergisms or antagonisms with different Notch inhibitors.

Upon interaction with a ligand, Notch undergoes two proteolytic cleavages that result in signal transduction activation (53, 54). These cleavages result in the release of active NICD and its nuclear translocation. Thus, NICD joins a multiprotein transcriptional complex, integrating MAML1 and CSL1 (55, 56), required for Notch-mediated transcriptional regulation. Notch signaling persists until NICD is phosphorylated by CDK8 and targeted for proteasomal degradation (57). NUMB promotes the degradation of NICD following receptor activation, targeting it for proteasomal degradation, and preventing its translocation to the nucleus acting as a tumor suppressor (20, 58). Therefore, sequestration of NUMB allows for an increase in NICD and its transcriptional activity.

MAP17 overexpression, a protein not present in most somatic cells, but commonly deregulated in tumors (7, 9, 10), binds and sequestrates NUMB, activating Notch pathway and allowing nuclear NICD translocation. Therefore, through this mechanism, MAP17-expressing tumors activate Notch-dependent pro-proliferative genes and repress Notch-dependent tumor suppressor genes (Supplementary Fig. S8).

Through our described NUMB sequestration by MAP17, a complete set of stem cell factors, such as SOX9, KLF4, NANOG, and OCT4, showed an increased transcription, increasing cell percentage with CSC-like physiology. MAP17 is frequently overexpressed in advanced stage tumors, correlating with higher levels of cancer-initiating cells. We can now suggest that these higher numbers of CSC-like cell populations are the consequence of Notch pathway activation caused by MAP17 expression.

In mammalian cells, NUMB acts as an adapter between NICD and the E3 ubiquitin ligase ITCH (21, 59). It has been proposed that NUMB intracellular levels may act as a molecular sensor that determines Notch fate and, hence, the responsiveness of cells to Notch ligands (20). Deregulation of this Notch switch by MAP17 overexpression may lead to unwanted deregulation of stem cell properties. Similar effects have been reported for the E3-ligase Siah1 and LNX. The E3-ligase Siah1 binds and ubiquitinates NUMB, leading to its proteasomal degradation (60). SIAH1 overexpression causes the relocalization of NICD from the cell surface to the nucleus, which is indicative of Notch activation. Similar results were obtained for the E3-ligase, LNX, which also targets NUMB for ubiquitin-dependent degradation (61).

In both breast cancers and NSCLCs, loss of NUMB expression is due to its exaggerated ubiquitination and ensuing degradation without genetic alterations in NUMB locus and normal mRNA levels (58, 62). The genetic lesion upstream of NUMB in these tumors remains unknown. It is tempting to speculate that MAP17 is responsible for these effects because a wide expression screening of E3-ligases in breast cancer did not reveal alterations immediately compatible with NUMB status of these cancers (63). Reduced NUMB levels also correlate with poor prognosis in salivary gland carcinomas, although no additional molecular or mechanistic details are currently known for these tumors (64). In breast tumors, NUMB expression is frequently lost (58). This event correlates with a poor prognosis (65), a less-differentiated phenotype compared with NUMB-expressing tumors (58, 65, 66), and CSC marker expression (66). This latter result is interesting in light of the recent finding that poorly differentiated breast tumors harbor a higher CSC-content than well-differentiated tumors (67). Furthermore, an in vivo RNA interference screen in a mouse lymphomagenesis model identified NUMB as a putative tumor suppressor whose ablation can accelerate the onset of lymphomas (68).

Therefore, the oncogenic function of MAP17 is easily explained at the biochemical and molecular level by the role of NUMB as a Notch pathway inhibitor. In the absence of NUMB, Notch pathway is activated, with pro-proliferative and antidifferentiative effects (20),

In summary, we report here that NUMB interacts with the C-terminal domain of MAP17. This physical interaction leads to an increase in nuclear NICD and consequent Notch pathway activation. As a result, MAP17-expressing tumor cells have an increased stem-related transcription factors along with a higher stemness. Increased MAP17 levels might therefore additionally contribute to the progression of cancer by increasing the plasticity of the CSC compartment and by resulting in higher rates of conversion from progenitors to CSCs.

No potential conflicts of interest were disclosed.

Conception and design: A. Carnero

Development of methodology: J.M. Garcia-Heredia, A. Lucena-Cacace, E.M. Verdugo-Sivianes, M. Pérez

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.M. Garcia-Heredia, A. Lucena-Cacace, E.M. Verdugo-Sivianes, M. Pérez, A. Carnero

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.M. Garcia-Heredia, A. Carnero

Writing, review, and/or revision of the manuscript: J.M. Garcia-Heredia, A. Carnero

Study supervision: A. Carnero

We thank Francisco Ramos (Genetics Department, Faculty of Biology, University of Seville, Seville, Spain) for all the materials needed to perform Yeast 2-Hybrid experiments.

A. Carnero's laboratory was supported by grants from the Spanish Ministry of Economy and Competitiveness, Plan Estatal de I+D+I 2013-2016, ISCIII (Fis: PI15/00045) and CIBER de Cáncer (CD16/12/00275), cofunded by FEDER from European Regional Development Funds (European Union), Consejeria de Ciencia e Innovacion (CTS-1848), and Consejeria de Salud of the Junta de Andalucia (FPS: PI-00-96-2014 and PI-0306-2012). The laboratory is also funded by the Fundacion BBVA.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.