The neuronal protein Neuroligin 1 promotes colorectal cancer progression by modulating the APC/β-catenin pathway

NLGN1 expression in colorectal cancer

We initially analyzed the expression of NLGN1 in a public CRC gene expression database (TCGA, PanCancer Atlas- Colorectal Adenocarcinoma) through the cBioportal platform [26, 27]. High NLGN1 expression was present in a small percentage of samples (3.5%, z-score threshold 1.6) (Fig. 1A). Survival analysis to generate Kaplan Meier, and log rank test to estimate significance, demonstrated that NLGN1 overexpressing patients had a significant reduction in both overall survival and progression free survival (p-values: 1.838e-3 and 0.0339 respectively, Fig. 1B, C). We thus investigated the expression of NLGN1 in CRC at the protein level by immunohistochemistry (IHC) on tissue microarrays (TMA) and on a subset of colon cancers available in our Institution. The specificity of the used antibody (Neuromab -clone N97A/31) for NLGN1 was verified on brains of NLGN1 and NLGN2 -null mice (Supplementary Fig. S1). The analysis on TMA (Fig. 1D-F) showed that NLGN1 was expressed by tumor cells as follow: about half of the TMA’s cores (75, 52%) were positive for NLGN1: 5 with score 3 (3,3%), 21 with score 2 (14%) and 49 with score 1 (32,7%). Sixty-nine cores were negative (48%) and 5 were not assessable. Histograms in Fig. 1D and E present respectively the percentage of NLGN1 positive cases for each tumor subtype and the distribution of NLGN1 positive cases among different grades of the adenocarcinoma subset (with grade 2 as the most represented).

Fig 1figure 1

NLGN1 expression in human colorectal cancer. A Tatsuky plot displaying the Z-Score distribution of NLGN1 in CRC primary tumors (TCGA, PanCancer Atlas). Red dots indicate the samples with Z-Scores above 1.6. B, C Kaplan-Maier plots showing that higher NLGN1 expression predicts worse overall survival (B) and disease free survival (C) in CRC patients from cBioportal. P value was calculated as logrank test. D-F IHC was performed to evaluate NLGN1 expression on a TMA of colorectal cancer (CRM1505 - Biomax), which contains the following specimens: 88 cases of adenocarcinoma, 29 mucinous adenocarcinoma, 10 tubular adenocarcinoma, 4 tubular adenocarcinoma partly mucinous, 6 tubular+papillary adenocarcinoma and 4 adenosquamous adenocarcinoma, single cores per case. D Distribution of NLGN1 protein expression according to the IHC positivity (present or absent) in terms of pathology diagnosis. NLGN1 is mainly expressed in human adenocarcinoma. E. Distribution of NLGN1 protein expression among cores from human adenocarcinoma according to the IHC score, as present (1, 2, 3) or absent (0). NLGN1 is expressed at a higher percentage and intensity in cores from tumor grade 2. F Representative images of NLGN1 protein expression by IHC in CRC TMA cores and their relative score (original magnification 10X). G-I Representative images of NLGN1 expression in human CRC from a patient case study available in our institute. Immunohistochemistry reactions with antibodies anti-NLGN1 showed intense, diffuse positivity in a subset of colon adenocarcinoma. Moreover, NLGN1 positivity was even stronger in single-cell high grade budding (G) and in vascular neoplastic emboli (H-I). Scale bar: 10 μm

We next focused on the possibility that NLGN1 was expressed by the most infiltrative tumor cells by analyzing the invasive front of the tumor (tumor cell budding) and tumor emboli. Immunohistochemical evaluation of neoplastic lymphovascular invasion /embolism is a difficult task since emboli can be present in one histological section but disappear in the subsequent one, even if sub-seriated. To bypass this obstacle, we selected a subset of 16 CRC cases available at our institute that were described in the pathology report as positive for lymphovascular invasion. Of these 16 cases, 6 showed NLGN1 score 3 positivity, 5 were score 2 and 3 were score 1, while only 2 cases were NLGN1 negative. These numbers, which come from a subset of aggressive tumors, globally represent a much higher level of NLGN1 expression/positivity than that resulting from the TMA analysis discussed above. We then concentrated on NLGN1expression in high grade tumor budding single cells, assuming that those cells would be the same to penetrate the lymphovascular vessel and embolize. The overall analysis showed that NLGN1 positive tumors showed systematically strong staining in high grade tumor budding single cells (Fig. 1G) and lymphovascular emboli (Fig. 1H, I), proving our initial hypothesis.

NLGN1 promotes crossing of an endothelial monolayer in vitro

In order to dissect the cellular activities of NLGN1 we initially explored its expression in a database of gene expression profiles of our institutional CRC cell line collection (N=151) [28]. The analysis, through a c-bioportal interface [26, 27], showed that NLGN1 expression is very low or absent in almost all cell lines, except in 7 of these (NCI-H716, MDST8, SNU-C2A, COLO320DM, HuTu 80, SNU-175, SNU-503) that we labelled as “NLGN1 upregulated CRC cells” (Fig. 2A).

Fig 2figure 2

NLGN1 promotes transendothelial migration in vitro A Tatsuky plot displaying the Z-Score distribution of NLGN1 in CRC cell lines (dataset GSE59857). Red dots indicate the samples with Z-Scores above 1.6, p <0.05. B qRTPCR of NLGN1 expression level in CRC cell lines selected from the cell bank available in our institute. Fold-change is calculated with respect to HUVEC cells, known to express high endogenous levels of the protein. C Immunoprecipitation assay was performed on HUVEC, NCI-H716, HuTu 80 and SNU-C2A cells, using an antiNLGN (L067) antibody and immunoblotting was conducted with a monoclonal antibody (4C12) able to recognize NLGN1. The band detectable at 120-kDa corresponds to NLGN1. D-H Transendothelial migration of human CRC cell line in which NLGN1 is overexpressed, HCT8 (D), HCT116 (E) and HT-29 (F), or silenced, HuTu 80 (G) and SNU-C2A (H). Cell migration was recorded in real time through the X-Celligence System for 12 hour. Histograms show the cell index in terms of percentage relative to start, at the indicated time points. A lower cell index in this representation means an increased capacity of trans-endothelial crossing. Values are expressed as mean ± SD (n=3 independent experiments performed in triplicate). Two-way ANOVA with Bonferroni test: * p < 0.05, ** p < 0.01, *** p < 0.001

To confirm these data, we performed a qRT-PCR on cell lines from the database, thus isolating two groups of cells for our purposes: one in which NLGN1 was very low or absent (“NLGN1 null”) and another constituted by the “NLGN1 upregulated” mentioned above (Fig. 2B). Among the NLGN1 upregulated we chose three cell lines (HuTu 80, SNU-C2A and NCI-H716) for follow-up experiments and tested their NLGN1 protein expression by immunoprecipitation/western blot (Fig. 2C). Similarly, among the NLGN1 null cells we chose the cell lines (HCT8, HT-29 and HCT116) for the prosecution of the experiments. The identity of all cell lines was verified through the Cell ID system (Supplementary Table S1), while Supplementary Fig. S2 presents the data relative to NLGN1 expression modulation in all cell models.

Prompted by the presence of NLGN1-positive cells in tumor emboli (Fig. 1H, I) and given that crossing the endothelial monolayer is a crucial event for both intravasation and extravasation during metastatization, we evaluated the ability of NLGN1 to promote the process of trans-endothelial migration (TEM). Fig. 2D-F shows that three NLGN1 null cell lines -HT-29, HCT116, HCT8-, crossed the EC barrier with much higher efficiency when artificially overexpressing NLGN1. Conversely, two NLGN1 upregulated cell lines, SNU-C2A and HuTu 80, were impaired to cross the barrier when NLGN1 was silenced (Fig. 2G, H). To verify the specificity of the NLGN1 effects a “rescue” experiment was performed on HuTu 80 and SNU-C2A. Results show that overexpression of mouse NLGN1 (not targeted by the shRNAs) completely reversed the effects of NLGN1 downregulation (Supplementary Fig. S3).

These TEM data were obtained through an automated system (Xcelligence ) that evaluates the integrity of the HUVEC monolayer by measuring its impedance in real time as described [25]. While the effects of NLGN1 on TEM were specific, reproducible and statistically significant, their magnitude was not very large (Fig 2D-H). Hence, to corroborate these data we reproduced the TEM assay with the same architecture (a HUVEC monolayer was formed on matrigel and next challenged with tumor cells) but performed it in a culture well, and manually counted tumoral cells adhering, interacting and breaking the HUVEC monolayer. Results showed that NLGN1 increased TEM for HuTu 80, SNU-C2A, HCT116, HCT8 and HT-29 cells with a significantly larger amplitude than the readings of the Xcelligence system (Supplementary Fig. S4). Our interpretation of this result is that in the impedance based assay, tumoral cells do destroy the integrity of the HUVEC monolayer, initially lowering the impedance, but next tend to increase the impedance once they adhere to the plate [25] , blunting the overall magnitude of the effect.

To further characterize the biological activity of NLGN1 on tumor cells , in a parallel set of experiments we evaluated if NLGN1 affected the proliferation of CRC cells. Results (Supplementary Fig. S2, panels C, G, K, O, S) show how in HT-29 and HCT116 cells, NLGN1 overexpression reduced proliferation, in HCT8 and SNU-C2A no effects of NLGN1 were visible, while only in HuTu 80, NLGN1 appeared to slightly promote proliferation. Overall, the effects of NLGN1 on proliferation appeared inconsistent and heavily cell-line-dependent. This situation completely reflects the in vivo subcutaneous tumor growth experiments that we performed with the different cell lines (Supplementary Fig. S2, panels D, L, P, T).

Globally, the only consistent effect of NLGN1 on tumor cells was to stimulate their ability to cross the endothelial barrier in all cell lines used.

NLGN1 promotes lung invasion in the tail vein colonization assay and increased metastatization in the CRC orthotopic mouse model

To expand on above in vitro results, which focused on a single step of the metastatization process, we examined the in vivo extravasation/colonization capability of CRC cells, through the so-called tail vein colonization assay [29]. We injected luciferase-infected cells in the lateral tail vein of NOD-SCID mice and followed lung colonization through the IVIS recording system. As can be seen from the luciferase signal in mouse lungs, the NLGN1 null cell lines HT-29 and HCT8, dramatically increase organ invasion upon NLGN1 exogenous expression (Fig. 3A-D). On the other hand, NLGN1 upregulated cells, HuTu 80 and SNU-C2A, reduced their invasive capacity when NLGN1 was silenced (Fig. 3E-H).

Fig 3figure 3

NLGN1 modulates lung metastasis outgrowth in a murine tail vein metastasis assay. 1,5x106 cells (A-D) HT-29 or HCT8 cells, pLVX and pEZNLGN1, and (E-H) HuTu 80 or SNU-C2A cells, shCTRL and shNLGN1, further infected with a CMV-Luc vector were intravenously inoculated into the tail of 6 weeks old NOD/SCID mice. After 4 weeks mice were subcutaneously inoculated with 15 mg/ml luciferine 5 minutes before the sacrifice and the lungs were surgically excised. Luciferine bioluminescence was recorded through IVIS Lumina II apparatus. Images shown in A, C, E and G are representative of 5 mice. Graphs in B and D show the bioluminescence of tumor cells as total flux while graphs in F and H show the luciferin bioluminescence as % of the area occupied by the metastatic foci respect the total organ area, measured on the explanted lungs. Values are expressed as mean ± SD, n=5 (B and D), n=12 (F), n=20 (H). Mann-Whitney test, two tailed: * p < 0.05 **, p < 0.01, *** p < 0.001

We next moved to a more comprehensive assessment of the metastatic capabilities of tumoral cells upon NLGN1 modulation through an orthotopic CRC implant model [30]. This model avoids the problem of primary tumor development in an exogenous environment, typical of xenografts, and better recapitulates the whole process of tumor progression: from growth to cell release from the primary tumor, to the formation of foci at distant sites. Cells were inoculated in the cecum and tumor growth was weekly monitored by in vivo bioluminescence. At week 4, after mice euthanasia, organs were explanted. A significant increase in stomach metastases (Fig. 4C, D) upon NLGN1 overexpression was recorded for the NLGN1 null HCT116 cells. On the other hand, the NLGN1 upregulated SNU-C2A cells reduced their metastatic capabilities in the spleen/pancreas upon NLGN1 silencing (Fig. 4G, H). Primary tumor growth (Fig. 4A, B, E, F), as well as metastatization to liver and lungs (Supplementary Fig. S5), was unaffected in all cases by NLGN1, indicating a site-specific activity in the metastatic promotion.

Fig 4figure 4

NLGN1 induces metastasis formation by CRC cells in vivo. 1,5 × 106 cells (A-D) HCT116 pLVX and pEZNLGN1 cells and (E-H) SNU-C2A cells, further infected with a CMV-Luc vector were orthotopically inoculated into cecum of 6 weeks old NOD/SCID mice. After 8 weeks mice were subcutaneously inoculated with 15 mg/ml luciferine 5 minutes before the sacrifice and the cecum, intestin, stomach, spleen/pancreas, liver and lungs were surgically excised. Luciferine bioluminescence was recorded through IVIS Lumina II apparatus. Images of the primary tumor (cecum) in A and E andof the stomach (C) and spleen/pancreas (G) metastases are representative of 5 mice. Graphs in B, D, F and H show the bioluminescence of tumor cells as total flux. Values are expressed as mean ± SD, n=10. Mann-Whitney test, two tailed: * p < 0.05, *** p < 0.001

NLGN1 expression affects the APC β-cat pathway.

When exploring the intracellular pathways on which NLGN1 could impact, we first considered that the tumor suppressor APC is required for localizing NLGN1 to neuronal nicotinic synapses [31], implicating a functional interaction at the cell membrane of both proteins. APC is a large multi-domain protein with a wide array of functions, and, along with its upstream regulator WNT, is heavily involved in the pathogenesis of CRC [22, 32]. Interestingly, when we explored the co-expression of NLGN1 with a series of proteins of the WNT pathway (Supplementary Table S2) in both the TCGA, PanCancer Atlas- Colorectal Adenocarcinoma dataset, and our internal cell line expression database [28] we found that CXXC4 (a negative regulator of the WNT signaling pathway via its interaction with DVL), and FZD1 (a 7-transmembrane domain protein that is a receptor for WNT ligands) were significantly correlated with NLGN1, potentially suggesting their cooperation at the membrane and β-cat destruction complex levels.

On this background, we set to explore the interaction of NLGN1 with the APC pathway. We tested by immunofluorescence whether NLGN1 could modulate APC localization at the plasma membrane. The results show that APC preferentially localizes at the plasma membrane in HuTu 80 and SNU-C2A,which carry a native form of APC (Supplementary Table S3) and express high endogenous levels of NLGN1, while NLGN1 downregulation releases APC from the cortical region (measured as described in Supplementary Fig. S) into the cytoplasm (Fig. 5A-D). Conversely, “NLGN1 null” cells (HCT8, HCT116, HT-29) relocalize APC to cell membrane upon NLGN1 overexpression (Supplementary Fig. S7). It is important to note that HCT8 and HT-29 carry mutated forms of APC (Supplementary Table S3) that are nevertheless recognized by the anti APC antibody used in this experiment and raised against the N-terminus of APC (see discussion for comments on the role of mutant APC in NLGN1 activity).

Fig 5figure 5

NLGN1 modulates APC localization at the plasma membrane. Confocal microscopy analysis of NLGN1 and APC co-staining in HuTu 80 (A and B), and SNU-C2A cells (C and D). A and C Cells were immunostained with anti APC (magenta) and anti NLGN1 (red) antibodies. Images shown are representative of 1 out of three reproducible experiments. Scale bar: 20 μm. The graphs in B and D show the quantification of APC signal at the plasma membrane using the imageJ Software and values are expressed as mean ± SD (HuTu 80: n = 17; SNU-C2A: n = 15). Mann-Whitney test, two tailed: ** p < 0.01

In a separate set of experiments (Supplementary Fig. S8), we evaluated the existence of a physical interaction between APC and NLGN1 by co-immunoprecipitation experiments. Being APC such a large protein with different splice variants [33] and numerous mutated forms, the study of its pattern of appearance upon separation on a gel can be challenging, especially while analyzing its interaction with a new protein. Nevertheless, by immunoprecipitating NLGN1 from cells physiologically expressing the protein (the “NLGN1 upregulated” HuTu 80 and SNU-C2A cells), and blotting for APC, we obtained a pattern of APC bands highly similar to that visible when APC itself was immunoprecipitated and blotted, including the full length APC, as expected in cells carrying the native APC (Supplementary Fig. S8, panel B and C). When we immunoprecipitated NLGN1 from the APC mutated, exogenous NLGN1 overexpressing HT-29 cells, we detected different APC bands (Supplementary Fig. S8 panel D) including one that could theoretically correspond to the product of the non-sense mutation E853*, present in HT-29 cells (Supplementary Table S3). We were not able to detect NLGN1 after immunoprecipitation of APC (i.e. in the “reverse direction”). In our experience, this is often the case because of different reasons, namely the fact that an epitope may be hidden in the complex NLGN1- APC. An extended study involving more antibodies targeting different APC regions would be needed to understand this point. We believe nevertheless that our experiments provide proof of physical interaction between NLGN1 and APC.

Obviously, other protein participants to the destruction complex or other modulators of the WNT pathway, could interact with NLGN1. To approach this issue we analyzed the membrane localization of CXXC4 which as described above, is significantly co-expressed with NLGN1, by confocal immunofluorescence (Supplementary Fig. S9). Results show that in HuTu 80, HT-29, HCT8 and HCT116 cells, NLGN1 promotes CXXC4 membrane cortical localization, indicating a potential functional interaction between these two proteins, similarly to that seen with APC and described above (Fig 5, and Supplementary Fig. S7)

As explained in the introduction, activation of the WNT pathway normally causes the recruitment of APC and the other components of the “destruction complex” to the plasma membrane and stabilization of β-cat, which then translocates to the nucleus, where it promotes transcription of the WNT target genes [34]. Hence, we next tested if NLGN1 modulated β-cat localization in the nucleus. Immunofluorescence (Fig. 6A-D) and subcellular fractionation (Fig. 6E, F), clearly showed that NLGN1 downregulation in HuTu 80 and SNU-C2A reduced nuclear β-cat. A “rescue” experiment verified the specificity of this effects for NLGN1 (Supplementary Fig. S10).

Fig 6figure 6

NLGN1 influences the APC/β-cat pathway. A and C Confocal microscopy of nuclear β -catenin into HuTu 80 (A) and SNU-C2A (C) cells. Cells were immunostained with anti α-catenin (magenta) and anti β-catenin (red) antibodies. Nuclear β-catenin was calculated as indicated in the methods section. The images in A and C are representative of 1 out of 3 reproducible experiments. Scale bar: 20 μm. The graphs in B and D show the fluorescent intensity of nuclear β-catenin and values are expressed as mean ± SD (HuTu 80: n=40; SNU-C2A: n=70). Mann-Whitney test: ** p < 0.01, *** p < 0.001. E and F Western blotting analysis of nuclear β-catenin into HuTu 80 (E) and SNU-C2A (F) control (shCTRL) or NLGN1-downregulated (shNLGN1b and c) cells. Cell lysates were fractionated using the NEPER nuclear and cytoplasmic extraction kit (ThermoFisher) Immunoblottings were carried out using antibodies specifically recognizing β-catenin (95 kDa), β -actin (b 45kDa), as housekeeping, and GAPDH (37 kDa), as a control of the purity of the fractions, and the images are representative of 1 out of 3 reproducible experiments. Graphs in E and F show the densitometry of nuclear, cytoplasmic and total β-catenin, normalized on β-actin used as housekeeping. Fold-change is calculated with respect to shCTRL cells and values are expressed as mean ± SE (n = 3 independent experiments). Kruskal-Wallis test with Dunn’s posttest: *, p < 0.05 **, p < 0.01. G-I qRT-PCR of NLGN1, c-MYC, Cyclin D2, Twist (undetectable in NCI-H716), N-Cadherin, MMP2, L1CAM and LAMC2 expression level in HuTu 80 (G), SNU-C2A (H) and NCI-H716 (I) control (shCTRL) or NLGN1-downregulated (shNLGN1) cells. Fold-change is calculated with respect to shCTRL cells for each gene and values are expressed as mean ± SE (n = 3 independent experiments). Mann-Whitney test, two tailed: *, p < 0.05 **, p < 0.01, ***, p < 0.001

We next tested whether NLGN1 expression influenced the expression of WNT β-cat/TCF target genes including some commonly used markers of EMT. We demonstrated that c-MYC, Cyclin D1, Twist, N-Cadherin, MMP2, L1CAM, and LAMC2 expression levels in the NLGN1 upregulated cell lines HuTu 80, SNU-C2A and NCI-H716, decreased significantly upon NLGN1 silencing (Fig. 6G, H, I).

Prompted by the data on EMT markers being modulated by NLGN1, we performed a final set of experiments aimed at elaborating on the link between NLGN1 and EMT . Results showed that NLGN1 induced an EMT phenotype [35] (Supplementary Fig. S11) in HuTu 80 and HCT116 cells. This phenotype consisted of elongated spindle -like shaped cells, numerous isolated migrating cells, a different shape of the colonies in HCT116 cells, and in many instances the presence of evident actin stress fibers (a characteristic of mesenchymal-migrating cells, which are thought to mediate cell contractility in close cooperation with focal adhesions), as opposed to the epithelial -polygonal shape in cells devoid of NLGN1. EMT is highly relevant for TEM [36] and the mechanism by which NLGN1 induces TEM could rely on the activation of an EMT program, by which cancer cells adhere to the endothelium and cross the vessel wall, perhaps by forming pseudopodia and invadopodia, as it is the case , for example, for Vascular cell adhesion molecule 1 (VCAM1) [36]. EMT is also directly linked to tumor budding [15, 37] and the NLGN1-induced EMT phenotype could explain our observation of NLGN1-stained budding cells in human samples. Finally, NLGN1 promotes cell migration, a mesenchymal trait which is also linked to tumor budding [15], in HuTu 80 and HCT116 cells (Supplementary Fig. S11).

留言 (0)

沒有登入
gif