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

Different proteins that are used in a controlled manner in neurons are unleashed by malignant transformation. These neuronal markers are part of various families: nuclear, cytosolic, membrane and vesicular proteins, neurotrophic factors and their receptors, developmental antigens, guidance proteins (reviewed in [1, 2]). In the case of purely synaptic proteins, such as Neuroligins (NLGNs), the knowledge on their involvement in tumor biology is limited to their prognostic value [3].

The family of NLGNs is composed of transmembrane post-synaptic proteins of the CNS that function in the fine-tuning of the synaptic activity. We have previously shown that a member of this family, Neuroligin 1 (NLGN1), is expressed in endothelial cells and modulates angiogenesis [4‐7]. Colorectal cancer (CRC) is a multifaceted and highly heterogeneous disease both at the inter-patient and intra-tumoral levels [8]. While a few basic pathway alterations have been identified and serve as predictive markers of resistance to therapy [9], prognostic markers in CRC are largely missing [10].

Despite the use of chemotherapy, targeted therapy and immunotherapy, the prognosis for metastatic CRC patients remains dismal. Indeed, the spread of most tumors, including CRC, is the culprit for basically all cancer-related deaths, and the discovery of actionable targets in this process is highly needed. The metastatization process is guided by a long and incompletely understood sequence of events. We can postulate that such sequence starts at the so-called invasive front in the primary tumor where tumor budding, i.e. the detachment of single or small groups of cells (up to 5) from the tumor mass can represent a step towards lymphatic and hematogenous invasion [11, 12]. Tumor budding is today broadly presented in standardized CRC pathology reports, and its role as a risk factor of metastatic disease is widely accepted. Moreover, several efforts are being made to sub-categorize budding itself, to better stratify metastatic risk [13]. Mechanistically, the process may represent the observable correlate of an epithelial-mesenchymal transition (EMT)-related process in the tumor microenvironment of CRC [14, 15], as well as the first visible connection between the primary tumor mass and the final establishment of metastatic disease.

Intravasation is another complicated process that follows tumor budding in time. It requires the crossing of the endothelial monolayer and appears to be dependent on microvessel density and diameter of the vasculature, as well as the presence of associated cells (platelets, fibroblasts, macrophages, neutrophils) that play a role in promoting tumor cell intravasation. Molecularly, TGFβ, EGF, uPA/uPAR, MMPs, integrins, all appear to influence intravasation [16]. Hereafter, tumor cells enter circulation as either single cells or aggregates (that are referred to as ‘emboli’ or ‘microemboli’). Studies dating back to the 1970s have suggested that out of thousands of cells that enter the bloodstream, only a few develop into metastases [17] and that cells die within 1–2 days in circulation [18]. This is explained by the severe stress to which cells are subject: loss of adhesion, hemodynamic shear forces, and attacks of the immune system [19].

The ability to form a metastasis still depends upon exiting the vasculature (extravasation) and colonization of distant organs. Extravasation requires occlusion in a small capillary or bifurcation, or rolling adhesion to the endothelium in a larger vessel, alterations in the endothelial barrier with breakage of the inter-endothelial cell-cell junctions, and transendothelial migration to colonize the organ [19].

In the last part of the process tumor cells colonize the host organ. This is again a very inefficient series of events: infiltration, evasion of immune defenses, adapting to the new environment, and finally, growing to substitute the target tissue [20]. Overall, the metastatic process appears to be impaired by many hurdles, yet once established, it is seldom curable.

As mentioned above, one of the early events in the metastatic process is EMT. During EMT, polarized epithelial cells transform into migratory mesenchymal cells with invasive properties. A key modulator of EMT is the WNT pathway, which induces the expression of EMT genes. This pathway is extraordinarily complex and can be classified as canonical and non-canonical. In the canonical pathway, activated WNT signaling inhibits the degradation of β-catenin (β-cat) which can translocate from the cytoplasm to the nucleus and regulate transcription of many genes. The WNT proteins family activate surface receptors such as the 7 transmembrane FZD (frizzled) proteins and the LRP5/6 (low-density lipoprotein receptor). Upon this activation, Dishevelled (DVL), a scaffolding protein required for the stabilization of β-cat, is activated, causing the recruitment of the so-called “destruction complex” (constituted by Axin, GSK-3 β, CK1, Adenomatous polyposis coli -APC) to the membrane receptors. GSK-3 β is now incapable of phosphorylating β-cat which migrates from cytosol to the nucleus, there interacting with T cell-specific factor (TCF)/lymphoid enhancer-binding factor (LEF) to turn on the WNT target genes such as c-Myc, cyclin D1, and Cdkn1a [21].

Among the members of the destruction complex, APC is regarded as a CRC tumor suppressor that can be mutated both in the germline and at the somatic level. Germ-line mutations in the APC gene result in familial adenomatous polyposis (FAP), which is characterized by numerous polyps in the intestines, but mutations in APC have been also found in up to ~80% of sporadic carcinomas and adenomas [22]. One of the effects of APC loss of function through mutation is the destabilization of the destruction complex with consequent translocation of β-cat to the nucleus [22].

Because of its pervasive activity, the interest in WNT signaling has remained high since the discovery of the first mammalian WNT proto-oncogene almost 40 years ago. In the meantime, a plethora of drugs targeting WNT downstream effectors has been produced, unfortunately without clear-cut success in the clinic to date [23].

In this paper, we investigated the hypothesis that NLGN1 could mediate the aggressive/invasive behavior of CRC cells and increase the overall metastatization capacity. We studied its mechanism of action by evaluating its ability to affect the WNT/APC/β-cat pathway. Results show that NLGN1 is expressed by cancer cells of the primary tumor, by migrating “budding” tumoral cells, and by intravasated tumor emboli in human CRC samples. Moreover, NLGN1 promotes trans-endothelial migration in vitro, in vivo lung colonization, and metastatization in animal models, it induces β-cat translocation to the nucleus, and β-cat target gene expression.

Our study of NLGN1, along with its partner Neurexin, started with its role in the vascular system, including in tumor angiogenesis [7] and subsequently, based on some preliminary observations , we considered the possible “tumor autonomous” role of NLGN1. In this paper we show that NLGN1 a) is expressed by CRC tumor cells in vitro and in vivo, including high grade tumor budding single cells and tumor vascular emboli b) promotes crossing of an endothelial monolayer in vitro c) promotes lung invasion in the tail vein colonization assay and metastatization in the CRC orthotopic mouse model d) anchors the tumor suppressor APC to the membrane, and physically interacts at least with some isoforms of it, e) stimulates β-cat translocation to the nucleus and upregulates mesenchymal markers and WNT target genes f) anchors the WNT signaling protein modulator CXXC4 to the membrane g) induces an “EMT phenotype” in CRC cell lines. To reach these goals, we used both cells that express high levels of NLGN1 physiologically, and cells that were exogenously induced to express NLGN1.

Our results in CRC clinical specimens and cell lines, along with those previously published by Qian et al. [3], consistently show that NLGN1 is expressed in colorectal tumors and may have prognostic implications. Notwithstanding the small percentage of CRC cell lines and clinical samples that display NLGN1 overexpression in our cohorts, we believe that this small subset deserves further characterization. Indeed, NLGN1 may represent a determinant of colorectal metastatic progression in individual patients whereby it could eventually be considered as a novel therapeutic target. In this regard, we note that precision oncology is effective in small subset of patients with colorectal tumors harboring infrequent molecular alterations. This is exemplified by the occurrence of ERRB2 gene amplification, or KRAS G12C mutations, which affect less than 5% and 3% CRCs [9]

Our work is the first to demonstrate that NLGN1 immunohistochemical expression is sustained in high grade tumor budding single cells and in lymphovascular emboli (Fig. 1G, H, I). This is particularly relevant since lymphovascular embolization is a crucial phenomenon, but due to its ephemeral nature, it is difficult to detect on histological sections. Certainly, studies on larger cohorts are needed to validate NLGN1 as a routine marker of tumor budding/EMT and lymphovascular invasion. Regarding the in vitro activity of NLGN1, few reflections must be made. The disruption of endothelial junctions to allow cancer cells to cross the endothelium (TEM) is required for both intravasation and extravasation, even though the two processes are essentially different because the cancer cells approach the endothelium from highly different microenvironments. At the cellular level, our in vitro model of TEM is set up so that cancer cells encounter first the apical side of endothelial cells, mimicking extravasation, in line with the tail vein assay, which shows an increased ability of NLGN1 expressing cells to extravasate and colonize the lungs. However, NLGN1 could also promote intravasation, as it is expressed by single cells in high-grade tumor budding and intravasated emboli (Fig. 1) and it may promote EMT (a necessary step for primary tumor spread and intravasation), as seen by β-cat nuclear translocation and gene expression modulation. Finally, the orthotopic model (Fig. 4) recapitulates all the stages of metastatization and, altogether, we can infer that NLGN1 may be important for both intra and extravasation.

Our next step was to analyze the ability to diffuse and colonize distant organs of CRC cells upon NLGN1 modulation. The tail vein assay has limits as a metastatization assay (tumor cells do not diffuse from a primary tumor but are injected in large numbers into the circulation) but it can be considered an in vivo counterpart of the TEM assay that also evaluates the capacity to colonize an organ. The orthotopic model of CRC presents otherwise the natural progression of a tumor to metastases. Our results show that while the growth of the primary tumors was unaffected, metastatization at particular sites was modulated by NLGN1. The metastatization sites affected by NLGN1 appeared to be different than the liver and lungs, two of the most common sites of CRC metastasis in humans. Given the extreme complexity of the metastatic process, it is impossible to infer on the reasons for the metastatic behavior of our cell models. However, given the vast heterogeneity of CRC, it does not appear surprising that different cell lines may respond differently to NLGN1 modulation in terms of the metastatic site that they colonize. In agreement with this notion, Hugen et al. [38] provide a retrospective review of pathological records of 5817 patients diagnosed with CRC and disclose large differences in metastatic patterns that depend on the histological subtypes and the localization of the primary tumor. Moreover, Riihimäki et al. [39] provide an epidemiologic study of metastatic colon and rectal cancer and reveal the same heterogeneity in the localization of metastatic foci.

Concerning the molecular mechanism of action of NLGN1, we initially evaluated the notion coming from one report from the neurobiology field that APC is required for localizing NLGN1 to neuronal nicotinic synapses [31]. While this function at the synapse may be mechanistically different than that exerted in CRC cells, APC, because of its large size and multi-domain organization, is also a key component of the WNT signaling, which has been the object of high interest in oncology since the first mammalian WNT proto-oncogene was discovered [40].

Hence, we first studied the NLGN1/APC membrane localization and NLGN1/APC physical interaction, which supported the result in neurons. Secondly, we studied the nuclear expression of β-cat, the hallmark of active canonical WNT signaling, that is predominant at the invasive front of carcinomas in dissociated, dedifferentiated tumor cells that are at a tumor-host interface and have undergone EMT [41]. As a final proof of NLGN1 involvement in the APC/β-cat pathway we demonstrated that this protein modulates β-cat target genes expression, including EMT markers, while promoting an “EMT phenotype”. Regarding the APC/β-cat pathway it is worth to consider that we used both cells carrying a native form of APC (HuTu 80, SNU-C2A, NCI-H716, HCT116 ) and cells that carry a mutant oncogenic APC (HCT8 cells, carrying missense and non- sense mutations, and HT-29 cells, carrying non-sense mutations, see Supplementary Table S3). In the first group, APC is able to fully participate in the function of the “destruction complex” [22], and in light of our whole set of experiments, NLGN1 appears as an upstream WNT canonical pathway modulator. For the second group, the activity of NLGN1 in the context of mutant oncogenic APC requires some more considerations. First, no noticeable differences in the biological effects of NLGN1 in vitro or in vivo was present between native or mutant APC-carrying cells, APC cortical localization was still induced by NLGN1 in mutant APC bearing cells (our antibody recognizes the mutant forms in HT-29, HCT8, see results section), and in HT-29, NLGN1 co-immunoprecipitated with a band of about 95-100 kDa that is a potential product of the APC mutant E853*. Although lacking mechanistic details, our data point to the idea that NLGN1 and mutant APC could work in synergy, whereby the activities of the APC mutants and NLGN1 would cooperatively increase the aggressivity of cells. Indeed, although mutations in APC are classically thought to lead to a constitutively active WNT pathway that renders cells insensitive to upstream regulation, some experimental data challenge this idea. For example, the WNT antagonists secreted Frizzled-related proteins (SFRPs) are able to inhibit Wnt/β-cat signalling in colorectal cancer cell lines carrying APC mutations [42, 43]. Furthermore, previous reports [44, 45] show that in different cell lines, including HT-29, regardless of the extent of the truncation, the APC variants still promote a significant recruitment of β-cat and members of the destruction complex, supporting the hypothesis that the biological effects of NLGN1 in mutant APC carrying cells could still go through the APC/β-cat pathway. It should be considered, for the prosecution of the project, that besides the role in β-cat translocation to the nucleus, NLGN1/APC could affect other cellular functions related to migration/metastatization. Studies have indeed suggested that APC plays roles in cell adhesion and migration and organization of the actin networks through varied mechanisms, possibly independent of the canonical pathway [22, 46]

One final comment on our data can be made concerning the role of neuroligin-3 (NLGN3) in glioma growth. This protein is a member of the NLGN family and is closely related to NLGN1 in terms of structure (72% aa identity and 81% conservation) and function (as they are both present at excitatory synapses). Neuronal activity stimulates NLGN3 extracellular domain shedding into the tumor microenvironment by the protease ADAM10 (A Disintegrin and Metalloprotease 10), thus inducing glioma growth [47]. ADAM10 represents a promising therapeutic target and a clinical trial (#NCT04295759) is ongoing for the inhibition of ADAM10 in glioma with the drug INCB7839 (https://​clinicaltrials.​gov/​). Very interestingly, NLGN1 also undergoes activity-dependent ectodomain shedding by ADAM-10 in the brain [48]. Hence, future research efforts may investigate whether soluble NLGN1 could play a functional role in mediating CRC growth and metastatization.

We have unveiled NLGN1 as a novel modulator of CRC aggressiveness. We believe that our findings, although do not describe a reciprocal influence between cancer and the nervous system, could be considered part of the rapidly developing “cancer neuroscience” field [49]. Indeed, the expression of neuronal markers in cancer cells facilitates nerve-cancer cross-talk through sharing of common molecular codes, and this perspective is certainly a major possible development of the project.