The dark side of synaptic proteins in tumours

Abstract

Research in the past decade has uncovered the essential role of the nervous system in the tumour microenvironment. The recent advances in cancer neuroscience, especially the discovery of neuron–tumour synaptic/perisynaptic structures, have revealed the dark side of synaptic proteins in the progression of brain tumours. Here, we provide an overview of the synaptic proteins expressed by tumour cells and analyse their molecular functions and organisation by comparing them with neuronal synaptic proteins. We focus on the studies of neuroligin-3, the glutamate receptors AMPAR and NMDAR and the synaptic scaffold protein DLGAP1, for their newly discovered regulatory role in the proliferation and progression of tumours. Progress in cancer neuroscience has brought novel insights into the treatment of cancers. In the last part of this review, we discuss the therapeutical strategies targeting synaptic proteins and the current challenges and possible toolkits regarding their clinical application in cancer treatment. Our understanding of cancer neuroscience is still in its infancy; deeper investigation of how tumour cells co-opt synaptic signaling will help fulfil the therapeutical potential of the synaptic proteins as promising anti-tumour targets.

Introduction

Tumour cells are notorious for taking advantage of vital systems in every specific microenvironment for survival and progression. Recent studies have found that the nervous system plays a critical role in tumours [1]. Cancer neuroscience, which focuses on the interactions between cancer and the nervous system, has emerged as a burgeoning research field and brings new insights into our understanding of the basis of tumour initiation and progression [2, 3].

Crosstalk between the nervous system and tumours is ubiquitous (Fig. 1). Systematically, the nervous system exerts a global influence on tumour survival and growth via the regulation of the immune system and the secretion of circulating hormones and neurotransmitters [1, 4]. In the local microenvironment, interactions between neurons and tumours are usually tissue-specific, but a common phenomenon observed across several tumours is increased innervation. [5, 6]. The neural activities trigger the release of many pro-tumorigenic factors, such as glutamate [7–10], acetylcholine [6, 11, 12] and neuroligin-3 [13–15], from neurons, tumour cells or other cells nearby, promoting tumour initiation and growth. These cellular interactions usually occur between neurons and tumour cells, yet in some cases, endothelial and stromal cells are involved, mediating the neuronal effect on tumour cells [2, 16–18]. A newly discovered mechanism includes synaptic/perisynaptic integration of tumour cells into the neural circuits to promote cancer progression [7, 8, 19].

Fig. 1. Interactions between the nervous system and tumours.

a From a systematic view, the nervous system could influence tumour growth and development by regulating the immune response, angiogenesis, and the secretion of hormones and neurotransmitters, while the effects of tumours on the nervous system mainly include the invasion, destruction of the brain structures and the treatment-associated neurotoxicity; b The crosstalk between neurons and tumour cells in the tumour microenvironment mainly involves neuronal activity-related autocrine/paracrine release of pro-tumorigenic factors, tumour-induced neurogenesis and neuronal hyperexcitability, synaptic/perisynaptic communication and the propagation of electrical signals through tumour networks.

Tumours affect the structure and function of the nervous system as well (Fig. 1). One study has reported that non-metastatic breast cancer could act remotely to disrupt normal brain function of sleep via lateral-hypothalamic hypocretin/orexin neurons [20]. In the microenvironment, secreted factors from glioma cells could induce aberrant synaptogenesis and increase the excitability of surrounding neurons, which in turn facilitated the activity-dependent release of mitogens that drive glioma growth [7, 8, 14, 21]. Similar feedforward mechanisms have been documented in gastric and prostate cancers to promote tumour progression through the induction of axonogenesis [6] and neurogenesis [22], respectively. Furthermore, invasion of gliomas can disrupt neurovascular coupling at the tumour's infiltrative margins, causing various patterns of high amplitude discharges and seizures in mice models [23].

Moreover, cancer therapies can impact the nervous system, mainly in neurotoxicity. Chemotherapies may cause central and peripheral neuropathies, including cognitive impairment, sensory loss, motor weakness and pain [24, 25]. Figure 1 depicts the complex interactions between the nervous system and tumours from systematic and microenvironmental perspectives.

Several insightful reviews have summarised the multifaceted interactions between the nervous system and cancers and their clinical implications [1, 2, 26]. In all forms of neuron–tumour communication, the synaptic/perisynaptic structure between neurons and tumour cells represents a novel mechanism in which tumour cells adapt to co-opt the microenvironment to survive and thrive [27]. Once linked with learning and memory, synaptic proteins have attracted much attention for their previously unknown dark side in tumour progression. Based on our experience in the research area of synaptic signaling networks, we are here to provide an overview of synaptic proteins expressed by tumours, with emphasis laid on some of the critical proteins and their potential therapeutic values.

Synaptic proteins expressed by tumour cells

The expression of synapse-related proteins connoted that tumour cells acquire the ability to crosstalk with the nervous system. Many synaptic proteins have been found in several cancer types, especially neural tumours [7, 14]. Single-cell transcriptome analysis of major classes of adult and pediatric high-grade gliomas revealed the broad expression of postsynaptic receptor and scaffold genes in malignant glioma cells, and unsupervised principal component analysis highlighted a malignant subgroup, mainly consisting of oligodendrocyte precursor cell (OPC)-like glioma cells, in which synaptic genes were enriched [7]. It is already known that synapses can form between neurons and normal OPCs, and synaptic signaling plays an essential role in regulating the proliferation and cell fate of the OPCs [28–32]. These findings suggested possible synaptic communication between neurons and glioma cells, and this hypothesis was later validated by well-designed experiments [7, 8].

How similar are the neuron-OPC synapses and the neuron–glioma synapses? In terms of morphology, neuron–glioma synapses are pretty identical to neuron-OPC synapses and even the neuronal synapses, showing a presynaptic structure with docked vesicles, an elongated synaptic cleft with electron-dense material and a postsynaptic density area. Similar to the OPCs, glioma cells always took the postsynaptic position [7, 8]. Unlike other types of synapses, contacts between neurons and glioma cells can occur in three morphological forms: 60% of them showed a single synaptic connection on a glioma tumour microtube (TM), about 30% took the form of a multisynaptic contact to both a glioma TM and a neuron, and less than 10% present as a TM approaching the synaptic cleft of an existing neuronal synapse [8]. The functional characteristics associated with these different synaptic forms, if there are any, are yet to be determined.

Regarding molecular composition, transcriptomic analysis has shown that OPCs express more than 300 synaptic proteins, including receptors, ion channels, scaffold proteins, adhesion proteins and signaling molecules [33, 34]. Regarding glioma cells, single-cell gene expression data obtained in primary glioma and patient-derived xenografts showed extensive expression of glutamate receptor and postsynaptic structural genes in high-grade glioma cells, distinguishing the malignant tumour cells from non-malignant control cells. Looking at the synaptic protein expressed by OPCs [33] and gliomas [7], there were many overlaps, such as the relevant synaptic scaffold protein PSD95 (postsynaptic density protein 95), AMPAR (alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor) subunits and the cell adhesion molecular neuroligin family members. Differences were also found: for example, in OPCs, the expression level of NMDAR (N-methyl-D-aspartate receptor) subunits Grin2C and Grin2D were significantly higher than Grin2A and Grin2B, while in glioma cells, Grin2A and Grin2B were the mainly expressed subunits, and Grin2C was not present [7, 33]. Such differences warrant further verification and investigation, as they hint toward possible therapeutic targets which may act on tumour cells while sparing normal neuronal cells.

Since glioma cells always take the postsynaptic side in neuron–glioma synapses [7, 8], we sought to compare the glioma single-cell transcriptomic data reported by the Monje lab [7] with the postsynaptic proteomic data we retrieved from our previous studies [35]; the matched genes were classified according to their Gene Ontology annotations. Four hundred and twenty-three genes encoding postsynaptic proteins were found in the transcriptome of glioma cells, containing various membrane receptors and ion channels, scaffold proteins, cytoskeleton proteins and regulators (Fig. 2a). The top three functional groups containing most postsynaptic proteins were:

Fig. 2. An overview of synaptic proteins expressed in glioma cells.

a Functional classification of postsynaptic proteins expressed by glioma cells. We compared the glioma single-cell transcriptome data reported by the Monje lab [7] with our PSD proteomic data of the mouse prefrontal cortex [36], and the matched proteins were further filtered for postsynaptic location and classified according to the annotations of the synaptic protein database SynGO [85]; b Conjectural organisation of the postsynaptic structure of the glioma cells. Distinct colours represent proteins belonging to different functional groups, consistent with the functional groups in panel a.

Synaptic signaling and regulatory protein group (represented by family members of the guanine nucleotide exchange factors and GTPase-activating proteins);

Metabolism-related protein group (represented by dozens of ribosomal subunits);

Group of proteins functioning in synapse assembly and organisation (represented by scaffold proteins such as PSD95 and HOMER1).

In neuronal synapses, the existence of various scaffold proteins is regarded as one of the essential features of postsynaptic density (PSD). With multiple protein-binding domains, these proteins connect membrane receptors and ion channels with various enzymes, downstream effectors, cytoskeleton proteins and transporters, playing a critical role in synapse organisation and synaptic signaling. Glioma cells express a repertoire of neuronal PSD scaffold proteins, including members of the DLG (discs-large), DLGAP (discs-large-associated protein), SHANK (SH3 domain and ankyrin repeat-containing) and HOMER families, indicating that neuron–glioma synapses are likely to be a highly connected and ordered structure, similar to the neuron–neuron synapses. As little is currently known about the protein–protein interactions in neuron–glioma synapses, we referred to the PSD interactome data we obtained from neurons [35] and conjectured the organisation of the postsynaptic structure in glioma cells (Fig. 2b).

We noticed that a few postsynaptic proteins present in our mouse PSD dataset were not in the single-cell transcriptomic data of the human glioma samples [7, 35]. Many of these proteins are subunits of synaptic receptors and ion channels, such as the calcium voltage-gated channel subunit CACNA1C and CACNG2, the NMDAR subunit Grin2C and Grin3A, the hyperpolarisation-activated cyclic nucleotide-gated potassium channel HCN1 and the potassium voltage-gated channel KCNA1. Besides, the PSD scaffold protein DLGAP2 was not found in the glioma, either. Such differences may provide important hints for the development of targeted therapy; however, they must be treated with caution at the current stage, as those data were from two different types of experiments (proteomic analysis vs. single-cell transcriptomic analysis), and there were intrinsic abundance differences of postsynaptic proteins in human and mouse samples [36]. A direct comparison of the human glioma and its nearby normal tissue will be essential to determine the differentially expressed proteins; however, as the neuron–glioma synapses form at the invasion margins of the tumours, an accurate sample collection remains a technical challenge.

It is worth noting that the expression of synaptic proteins reflects the high degree of cellular heterogeneity typically found in gliomas. Many synaptic proteins are present in only a small population of glioma cells. Different subpopulations of glioma cells express distinct sets of synaptic proteins [7], echoing the discovery that only a subpopulation of glioma cells possess the ability to form synapses with neurons [7, 8]. The molecular heterogeneity is associated with the electrophysiological behaviour of gliomas. Among the several glioma models developed in the Monje lab, the SU-DIPGVI model had a more significant portion of OPC-like cells, expressed more synapse-associated genes and ion channels and generated more synaptic spontaneous excitatory postsynaptic currents (EPSCs) than other glioma models upon axonal electrical stimulation [7]. Besides EPSCs, spontaneous slow inward currents (SICs) were also found in human glioblastoma-neuron co-culture models and showed distinct responses to various ion channel blockers, suggesting their unique current compositions [8]. This difference in electrophysiological response is considered a consequence of human glioma cells' extensive molecular and anatomical heterogeneity, including the reported heterogeneity in the expression of AMPAR subunits in glioblastoma cells [8].

The discovery that the molecular heterogeneity of glioma cells correlates with their electrophysiological properties points out a new way for precisely targeted therapy. Increasing evidence has shown that tumour invasion, metastasis and recurrence are often driven by a small subpopulation of tumour cells [37]. In gliomas, neuronal electrical activities promote tumour initiation and progression [7, 8, 14, 15], so identifying and targeting those electrophysiologically active cells is vital for improved diagnosis, classification and treatment for gliomas. In this regard, expression analysis of synapse-related genes provides a simpler and more feasible alternative than electrophysiological tests. A comprehensive understanding of synaptic protein expression and interactions in subdivided glioma cell groups will help identify the most malignant and aggressive subpopulations and provide candidates for targeted intervention.

In addition to gliomas, synaptic proteins have also been reported to promote tumour development in other cancer types, although no neuron–tumour synapse has been found outside the central nervous system so far. Genetic deletion of stromal β2- and β3-adrenergic receptors in prostate cancer hindered the early phase of tumour development [17], and the β2-adrenergic-neurotrophin feedforward loop drove tumour progression in pancreatic cancer [38]. Scaffold protein DLGAP1 influenced the invasive tumour growth of pancreatic cancer through the NMDAR-DLGAP1-HSF1/FMRP axis, and the low-NMDAR activity transcriptomic signature was associated with favourable patient prognosis in several cancer types [39]. Moreover, researchers found that breast-to-brain metastatic tumour cells could form pseudo-tripartite synapses with neurons to drive tumour metastasis, a process highly associated with the expression of NMDAR subunit Grin2B and its interactor DLGAP1 [19]. Table 1 summarises some of the postsynaptic proteins involved in tumour growth and invasion, and a more detailed description will be provided for neuroligin-3, AMPAR, NMDAR and DLGAP1 in the following sections.

Table 1.

Postsynaptic proteins involved in tumour growth and invasion.

Gene symbol	Protein name	Subcellular locationa	Molecular functiona	Reported cancers
ADRB2	Adrenoceptor beta 2	Plasma membrane, Golgi apparatus, endosome	Adrenergic receptor; mediates the catecholamine-induced activation of adenylate cyclase through the action of G proteins.	Prostate cancer [17]; pancreatic cancer [38]
CACNG4	Voltage-dependent calcium channel gamma-4 subunit	Plasma membrane	Calcium channel subunit; regulates the activity of L-type calcium channels; regulates the trafficking and gating properties of AMPARs, including GRIA1 and GRIA4.	Gliomas [7, 86]
CNTN1	Contactin-1	Plasma membrane	Mediate cell-surface interactions; Involved in the formation of paranodal axo-glial junctions in myelinated peripheral nerves.	Gliomas [7, 87]
DLGAP1	Disks large-associated protein 1	Plasma membrane; cell junction;	Part of the postsynaptic scaffold in neuronal cells.	Pancreatic cancers [39]; Breast-to-brain metastasis [19]
GRIA2	Glutamate ionotropic receptor AMPA type subunit 2	Plasma membrane; endoplasmic reticulum	Receptor for glutamate that functions as ligand-gated ion channel in the central nervous system; plays an important role in excitatory synaptic transmission.	Gliomas [7, 87]
GRIA3	Glutamate ionotropic receptor AMPA type subunit 3	Plasma membrane	Receptor for glutamate that functions as ligand-gated ion channel in the central nervous system; plays an important role in excitatory synaptic transmission.	Gliomas [7, 39]
Grin2B	Glutamate receptor ionotropic, NMDA 2B	Endosome; cytoskeleton; lysosome; plasma membrane	Component of NMDA receptor complexes that function as heterotetrameric, ligand-gated ion channels with high calcium permeability and voltage-dependent sensitivity to magnesium.	Pancreatic cancers [39]; Breast-to-brain metastasis [19]
NLGN3	Neuroligin-3	Plasma membrane	Cell-surface protein involved in cell–cell interactions via its interactions with neurexin family members; plays a role in synapse function and synaptic signal transmission.	Gliomas [7, 13, 86, 87]

^aData are retrieved from the Uniprot database (www.uniprot.org).

Neuroligin-3

Neuroligin-3, encoded by the gene NLGN3, is a member of the neuroligin family of cell adhesion proteins located in the postsynaptic membrane of neurons [40]. As one of the fundamental synaptic organisers, neuroligin proteins link to many synaptic proteins and interact with the presynaptic neurexin proteins, thus promoting synapse formation and stabilisation and further influencing the balance between excitatory and inhibitory neurotransmission and synaptic plasticity [41].

In 2015, Venkatesh and her colleagues demonstrated that neuronal activity robustly promoted the growth and progression of high-grade glioma, with neuroligin-3 as the most potent glioma mitogen functioning in the microenvironment [14]. The protein acts through the classical PI3K-AKT-mTOR pathway to elicit the expression of a cluster of synapse-related genes, including neuroligin-3 itself, thus revealing a malignant feedforward loop for tumour progression. Further investigation by the same group demonstrated that neurons, oligodendrocyte precursor cells and glioma cells all contributed to the extracellular pool of neuroligin-3 together in an activity-regulated manner, and the sheddase ADAM10 was determined as the enzyme responsible for the cleavage of neuroligin-3 [15].

In the following study, the authors utilised phosphoproteomics to analyse the phosphorylation changes induced by neuroligin-3 and revealed that FAK (focal adhesion kinase) phosphorylation was an earlier event upon the neuroligin-3 exposure in glioma cells, placing FAK activation upstream of the PI3K-AKT-mTOR axis [15]. Besides, additional phosphorylation events downstream of FAK were also detected, including the elements of SRC kinase cascade and SHC-RAS-MEK-ERK cascade, suggesting the mitogenic effect of neuroligin-3 was mediated by the orchestration of multiple signaling cascades [15] (Fig. 3). However, the direct interactor of neuroligin-3 to elicit the cascades remains to be determined.

Fig. 3. Extracellular soluble neuroligin-3 acts as a mitogen to promote glioma cell proliferation.

The ectodomain of neuroligin-3 is cleaved by ADAM10 and released from neurons, OPCs or glioma cells into the microenvironment, where it functions as a stimulus to activate the Fak and downstream pathways in glioma cells, inducing the expression of many mitogens and synaptic proteins, including neuroligin-3 itself.

In the meantime, the discovery that neuroligin-3 upregulated the expression of a cluster of synaptic proteins in high-grade gliomas sparked the hypothesis of the possible link between synaptic signaling and tumour growth. Efforts in this direction led to the significant finding that glioma cells could co-opt the neural circuits to drive tumour growth and invasion by forming structural and functional synapses with neurons [7]. Microenvironmental neuroligin-3 played a critical role here by contributing to the formation of neuron-to-glioma synapses, besides its function to stimulate the canonical oncogenic pathways [7]. This finding further highlighted the fundamental role of neuroligin-3 in glioma pathophysiology, raising new questions such as the mechanism underlying the pro-synaptogenesis effect of extracellular neuroligin-3 and how the cleavage process is regulated to achieve an optimal balance between self-retention and secretion into the surroundings.

The latest progress in this series of excellent work is the finding that increased optic nerve activity-dependent secretion of neuroligin-3 was associated with the initiation and progression of Nf1 (neurofibromatosis type 1)-driven optic pathway in gliomas [13]. Besides, neuroligin-3 was also involved in neuroblastoma proliferation and growth [42], recurrence of glioblastoma [43] and progression-free survival in patients with diffuse glioma [44]. These findings suggested that gliomas widely exploit the neuroligin-3 signaling to drive tumour progression. In-depth investigations of the detailed mechanisms, especially the determination of the specific binding partner of the microenvironmental neuroligin-3, are warranted to fulfil its potential as a promising therapeutic target in gliomas.

AMPAR

AMPARs are among the most intensely studied glutamate receptors in the brain. The core part of AMPARs are tetramers made up of different combinations of the four subunits GluA 1-4, which constitute a glutamate-gated cation channel allowing sodium and potassium ions to pass through [45]. The subunit composition of AMPAR determines its permeability to Ca²⁺ ions. Generally, GluA1 and GluA4 are permeable to Ca²⁺; most mature GluA2 subunits in neurons would be impermeable to Ca²⁺ after an RNA editing process; some GluA2 subunits without mRNA editing remain permeable to Ca²⁺ [46]. AMPARs mediated most of the fast excitatory transmission in the mammalian brain. The regulation of their abundance, subunit combination, subcellular location and post-translational modifications are critical for the normal development and function of the brain [47, 48].

Many studies found that AMPARs are expressed in glioma and glioblastoma cells and play an essential role in cancer development. The GluA2 mRNAs in human glioblastoma tissues were substantially under-edited compared with those in normal brain tissues [49]. The expression of Ca²⁺ permeable AMPARs facilitated tumour cell migration and proliferation, and converting the Ca²⁺ permeable receptors to calcium impermeable ones decreased tumour cell motility and induced apoptosis [50]. Further investigation revealed that the calcium currents mediated by AMPAR promoted the phosphorylation of Akt at Ser-473 and activated the downstream pathways, thus facilitating the proliferation and migration of glioblastoma cells [51]. Another independent study using Ca²⁺ imaging experiments showed that Ca²⁺ permeable AMPARs mediated the intracellular Ca²⁺ oscillations and the multiple downstream cellular processes [52]. In this case, specific blockers of Ca²⁺permeable AMPARs, like GYKI or Joro spider toxin, could inhibit tumour cell migration and invasion, while the general AMPAR inhibitor CNQX was ineffective [52]. A more recent study indicated that the MEK-ERK1/2 signaling was involved in the transcriptional modulation of the expression of Ca²⁺ permeable AMPAR subunits [53].

The pro-tumoural effects of AMPARs were attributed to paracrine/autocrine AMPAR signaling in the above studies. However, recent studies found that glioma cells, especially incurable high-grade gliomas or glioblastomas, also sought to co-opt the neural circuits synaptically and electrically. AMPARs functions in both synaptic and non-synaptic current transmission between neurons and glioma cells [7, 8]. In brain slices and glioma cell-neuron co-culture models, targeted patch-clamp recordings revealed the existence of EPSCs featuring AMPAR-mediated synaptic currents in postsynaptic glioma cells [7, 8]. Besides, AMPARs were also the primary contributor to the non-synaptic SICs found in glioma cells in a neuronal microenvironment [8].

The expression of AMPAR delivers a profound influence on the development of gliomas. Ca²⁺ permeable AMPARs are part of the calcium communication machinery of the glioma cells. The Ca²⁺ permeability-determinant GluA2 is broadly expressed and under-edited in gliomas, potentially rendering many AMPARs permeable to the calcium ions in glioma cells. These Ca²⁺-permeable AMPARs contribute to synaptic EPSCs and non-synaptic SICs, as these currents were partially decreased by the specific antagonist NASPM (1-naphthyl acetyl sperminethe) [7, 8]. Thus, the AMPARs promote glioma invasion and proliferation partly through the mediation of calcium response to neuronal activities, indicating the therapeutical potential of targeting those AMPAR-mediated calcium transients in clinical settings. Studies showed that the expression of the dominant-negative GluA2 subunit reduced the migration and growth kinetics of glioma cells, concordant with the findings that the pharmacological blockade of the AMPAR-induced calcium currents by the approved drug perampanel attenuated tumour proliferation in glioma mice models, making this anti-epileptic drug a promising anti-glioma drug candidate [7, 8, 52]

Furthermore, AMPAR mediates the electrical integration of glioma cells into the neuronal circuits, and optogenetic evidence has demonstrated that cell membrane depolarisation promotes tumour cell proliferation [7]. As glioma cells form connected networks through TMs and gap junctions, glioma currents can be propagated and calcium waves can be transmitted to a cluster of glioma cells, thus amplifying the tumour-promoting effects of neuronal activities [7, 8].

NMDAR

The NMDAR is a postsynaptic glutamate receptor and cation-channel permeable to Ca²⁺ ions and plays a critical role in neuron development and function [54]. In neurons, the NMDAR signaling can activate the Ca²⁺-dependent CaMK (calcium/calmodulin-dependent protein kinase) pathway and the Ca²⁺-independent MAPK (mitogen-activated protein kinase) pathway, and both pathways lead to the activation of the CREB (cAMP-responsive element-binding) transcription factor, resulting in the expression of several transcription factors, i.e. the early response genes [55, 56]. These genes further modulate the gene expression profile of the neuron and exert significant influence on neuronal growth, proliferation, survival, migration and synaptic plasticity [57, 58].

Mounting evidence has indicated that NMDAR signaling functions in many tumour types and cancer cell lines, including prostate cancer, gastric cancer, breast cancer, laryngeal cancer, lung cancer, pancreatic cancer, kidney cancer and glioblastomas [39, 59–65]. Generally, NMDAR expression and activation could promote cancer progression, while its antagonists dampen tumour growth and invasiveness [9, 61, 62, 64]. The transcriptomic profile associated with low-NMDAR activity was a predictive signature of favourable patient prognosis in a few cancer types [39]. However, the precise mechanism of NMDARs' tumour-promoting effect is still not fully understood. Some studies found that the activation of NMDAR and downstream MEK-MAPK and CaMK pathways led to CREB activation in tumour cells, analogous to what happens in neurons [64, 66]. CREB further elicits the expression of the early response genes, including the proto-oncogene cFos, via Top2β (topoisomerase IIβ) induced DNA double-strand breaks in their promoter regions [67, 68]. The breaking of DNA strands can facilitate the fast transcription of the early response genes, yet they also increase the instability of the genome and are considered a possible cause of tumorigenesis [68, 69]. Besides, there have been reports of other pathways and effectors in NMDAR signaling in different cancer types, including ERK1/2 in lung cancer [65] and MMP-2 in glioma cells [70], and crosstalk between NMDAR and AMPAR might also exist in glioblastoma [71].

Autocrine/paracrine release of glutamate act as extracellular stimuli to activate NMDAR in the above cases. However, in patients and mouse models of breast-to-brain metastasis, it is found that the expression and activation of NMDAR in tumour cells promoted brain metastasis and was associated with poor prognosis. However, the glutamate in the microenvironment was insufficient to trigger NMDAR signaling [19]. Activation of NMDAR took place at the invasive fronts of the tumour, and super-resolution microscopy revealed that the metastatic cancer cells exposed themselves to enriched glutamate supply by forming pseudo-tripartite synapses with neurons [19]. NMDAR and the resultant calcium signaling pathways in the tumour cells were thus activated by the synaptic release of glutamate, which promoted the colonisation and growth of the metastatic tumour in the brain [19]. An earlier study had reported that metastatic breast cancer cells connected with astrocytes via gap junctions in the brain [72]; yet this NMDAR-dependent synaptic signaling represented a novel mechanism adopted by metastatic brain tumours to functionally interact with the neuronal microenvironment for their growth and invasion.

DLGAP1

DLGAP1, also called GKAP, is a scaffold protein mainly located in the PSD of neurons. Our previous work, which focused on the PSD protein interaction networks, revealed DLGAP1 as one of the core scaffold proteins, serving as a high-connectivity hub in the PSD signaling network by linking glutamate receptors and the MAGUK-family members (PSD top-layer scaffolds) to the SHANK-family members (PSD bottom-layer scaffolds) and downstream effectors [35]. DLGAP1 knockout mice exhibit altered PSD arrangements and deficits in sociability [73]. Genetic variants of DLGAP1 have been implicated in a series of neuropsychiatric, neurodevelopmental and neurodegenerative disorders [74–76].

Genetic linkage analysis indicated that DLGAP1 might modify the progression of pancreatic neuroendocrine cancer [77]. Further investigation revealed that DLGAP1 modulates the invasive tumour growth by participating in the regulation of NMDAR pathway activity and identified FMRP and HSF1 as downstream effectors of the NMDAR-DLGAP1 axis [39]. This study also reported a multigene transcriptomic signature of low/inhibited NMDAR-DLGAP1 pathway activities, which predicted better survival of patients in many cancer types, including pancreatic cancers, brain cancers, kidney cancers and uveal melanoma [39]. The subsequent work of the same group showed that higher expression of DLGAP1 was associated with the more malignant subtype of breast cancer as well [19].

There has been a growing appreciation that DLGAP1 promotes tumour growth and invasion via its interaction with NMDAR. However, the precise molecular mechanism is still unknown. In one of our previous works, we studied the synaptic phosphorylation networks modulated by long-term potentiation in the mouse hippocampus, a process that required the participation of NMDAR. We found that DLGAP1 increased its structural association with the NMDAR signaling machinery upon long-term potentiation induction and exhibited increased binding with kinases and phosphatases [78], suggesting a possible regulatory role in phosphorylation-dependent activation of NMDAR. Whether DLGAP1 adopts a similar approach in tumours, regulating the activity of NMDAR and the resultant aggressive growth of tumours by recruiting and organising a cluster of kinases and phosphatases is a question worth further exploring.

Clinical Implications of synaptic proteins in tumours

The discoveries of synaptic protein's role in cancer bring new insights into cancer therapy. Recently proposed therapeutic strategies targeted the microenvironmental pro-tumorigenic factors and the synaptic communication between neurons and tumour cells.

Blocking the release of neuroligin-3

Studies have determined that neuroligin-3 as one of the secreted mitogens to promote tumour growth and the ADAM10 sheddase played a critical role in the cleavage and release of neuroligin-3 into the tumour microenvironment; ADAM10 inhibition by GI254023X abrogated all cellular sources of neuroligin-3 and attenuated tumour proliferation and growth in mice models bearing pediatric glioblastoma or diffuse intrinsic pontine glioma orthotopic xenografts [14, 15]. Lab [79] and clinical data (clinical trial NCT02141451) suggested that the ADAM10 inhibitors developed for clinical application were well tolerated by patients, supporting their use in the treatment of high-grade gliomas. ADAM10 inhibitor INCB7839 is currently undergoing clinical trial (NCT04295759) to determine its effect in children with recurrent/progressive high-grade gliomas. Further investigation is warranted to comprehensively evaluate the long-term effects of this therapeutical strategy on brain function.

Targeting the glutamate receptors on tumour cells

The pro-tumorigenic effects of glutamate have drawn attention to its receptors as possible anti-tumour drug targets. The discoveries that AMPARs drive glioma progression via participating in neuronal activity-induced glioma cell membrane depolarisation and calcium communication have highlighted the translational potential of targeting the AMPAR signaling in treating high-grade gliomas [7, 8]. While genetic perturbation of AMPAR signaling reduced glioma growth and invasiveness, lowering neuronal activity through anaesthesia or using AMPAR antagonist perampanel could also achieve a similar promising result [7, 8, 80]. The approved anti-epileptic drug perampanel is a noncompetitive and highly selective AMPAR antagonist, inhibiting AMPA-induced calcium influx, and has shown potential anti-tumour effects in glioma patients [80]. In xenografted glioma mice models, perampanel treatment had decreased the proliferation of both pediatric glioma and glioblastoma cells [7, 8]. In 2020, researchers launched two clinical trials (NCT04497142 and NCT04650204) to determine the effect of perampanel on peritumoural hyperexcitability and seizure frequency in patients with high-grade glioma, which would be a step forward for the clinical application of this AMPAR antagonist.

Another glutamate receptor NMDAR has also gained attention for its critical role in tumour growth and brain metastasis [19, 39]. Transcriptomic signatures of low or inhibited NMDAR activity have shown predictive of patient prognosis and therapeutic responsiveness [39], suggesting its potential clinical value in personalised precision medicine. Meanwhile, preclinical trials in pancreatic cancer mouse models indicated that targeting the NMDAR-DLGAP1 axis by NMDAR inhibitors MK801 and memantine might have the beneficial effect of delaying cancer progression [39]. However, caution should be taken when using inhibitors targeting NMDAR or other synaptic receptors on patients, as these proteins are essential for normal brain functions. Direct inhibiting the relevant synaptic signaling in tumour-affected brain areas may elicit severe neurotoxicity in the central nervous system [81]. A plausible solution to avoid neurotoxicity is to bypass the receptors and target the downstream effectors of the signaling pathways exclusively found in tumour cells. The prerequisite is to identify these tumour-specific targets first.

Conclusion and outlooks

The nervous system has increasingly been recognised as an essential player in cancer biology. Recent studies have revealed new forms and mechanisms of neuron–tumour interactions and provided a candidate list of potential anti-tumour therapeutical targets, in which many are synaptic proteins, for future preclinical and clinical research. However, our understanding of synaptic proteins' role in tumours is still in its infancy, with many questions waiting for answers. Are there other types of tumours that can form synaptic or perisynaptic structures with neurons? Are neuron–tumour synapses also present in the peripheral nervous system? What are the downstream effectors in the neuron–tumour synaptic signaling pathways?

However, a few challenges need to be addressed before answering these questions. Firstly, tumours, especially brain tumours, are notoriously known for their heterogeneity. Transcriptional data revealed highly heterogeneous expression of synaptic proteins among different tumour subtypes and within the tumours, and only the cells at the tumour invasive front form synaptic/perisynaptic structures with neurons. To unravel the molecular mechanisms of neuron–tumour synaptic signaling, we need to take full advantage of the fast-developing single-cell omics technologies to identify changes happening in specific cell types. Moreover, as the neuron–tumour synapses share many synaptic molecules with the neuronal synapses, investigation of the relevant molecular mechanisms might require subtle discrimination refined to even single-synapse level. Emerging synaptomic technologies, including cell-surface proximity labelling [82] and single-synapse resolution proteome profiling [83], can be utilised to map the molecular components of individual synapses, providing candidate targets that are only involved in neuron–tumour synaptic signaling.

The second challenge relates to the precise genetic manipulation of different cell types. The tumour microenvironment is a highly complex system with multiple cell types; gene editing and expression regulation tools targeting specific cell types are required to conduct functional studies of potentially essential proteins in tumour cells. In this regard, the recently launched the U.S. Brain Research Through Advancing Innovative Neurotechnologies (BRAIN) Initiative transformative project—precision access to brain cell types endeavours to provide an advanced toolkit to probe neural circuits and different brain cell types accurately [84]. With the increasingly rich resources of the human brain cell atlas and single-cell genetic and epigenetic data, the progress in the modification of viral vectors and nanoparticle delivery systems holds promise for precise access to distinct brain cell subpopulations [84]. The success achieved in brain cells can be readily adopted in the research of tumour cells in a neural microenvironment, paving the way for targeted therapy and shedding light on the treatment of the devastating brain tumours.

Author contributions

JL conceived the manuscript, JL, YX and HZ wrote the manuscript, YX and DW drew the figures, DW, YW and PL revised the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (No. 32101020 and 91849209), Natural Science Foundation of Shandong Province (CN) (No. ZR2020MC071 and ZR2019LZL001) and the People's Livelihood Science and Technology Project of Qingdao (CN) (No. 20-3-4-41-nsh).

Data availability

The statistical data for Fig. 2a are available upon request.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.