Tunneling nanotubes: A bridge for heterogeneity in glioblastoma and a new therapeutic target?

Abstract

Background

The concept of tumour heterogeneity is not novel but is fast becoming a paradigm by which to explain part of the highly recalcitrant nature of aggressive malignant tumours. Glioblastoma is a prime example of such difficult‐to‐treat, invasive, and incurable malignancies. With the advent of the post‐genomic age and increased access to next‐generation sequencing technologies, numerous publications have described the presence and extent of intratumoural and intertumoural heterogeneity present in glioblastoma. Moreover, there have been numerous reports more directly correlating the heterogeneity of glioblastoma to its refractory, reoccurring, and inevitably terminal nature. It is therefore prudent to consider the different forms of heterogeneity seen in glioblastoma and how to harness this understanding to better strategize novel therapeutic approaches. One of the most central questions of tumour heterogeneity is how these numerous different cell types (both tumour and non‐tumour) in the tumour mass communicate.

Recent findings

This chapter provides a brief review on the variable heterogeneity of glioblastoma, with a focus on cellular heterogeneity and on modalities of communication that can induce further molecular diversity within the complex and ever‐evolving tumour microenvironment. We provide particular emphasis on the emerging role of actin‐based cellular conduits called tunnelling nanotubes (TNTs) and tumour microtubes (TMs) and outline the perceived current problems in the field that need to be resolved before pharmacological targeting of TNTs can become a reality.

Conclusions

We conclude that TNTs and TMs provide a new and exciting avenue for the therapeutic targeting of glioblastoma and that numerous inroads have already made into TNT and TM biology. However, to target TMs and TNTs, several advances must be made before this aim can become a reality.

1. INTRODUCTION TO THE CONCEPT OF TUMOUR HETEROGENEITY IN GLIOBLASTOMA

Application of the term “heterogeneity” to glioblastoma is not new, as the disease was originally named glioblastoma “multiforme” by Bailey and Cushing in 1922 due to the highly heterogeneous gross presentation.1 The advent of the post‐genomic age has allowed the scientific community to discover the vast extent of molecular heterogeneity of this disease. In recent years, the classification of glioblastoma has evolved from a traditional histopathologic approach to a more molecular one based on driver and associated passenger mutations. This means that where once glioblastoma could only be classified by world health organization diagnosis criteria as glioblastoma or giant cell glioblastoma, it is now classified genetically with several different categories such as Isocitrate dehydrogenase (IDH) wild type (wt) and IDH mutant subtypes of glioblastoma.2 Tumour heterogeneity is a broad term that describes the variability seen both between tumour and within tumours with regard to several phenotypic and genotypic factors. However, this cursory definition only begins to describe the depth of heterogeneity present in glioblastoma.

Heterogeneity as a biological feature is highly contributory to the refractory nature of glioblastoma as detailed in great depth by Qazi et al.3 Simply put the highly heterogeneous expression of commonly dysregulated pathways in glioblastoma tumour cells from the same tumour means that while a single targeted therapy may be effective against one population of cells, the expression of several other oncogenic signalling pathways enables targeted therapy resistance.4 This was clearly exemplified by Reinhartz et al5 whom showed the single cell subclones derived from the same tumour exhibit marked heterogeneity in response to drug treatments and ergo must be considered separately for pharmacological intervention due to their difference in drug resistance. Heterogeneity is not only found in gene expression but also in the number and types of cells present in a given region of glioblastoma. High heterogeneity of both cancer stem cells (CSCs) and subsequently differentiated cells may give rise to new tumours despite effective treatment interventions.6 This concept has been extensively reviewed elsewhere7 and is well exemplified in the paper by Piccirillo et al, which reported two different types of CSCs in the same glioblastoma with strikingly different tumourigenicity and gene expression.8 Follow‐up papers from the Watts lab demonstrates that the validity of this assertion showing different tumour initiating cells (TICs) present in the same tumour has different tumour initiating capacities.9, 10 These different aspects of heterogeneity that are contributory to treatment resistance (refractory) and tumour reoccurrence lead to glioblastoma being a terminal disease.

Heterogeneity may be categorized in several ways; however, the two traditional overarching categories are intratumoural heterogeneity (which exists within a given individual tumour), and intertumoural heterogeneity, which exists between tumours of the same histopathologic type or subtype between different patients. More recently, however, with greater recognition that tumours evolve and adapt over time in response to therapeutic attacks on cell integrity, a new category termed “post‐therapy intratumoural heterogeneity” can also be proposed in order to reflect genetic and cellular alterations to tumour composition after diagnosis and treatment.3, 5, 11, 12, 13, 14 These different types of heterogeneity are summarized below in Figure 1.

Figure 1.

A summary of the key forms of heterogeneity in glioblastoma. Panels A, B, and C each represent differing composition of intact glioblastoma tumours. Strong intertumoural heterogeneity is seen between tumours (Panel A and B) from different patients, despite similar histopathologic diagnosis and identification of high‐grade (Grade IV) gliomas per current criteria. Within each tumour, there is significant intratumoural heterogeneity as represented by variable proportions of stromal components and genetic heterogeneity among the malignant tumour cells residing within the same matrix. The ability of glioblastomas to recur after attempt at safe maximal tumour resection and treatment leads to a completely different post‐treatment heterogeneity as represented by Panel D. Finally, the communication between cells by tunnelling nanotubes (TNTs) or tumour microtubes (TMs) (Panel C) represents a communication method that links the heterogeneous population of cells in glioblastoma

2. THE TYPES AND IMPLICATIONS OF HETEROGENEITY

There are different types of intratumoural and intertumoural heterogeneity contingent on the phenotypic or genotypic trait examined. With significant advances in next‐generation sequencing as well as imaging techniques in the past decade, the roles of genetic, cellular, and microenvironmental heterogeneity have more strongly been implicated in the pathogenesis and outcome of glioblastoma.15, 16

2.1. Genetic heterogeneity

Completion of The Cancer Genome Atlas (TCGA) and examination of the molecular profiles of glioblastoma have helped redefine this disease and characterize the genetic heterogeneity seen in glioblastoma.17 Specifically, this analysis resulted in the identification of several molecular subtypes of glioblastoma correlated to corresponding phenotypes termed here as “Verhaak subtypes”18; these primarily consist of proneural, mesenchymal, neural, and classical; however, further refinements to these classifications have been suggested.18, 19, 20 Studies utilizing the TCGA opened the door to understanding the extent of intertumoural genetic heterogeneity in glioblastoma.

Understanding of the complex intertumoural and intratumoural heterogeneity in glioblastoma was developed even further with the discovery that individual tumour may contain multiple subtypes. In 2013, Sottoriva et al reported their findings that different glioblastoma subtypes may be found within the same tumour.21 The authors achieved this by sampling 10 different patient‐isolated glioblastoma tumours and taking regional samples from each glioblastoma. When the regional samples from each of the 10 glioblastoma patients were examined using microarray analysis, it was found that 6 of the 10 glioblastoma samples exhibited two or more Verhaak glioblastoma subtypes. For contextual understanding, each glioblastoma subtype has a 210 unique gene set classifiers (thus 840 genes overall) with each gene set being unique18; this finding demonstrates the significant heterogeneity in gene expression in very closely linked spatio‐temporal regions.

This discovery is one of many that allude to the significant intratumoural heterogeneity present in glioblastoma and begs the question: What is the biological significance of intratumoural heterogeneity in glioblastoma? Qazi et al have written an excellent review on intratumoural heterogeneity in glioblastoma and its biological significance3 with the key conclusion that intratumoural heterogeneity may contribute to the recalcitrant and relapsing nature of glioblastoma. This conclusion indicates that understanding the contribution of intratumoural heterogeneity to glioblastoma pathogenesis is important.

2.2. Cellular heterogeneity

Cellular heterogeneity may be defined as the vast mixture of both tumour and non‐tumour cells that interact within glioblastoma22 and has been heavily implicated in the recalcitrant nature of glioblastoma.23 The cellular heterogeneity of glioblastoma has been evidenced by the numerous single‐cell RNAseq publications describing the variability of cell types in glioblastoma,4, 24, 25 which has been well reviewed elsewhere.26 Within these two categories of cells, there is a range of different cell types that contribute to glioblastoma pathogenesis independently.

Within the context of non‐tumour cells, differing presentations (ie. heterogenous gene expression) of the same cell contribute to glioblastoma pathogenesis. For example, differing subpopulations of astrocytes were shown to contribute to glioblastoma progression.27 Another example is the heterogeneous presentation of tumour associated macrophages (TAMs) and microglia, which together contribute to glioblastoma pathogenesis.28, 29 Most recently, the cross talk between these two cell types has alluded to another important aspect of glioblastoma pathogenesis that is conferred by cellular heterogeneity.30

However, the heterogenous presentation of tumour cells is arguably more important to glioblastoma. One of the most critical ways in which cellular heterogeneity is maintained is through the presence of TICs otherwise termed as glioblastoma CSCs (GSCs),31 which represent a subpopulation of glioblastoma tumour cells that can heterogeneously repopulate the tumour mass of glioblastoma after surgical removal of gross disease. The seminal work of Stoltz et al displayed the importance of TICs using a Sox‐2‐EGFP reporter system, which demonstrated, upon implantation, that TIC cells would give rise to a highly heterogenous tumour.23, 32 While able heterogeneously repopulate the tumour mass, different GSCs provide different starting points for the hierarchical structure of cells in glioblastoma.23, 33, 34 This concept is described in detail elsewhere,7 but a key paper from Iwama and Park showed tumours from different subclones derived from the same patient tumour sample result in completely different morphologies and self‐renewal capabilities.35 It is evident from the vast body of literature that cellular heterogeneity is important to glioblastoma pathogenesis, but how these numerous cell types communicate and interact is arguably a key, but understudied, feature of tumour heterogeneity.

2.3. Heterogeneity of the tumour microenvironment

While there are numerous different tumour cell types and genetic variants that contribute to glioblastoma heterogeneity, the tumour microenvironment (TME) in which these cells reside is also critically contributory to glioblastoma intratumoural heterogeneity.36 The TME can be defined as the aggregate of normal cells, stroma, blood vessels, and matrix,37 which when interacting with the tumour generate permissive tumoural growth conditions. TMEs can vary greatly throughout the different regions of glioblastoma. For example, the pseudopalisading cells surrounding necrotic regions are highly hypoxic38 in contrast with highly vascularized regions in glioblastoma, which are better oxygenated.39 Certain TMEs provide “niches” that give cues and support to particular cells such as GSCs40, 41 and are, as such, contributing to glioblastoma pathogenesis.42

The complex milieu of molecular and cellular intratumoural heterogeneity linked by the heterogeneous TMEs that these cells reside in paints a picture of a complex social network,43 but this leads to the logical question, what are the cellular modes of communication that facilitate this dynamic process?

3. MODES OF CELLULAR COMMUNICATION: IMPORTANCE TO TUMOUR HETEROGENEITY

If we were to consider the cellular intratumoural heterogeneity of glioblastoma to be a social network of different cell types communicating to propagate the tumour, then the different modalities of communication such as growth factor release, extracellular vesicles (EVs), membrane protrusions, and gap junctions (GJs) could be considered the various “social media tools” that are key to their communication and coordination.43 This section seeks to highlight different methods of cellular communication that can link heterogeneity with a focus on tunnelling nanotubes (TNTs).

3.1. Soluble secreted factor release

The secretion and transmission of soluble factors such as cytokines is a common way by which tumour cells may communicate with stromal cells and vice versa. Furthermore, communication among heterogeneous tumour cells has also been shown to be mediated by the secretion of soluble factors. Several publications from the Furnari lab, detailed subsequently, have demonstrated paracrine signalling systems between tumour and stromal or tumour and tumour cells through the secretion of interleukins (IL) that promote the growth of glioblastoma.44, 45 The key example of this from the Furnari lab shows that epidermal growth factor receptor (EGFR) wt and EGFRvIII variant cell types are both present in glioblastoma; the authors demonstrated that EGFRVIII cells could mediate the proliferation of the EGFRwt cells through secretion of IL‐6.46

3.2. Gap junctions

GJs are direct communicative points between cells that enable the transfer of cellular factors such as ions and secondary messenger molecules between adjacent cells. They have been heavily implicated in glioblastoma pathogenesis.47, 48 Structurally, GJs are composed primarily of connexin (Cx) lined channels; specific connexins such as Cx43 have been reported, controversially, to inhibit the growth of glioblastoma.49, 50 Connexins such as Cx43 are particularly salient to TNTs because they have been found to localize at the tips of the structures facilitating the transfer of intracellular contents.51, 52

3.3. Extracellular vesicles

EVs are membrane‐bound compartments released from cells that enable the transfer of intracellular contents between cells; their physiology is adeptly reviewed elsewhere.53, 54 EVs from glioblastoma have been reported to be important for maintenance of intratumoural heterogeneity55 and the transfer of intracellular components between cells that promotes tumour growth.56 Interactions between tumour and stromal cells via EVs are also reported, such as, between tumour associated mesenchymal stem cells (MSCs) and tumour cells.57 Specific transmitted factors of importance include microRNAs (miRNA),58, 59 immune suppressive molecules like programmed death ligand 1 (PD‐L1),60 and myriad other molecules reviewed elsewhere,61 which indicate that EVs are critical to glioblastoma progression through signalling.

3.4. Membrane protrusions: an emerging alternative route for intercellular communication in glioblastoma

The above forms of cellular communication have shortcomings when direct contact over a longer distance than two adjacent cells is required, as the above communication methods rely on secretion or direct contact. Thus, alternative methods of cell‐cell communication are required beyond GJs and EVs for cells to adequately interact and exchange signals at long range with optimal efficiency. One of the most recent discoveries in cancer biology was discovered approximately a decade ago. Long, thin, filamentous membrane protrusions acting as delivery tubes of intracellular cargo have been reported to be a more direct and effective method of tumour‐tumour and tumour‐stromal cell communication.62 The two most predominant forms of this membrane protrusions are TNTs51, 63, 64 and TMs62, 65, 66; these may be collectively termed as membranous tunnelling tubes (TTs).67

TNTs were initially discovered by Rustom et al. in 2004, dubbing them “TNT” described as “nanotubular highways”68 due to their ability to act as transport conduits for intercellular signals. They have since been physically characterized in numerous cell types, including cell types of interest to the field of general neuroscience as well as neuro‐oncology69, 70, 71 more specifically, rat astrocytes and C6 glioma cells,72 PC12 cells,73 and murine neuronal cells.74 The initial definitions of the physical characteristics were based on observations of morphology and structural features of TNTs in cancer and non‐cancer cell lines, in addition to their functional nature. These morphologic features include but are not necessarily limited to the following: approximately ≤1000‐nm diameter (thus the “nano” designation in “nanotubes”); the presence of predominantly filamentous actin; open‐ended (rather than closed at the tips), a unique feature that distinguished TNTs from other forms of membranous protrusions; and also, characteristically, the fact that TNT protrusions hover above the in vitro substrate without establishing attachment. Moreover, the transfer of intracellular content is a critical aspect of TNT function. TNTs have been reviewed by our group and also75 many others that have performed research in this growing area.63, 76, 77, 78 In addition, in the past several years, understanding and definitions of this type of cell protrusion have expanded considerably with the discovery and report of TMs, which were first described in an in vivo orthotopic animal model of glioblastoma by the laboratory of Dr. Frank Winkler at the University of Heidelberg. These TMs were found to be quite similar to TNT in function but morphologically and ultrastructurally were on average thicker (up to several micrometers in width), longer, and displayed more prolonged temporal periods of connection.66 Additionally, TMs were purported to transfer the GJ permeable dye sulforhodamine 101 inferring a functional connection like TNTs.66 TMs will be covered only briefly here as Chapter 8 of this special issue will encompass a deeper firsthand account of these findings by Drs. Osswald and Winkler.

One of the most intriguing aspects of TM biology that the Winkler lab has established was the characterization of TM density in differing glioma. The Winkler lab showed that there were fewer TMs present in oligodendrogliomas. Furthermore, they correlated the number of TMs with a genotypic marker, co‐deletion of the short arm of chromosome 1, and the long arm of chromosome 19 (1p/19q codeletion). This reduction of TM presence in oligodendroglioma was proposed to be due to the loss of genes salient to TM outgrowth such as Cx43 and GAP 43, which may be present in 1p/19q non‐codeleted tumours.62 As oligodendroglioma has a better prognostic outlook, this begs the question, do TTs promote a poor prognosis and does targeting of TTs provide a therapeutic avenue, whether by pharmacological targeting or simply as a prognostic marker. However, this finding may not be as straightforward or widely applicable as of yet, as while Winkler et al suggest an important role of Cx43 for TM formation, Yu et al50 have reported a downregulation of Cx43 in GSCs due to hypermethylation of the GJ protein a 1 gene. When Cx43 expression was upregulated in GSCs though a viral vector, orthotopic tumourigenicity was lost.50, 79 It is unclear why this discrepancy exists80; however, Osswald et al have stated that this discrepancy is possible and it was not clear, what the influence and function of Cx43 on TMs are.81 Further work would be required to account for this possible discrepancy.

It is pertinent at this point to note that the classification of TNTs and TMs is not fixed. Currently, these two structures serve, arguably, the same purpose with some heterogeneity with regard to physical parameters. However, currently, TNTs and TMs are regarded as different phenomena rather than the normal variance of a single phenomenon. In time, the terms used to describe TTs may change. This self‐correcting of the field may be required soon for TTs as it has so recently occurred for EVs with the publication of the 2018 minimal information for studies of EVs.82 In this publication, the authors outline the minimum information required for reporting as well as the appropriate recourse to characterizing EVs. One example of this was a biochemical approach to determining the expression of an outer surface EV marker. This was done by recommending four aliquots: (1) untreated, (2) treated with an enzyme that degrades the target surface EV component, (3) a detergent and enzyme combination to ensure total degradation of the target EV component within and outside the cell and, and (4) detergent only (to determine downstream nonspecific detergent effects). Subsequently, all four aliquots are analyzed by the salient technique, ie, by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS‐PAGE) or polymerase chain reaction (PCR) to determine if the surface EV component is degraded or not.82 Similar biochemical testing may be highly relevant for the characterization of TTs. The need to better understand the heterogeneity of TNTs has already been broached by some of the key researchers in the field of TNT biology in a list of outputs required to advance the field of TNTs51; we also suggest one mechanism by which this could occur, below.

The identification of TTs has occurred in glioblastoma in both in vivo and in vitro models (summarized in Table 1 below), but there are many questions yet to be answered. The publications in Table 1 suggest that TTs may have a role in glioblastoma pathogenesis. Below we discuss potential mechanisms by which these TTs may be involved in the heterogeneity of glioblastoma.

Table 1.

Select papers on glioma membrane protrusions

Title of Paper	Brief Description	TNT Model	Reference
Tunnelling nanotubes between rat primary astrocytes and C6 glioma cells alter proliferation potential of glioma cells	Establishment of an in vitro rat co‐culture model that demonstrated the tunnelling nanotubes (TNTs) were free floating in medium between two cells rather than adherent. H2O2 stimulated the production of TNTs through a P53 dependent mechanism.	In vitro co‐culture of rat glioma cells and astrocytes	Zhang and Zhang72
In vitro effects of cocaine on tunnelling nanotube formation and extracellular vesicle release in glioblastoma cell cultures	Low dose (150 nM) causes a nearly twofold increase in TNT formation over higher (300 nM and 150 μM) concentrations of cocaine.	In vitro culture of U87MG cells	Carone et al83
Structure and elastic properties of tunnelling nanotubes	Characterized presence of TNTs in U87MG cells using the Rustom characteristics and determined structural and elastic properties of the TNTs.	In vitro cultured U87MG cells	Pontes et al84
Exposure to ALS‐FTD‐CSF generates TDP‐43 aggregates in glioblastoma cells through exosomes and TNT‐like structure	Determined that cerebrospinal fluid derived from patients with amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) can induce the formation of TNTs	In vitro cultured U251	Ding et al85
Brain tumour cells interconnect to a functional and resistant network.	Specifically describes the presence of tumour microtubes in vivo in mouse glioma. Authors demonstrated the importance of tumour microtubes (TMs) in recovery after injury.	Primary cell lines orthotopically implanted into mouse cranium	Osswald et al66
Tweety‐homolog 1 drives brain colonization of gliomas	Tweet‐homolog 1 was reported to be expressed in and is important to TM generation in glioma but not expressed in TNTs		Jung et al86
Tunnelling nanotubes evoke pericyte/endothelial communication during normal and tumoural angiogenesis	Pericyte originating TNTs contribute to pathological angiogenesis in glioblastoma. Evidence of TNTs in glioblastoma tissue samples between endothelial cells and pericytes	Glioblastoma tissue samples	Errede et al87
Tumour microtubes convey resistance to surgical lesions and chemotherapy in gliomas.	Tumour microtubes likely confer resistance to conventional treatments of glioblastoma as well as assist in the reoccurrence of glioblastoma at the tumour resection margin	Orthotopic glioblastoma tumours from resected tissue samples	Weil et al65

In Figure 1 (above), we described three types of heterogeneity, while intertumoural and post‐therapy intratumoural heterogeneity are critical to the pathogenesis of glioblastoma, the significant intratumoural heterogeneity seen within the tumour (Figure 1A), which is depicted by the various tumour and stromal cells types in different areas of the tumour, requires a direct communication method. We hypothesize that TNTs form a communication path between both tumour‐tumour and stromal‐tumour cells in glioblastoma. This is depicted above (Figure 1C) as a TAM, an astrocyte, and two populations of tumour cells all interconnected by TNTs. While these interactions are theoretical, unpublished data have suggested macrophage‐tumour TNT interactions88 and published in vitro evidence has shown astrocyte‐glioma interactions72 and glioblastoma tumour‐tumour cell interactions,84 indicating the potential for such a paradigm. We postulate that TNTs may have a role in communication within the highly intratumourally heterogeneous glioblastoma and, as such, looked to other reports of TNTs in tumours that may allude to analogous roles for TNTs in glioblastoma; these are discussed below.

4. PUTATIVE MECHANISMS BY WHICH TNTs MAY HAVE A ROLE IN GLIOBLASTOMA HETEROGENEITY

4.1. TNTs in tumour‐tumour and stromal‐tumour interactions

The pathogenesis of highly heterogeneous tumours such as glioblastoma is multifaceted. The interaction between tumour cells and tumour associated stromal cells within a permissive TME is an important factor that mediates therapy resistance and the aggressive progression of glioblastoma. Numerous examples of communication through a non‐TNT‐based mechanism between tumour cells and stromal cells enabling the progression of glioblastoma suggest that this tumour‐to‐stromal cell communication is critical. Examples of this include the following: tumour‐astrocyte transfer of miRNAs through GJs, which facilitates invasiveness89; tumour‐associated MSCs transferring miRNA to glioblastoma stem cells (GSCs) via exosomes to promote tumourigenicity of GSCs57; and secretion of colony stimulating factor 1 (CSF‐1) from glioma cells in vivo to stimulate the recruitment of TAMs, which facilitates glioblastoma invasion.90 In addition, there are also reports of tumour‐tumour cell communication via GJs, exosomes, and soluble factor release that also facilitate glioblastoma progression.46, 91, 92 It is clear that tumour‐tumour and tumour‐stromal cell communication is important to tumour progression93 and that this communication can occur through several of the mechanisms described above, but this begs the question of whether these interactions occur in glioblastoma through TNTs.

Such interactions through TNTs have been reported in numerous other tumours. For example, B‐cells derived from acute lymphoblastic leukaemia have been shown to connect to MSCs with TNTs, thus promoting the secretion of pro‐survival cytokines.94, 95 Our group and others have previously demonstrated the transfer of intracellular contents including mitochondria and miRNAs between malignant cells in several different forms of cancer.96, 97, 98, 99 Key reports related to TNTs in glioblastoma are summarized below in Table 1.

TNTs have also been reported to transfer numerous forms of intracellular cargos between both tumour and stromal cell types. The breadth of potential cellular cargo transmitted from cell to cell via TNTs implicated a potential role for TNTs in the pathogenesis of cancers, but the further determination of the downstream effects is warranted before this conclusion can be solidified. Transferred molecules are wide‐ranging but include nucleic acids such as miRNAs, differentially expressed in chemoresistant forms of cancer,100 and whole organelles, most prominently mitochondria.73 TNT‐mediated intercellular transfer of mitochondria has been reported to rescue apoptotic cells73 and promote invasiveness in cancers.97 Pasquier et al demonstrated the transfer of mitochondria from endothelial cells and cancer cells via TNTs and have associated this transfer with the development of chemoresistance.101 It is clear from the vast number of papers, well‐reviewed elsewhere,77 that TNTs provide a transfer route for intracellular components; however, this has not been investigated in glioblastoma.

Separately to the transfer of intracellular contents, TNTs were recently hypothesized to promote pathological angiogenesis though connecting pericytes‐to‐pericytes as well as pericytes‐to‐endothelial cells.87 Through intravital microscopic imaging of an in vivo animal model of gliomas, TMs were also recently shown to correlate with the reoccurrence of tumours at surgical lesion sites.65 These reports suggest that TTs could have other roles outside of transfer of intracellular context, but critically are important for facilitating communication between different cells of the tumour.

These publications and many others illustrate the importance of TTs in connecting tumour cells to other tumour and stromal cells. Moreover, these observations have set the stage for the next set of questions: in a highly heterogeneous tumour type such as glioblastoma, what roles do TTs play in the early and later stages of pathogenesis? Even further, and most intriguing from a translational standpoint, is the ultimate translational question: Are these structures pharmacologically targetable, and would elimination or reduction of TT‐mediated communication have a significant effect on reducing invasion and viability of gliomas and other similarly invasive cancers?

4.2. TNTs in glioblastoma microenvironments

The TME of glioblastoma is made up by numerous cell types as previously described; however, it is critical to note that the presence of several anatomically distinct sites provide “niches” enabling the maintenance of glioblastoma heterogeneity.39 These niches have several roles including the maintenance of GSCs.40, 102 There are several overlapping niches within glioblastoma including the perivascular103 and perinecrotic niches,104 and within the niches are several factors that maintain GSCs and promote glioblastoma growth. These include secreted soluble factors such as vascular endothelial growth factor, fibroblast growth factor 2,105 and EVs61; reports of TNTs in the maintenance of these niches are few and limited to date despite there being clues in other cancers alluding to this possibility. The most critical report is from the den Boer lab who reported that TNTs maintain the leukemic niche through the transfer of autophagosomes to MSCs, thus stimulating cytokine release.95 The roles of TNTs in transferring cargo between MSCs and target cells have been reviewed thoroughly elsewhere106; however, it is important to note that MSCs can play a key role in the GSC niche,107 and this concept of MSC‐to‐glioblastoma cell transfer via TNTs has been suggested but not conclusively demonstrated.98

There is a growing body of evidence that the formation of TNTs occurs in response to microenvironmental conditions characteristic of the evolving TME. Hypoxia is a common hallmark of glioblastoma, in areas such as the perinecrotic niche.108 Our lab has reported on hypoxia promoting TNT formation in ovarian cancer cells.109 This hypoxia‐mediated increase of TNTs may also occur in the highly hypoxic environment typical of glioblastoma. The roles of TNTs in facilitating communication in the TME, and the implications for the development of potential novel therapeutic strategies in the clinical setting, have been discussed previously by our lab.75, 110 The evolution of the field of TNT biology from one of basic discovery to one with implications for human diseases is still up for debate, but with a groundswell of increasing data supporting the role of this form of cell communication in cancer, it is becoming increasingly conceivable that the effects in glioblastoma and other aggressive cancers may be disrupted for therapeutic gain.

5. LIMITATIONS TO THE FIELD OF TNTs

5.1. The continued search for TNT‐specific markers

The targeting of TNTs is not beyond the realm of possibility, but there are several substeps required before inroads can be made into determining if there is a potential for pharmacologically targeting TNTs.

One of the current holy grails in TNT biology is the identification of a unique marker of TNTs. While TNTs have been shown to have dominant features such as being predominantly made up of actin or largely being less than 1 μM in diameter, there is not one specific, measurable cell surface or endogenous cytoplasmic marker, although some candidates have been suggested thus far. One of the more promising markers is M‐Sec, which is also referred to as tumour‐necrosis factor alpha induced protein 2 (TNFaip2).111 In the context of TNTs, the Ohno lab has produced several publications regarding the central role of M‐Sec in nanotube formation,112, 113, 114 suggesting M‐Sec as a central mediator of TNT formation. Wang et al have also demonstrated induction of M‐Sec under TNT promoting conditions.115 However, it is unclear if M‐Sec is expressed in other TTs as Chen et a. demonstrate a more ubiquitous expression of M‐Sec in nasopharyngeal carcinoma cells.116 One approach to determine whether this is a putative universal marker would be to perform a high‐throughput confocal characterization of several cancer cell lines and confirmed using primary cancer cells, under TNT‐inducing conditions for M‐Sec (as well as other markers) to determine if expression is localized completely to TNTs. In the future, the identification of a unique marker or set of markers, such as M‐Sec,117 would be invaluable both for the identification of TNTs and then allowing the monitoring of TNTs under different treatment conditions or time points.

The predominant bulk of TNTs are made up of f‐actin,118 so another potential future source of markers for in vitro experiments could be within f‐actin binding molecules such as Drebrin, which has already been shown to co‐localize with f‐actin positive membrane protrusions.119, 120 Alternatively, another approach to determining unique markers of TNTs may be using mass spectrometry imaging, ie, MALDI IMS,121 or utilizing micro‐proteomic techniques, such as the Gousset lab is utilizing,122 to specifically examine the proteins present in TNTs and correlate this to the physical characteristics of TNTs. The proteomic analysis of TNTs will allow an understanding of the spatio‐temporal expression of the proteins in TNTs, thus enabling a better characterization of TNTs under treatment conditions.

It is plausible that a single solitary marker will never be identified, due principally to the molecular and cellular heterogeneity that distinguish different cancer cell types. Potential heterogeneity in TNT expression between cell types or tumour types infers that classification of TNTs by other mechanism(s) would be critical. The ability of cells to form TNTs can be measured by what we call a “TNT index,” reported as the average number of TNTs/cell or TNTs/100 cells; the TNT index could serve as a reference point to quantifying relative potential of any given cell line or cell type for forming TNTs and has been similarly discussed by other labs.123 This measure would have to be validated but if so, could be used to correlate other feature such as prognosis or malignancy of a tumour. The ability of different cell types to form TNTs between similar cells (homotypic) other cells (heterotypic) under specific culturing conditions may prove to be an important marker in the future. A TNT potential ratio could also be developed to quantify this. First, determining a series of markers such as previously described by Rustom (2004) and our lab117 would be required to identify TNTs. Briefly, these physical parameters included lack of adherence to culture substrate, TNTs connecting two cells, very narrow <1‐μM width, a narrow base at the extrusion site, and the addition of molecular characteristics such as predominant f‐actin composition and M‐Sec enrichment may also help. This could be combined with an experimental test such as culturing cells in TNT permissive and control media and determining the TNT index for each condition. The TNT potential ratio could then be determined by the following equation to express the ability of a cell to form TNTs.

$$\mathrm{TNT}\text{potential ratio}=\frac{\mathrm{TNT}\text{index in}\mathrm{TNT}\text{permissive media}}{\mathrm{TNT}\text{index formed in control media}}.$$

The development of a clear set of tools to identify and quantify TNTs and describe the changes in ability to form TNTs is an important step towards harnessing TNTs in a translational capacity. However, quantification is insufficient. The study of TNTs in salient models is also imperative to draw accurate conclusions on TNT biology.

5.2. The death of 2D cell culture models and the advent of 3D models and more representative in vivo models that recapitulate heterogeneity

Arguably, the most important future step to treating glioblastoma via TNTs is adopting more representative models of glioblastoma that accurately reflect its heterogeneity as well as the 3D communication of TNTs. This requires a shift from the traditional 2D culture approach currently used towards more 3D models.

Two‐dimensional (2D) in vitro cell culture is the most widely used culturing system, whereby cells are cultured in a flat layer on plastic dishes.124 Two‐dimensional in vitro cell culture models are commonly used to elucidate fundamental molecular interactions due to their ease of culture, reproducibility, predictability, and cost‐effective nature. Conversely, 2D in vitro models lack several features observed in situ or within in vivo models, including hypoxic cores and multicellular interactions common in stem cell and perivascular niches. Two‐dimensional models can further be extended into three‐dimensional (3D) models in which in vitro cells are cultured in 3D structures such as spheroids or using unique scaffolds.125 Recently, MSC spheroids have already been shown to have TNTs form between cells.126 This presents a cost‐effective and more salient model of TNT biology and is, in our opinion, an approach that would provide the stronger potential for more accurate assessment of TNTs in vitro, and one that could be universally adopted to ensure better reproducibility of results across cell types and between labs.

The interaction of TNTs between glioblastoma tumour cells and stromal cells can be investigated by using 3D models, such as multicellular spheroids.127 Multicellular spheroids have been used in a variety of ways, including the investigation of the interaction of endothelial and tumour cells128 and the assessment of tumour cell invasion.129 Spheroid tumour models have been reviewed completely elsewhere130 and represent an exciting direction in TT research.

A specific alternative and potentially even more promising 3D model is the 3D organotypic culture model that utilizes a slice of brain tissue embedded in Matrigel and embedded a tumour spheroid into it that replicates the microenvironment and multicellular interactions.131, 132

Of course, in vivo models would even more so accurately reflect the biology of the multicellular microenvironment of glioblastoma and could be used as demonstrated by the Winkler lab.62 One challenge to this approach, however, is that these animal models represent a greater cost and requirement of the necessary equipment and technical expertise for intravital microscopy. In addition, the reductionist system of in vitro 3D models represents a far more modifiable system by which to study TNTs.

Irrespective of which types of model are used, a shift from 2D to 3D cultures appears critical not only to image TNTs in a more in situ setting but also to investigate TNTs functionally as they transfer their cargo.133 This change towards 3D models also requires a shift in imaging methodology as thicker tissue sections and 3D models surpass the limitations of traditional light microscopy techniques; additionally, better resolution offered by newer microscopy techniques will enable clearer discerning of the structure of TNTs134 as well as the functions of TNTs, ie, in the transfer of cellular contents.135 The recent use of lattice light sheet microscopy to imagine TNTs in vitro and in vivo136 represents a step toward the adaptation of next‐generation microscopy techniques for TT research. In addition to these modalities, other advanced microscopy techniques that enable imaging of thicker 3D models include multiphoton microscopy, super‐resolution microscopy, and tissue clearing methodology that would be beneficial to adopt. The Winkler lab's use of intravital imaging66 and cranial windows in conjunction with multiphoton microscopy65 represent a new gold standard for TT biology. Finally, the Zurzolo lab recently produced a paper demonstrating the ultra‐structures of neuronal TNTs using novel using a novel correlative light‐ and cryo‐electron tomography (Cryo‐ET) microscopy combined with correlative focus ion beam scanning electron microscopy (FIB‐SEM) approaches. The paper answered several longer standing questions of TNT structure, including, significantly, that TNTs are bundles of open‐ended tubes held together with N‐cadherin containing strands.135 Thus, striking a balance between a representative model while still allowing cost‐effective research into TTs is imperative.

6. TNTs IN GLIOBLASTOMA: A THERAPEUTIC TARGET?

The potential for TTs (both TNTs and TMs) to serve as a therapeutic or prognostic target is exciting. The first effective correlation of molecular markers to TTs as cellular markers in the context of prognosis was the finding that TMs inversely correlated with co‐deletion of chromosomes 1p and 19q in oligodendroglioma; the implication that fewer TMs leads to a less aggressive tumour, and conversely that more TMs leads to a more aggressive and treatment‐refractory form of this disease, provides a basis for further exploration of TTs as putative predictive and/or prognostic biomarkers of glioblastomas and other malignant tumours.137 If oligodendroglioma exhibits less TMs, could TTs be used as a prognostic indicator of metastatic potential in other cancers? Once an appropriate marker of TTs is found, this would make correlating TT numbers to the grade, severity, or prognosis of tumours relatively trivial.

If we as a scientific community are unable to define a TNT marker, which is a highly plausible possibility, using a machine learning approach to be able to discern TNTs in images, thus allowing us to screen for compounds that are effective at inhibiting TNT formation could be a novel approach.138 The pharmacological targeting of TNTs in vitro has already been extensively performed with numerous papers published (summarized in Table 2 below); however, whether these compounds hold clinical relevance is yet to be determined. This requires the study of TNTs in in vivo and 3D in vitro models in conjunction with accurate measures of TNTs.

Table 2.

A review of the pharmacological inhibitors known to abrogate TNT formation

Drug	Target/Model	Reference
CK‐666	Arp 2/3 inhibitor on in vitro trabecular meshwork cells. *Has been recently reported to induce TNT formation.	Sartori‐Rupp et al and Keller et al135, 139 *
Everolimus, metformin	mTOR inhibitor ovarian carcinoma cells in vitro	Desir et al109
Cytochalasin B	Actin polymerization inhibitor in vitro MSCs and PC12 cells.	Jackson et al and Bukoreshtliev et al140, 141
Latrunculin B	Inhibition of actin polymerization and enhancing depolymerization in astrocytes in vitro	Rostami et al142
Cytarabine (AraC)	Inhibits DNA synthesis resulting in reduced TNTs in AML and HTLV‐1 cells	Omsland et al143, 144
Migrastatin core ether	Targets Fascin	Ady et al117

Abbreviations: MSCs, mesenchymal stem cells; TNT, tunnelling nanotube.

7. CONCLUSIONS

In the battle against highly recalcitrant tumours with dismal prognosis such as glioblastoma, novel targets such as TNTs are desperately needed. Heterogeneity is often considered to be a key reason for glioblastoma's recalcitrance, and TTs appear as key mediators of communication and promoters of tumour heterogeneity. With the advent of novel models and microscopy techniques, there is a ripe opportunity to better elucidate the structural and morphological features that not only distinguish TNTs and TMs from other membranous protrusions but also to identify characteristics that are either common in all TTs regardless of cell of origin and characteristics that differ based on individual cell types. Once these factors and differences have been clarified, it will be more feasible to determine whether targeting of TTs is a realistic goal. Research work to date in this exciting emerging field of cancer cell biology marks TTs as a novel and direct form of intercellular communication that has a high level of potential impact on our understanding of glioma cell biology and the evolution of these highly aggressive tumours into heterogeneous structures that have proven to be refractory to current standards of treatment.

CONFLICT OF INTEREST

V. Venkatesh has no relevant disclosures. E. Lou discloses the following interests: collaborator, consultant, and clinical trial site principal investigator (University of Minnesota) for Novocure, LLC; consultant for Nomocan Pharmaceuticals; Scientific Advisory Board Member for Minnetronix, LLC; honorarium and travel expenses for research presentation for GlaxoSmithKline, LLC (2016); site principal investigator (University of Minnesota) for a multi‐site clinical trial sponsored by Celgene Incorporated (BTCRC‐GI15‐015).

AUTHORS' CONTRIBUTION

All authors had full access to the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Conceptualization, V.V., E.L.; Methodology, V.V., E.L.; Investigation, V.V., E.L.; Formal Analysis, V.V., E.L.; Resources, V.V., E.L.; Writing ‐ Original Draft, V.V.; Writing ‐ Review & Editing, V.V., E.L.; Visualization, V.V.; Supervision, E.L.; Funding Acquisition, E.L.

Venkatesh VS, Lou E. Tunneling nanotubes: A bridge for heterogeneity in glioblastoma and a new therapeutic target?. Cancer Reports. 2019;2:e1185. 10.1002/cnr2.1185