Perspectives of cellular communication through tunneling nanotubes in cancer cells and the connection to radiation effects

To date, a clear and totally agreed definition of TNTs does not exist. This is a consequence of numerous observations of similar structures which show on the one hand comparable but on the other hand different properties. However, some key characteristics can be satisfied about TNTs.

TNTs are thin cytoplasmic membrane bridges with a diameter ranging from 50 nm to 1500 nm that interconnect cells over long distances up to several cell diameters length [8] (see Fig. 1). This allows the direct cell-to-cell transfer of signals as well as cellular compounds [8, 24]. They often appear as straight lines in-vitro, but in tissue or in three dimensional extracellular matrix they can exhibit a curved morphology [11, 25, 36]. In Fig. 2 a 3D rendering of a TNT connection between U87 glioblastoma cells is shown which has kinks and its middle part lies on the substrate. Due to their flexible shape, TNTs are also able to connect cells even if the nearest distance between them is blocked e.g. by other cells [11]. Between cell lines or even in the same cell line the morphology and cytoskeletal composition of TNTs vary [24]. Whereas F-actin is found in most TNTs, usually only the thicker TNTs contain microtubules [25] or cytokeratin filaments [37]. The length varies in a range of just a few to over 100 μm [11] and can be dynamically regulated if the interconnected cells migrate until the distance becomes too large and the tube disappears [24]. The lifetime is ranging from a few minutes [11, 38] up to several hours [12, 26]. Both length and life time might be determined by the available membrane reservoirs and migration speed [26, 39]. Additionally, TNTs which contain microtubules may be more stable than those composed only of F-actin since microtubule-filaments exhibit a higher degree of stiffness [24]. Besides the differences regarding the cytoskeleton content, also the connection of a TNT to the cell body varies among different cell types. Some TNTs are open-ended at both ends and thus exhibit membrane continuity [23, 40], but there are also close-ended TNTs containing a junction [11, 41, 42] or immune synapse [43] as gating mechanism. A sketch of open- and close-ended TNTs is displayed in Fig. 3.

TNTs have to be distinguished from other similar membrane structures such as cytonemes [44, 45] and filopodia [46] (see Fig. 1 and Fig. 2), which do not have cell-to-cell contact. Additionally, there are membrane connections which also allow cell-to-cell communication, but are structurally distinct from TNTs, the gap junctions [47] and epithilial bridges [48]. The key features of these connections and protrusions are shown in Table 1.

Based on the observations of TNT formation in research two ways of TNT establishment were identified [49]. The first is the de novo generation of nanotubes from filopodia-like protrusions by an actin-driven process within several minutes [10, 23, 38] (Fig. 4 left side). The main differentiation between filopodia and not yet fused TNTs in this case is, that filopodia seem to be more branched and with contact to the surface in cell culture, which TNTs normally don’t have. But a clear distinction is difficult or even impossible (see Fig. 1), which gives an additional hint on the complexity of the function of cellular membrane protrusions. Here, a protrusion of membrane, probably initiated by Rho-family GTPases, elongates by actin polymerization. If the tip of the protrusion reaches the target cell or another protrusion, a physical contact will be establish by adhesion and possibly membrane fusion [23] (Fig. 4 left side). However, it is uncertain whether the filopodia growth is a stochastic process or driven by a chemoattractant.

The second way is the TNT formation at the detachment of cells after direct cell-to-cell contact (Fig. 4 right side). This mechanism has been observed in immune cells as T cells, natural killer cells or macrophages as well as in normal rat kidney cells [11, 12, 41‐43]. Here, cells form an immune synapse or fuse when they come into direct contact. With the dislodgement of the cells a nanotube is pulled out (Fig. 4 c). Whether one or both cells contribute to the establishment of the TNT is uncertain. Additionally, the involvement of adhesion, fusion or actin polymerization remains unknown in this formation mechanism [49]. The time of cell-to-cell contact seems to play a key role, since TNTs are only hardly established if the cell-to-cell contact lasts under 4 min in T cells [11, 43]. Even though, these two models might appear to be very distinct, they are not mutually exclusive [49, 50]. Several stimuli of TNT formation such as Fas-ligand receptor in the immune system or M-sec, a protein associated with the component Sec6 of the exocyst complex required for the docking of exocytic vesicles on the plasma membrane [51, 52], have been discovered. These findings suggest that the TNT formation might be like the mechanisms of filopodia and lamelipodia regulation. Involved molecules in the TNT formation machinery are Rho and Ras small GTPases families such as Cdc42, Ral or the exocyst effector [52] as well as myosin X [49] and the transmembrane MHC class III protein LST1 [53]. These molecules are responsible for the actin cytoskeleton remodelling along the protrusion steps of TNT formation. Furthermore, lipid draft proteins such as I-Bar [54] which generate cell membrane curvature are required for TNT stabilization and formation. For TNT guidance and initiation of cell-to-cell contact, adhesion molecules such as N-cadherin and β–catenin [37, 55] as well as receptor-ligand interactions [43] are crucial. For more detailed research on the molecular basis of TNT formation and the inception mechanisms in the TNT development please refer to [31, 34, 56].

TNTs can be used as highways to transfer and exchange cellular compounds from one cell to another. Organelles and particles, which are observed to be interchanged between cells are mitochondria [4‐6, 12, 16, 29, 57‐60], vesicles [12, 23, 37, 38], membrane as well as cytoplasmic components [12, 23, 40, 51], nanoparticles [61‐64] and even more [17, 28, 49, 57, 65]. Furthermore, the research on TNTs pointed out that there are different transport mechanisms which may be dependent on the cytoskeletal content, diameter as well as membrane continuity of the considered nanotube. For instance, in human macrophages two different types of TNTs are found. Here, intracellular components including endosomes, lysosomes and mitochondria are only exchanged within thick, microtubulin containing TNTs, whereas surfing of bacteria on the membrane surface is only detectable in thin actin-based TNTs [12].

It has been proven that TNTs have a prominent role in the propagation of signals. For instance, Ca²⁺ signals can be transferred via TNTs between remote cells [29, 40‐42, 52, 66]. Usually, electrical signals are transmitted through neuronal synapses or gap junctions, latter need a close proximity for cell-to-cell communication. In contrast, these kinds of signals can be quickly transferred over long distances when cells are connected by TNTs. That phenomenon was initially reported in 2005 [40]. Treatment with α-glycyrrhetinic acid, an inhibitor for the functionality of gap junctions, does not block the Ca²⁺ flux transfer, suggesting that gap junctions are not involved in this transmission. Recent studies using the sensitive membrane potential probe DiBAC4 [3] reveal that several cell types can be electrically coupled by TNTs, where gap junctions interposed at the membrane interface in one end of the connection allowing bi-directional passage of electrical currents in a selective way [41, 42]. The strength of signal depends on the length and diameter of the nanotubes, the open probability of present gap junctions as well as the number of involved TNTs per connection. The fact, that there are both gap junction independent and gap junction dependent electrical transmissions by TNTs implies a significant diversity of TNTs with different properties and functions.

Besides electrical signals, death signals can also be transferred through TNTs [43, 51]. The involvement of TNTs in apoptosis signalling was found in T cells, in which death signals are propagated by TNTs and immune stimulation leads to an enhanced TNT formation among cells. In fact, it was shown that phagocytosis signals are transferred from apoptotic to viable cells in order to help the immune system to detect a damaged cellular region [67]. Furthermore, it was reported that activated natural killer cells more frequently form TNTs and connected target cells are more often lysed than unconnected [43]. This suggests that TNTs can help the immune system to transfer cytotoxic chemicals to distant target cells.

The research on TNTs uncovers a significant diversity of transport processes. One observed and suspected transport mechanism is the exchange of molecules by molecular motors [23, 27] as several transferred molecules and particles such as HIV-1 viral particles or lysosomal vesicles exhibit a velocity in a range similar to those of actin-driven molecular motors [49]. Furthermore, this myosin-driven transport mechanism is in the agreement with the unidirectional one-way street feature, since this could be established by assuming that the actin filaments are of the same polarity [68]. It could also be possible, that the cargoes are linked to actin themselves and transported by actin polymerization [23]. Here, the actin could be seen as a rope on which the cargoes are anchored and dragged along the tube driven by actin polymerization at one end [27]. Bidirectional transport of cargoes was only observed in TNTs which additionally contain microtubules as cytoskeletal content suggesting that a microtubule molecular motor could be responsible for this behaviour [49]. It was also observed that bidirectional transfer of cargoes can change into an unidirectional transport proceeding after stress situations such as injury [28]. It might also be possible, that there are transport mechanisms which are cytoskeleton independent such as the transfer by gondolas [38]. Gondolas, are moving distensions in TNTs which can carry enclosed organelles which are bigger than the diameter of the respective nanotube itself [69]. The formation and the generating force needed for their movement are unexplored. It is possible that the movement is driven by differences in chemical potential regarding to the molecules inside the bulk solution and the interior of the target cell or to the compositions of the gondola membrane and the target cell membrane [37]. Further investigations must be performed in order to resolve how the carriage of cargoes occurs within TNTs and which molecular motors or other proteins are involved in such a transfer system. Nevertheless, many reports demonstrate that the transfer of molecules and particles occurs in an active manner and not due to diffusion. For instance, studies with ATP reveal a blockage of organelle transfer via nanotubes pointing out that ATP is necessary for the respective transferring processes [10, 12, 26, 28]. Same results were obtained by considering the transfer of electrical signals across TNTs, here simulations showed that a passive transfer is inefficient and experiments indicate that the signals are actively generated and propagated within TNTs [66].

Several studies point out that the presence of TNTs as cellular network correlates with diverse stress factors such as hydrogen peroxide (H₂O₂) exposure [70], hypoxia [71], UV [4], x-ray [72] and particle radiation [73] as well as serum starvation [59], temperature, toxin B [54], infections [14, 74] or inflammation [36, 39, 56]. Furthermore, it is revealed that healthy cells are able to rescue apoptotic cells by the exchange of functional mitochondria via TNTs [4, 5]. These findings suggest that TNTs have a special relation to stress. In his recent paper, Amin Rustom used these findings to introduce a new mechanistic model of reactive oxygen species-dependent tunneling nanotube formation [75]. He describes the formation of TNTs according to the increase of reactive oxygen species (ROS) level in stressed cells. These stressed cells transmit “call-for-help” signals to their surroundings. According to this model, TNTs will be formed by unstressed cells in order to establish an open communication channel to the stressed cell. Followed by the exchange of particles such as mitochondria to rescue the apoptotic cell or by the isolation and removal of the cells whose ROS level is too excessive. Based on this model, TNTs are a communication tool used for the cellular organization and survival during stress.

However, the exact role of TNTs in stress situations remains obscure. Although, more and more reports of TNTs under stressful conditions become published, it is still unclear under which circumstances the TNTs are established from the non-stressed to the stressed cells as in the model of Rustom [75] or vice versa as reported in other studies [4, 76]. Wang et al. [76] found that the transcription factor p53 which regulates several genes in response to various harmful stress signals including DNA damage and hypoxia [77], plays an important role for the development of TNTs in astrocytes after H₂O₂ treatment and serum depletion. Nevertheless, it is still unknown which mechanisms or signals activate the cells to form an open communication channel and thus respond to stress. Additionally, the role of stress proteins such as heat shock proteins in the formation of TNTs is widely unexplored. A recent study showed that the membrane-bound heat shock protein mHsp70 is located on TNTs in human U87 glioblastoma, mouse GL261 glioma and mouse 4 T1 mammary carcinoma cell lines [73]. This finding provides the first evidence of the presence of stress proteins as structural component in the lipid composition of TNTs and therefore the support of TNTs by stress proteins.

TNTs have been found in several cancer cell lines including glioblastoma [72, 73, 78‐80], carcinomas [38, 62, 63, 73], ovarian cancer [57, 71, 81‐84], breast cancer [17, 57, 59, 81, 82], bladder cancer [16, 54], HeLa [15, 52], human neuroblastoma [17] and mesothelioma cell lines [58, 59, 85‐87]. Furthermore, Ady et al. [85] showed that malignant mesothelioma cells exhibit 20-fold to 80-fold more TNTs as compared to normal mesothelial cells after 72 h in vitro cell culture. Thus, TNTs are a preferred communication mechanism in malignant cell growth. In addition, TNTs are not only found in in vitro cancer cell cultures but also in vivo. The first evidence for the existence of TNTs in vivo was provided by Chinnery and colleagues in 2008 [36]. They showed that MHC class II⁺ cells in the mouse corneal stroma are interconnected by TNTs and that this network is more distinct under inflammatory conditions. In 2012, Lou et al. [59] demonstrate the occurrence of TNTs in solid tumor samples from patients with mesothelioma and lung adenocarcinoma. TNTs were also found in tumor explants of murine orthotopic osteosarcoma and human patients with ovarian cancer [57, 71, 83]. Osswald et al. [72] discovered that membrane tubes are jointly responsible for the high resistance and progression of brain tumors when acting as a functional multicellular network.

This statement becomes even clearer when considering the so far observed exchanges of different cargoes via TNTs in cancer cells. It has been reported, that mitochondria traffic through TNTs supports the invasiveness of bladder cancer cells or modulates chemoresistance [16, 57]. Furthermore, TNTs are also suspected to transfer P-glycoproteins in cancer cells, a protein which can cause multidrug resistance [17, 82]. Chemoresistance can also be accomplished by the transfer of genetic material such as microRNA through TNTs [83, 88]. Although, cellular communication via TNTs can cause redistribution of drugs which in turn leads to increased resistance to chemotherapy or radiotherapy, the opposite can also be true. A recent study on the bystander effect showed that the herpes simplex virus NV1066 can be delivered to non-infected recipient cells by TNTs [86]. Here, the authors used a modified transwell assay that physically separates the cells by a 0.4 μm thick membrane to prevent contact-dependent cellular communication via gap junctions and to reduce the exchange of diffusing exosomes by > 80% [89]. This finding demonstrates that cell-to-cell communication via TNTs can induce apoptosis in non-targeted cells and thus may also be a promising tool to enhance the effectivity of a therapeutic treatment.

Radiotherapy is one of the four pillars in cancer therapy as approx. 50% of all tumors worldwide are treated using radiotherapy [90, 91]. In particular, patients with central nervous system, breast, oesophageal, lung, head and neck cancers frequently receive radiotherapy during their course of illness [90]. Although less is known about the role of TNTs after irradiation first studies point on the one hand to a protecting effect of TNTs after x-ray radiation [72] and on the other hand to a decrease of connections after alpha-particle irradiation [73]. This points to a versatile role of TNTs in the radiation response of a cell composite. Additionally, TNTs were found in several cell lines which originate from the tumor types which are treated using radiotherapy. Furthermore, tumors such as glioblastoma which are well known to be highly migrative and invasive [92] show low response to treatment resulting in low 5-year survival rate [93, 94]. Moreover the invasive potential of these tumors is enhanced upon conventional radiation treatment using x-rays [95, 96] but not after alpha-particle treatment [95]. Additionally, it is well known that photon radiation induces cellular stress via the induction of ROS such as hydrogen peroxide (H₂O₂) which are suspected for triggering TNT formation [70, 75]. Particle radiation on the other hand predominantly directly interacts with the DNA and therefore ROS induction might be reduced [97].