Tunneling nanotubes and cytonemes in cancer and development

Like people, cells communicate with each other in many different ways. Face-to-face sharing of information can occur via cellular gap junctions, public speaking can be accomplished by emitting growth factors and cytokines, and personal objects can be mailed using extracellular vesicles and endocytosis (Alberts et al., 2022). Interestingly, just as the establishment of telephone lines permitted direct, private communication between people at long distances, cells can also generate direct interconnections to distant cells or tissues for the focused transfer of information and cellular contents using long, slender membrane tubes termed cytonemes or tunneling nanotubes (TNTs). Even social media can be mimicked by certain cells that establish a TNT-mediated network of interconnections between multiple cells.

These thin membranous tubules have been termed cytonemes, TNTs, TNT-like structures, tumor microtubes, membrane nanotubes, nanotubular networks, and related names in over a thousand publications (Cervantes and Zurzolo, 2021; Dagar et al., 2021a; Osswald et al., 2015; Pinto et al., 2020; Rustom et al., 2004; Wang and Gerdes, 2015; Watson et al., 2023; Yamashita et al., 2018). These direct nanotubular cellular interconnections (Box 1) can transfer a wide range of cytoplasmic contents from 1 cell to another distant cell. Depending on the type of nanotubular link, they can transfer molecules ranging in size from small ions, including calcium fluxes, to larger growth factors, vesicles, and even mitochondria (Figs. 1 and 2). In this review, we primarily discuss TNTs in comparison to cytonemes and other tubular structures and review their roles in cancers, other diseases, and embryonic development, focusing on recent intriguing primary publications alongside some classical studies. Finally, we discuss eight unresolved classes of broad conceptual or specific functional issues that should provide abundant exciting opportunities for future research.

Cytonemes are extensively studied nanotubular cytoplasmic extensions that interconnect cells to mediate intercellular signaling (Fig. 1 and Fig. 2 A), particularly during embryonic development (Table S1 [Hall et al., 2024; Ma et al., 2025; Wang et al., 2025b]). A wide range of developmental regulators can be targeted to other cells via cytonemes to regulate tissue patterning and differentiation. Cytonemes in Drosophila that transport the TGF-β family member DPP (decapentaplegic) were originally shown to require, besides actin, the expression of multiple genes including a formin, cell adhesion proteins (neuroglian and capricious), and a dynamin (Roy et al., 2011). Since then, other molecules have been implicated in cytoneme structure and their intercellular transport function, including Wnt, Notch, and fibroblast growth factor receptors (Hu et al., 2024; Ma et al., 2025; Patel et al., 2022; Sherer and Mothes, 2008; Yamashita et al., 2018).

Cytonemes have been particularly well-characterized in Drosophila, but they also perform critical functions in mammalian systems. For example, during mouse neural tube development, cytonemes were shown to be essential for generating normal neural tubes (Hall et al., 2024). Long cytonemes are reportedly generated through myosin-X (MYO10) activity, and they transport critical developmental signaling molecules including sonic hedgehog and Wnt for the regulation of neural cell fate during neural tube patterning. Cytonemes are also required for Notch-mediated epidermal patterning and differentiation in zebrafish (Wang et al., 2025b).

TNTs and TNT-like structures are also long tubular membranous structures, but they are generally larger in diameter than cytonemes (Fig. 1 and Fig. 2 B)—approximate diameters are 50–400 nm for TNTs versus 50–200 nm reported for cytonemes, but with considerable overlaps in size (Cervantes and Zurzolo, 2021; Dagar et al., 2021a; Hu et al., 2024; Jung et al., 2020; Korenkova et al., 2020; Lou et al., 2024; Sherer and Mothes, 2008; Yamashita et al., 2018). Both TNTs and cytonemes can sometimes extend large distances—in some cases up to 500 µm or more between cells for large TNTs termed tumor microtubes. In vitro studies describe two distinct mechanisms by which TNTs are generated: either by the extension of a slender cell process outward to contact and then to interconnect with another cell or by the generation of residual thin cellular fibers as cells separate from one another (Ljubojevic et al., 2021; Matejka et al., 2024).

TNTs can be homotypic or heterotypic: homotypic TNTs link cells of the same type, such as neuronal cells in a network, whereas heterotypic TNTs provide nanotubular connections between different cell types (Cervantes and Zurzolo, 2021). Homotypic TNTs can provide coordination between related cells, such as within a tumor for protection against stressors such as hypoxia (Desir et al., 2016), whereas heterotypic TNTs can link distinct cell types, for example, between cancer and immune cells for mitochondrial transfer (Ikeda et al., 2025). Some TNTs are “closed-ended” in that they are open at one end but closed at the other end by a gap junction containing connexin 43 at the target cell surface to permit the transfer of small molecules. In addition, TNTs have been described as “thin” or “thick” depending on their diameter of ∼50–400 nm for thin TNTs and 700–1,500 nm for thick TNTs (Fig. 1).

TNTs can provide interconnections between not only normal, nonmalignant embryonic and adult cells, e.g., between neuronal and immune cells, but also between multiple other cell types (see following sections, Fig. 3, and Cervantes and Zurzolo, 2021). TNTs have been studied most intensively in cancer, especially in brain cancers but also in a broad range of hematologic and solid cancers, both in vitro and in vivo (Table S2). TNTs have also been implicated in the cell-to-cell transfer of pathogenic proteins and viruses. TNTs can overlap morphologically and conceptually with other intercellular transfer structures such as tumor microtubes (which are larger than thin TNTs), long filopodia, microtentacles, cytokinetic bridges, and membrane nanotubes (Cervantes and Zurzolo, 2021; Dagar et al., 2021a; Lou et al., 2024; Osswald et al., 2015).

Cytonemes and TNTs both contain a structural core of F-actin filaments. As depicted in Figs. 1 and 2, morphological studies of these nanotubular structures ranging from thin (50 nm) to thick tubules (1 µm or more) reveal that their contents can include a wide range of molecules and structures, including ions and proteins, microtubules, intermediate filaments such as vimentin, vesicles including autophagosomes, rough ER, and even mitochondria (Cervantes and Zurzolo, 2021; Jung et al., 2020; Lou et al., 2012; Wang et al., 2025a). Although classical filopodia lack vesicles and are shorter than cytonemes and TNTs, cellular extensions termed “long filopodia” and cytonemes can contain organelles including mitochondria—underscoring the overlap in contents between different cytoneme- or TNT-like structures (Matejka et al., 2024; Sartori-Rupp et al., 2019; Wood et al., 2021).

High-resolution microscopy revealed that classical Drosophila cytonemes can be initiated as very long filopodia that extend outward to contact invaginations in target cells and ultimately to interconnect them. Importantly, they can contain rough ER and even mitochondria in bulges of these long filopodia, though only rarely vesicles (Wood et al., 2021). Some cancer cells, including breast cancer cells, can produce long, strongly contractile cell protrusions containing mitochondria, which suggests they might be different from filopodia and TNTs (Nazari et al., 2023). However, an ultrastructural study of a breast cancer cell line describes morphologically similar long filopodia suggested to be precursors to TNTs (Franchi et al., 2020), and some long filopodia have been shown to be contractile during embryonic development (Chauhan et al., 2009). Nevertheless, whether some TNTs are contractile remains to be determined. Fig. 1 depicts the full range of membranous tubular structures; it is, however, important to emphasize that their relationships with each other remain to be further elucidated. Although we speculate that long filopodia and the thick, long cell protrusions that extend outward from cells to contact and remodel surrounding extracellular matrix may potentially be precursors of thick TNT-like structures including tumor microtubes, this possibility needs direct experimental evaluation.

While the vast majority of studies have focused on characterizing TNT and cytoneme composition as well as their intercellular transfer functions, their biophysical roles in cell-to-cell physical interactions have been examined much less. A recent in vitro study characterized TNT stiffness and internal tension by atomic force microscopy, reporting that long thin TNTs displayed higher internal tension than shorter TNTs (Ota and Nagayama, 2025). We speculate that such forces between connected cells might promote their translocation. TNTs have been implicated in the compaction of spheroids of MCF7 cancer cells in vitro (Pulze et al., 2020), although whether their role depends on direct cell–cell forces or indirect effects of signaling is unclear. It is also not clear whether such biophysical interactions and nanotubular contractility play a significant role in vivo, e.g., by influencing the proximity of 1 cell to another, such as to promote cancer cell interactions with endothelial cells during extravasation or intravasation, or to pull immune cells toward their targets.

In vivo studies in zebrafish embryos demonstrate the importance of TNTs in normal embryonic development (Korenkova et al., 2025). TNTs in zebrafish embryos can transfer cytosolic proteins and organelles such as mitochondria and early endosomes to other migrating gastrula cells (Fig. 3 D). TNT-like extensions in migrating neural crest cells have been suggested to play a role in the directionality of the neural crest cell migratory stream (Teddy and Kulesa, 2004) (Fig. 3 A). Recently, TNT-like tubular interconnections have been identified as physical interconnections extending between cardiomyocytes of the myocardium through the cardiac jelly extracellular matrix to insert into endocardial cells in the endocardium for mediating long-distance communication needed for normal heart formation (Miao et al., 2025). These TNT-like structures depend on Cdc42 activity for their formation and can transport proteins such as signaling molecules and vesicles. For example, they permit developmental signaling via the Notch1 pathway. Disruption of these TNT-like linkages impairs myocardial morphogenesis (Miao et al., 2025). These studies of nanotubular structures in different embryonic systems suggest that the morphological and functional roles of cytonemes and TNTs can overlap.

In mature organisms, TNT interconnections between cells within a tissue or between cells of different tissues have been suggested to play roles in inter-tissue regulation. For example, TNTs can interconnect pericytes that surround capillaries in mouse retina and communicate calcium waves to regulate blood flow, mediating a critical role in neurovascular coupling in vivo (Alarcon-Martinez et al., 2020) (Fig. 3 D). These pericyte-to-pericyte TNTs are disrupted by elevated intraocular pressure in a mouse model of glaucoma, which induces calcium influx-mediated neuronal damage in the retina (Alarcon-Martinez et al., 2022). Long TNTs interconnecting pericytes have also recently been characterized in the macula of human eyes (Hein et al., 2024).

TNTs can also interconnect pericytes in the retina with endothelial cells, and in vitro treatment of pericytes with PDGF can induce such TNTs (Fig. 3 D). These TNTs transport mitochondria unidirectionally from pericytes to endothelial cells. In a metabolic rescue process, this mechanism restored a functional mitochondrial network after experimental depletion of endothelial cell mitochondria (Kempf et al., 2024).

Another TNT-mediated mitochondrial rescue process has been reported involving a transfer from stem cells to adipose tissue macrophages to alleviate a mouse model of pregnancy-induced diabetes (Chen et al., 2025a). Similarly, TNT-mediated transfer of mitochondria has been reported to contribute to therapeutic angiogenesis in a mouse model of hindlimb ischemia (Che et al., 2025b). In a model of cardiac sepsis, cardiomyocytes form TNTs with neighboring cells such as endothelial cells and fibroblasts to transfer damaged mitochondria to these neighboring cells, which can help the cardiomyocytes but contributes to sepsis-induced cardiac damage (Song et al., 2026). Consequently, the use of TNTs to interconnect cells with neighbors or with other cell types appears to be widespread in both embryonic and adult organisms.

Within the brain, TNTs can interconnect neuronal and microglial cells to permit exchanges of mitochondria and α-synuclein (Chakraborty et al., 2023). TNTs also contribute to the generation of a network of interconnected neurons (Khattar et al., 2022; Palese et al., 2025; Rakotobe and Zurzolo, 2025). Short TNT-like structures extending outward from neuronal dendrites provide interconnections termed dendritic nanotubes. Dendritic nanotubes in the visual cortex (shown to be different from synaptic connections) provide pathways for not only calcium signaling but also for transporting Alzheimer’s disease human β-amyloid in mouse brains. This dendritic nanotube network is altered in a mouse model of early Alzheimer’s disease prior to amyloid plaque formation, suggesting a role in the progression of Alzheimer’s pathology (Chang et al., 2025). In fact, TNTs have also been implicated in the spread of prion, Parkinson’s, and Huntington disease proteins in the central nervous system (Victoria and Zurzolo, 2017). Not surprisingly, there are also multiple examples of roles for TNTs in the initial cell-to-cell spread of pathogenic viruses including HIV, influenza virus, herpesvirus, SARS-CoV-2, and others (Jansens et al., 2020; Lv et al., 2024; Pepe et al., 2022; Sowinski et al., 2008; Weir et al., 2025), as well as even nanoplastic particles (Chang and Wang, 2025).

TNTs/tumor microtubes can interconnect cancer cells (Fig. 3 E) and have been identified in a wide range of cancers (Table S2). They have been studied most extensively in glioma brain cancers but also in many other blood and solid cancers, including other brain, head-and-neck, lung, colon, breast, ovarian, and prostate cancers, as well as leukemias (Cervantes and Zurzolo, 2021; Chen and Zhao, 2024; Osswald et al., 2015; Pinto et al., 2021; Sáenz-de-Santa-María et al., 2017; Venkataramani et al., 2022). TNTs are thought to play important roles in promoting cancer progression or in the protection of tumor cells against stressors, including hypoxia and metabolic stress (Desir et al., 2016; Kretschmer et al., 2019). They can contribute to cancer promotion, protection from cancer chemotherapy and radiation therapy, and maintenance of tumor fitness in the face of hypoxic and chemical stresses. For example, TNTs have been implicated in the development of resistance to chemotherapy in a variety of cancers (Desir et al., 2018; Sarkari and Lou, 2024). Mechanistically, tumor cells suffering from a high concentration of a chemotherapeutic agent are suggested to be able to reduce toxic levels by efflux to other interconnected cells having lower levels of the agent. Alternatively, TNTs can mediate the sharing of the multi-drug resistance membrane protein P-glycoprotein (MDR1) with other interconnected cancer cells (Pasquier et al., 2012).

In human and mouse in vivo tissues, malignant glioma cells interconnected by TNT-like tumor microtubes were found to differ from non-interconnected cells by the elevated expression of stem cell markers, especially nestin along with Musashi and Sox2; they also showed higher levels of the proliferation marker Ki67, all consistent with cancer promotion (Xie et al., 2021). Cancer cells that had been integrated into this TNT-mediated network displayed enhanced resistance to radiation therapy (Xie et al., 2021). In addition, TNTs are thought to contribute to the general distribution-redistribution of vesicles, including lysosomes and autophagosomes, between cancer cells, as well as mitochondria in tumors (Fig. 4), which could enhance overall tumor metabolism and fitness (Libring et al., 2024; Sáenz-de-Santa-María et al., 2017; Zampieri et al., 2021). For example, pheochromocytoma PC12 cells that would otherwise undergo apoptosis after UV treatment can be rescued by the transfer of fresh mitochondria via TNTs from untreated cells (Wang and Gerdes, 2015).

In an alternative mechanism, the formation of TNTs between glioblastoma cells and adjacent nonmalignant astrocytes could modify their tumor microenvironment. TNT-mediated transfer of tumor cell–modified mitochondria to neighboring astrocytes is suggested to promote a more tumor-supportive local environment with altered astrocyte glutamine metabolism and resistance to hypoxia (Valdebenito et al., 2021). Conversely, malignant multiple myeloma cells can use TNTs to hijack mitochondria from adjacent nonmalignant bone marrow stromal cells to promote their own bioenergetic plasticity and fitness (Marlein et al., 2019). Besides mitochondria, other cellular components such as vesicles are known to be redistributed, and even cancer-promoting mutated oncogenes such as oncogenic KRAS can be shared between interconnected cancer cells (Desir et al., 2019). Similarly, the microRNA miR-115 that promotes cell proliferation and invasion can also be transferred between bladder cancer cells via TNTs (Lu et al., 2019). Conceptually, TNTs can enhance tumor progression by: (1) sharing mitochondria with other cancer cells to enhance overall metabolic fitness, (2) helping to protect against stressors resulting from chemotherapy, radiotherapy, and hypoxia, and (3) spreading oncogenic factors. Yet another variation on the theme of molecular sharing to enhance tumor aggressiveness is the ability of breast cancer cells to “borrow” a metastasis-promoting protein from adjacent mesenchymal stromal cells to enhance tumor dissemination (Sinha et al., 2024).

As reviewed elsewhere, TNTs interconnect multiple different types of immune cells, playing roles in their differentiation and subsequent overall immune surveillance (Cervantes and Zurzolo, 2021; Watkins and Salter, 2005; Zhu et al., 2021). TNTs also play roles in cancer immunosurveillance; for example, TNTs can promote long-distance linkage between natural killer (NK) cells and target cells, and physically disrupting TNTs reduces the lysis of targeted tumor cells (Chauveau et al., 2010). Particularly intriguing recent findings have focused on the roles of TNT-mediated transfer of mitochondria between normal, malignant, or immune cells to enhance or cripple immune cell function and/or support tumor cell metabolism (Fig. 4). TNTs can be used by T cells to obtain fresh mitochondria from bone marrow stem cells, thus enhancing their supply of mitochondria and overall metabolic fitness and T cell efficiency (Baldwin et al., 2024). Analogously, various cancer cells can hijack immune cell mitochondria from T or NK cells to both enhance cancer cell metabolism and impair immune cell function, both in vitro and in vivo (Saha et al., 2022). Conversely, cancer cells can transfer faulty mitochondria to T cells to blunt immune killing (Ikeda et al., 2025; Liang et al., 2025). Consequently, the transfer of mitochondria through TNTs can enhance both immune and malignant cell metabolism, deplete immune cell mitochondria, or even transfer faulty mitochondria from tumor cells to cripple immune cells.

Tumor cells can interact with host macrophages using potentially bidirectional TNT-inductive mechanisms. Conditioned media from macrophages containing epidermal growth factor (EGF) can induce TNTs in various human cancer cell lines. Depletion of the TNT-promoting protein M-Sec (also known as the TNF-α–induced protein TNFAIP2, repeatedly associated with TNT induction) resulted in inhibition of 3D in vitro tumor cell invasion (Carter et al., 2019). Conversely, nonmalignant macrophages can generate TNTs in a process involving Cdc42 and Rac1 (Hanna et al., 2017). Interestingly, TNTs linking tumor cells and macrophages can promote tumor cell invasion in an in vivo zebrafish model system of metastatic invasion stimulated by EGF and dependent on M-Sec (Hanna et al., 2019). A novel, intriguing physical mechanism promoting extravasation was found to be based on long, thin, membranous TNT-like structures that provide physical links extending from macrophages to tumor cells on the other side of endothelia; the TNTs were apparently from both macrophages and tumor cells (Genna et al., 2023). These macrophage-mediated TNT-like structures spanning the endothelium can reportedly assist tumor cell translocation through the endothelial cell barrier during extravasation for seeding of metastases. In fact, mice deficient in M-Sec, including in their macrophages, showed reduced experimental metastasis (Genna et al., 2023). If these TNT-like structures linking tumor cells within blood vessels lined with a basement membrane to host macrophages on the other side of endothelia were to possess active contractility, we speculate that the tumor cell could be pulled physically through the endothelial barrier. In fact, long prehensile cell protrusions of tumor cells previously shown to be contractile for pulling cells through basement membrane barriers (Nazari et al., 2023) might be precursors of TNT-like structures that would contract to help pull tumor cells into tissues during extravasation—a testable speculative hypothesis.

Increasing data indicate the nervous system can make substantial contributions to tumorigenesis, the cancer immune environment, and even metastasis of a wide variety of cancer types, including lung, breast, and colorectal cancers (Huang et al., 2025; Magnon and Hondermarck, 2023). Although neuronal-cancer cell input can occur via synaptic connections (Sakthivelu et al., 2025; Savchuk et al., 2025), neuronal cell-to-cancer cell TNTs or TNT-like interconnecting structures have also been documented (Hoover et al., 2025) (Fig. 3 F). Such TNT-like connections can provide fresh mitochondria to cancer cells to promote more effective cancer cell function (Hoover et al., 2025). Approximately 30% of breast cancer cells showed such transfers in vitro, and disruption of TNTs with cytochalasin inhibited the transfer, though more specific inhibitory methods were not tested. Besides neuronal cells, primary fibroblasts and a pre-adipocyte cell line could also transfer mitochondria to recipient cancer cells in vitro. Interestingly, in vivo xenograft analyses revealed enrichment in metastases: starting from a low 5% initial proportion of cells displaying a genetic mitochondrial transfer marker in the primary tumor, the percentage rose dramatically to a quarter of total cancer cells in brain metastases and to nearly half in lung metastases (Hoover et al., 2025). These findings were partially confirmed using a melanoma cancer cell model. Human breast cancer metastases also showed increased mitochondrial loads after pathological analysis. Taken together, these findings support a key role for TNTs and mitochondrial transfer from neuronal and possibly other cell types to promote cancer metastasis (Hoover et al., 2025).

Two noteworthy features of the literature on TNT formation and maintenance are the wide variety of molecules implicated in TNT formation and the conceptual difficulty in identifying the molecules needed for induction of TNTs versus their maintenance. As for other complex biological processes, multiple cellular molecules could be required for TNT maintenance, making it difficult to determine which molecule(s) are essential for their initial formation. Box 2 provides a partial listing of molecules implicated in the formation or maintenance of TNTs and related structures. Many of the molecules implicated to date appear to differ between cytonemes and TNTs, but it is not clear whether this distinction is definitive, especially considering the striking heterogeneity of the reported molecular drivers of TNTs in different publications and the capacity of both TNTs and cytonemes to carry mitochondria and rough ER.

Plausible candidates for inducers of TNTs in Box 2 include growth factors such as EGF and HGF, but each is well-known to stimulate multiple other cellular processes—that is, they can stimulate TNT formation, but as signaling molecules they may not be the direct molecular initiators/mediators of TNT formation. At an intracellular mechanistic level, a logical driver for generating thin cellular protrusions would be the Rho GTPase Cdc42, which is well-known to induce filopodia (Nobes and Hall, 1995). Actin polymerization is required, since the core of TNTs includes F-actin, and the actin remodeling protein cofilin has been implicated in TNTs (Chen et al., 2025b; Dagar et al., 2021b). Another plausible candidate, Arp2/3, was implicated in TNT formation in macrophage RAW cells (Hanna et al., 2017), but not in neuronal CAD cells, where inhibition of its activity in promoting F-actin branching instead drove the formation of longer filopodial precursors of TNTs via unbranched linear F-actin filaments (Henderson et al., 2023). Whether these differences reflect cell type specificity or differing roles in actin regulation remains to be determined. What is even less clear is which specific molecular drivers are involved in generating the unusual lengths of the tubular cell extensions characteristic of both TNTs and cytonemes compared with filopodia.

A particularly attractive molecular mechanism for TNT induction involves the regulatory protein M-Sec (TNFAIP2), which, when overexpressed, can promote the formation of long cellular protrusions and TNTs; M-Sec has also been implicated in TNT formation in multiple cancers (Che et al., 2025a; Fu and Yuan, 2024; Hase et al., 2009). M-Sec supports TNT elongation by translocating to the plasma membrane with the help of nucleolin and interacting with Ral, LST1, cofilin and finally the exocyst complex for stimulation of actin filaments and insertion of the plasma membrane needed for the TNT (Dagar et al., 2021b; Raghunathan et al., 2001; Schiller et al., 2013). However, how universal this plausible sequence may be for other cell types remains to be clarified, as well as the exact sequential molecular steps and mechanisms—ideally established by future in vitro reconstitution experiments.

A recent microscopy study has identified an additional biophysical mechanism for TNT formation based on filopodia: a filopodium from 1 cell reaches out and adheres to a filopodium from another cell using N-cadherin. One of the filopodia then retracts with the help of a helical twist in the filopodial shaft facilitated by myosin V, resulting in a single long TNT linking the 2 cells. In this study, most TNTs were closed-ended, leaving unanswered how the open-ended TNTs that can transport mitochondria between cells are generated (Chang et al., 2022).

What mechanisms/motor proteins mediate the cell-to-cell translocation of organelles such as mitochondria, and how is directionality of transfer determined? Candidates include the microtubule motor protein kinesin KIF5B and contractile protein myosin VI, but not the myosin II molecule implicated in many other cellular contractile processes (Halasz et al., 2024), and possibly myosin XIX (Qin et al., 2021). A key regulator of mitochondrial translocation is Miro1, which is a Rho GTPase on mitochondria that regulates mitochondrial transfer along TNTs (Nahacka et al., 2022). The processes of active transport of organelles within TNTs can be mimicked and reversed experimentally (Gong et al., 2025). External manipulation of TNT cargoes may ultimately be able to determine more definitively the endogenous biophysical forces and motor mechanisms used to transport organelle cargo, including mitochondria.

• •

  TNTs, cytonemes, and their morphological relatives mediate distant cell–cell interconnections; they can be up to hundreds of micrometers long and have a wide range of diameters, ranging from slender cytonemes to thick TNTs and TNT-like structures.

• •

  They can be initiated and maintained by a wide range of molecules, and a current dilemma is that neither cytonemes nor TNTs have a specific molecular marker.

• •

  The intercellular communication systems based on cytonemes and TNTs are needed for normal embryonic development and perhaps adult tissue homeostasis.

• •

  TNT-mediated communication becomes hijacked in diseases that include cancer and neurological diseases.

• •

  Cancer cells use TNT communication and organelle exchange systems to support cancer progression and to defend against a variety of attacking stressors, e.g., hypoxia, host immune defenses, chemotherapeutic agents, and other stressors

Many challenging questions and research opportunities remain as this field continues to expand rapidly.

Although long filopodia, cytonemes, TNTs, TNT-like structures, and tumor microtubes might comprise a spectrum of tubular structures, distinctive features of each exist as indicated in Box 1 and Fig. 1, making it important to answer this question more definitively. For example, although cytonemes are classically considered to be closed-ended signaling conduits, TNTs can either be open-ended or closed-ended due to a gap junction, thereby blocking their function in organelle transport. But can TNTs ever be interconverted from closed- to open-ended? A technical problem is that many studies of each type of TNT or TNT-like structure have used a different cell or tissue system. One approach might be to find cell types that can make > 1 class of these structures, and then to characterize the molecular and morphological features that are shared between these TNTs versus those that are distinct, as well as whether interconversion is possible, for example, as documented by live-cell imaging.

A major challenge will be to establish exactly how TNTs are generated, ideally by using in vitro reconstitution approaches, such as those used to establish the pathways of protein secretion. The partial list of candidate molecules in Box 2 will need further classification into shared and cell type–specific molecules, as well as their mechanistic pathways.

It will be important to determine in the future which of the molecules listed in Box 2 are necessary as opposed to being able to initiate TMT formation. Although M-Sec/TNFAIP2 and a couple of other molecules have been examined for sufficiency by overexpression analysis, the latter type of approach will be valuable for all the other candidates with testing in multiple different cell types. Another caveat is that some growth factors such as HGF may trigger permissive cells to form TNTs, but the signaling and downstream mechanistic pathways may differ.

Although different TNTs and tumor microtubes can have substantially different diameters, a potential source of confusion is that TNTs that appear to be thick by confocal microscopy can in fact be comprised of a bundle of thin individual TNTs as clarified by careful ultrastructural analyses (Sartori-Rupp et al., 2019). Answering how diameters are regulated will therefore first require ultrastructural evidence that putative thick TNTs are authentic single tubular structures, followed by comparisons of mechanisms between different cell types. Also needing elucidation is how the choice of which types of cargo are transported is made, then identifying the motor mechanisms that transport mitochondria or other organelles, including how the directionality of transfer is determined. The processes of active transport of organelles within TNTs can be mimicked and reversed experimentally (Gong et al., 2025). External manipulation of TNT cargoes may help determine the endogenous cellular forces involved to facilitate characterization of the motor mechanisms used to transport organelle cargo such as mitochondria.

Although many molecules have been implicated in TNT formation or maintenance, and some in sufficiency, the studies are often based on only a single cell type. It will eventually be important to determine which of the putative initiators and those involved in elongation versus stabilization of TNTs are general for other cell types. It may be quite possible for these nanotubular structures to be regulated in different ways in different types of cells—an important point to establish definitively.

One overall possibility is that TNTs, cytonemes, and tumor microtubes are cell type specific (Cervantes and Zurzolo, 2021). It may help to consider an analogy with another complex field in cell biology, that of the modes and mechanisms of 3D cell migration. In that field, there turned out to be clearly distinct functional mechanisms depending on cell type, experimental condition, and microenvironment, and yet there were also frequently shared molecular mediators (Yamada and Sixt, 2019). Whether identifying this pattern of differences and overlaps of mechanisms will also be the case for TNT research will likely challenge the cytoneme and TNT fields for many years to come. One type of approach, though painful, would be to test each of the reported regulators of TNTs in each biological system, ideally within the same laboratory. For example, the reported requirements for Arp2/3 or p53 activation in TNT formation differ depending upon the cell type being examined (Andresen et al., 2013; Wang et al., 2011), possibly due to their multiple cell biological roles. This type of discrepancy contributing to the so-called “reproducibility crisis” could be alleviated by systematic tests of generalizability to separate out cell-specific mechanisms from more broadly shared molecular mechanisms.

Although the widespread use of in vitro culture and co-culture of various cell types has provided a powerful tool for establishing concepts of what TNTs and tumor microtubes can do, what these cells actually do in vivo is not as definitive. Although dynamics of tumor microtubes have been successfully imaged by two-photon microscopy in vivo for many days (Osswald et al., 2015), in-depth real-time imaging to characterize vesicle and mitochondrial transfers in vivo has been impeded by technical imaging limitations. Instead, some papers have used markers to provide quantitative evidence, for example, that mitochondrial transfer has occurred—the caveat is that some alternative or additive mechanism besides TNT transfer might have been involved in vivo. In fact, use of the mitochondrial marker MitoTracker to show cell-to-cell transfer has been challenged due to its capacity to transfer between cells independently of cell–cell contact, making the use of genetic markers important (Hole et al., 2026). As technology advances, a future opportunity will be to show directly by intravital imaging the existence, extent (rare or widespread), and mechanisms of TNT-mediated transfers of specific organelles such as mitochondria in vivo.

A related question involves determining how widespread the roles of cytonemes and TNTs are in the development and homeostasis of all tissues. It is conceivable, but not established, that cytoneme and TNT-mediated interconnections are widespread in both embryos and adult organisms; but until reliable markers can be identified, this question will remain open. In the meantime, application of automated machine-learning algorithms may help detect TNTs in microscopy images (Ceran et al., 2022), as could lattice light sheet imaging (Parker et al., 2017) and super-resolution radial fluctuation microscopy (Hoover et al., 2025).

Although TNT-like long filopodia linking embryonic vertebrate lens and retinal epithelia appear to be contractile (Chauhan et al., 2009), the possibility of TNT contractility needs further rigorous experimental evaluation. In addition, whether the long contractile cell protrusions that can contract to remodel collagen (Nazari et al., 2023) might be precursors of contractile TNTs/tumor microtubes will also need stringent testing. If TNT contractility can be established, we speculate that tumor cells reported to link to host cells such as macrophages on the other side of the endothelial barrier (Genna et al., 2023) might use TNT-like contractile structures for pulling cancer cells across endothelia during tumor cell extravasation. An additional speculative possibility is that TNT-facilitated crossing of the endothelium and its basement membrane might also be used by normal immune cells to enter and leave tissues.

A TNT that extends out to contact another cell must undergo membrane fusion with the target cell to open an intercellular conduit for transport. How that fusion occurs will need to be determined, e.g., possibly via some unknown targeting mechanism and a fusogen to create a fusion pore (Rakotobe and Zurzolo, 2025).

Considered from a clinical standpoint, disruption of TNT functions could potentially block TNT-mediated cancer cell defenses against chemotherapeutic agents and radiation therapy, mitochondrial defense mechanisms, and other TNT protective strategies. Proposed approaches have included inhibiting gap junctions, neuronal signaling, and even Wnt signaling (Beichert et al., 2025; Chen and Zhao, 2024; Feng et al., 2025). A major concern, however, is that in analogy to classical anti-proliferation chemotherapy approaches, targeting molecules and pathways necessary for not only TNTs but also other critical cellular functions will result in too many clinical side effects. Combining anti-TNT treatments with other modalities in combination could alleviate this problem. For example, inhibiting tumor microtubes by modulating protein kinase C can be combined with therapeutic radiation for anti-glioblastoma tumor therapy (Azorin et al., 2026).

An interesting alternative approach might be to turn their TNTs against cancer cell networks, i.e., using them to spread “nanomedicines” throughout the cancer cell system. For example, anticancer nanoparticles with shapes optimized for TNT-mediated transfer or with piezoelectric properties might spread quickly through TNT networks to kill cancer cells that would otherwise be difficult to reach by diffusion alone (Liu et al., 2025; Ottonelli et al., 2022; Sierri et al., 2025). Similarly, liposomes that could contain chemotherapeutic agents can also be distributed through TNTs (Formicola et al., 2019). A major challenge is how therapeutic liposomes or nanoparticles can be targeted to disrupt TNTs yet not disrupt other cellular processes. It will also be necessary to obtain robust and selective uptake: although cancer cells are known to ingest some nanoparticles and liposomes, particularly if targeted to a particular cancer marker, the magnitude of uptake and avoiding subsequent damage to non-cancerous cells will be concerns. It would be ideal if specific TNT or cancer microtube molecular targets can be identified, ideally some targetable cell surface marker to permit the use of antibody-toxin chimeras.

A physical approach to disrupting TNTs using low-intensity alternating electric fields could disrupt TNTs in malignant mesothelioma cells with altered biomarkers in vivo, though whether the effects were due solely to TNT disruption was not clear (Sarkari et al., 2023). These examples again point to the need for innovative approaches to alter TNT functions selectively in tumors without disrupting essential normal cell–cell interactions. On the other hand, because some particular cancers can be treated by a nonspecific treatment such as antimitotic chemotherapy, targeting TNTs in particularly susceptible cells might eventually become practical therapeutically in selected cancers, infections, or neurological diseases.

In summary, the fields of TNT and cytoneme research have expanded rapidly from pioneering studies on their roles in cancer cell self-protection and in signaling during embryonic development to discoveries of a remarkable range of cell–cell communication and organelle-sharing mechanisms. The growing numbers of tissues and diseases, including cancers, that use these nanotubular structures for a wide range of functions, combined with the multiple still-unanswered questions in this field, predict an exciting future for this increasingly important area of cell biological research.

Table S1 summarizes examples of developmental systems in which cytonemes have been characterized, and Table S2 lists examples of tunneling nanotubes (also termed cancer microtubes) in various cancers.

We apologize for being unable to cite more than a selected sample of the >1,500 papers in this field due to space limitations. We focused primarily on concepts and recent publications. We used the following search systems for finding relevant publications, but not for writing or editing this review: PubMed, Google Gemini versions 2 and 3, and Undermind Research Assistant. We thank the expert peer reviewers for their rigorous, in-depth critiques, which significantly enhanced this review.

Our research is supported in part by the Intramural Research Program of the National Institutes of Health (ZIADE000719 and ZIADE000524).

Author contributions: Kenneth M. Yamada: conceptualization, funding acquisition, investigation, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing. David A. Cruz Walma: conceptualization, visualization, and writing—original draft, review, and editing. Wakako Endo: visualization and writing—original draft, review, and editing.