The cytoneme connection: direct long-distance signal transfer during development

Embryonic patterning is key to the development of complex multicellular organisms by establishing the body plan. Our understanding of developmental patterning is based on the concept of positional information, which proposes that cells acquire positional values and differentiate accordingly to give rise to specific spatial patterns (Wolpert, 1969). In 1952, Alan Turing proposed a mathematical reaction-diffusion system to establish a precise spatial pattern through the interaction of activator and repressor molecules, later named morphogens. Here, positional information is acquired through gradients of diffusible substances (Turing, 1952; Wolpert, 1969; Crick, 1970) produced by a localized group of cells (organizers) (Spemann and Mangold, 1924). Progress in developmental genetics, and molecular identification of biological morphogens, highlighted the functional importance of morphogens for embryonic development (Driever and Nusslein-Volhard, 1988; Green and Smith, 1990). Indeed, several morphogenetic signal gradients that provide positional information in embryos, are presently known (reviewed by Rogers and Schier, 2011). However, increasing experimental evidence has challenged simple diffusion models (Kornberg, 2014; Wolpert, 2016).

Today, the most common theory for morphogen gradient dispersal integrates three parameters: morphogen diffusion from a localized source; the rate of morphogen production; and the rate of morphogen degradation. In addition, interaction of morphogens with extracellular molecules also plays an active role in the regulation of their transport (reviewed by Restrepo et al., 2014). However, how morphogens spread through complex three-dimensional structures in embryonic tissues and give rise to a precise spatial pattern remains an unresolved issue. Another significant challenge is to decipher how only some cells are targeted by a specific signal, especially in a naïve tissue with a confluence of different morphogenetic signals. Furthermore, it is necessary to understand how target cells are selected when signaling cells are either acquiring different fates or are physically separated by other tissues. Finally, and considering that most morphogens are insoluble lipid-modified or integral transmembrane proteins, it is of equal importance to understand how their movement through the extracellular space is achieved. Taking all these points into account, the simple diffusion of morphogens does not seem to be a viable mechanism for precise cell positional information.

On the other hand, an alternative mode of distant signaling has been put forward, thanks to advances in imaging technology and the availability of genetic tools, reporter genes and fluorescent-tagged components. This proposal demonstrates that cells can communicate at a distance through cellular membrane protrusions. In this Primer, we focus on signaling filopodia (using the term ‘cytonemes’ for simplicity) as a conduit for the morphogen spatial distribution that generates patterns of gene expression and cell differentiation. Cytonemes act as ‘spatial communication guides’, allowing direct physical contact at distance between signal-sending and signal-receiving cell membranes (Ramírez-Weber and Kornberg, 1999). They overcome the challenge of signal transport through tortuous structures within the tissue, disperse membrane-bound signaling molecules and can provide spatial restriction, as well as specificity, in signaling. Here, we examine recent progress and hypothesize that synaptic-like processes could be a common mechanism for direct signal transfer. Finally, we also summarize remaining questions regarding the cellular mechanisms involved in cytoneme formation, guiding and dynamics, as well as the mechanisms for signal transfer during contact with recipient cells.

Thin membrane protrusions are becoming increasingly relevant for intercellular communication. They are found in different organisms and tissues, and are implicated in several functions (cell-cell communication during early development, cell migration, stem cell-mediated homeostasis and regeneration) and in the progression of particular types of cancer (reviewed by Caviglia and Ober, 2018; Mattes and Scholpp, 2018) (see Table 1). These variable functional and structural characteristics of cell protrusions have led to various terminologies (Fig. 1). Cell membrane protrusions are structurally defined depending on their closed- or open-ended tip, and on the presence of microtubules (MT) and/or an actin cytoskeleton (reviewed by Yamashita et al., 2018). Closed-ended actin-based structures for cell-cell communication have been called signaling filopodia or cytonemes (Fig. 1A). In addition to actin cytoskeleton, tubulin has also been observed restricted to the base of some cytonemes (Sanders et al., 2013; Stanganello et al., 2015). MT-nanotubes are microtubule-based protrusions that contact cells through closed-ended protrusions (Inaba et al., 2015) (Fig. 1B), whereas tunneling nanotubes (TNTs) and membrane nanotubes (also known as cellular tubes, bridges or conduits) are actin-based open-ended connections, although in some cases they have also been described to contain microtubules (Yamashita et al., 2018). TNTs and membrane nanotubes allow the exchange between cells of soluble cytoplasmic components, membrane-associated molecules, intracellular vesicles and even larger organelles (Rustom et al., 2004; Gerdes and Carvalho, 2008; Rustom, 2016) (Fig. 1C). Thus, there is an increasing number of reported intercellular communicating filopodia with apparent structural and functional differences, and we are just beginning to understand the possible categories. In this Primer, we focus on signaling filopodia (cytonemes) and MT nanotubes as close-ended protrusions for cell signaling.

Thin signaling filopodia were first observed during sea urchin gastrulation (Miller et al., 1995). They localize at sites in which cell-cell signaling is known to take place, but no migration occurs, suggesting a function that is independent from cell migration. Later, the term ‘cytoneme’ was coined to define the thin signaling filopodia in Drosophila imaginal discs (Ramírez-Weber and Kornberg, 1999). Here, cytonemes act as connectors to mediate the interaction of the receptor cells with ligand-producing cells, allowing for specific signal transduction between distant cells. Importantly, cytonemes are highly dynamic structures that permit ligand-receptor binding in a spatially restricted manner, with different extension and retraction velocities and lengths (see Table 1). These dynamics could allow the formation of adaptable and specialized contact sites between juxtaposed membranes.

Until recently, it was difficult to functionally assess the role of cytonemes in cell signaling without affecting other cellular processes. New approaches using brief pulses of inhibitors or specifically targeting, using RNA interference, actin cytoskeleton regulators in a short time-window allow for perturbing cytoneme formation without grossly disrupting overall cell morphology. These approaches have revealed a close correlation between cytoneme formation and signaling function (see Table 1). To date, cytonemes have been identified as signaling-specialized structures in Drosophila and vertebrates (amphibians, chick, zebrafish and mouse), and in different developmental models such as during the formation of signaling gradients or in the differentiation of specialized cells in a pattern (see Table 1).

The presence of cytonemes is correlated with paracrine transport of fundamental signaling molecules, including Notch, Spitz (Spi)/EGF, Branchless (Bnl)/FGF, Decapentaplegic (Dpp)/BMP, Wingless (Wg)/Wnt and Hedgehog (Hh)/Sonic hedgehog (Shh) (for references and details, see Table 1). Importantly, cytonemes have been described to be specifically associated with components of a particular signaling pathway, even when emerging from the same cells (Roy et al., 2011). In addition, cytonemes can transport either the pathway ligand (or ligand complex) or the receptor (or receptor complex), depending on whether they emanate from receiving or signal-producing cells, or from both, but always achieve spatially restricted distribution of signals.

Cytonemes generate the spatial distribution of morphogen molecules for paracrine signaling and can facilitate the graded dispersion of signals according to their extent and dynamics (see Table 1, Fig. 2A). Cytoneme-mediated delivery of signaling ligands has been shown for Wnt in zebrafish (Stanganello et al., 2015; Mattes et al., 2018), Hh in Drosophila and Sonic hedgehog (Shh) in chick limb bud (Bischoff et al., 2013; Gradilla and Guerrero, 2013; Sanders et al., 2013; Chen et al., 2017; González-Méndez et al., 2017). Conversely, cytonemes emanating from signal-receiving cells (see Table 1, Fig. 2B) have been shown to be important for the graded distribution of the Drosophila FGF receptor homolog Breathless (Btl) (Du et al., 2018) in the developing air sac primordia (ASP), a structure analogous to the human lung.

Cytonemes can also spatially distribute signals to differentiate specific target cells within an otherwise apparently uniform field, e.g. to form a salt and pepper pattern. In Drosophila, signaling filopodia direct spatially biased Spitz/EGF enhanced signaling to determine the formation of a structural bract cell at the proximal side of each mechanosensory organ in the leg, generating spatially patterned cell fates (Peng et al., 2012). Similarly, signaling filopodia direct spatially restricted lateral inhibition of Notch signaling to generate the organized pattern of bristles in the notum (De Joussineau et al., 2003; Cohen et al., 2010) (see Table 1). Cytonemes have also been observed during butterfly wing patterning, which suggests that signal delivered by cytonemes might determine the characteristic color patterns (Ohno and Otaki, 2015; Iwasaki et al., 2017); indeed, such patterns are not explained by signal diffusion alone (Iwata et al., 2018). In addition, cytoneme-mediated communication promotes the pigmented stripe pattern in zebrafish. Here, cytoneme-like protrusions called ‘airinemes’ connect pigmented cells precursors at a distance to facilitate Delta/Notch-mediated signaling, which contributes to the clearance of pigmented cells from the developing interstripe and the consolidation of the stripes (Hamada et al., 2014; Eom et al., 2015; Eom and Parichy, 2017) (see Table 1).

Cytonemes can also coordinate morphogenesis of physically separated tissues, bypassing intermediate areas. The development of the ASP in Drosophila is one such example (see Table 1). Cytonemes emanating from the ASP allow the ASP cell membranes to contact the distant membranes of the ligand-producing cells of the wing disc. These contacts facilitate Dpp and FGF signal transfer, which are only produced in the wing disc and are received by the respective receptors at cytonemes of the ASP, promoting both cell migration and patterning (Roy et al., 2014; Du et al., 2018; Huang et al., 2019; Sohr et al., 2019). In this system, cytonemes bypass intermediate tissues, such as the mesodermal layer, between the ASP and the imaginal disc for targeted signaling (Huang and Kornberg, 2015) (Fig. 4).

In a similar process, signaling filopodia can also facilitate communication between immobile and migrating cells or moving cells during dynamic morphogenetic processes. It has recently been shown in vertebrates that cytonemes mediate long-distance cell contacts between epithelial and mesenchyme-like tissues to facilitate EphB3b-ephrin B1 signaling, which guarantees collective hepatoblast movement during asymmetric placement of liver and gut (Cayuso et al., 2016). See Table 1 for other examples of the role of coordinating morphogenesis by cytonemes in physically separated tissues.

In addition to the coordination of developmental patterning, cytonemes are essential in the maintenance of restricted paracrine signaling between stem cells and their supportive niche (see Table 1). In the Drosophila ovary, cytonemes maintain the female germline stem cell (GSC) niche by paracrine Hh signaling at short and long distances. Here, they enable the specificity of signal transport from the support cap cells to the adjacent population of niche cells, the escort cells, and activate the transcription of BMP family members: Dpp and Glass bottom boat (Gbb) (Rojas-Ríos et al., 2012). Similarly, protrusions defined as MT-nanotubes in the Drosophila testis restrict BMP signaling to maintain GSCs in an undifferentiated and self-renewing state. These protrusions extend from male GSCs and direct the Dpp receptor Tkv to the interior positions in the niche in order to bind the Dpp produced by hub cells, without influencing other non-stem cells nearby (Inaba et al., 2015). Paracrine signaling for stem cell niche maintenance has also been described for the intestinal stem cell (ISC) niche in the Drosophila midgut. ISC cells are maintained by the enteroendocrine signal Bursicon, which acts at distance on the Lgr2 G-protein-coupled receptor (Scopelliti et al., 2014); however, to date no mechanism for signal transfer has been reported. Interestingly, the Lgr2 vertebrate orthologs and stem cell markers Lgr4 and Lgr5 promote cell membrane protrusions that resemble cytonemes in stem cell cultures in vitro (Snyder et al., 2015). Thus, future research might indeed further implicate cytoneme-mediated signal delivery to restrict ISC populations, both during development and in regenerative processes. Another example is in the Drosophila hematopoietic organ, the lymph gland, in which a small cluster of posterior signaling cells (PSC) controls the balance between multipotent prohemocytes and differentiating hemocytes. Here, PSCs also exhibit extensive protrusions that project towards the lymph gland and appear to be required for the maintenance of the undifferentiated cell population, probably by mediating signaling between PSCs and hematopoietic progenitors (Mandal et al., 2007; Fuwa et al., 2015).

In epithelial tissues, cytonemes emanate from specific membrane territories that have an intrinsic basolateral polarization (Fig. 2A). Furthermore, in Drosophila, basolateral positioning of the signaling ligands Hh, Wg, Delta and Spz, as well as their reception processes, have also been described (Steinhauer et al., 2013; Cohen et al., 2010; Huang and Kornberg, 2015; Chen et al., 2017; González-Méndez et al., 2017). This subcellular localization implicates a mechanism or mechanisms that drive both signaling components and machinery for the initiation of protrusion to the basolateral side (Fig. 3). In the Drosophila wing disc, an apico-basal activity gradient of the Rho GTPase Rac has been described to generally regulate filopodial polarization (Georgiou and Baum, 2010; Couto et al., 2017). This Rac gradient is regulated by adherent-junction (AJ) proteins, and drives both the position and shape of epithelial filopodia. Likewise, a vesicle-sorting mechanism has been described to transport signaling ligands to the basolateral side (Callejo et al., 2011; Bilioni et al., 2013; Steinhauer et al., 2013; Yamazaki et al., 2016; Sohr et al., 2019; reviewed by Gradilla and Guerrero, 2013; Guerrero and Kornberg, 2014) (Fig. 3). However, the regulatory mechanisms for cytoneme cargo upload have yet to be determined and whether the vesicle-recycling mechanism could also contribute to cytoneme formation. In zebrafish, Wnt8a at the plasma membrane recruits transducer of CDC42-dependent assembly protein 1 (Toca-1) that locally activates cytoneme nucleation (Ho et al., 2004; Stanganello et al., 2015). Hence, intracellular trafficking of the Wnt ligand would be key for the spatial localization of membrane protrusion and signaling.

There are increasing reports of vesicles or dynamic puncta on cytonemes, which again support a general process of intracellular vesicle trafficking for localized signaling territories at membranes (see Table 1). The use of protrusions as tracks for membrane vesicles was first suggested in the sea urchin (Miller et al., 1995). More recently, several membrane-bound signaling proteins have been observed trafficking through cytonemes. Indeed, the extracellular transport of membrane-anchored morphogens has been associated with extracellular particles, such as argosomes (Greco et al., 2001), lipoprotein particles (Panáková et al., 2005), exosomes (Gross et al., 2012; Koles et al., 2012; Beckett et al., 2013; Gradilla et al., 2014) and exosome-like particles (Danilchik et al., 2013; Matusek et al., 2014).

Association of extracellular vesicles (EVs) to cytonemes has been described for the Hh morphogen in Drosophila. In this case, cytoneme-mediated Hh graded distribution is dependent on protein complexes involved in the formation of multivesicular bodies (MVBs), which are necessary for the biogenesis of signaling EVs (Gradilla et al., 2014; Matusek et al., 2014; Vyas et al., 2014) (Fig. 3). Moreover, live-cell imaging shows that vesicles containing Hh travel along cytonemes and mediate the transfer of signal between linked cells (Gradilla et al., 2014; González-Méndez et al., 2017). A related example is that of the zebrafish ‘airinemes’, which deliver membranous vesicles containing the Notch ligand Delta from xanthoblast to distant dark pigmented cells or melanophores. This initiates Notch signaling, which contributes to melanophore migration towards stripes and away from the interstripe space (Eom et al., 2015). Interestingly, these vesicles also seem to have a role in the formation of ‘airinemes’ by mediating the association of xanthoblasts surface blebs and moving macrophages. Motile macrophages engulf intact ‘airineme’ vesicles and drag the vesicles and filaments from the Delta-producing cell as they move away; in doing so, the macrophages pull and extend the ‘airinemes’ from the bleb-contact sites to the melanophores for direct signaling (Eom and Parichy, 2017). Other reported vesicle-like puncta associated with cytonemes include the motor protein myosin X, which is observed along cytonemes transporting Shh in chick limb bud (Sanders et al., 2013) and in Wnt-bearing cytonemes in zebrafish (Stanganello et al., 2015). Wnt localization to EVs has also been shown in vertebrates and Drosophila, although no link between this vesicle type and cytonemes has been reported (Gross et al., 2012; Koles et al., 2012; Beckett et al., 2013).

Taken together, cytonemes can be envisioned as extensions of membrane territories that couple mechanisms for signaling and protrusion initiation. Intracellular vesicle trafficking has a role in the cellular localization of signaling mediated by cytonemes. Cytonemes, in turn, can also serve as tracks for the long-distance transport of vesicle-like structures (Fig. 3). In addition, latest research has revealed sorting of the Drosophila FGF ligand Bnl towards the basal side of the wing disc-producing cells, where the ASP-emanating cytonemes carrying the FGF receptor Btl make contact for ligand uptake (Sohr et al., 2019). This ligand sorting mechanism requires an intracellular protein cleavage process, which ensures there is sufficient ligand to sustain the cytoneme uptake required for appropriate graded dispersion and ASP morphogenesis (Sohr et al., 2019). However, further research is needed to elucidate the various contributions of vesicle dynamics in cytoneme-mediated cell-to-cell communication.

Another key issue to address is how cytonemes recognize and contact the target cell for signal transfer. Several studies have explored this and have identified two types of contacts: the cytoneme-cell body contact (Stanganello et al., 2015) and the cytoneme-cytoneme contact (González-Méndez et al., 2017). These two types are also conditioned by whether cytonemes carry ligands from ligand-producing cells or specific receptor complexes from responding cells, or both (see Table 1). For example, in the zebrafish embryo, the tips of cytonemes from Wnt8a-producing cells transfer the ligand by contacting the cell body of responding cells (Stanganello et al., 2015). Conversely, in Drosophila (Fig. 4A-D), cytonemes carrying the Frizzled (Fz) receptor from the ASP cells contact the Wg-producing cell bodies in the wing disc for signal reception (Huang and Kornberg, 2015) (Fig. 4C). Establishing the Dpp gradient in the Drosophila wing disc also depends on cytonemes containing the receptor Thickvein (Tkv). These cytonemes extend from cells situated on both sides of the source territory containing Dpp from producer cells, but no specific contact site has been defined (Fig. 4B) (Ramírez-Weber and Kornberg, 1999).

Experiments using GFP reconstitution across synaptic partners (GRASP) have defined precise physical contact sites of cytonemes from the Drosophila ASP. In the ASP, these cytonemes carrying receptor complexes contact distant Dpp- and FGF-producing cells of the wing disc (Roy et al., 2014; Du et al., 2018) (Fig. 4D). On the other hand, GRASP reconstitution experiments show that contacts at cytonemes from myoblasts carrying the Notch ligand Delta also contact the ASP, and vice versa, suggesting a possible cytoneme-cytoneme interaction to activate Notch signaling (Huang and Kornberg, 2015) (Fig. 4C).

For Hh distribution within Drosophila epithelia, GRASP indeed revealed direct cytoneme-cytoneme contact sites between distant producer and receptor cells all along their length (González-Méndez et al., 2017) (Fig. 4A). Similar contacts have been suggested for Shh signaling in the developing chick limb bud, where cytonemes from Shh-producing mesenchymal cells interact at defined territories along cytonemes that carry the Shh co-receptors and adhesion molecules BOC and CDO (Sanders et al., 2013). In addition, the Drosophila ortholog of CDO, Interference hedgehog (Ihog), is detected in several contact points along overlapping cytonemes. These contact points might work as specialized sites for Hh reception and contain other components of the Hh reception complex, such as Patched (Ptc) and the glypicans Dally and Dally-like (Dlp) (González-Méndez et al., 2017). As Hh ligand is membrane bound, such contacts might also define the subcellular location of proteolytic processing or shedding of anchored Hh, either from cytoneme membranes or from exovesicles released at the site; this would allow the release of active morphogen only in response to a precise signal (reviewed by Manikowski et al., 2018).

A signal transfer mechanism similar to a neuronal synapsis has been hypothesized (Kornberg and Roy, 2014; Roy et al., 2014; Chen et al., 2017; González-Méndez et al., 2017; Huang et al., 2019). A ‘synapse’ is an intercellular communication mechanism that implies close cell-cell specialized contact, which allows for specific targeted signaling and implies the intracellular localization of presynaptic and postsynaptic molecular machinery in specific membrane areas. Synaptic-like signaling through cytonemes shares these general characteristics by facilitating direct contact, even at several cell diameters of distance. Furthermore, the ligand presentation complex could be compared with the presynaptic molecules, and the receptor/receptor complex with the postsynaptic molecules in specialized areas; thus, the term ‘morphogenetic synapsis’ has been coined (reviewed by Kornberg, 2017).

As in synapses, MVBs containing Hh move along cytonemes in Drosophila epithelia. These MVBs are the immediately previous step to the secretion of Hh in exosomes, which could take place specifically at contact points between cytonemes from signal producer and receptor cells (Gradilla et al., 2014) (Fig. 5). Supporting this theory, the contact sites at membranes along cytonemes revealed annular structures that resemble specialized presynaptic membrane swellings (González-Méndez et al., 2017), termed synaptic boutons, where vesicles are released at the end of neurons during synapsis. Interestingly, both the Hh ligand and its receptor, Ptc, and co-receptor, Ihog, are localized in these annular structures (González-Méndez et al., 2017). Moreover, by using the GRASP technique it has been suggested that the size of the contact between cytonemes or between cytonemes and the target cell is comparable with that of immune and neuronal synapses (Roy et al., 2014). Furthermore, synaptic transfer of morphogenetic signals has also been observed at Drosophila larval neuromuscular junctions, where Wg is transported across synapses by vesicles released from presynaptic cells (Korkut et al., 2009).

If signal transmission is confined to the cytoneme-mediated contact site(s) between sending and recipient cells, signaling must then be dependent on the frequency and stability of these contacts (Fig. 2A). Therefore, it is crucial to identify the molecules responsible for establishing these synaptic-like contacts. To date, proteins exposed to the cellular surface, including the adhesion molecule Ihog, have been reported to be modifiers of cytoneme dynamics (Bischoff et al., 2013; González-Méndez et al., 2017). Such proteins could enable cytonemes to interact with surrounding cells, sensing both the extracellular matrix (ECM) and cell membranes. Indeed, ASP cytoneme extensions use Dally and Dally-like (Dlp) glypicans as substrates for growth, and are also dependent on laminin-activated integrins (Roy et al., 2014; Huang and Kornberg, 2015). In addition, synaptic adhesion proteins such as Neuroglian, Capricious and Neuroligin2 are required for establishing functional contacts between ASP cytonemes and wing disc ligand-sending cells (Roy et al., 2014; Huang and Kornberg, 2015; Huang et al., 2019). Therefore, it is conceivable that these membrane and ECM proteins play a role in cytoneme guidance as well as in the recognition and establishment of physical contact with the distant cell. Interestingly, very recent research suggests that both the cytoneme establishment and their contact in the ASP are influenced by glutamate-mediated signaling, which produces intracellular calcium influx waves. This work can even distinguish a requirement for distinct sets of presynaptic (Synaptotagmine 1, Synaptobrevin, Vesicle Glutamate Transporter, etc.) and postsynaptic (Synaptotagmine 4, Glutamate Receptor II) regulators for cytoneme-mediated Dpp signaling (Huang et al., 2019).

In summary, increasing evidence shows that the cytoneme connection resembles a synaptic contact, supporting cell-cell signaling in diverse biological contexts (Fig. 5). Keeping in mind marked differences regarding the complexity of molecules involved, contact duration, which in the case of neuronal synapsis can last up to months, or the link to a propagating electrical impulse, some key features are shared, such as cell polarization, directed secretion for communication, and specialized membrane domains containing specific receptors and cell adhesion molecules at the contact sites.

Another intriguing issue is how the spatial and temporal establishment of cytonemes is regulated to generate diversity, e.g. to generate different signaling gradient shapes for different tissue morphologies. In addition, several signaling pathways can co-exist in the same tissues, such as in the ASP and the wing disc in Drosophila (Fig. 4). Supporting the hypothesis that specific cytomemes exist for each signaling pathway, recent work proposes self-regulatory mechanisms for cytoneme-mediated signaling, in which activation of a signaling pathway fuels the regulation of protrusions carrying its own pathway components (Du et al., 2018; Mattes et al., 2018).

In the Drosophila ASP, genome-edited constructs for the FGF ortholog Bnl (Bnl:GFP) and FGF receptor ortholog Btl (Btl:Cherry) have shown that FGF moves specifically from the wing disc-producing cells to the receiving cytonemes in the ASP (Fig. 4D), forming a long-range concentration gradient that adapts to the developing ASP-specific shapes (Du et al., 2018; Sohr et al., 2019). Interestingly, this adaptation involves two targets of the pathway, cut and pointed-P1, which respond to either low or high levels of the ligand FGF, respectively. In turn, these transcription factors have antagonistic effects: Pointed-P1 enhances the number and length of FGF-receiving cytonemes, while Cut acts to repress protrusions. Thus, the gradient is a consequence of initial protrusions exposed to high levels of signaling close to the source, which in turn enhances cytoneme contact and signaling that will further promote protrusion extension that maintains high signaling levels. In contrast, lower levels of signaling will repress protrusion extension, which will translate into lower levels of signaling over time (Du et al., 2018).

A second feedback loop mechanism has been described for Wnt signaling where cytoneme transport of the zebrafish ligand Wnt8a acts in both paracrine and autocrine signaling. For paracrine signaling, Wnt8a activates the canonical Wnt pathway (Stanganello et al., 2015), while it triggers the alternative planar cell polarity (PCP) pathway through autocrine signaling by binding the co-receptor Ror2 (Mattes et al., 2018). PCP activation leads to the activation of actin polymerization processes, which are dependent on the small GTPase Cdc42, followed by outgrowth of a subset of new Wnt8a-carrying cytonemes, which facilitate more Wnt8a transport to activate canonical Wnt signaling at a distance. This feedback loop also regulates Wnt cytoneme formation in zebrafish neural plate patterning and fibroblasts, and in human embryonic kidney and gastric cancer cells (Mattes et al., 2018).

In stem cells, there is also evidence of signal reception mediated by cytonemes, which in turn regulate cytoneme extension. For example, Hh cytonemes within the stem cell niche of the adult Drosophila ovary extend until they receive the ligand; impairment of signal reception causes cytonemes to extend until ligand is reached and wild-type signaling is activated (Rojas-Ríos et al., 2012). In this way, the GSC niche responds to insufficient Hh signaling, which suggests an ability of cellular niches to respond to challenging physiological conditions. However, in the wing imaginal disc and abdominal histoblasts, both cytoneme formation and dynamics appear to be independent of Hh signal reception, because the protrusions form and dynamically extend in the absence of either Hh ligand or its receptor, Ptc (González-Méndez et al., 2017). Thus, there are still unanswered questions regarding the fundamental mechanisms for initial cytoneme formation and its links to pathway specificity and adaptability.

Increasing evidence shows that cells master distant communication through filopodial protrusions. These structures can direct the patterning of tissues, and facilitate both graded and selective distribution of signaling molecules. In addition, filopodial protrusions serve as connectors of distant associated tissues and have a pivotal role both during the development of and in the maintenance of adult tissue homeostasis. However, the distinction between different protrusions that perform cell-cell communication is not completely clear; the current classifications might indeed change as further knowledge is acquired.

Key issues to address include, among many others: understanding directionality of targeting during cytoneme formation and their dynamic extension; how cytonemes intercalate between non target cells either within a tissue or to reach physically separated tissues; and how the signaling pathway specificity of cytonemes is achieved. Cytonemes are indeed cytoskeleton-driven extensions of specialized membrane territories, which could couple the presence of signaling components, cytoneme formation and probably even the factors needed for contact. All these components are most probably dependent on intracellular vesicular trafficking, although further research is needed to confirm this hypothesis.

Regarding the cytoneme contact for information exchange, this might be a synapse-like process. Neurons, axons and dendrites are extended to signal through synapses with diverse distant target cells (chemical synapses). Various types of synapses also exist for the vertebrate immune system cells (immune synapses), which either synapse with each other or with tumor or infected cells (reviewed by Alcover et al., 2016). Interestingly, recent research has shown that human immunodeficiency virus can propagate by inducing, in addition to TNT, filopodia formation between infected and uninfected cells (Sherer and Mothes, 2008); this is also considered to be a synaptic process (virological synapses) (reviewed by Sewald et al., 2016; Dufloo et al., 2018). The morphogenetic synapsis (reviewed by Kornberg, 2017) could be envisioned as part of an evolutionarily conserved mechanism for cellular communication that supports cell-cell signaling in diverse biological contexts. The finding that DPP signaling in the ASP could be glutamatergic as in the chemical synapses (Huang et al., 2019) supports this hypothesis.

In summary, studying signaling mediated by cytonemes has clear implications in our understanding of normal development, directing signals to create patterns of cell differentiation; in adult tissues cytonemes could have a crucial role in the maintenance of balanced stem cell populations, influencing processes such as tissue regeneration and cancer progression. Indeed, organoid studies show that cytoneme-mediated Wnt signaling aids the maintenance of a constant supply of Wnt signal that is necessary for tissue homeostasis in the mouse intestinal crypt (Mattes et al., 2018). Regarding cancer progression, it has been reported that glioblastoma cells extend ultra-long membrane protrusions, similar to TNTs and cytonemes, that interconnect tumor cells (Osswald et al., 2015). These tumor microtubes contribute to invasion, proliferation, effective brain colonization and radioresistance of glioblastome cells (Osswald et al., 2015; Weil et al., 2017). Further research on cytoneme formation and signal transfer mechanisms will allow us to better understand disease processes and facilitate the potential finding of more directed therapies.

We are grateful to the members of I.G.’s lab for discussions during the development of this review, to Pedro Ripoll and Robert Wilson for editing and comments on the manuscript, and to José Ignacio Belio for figure designs assistance.