Available via license: CC BY 4.0
Content may be subject to copyright.
1
Glypiation enables FGF to self-regulate its tissue-specific dispersion and
interpretation through cytonemes
Lijuan Du, Alex Sohr, Sougata Roy*
Department of Cell Biology and Molecular Genetics; University of Maryland, College
Park, MD 20742
*Corresponding author: sougata@umd.edu
Keywords: FGF, bidirectional signaling, GPI anchor, cytonemes, synapse, self-
organization, Bnl, cell-adhesion-molecule (CAM), Drosophila
Short Title: Bidirectional FGF-FGFR signaling at cytoneme contacts
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
2
ABSTRACT
During development, a handful of signals sculpt diverse tissue architectures. How the
same signal produces different tissue/context-specific information and outcomes is
poorly understood. We explored the basis that programs tissue-specific FGF dispersion
and interpretation by cytoneme-mediated contact-dependent communication. Although
a Drosophila FGF was thought to be freely secreted, we discovered that it is glypiated
and GPI-anchored on the source cell surface, which inhibits non-specific secretion but
facilitates tissue-specific cytoneme contact formation and contact-dependent release.
For long-distance signaling, source and recipient cells extend FGF-containing and
FGFR-containing cytonemes that contact and recognize each other by CAM-like
receptor-ligand binding. FGF-FGFR binding reciprocally induces forward and reverse
signaling in recipient and source cells, responses of which polarize their cytonemes
toward each other to mutually self-sustain contacts. FGFR-bound FGF’s subsequent
unanchoring hand-delivers FGF to receiving cytonemes and dissociates contacts. Thus,
while cytonemes spatiotemporally control FGF dispersion/interpretation, FGF self-
regulates its tissue-specific signaling by controlling cytonemes.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
3
Introduction
During morphogenesis, cells spatiotemporally coordinate their differentiation, function,
and patterns of organization by communicating with signaling proteins or morphogens
(1). Genetic and molecular characterization of pattern-forming genes revealed that there
are only a handful of morphogenetic signals, including Fibroblast Growth Factor (FGF),
Hedgehog (Hh), Wingless (Wg)/Wnt, Epidermal Growth Factor (EGF), and
Decapentaplegic (Dpp - a BMP homolog). These signals disperse across tissues to
form concertation gradients and, upon binding to receptors in recipient cells, activate
gene regulatory pathways in a dose-dependent manner to control coordinated
responses in cells (1). Strikingly, activity of this same set of signals and pathways is
sufficient to specify a plethora of cell types and tissue patterns in diverse contexts,
suggesting their adaptive, context-specific functions. This raises a question on the exact
nature of the morphogenetic cues, especially in the way they inform cells of their
spatiotemporal identity, activity, and dynamic organization. Despite myriads of examples
of context-specificity, the mechanisms by which a signal might program and induce
diverse tissue-specific information and outcomes remained an open question.
The ability for a tissue to interpret a specific signal or 'competence' is an actively
acquired state in responding cells (2). In contrast, the signals are traditionally
envisioned to be non-selectively secreted in the extracellular space and randomly
dispersed by passive diffusion. However, recent advances in live microscopy revealed
that during development, cells actively regulate target-specific signal dispersion by
dynamically extending actin-based signaling filopodia named cytonemes (3).
Cytonemes and cytoneme-like signaling filopodia were discovered in different vertebrate
and invertebrate morphogenetic, stem cell, and disease contexts (4-9). Prevalence of
signaling filopodia and their essential roles for most signals, including Hh, Dpp, FGF,
EGF, and Wnt suggest that the contact-based signaling is an evolutionarily conserved,
basic mechanism.
Despite being contact-dependent, cytoneme-mediated signaling can produce diverse
tissue-specific signal dispersion and signaling patterns (10, 11). For instance, a
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
4
Drosophila FGF, Branchless (Bnl) is expressed by a group of cells in the larval wing
imaginal disc, but it forms a long-range dispersion gradient only within the disc-
associated air-sac primordium (ASP) that expresses the Bnl receptor Breathless
(FGFR/Btl) (10, 12). The ASP extends Btl-containing cytonemes that directly contact
Bnl-expressing disc cells and take up Bnl (10). Recipient-specific shapes of Bnl
gradients emerge due to signaling through a graded number of Bnl-receiving cytonemes
along the distal-proximal axis of the ASP. On the other hand, varied Bnl signaling levels
in different ASP regions activate different target genes that differentially feedback
control formation of Bnl-receiving cytonemes, establishing the graded pattern of Bnl-
receiving cytonemes. Consequently, robust recipient-specific shapes of signal and
signaling patterns emerge in precise coordination with ASP growth and patterning.
Initiation of cytoneme-mediated patterning and its tissue-specificity depend on where
and when cytonemes establish signaling contacts (10). However, little is known about
how tissue-specific signaling contacts are established and controlled in space and time
and why and how secreted signals are release only via cytoneme contacts. In this
paper, we addressed these questions using Bnl/FGF signaling in the Drosophila ASP.
Bnl has a single receptor, Btl, and its activation induces MAPK signaling. Despite
apparent similar stimulation, genetic analyses have revealed multiple functions for Bnl,
such as acting as a mitogen, morphogen, or chemoattractant in diverse contexts,
including stem cells, trachea, neuron/glial, blood cells, and cancer (8, 10, 12-18). We
discovered that while cytonemes control tissue-specific Bnl distribution and signaling,
Bnl glypiation programs the spatiotemporal origin, distribution, and plasticity of
cytoneme contacts, thereby self-regulating its tissue-specific dispersion and functions.
This provides a new mechanistic insight into how the context-specific signaling is
programmed and realized by self-organizing the signal dispersion process.
Results
Bidirectional contact matchmaking of Bnl-sending and Bnl-receiving cytonemes.
Inter-organ Bnl signaling between the ASP and wing disc provides a simple genetic
system for an unbiased interpretation of tissue-specific dispersion as Bnl is produced in
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
5
the wing disc and travels across a layer of interspersed myoblasts to the overlaying
ASP, which expresses its only receptor, Btl (Fig. 1A,B) (10). Bnl is a secreted FGF
family protein, but externalized Bnl (Bnlex) detected by anti-Bnl immunostaining of non-
permeabilized imaginal disc preparations (Bnlex; Materials and Methods) is not broadly
distributed in the myoblasts and extracellular plane of its wing disc source. Instead,
Bnlex puncta were exclusively localized tissue-specifically only on the basal surface of
wing disc source cells, and recipient ASP and ASP cytonemes (Figs. 1A,C). These
images are consistent with tissue-specific transport of Bnlex to the ASP in a receptor-
bound form along the surface of ASP cytonemes (10). More strikingly, compared to the
relatively large bnl expression domain, Bnlex puncta on source cells were asymmetrically
congregated at the sites where Btl-containing ASP cytonemes established contacts.
To examine if the distribution of Bnl in source cells is uniform or spatially biased toward
the ASP, we examined wing discs that overexpress Bnl:GFP and mCherryCAAX driven
by bnl-Gal4. Although both constructs were expressed under the same driver, only
Bnl:GFP puncta were highly enriched in the ASP-proximal source area (Figs. 1D-D";
S1A,A'). Furthermore, wing discs that expressed Bnl:GFP either under endogenous
control (bnl:gfpendo allele; (10)) or under bnl-Gal4, revealed that source cells in the ASP-
proximal area extend Bnl:GFP-containing filopodia or cytonemes toward the ASP (Figs.
1D; S1B). These results suggest that the cellular components in source cells
responsible for Bnl secretion, display, and/or delivery are polarized toward the ASP.
Because Bnl-containing filopodia from source cells had not been reported before, we
first examined their functions. Source cytonemes were detected in unfixed wing discs
that expressed a fluorescent membrane marker (e.g., CD8:GFP or CherryCAAX) either
in all of the bnl-expressing cells or in small clones of cells within the Bnl expressing
area. Three-dimensional XZY image projections revealed that each of the Bnl-
expressing columnar cells proximal to the ASP extended ~2-4 short (<15 m)
cytonemes perpendicularly from their basal surface (Figs. 1F-H; S1C-E; Movie S1;
Table S1). The organization of the source cells therefore can be described as polarized
for basal Bnl presentation with basal projections extending toward the ASP. This
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
6
organization is mirrored in the ASP, which, as we previously reported, exhibits polarized
Btl synthesis, Bnl reception, and cytoneme orientation toward source cells (10). Thus,
signal-sending and -receiving components polarize to face each other, forming a tissue-
level niche at the ASP:source interface to promote cytoneme-mediated interactions.
Time-lapse imaging of ex vivo cultured wing discs revealed that the ASP and source
cytonemes orient toward each other and transiently contact each other's tips, bases, or
shafts as they dynamically extend and retract (Figs. 1I-M; S1H-O; Table S2; Movies
S2a,b). Both cytoneme types had dynamic and repeated cycles of contact association-
dissociation, but source cytonemes had a shorter lifetime than ASP cytonemes.
Importantly, such transient cytoneme:cytoneme contacts at the interface of ASP and
source cells are persistent throughout larval development (Fig. S1H-M). This is
consistent with the model of a cytoneme signaling niche at the interface of the ASP abd
wing disc source.
Based on our previous results on cytoneme-mediated Bnl uptake in the ASP (10), Bnl is
likely to be exchanged at the inter-cytoneme contact sites. However, it was technically
challenging to capture Bnl exchange via dynamic cytoneme interactions. Therefore, we
genetically ablated the source cytonemes and analyzed non-autonomous Bnl dispersion
into the ASP. Similar to ASP cytonemes (19), the formation of source cytonemes could
be induced by overexpressing an actin modulator formin, Diaphanous (Dia), and were
suppressed by dia knockdown in the source (Fig. 2A-E). Notably, a constitutively active
form of Dia, DiaDAD-GFP, selectively localized at the ASP-proximal source area and
at cytoneme tips, which is consistent with a localized increase in f-actin polymerization
activity in the cytoneme-producing source area (Figs. 2C; S1F,G). Ablating source
cytonemes in bnl:gfpendo larvae by dia RNAi expression in the source led to the non-
autonomous reduction of Bnl:GFPendo uptake in the ASP, leading to abnormally stunted
ASPs (Fig. 2F-H). This suggested that source cytonemes are important to selectively
deliver Bnl to ASP cytonemes and cells.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
7
Inter-cytoneme Bnl exchange is consistent with previous reports that Hh and Wg are
both sent and received by cytonemes (20-22). However, dynamic interactions of Bnl-
exchanging cytonemes that are convergently polarized toward each other might also
indicate a possibility of contact-dependent reciprocal guidance of source and recipient
cytonemes. To test this possibility, we first ablated source cytonemes by dia RNAi
expression and analyzed the effects on ASP cytonemes. The absence of source
cytonemes non-autonomously reduced the long, polarized ASP tip cytonemes that
make signaling contacts with the source (Fig. 2I-K). Since randomly oriented short ASP
cytonemes were unaffected, Bnl-sending cytonemes are required for the formation of
only polarized Bnl-receiving cytonemes from the ASP.
To determine if ASP cytonemes also influence the polarity of source cytonemes, we
removed ASP cytonemes by expressing a dominant-negative form of Btl (Btl:DN) in the
trachea, as reported previously (12). A complete loss of the ASP and ASP cytonemes
also led to a corresponding non-autonomous loss of polarized source cytonemes (Fig.
2L). Btl:DN expression occasionally produced partial dominant-negative effects, yielding
stunted ASPs with few polarized cytonemes. Strikingly, the appearance of polarized
cytonemes in each of these ASPs correlated with the appearance of equivalent
numbers of source cytonemes, making stable cytoneme:cytoneme contacts (Fig. 2M-
O). These results suggested that the inter-cytoneme contacts induce ASP cells to
extend Btl-containing cytonemes toward the source and induce source cells to extend
Bnl-containing cytonemes toward the ASP. Thus, ASP and source cytonemes
reciprocally guide each other to establish contacts and/or selectively stabilize these
contacts.
Glypiation tethers Bnl to the source cell surface.
Note that Btl:DN cannot signal due to the lack of its intracellular domain, but can bind to
Bnl with its extracellular portion (12). When we probed for the distribution of externalized
Bnl in the Btl:DN expressing samples, Bnlex were selectively enriched on the source and
recipient cytonemes that established stable contacts with each other (Fig. 2P-P"). The
enrichment of surface Bnlex at the inter-cytoneme contacts suggested an interesting
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
8
possibility that the Btl-Bnl binding that is required for intercellular Bnl exchange might
also be the molecular basis for the source and recipient cytoneme contact formation.
However, this model assumes that Bnl is retained on the source cell surface to act as a
cell recognition molecule. How might a secreted protein be retained exclusively on the
source cell surface, without being randomly dispersed in the extracellular space?
A possible mechanism emerged from studies of the surface distribution of various
chimeric Bnl:GFP constructs expressed in cultured Drosophila S2 cells. Bnl is
synthesized as a 770 amino acid protein and is cleaved in the Golgi by Furin1 at residue
164 prior to the externalization of its truncated C-terminal signaling portion (Fig.3A;
(23)). Therefore, when S2 cells expressed a chimeric Bnl:HA1GFP3 construct with HA
(site 1) and GFP (site 3) flanking the Furin cleavage site, the HA-tag localized to the
Golgi and the truncated Bnl:GFP3 fragment was externalized. These observations led
us to hypothesize that a Bnl:GFP3Cherryc construct (Fig. 3A), which has an in-frame C-
terminal mCherry fusion, would externalize the truncated C-terminal fragment marked
with both GFP and mCherry. Unexpectedly, most membrane-localized Bnl:GFP3 puncta
(detected with GFPex) lacked mCherry, indicating the possibility of a second
intracellular cleavage at the C-terminus of the Bnl protein prior to externalization (Fig.
3B-B"').
Bioinformatic analyses (see Methods) revealed that the Bnl C-terminal tail has a 20
amino acid hydrophobic segment immediately adjacent to a hydrophilic region (Fig.
3A,A'), similar to signal sequences of pro-GPI-anchored proteins (pro-GPI-APs). The
signal sequences of pro-GPI-APs are cleaved and replaced with a GPI moiety in the
endoplasmic reticulum (ER) prior to trafficking to the cell surface, where these proteins
are anchored to the outer leaflet of the plasma membrane (24). Because the presence
of C-terminal tags does not prevent glypiation of pro-GPI-APs (25), we surmised that
Bnl glypiation might explain C-terminal cleavage in Bnl:HA1GFP3Cherryc and surface
distribution of only Bnl:GFP3. To investigate Bnl glypiation, we followed a standard
phosphoinositide phospholipase C (PI-PLC) assay (Figs. 3C-E; S2B-I; see Methods).
PI-PLC specifically cleaves most GPI moieties between the phosphate and glycerol
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
9
backbone, leading to the shedding of GPI-APs from the cell surface. When S2 cells
expressed Bnl:GFP3, Bn:HA1GFP3 (henceforth Bnl:GFP), untagged Bnl (detected with
Bnlex), and a control GFP-GPI construct (26), all constructs localized on the cell
surface and were shed by PI-PLC treatment (Figs. 3C,C’,E; S2B,F,I). In contrast, PI-
PLC did not shed a constitutively active Drosophila EGF construct, cSpitz:GFP (Fig.
3C,C’,E), which is tethered to the cell membrane due to palmitoylation (27).
In silico analyses (see Methods) identified S741 of Bnl as a probable glypiation site (-
site). In S2 cells, Bnl:GFP mutant constructs that either lacked the C-terminal 40 amino
acid residues (Bnl:GFPC), or contained mutated + and + sites (Bnl:GFP-m)
failed to localize on the producing cell surface, even in the absence of PI-PLC treatment
(Figs. 3C-E’; S2G,I). However, when we added the transmembrane domain of the
mammalian CD8a protein to the C-terminus of Bnl:GFPC, the protein (henceforth
Bnl:GFPC-TM) was tethered to the cell surface irrespective of PI-PLC treatment (Figs.
3C-E; S2H,I). To test the functionality of the Bnl C-terminal signal sequence, we fused it
to the C-terminus of a secreted sfGFP that also has the Bnl N-terminal signal peptide
(10). The resultant construct, bGFP-GPI, had the same localization as a GPI-AP and
was sensitive to PI-PLC treatment (Fig. 3C-E). Together, these results confirmed that
Bnl is a GPI-AP and that it has a C-terminal signal sequence that is replaced by a GPI
moiety after cleavage at S741.
To determine if Bnl is also glypiated in native wing disc cells, we performed
surface Bnlex immunostaining on ex vivo cultured wing discs (10) before and after PI-
PLC treatment (Fig. 3F-R). Native Bnlex puncta were concentrated on the disc source
cells near ASP cytonemes, but their levels were significantly reduced after PI-PLC
treatment (Fig. 3F-H). When Bnl and the Bnl:GFP, Bnl:GFPC, and Bnl:GFPC-TM
constructs were overexpressed under bnl-Gal4, PI-PLC treatment significantly reduced
levels of Bnlex and Bnl:GFPex, but not Bnl:GFPC-TMex (Fig. 3I-R). Similar to the
observations in S2 cells, Bnl:GFPCex, was not concentrated on the surface of source
cells even without PI-PLC treatment (Figs. 3N-O’,R). Instead, randomly dispersed
Bnl:GFPCex puncta surrounding the disc source suggested that Bnl:GFPCex is
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
10
readily secreted into extracellular space. These results confirmed that Bnl is GPI-
anchored to the wing disc source cells.
Bnl's GPI anchor ensures tissue-specific Bnl dispersion and signaling.
To investigate if GPI anchoring on the source surface specifies Bnl's tissue-specific
distribution, we imaged GPI-modified (Bnl:GFP) and non-GPI-modified (Bnl:GFPC and
Bnl:GFPC-TM) constructs expressed from the mCherrryCAAX-marked wing disc Bnl
source. Overexpressed Bnl:GFP was localized almost exclusively in disc producing
cells and target-specifically dispersed into the ASP (Fig. 4A; Movie S3). In contrast,
Bnl:GFP mutants that lack a GPI anchor (Bnl:GFPC and Bnl:GFPC168) were also
dispersed around disc source cells, without apparent spatial specificity (Fig. 4C; Movie
S4). On the other hand, Bnl:GFPC-TM puncta were restricted to only source cells and
a few ASP cells that were directly juxtaposed to the source (Fig. 4B; Movie S5). Unlike
Bnl:GFP in ASP cells, these Bnl:GFPC-TM puncta in the ASP were abnormally
colocalized with the source cell membrane, indicating a defect in release from their TM
tether (Figs. 4B; S3A-B'). These results suggest that Bnl's GPI anchoring inhibits only
random non-specific release but facilitates its tissue-specific dispersion.
To investigate if tissue-specific Bnl dispersion is critical for signaling outcomes, we
imaged the correlation of the spatial distribution patterns of GPI-modified and non-GPI-
modified Bnl:GFP variants and that of the MAPK signaling. When Bnl:GFP was
overexpressed from the wing disc source under bnl-Gal4, all of the ASP cells received
Bnl:GFP and induced MAPK (nuclear dpERK localization) signaling (Figs. 4D,K; S3C-
E'). As expected, Bnl:GFPC-TM distribution and activity was restricted to only a small
number of ASP tip cells (Figs. 4E,K; S3C). In contrast, a significant number of cells that
received Bnl:GFPC in the ASP failed to activate nuclear MAPK signaling (Figs. 4F,K;
S3F,G). It is clear that the precise coordination of signal dispersion, signaling, and
growth was lost with Bnl:GFPC expression. Interestingly, similar to Bnl:GFP,
Bnl:GFPC could also induce ASP overgrowth and it is likely that the growth is
controlled by unknown interactions of multiple signaling pathways.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
11
To analyze these signaling anomalies further, we generated ectopic gain-of-function
(GOF) clones of Bnl:GFP, Bnl:GFPC-TM, or Bnl:GFPC directly within the ASP
epithelium. We scored only small sized clones (2-3 cell diameter) within the ASP stalk
or transverse connective (TC) that normally do not receive Bnl or activate MAPK
signaling. Ectopic Bnl:GFP GOF clones in the ASP stalk induced dpERK within a
uniform distance of 2-3 cells surrounding the clone (Figs. 4H,H’,K; S3K). In contrast,
Bnl:GFPC-TM moved from its clonal source to only a few of the juxtaposed cells to
induce MAPK signaling in them and to organize them into an ectopic tracheal outgrowth
(Figs. 4I,K; S3H,I,K). On the other hand, Bnl:GFPC GOF clones had wider signal
dispersion than Bnl:GFP, but only a few random signal-receiving cells had dpERK (Figs.
4J,K; S3J,K). Collectively, all these results show that the precise spatial correlation of
signal dispersion, interpretation, and growth require GPI anchored retention of Bnl on
the source surface. These results also show that the GPI anchor enables Bnl to
generate adaptive, context-specific spatial patterns depending on the location of the
source and recipient cells.
To further test if the presence or absence of a surface tether is critical to determine
context-specific outcomes, we took advantage of Bnl's ability to attract tracheal branch
invasion on its source cell surfaces (14). According to the chemo-gradient model that
has been proposed for Bnl's function as an inducer of tracheal branch migration,
random extracellular Bnl:GFPC dispersion might promote tracheal chemotaxis toward
its source. To examine, we ectopically expressed the GPI-modified and non-GPI-
modified Bnl:GFP variants in the larval salivary gland, a non-essential, trachea-free
organ that normally does not express Bnl (Fig. 4L-O; see Methods) (23). Surprisingly,
although Bnl:GFP and Bnl:GFPC-TM induced tracheal invasion and branching into the
signal-expressing salivary glands, Bnl:GFPC did not. Bnl:GFPC-TM is a TM-tethered
Bnl:GFPC, yet it induced extensive tracheation on salivary glands. Thus, tracheal
migration and branching phenotypes do not correlate with the random extracellular
presence of extracellular Bnl. On the contrary, increased tracheal invasion and
branching on producing cells do correlate with the levels of Bnl available on the basal
surface of source cells. As shown in Figure 4P-S, the amount of Bnl:GFPC on the
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
12
surface of the salivary gland is significantly less than either Bnl:GFP or Bnl:GFPC-TM.
These results show that retention of Bnl on the source surface is the key factor for Bnl
target-specific dispersion and interpretation. We proposed that cytoneme-mediated
exchange between source and target cells is the mechanism that links source surface
Bnl retention with its long-range context-specific dispersion and interpretation and that
GPI-anchored Bnl is the driver of the contact-dependent signaling through cytonemes.
GPI-anchored Bnl acts as a CAM to drive bidirectional matchmaking of source
and ASP cytonemes.
How does the addition of a GPI moiety enable Bnl to drive dynamic assembly-
disassembly of cytoneme contacts and mediate contact-dependent signal release? We
tested an idea that GPI-anchored Bnl acts as a cell surface CAM that mediates
heterotypic receptor-ligand-dependent target recognition and bidirectional contact
matchmaking of cytonemes. This model also predicts that, like other CAMs (28),
contact-dependent Btl-Bnl signaling is bidirectional, which would reciprocally activate
source and recipient cells to project cytonemes toward each other to form signaling
contacts. Finally, binding by Btl present on recipient cytonemes also would release Bnl
from its anchor to the source membrane and dissociate cytoneme contacts, thereby
restricting Bnl release only to the target-specific cytoneme contacts.
Results from Btl:DN experiments corroborated with Btl's role as a CAM (Fig. 2L-P’’).
Although Btl:DN cannot activate MAPK signaling, it retains the extracellular Bnl-binding
domain (12), and therefore Btl:DN-containing ASP cytonemes were capable of
establishing inter-cytoneme contacts and inducing polarized source cytoneme formation
(Fig. 2N-P’’). To test if surface-tethered Bnl has CAM-like bidirectional activity, we
investigated how GPI-modified and non-GPI-modified Bnl:GFP variants affect source
and ASP cytonemes (Figs. 5A-I; S4A-I). Although Bnl:GFP overexpression abnormally
brought a large number of source and recipient cells into close proximity due to the ASP
overgrowth, both tissues extended polarized cytonemes toward each other (Figs. 5A-C;
S4H,I). Increased extension-retraction dynamics of ASP cytonemes suggested high
signaling activity (Movie S6; Table S2). In contrast, overexpression of Bnl:GFPC
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
13
lacked CAM activity and significantly suppressed the formation of long polarized
cytonemes in both the source and ASP cells (Figs. 5D-F; S4A,B,H,I). Short cytonemes,
when detectable, lacked any directional bias and rarely contained Bnl:GFPC. These
results are consistent with the idea that GPI-anchored Bnl induces CAM-like
bidirectional contact matchmaking.
Importantly, unlike freely dispersed Bnl:GFPC, Bnl:GFPC-TM induced both ASP and
source cells to extend large numbers of long cytonemes that were stably connected with
each other via their tips and lateral sides (Figs. 5G-J’; S4C-K; Movies S8-10). The
polarized responses induced by Bnl:GFPC-TM were stronger than either Bnl:GFP or
WT (native Bnl). Bnl:GFPC-TM puncta localized at multiple lateral contact interfaces of
the source and recipient cytonemes, indicating the increased number and stability of
contact sites (Figs. 5G,J-J’; S4E-F; Movies S9-10). The increase in the stability of the
contacts might also account for an increase in the intensity of their bidirectional
responses. These results show that Bnl surface anchoring, irrespective of the nature of
the anchor, is sufficient for its CAM-like bidirectional signaling and contact
matchmaking.
To examine whether heterotypic CAM-like Bnl-Btl interactions establish inter-cytoneme
contacts, we took advantage of the relatively stable cytonemes produced by
Bnl:GFPC-TM. Bnl:GFPC-TM was expressed in CD4:IFP2-marked wing disc source
cells of larvae that also harbored a btl:cherryendo knock-in allele (10) expressing
endogenous Btl:Cherryendo from the ASP. BtlCherryendo puncta localized in close
proximity to Bnl:GFPC-TM puncta at multiple sites along the length of the juxtaposed
source and recipient cytonemes (Fig. 6A-B). These colocalized puncta mimicked the co-
clustered organization of receptors and ligands at the inter-cytoneme junctions (Fig. 2P-
P’’). These results, together with the previous evidence (Fig. 1A-C), show that the CAM-
like Btl-Bnl-binding through cytoneme contacts is the basis of contact recognition and
bidirectional matchmaking.
GPI anchor ensures Bnl's contact-dependent release and contact plasticity.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
14
Although Bnl:GFPC-TM acts as a CAM to induce strong bidirectional responses that
manifest in target-specific cytoneme polarity and inter-cytoneme contacts, Bnl:GFPC-
TM-exchanging cytonemes had a significantly longer lifetime than WT or Bnl:GFP-
exchanging cytonemes (Figs. 6C,D; S4J,K; Movies S8-S11; Table S2). Time-lapse live
imaging of source and ASP cells showed that the Bnl:GFPC-TM-exchanging
cytonemes made stable connections that resist signal release, and therefore also resist
contact dissociation, leading to cytoneme breakage (Movie S10). Thus, the GPI anchor
is required for the release of Btl-bound Bnl at the contact sites. This process is also
likely to trigger cytoneme contact dissociation.
This selective release mechanism suggests that the CAM-like function of GPI-anchored
Bnl is a prerequisite for Bnl release and subsequent morphogen-like interpretation. As a
result, Bnl:GFPC-TM exhibits efficient CAM-like cytoneme contact assembly, but is
restricted in functional range due to its lack of release at the contact sites (Figs. 4D-K;
6E). Similarly, this explains why Bnl:GFPC that lacks CAM activity also fails to induce
subsequent morphogen-like patterning of Bnl signaling (Fig. 4D-K). Only the GPI
modification integrates both tissue-specific contact formation and contact-dependent
release, leading to the diverse context-specific signaling patterns.
DISCUSSION
This study uncovered an elegant programming of a tissue-specific signaling mechanism
that is encoded by the lipid-modification of an FGF family morphogen, Bnl and
orchestrated by cytoneme-mediated contact-dependent communications. On one hand,
it explains how cytonemes can select a specific target to establish signaling contacts,
exchange signals at these contact sites, and inform cells where they are, what they
should do, and when (Fig. 6E). On the other hand, it shows how glypiation enables the
signal to program and drive these cytoneme-mediated events to self-regulate its diverse
tissue-specific journey and signaling. We provided evidence that the self-regulatory
interplay between signals and signaling cells, controlling each other's location and
activity through cytonemes, is essential for context-specific self-organization of signal
dispersion and interpretation in coordination with tissue growth.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
15
Traditionally, most secreted signals are envisioned to orchestrate morphogenesis by
diffusing away from the source and then activating the receptor and gene-regulatory
pathway only in recipient cells. According to this one-way directive from a signal to
signaling outcomes, signal retention in the source is inhibitory for its subsequent long-
range signal spread and interpretations. In contrast, we discovered that Bnl is GPI
anchored to the source cell surface and although this modification inhibits its non-
specific release and random diffusion, it facilitates Bnl's target-specific cytoneme-
mediated long-range spread and bidirectional signaling.
Mechanistically, GPI-anchored Bnl modulates three interdependent events: cytoneme-
mediated target selection, target-specific signal release, and feedback regulations of
these processes. We showed that GPI-anchored Bnl on the source cell surface and Btl
on the receiving cell surface act as CAMs. This enables Bnl-producing and receiving
cells to adhere with and recognize each other by Btl-Bnl binding. However, these cells
are situated in two different tissues and are separated in space. Therefore, both Bnl
source and recipient cells dynamically extend cytonemes to present Bnl and Btl on their
surfaces and recognize each other through Btl-Bnl interactions at their contact sites.
The CAM-like Btl-Bnl binding induces forward signaling in the ASP and retrograde
signaling in the source, feedback responses of which promote these cells to extend
more Bnl-receiving and Bnl-sending cytonemes toward each other and selectively
stabilize their mutual contact sites. Our results suggest that by employing these
reciprocal cytoneme-promoting feedbacks, source and ASP cells inform each other of
their relative locations, polarity, and signaling activities and establish a convergently
polarized signaling niche to self-sustain contact-mediated interactions. Here, the
molecular, cellular, and tissue level events are interdependent and are integrated by
cytoneme contacts.
The process of cytoneme contact assembly is reminiscent of CAM-mediated
bidirectional matchmaking of pre-synaptic and post-synaptic filopodial matchmaking that
organizes Drosophila neuromuscular junctions (29). Bidirectional transmission of
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
16
information is the basis for neuronal synaptic assembly, plasticity, and functions.
However, the purpose of Btl-Bnl-dependent contact recognition is to subsequently
release the Btl-bound Bnl at the cytoneme contact sites. Bnl release not only initiates
target-specific Bnl delivery and long-range morphogen-like functions in the ASP (10) but
also initiates cytoneme contact disassembly. Based on functional attributes of GPI-
modified and non-GPI-modified Bnl variants, we predict that the plasticity of signaling
contacts modulate the levels and duration of Bnl signaling and is also the key for
spatiotemporal adaptability/plasticity in the emerging shapes.
Importantly, the GPI anchor enables Bnl to simultaneously modulate both the
development of target-specific cytoneme contacts and contact-dependent signal
release, thereby mechanistically linking these two events to be interdependent in time
and space via signaling contacts. A consequence is that the cause and effects of Bnl
signaling can control each other via cytonemes. The same cytoneme contacts that Btl-
Bnl binding helps to establish also bring Btl and Bnl molecules together to interact.
Thus, not only is the signal exchange cytoneme or contact-dependent, but the
cytoneme contacts are also formed signal- or tissue-specifically. Similarly, Bnl and Btl
each can act as both a ligand and a receptor for the other and relay information from
inside-out and outside-in across the cell membrane. Consequently, contact-dependent
Btl-Bnl signaling is bidirectional, specifying functions in both source and recipient cells.
How the local contact-dependent Bnl exchange might self-organize large-scale tissue-
specific dispersion/signaling patterns can be explained from our previous findings (10).
Collectively, these results suggest that cytoneme-mediated signal exchange is designed
to induce context-specific self-organization of signaling patterns. Self-organization is an
inherent property of living systems. How diffusion and interactions of morphogens might
generate self-organized patterns were first theorized by Alan Turing in 1952 (30).
However, the evidence for a mechanism by which signal dispersion might program self-
organizing outcomes remained elusive. This study provides an evidence for a
mechanism that is orchestrated by cytoneme-mediated interactions and is encoded by a
lipid-modified morphogen.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
17
Post-translational modifications are common to most secreted morphogens. Hh, Wnt,
and EGF/Spi are known to be lipid-modified and, therefore, are restricted for secretion
and spread (27, 31, 32). Signals that are not known to be lipidated, including BMPs and
many FGFs, need to interact with GPI-anchored glypicans to produce morphogenetic
outcomes (33, 34). Proteoglycans restrict free release of the interacting signals and
show biphasic activation and inhibition effects on signal spread (35). Similar to Bnl, non-
lipidated Hh, Spi, and Wnt have unrestricted spread but poor tissue organizing potency,
but their TM-tethered forms can induce morphogen-like patterning despite a restricted
functional range (27, 36-40). These results suggest a significant role of signal retention
in morphogenesis. However, the functional links of the diverse signal retention
strategies to either cytoneme-mediated tissue-specific dispersal or context-specific self-
organization of signaling need to be examined in the future.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
18
References
1. N. Perrimon, C. Pitsouli, B.-Z. Shilo, Signaling mechanisms controlling cell fate
and embryonic patterning. Cold Spring Harbor Perspectives in Biology. 4,
a005975–a005975 (2012).
2. A. Sagner, J. Briscoe, Morphogen interpretation: concentration, time,
competence, and signaling dynamics. Wiley Interdiscip Rev Dev Biol. 6, e271
(2017).
3. F. A. Ramírez-Weber, T. B. Kornberg, Cytonemes: cellular processes that project
to the principal signaling center in Drosophila imaginal discs. Cell. 97, 599–607
(1999).
4. T. A. Sanders, E. Llagostera, M. Barna, Specialized filopodia direct long-range
transport of SHH during vertebrate tissue patterning. Nature. 497, 628–632
(2013).
5. T. B. Kornberg, S. Roy, Cytonemes as specialized signaling filopodia.
Development. 141, 729–736 (2014).
6. D. S. Eom, D. M. Parichy, A macrophage relay for long-distance signaling during
postembryonic tissue remodeling. Science. 355, 1317–1320 (2017).
7. Y. M. Yamashita, M. Inaba, M. Buszczak, Specialized Intercellular
Communications via Cytonemes and Nanotubes. Annu. Rev. Cell Dev. Biol. 34,
59–84 (2018).
8. S. Fereres, R. Hatori, M. Hatori, T. B. Kornberg, Cytoneme-mediated signaling
essential for tumorigenesis. PLoS Genet. 15, e1008415 (2019).
9. C. Zhang, S. Scholpp, Cytonemes in development. Current Opinion in Genetics &
Development. 57, 25–30 (2019).
10. L. Du, A. Sohr, G. Yan, S. Roy, Feedback regulation of cytoneme-mediated
transport shapes a tissue-specific FGF morphogen gradient (2018),
doi:10.7554/eLife.38137.
11. R. Hatori, T. B. Kornberg, Hedgehog produced by the Drosophila wing imaginal
disc induces distinct responses in three target tissues. Development. 147,
dev195974 (2020).
12. M. Sato, T. B. Kornberg, FGF is an essential mitogen and chemoattractant for the
air sacs of the drosophila tracheal system. Developmental Cell. 3, 195–207
(2002).
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
19
13. D. Sutherland, C. Samakovlis, M. A. Krasnow, branchless encodes a Drosophila
FGF homolog that controls tracheal cell migration and the pattern of branching.
Cell. 87, 1091–1101 (1996).
14. A. Ochoa-Espinosa, M. Affolter, Branching morphogenesis: from cells to organs
and back. Cold Spring Harbor Perspectives in Biology. 4, a008243–a008243
(2012).
15. F. Chen, M. A. Krasnow, Progenitor outgrowth from the niche in Drosophila
trachea is guided by FGF from decaying branches. Science. 343, 186–189
(2014).
16. L. Du et al., Unique patterns of organization and migration of FGF-expressing
cells during Drosophila morphogenesis. Dev. Biol. 427, 35–48 (2017).
17. J. V. Dos Santos, R. Y. Yu, A. Terceros, B. E. Chen, FGF receptors are required
for proper axonal branch targeting in Drosophila. Mol Brain. 12, 84 (2019).
18. M. Destalminil-Letourneau, I. Morin-Poulard, Y. Tian, N. Vanzo, M. Crozatier, The
vascular niche controls Drosophila hematopoiesis via fibroblast growth factor
signaling. Elife. 10, 45 (2021).
19. S. Roy, H. Huang, S. Liu, T. B. Kornberg, Cytoneme-mediated contact-dependent
transport of the Drosophila decapentaplegic signaling protein. Science. 343,
1244624 (2014).
20. L. González-Mendez, I. Seijo-Barandiarán, I. Guerrero, Cytoneme-mediated cell-
cell contacts for Hedgehog reception. Elife. 6, 605 (2017).
21. W. Chen, H. Huang, R. Hatori, T. B. Kornberg, Essential basal cytonemes take up
Hedgehog in the Drosophila wing imaginal disc. Development. 144, 3134–3144
(2017).
22. B. Mattes et al., Wnt/PCP controls spreading of Wnt/β-catenin signals by
cytonemes in vertebrates. Elife. 7, 180 (2018).
23. A. Sohr, L. Du, R. Wang, L. Lin, S. Roy, FGF cleavage is required for efficient
intracellular sorting and intercellular dispersal. J Cell Biol (2019),
doi:10.1083/jcb.201810138.
24. S. Saha, A. A. Anilkumar, S. Mayor, GPI-anchored protein organization and
dynamics at the cell surface. J. Lipid Res. 57, 159–175 (2016).
25. I. W. Caras, An internally positioned signal can direct attachment of a
glycophospholipid membrane anchor. J Cell Biol. 113, 77–85 (1991).
26. V. Greco, M. Hannus, S. Eaton, Argosomes: a potential vehicle for the spread of
morphogens through epithelia. Cell. 106, 633–645 (2001).
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
20
27. G. I. Miura et al., Palmitoylation of the EGFR ligand Spitz by Rasp increases Spitz
activity by restricting its diffusion. Developmental Cell. 10, 167–176 (2006).
28. R. O. Hynes, Integrins: bidirectional, allosteric signaling machines. Cell. 110, 673–
687 (2002).
29. H. Kohsaka, A. Nose, Target recognition at the tips of postsynaptic filopodia:
accumulation and function of Capricious. Development. 136, 1127–1135 (2009).
30. J. B. A. Green, J. Sharpe, Positional information and reaction-diffusion: two big
ideas in developmental biology combine. Development. 142, 1203–1211 (2015).
31. J. A. Porter, K. E. Young, P. A. Beachy, Cholesterol modification of hedgehog
signaling proteins in animal development. Science. 274, 255–259 (1996).
32. K. Willert et al., Wnt proteins are lipid-modified and can act as stem cell growth
factors. Nature. 423, 448–452 (2003).
33. D. Yan, X. Lin, Shaping morphogen gradients by proteoglycans. Cold Spring
Harbor Perspectives in Biology. 1, a002493–a002493 (2009).
34. R. Balasubramanian, X. Zhang, Mechanisms of FGF gradient formation during
embryogenesis. Semin. Cell Dev. Biol. 53, 94–100 (2016).
35. K. Dejima, M. I. Kanai, T. Akiyama, D. C. Levings, H. Nakato, Novel contact-
dependent bone morphogenetic protein (BMP) signaling mediated by heparan
sulfate proteoglycans. J. Biol. Chem. 286, 17103–17111 (2011).
36. F. R. Taylor et al., Enhanced potency of human Sonic hedgehog by hydrophobic
modification. Biochemistry. 40, 4359–4371 (2001).
37. A. Gallet, L. Ruel, L. Staccini-Lavenant, P. P. Thérond, Cholesterol modification is
necessary for controlled planar long-range activity of Hedgehog in Drosophila
epithelia. Development. 133, 407–418 (2006).
38. A. Callejo, C. Torroja, L. Quijada, I. Guerrero, Hedgehog lipid modifications are
required for Hedgehog stabilization in the extracellular matrix. Development. 133,
471–483 (2006).
39. K. F. Speer et al., Non-acylated Wnts Can Promote Signaling. CellReports. 26,
875–883.e5 (2019).
40. C. Alexandre, A. Baena-Lopez, J.-P. Vincent, Patterning and growth control by
membrane-tethered Wingless. Nature. 505, 180–185 (2014).
41. C. Cabernard, M. Affolter, Distinct roles for two receptor tyrosine kinases in
epithelial branching morphogenesis in Drosophila. Developmental Cell. 9, 831–
842 (2005).
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
21
42. N. M. Hooper, Determination of glycosyl-phosphatidylinositol membrane protein
anchorage. Proteomics. 1, 748–755 (2001).
43. N. A. Dye et al., Cell dynamics underlying oriented growth of the Drosophila wing
imaginal disc. Development. 144, 4406–4421 (2017).
Acknowledgments
We thank Drs. N. Andrews and T.B. Kornberg for reading and comments on the
manuscript; Drs. T.B. Kornberg and G. O. Barbosa for sharing the design of the culture
chamber for live imaging; the Bloomington Stock Center for Drosophila lines; the
Developmental Studies Hybridoma Bank for antibodies; Dr. A.E. Beaven for the help in
the UMD imaging core facility. Funding: NIH grant R35GM124878 to SR. Author
Contributions: A. Sohr discovered GPI-anchored Bnl and L. Du discovered its roles; S.
Roy supervised the work and designed the project; L. Du and A. Sohr designed and
conducted the experiments: S. Roy, L. Du, and A. Sohr wrote the paper. Competing
interests: None declared.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
22
Legends
Fig. 1. Bnl source and recipient cells extend cytonemes to contact each other. (A)
Drawing depicting organization of the ASP, wing disc, myoblasts, and Btl-containing
ASP cytonemes that directly receive Bnl from the disc bnl source. (B) Polarized ASP
cytonemes (green; arrow) establish contacts with the wing disc bnl source (red). (C)
Externalized Bnlex (red, Bnlex) is asymmetrically enriched on the source (dashed lined)
and Btl:GFP-containing ASP cytoneme contact sites (arrow). (D-D") The wing disc bnl
source (red) co-expressing Bnl:GFP and mCherryCAAX (bnl-Gal4 X UAS-Bnl:GFP,
UAS-mCherryCAAX); arrowhead, ASP-proximal area; arrow, Bnl:GFP containing
cytonemes; dashed line, ASP; D", Bnl:GFP intensity profile within the boxed source
area in the direction of the arrow in D' (see Fig. S1A-B). (E,F) 3D-rendered XZY views
of the nlsGFP-marked ASP and mCherryCAAX-marked source cytonemes (arrow). (G)
3D-rendered views of two CD8:GFP-expressing clones within the bnl source area;
arrow, cytonemes (see Fig. S1D,E; Table S1). (H) Violin plot showing the source
cytoneme length distribution. (I-K) Dynamic contact-based interactions of source (red)
and ASP (green) cytonemes; K, illustration of the results; arrowhead, contact sites.
(L,M) Plots comparing ASP and source cytoneme dynamics (see Material and Methods;
Fig. S1N,O; Table S2). All except C, live imaging. Scale bars, 20 m; 5 m (G).
Fig. 2. Bidirectional contact matchmaking of Bnl sending and receiving
cytonemes. (A-E) Autonomous effects of Dia:GFP, DiaDAD:GFP and diaRNAi
expression on cytonemes in mCherryAAX-marked source cells; all panels, 3D-rendered
XZY views; E, graph showing source cytoneme numbers under indicated conditions; p <
0.01 for WT (n = 6) vs Dia (n = 6) or DiaDAD (n = 8) or dia-i (n = 6); (UAS-
mCherryCAAX;bnl-Gal4 X w- for control, or UAS-"X"). (F-H) Bnl:GFPendo uptake in the
ASP (dashed line) in control (bnl:gfpendo X bnl-Gal4) and source cytoneme-depleted
conditions (UAS-dia-i, bnl:gfpendo X bnl-Gal4); H, violin plot showing levels of
Bnl:GFPendo uptake in the ASP under indicated conditions; p < 0.0001, n = 7 (w-) and 17
(dia-i); red, phalloidin-Alexa-647. (I-K) Non-autonomous effects of source cytoneme-
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
23
depletion on ASP cytonemes (CD2:GFP-marked); K, Plots comparing ASP cytoneme
numbers under indicated conditions, p < 0.0001 for cytonemes >15 m, n = 6 ASPs
(control), 14 ASPs (dia-i); dashed arrow, ASP tip cytonemes. (L-O) Tracheal Btl:DN
expression that removed ASP cytonemes also non-autonomously removed source
cytonemes (red); O, graph showing correlation of source and ASP cytoneme numbers
under partial Btl:DN conditions. (P-P") Bnlex (blue, Bnlex) localization at inter-cytoneme
contacts in samples with partial Btl:DN phenotypes; P', split Bnlex channel; P",
zoomed-in part of P. (F-N) dashed arrow, non-autonomous effects of the indicated
genetic manipulation (arrow); arrowheads, inter-cytoneme contacts. All panels except
F,G, P-P’’, live imaging. Scale bars, 20 m.
Fig. 3. Bnl is GPI anchored to the source cell surface. (A,A') Schematic map of the
Bnl protein showing FGF domain, secreted signaling portion (left-right arrow), signal
peptide (SP), signal sequence (SS), and sites for: furin cleavage (arrow), HA-tag (site
#1), GFP-tag (site # 3), mCherry-tag, hydrophobicity plot (A'), and numbers, amino acid
residues. (B-E’) S2 cells co-transfected with actin-Gal4 and UAS-"X" constructs and
surface immuno-probed in non-permeabilized condition with either GFP or Bnl antibody,
as indicated. (B-B’’’) In S2 cells, Bnl:HA1GFP3Cherryc (arrowheads) is cleaved
intracellularly, and Bnl:GFP3ex portion (arrow; blue) is surface localized; B’-B’’’, split
channels. (C-E) Surface localization (red) of various constructs under pre-PIPLC and
post-PIPLC conditions; CD8:GFP was co-expressed for untagged Bnl as an internal
control; D, schematic maps of different Bnl constructs; E, E’, Graphs comparing the ratio
of surface (red) to total protein (GFP) expressed for indicated constructs and conditions
(also see Fig. S2); ***, p<0.001, n=15 cells. (F-K) Surface Bnlex (red) levels on the wing
disc bnl source (arrows) before and after PIPLC treatment; F-H, native Bnl in w- control
larvae; I-K, overexpressed Bnl under bnl-Gal4. (L-R) Source surface levels (red) of
Bnl:GFP, Bnl:GFPC and Bnl:GFPC-TM on wing discs when expressed under bnl-
Gal4 (control) before and after PIPLC; R, graph comparing the fraction of source
surface levels (red, GFPex) to the total expression levels (green) of different variants;
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
24
n= 5 for each; only merged and red channels shown; arrows, source area; *, source-
surrounding disc area; dashed line, ASP; Scale bars, 30 m; 10 m (B-C’).
Fig. 4. GPI anchor ensures tissue-specific Bnl dispersion and interpretation. (A-C)
Dispersion patterns of Bnl:GFP, Bnl:GFPC-TM (TM), or Bnl:GFPC (C) from the
wing discs bnl source (UAS-mCherryCAAX; bnl-Gal4 X UAS-"X"); B, arrowhead,
abnormal source membrane colocalized Bnl:GFPC-TM puncta; insets, zoomed-in ROI
(box); dashed line, ASP; *, source-surrounding disc area; Discs large (Dlg), cell
outlines. (D-K) Dispersion and signaling (dpERK, red) patterns of Bnl:GFP, TM, and C
when expressed from either the wing disc bnl source (D-F; bnl-Gal4 x UAS-"X") or
ectopic GOF clones within the ASP (H-J; hsFlp; btlenh>y+>Gal4, btlenh-mRFPmoe x
UAS-"X"); H-J insets, clone positions and signaling patterns (also see Fig. S3H-K); D-J;
arrow and arrowhead, signal-recipient cells with and without nuclear dpERK,
respectively; G, average intensity plots comparing Bnl:GFP, TM, and C distribution
along the ASP D-P axis; K, violin plots comparing the percentage of signal-receiving
ASP cells with nuclear dpERK from overexpression (OE) and clonal expression; p <
0.01: C (n = 16) vs either Bnl:GFP (n = 17) or TM (n = 13) under OE; see Fig. S3K for
p values for clonal assays. (L-O) Levels of tracheal branch invasion (arrows) on larval
salivary glands expressing either Bnl:GFP, TM, or C (bnl-Gal4 X UAS-X); L-N,
brightfield, 10X magnification; O, Sholl analyses graph comparing frequency of terminal
branching. (P-S) Fraction of Bnl:GFP, TM, and C displayed on the basal surface (red,
GFPex) when expressed from salivary glands; S, graph comparing the fraction of
surface displayed signals (arrowhead); arrow, cell junctions; p < 0.05: Bnl:GFP (n = 17)
vs TM (n = 26), p < 0.01: C (29) vs either Bnl:GFP or TM. Scale bars, 30 m; 100 m
(L-N); 20 m (P-R').
Fig. 5. GPI-anchored Bnl acts as a CAM to drive bidirectional cytoneme contact
matchmaking. (A-I) Effect on the reciprocal polarity and numbers of ASP and source
cytonemes (arrows) when either Bnl:GFP, Bnl:GFPC, or Bnl:GFPC-TM were
expressed from the disc source (also see Fig.S4A-G); A,D,G, extended Z- projection,
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
25
both the ASP (red) and source (blue) are genetically marked; A, inset, ROI (box) in
green and blue channels; B,E,H, 3D-rendered views, only source membrane marked
(red); dashed lines, ASP; B', only red channel of B; D,E, dashed arrows, randomly
oriented short cytonemes devoid of signal localization; C,F,I, R-plots comparing
numbers, length, and polarity of ASP and source cytonemes (also see Fig. S4H,I). (J-J’)
3D projection of cytonemes from the ASP (btl>CherryCAAX) and the Bnl:GFPC-TM -
expressing wing disc bnl source (blue, bnl>CD4:IFP2); J', orthogonal views showing a
Bnl:GFPC-TM puncta (arrow) at an inter-cytoneme junction. All panels, live imaging;
dashed line, ASP outline. Scale bars, 20 m.
Fig. 6. Btl-Bnl-interactions mediate cytoneme contact assembly and plasticity. (A-
B) 3D-projected images of cytonemes from the Bnl:GFPC-TM-expressing wing disc
bnl source (blue, bnl>CD4:IFP2) in btl:cherryendo knock-in larvae; Co-clustering of
Btl:Cherryendo puncta (red) from the ASP and Bnl:GFPC-TM puncta (arrows) from the
source at the inter-cytoneme contact sites. (C,D) Violin plots showing the maximum
extension and lifetime (C) and average extension and retraction velocity (D) of ASP
(recipient) and source cytonemes while source cells expressed Bnl:GFPC-TM (also
see Table S2 for comparison and statistical significance). All panels, live imaging. (E)
Proposed model showing how GPI-anchored Bnl ensures tissue-specific cytoneme
contact matchmaking, contact-dependent exchange, and bidirectional signaling
feedback to self-sustain signaling sites. Scale bars, 20 m.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.23.432493doi: bioRxiv preprint
