RESEARCH ARTICLE◥

NEURODEVELOPMENT

Glia relay differentiation cues tocoordinate neuronal developmentin DrosophilaVilaiwan M. Fernandes,*† Zhenqing Chen,†‡ Anthony M. Rossi,Jaqueline Zipfel,§ Claude Desplan

Neuronal birth and specification must be coordinated across the developing brain togenerate the neurons that constitute neural circuits.We used the Drosophila visual systemto investigate how development is coordinated to establish retinotopy, a feature of allvisual systems. Photoreceptors achieve retinotopy by inducing their target field in the opticlobe, the lamina neurons, with a secreted differentiation cue, epidermal growth factor(EGF). We find that communication between photoreceptors and lamina cells requires asignaling relay through glia. In response to photoreceptor-EGF, glia produce insulin-likepeptides, which induce lamina neuronal differentiation. Our study identifies a role for glia incoordinating neuronal development across distinct brain regions, thus reconciling thetiming of column assembly with that of delayed differentiation, as well as thespatiotemporal pattern of lamina neuron differentiation.

Akey challenge during neural developmentis to coordinate the birth and specificationof diverse neuronal and glial cell typesacross different brain regions. To probe thisprocess, we focused on the visual system

ofDrosophila. Like vertebrate visual systems, thefly visual system is organized retinotopically intorepeated modular circuits that process sensoryinput from the entire visual field (1). The laminais the first ganglion in the optic lobe to receiveinput from photoreceptors (1). For each of the800 unit eyes (ommatidia) in the retina, thereis a corresponding lamina unit (cartridge) inthe optic lobe, made up of five lamina neuronaltypes and multiple glial subtypes. Populatingthese circuits with the correct number of cellsand cell types, and organizing them spatially,requires that photoreceptor, lamina neuronal,and glial development be precisely coordinated.Photoreceptors develop progressively as awave

of differentiation sweeps across the developing eyeimaginal disc, from posterior to anterior (Fig. 1A)(2). Newborn photoreceptors promote wave prop-agation by expressing Hedgehog (Hh), and recruitadditional photoreceptor subtypes to developingommatidia by expressing the epidermal growthfactor (EGF), Spitz (Spi) (2). As photoreceptoraxons grow into the optic lobes, they are en-sheathed by a population of glia, called wrapping

glia. Wrapping glial morphogenesis and photo-receptor axon ensheathment occur in responseto fibroblast growth factor (FGF) from photo-receptors (3). Upon arrival in the optic lobes,photoreceptors redeploy Hh and Spi to inducetheir first target field, the lamina (4–7). Thus, thedifferentiation wave in the eye disc ultimatelydrives the development of photoreceptors, wrap-ping glia, and the photoreceptor target field (lam-ina neurons).In the optic lobes,Hh fromphotoreceptor axons

promotes terminal divisions of neuroepithelialcells into equipotent postmitotic lamina precursors,which express Dachshund (Dac) (Fig. 1, A and B)(4, 7). These precursors assemble into columnsof six to seven cells along photoreceptor axons,also in anHh-dependent manner (Fig. 1A) (4, 7–9).Hh signaling promotes EGF receptor (EGFR) ex-pression inprecursors and, according to the currentmodel, makes them competent to respond to Spi,which photoreceptor axons also deliver (Fig. 1A)(5, 6). EGF from photoreceptors drives precursordifferentiation into the five lamina neuronal types,L1 to L5, in each column [marked by embryoniclethal abnormal vision (Elav), a pan neuronalmarker] (Fig. 1, A and B) (5). Although photo-receptors concomitantly express Hh, which con-trols precursor cell divisions and column assembly,and EGF, which controls differentiation, precursorsdifferentiate only after column assembly is com-pleted (2, 4, 5). Lamina precursors in each columndifferentiate according to an invariant spatio-temporal pattern, despite an apparently homog-enous differentiation signal from photoreceptors(EGF). In each assembled column of seven laminaprecursors, the most proximal (bottom) and mostdistal (top) cells differentiate first into L5 and L2,

respectively; differentiation then proceeds ina distal-to-proximal (top-to-bottom) sequence,L3 forming next, followed by L1, then L4. Thetwo “excess” cells are later cleared by apoptosis(Fig. 1A) (5).We explored the possibility that other cell types,

such as glia, may be involved in coordinatinglamina neuronal differentiation with photo-receptors. We found that EGFR in lamina pre-cursors is dispensable for their differentiationinto neurons. Instead, photoreceptors signal towrapping glia with EGF and, in response, wrap-ping glia induce L1 to L4 neuronal differentiationby secreting insulin-like peptides. This intercellularsignaling relay couples neuronal differentiationin the lamina with the timing of wrapping glialmorphogenesis. We suggest that it accounts forthe spatiotemporal pattern of differentiation,which is linked to fate specification of laminaneurons.Moreover, because glial processes arrivein the lamina after photoreceptors, theymay relaythe differentiation signal to the lamina with a lagrelative to the photoreceptor-delivered signal forcolumn assembly. In this way, glia help reconcileboth the timing of column assembly with thatof delayed differentiation, as well as the spatio-temporal pattern of lamina neuron differentia-tion. Glia thus coordinate neuronal developmentacross different ganglia.

Glial morphogenesis instructs laminaneuron differentiation

To explore the coordination of glial morphogen-esis with lamina and photoreceptor development,wemarkedwrapping glia and their processes byusing a wrapping glia–specific driver to expressmembrane-targeted green fluorescent protein(GFP) (Fig. 1B and movie S1) (10). Wrapping gliaare basal to photoreceptor cell bodies in the eyedisc (3, 11–13). Their processes wrap photoreceptoraxons through the optic stalk and in the develop-ing lamina (Fig. 1B) (12). Wrapping glial ex-tension along photoreceptors into the optic lobeswas progressive, such that their processes invadedthe lamina further in older columns, progressingas did differentiation (Fig. 1B). Therefore, theleading edge of wrapping glial processes arrivingin the optic lobes correlated with the front ofneuronal differentiation in the lamina (Fig. 1B).Whenwedisruptedwrapping glialmorphogenesisand extension into the optic lobes by expressinga dominant negative form of the FGF receptorHeartless (HtlDN) (3) in wrapping glia, we ob-served that the triangular front of differentiationindicative of sequential L1 to L4 differentiationwas disrupted (Fig. 1C). Differentiating laminaneurons only occupied the distal (top)–mostpositions in columns. Presumptive L5 neuronswere still present but differentiated with a de-lay of about three columns (Fig. 1C). This sug-gests that wrapping glia are involved in laminadifferentiation.

EGFR is dispensable in laminaprecursors for differentiation

Because both wrapping glial morphogenesisand EGF from photoreceptors are required for

RESEARCH

Fernandes et al., Science 357, 886–891 (2017) 1 September 2017 1 of 6

Department of Biology, New York University, 100 WashingtonSquare East, New York, NY 10003, USA.*Corresponding author. Email: vilaiwan@nyu.edu†These authors contributed equally to this work. ‡Present address:Carl R. Woese Institute for Genomic Biology, Department of Cell andDevelopmental Biology, University of Illinois, Urbana, IL 61801, USA.§Present address: Institut für Neuro- und Verhaltensbiologie,Badestraße 9, 48149 Münster, Germany.

on March 28, 2020

http://science.sciencem

ag.org/D

ownloaded from

lamina neuron differentiation (5), we revisitedthe originalmodel to further characterize the roleof EGF from photoreceptors. Rhomboid (Rho)proteins cleave Spi, making it active for secretion(6, 14). Rho3 is specifically localized to photo-receptor axons and rho3 mutants lack photo-receptor axon-derived Spi, and lamina precursorsfail to differentiate, although they are recruitedinto columns (6). In these mutants, Rho1 main-tains normal Spi secretion from the cell bodysuch that photoreceptors are specified normallyand project appropriately to the lamina (6). Res-cue experiments have demonstrated that therho3mutant phenotype can be entirely attributedto loss of Spi secretion from photoreceptor axons(6). As expected, differentiating lamina neuronsexhibited dually phosphorylatedmitogen-activatedprotein kinase (dpMAPK), a readout for receptortyrosine kinase (RTK) activity, which was lost inrho3mutants. L1s to L4swere eliminated in rho3mutants, as previously reported; however, L5sdifferentiated but with a delay, indicating thatthey follow a distinct differentiation programfrom L1 to L4 and do not require EGF fromphotoreceptor axons (fig. S1, C to G). We there-after focused only on L1 to L4 differentiation,which is abolished in the absence of EGF fromphotoreceptor axons.Although it is unambiguous that EGF from

photoreceptors is required for lamina neuronaldifferentiation, the lack of cell type–specific toolsat the time the original model was formulatedprecluded testing whether EGFR and MAPK arerequired specifically in the lamina for L1 to L4 todifferentiate (5). To test this, we used a lamina-specific Gal4 (10) to drive a dominant negativeform of EGFR (EGFRDN) in lamina precursorsand in differentiating lamina neurons (fig. S2A).This did not prevent lamina neuron differenti-ation as Elav-positive cells were observed. How-ever, lamina morphology was disrupted, likelydue to apoptosis of lamina cells (fig. S2A). Pre-venting cell death by expressing the baculovi-rus caspase inhibitor P35 along with EGFRDN

restored lamina morphology, revealing that thepattern of differentiationwas unaffected by block-ing EGFR activity in the lamina (Fig. 1, E and H).Thus, although EGFR appears to be required inlamina cells for their survival, they do not re-quire it to differentiate, which implies that photo-receptors donot communicate directlywith laminaprecursors through EGF.Although EGFR is not required for differentiat-

ing lamina neurons, MAPK signaling was activein these cells (fig. S1A), likely downstream ofanother RTK (15). To investigate whether MAPKsignaling is required for lamina neuron differen-tiation, we blocked transcription downstream ofMAPK by expressing an activated form of thenegative regulator of the pathway, anterior open(AopACT), in lamina precursors (fig. S2, B and C).Differentiation was blocked, and lamina mor-phology was also disrupted (again, likely due toapoptosis) (fig. S2C). Blocking cell death by co-expressing P35 with AopACT restored laminamorphology. However, lamina neuron differen-tiation was still prevented (Fig. 1, F and H). To

Fernandes et al., Science 357, 886–891 (2017) 1 September 2017 2 of 6

GFPDacElavHRP

GFPElav

wr. glia>GFP

lamina furrow

LPC

neuro-epithelium

eye disc

Hh dependent column assembly

EGF dependent differentiation

L5

eyedisc

L5

L4L1L3L2

X

young old

wr. glia>HtlDN

wrappingglia

DacElavHRP

GFPDac

ommatidium

Photoreceptor axons

ElavHRP

Lamina neuron

Distal

Proximal

PosteriorAnterior

GFPGFPDacDacElavHRP

ElavHRP

ElavHRP

ElavHRP

ElavHRP

lamina>MAPKACT

L5

lamina>GFP lamina>EGFRDN, P35

lamina>AopACT, P35

20

40

60

% d

iffe

ren

tiat

ed 80

0Lamina >GFP >EGFRDN

+P35>AopACT

+P35>MAPKACT

Lamina column

100

(10) (10)(6)

(6)

GFPDacHRP

*

*

R1-R6

R7-R8

L1-L4

L1-L4

Fig. 1. Photoreceptors do not communicate directly with lamina precursors through EGF.(A) Schematic of lamina development in the optic lobes, which is coupled to photoreceptor developmentin the eye disc. Hh from photoreceptors drives lamina precursor (purple) birth and assembly intocolumns. Photoreceptor-EGF is required for precursor differentiation into neurons (yellow). Columnsconsist of 6 or 7 precursors, which differentiate in an invariant spatiotemporal pattern (yellow). (B, B′, andB′′) A horizontal view of an early pupal (10 to 15 hours APF) eye disc and optic lobe showingphotoreceptor axons marked by horseradish peroxidase (HRP) (cyan). In the optic lobe, laminaprecursors express Dac (magenta) and differentiated photoreceptors, and neurons express Elav (yellow).Lamina cell bodies (magenta) are organized into columns that associate with photoreceptor axons.Wrapping (wr.) glia, marked by membrane-targeted GFP (white) driven by a wrapping glia–specific Gal4,extended processes through the optic stalk and into the lamina, where they encapsulate lamina cells andphotoreceptors progressively [inset in (B′′); arrowheads mark location of photoreceptors between glialprocesses and lamina cells]. (C, C′, and C′′) Expressing HtlDN in wrapping glia disrupted glial processinfiltration into the lamina. Only cells immediately below glial processes differentiated [arrowhead in (C′′)].(D and D′) Lamina-specific Gal4 driving GFP showed normal lamina neuron differentiation. (E) Lamina-specific EGFRDN and P35 coexpression did not affect neuronal differentiation. (F) Lamina-specific AopACT

and P35 coexpression led to loss of differentiated neurons (dashed bracket). (G) Lamina-specificMAPKACT expression led to premature Elav expression in columns. (H) Quantification of (D) to (G) as apercentage of differentiated cells in the six youngest lamina columns. Asterisks indicate significance withMann-Whitney U test; P < 0.01; no. of optic lobes examined indicated in parentheses. Scale bar, 10 mm.

RESEARCH | RESEARCH ARTICLEon M

arch 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

determine whether MAPK activation could driveectopic neuronal differentiation, we expressedan activated form of MAPK (MAPKACT) in thelamina (Fig. 1G). Instead of a triangular front ofdifferentiation, indicative of sequential differ-entiation, most lamina columns differentiatedimmediately after formation, and many moredifferentiated cells were present (Fig. 1, G andH). Importantly, MAPK was sufficient to drivelamina differentiation even in the absence ofEGF from photoreceptors in a rho3mutant (sup-plementary text and fig. S2). These data showthat MAPK signaling in lamina precursor cellsis both necessary and sufficient for lamina neu-ronal differentiation.

Lamina differentiation requiresphotoreceptor-activated EGFR in glia

Because photoreceptors signal through EGF butlamina precursors do not respond to it, we hy-pothesized that photoreceptors signal to wrappingglia, which relay cues to lamina precursors. Weinvestigated whether glia respond to Spi fromphotoreceptor axons: Wrapping glial nuclei (lo-cated in eye discs) had reduced levels of dpMAPK

in rho3mutants relative to controls (Fig. 2, A to C)(remaining activity likely due to FGFR signaling),indicating that photoreceptor axon–derived Spiactivated the EGFR pathway in wrapping glia.To evaluate the function of active EGFR

signaling in wrapping glia, we used a wrappingglia–specific Gal4 line to express EGFRDN. Glialensheathment of photoreceptor axons was notaffected by this manipulation (movies S1 to S3).However, the L1 to L4 triangular front of neu-ronal differentiation was absent (Fig. 2, D, E,and J). L5 differentiation was unaffected (Figs. 2,D, E, and J), as L5-specific markers were ex-pressed in the proximal row of the developinglamina (fig. S2H) (see table S1 for description ofneuronal subtype-specificmarkers) (16). Thesedatashow that active EGFR signaling in wrapping gliais necessary for L1 to L4 but not L5 differentiation.Together our data suggest that photoreceptors

do not signal directly to lamina precursors. Rather,wrappingglia respond toEGF fromphotoreceptorsto induce L1 to L4 differentiation. We thereforeasked whether activating the EGFR pathway inwrapping glia alone could bypass the require-ment for EGF from photoreceptors and rescue L1

to L4 differentiation in the lamina.We expressedan activated form of EGFR (EGFRACT) in wrap-ping glia in a rho3 mutant background. In thisgenotype, photoreceptor axons could not secreteEGF, but EGFR signaling was activated only inwrapping glia. Lamina differentiationwas rescued,and all L1 to L4 cell types were recovered (Fig. 2,G, I, and J; fig. S4; and table S1). Similar resultswere obtained when we expressed activated Ras(RasV12) in wrapping glia in rho3 mutants (fig.S2, I and J). These results argue that EGF fromphotoreceptor axons activates EGFR inwrappingglia, which is both necessary and sufficient toinduce L1 to L4 differentiation.

Glial insulin-like peptides inducelamina differentiation

Because lamina precursors require active MAPKsignaling to differentiate into neurons (Fig. 1, Fto H), we reasoned that the differentiation signalfrom wrapping glia must act through an RTKupstream of MAPK. The Drosophila genomeencodes 20 RTKs, although only 10 lie upstreamof MAPK signaling (15). Of these, we focused onthe insulin receptor (InR) as, in other instances,

Fernandes et al., Science 357, 886–891 (2017) 1 September 2017 3 of 6

Fig. 2. L1 to L4 differentiationrequires photoreceptor-induced EGFR signaling inwrapping glia. (A to C) Eyediscs with wrapping glia markedby the panglial nuclear markerRepo (magenta) and dpMAPK(yellow) in (A) rho3–/+ and (B)rho3−/− animals, quantified in(C). P < 0.001; Mann-WhitneyU test; no. of discs indicated inparentheses. (D to G) Optic lobesstained for Elav (yellow), Dac(magenta), HRP (cyan), and GFP(white). (D) A control wr.glia>GFP lamina. (E) Whenwrapping glia express EGFRDN,only presumptive L5s differenti-ated (arrow head). (F) In arho3−/− animal, there was only alate differentiating presumptiveL5 (see also fig. S1, C to G).(G) When wrapping glia expressEGFRACT and GFP in a rho3−/−

background, the L1 to L4 front ofdifferentiation is restored(bracket). (H and I) Develop-mentally expressed subtype-specific markers used incombination to identify neuronalsubtypes (16): Sloppy paired 2(Slp2) alone marks L2 and L3;Slp2 and Seven up (Svp)together mark L1, Brain-specifichomeobox (Bsh) alone marksL4, and Slp2 and Bsh togethermark L5 (dashed line indicateslamina plexus). (H) In a control rho3–/+ brain and (I) when wrapping glia drive EGFRACT (and GFP; not shown) in a rho3−/− background, all cell typeswere recovered. (J) Quantification of (D) to (G) as a percentage of differentiated cells in the six youngest lamina columns. Asterisks indicate significancewith Mann-Whitney U test, P < 0.01; no. of optic lobes examined indicated in parentheses. Scale bar, 10 mm.

RESEARCH | RESEARCH ARTICLEon M

arch 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

glia can use insulin/insulin-like growth factorsignaling to communicate with neural progen-itors (17, 18). In Drosophila, seven insulin-likepeptides (Ilp1 to Ilp7) bind to and activate thesoleDrosophila InR (19). Chico is the only insulinreceptor substrate in Drosophila (20) and acts tostabilize binding of activated InR to PI3-Kinase(PI3K) and growth factor receptor–bound pro-tein 2 (Grb2), leading to activation of PI3K andMAPK signaling, respectively (fig. S3A) (21, 22).Thus, chico mutants have low levels of insulinsignaling. Differentiating L1 to L4s were missingin large regions of all chicomutant laminas (~20%of the lamina) (Fig. 3B). In other regions of thesame mutant brains, no or partial loss of differ-entiated lamina neurons was evident. A viablehypomorphic allelic combination for InR showedsimilar differentiation defects (fig. S3B) (21, 22).To determine whether InR and Chico were signal-ing through PI3K or through MAPK, we usedChico constructs that rescue both or only one orthe other of the downstream signaling pathways(21): InR requires signaling through MAPK butnot PI3K for lamina neuronal differentiation(supplementary text and fig. S3, C to E).To activate InR in lamina precursors and

induce L1 to L4 neuronal differentiation, wrappingglia must secrete Ilps in response to EGF fromphotoreceptors. Although the central brain insulin-producing cells secrete several Ilps that act sys-temically, Ilps can also be developmentally andregionally expressed (17, 18, 23). Ilp2, Ilp3, andIlp5 are only expressed in insulin-producing cells

in the central brain complex (23, 24). We wereunable to test for wrapping glial expression ofIlp1 and Ilp4 due to a lack of reporters. However,reporter constructs for Ilp6 and Ilp7 both droveGFP expression in wrapping glia (Fig. 3, C and D),and Ilp6-Gal4 expression in wrapping glia wasdramatically decreased in rho3 mutants (fig. S3,F and G). Neither Ilp6 nor Ilp7 single mutantsshowed defects in lamina neuronal differentia-tion (fig. S3, H and I); however, Ilps are known toact redundantly, and removal of some Ilps canlead to compensatory regulation by others (24).Therefore, to disrupt Ilp function, we ectopicallyexpressed a secreted antagonist of Ilps, imaginalmorphogenesis protein L2 (Imp-L2) (25–27), inlarge actin-flip-out clones (fig. S3J). Consistentwith the chico and InRmutant data, blocking Ilpactivity by ImpL2misexpression led to an almostcomplete loss of L1 to L4 neuronal differentia-tion (fig. S3J). Thus, secreted Ilps are required forlamina differentiation.Because Ilp6 in wrapping glia is lost in the

absence of EGF from photoreceptors, we askedwhether restoring Ilp6 in wrapping glia was suf-ficient to induce L1 to L4 differentiation. Ex-pressing Ilp6 in wrapping glia in a rho3mutantbackground rescued the L1 to L4 triangular frontof differentiation (Fig. 3, F and H). Moreover,all L1 to L4 subtypes were recovered (figs. S3Kand S4 and table S1). dpMAPK expression in thelamina was also restored (fig. S3, L and M), fur-ther confirming that Ilp6 is sufficient to activatethe MAPK branch of insulin signaling during

lamina neuronal differentiation. We also testedwhether ectopically activating InR (with InRACT)in the lamina could bypass all exogenous cues torescue lamina neuronal differentiation in a rho3mutant background (Fig. 3, G and H). Althoughseveral rows of Elav-positive cells were recovered,they were disorganized (Fig. 3, G and H). Thesecells included L1s, L2/3s, and L5s but no L4s(figs. S3N and S4 and table S1). Altogether, ourdata show that wrapping glia receive EGF fromphotoreceptors and respond to produce insulin-like peptides that induce differentiation of laminaprecursors by activating MAPK.

The signaling relay may serve todelay differentiation

Because photoreceptors could signal directly tolamina precursors (4, 7), but instead act throughglia, we sought to understand the advantages ofthis relay mechanism. Glial processes arrive inthe optic lobes after photoreceptor axons (Fig. 1B).Thus, the relay may delay the differentiation cueto ensure that column assembly is completedbefore differentiation initiates. To test this, wesupplied Ilp6 directly from photoreceptors inrho3 mutants to bypass glial signaling (Fig. 4).We expressed Ilp6 with two panphotoreceptordrivers that differed in the onset of their ex-pression: The first was expressed in early-bornphotoreceptors, and the second was delayed rel-ative to photoreceptor birth (Fig. 4, C and D). Wepredicted that column assembly (6 to 7 laminaprecursors/column) would not be completed

Fernandes et al., Science 357, 886–891 (2017) 1 September 2017 4 of 6

Fig. 3.Wrapping glial insulin-like peptides induce lamina neuronal dif-ferentiation. (A) Normal lamina neuronal differentiation in a control. (B) Achico−/− brain lacked L1 to L4 differentiation (dashed bracket). (C) Ilp6-Gal4and (D) Ilp7-Gal4 drove expression of GFP (membrane or cytoplasmic,respectively) in wrapping glia and their extensions into the optic stalk (yellowarrows). (E) A rho3−/− lamina. (F) A rho3−/− animal with wrapping glia expressing

Ilp6 showed L1 to L4 differentiation (bracket). (G) A rho3−/− animal with thelamina expressing InRACT showed neuronal differentiation (bracket). Elav(yellow), Dac (magenta), HRP (cyan) and GFP (white). (H) Quantification of (F)and (G) as a percentage of differentiated cells in the six youngest laminacolumns. Asterisks indicate significance with Mann-Whitney U test, P < 0.01;no. of optic lobes examined indicated in parentheses. Scale bar, 10 mm.

RESEARCH | RESEARCH ARTICLEon M

arch 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

reliably when photoreceptors delivered assemblyand differentiation cues simultaneously (earlydriver), because lamina precursors would differ-entiate too early. However, if the differentiationcue was delayed, the correct number of laminaprecursors would assemble into columns beforedifferentiating. Although early photoreceptor-delivered Ilp6 expression rescued lamina neu-ronal differentiation in a rho3 mutant, fewerlamina precursors incorporated into columnson average (4.4 lamina precursors per column± 0.99 SD;N = 4 optic lobes) (Fig. 4C). Moreover,Elav expression initiated in the youngest columnthat contained four or fewer lamina precursors,indicating that they were still being assembled(Fig. 4C). When Ilp6 was expressed with thedelayed-onset photoreceptor driver in rho3mu-tants, lamina neuronal differentiation had onlyinitiated in old columns at ~10 hours after pu-parium formation (APF) (Fig. 4D). However, col-umns contained 6 to 7 precursors each (±0.88SD; N = 4 optic lobes) (Fig. 4C), indicating thatcolumn assembly was less disrupted comparedwith early-onset Ilp6 expression. By ~15 hoursAPF, all photoreceptors expressed Ilp6, and neu-

ronal differentiation was widespread, suggestingthat differentiation could “catch up” (Fig. 4E).Nonetheless, the pattern was disrupted, becausethe number of neurons in each column did notreflect the age of the column (Fig. 5E). These datasuggest that the relay from photoreceptors towrapping glia to lamina precursorsmay functionto segregate column assembly from differentia-tion in time.

Discussion

In Drosophila, photoreceptors establish reti-notopy between the retina and their target field,the lamina, by inducing lamina units, each con-taining five neurons. This is a multistep processrequiring lamina precursor generation and as-sembly into naïve columns, followed by theirdifferentiation into L1 to L5 according to aninvariant spatiotemporal pattern. We showedthat L1 to L4 differentiation is the consequenceof an intercellular signaling relay from photo-receptors to wrapping glia and then to laminaprecursors (Fig. 5). Rather than instructing laminaneuronal differentiation directly, EGF secretedfrom photoreceptor axons activates the EGFR

pathway in glia (Fig. 5); in turn, glia induce laminaprecursors to differentiate into L1 to L4 throughlocal insulin andMAPKsignaling (Fig. 5). Althoughphotoreceptor axons require InR to target thelamina (28), targeting was unaffected in rho3mutants (where L1 to L4 do not differentiate dueto reduced glial Ilps). It is therefore unlikely thatwrapping glial Ilps also guide photoreceptor axontargeting.Intercellular signaling relays are used in various

contexts during development (17, 18, 29, 30). Inthe context of the lamina, the glial relay servesseveral purposes: (i) The delayed arrival of glialprocesses into the optic lobes relative to photo-receptor axons temporally segregates columnassembly from differentiation (Fig. 4). The relayfrom photoreceptors to glia to lamina precursorscould therefore be a mechanism to ensure thatcolumn assembly is completed before differenti-ation initiates, leading to reproducible numbersof precursors in each column (Fig. 4). (ii) Thespatiotemporal pattern of lamina neuronal dif-ferentiation is likely a consequence of beingcoupled to progressive glial morphogenesis. Glialwrapping is itself coordinated independently

Fernandes et al., Science 357, 886–891 (2017) 1 September 2017 5 of 6

Fig. 4. The signaling relay may serve to delay differentiation to ensureconsistent column assembly. (A to D) Early pupal (stages indicated)eye-optic lobe complexes stained for Elav (yellow), Dac (magenta), HRP(cyan), and [(C) and (D)] GFP (white). Cyan dashed line marks the youngestphotoreceptors. (A) Control. (B) rho3−/−. (C) An early-onset panphotoreceptorGal4 driving GFP and Ilp6 in a rho3−/− background. Differentiation waswidespread and initiated in the youngest column (arrowhead), which

contained about four lamina precursors. (D and E) A late-onset panphoto-receptor Gal4 driving GFP and Ilp6 in a rho3−/− background. (D) At ~10 hoursAPF, differentiation initiated only in old columns (arrowheads), but columnsassembled 6 or 7 lamina precursors/column. (E) At ~15 hours APF, GFPand Ilp6 were expressed in all photoreceptors. Differentiation was widespreadbut variable because some columns contained more differentiated neuronsthan their older neighbors (arrowhead). Scale bar, 10 mm.

RESEARCH | RESEARCH ARTICLEon M

arch 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

by FGF from photoreceptors (Fig. 5) (3). Thus,photoreceptors independently regulate the abilityof wrapping glia to induce differentiation in thelamina as well as the timing and pattern of thisinduction. All wrapping glia–driven rescues ofthe rho3 mutant generated all lamina neuronsubtypes (fig. S4 and table S1). However, this wasnot the case for lamina-driven rescues of rho3,which produced aberrant subtypes while some-times lacking others (fig. S4 and table S1). By

signaling through glia, photoreceptors may betranslating a homogeneous cue (EGF) into aspatiotemporally graded one, which appears es-sential for diversifying (L1 to L4) neuronal fates.(iii) Glial cells may be well suited for integratingsparse cues to interpret them into stronger ormore robust signals (31). Thus, by amplifyingcues from photoreceptors, glia may help reducenoise or variability of the signaling outcome.

REFERENCES AND NOTES

1. D. Hadjieconomou, K. Timofeev, I. Salecker, Curr. Opin.Neurobiol. 21, 76–84 (2011).

2. J.-Y. Roignant, J. E. Treisman, Int. J. Dev. Biol. 53, 795–804(2009).

3. S. R. Franzdóttir et al., Nature 460, 758–761 (2009).4. Z. Huang, S. Kunes, Cell 86, 411–422 (1996).5. Z. Huang, B. Z. Shilo, S. Kunes, Cell 95, 693–703 (1998).6. S. Yogev, E. D. Schejter, B. Z. Shilo, PLOS Biol. 8, e1000505

(2010).7. Z. Huang, S. Kunes, Development 125, 3753–3764 (1998).8. A. Sugie, D. Umetsu, T. Yasugi, K.-F. Fischbach, T. Tabata,

Development 137, 3303–3313 (2010).9. D. Umetsu, S. Murakami, M. Sato, T. Tabata, Development 133,

791–800 (2006).10. B. D. Pfeiffer et al., Proc. Natl. Acad. Sci. U.S.A. 105,

9715–9720 (2008).11. R. Rangarajan, Q. Gong, U. Gaul, Development 126, 3285–3292

(1999).12. T. N. Edwards, A. C. Nuschke, A. Nern, I. A. Meinertzhagen,

J. Comp. Neurol. 520, 2067–2085 (2012).13. M. Silies, Y. Yuva-Aydemir, S. R. Franzdóttir, C. Klämbt, Fly 4,

71–79 (2010).14. S. Urban, J. R. Lee, M. Freeman, EMBO J. 21, 4277–4286 (2002).15. R. Sopko, N. Perrimon, Cold Spring Harb. Perspect. Biol. 5,

a009050 (2013).16. M. Y. Pecot et al., Neuron 82, 320–333 (2014).17. J. M. Chell, A. H. Brand, Cell 143, 1161–1173 (2010).18. R. Sousa-Nunes, L. L. Yee, A. P. Gould, Nature 471, 508–512

(2011).

19. W. Brogiolo et al., Curr. Biol. 11, 213–221 (2001).20. R. Böhni et al., Cell 97, 865–875 (1999).21. C. Slack et al., Cell 162, 72–83 (2015).22. S. Oldham, E. Hafen, Trends Cell Biol. 13, 79–85 (2003).23. D. R. Nässel, J. Vanden Broeck, Cell. Mol. Life Sci. 73, 271–290

(2016).24. S. Grönke, D. F. Clarke, S. Broughton, T. D. Andrews,

L. Partridge, PLOS Genet. 6, e1000857 (2010).25. N. Alic, M. P. Hoddinott, G. Vinti, L. Partridge, Aging Cell 10,

137–147 (2011).26. N. Arquier et al., Cell Metab. 7, 333–338 (2008).27. B. Honegger et al., J. Biol. 7, 10 (2008).28. J. Song, L. Wu, Z. Chen, R. A. Kohanski, L. Pick, Science 300,

502–505 (2003).29. Z. Ma, T. Stork, D. E. Bergles, M. R. Freeman, Nature 539,

428–432 (2016).30. J. L. Leatherman, S. Dinardo, Nat. Cell Biol. 12, 806–811

(2010).31. D. K. Ma, G. L. Ming, H. Song, Curr. Opin. Neurobiol. 15,

514–520 (2005).

ACKNOWLEDGMENTS

We thank M. Amoyel, G. Odell, K. Menon, S. Kunes, and currentand former laboratory members for insightful commentsand suggestions. We thank B. Shilo, L. Partridge, E. Hafen,P. Léopold, Y. N. Jan, and E. Bach for reagents. This work wassupported by NIH grant EY13012 to C.D.; V.M.F. was supportedby Natural Sciences and Engineering Research Council ofCanada and Canadian Institutes of Health Research–Bantingpostdoctoral fellowships. The supplementary materials containadditional data.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/357/6354/886/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S4Tables S1 and S2Movies S1 to S3References

27 March 2017; accepted 27 June 201710.1126/science.aan3174

Fernandes et al., Science 357, 886–891 (2017) 1 September 2017 6 of 6

Lam

ina

Pre

curs

ors

PR

s Wrap

pin

g g

lia

HhEGF

Terminaldivisions

EGFR

InsulinsInR

MAPK

L1-L4Differentiation

Columnassembly

FGFRFGF

glial

morp

hogenes

is

Fig. 5. A signaling relay from photoreceptorsto glia to lamina precursors instructs laminadifferentiation.Model: Photoreceptors secreteEGF and FGF, which activate EGFR and FGFR,respectively, in wrapping glia. EGFR activationis required for glial expression of Ilps, whichactivate InR and MAPK in lamina precursorsleading to L1 to L4 differentiation. FGFR signalingregulates glia morphogenesis and processextension into the brain (3) and thereforeindirectly regulates the timing and patterningof L1 to L4 differentiation.

RESEARCH | RESEARCH ARTICLEon M

arch 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

DrosophilaGlia relay differentiation cues to coordinate neuronal development in Vilaiwan M. Fernandes, Zhenqing Chen, Anthony M. Rossi, Jaqueline Zipfel and Claude Desplan

DOI: 10.1126/science.aan3174 (6354), 886-891.357Science 

, this issue p. 886; see also p. 867Sciencepatterning of neurogenesis.

visual system. Thus, glia can play an instructive role in differentiation, helping to direct the spatiotemporalDrosophilain the −−the so-called lamina neurons−−photoreceptors to induce the differentiation of the photoreceptor target field

Perspective by Isaacman-Beck and Clandinin). The authors show that glial cells that ensheath axons relay cues from eye develops these maps (see the Drosophila now open a window into how the et al.within a circuit. Fernandes

During development, sensory systems must build topographic maps by connecting neurons at different levelsWiring up the eye

ARTICLE TOOLS http://science.sciencemag.org/content/357/6354/886

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2017/08/31/357.6354.886.DC1

CONTENTRELATED http://science.sciencemag.org/content/sci/357/6354/867.full

REFERENCES

http://science.sciencemag.org/content/357/6354/886#BIBLThis article cites 31 articles, 8 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2017, American Association for the Advancement of Science

on March 28, 2020

http://science.sciencem

ag.org/D

ownloaded from