Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2019 Oct 18;366(6463):eaay6727.
doi: 10.1126/science.aay6727. Epub 2019 Oct 3.

Coordination Between Stochastic and Deterministic Specification in the Drosophila Visual System

Free PMC article

Coordination Between Stochastic and Deterministic Specification in the Drosophila Visual System

Maximilien Courgeon et al. Science. .
Free PMC article


Sensory systems use stochastic fate specification to increase their repertoire of neuronal types. How these stochastic decisions are coordinated with the development of their targets is unknown. In the Drosophila retina, two subtypes of ultraviolet-sensitive R7 photoreceptors are stochastically specified. In contrast, their targets in the brain are specified through a deterministic program. We identified subtypes of the main target of R7, the Dm8 neurons, each specific to the different subtypes of R7s. Dm8 subtypes are produced in excess by distinct neuronal progenitors, independently from R7. After matching with their cognate R7, supernumerary Dm8s are eliminated by apoptosis. Two interacting cell adhesion molecules, Dpr11 and DIPγ, are essential for the matching of one of the synaptic pairs. These mechanisms allow the qualitative and quantitative matching of R7 and Dm8 and thereby permit the stochastic choice made in R7 to propagate to the brain.

Conflict of interest statement

Competing interests: Authors declare no competing interests;


Fig. 1.
Fig. 1.. Identification of three Dm8 subtypes corresponding to the three R7 subtypes:
(A) Schematic representation of the three different subtypes of ommatidia. (B) Regulatory network controlling R7 and R8 fate specification. (C) Schematic of the Drosophila visual system with R7 axons and their postsynaptic target Dm8 neurons in the medulla. (D) Dpr11MI02231 gene-trap expression in retina photoreceptors (Elav, blue) at 25 hours After Puparium Formation (APF). Dpr11-GFP (green) is strongly expressed in yR7, labelled by Ss (red, outline in yellow circles) but absent from pR7 (grey circles). (E) DIPγMI03222 gene-trap drives expression of GFP (green) in the adult medulla (Neuropile labelled using NCad, blue). A subset of Dm8s (labelled by CD8::RFP, in red) expresses DIPγ (DIPγ-expressing Dm8s, arrowhead, DIPγ-negative Dm8s, arrow). (F-G) DIPγ-expressing Dm8s always contact a yR7 in their home column. (F) Dorsoventral view of DIPγ-expressing Dm8 sparsely labeled with myr::GFP (green) extending a single process to the M4 layer in its home column (arrowhead). (G) Proximodistal view of a DIPγ-expressing Dm8. The yellow circle represents the center of the Dm8 dendritic field where the home column is located. yDm8s have a yR7 axon terminal in their home column (31 out 31 clones). Photoreceptors are labeled with GMR-RFP (red) and yR7 with Rh4-lacZ (blue). (H-I) pDm8s don’t express DIPγ and always contact a pR7 in their home-column. (H) Dorsoventral view of DIPγ-negative Dm8 with its single process in its home column (arrowhead). (I) Proximodistal view of a pDm8. pDm8s have a pR7 in their home column (grey circle, 33 out of 33 clones). (J) Tukey boxplots representing the number of R7s contacted per yDm8s and pDm8s outside of their home column, and the percentage of these contacts being with yR7s. Edges of the box indicate the first and third quartiles and the line the median. Mean (m) is represented by a cross. Whiskers represent the highest and lowest data point within 1.5 IQR of the first or third quartile respectively. ns (non-significant), **p>0.005; Student’s t-test. (K-L) A second type of DIPγ-negative Dm8 only contacts draR7s. Unlike y and pDm8s, they do not have a well-defined home-column (arrowhead) and their lateral processes do not contact non-draR7s (arrow). (M) The R13E04-Gal4 line specifically labels draDm8s. Proximodistal view of the entire medulla with draDm8s labelled with myr::GFP (in green), photoreceptors with GMR-RFP (red) and yR7 axons by Rh4-lacZ (blue). Note that draDm8s projections are only located at the edges of the dorsal half the medulla where draR7s axons are. Scale bars: (D), (E’-I) and (K and L) 5 μm, (E and M) 20μm.
Fig. 2.
Fig. 2.. Dm8 subtypes are pre-specified and have distinct lineages:
(A) In adult, both p and yDm8s, labeled by myr::RFP (in red), and DIPγ-GFP only for yDm8s (in green), express Dac and Tj (in grey and blue respectively). (B) DIPγ-GFP expression in late L3 optic lobe. DIPγ is expressed in three clusters of neurons (arrowheads). HRP labels neurons membrane (in blue). (C) In L3 stage, only the smaller cluster labelled by DIPγ-GFP also express Tj and Dac (in blue and red). Optix lineage trace with nuclear β-Galactosidase (outlined in C and cyan in C’) revealed that this cluster is coming from the ventral part of the Optix domain (C and C3). Three other clusters of Tj+Dac+ neurons are found in the larval optic lobe: one in the most dorsal part of the Optix region (C1), one in the ventral half of the mOPC (left in C2) and one adjacent in the ventral Optix domain (right in C2). (D) The split-Gal4 line VGlut-VP16∩DIPγ-Gal4DBD specifically labels the DIPγ+Dac+ cell cluster in late L3 optic lobes (labeled with myr::GFP, in green). (E) At 25h APF, VGlut-VP16∩DIPγ-Gal4DBD is still specific to the same cluster of cells expressing Dac (in red) and Tj (in blue). (F) VGlut-VP16∩DIPγ-Gal4DBD driving myr::GFP shows that this cluster of cells are Dm8s based on morphology. Photoreceptors are labelled in red by Chaoptin, and the neuropile in blue by NCad. (G) pxb lineage trace in late L3 optic lobe labeled all mOPC derived neurons with nuclear β-Galactosidase (cyan). Note that none of the DIPγ-GFP neurons are labelled by β-Gal. (H) Same lineage tool in adult in combination with a R24F06-LexA driving myr::GFP in p/yDm8s (red). None of the p/yDm8s express β-Gal (arrowheads). (I) hh lineage trace in late L3 optic lobe labeled all neurons derived from the ventral half of the OPC with nuclear β-Galactosidase (cyan). (J) hh lineage tool in combination with DIPγ-GFP labels both p (arrow) and yDm8s (arrowhead). hh-Gal4 drives the expression of the Flip recombinase that will lead to the excision of a stop cassette within a LexAop-RFP reporter. Thus, only cells coming from the ventral hh+ region and expressing the p/yLexA driver will be labelled by RFP. (K) hh lineage trace in adult does not label draDm8s (lineage: β-Galactosidase in cyan, and draDm8s in green). (L) Schematic of the distinct lineages of the three Dm8 subtypes. Scale bars: (A-J) 20μm, (A’) 5μm.
Fig. 3.
Fig. 3.. Dm8 subtype specification is coordinated with R7 stochastic specification:
(A) Schematic representing the R7 subtypes in the different mutant conditions. In the spineless (ss) eye specific mutant the retina is only composed of pale and DRA R7s, in sevenless (sev) mutants R7 photoreceptors are not specified and thus absent. lGMR-ss and lGMR-hth are photoreceptors specific gain-of-function where all R7s are either all o›f the yellow or DRA type respectively. (B-F) yDm8s labeled with CD8::GFP in WT (B), ss (C), sev (D), lGMR-hth (E) and lGMR-ss (F). Photoreceptors are labeled with GMR-RFP (red) and yR7 with Rh4-lacZ (blue). (G-H) Quantification of the number of yDm8s and pDm8s per optic lobe in Dm8-LexA>LexAop-RFP;DIPγ-GFP animals. (G) The number of yDm8s is plotted using two different quantification: the number of DIPγ-GFP+Dac+ cells or as the number of RFP+GFP+. (H) Quantification of the number of pDm8s (RFP+GFP). Bars show the mean +/− SD. ****p<0.0001, one-way ANOVA, Tukey test. (I) Single cell clone of a yDm8 with two home-columns in ss gain-of-function (lGMR-ss). (J) Distribution of yDm8s with 2 home-columns (WT, n=31; lGMR-ss, n=22). p=0.0302, Fischer’s exact test. Scale bars: 20μm in B for (B-F), 5μm in B’ for (B’-F’) and (I) 5μm.
Fig. 4.
Fig. 4.. Apoptosis of excess Dm8s ensures the numerical matching of R7s and Dm8s:
(A-D) DIPγ-GFP expression in the optic lobe in WT (A,B) and ss mutant (C,D) at 20 hours (A and C) and 40 hours After Pupa Formation (APF) (B and D). yDm8s were labelled by segmenting the Dac staining from the GFP staining (A’-D’, in grey). Scale bar: 20 μm in A for (A-D). (E) Number of yDm8s per optic lobe (DIPγ-GFP+,Dac+ cells, n=4-6 optic lobes per genotype). Bars show the mean +/− SD. ****p<0.0001, one-way ANOVA, Tukey test. (F) Proximodistal view of yDm8s labelled with CD8::GFP in WT upon cell death inhibition by mis-expressing P35 in DIPγ expressing neurons. A single yDm8 mis-paired with a pR7 (circled in grey) showed by dense GFP staining at the level of the M6 layer (E’) and its home column at the level of the M5 layer (F” and F”’) (mis-paired yDm8s=0.7%, n=428). Scale bar: 5 μm.
Fig. 5.
Fig. 5.. yDm8 morphology and survival are affected in DIPγ and dpr11 mutants:
(A and B) dpr11-GFP expression in 25h APF retinas (green) in ss mutant (A) or ss gain-of-function (B). R7 cells are labelled by Prospero (red) and all photoreceptors by Elav (blue). Large ellipse indicates a single ommatidia and the smaller ellipses the R7 cells. Note that dpr11-GFP expression is not lost in a single outer photoreceptor in ss mutant whereas weak dpr11-GFP is also seen in a few outer photoreceptors in the ss gof. (C) dpr11-GFP expression in 25h APF medulla (green). The split-gal4 line DIPγ-Gal4DBD∩OK371-Vp16 drives CD8::RFP in yDm8s (red). In the M6 layer, dpr11-GFP is mainly seen in a subset of photoreceptors (blue) that corresponds to the yR7s (arrowheads). (D) DIPγ mutant yDm8s labelled with CD8::GFP (red) at 25h APF. (E-F) Dorsoventral view of yDm8s labelled with CD8::GFP in DIPγ (E) and dpr11 (F) mutants. Arrowheads indicate morphological defects in yDm8 home columns. (G-I) Sparsely labeled yDm8s in WT (G), DIPγ (H) and dpr11 whole mutant animals (I). Arrow in G’ indicates the Dm8 process in its home column, and arrowheads in H’ and I’ indicate the defective process in the home column. (J-K) Number of yDm8s per optic lobe (DIPγ-GFP+,Dac+ cells, n=4-7 optic lobes per genotype). Bars show the mean +/− SD. ****p<0.0001, one-way ANOVA, Tukey test. Scale bars: (A and B, C’ and D, G-I) 5μm, (C, E and F) 20μm.
Fig. 6.
Fig. 6.. DIPγ and Dpr11 regulate pairing of yR7 and yDm8:
(A-C) Proximodistal view yDm8s labelled with CD8::GFP in WT (A), DIPγ mutant (B) and dpr11 mutant (C), either at the level of the M6 layer (A-C and A’-C’) or at the level of the M5 layer (A”-C”). Some yR7 or pR7 columns were highlight by yellow and grey circles respectively. In WT, yDm8s occupy most yR7 columns but never occupy pR7 columns (A”). In DIPγ and dpr11 mutants many yR7 columns are devoid of yDm8s, and some yDm8s contact pR7s (grey circle in B” and C”). (D) Quantification of yDm8s mis-pairing with pR7s (total number of yDm8s counted: WT; n=1046, DIPγ; n=223, dpr11; n=478, DIPγ;DIPγ-Gal4>P35; n=251 and WT;DIPγ-Gal4>P35; n=428). (E) MARCM clone of a pDm8 overexpressing DIPγ. DIPγ overexpression is sufficient for mis-pairing of pDm8s with yR7s (n=12/12). (F) yDm8 Flip-out clone labelled with myr::GFP with dpr11 overexpressed in all photoreceptors using lGMR-dpr11. Some yDm8s are mis-paired with pR7s (grey circle; n=5/23). (G) yDm8 Flip-out clone expressing CD8::GFP DIPγ background. 100% of the yDm8 clones obtained were paired with yR7s (yellow circle; n=15/15). (H) DIPγ mutant MARCM clone of a yDm8 expressing myr::GFP and P35 in an otherwise heterozygous background. 50% of the yDm8 clones obtained were mis-paired with pR7s (grey circle; n=7/14). Scale bar: 5μm for all micrographs.

Similar articles

See all similar articles

Cited by 1 article

Publication types

MeSH terms


LinkOut - more resources