Figure 1.
Overview of oogenesis in Drosophila melanogaster.
(A) Schematic of an ovariole. Egg chambers, displayed at progressively later stages from anterior (left) to posterior (right), are formed in the germarium, and consist of three main cell types: nurse cells and the oocyte, both germ line, enveloped by a layer of somatic follicle cells (FC). After stage 9, the FCs have remodeled to form a columnar epithelium over the oocyte, and a squamous epithelium over the nurse cells. (B–B′) At early stages, ligand Gurken (Grk; in yellow) co-localizes with the oocyte nucleus to the posterior pole of the oocyte. It signals to EGFR in the overlying FC, activating the EGF pathway in a posterior-anterior gradient. (C–C′) After oocyte repolarization, Grk and the oocyte nucleus are located at the dorsal-anterior cortex of the oocyte. The EGF pathway is locally activated in overlying FC. (D–D′) Dpp ligand produced in the anterior FC establishes a steep anterior-posterior gradient of BMP signaling activity in the columnar FC. (E–E″) The appendage primordia are defined at stage 10 and consist, on either side of the midline, of two groups of cells, roof and floor. The eggshell deposited between the oocyte (Oo) and the follicle cells (FC) contains the operculum (OP), the micropyle (MP), and two dorsal appendages (DA); and is constituted by the vitelline membrane (VM), the inner chorionic layer (ICL), an endochorion (EnC) and an exochorion (ExC) [31].
Figure 2.
Phenomenological model: rules and result.
(A) Regulatory graph: the model links three distinct follicle cell fates, Operculum, Floor and Roof, to a combination of input components EGF, BMP, anterior, and Roof_adj. Oval nodes are Boolean (0 or 1) and the rectangular node (EGF) is associated to a multi-valued variable, which here takes values between 0 and 2 (absent, intermediate and high level). EGF directly influences the position of the three domains on the dorsal-ventral axis. BMP establishes the anterior border of the roof, while anterior defines the anterior competence region. Roof_adj is an input variable accounting for the differentiated state of neighboring cells. Green and red edges denote positive and negative effects, respectively. The edge in purple denotes a dual effect, i.e. activating or repressing, depending on the level of its source. (B) Logical functions driving the dynamics of the model: Each rule specifies under which conditions the variable evolves to value 1 (otherwise, the variable tends to 0). The condition of the presence of EGF is simply denoted as “EGF”, and “EGF:1” or “EGF:2” whenever distinction between levels is required. Logical connectors are: & for a conjunction (and), | for a disjunction (or) and ! for a negation (not). (C) Epithelial model: left, patterns for the inputs EGF (yellow), BMP (purple) and anterior (pink) as used during the simulation. Right, final cell fates are shown in green (operculum), blue (floor) and red (roof).
Figure 3.
(A) Regulatory graph: Grey nodes combine para- and/or autocrine signals. Dotted edges on Br and on Mid represent ad hoc interactions to account for protein maintenance. The grey box encompasses the module defining the anterior competence region (see text and Supplementary Text S1 for details). Other graphical conventions are as in Figure 2. (B) Logical functions driving model dynamics: (see Figure 2 for notation).
Figure 4.
Mechanistic epithelial model, simulation.
The simulation starts with a naive configuration (i.e. all cells are undifferentiated) upon which Grk, Dpp, and Mid input levels are applied. All components are updated synchronously except dpERK and the integration variables (not shown), which are always updated earlier (see text). From left to right are depicted the successive states of each component in the epithelium (gene expression patterns), before Grk extinction. The right panels show the components' states after Grk extinction. Color intensities are used for multi-valued components (Grk, Rho and EGRF, see Figure 3B). Red frames denote pattern changes.
Figure 5.
(A) Schematic dorsal view of the follicular epithelium, showing the 12 regions defined by combinations of input levels: Grk (4 levels, 0 to 3), Dpp (2 levels, 0 and 1), and Mid (2 levels, 0 and 1). (B) Final patterns before and after Grk extinction and description of the stable fates (F1 to F8). In regions R1 to R12, cells may adopt one of eight fates (F1 to F8) according to the values of the genetic network components (dpERK, Mirr, Pnt, Rho, Aos and Br; see also Supplementary Figure S1). The left diagram shows the final patterns obtained before and after Grk extinction (see Figure 4). Each row of the table describes the expression state of each component for a given fate. (C) Reachability analysis under the asynchronous update. In each region, we simulate the behavior of a naive cell inserted into the epithelium in its configuration just before the Grk extinction (the position of insertion determines the input values) and determine which stable state is reached (yellow arrows). The fate adopted by the cell follows the color code indicated in panel B. Upon Grk extinction, the simulation starts from a cell carrying the cell fate of the previous phase with now Grk levels set to 0, possibly leading to a new cell fate (grey arrows). In a few cases, more than one solution is attainable, such as in R5, R6 and R7. Full arrows represent trajectories towards fates matching the wild type situation, and dotted arrows indicate trajectories leading to alternative fates: e.g in R6 and R7, in addition to F5, the Br expressing pattern F8 is also reachable, unless a delay is assigned to Pnt (see text).
Figure 6.
Perturbations of the anterior domain.
(A) Simulation of wild type inputs, and of dpp mild and strong overexpression (using GAL4 drivers GR1 and CY2) [43], [63], [75], [83], and grk overexpression (using the Tub Gal4 driver). Asymmetry is maintained in GR1 and Tub driven inputs, taking into account the cumulative effect of the Gal4 driver and wild type expression. (B) Simulation results, pre- and post-Grk extinction, obtained from combinations of the input conditions described above. The boxes show the resulting patterns of Br (roof, red) and Rho (floor, blue). Compare to Shravage et al. 2007, Figure 3, panels Ea/Ec; Fa/Fc; Ga/Gc; Ha/Hc; Ia/Ic; Ja/Jc respectively [43]. (C) Perturbations of the BMP pathway and of Mid; LOF = loss-of-function; GOF = gain-of-function. BMP LOF was simulated by setting both the early_BMP and Dpp inputs to 0, to simulate a disruption of the BMP pathway before it could repress Mid; BMP LOF (late only) was simulated by setting only the Dpp input to 0, while keeping the early_BMP unchanged, to simulate a disruption of the BMP pathway after it could repress Mid. BMP GOF was simulated by setting both early_BMP and Dpp to 1 in the highlighted region. Mid LOF and GOF were simulated by setting Mid to 0 or 1, respectively, in the highlighted cells.
Figure 7.
(A) Loss-of-function and (B) Gain-of-function analyses. Wild type patterns are shown on the left for comparison. Each of the boxes shows the resulting patterns of Br and Rho or dpERK, under gain-of-function or loss-of-function situations for multiple elements (corresponding to genes) of the model. The outcomes are shown both before and after Grk extinction on the top and lower rows of each panel, respectively. In the case of Rho GOF (for which the level is constrained between levels 1 and 2), a cyclic attractor is reached before Grk extinction, which resolves into a stable state afterwards (see text).
Figure 8.
Mechanistic epithelium model, clonal analyses.
(A) Loss-of-function clones. (B) Gain-of-function clones. Before Grk extinction (top row) and after Grk extinction (lower row). Row organization and color codes as in Figure 7. In the Br GOF case, the oscillatory attractor obtained before Grk extinction is due to the synchronous simulation scheme (see text). Here, we show the most consistent pattern of the two stable states resulting from the Grk removal.