Brat Promotes Stem Cell Differentiation via Control of a Bistable Switch that Restricts BMP Signaling

Summary Drosophila ovarian germline stem cells (GSCs) are maintained by Dpp signaling and the Pumilio (Pum) and Nanos (Nos) translational repressors. Upon division, Dpp signaling is extinguished, and Nos is downregulated in one daughter cell, causing it to switch to a differentiating cystoblast (CB). However, downstream effectors of Pum-Nos remain unknown, and how CBs lose their responsiveness to Dpp is unclear. Here, we identify Brain Tumor (Brat) as a potent differentiation factor and target of Pum-Nos regulation. Brat is excluded from GSCs by Pum-Nos but functions with Pum in CBs to translationally repress distinct targets, including the Mad and dMyc mRNAs. Regulation of both targets simultaneously lowers cellular responsiveness to Dpp signaling, forcing the cell to become refractory to the self-renewal signal. Mathematical modeling elucidates bistability of cell fate in the Brat-mediated system, revealing how autoregulation of GSC number can arise from Brat coupling extracellular Dpp regulation to intracellular interpretation.

, there is no evidence of a "cystocytoma" type expansion of cystocytes at the posterior germarium in any brat mutant allele examined. GSCs (arrowheads) are recognized by round fusomes shown by α-Spectrin staining (green) and a lack of Bam (red). Scale bars represent 10µm. (C) brat mutants exhibit anterior-posterior patterning defects, and therefore the allelic strength of each brat mutant can be evaluated by the number C A of segments present in brat germline clone embryos. Quantitation of segment number from cuticle preparations (n=50) and a representative embryo cuticle of each genotype is shown. brat 11 and brat 192 demonstrate the strongest defects, most commonly bearing only 2 segments. brat 192 has a greater proportion of weaker defects potentially due to increased paternal rescue. brat k06028 produces a much milder phenotype. The strength of the alleles is consistent with the nature of each mutation, as brat 11 and brat 192 encode truncations of the Brat protein before the vital C-terminal NHL domain (Arama et al., 2000;Betschinger et al., 2006), while brat k06028 is an insertion in the 5'UTR that potentially produces unstable mRNA, but still expresses low levels of wild type protein (Arama et al., 2000).  showing that an increase in Nos protein through removal of RNAi does not affect the pattern of GFP-mad3' reporter repression by Pum and Brat, but the level of repression is reduced. This is likely due to the Nos protein sequestering Pum, inhibiting the formation of a Pum-Brat repressive complex. In addition, this increased level of Pum-Nos complex can act to repress Brat, thus further inhibiting the GFP-mad3' reported regulation. Quantification shows mean relative expression of independent repeats (n=3), error bars are +SEM, *p<0.05. , yielding the observed correlations among production (φ), degradation (τ), and regulatory (Κ) parameters. As expected, a qualitative correlation is observed between the lumped regulatory parameters and cooperativity coefficients (ν). Bistability of the system was relatively insensitive to the remaining parameters. (B) To visualize the effects of parameter variation on the system bistability in the model with dMyc regulated endocytosis, families of bifurcation diagrams were plotted with one parameter varied. (a-e) Families of bifurcation curves plotted for Brat, Bam, Mad, Nos and dMyc for various production rates (0, 0.0001, 0.0002, 0.00025, 0.0003 nM*sec-1). All curves are plotted against the bound receptor (BR) parameter. (C) Further views of the progression of cells over time are plotted against the internalized bound receptor concentration (brin) and against the total bound receptor. The cell trajectories for a GSC, CB, 2-cell and 4-cell cyst are shown plotted against internalized bound receptor (a) and total bound receptor, the sum of both surface and internalized bound receptor (b). The total bound receptor quantity gives the most direct approximation of BMP signaling to the cell by representing both surface-bound and internalized signal sources. (D) As a second visualization of the progress of a 4xdMyc clone over time, the cell trajectories for the clone and the adjacent wild type GSC were plotted through time point t2 (since the positional definitions change afterward, with the displacement of the wild type cell from the niche). The two GSCs from Figure 6f follow different paths, resulting in differentiation of the wild type cell. The 4xdMyc clone (red trajectory) varies only slightly by t1, where the wild type cell's (green trajectory) pMad level decreases significantly. The trajectories diverge further with the decrease and subsequent return of Dpp production until, at t2, the clone is on the high-pMad branch while the wild type cell has fallen to the low-pMad branch. (E) Bifurcation diagram for the alternate dMyc model. The alternate model still supports bistability, though for a different region of the parameter space, illustrated by the bifurcation diagram.

Immunofluorescence
Adults were dissected after maturing on yeast and apple juice agar plates. Ovaries were fixed and stained using standard techniques (detailed methods available on request). The following primary antibodies were used: mouse anti-α-Spectrin (1:20, DSHB), rabbit anti-Vasa (Hay et al., 1990) (1:500), rabbit anti-Brat (Betschinger et al., 2006) (1:250) (1:500). The following secondary antibodies were used: goat anti-mouse and anti-rabbit FITC (1:100, Jackson Labs), goat anti-mouse Alexa-555, anti-mouse and anti-rabbit Alexa 594 (1:500 Invitrogen). Samples were counterstained with Dapi (1:300, Invitrogen), mounted using Prolong Anti-Fade Mounting Kit (Invitrogen) and visualised with a Leica SP5 Confocal Microscope, either as single focal plane or compiled Z-stacks. Cell death was detected via TUNEL assay, using in situ Cell Death Detection Kit, TMR Red (Roche) according to manufacturers' instructions, viability staining was performed using Image-iT DEAD green (Invitrogen) according to manufacturers' instructions. As a positive control for cell death hs-grim adult flies were heat shocked for 30 minutes at 37 o C, dissected after an hour and stained immediately. RNA in situ was performed using standard methods, using E.L.F. substrate (Invitrogen) according to manufacturers' instructions.

RNAi
dsRNA was generated by MegaScript T7 Kit (Ambion), using fragments of brat (BDGP: LD16270) and nos amplified from genomic DNA as templates, designed using the E-RNAi tool (http://www.dkfz.de/signaling/e-rnai3/) to prevent off target effects. 10ug of dsRNA was added to 2mls serum free cells per sample, followed by addition of serum after 30 mins.

Transgenic expression assessment
Level of brat transgene expression was measured in adult ovaries by dissecting 20 ovaries from newly eclosed flies of each genotype, removing embryos that were present and homogenizing the remaining germarium tissue in 100µl sample buffer. 30µl of each preparation were analysed by western blot, and to account for any loss of germline due to differentiation caused by ectopic brat expression, levels were equalized to the germline marker Vasa.

Cuticle preparations
Flies were allowed to lay for 3 hours and embryos aged for 18 hours at 25 o C, after which they were collected by washing in 0.1% Triton-X, bleached for 2 minutes, rinsed thoroughly in distilled water, and added to a 1:1 heptane/MeOH solution. Samples were shaken vigorously for 30 seconds, allowed to settle and the supernatant and floating embryos removed. Samples were rinsed 3 times in MeOH, 2 times in 1:1 MeOH/0.1% Triton-X and pipetted gently onto a slide. Excess liquid was removed with blotting paper, several drops of 3:1 Lactic acid/H 2 O added and a coverslip gently applied. Cuticles were cleared overnight at 65 o C, slides sealed with nail varnish and images taken with a Leica compound microscope.

Mathematical Modelling
To investigate local and spatial interactions mediated by Brat repression, we developed: (1)  Equations for the local model are given below (1-6). The local analysis is performed by continuation of the system across values of a bound receptor, which provides the input level for signaling, the concentration of which is represented as [ ] BR below. In the local system where BR is treated as a parameter, it is capitalized. When it is a variable solved for in the spatial model it is given as lowercase br.
The local model equations were expanded to develop a three dimensional spatiotemporal model of the germarium, for which the non-dimensionalized equations are given below. The BMP source is defined as the anterior boundary. This model includes dMycmediated endocytosis, such that the rate of endocytosis is k e =f(dMyc) with units sec -1 . Note that in the scaled equations, the repression behaviour depends on the balance between production, decay, and the saturation parameter K i . While we have rough estimates for the decay rates, we do not have an estimate for production or the saturation parameter.
However, the dimensionless quantity: (production/decay * 1/K i ) -1 is the scaled half-maximal saturation parameter, which is between 0.2 and 0.5 for the species in the model.  The system is solved via the Galerkin Finite Element method using the broadly available

Specific parameter value determination
The diffusion rate of free, extracellular Dpp was approximated by two measured rates for different molecules involved in embryo development: Dpp in the wing imaginal disc (Kicheva et al., 2007), and Bcd in the syncytial embryo (Gregor et al., 2007) have measured diffusion rates that range from 0.1-0.3 microns 2 /second. Intracellular diffusion rates were assumed to be greater than intracellular transport rates, which allows for easier analysis of the model and direct correspondence between the local cellular model analysis and the full 3D germarium model.
The diffusion of receptors in the membrane of a cell is approximated as 0.1 microns 2 /second, which is in the range of other transmembrane receptors (Lauffenburger and Linderman, 1993).
Kinetic parameters for Dpp binding to type I receptor fall in the middle of the expected range (Umulis et al., 2009) and lead to a dissociation constant of 3nM. Measured dissociation constants typically fall between 0.5 and 10nM for the binding between BMPs and type I BMP receptors.
Kinetic parameters for the intracellular network have not been carefully measured to date. We estimated the phosphorylation rate for surface localized BMP-bound receptors and internalized bound receptors as k S = 2e-4nM -1 *sec -1 , which is nearly the same as estimated values for Smad2 phosphorylation in HaCaT cell culture stimulated with TGF-β k phos~3 e-4nM -1 sec -1 (Schmierer et al., 2008). For our initial screen, we selected a receptor density of 100 receptors/square micron or a total of about 7000 receptors per cell, which is in the range of expected values (Lauffenburger and Linderman, 1993) for other systems. All protein degradation parameters were selected to be in the range of typical protein lifetimes and in accordance with the timescale for processes in the germarium. Production rates for intracellular molecules were selected so that the steady-state level of intracellular components is one nM, but the specific values for decay and production do not change the regulation inherent in the network as can be easily seen in the dimensionless forms of the equations. The regulation depends on the ratio of maximum concentration for a particular species to the half maximal concentration hill parameter ( ! x " x / K x ).

Alternate dMyc mechanism
To describe the alternate dMyc model, wherein protein synthesis rates for many intracellular processes are upregulated by dMyc, production terms for