Supplementary Figure 1 Neuron class-specific arrangements of Khc::nod::lacZ label in dendrites. Staining with fluorescence antibodies to detect GFP (Green), β-galactosidase (magenta/white). (a, b) Class I 3 rd instar. (c) Coverage index (sum of label length/arbor length) at 3 rd instar for Khc::nod::lacZ label foci (<3 m) (n (neurons) = 7, 7, 6; class IV WT vs class IV UAS-ab t = 3.55 P < 0.01) (d-g) Class IV 3rd instar (magnification of panels in Fig. 1); (h-k) class I st17 (magnification of panels in Fig. 1). Data drawn as scatter plots with the mean ± standard deviation (SD). Scale bars: (a-b, d-g) 50 m: (h-k) 5 m. Statistical test used: (c) one-way ANOVA followed by Bonferroni s Multiple Comparison Test.
Supplementary Figure 2 Neuron class-specific arrangements of Lis1 label in dendrites. Coverage index (sum of label length/arbor length) for Lis1 label with a length of >9 m (a) (n (neurons) = 6, 4, 5; class I WT vs class IV WT t = 6.07 P < 0.001, class IV WT vs class IV UAS-ab t = 5.34 P < 0.001) and <3 m (b) (n (neurons) = 6, 4, 5; class I WT vs class IV WT t = 3.27 P < 0.05) at 3 rd instar, and embryo st17 (c) (n (neurons) = 12, 11; t(21) = 3.53 P = 0.0020). (d-f) Class I (yellow arrrowheads) and class IV (green arrowheads) neurons labelled with mcd8::ko and EGFP::Lis1. Data are shown as percentages or as scatter plots with the mean ± SD. Statistical tests used: (a, b) one-way ANOVA followed by Bonferroni s Multiple Comparison Test, (c) 2-tail t-test.
Supplementary Figure 3 Gene expression in ab mutant da neurons. (a-c) Gene expression in WT and ab mutant neurons (stage16-17, Rluv3>mCD8::GFP, n = 5, 6). Expression levels of (a) cnn are significantly reduced in ab mutants (t(9) = 3.49 P = 0.0069). However, those of (b) elav and (c) plp remain unchanged between genotypes. Expression levels are shown relative to gapdh (and also validated relative to rp49). Data collected at st16-17, and shown as scatter plots with the mean ± SD. Statistical tests used: t-test.
Supplementary Figure 4 Analysis of branch order and arbor complexity in cnn mutant da neurons. (a) Cartoon illustrating how branch orders were defined for panels (b) and (c); the ordering is based on the Strahler method, and combines consecutive branch segments with the same order into a single definitive branch (as indicated with *). In this cartoon branch orders are show in different colors and numbers (1-black; 2-red; 3-blue). (b, c) The total branch number for each branch order at st17 for (b) class I (n (neurons) = 10, 16) and (c) class IV neurons (n (here we show tracing data for the single largest tree in the dorsal half of a ddac arbor; one tree per neuron) = 17, 18). Branch number increased at each consecutive branch order in cnn mutants. Upon testing (with Bonferroni s multiple test correction), in class I this difference reached statistical significance at the highest two branch orders (1 st order t = 8.18 P < 0.0001, 2 nd order t = 3.21 P < 0.01); and in class IV at the first 1 st order (t = 5.05 P < 0.0001). (d, e) Sholl analysis reveals, for (d) class I (n (neurons) = 10, 20; F(1, 252) = 6.867 P = 0.0093) and (e) class IV (n (neurons: dorsal half of the arbor only) = 17, 17; F(1, 323) = 11.70 P = 0.0007), an increase in arbor complexity in medial and distal arbor regions in cnn mutants. Data collected at st17 shown as scatter and line plots with mean ± SD. Statistical tests used: (a, b) one-way ANOVA followed by Bonferroni s Multiple Comparison Test, (d, e) two-way ANOVA.
Supplementary Figure 5 Cnn controls VS1 dendrite branching. (a, b) Sequential z-images taken of (a) WT and (b) cnn VS1 neurons. Secondary branches (magenta asterisks) leave the primary branch (green asterisk). These form side trees that terminate in dendrite spine-like structures (orange arrowheads). In cnn mutants, the frequency of secondary branching is increased. Scale bar: 5 m, z step: 0.5 m. (c, d) In the complex VS1 interneuron, Cnn appears to repress branch formation in the early scaffolding phase, and later mechanisms limit the effect of this early over-branching on final spine-like structure distribution. The number of secondary trees was increased in the cnn mutant, yet the number of spine-like structures per secondary tree was concomitantly reduced. (c) Within the trees the density of spine-like structures in cnn mutants is indistinguishable from WT (n (neurons) = 14, 14, P = 0.1664). (d) However the side trees are smaller and the total number of spine-like structures per tree (counted in the largest tree in each neuron imaged) is reduced in the cnn mutants (n (neurons)=14, 14, t(27) = 5.77 P < 0.0001). Data collected in the adult. Data presented as scatter plots with the mean ± SD. Statistical test used: (c, d) 2-tail t-test.
Supplementary Figure 6 RhoA controls F-Actin distribution in dendrites. (a, b) The distribution of F-actin (measured by GMA fluorescence intensity (GMA: the 137 amino acid actin binding domain of Moesin tagged with GFP)) in developing class I dendrites, st17. Scale bar: 5 m. (c) WT neurons show a strong concentration of F-actin at dendrite tips. This concentration gradient is reduced upon expression of a dominant negative RhoN19 (n (branches)=153, 153). Data collected at st17, and represented as mean ± SEM.
Supplementary Figure 7 Additive Futsch- and Cnn-mediated branch repression. For dendrite tip number in class I neurons at 3rd instar, an additive effect between cnn and futsch was detected (a) F(1, 73) = 1.18 P = 0.2817, (b) F (1, 58) = 0.50 P = 0.4823 therefore showing no interaction (n (neurons)= 20, 20, 17, 20, 19, 21, 11, 12). Data collected at 3 rd instar and drawn as scatter plots with the mean ± SD. Statistical test used: two-way ANOVA.
Supplementary Figure 8 Rescue of extra branching in cnn mutants by co-expression of GFP::Cnn. Dendrite tip number in class I neurons at st17 (n (neurons) = 11, 19, 21, 16; interaction F(1, 63) = 12.08 P = 0.0009). Data collected at st17 and drawn as scatter plots with the mean ± SD. Statistical test used: two-way ANOVA.
Supplementary Figure 9 Distribution of GFP::Cnn foci on the Golgi surface. (a) A regular distribution of GFP::Cnn across the Golgi cis face. Scale bar: 1 m. (b,c,d) 1, 2, and 3 foci, respectively. Foci refers to a point where the uneven distribution of signal leads to an increased accumulation of the signal at one position relative to the surrounding region. Data collected at st17.
Supplementary Figure 10 No alteration in arbor-wide density of GFP::Cnn foci in plp mutants. The density of GFP::Cnn foci counted through the entire dendrite arbor (n (neurons) = 8, 7). Data collected at st17 and drawn as scatter plots with the mean ± SD.
Supplementary Figure 11 Differential relationship between Khc::nod::lacZ label and Golgi outposts in neurons with simple and complex arbors. (a) Khc::nod::lacZ signal is adjacent to Golgi outposts in class I neurons. Scale bar 10 m. (b) Khc::nod::lacZ signal does not show preferential localization to Golgi outposts along the dendrites of class IV neurons. (c) A bar chart comparing the relationship between Khc::nod::lacZ label and Golgi outposts in these two neuron classes (n= (outpost) 64, 63; Chi-square(2, N = 127) = 72.00 P < 0.0001). i magnification of the signals, ii path of the arbor. Abbreviation: p - proximal. Data collected at 3 rd instar and represented as a percentage. Statistical test used: (c) Chi-square test.
Supplementary Figure 12 Class I Golgi outposts have a single-compartment structure. (a) A Sholl analysis showing the distribution of Golgi compartments in the dendrites of late stage 17 class I neurons,(n(neuron) ManII=8, GalT=6). Data collected at st17 represent mean ± SD. (F(1, 108) = 51.11 P < 0.0001). (b) A comparison between the outpost compositions in complex (class III) neurons reported by Zhou and colleagues and the findings of this study in class I. Statistical test used: (a) two-way ANOVA.
Supplementary Figure 13 A role for Golgi outposts in regulating the polarity of microtubules in termini. (a) The relative contributions of Golgi and non-golgi derived polymerizing microtubules to the dendrite termini. In the class I arbor, microtubules derived from Golgi outposts favor polymerization in the retrograde direction (n (comets)= 50). (b) A cartoon to illustrate the consequence of linking microtubule nucleation events to a Golgi outpost in a nascent branch of a class I neuron. Arrows represent polymerizing microtubules, purple circles represent outposts. In mutants an outpost that is not supporting microtubule nucleation is shown as a pale circle. Data collected at st17.
Supplementary Figure 14 A model to address the outcome of changing branching frequency on class I dendrite arbor complexity. The simulation entails a straightforward model of branching in class I da neurons. (a) A soma gives rise to a number of initial segments; this distribution was estimated by counting primary dendrites in WT experimental data at stage 17. Further branching events were estimated from analysis of tracings of wildtype stage 17 neurons. In the model, subsequently, the end points are increased in three growth phases. In the first growth phase, each end point gives rise to an average of 4.2 new end points (normal distribution mu=4.2, sigma=2). This is the phase of rapid expansion. In the second and third growth phase, the end points can sprout once more with a probability of 1/4 and 1/10, respectively. In the non-wt case, the initial number of stems is 15% higher, and subsequently all branching probabilities are increased by 15% as well. (b) We ran this model 1000 times for both WT case and non-wt case. (c) The average number of end points was 20 and 28 in WT and non-wt case, respectively (D=0.426, P << 0.05, two-tailed Kolmogorov-Smirnov test).
Supplementary Figure 15 A comparison between Anti-Ab ChIP-seq and Ab-Dam ID at the cnn locus. Anti-Ab ChIP-Seq peaks were called peaks using MACS. Optimal peak calling (purple bar) called 2,397 peaks. Conservative peak calling (magenta bar) called 515 peaks. Binding at the cnn locus was comparable between Anti-Ab ChIP-Seq and a recently published DamID analysis of Ab (green bar).