a c activity Midbrain microdissection Cacna1c Cav1.2 Collected material Cacna1d Cav1.3 Cacna1g Cav3.1 SNc Hcn2 HCN2 SNr VTA Hcn4 HCN4 Scn2a1 Nav1.2 + Scn5a Nav1.5 TaqMan assays Scn8a Nav1.6 Kcna2 Kv1.2 Dissociated midbrain neurons Kcnb1 Kv2.1 Kcnd2 Kv4.2 Kcnd3 Kv4.3 Targeted reverse Kcnip3 KCHIP3 and preamplification Kcnj11 Kir6.2 Abcc8 SUR1 Abcc9 SUR2B Fluorescence imaging Kcnj5 GIRK4 Kcnj6 GIRK2 GFP Kcnn3 SK3 DA metabolism & signaling Non-GFP Microfluidic quantitative PCR Th TH Slc6a3 DAT Assays Samples Slc18a2 VMAT2 Pipette harvesting Drd2 D2R Glia-specific markers Gfap GFAP Aldh1l1 FDH -ion-binding Calb1 CB Pvalb PV Other neuronal markers b 40 Slc17a6 VGLUT2 30 Gad1 GAD67 20 Gad2 GAD65 10 Chat CHAT Cell count 0 Penk ENK 16 GFP Neuronal structure Non-GFP 14 Ncam2 NCAM2 WT 12 Map2 MAP2 10 Nefm NEF3

Th (TH) 8 Neuronal activation x

E Creb1 CREB

2 6 DA neurons Fos C-FOS 4 L o g (n=111) Bdnf BDNF 2 nDA neurons Housekeeping/ 0 (n=37) transcriptional factors 0 2 4 6 8 10 12 14 16 0 10 20 30 40 Hprt HGPRT Tbp TBP Log2Ex Slc6a3 (DAT) Cell count Tbx3 TBX3

Supplementary Figure 1. Experimental workflow and classification of DA and nDA neurons. a, schematic showing the workflow for the single-cell profiling. Left, neurons were manually collected after acute dissociation from microdissected midbrain slices containing the substantia nigra pars compacta and reticulata (SNc, SNr) and part of the ventral tegmental area (VTA) obtained from TH-GFP mice. Right, specific targeted reverse transcription and amplification was then performed in the same tube, and microfluidic qPCR was run by using 96.96 Dynamic arrays on a BiomarkTM HD System. b, DA and nDA were classified based on the expression levels of Th (TH) and Slc6a3 (DAT), nDA neurons corresponding to neurons expressing 0 Th and/or 0 Slc6a3. c, list of the selected for the single-cell study with the gene names (left) and corresponding names (right). Genes are classified based on the function of the corresponding (colored text). a b acute slices neuron GF P

acute dissociation neuron P non-GF

c *** 2.5 n=10 n=9 ** -30 30 *** 80 2.0 -35 20 1.5 -40 60 -45 10 p litude (mV) 1.0 e s h o l d (mV)

40 half-width (ms)

a m -50 acute slices t h r P

P 0.5 P A A

Sag amplitude (mV) 0

A -55 20 0.0 ** n=10 n=10 n=10 n=9 n=10 n=9 -60

* 2.5 n=13 n=8 * -30 30 *** 80 2.0 -35 20 1.5 -40 60 -45 1.0 p litude (mV) 10 e s h o l d (mV) 40 -50 half-width (ms) t h r a m 0.5 P P P A A

0 A Sag amplitude (mV) -55 20 0.0 acute dissociation n=12 n=5 n=13 n=8 n=13 n=8 -60 **

Supplementary figure 2. Electrophysiological properties of GFP and non-GFP midbrain neurons in acute dissociation and in acute slices. a, bright field and fluorescence images corresponding to recorded GFP (top pictures) and non-GFP neurons (bottom pictures). b, representative recordings obtained from GFP (green traces) and non-GFP midbrain neurons (dark red traces) in acute slices (top traces) and after acute dissociation (bottom traces). Please note the differences in sag amplitude and rebound kinetics (left traces) and the significantly broader action potential (right traces) in GFP neurons. c, bar and scatter plots summarizing the electrophysiological differences observed between GFP and non-GFP neurons in acute slices (top row) and after acute dissociation (bottom row). As expected for midbrain DA neurons, GFP neurons were characterized by significantly larger sag amplitude (left plot), smaller and broader action potential (middle plots), and more depolarized action potential threshold (right plot). Scale bars in b: 20mV, 500ms for left traces; 20mV, 2ms for right traces. Dotted lines indicate -60mV. Asterisks in c indicate statistical significance (Mann-Whitney). * p<0.05, ** p<0.01, *** p<0.001. 60 Scn2a1 (Nav1.2) 60 Kcnj6 (GIRK2) 60 Kcna2 (Kv1.2) 60 Slc17a6 (VGLUT2) 60 Ncam2 (NCAM2)

40 40 40 40 40

20 20 20 20 20

Cell count (%) 0 0 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16

60 Scn8a (Nav1.6) 60 Kcnj11 (Kir6.2) 60 Kcnb1 (Kv2.1) 60 Gad1 (GAD67) 60 Map2 (MAP2)

40 40 40 40 40

20 20 20 20 20

Cell count (%) 0 0 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16

60 Scn5a (Nav1.5) 60 Abcc8 (SUR1) 60 Kcnn3 (SK3) 60 Gad2 (GAD65) 60 Nefm (NEF3)

40 40 40 40 40

20 20 20 20 20

Cell count (%) 0 0 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16

60 Cacna1c (Cav1.2) 60 Abcc9 (SUR2B) 60 Th (TH) 60 Penk (PENK) 60 Creb1 (CREB)

40 40 40 40 40

20 20 20 20 20

Cell count (%) 0 0 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16

60 Cacna1d (Cav1.3) 60 Kcnd2 (Kv4.2) 60 Slc6a3 (DAT) 60 Calb1 (CB) 60 Fos (C-FOS)

40 40 40 40 40

20 20 20 20 20

Cell count (%) 0 0 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16

60 Cacna1g (Cav3.1) 60 Kcnd3_2 (Kv4.3) 60 Slc18a2 (VMAT2) 60 Pvalb (PV) 60 Bdnf (BDNF)

40 40 40 40 40

20 20 20 20 20

Cell count (%) 0 0 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16

60 Hcn2 (HCN2) 60 Kcnd3_1 (Kv4.3) 60 Drd2 (D2R) 60 Gfap (GFAP) 60 Hprt (HGPRT)

40 40 40 40 40

20 20 20 20 20

Cell count (%) 0 0 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Log2Ex Log2Ex 60 Hcn4 (HCN4) 60 Kcnip3 (KCHIP3) 60 Aldh1l1 (FDH)

40 40 40

20 20 20

Cell count (%) 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 DA neurons nDA neurons Log2Ex Log2Ex Log2Ex

Supplementary Figure 3. profiles in DA and nDA neurons. Graphs showing the frequency distribution of gene expression levels in DA (green) and nDA (dark red) neurons. The three genes with the lowest expression (Tbp, Kcnj5 and Tbx3) are not represented. The Y-axes correspond to normalized cell count (%) and the X-axes to relative gene expression levels in logarithmic (Log2Ex) scale. DA neurons

16 16 16 r=0.867, p<0.001, n=111 r=0.855, p<0.001, n=111 r=0.777, p<0.001, n=111 14 14 r= 0.783, p<0.001, n=111 14 r=-0.145, p=0.622, n=14 r=0.235, p=0.382, n=16 12 12 12 10 10 10 8 8 8 6 6 6 4 4 4 Kcnd3_2(Kv4.3) Slc6a3 (DAT) 2 r=0.695, p< 0.001, n=105 2 2 0 0 Kcnd3_2 (Kv4.3) 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Th (TH) Kcnn3 (SK3) Slc6a3 (DAT) Kcnj6 (GIRK2) Scn2a1 (Nav1.2) 16 16 16 r=0.852, p<0.001, n=111 r=0.706, p<0.001, n=106 r=0.786, p<0.001, n=105 r=0.721, p<0.001, n=105 14 14 n=1 n=2 14 r=0.542, p=0.055, n=13 r=-0.07, p=0.816, n=14 12 12 12 10 10 10 8 8 8 6 6 6 4 4 4 Kcnj6 (GIRK2) Kcnn3 (SK3) 2 2 Kcnj6a (GIRK2) 2 r=0.768, p<0.001, n=109 0 0 0 r=-0.521, p=0.1235, n=10 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Slc6a3 (DAT) Drd2 (D2R) Kcnj6 (GIRK2) Scn2a1 (Nav1.2) Slc18a2 (VMAT2) 16 16 16 r=0.785, p<0.001, n=110 r=0.745, p<0.001, n=111 r=0.783, p<0.001, n=106 r=0.711, p<0.001, n=106 14 14 14 r=0.686, p< 0.001, n=30 r=0.38, p=0.024, n=35 r=0.37, p=0.044, n=30 r=0.451, p=0.046, n=20 12 12 12 10 10 10 2R) 8 8 8 6 6 6

4 Drd2( D 4 4 Kcnd2 (Kv4.2)

Scn2a1 (Nav1.2) 2 2 2 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Map2 (MAP2) Ncam2 (NCAM2) Th (TH) Slc6a3 (DAT) Slc17a6 (VGLUT2)

16 16 16 r= 0.757, p<0.001, n=111 r=0.736, p<0.001, n=109 r=0.708, p<0.001, n=109 r=-0.848, p=0.033, n=6 14 14 14 n=3 n=1 n=3 12 12 12 10 10 10 8 8 8 6 6 6 4 4 4 r= 0.775, p<0.001, n=111 Kcnj6 (GIRK2)

Slc18a2 (VMAT2) 2 2 r=0.436, p=0.013, n=32 2 Cacna1d (Cav1.3) 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Scn2a1 (Nav1.2) Th (TH) Slc6a3 (DAT) Th (TH) Scn8a (Nav1.6)

nDA neurons

16 16 16 r=-0.054, p=0.653, n=71 r=0.107, p=0.362, n=74 r=0.285, p=0.045, n=50 r=0.306, p=0.003, n=92 r=0.405, p<0.001, n=92 14 14 14 r=-0.784, p=0.007, n=10 r=0.710, p=0.048, n=8 r=0.735, p=0.004, n=13 r=0.631, p<0.001, n=31 r=0.609, p<0.001, n=30 12 12 12 10 10 10 8 8 8 6 6 6 4 4 4 Hcn2 (HCN2) Kcna2 (Kv1.2)

Kcnd3_2 (Kv4.3) 2 2 2 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Calb1 (CB) Kcnj11 (Kir6.2) Abcc8 (SUR1) Nefm (NEF3) Nefm (NEF3) 16 16 16 r=0.366, p=0.008, n=51 r=0.331, p<0.001, n=101 r=-0.401, p=0.001, n=62 r=0.48, p<0.001,n=51 r= 0.492, p<0.001, n=62 14 14 14 r=0.628, p=0.016, n=14 r=0.622, p<0.001, n=26 r= 0.616, p<0.001, n=28 r=0.647, p=0.012, n=14 r=-0.634, p=0.002, n=23 12 12 12 10 10 10 8 8 8 6 6 6 4 4 4 Map2 (MAP2) 2 2

2 Scn2a1 (Nav1.2) Ncam2 (NCAM2) 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Abcc8 (SUR1) Kcnb1 (Kv2.1) Slc17a6 (VGLUT2) Abcc8 (SUR1) Slc17a6 (VGLUT2) 16 16 16 r=0.501, p<0.001, n=95 r=0.376, p=0.006, n=51 r=0.308, p=0.028, n=51 r=0.164, p=0.295, n=43 r=0.36, p=0.006, n=58 14 14 14 r=0.616, p=0.002, n=23 r=0.653, p=0.016, n=13 r=0.875, p=0.010, n=7 r=0.621, p=0.023, n=13 r=0.663, p=0.01, n=14 12 12 12 10 10 10 8 8 8 6 6 6 4 4 4 Hcn2 (HCN2) 2 Abcc8 (SUR1) 2 2 Ncam2 (NCAM2) 0 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Fos (C-FOS) Abcc8 (SUR1) Kcnd3_2 (Kv4.3) Scn8a (Nav1.6) Creb1 (CREB)

16 16 16 r=0.663, p<0.001, n=38 r=0.078, p=0.521, n=69 14 14 r=0.668, p=0.049, n=9 14 r=0.61, p=0.009, n=17 12 12 12 10 10 10 8 8 8 6 6 6 4 4 4 Nefm (NEF3) Gad2 (GAD2) r= 0.55, p<0.001, n=47 r=0.53, p<0.001, n=54

2 2 Kcnj6 (GIRK2) 2 r=-0.676, p=0.008, n=14 r=0.662, p=0.007, n=15 r=-0.084, p=0.517, n=62 0 0 0 r=0.645, p<0.001, n=29 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 Fos (C-FOS) Gad1 (GAD1) Creb1 (CREB) Hcn4 (HCN4) Slc17a6 (VGLUT2)

Supplementary Figure 4. Scatter plots of the 20 most significant correlations in expression levels in DA and nDA neurons. Axes indicate relative gene expression level in logarithmic (Log2Ex) scale. Green and dark red dots correspond to DA and nDA neurons, respectively. Plain and dotted lines indicate significant and non-significant Pearson correlations, respectively. r, p, and n values are displayed for each correlation test. Protein names corresponding to each gene are given in parentheses. a P(Th)

50/111 40/ 11 1 0/ 11 1 0/111 14

12

10 P(Calb1) 8

6 Ex Calb1 2 4 Log 2

0

4 6 8 10 12 14 16

Log2Ex Th P(Th,Calb1)

0/111 22/111 b Raw heatmap 111 DA neurons 15 Log 2 Ex 21 genes

0

Resampled heatmap 111 DA neurons 7 a.u. 21 genes

0

Supplementary Figure 5. Principles of probability estimation for 2 random variables. a, illustration of the basic procedure used in practice to estimate the probability densitiy for the two genes (n=2) Th and Calb1 in 111 DA neurons (m=111) using a graining of 8 (N1=N2=8). The data points corresponding to the 111 observations are represented as red dots, and the graining is depicted by the 64-box grid (N1*N2). The borders of the graining interval are obtained by considering the maximum and minimum measured values for each variable, and data are then sampled regularly within this interval with Ni values. Projections of the data points on lower dimensional variable subspaces (X and Y axes here) are obtained by marginalization, giving the marginal probability laws for the 2 variables X1 and X2 (PXi,Ni,m ; represented as histograms above the X-axis for Th and on the right of the Y-axis for Calb1). b, heatmaps representing the levels of expression of the 21 genes of interest on a Log2Ex scale (top, raw heatmap) and after resampling with a graining of 8 (bottom, N1=N2=...=N21=8). a b Four maxima of I3(X1; X2; X3; P)=log2(2)=1 bit Two minima of I3(X1; X2; X3; P)=-log2(2)=-1 bit Subsimplex of the probability 7-simplex Subsimplex of the probability 7-simplex P P P 001 001 0 0 001 0 P 0 P 1/2 P 1/4 P P 000 010 0 0 010 0 0 0 010 000 0 1/2 0 0 1/2 1/4 1/4

P111 P011 0 P011 P111 P011 0 1/2 0 0 0 1/2 1/4 0 1/4 0 0 P P P P P P 110 100 110 0 100 0 100 110 0 0 1/2 0 0 1/2 1/4 1/4 0 P P 0 P 101 0 101 0 0 101 0 1/2 1/4 Representation in 3-variable space Representation in 3-variable space

P011 =1/4 P111 =1/4 P010 P110 2 =1/4 =1/4 P X 101 P X 001 P P =1/4 3 =1/4 100 000 =1/4 =1/4 X 1 Representation in 2-variable subspaces Representation in 2-variable subspaces

X1,X2 X1,X3 X3,X2 X1,X2 X1,X3 X3,X2 c d Information landscapes 2-simplex of X1, X2, X3

Maxima of I3 Minima of I3 Maxima of I3 Minima of I3 Imax =log (N) 3 2 Count Count Count Count =1 = = = = Count C(3,1)=3 C(3,2)=3 C(3,3)=1 C(3,1)=3 H(X2) I H(X2) I 2 2 = (X (X Count =1 =1 C(3,2)=3 )=1 )=0 2 2 2 ;X 2 ;X 0 ;X ;X

(bits) 3 3 1 )=1 1 )=0 k I Count (X (X I 2 I 2 = I3(X1;X2;X3) I3(X1;X2;X3) Imax C(3,3)=1 =1 =-1 H(X1) H(X3) H(X1) H(X3) =-log2(N) =1 =1 =1 =1 =-1 1 I2(X1;X3)=1 I2(X1;X3)=0 1 2 3 1 2 3 k k

Supplementary Figure 6. Examples of mutual information positive maxima and negative minima. The simplest case of maxima and minima of I3 for 3 binary random variables X1,X2,X3 (n=3,N1=N2=N3=2) is represented. The corresponding probability simplex is 7-dimensional (N1*N2*N3-1), and is depicted at the top of panels a and b. The 8 vertices are labeled by the atomic probabilities noted P000,P001,...,P111. a, for this simple setting, there are 4 maxima of I3, I(X1;X2;X3)= 1 bit, corresponding to the 4 subsimplex of probabilities where all atomic probabilities are 0 except for the pairs {P000=1/2,P111=1/2}, {P110=1/2,P001=1/2}, {P101=1/2,P010=1/2}, {P011=1/2,P100=1/2} (top row). The corresponding configurations in the 3-variable space (middle row) correspond to the diagonals of the cube, such that the 3 projections on the 2-variable subspaces (bottom row) correspond to the diagonals for which I2 is maximal, i.e. I(X1;X2)=I(X1;X3)=I(X2;X3)= 1 bit, and that the projections on marginal single-variable subspaces are maximally uncertain (or informative; H(X1)=I(X1)=I(X2)=I(X3)= 1 bit). b, the same representations as in panel a are presented for the two minima of I3, I(X1;X2;X3)= -1 bit, corresponding to the configurations where all atomic probabilities are 0 except for the 4-tuples {P001=1/4,P010=1/4,P100=1/4,P111=1/4}, {P000=1/4,P011=1/4,P101=1/4,P110=1/4}. The corresponding configurations in the 3-variable space (middle row) consist of alternated volumes, such that the projections on the 2-variable subspaces (bottom row) are maximally uncertain, i.e. I(X1;X2)=I(X1;X3)=I(X2;X3)= 0 bit, and that the projections on marginal single-variable subspaces are also maximally uncertain (or informative; H(X1)=I(X1)=I(X2)=I(X3)= 1 bit). c, information landscapes associated with the maxima and minima presented in a and b. Count(n,k) gives the number of combinations for each degree k. d, equivalent simplicial representation with the k-faces labeled with their corresponding Ik values. a H(2) b =I(2) Randomly equidistributed variables X1 X2 X2 H(23) max(I) I(23) =log2(N) H(12) H(24)

I(24) I(12)H(13) I(13) H(1) X H(3) X 3 (bits) 1

=I(1) =I(3) k 0 I H(14) I(14) H(34) X4 X3 X4 I(34) 0 2 4 6 n/2 n-2 n H(4) k =I(4) Strictly redundant variables Entropy landscape n=4 Information landscape n=4 max(I) Simplicial information structure / Lattice Simplicial information structure / Lattice =log2(N) max(H) max(I) I(12) =n.log2(N) =log2(N) I(13) I(1) I(14) (bits) H(1234) I(2)

H(1) I(23) k 0 I(123) I H(2) H(123) I(3) I(24) I(124) H(3) H(124) I(4) I(34) H(12) I(134) (bits) H(4) H(134)

(bits)

k I(234) H(13) k 0 2 4 6 n/2 n-2 n

H(234) I H H(14) I(1234) k 0 H(23) H(24) 1 C(n,n/2) min(I) H(34) 0 =-log2(N) 0 1 2 3 4 0 1 2 3 4 Count k k c DA neurons nDA neurons 21 “relevant” genes (Figure 2) Count 6 105 104 4 103 (bits)

k 102 H 2 10 1 0

Count 2 105 4 1 10 103

(bits) 0 2

k 10 I 10 -1 1 -2 20 other “non-relevant” genes Count 2 105

104 1 103

(bits) 0 2 10 k I 10 -1 1 -2 0 5 10 15 20 0 5 10 15 20 k k

Supplementary Figure 7. Entropy and information landscapes. a, top, Venn diagram (left) and simplicial representation (3-simplex, right) of a system of 4 random variables (n=4). Bottom, entropy landscape (left) and information landscape (right) for the system of 4 random variables depicted in the top row, corresponding to the lattices or simplicial information structures. Entropy and information values for k=1 (single variables) and k=2 (joint-entropy and mutual information shared by pairs of variables) are color-coded with colors matching the schematic presented in the top row. While the joint-entropy always increases with k (with a varying slope), mutual information shows variable profiles, with decreases and increases depending on the variables and on k. b, theoretical examples of information landscapes for randomly equidistributed variables (i.e. of maximum entropy and independent, top) and strictly redundant equidistributed variables (such as maximally correlated variables, bottom). c, entropy landscape (top row) and information landscape (middle row) for the 21 “relevant” genes presented in Figure 2, and information landscape for the 20 other genes (bottom row) not presented in Figure 2. Please note that the information landscape for the “non-relevant” genes contains low values of mutual information (both positive and negative), and is dominated by independence (Ik=0), especially in DA neurons, in contrast with the rich information landscape obtained for the 21 “relevant” genes (middle row) analyzed in Figure 2 and 3. Original setting N = 8, m = 111

Calb1 Gad2 Slc17a6 Map2 Slc18a2 Nefm Slc6a3 Cacna1c Th Cacna1g Drd2

Scn2a1 Kcnj11

Hcn4 Abcc8 Hcn2 Kcnj6 Kcnn3 Kcna2 Kcnd3_2 Kcnb1

Variable sample size (m) Variable graining (N) N=8, m=28 N=4, m=111

Calb1 Gad2 Slc17a6 Slc18a2 Slc6a3 Slc6a3 Th Th Drd2

Scn2a1

Hcn4 Hcn4 Kcnj6 Kcnj6 Kcnn3 Kcnd3_2 Kcnd3_2

N=8, m=56 N=6, m=111

Slc6a3 Slc6a3 Th Drd2

Scn2a1

Kcnj6 Hcn2 Kcnj6 Kcnn3 Kcnd3_2 Kcnb1 Kcnd3_2

N=8, m=84 N=10, m=111

Slc6a3 Slc6a3 Th Th Drd2

Scn2a1

Kcnj6 Kcnj6 Kcnn3 Kcnn3 Kcnd3_2 Kcnd3_2

Supplementary Figure 8. Effect of changing sample size and graining on the identification of gene modules. The positive Ik paths of maximum length (see Figure 3b-c) were computed for a variable number of cells (m, left column) and a variable graining (N, right column). For clarity, only the two positive paths of maximum length are represented (first in red, second in black) for each parameter setting and the direction of each path is indicated by arrowheads. The two positive paths of maximum length for the original setting (N=8, m=111) are represented on the scaffold at the top of the figure for comparison. Smaller samples of cells (one random pick of 28, 56 and 84 cells) and larger (N=10) or smaller (N=4, N=6) graining than the original (N=8) were tested. Although slight differences in paths can be seen (especially for m=28), most of the parameter combinations identify gene modules that strongly overlap with the module identified using the original setting, which includes the DA metabolism and signaling genes (Slc6a3/DAT, Th/TH, Drd2/D2R) and ion channels (Kcnj6/GIRK2, Kcnd3_2/Kv4.3, Kcnn3/SK3). a b dendrite + Pitx3 ? Tyrosine DA Nurr1 + TH VMAT2

+ + + + + + + + Th Slc18a2 Slc6a3 Drd2 Kcnj6 Kcnd3_2 Kcnn3 (TH) (VMAT2) (DAT) (D2R) (GIRK2) (Kv4.3) (SK3) DAT GIRK2 D2R dopaminergic identity electrical phenotype

c d

Negative Ik-sharing Documented heterogeneity genes in midbrain DA neurons VTA SNc

CB+ in VTA and SNc subpopulations Calb1 / CB Damier et al. (1999) González-Hernández et al. (2000) D2R- subpopulations in VTA Drd2 / D2R Lammel et al. (2008) Margolis et al. (2008) GAD65+ cells in VTA and SNc subpopulations Gad2 / GAD65 González-Hernández et al. (2001) Slc6a3 Kcnj6 Abcc8 / SUR1 Higher mRNA expression in SNc than in VTA Kcnj11 / Kir6.2 Liss et al. (2005) Th Kcnn3 Higher Cav1.2 expression in lateral than in medial SNc. Differential expression of L-type Drd2 Kcnj11 Cacna1c / Cav1.2 channels in VTA and SNc Takada et al. (2001) Calb1 Gad2 Philippart et al. (2016) Larger T-type current density in CB- SNc cells low high Cacna1g / Cav3.1 Evans et al. (2017)

Supplementary Figure 9. Transcriptional control of dopaminergic metabolism and electrical phenotype. a, schematic representing the DA signaling pathway in dendrites of midbrain DA neurons. After synthesis (notably via the action of TH on tyrosine), DA is taken up in synaptic vesicles (via the action of VMAT2), released at dendro-dendritic synapses, acts on the D2 autoreceptor coupled to GIRK2, and is finally taken up by the plasmatic transporter DAT. b, schematic representing the transcriptional control of the genes involved in DA metabolism and signaling by the pair of transcription factors Nurr1 and Pitx3 (left, modified from reference 23), and the proposed common transcriptional control of ion channel genes involved in the definition of the electrical phenotype of midbrain DA neurons (right), based on the conditional mutual information analysis presented in Figure 3c. The possible involvement of an unidentified regulatory factor is also included (question mark). c, negative Ik-sharing genes have heterogeneous profiles of expression in midbrain DA neurons. The table shows 7 genes sharing strong negative Ik and the list of articles reporting their heterogeneous expression (measured using PCR, immunohistochemistry or electrophysiological recordings). d, schematic representing the superimposition of profiles of expression of positive Ik-sharing genes (co-varying) and negative Ik-sharing genes (heterogeneous, clustering) in midbrain DA neurons. While the levels of expression of positive Ik-sharing genes vary in a homogeneous manner, the levels of expression of negative Ik-sharing genes are extremely heterogeneous, producing a mosaic-like pattern. The precise patterns of expression of the negative Ik-sharing genes are based on the articles listed in panel c. The results on Kcnn3/SK3 gradient in the SNc obtained by Wolfart and colleagues (reviewed in reference 12) have been used to extrapolate the profile of expression of positive Ik-sharing genes.