1. Continuum of Gene-Expression Profiles Provides Spatial Division of Labor within a Differentiated Cell Type 
Miri Adler, Yael Korem Kohanim, Avichai Tendler, Avi Mayo, Uri Alon 
Cell Systems 
Volume 8, Issue 1, Pages e5 (January 2019)
DOI: /j.cels
Copyright © 2019 The Authors Terms and Conditions

2. Cell Systems 2019 8, 43-52.e5DOI: (10.1016/j.cels.2018.12.008)
Copyright © 2019 The Authors Terms and Conditions

3. Figure 1
Theory of Optimal Total Tissue Performance in the Presence of Spatial Gradients Yields a Continuum of Gene Expression inside a Polyhedron
(A–E) (A) The total tissue performance function is based on the sum of the contributions of individual cells. Without spatial gradient effect, cells either share the same gene-expression profile, performing all tasks in the case of concave (or linear) performance functions (B and C) or specialize in one of the tasks forming discrete cell types in the case of convex performance functions (D and E).
(F) Total tissue performance function considering spatial gradients of performance across the tissue, ϕi(xj).
(G) We consider three linear one-dimensional (1D) gradient effects.
(H) The optimal gene expression profile lies along a 1D curve in gene expression space bounded inside a triangle (color denotes gene expression, with red, blue, and green indicating specialization at tasks 1, 2, and 3). The simulated cells colored according to their gene expression show a continuous spatial zonation along the 1D tissue space (upper right inset).
(I) The resulting monotonic and non-monotonic zonation patterns are plotted along the spatial axis of the tissue, x.
(J and K) 2D performance gradients lead to a full 2D triangle in gene expression space similarly to the 1D case.
(L) The resulting zonated gene-expression patterns are plotted as a fucntion of the spatial axes of the tissue. Color denotes gene-expression level.
Cell Systems 2019 8, e5DOI: ( /j.cels )
Copyright © 2019 The Authors Terms and Conditions

4. Figure 2
Enterocytes Fall in a 1D Continuum in Gene-Expression Space inside a Triangle, Suggesting Trade-off between Three Tasks
(A) The intestinal villus has a 1D structure where enterocytes are formed at the bottom and shed off at the top.
(B) T-SNE representation of the enterocyte and progenitor populations colored according to the inferred villus zone (replotted from Moor et al. (2018)).
(C) Enterocyte gene expression plotted on the plane of the first two PCs shows a 1D curve that can be bounded by a triangle. The colors indicate the position in the villus from the bottom (blue) to the tip (red). The expression of three enriched genes is shown (green color map denotes z-scored expression).
(D) The mean (black line) and STD (gray) of the expression of enriched genes in each of the archetypes are plotted as function of the height of the cells in the villus (height according to Moor et al. (2018)).
Cell Systems 2019 8, e5DOI: ( /j.cels )
Copyright © 2019 The Authors Terms and Conditions

5. Figure 3
Hepatocytes Fill a 3D Tetrahedron in Gene-Expression Space, Suggesting a Trade-Off between Four Complexes of Tasks
(A) T-SNE representation of the hepatocyte (circled with a black line) and non-hepatocyte populations colored according to Cyp2f2 expression (replotted from Halpern et al. (2017)).
(B) Hepatocyte single-cell gene-expression in the space of the first 3 PCs shows a continuum that can be enclosed by a 3D tetrahedron. At the vertices are ellipses that indicate STD of vertex position from bootstrapping. Projections of data on the tetrahedron faces are plotted in gray. The expression of four enriched genes is shown (green color map denotes z-scored expression).
(C) Shuffled data show a more spherical cloud compared to the real data. Ellipses at vertices indicate that the best fit tetrahedron varies widely with bootstrapping.
(D) Individual hepatocytes are colored based on their inferred position (r) along the CV/PN axis of the repeating liver unit, the liver lobule (position according to Halpern et al. (2017)).
(E) Enriched genes show zonation along the CV/PN axis (mean in black line and STD in gray). For the third archetype (third panel) the 30% most non-monotonic enriched genes are plotted and all enriched genes near the third archetype are considered in the inset.
Cell Systems 2019 8, e5DOI: ( /j.cels )
Copyright © 2019 The Authors Terms and Conditions

6. Figure 4 The Theory Predicts 3D Zonation Patterns in the Liver Lobule
(A) The coordinates of the liver lobule are radial distance from the CV (r), angle from the CV/PN axis (α), and height (z).
(B–D) (B) A demonstration of 3D performance gradients for the tasks that yield a 3D tetrahedron in gene-expression space (C), and a flower-shaped continuous zonation patterns in the tissue space, shown in two views in which the z axis is flipped (D) (color denotes gene expression, with red, blue, yellow and green indicating specialization at tasks 1, 2, 3, and 4).
(E) This example yields similar zonation patterns along the radial axis of the lobule as in the real data (compare to Figure 3E) when averaging over α and z.
Cell Systems 2019 8, e5DOI: ( /j.cels )
Copyright © 2019 The Authors Terms and Conditions