A role of stochastic phenotype switching in generating mosaic endothelial cell heterogeneity

Previous studies have shown that biological noise may drive dynamic phenotypic mosaicism in isogenic unicellular organisms. However, there is no evidence for a similar mechanism operating in metazoans. Here we show that the endothelial-restricted gene, von Willebrand factor (VWF), is expressed in a mosaic pattern in the capillaries of many vascular beds and in the aorta. In capillaries, the mosaicism is dynamically regulated, with VWF switching between ON and OFF states during the lifetime of the animal. Clonal analysis of cultured endothelial cells reveals that dynamic mosaic heterogeneity is controlled by a low-barrier, noise-sensitive bistable switch that involves random transitions in the DNA methylation status of the VWF promoter. Finally, the hearts of VWF-null mice demonstrate an abnormal endothelial phenotype as well as cardiac dysfunction. Together, these findings suggest a novel stochastic phenotype switching strategy for adaptive homoeostasis in the adult vasculature.

FISH staining of vWF (red) and VE-cadherin (green) mRNA (blue: DAPI) in primary human ECs, including cardiac microvascular ECs (HMVEC) (left), dermal microvascular ECs (HMVEC-D) (middle), and aortic ECs (HAoEC) (right). n=10, with 3 replicates. Scale bar: 20μm. b. vWF mRNA distribution in populations of HMVEC, HMVEC-D and HAoEC. c. First, EC pairs at all distances have similarly heterogeneous vWF mRNA expression, as shown by difference in vWF mRNA between pairs of mouse heart ECs as a function of distance (vWF mRNA count was estimated as the sum of vWF and LacZ staining; EC distance was normalized to average nuclear diameter of ~50 ECs). Red line/error bars: averages/standard deviation within 1 nuclear-diameter windows. Second, low and high vWFexpressing cells reside in neighborhoods show comparable neighbor density, as indicated by the number of ECs within a radius of 2, 3, 4 and 5 nuclear diameters from an EC as a function of log vWF mRNA in the central cell. Solid lines: averages in windows of increasing size: (10 [1-10], 20 , 40 , 80 [70-150] and 160 ). Third and fourth, Low/high vWFexpressing cells have neighborhoods with comparable average vWF expression (third) and heterogeneity (fourth), as indicated by the average (third) and standard deviation (fourth) of vWF in ECs within 2, 3, 4 and 5 nuclear diameters from an EC, as a function of log vWF mRNA in the central cell. Solid lines: averages within the above windows.   replicates. * p<0.05; ** p<0.01; *** p<0.001; **** p<0.0001; n.s. p>0.05. e. Probability of observing n strands with patterns that do not match the two highest-frequency patterns due to experimental error (in silico), compared to in vitro results (black/red bars: probability of n strands with 50% mismatching CpGs). Figure 8. Absence of vWF is associated with abnormal cardiac endothelial phenotype. a-d. One-micron Giemsa-stained sections of the heart from wild-type (WT) vs. VWF knockout (KO) mice. Heart capillaries are cut longitudinally in all panels. Note the loss of capillary space in the KO sections, and the frothy appearance of cardiomyocytes in KO-2. Scale bar: 150μm. e-h. EM of the wild type heart capillary shows a thin attenuated endothelium. The lumen contains 2 red blood cells (RBC). The heart capillary from the knockout mouse shows electron-lucent endothelial cells with a well-defined lateral border (arrow). The lumen is filled with sloughed membrane bound vesicles. Underlying the endothelium are myocytes containing electron-lucent (injured) mitochondria (M). i-l. The aortic endothelium of wild type and knockout mice show abundant caveolae and vesicles. L, lumen. k-l. Liver sinusoids in wild type and knockout mice show normal fenestrated endothelium (arrows). m-n. The kidney glomerulus of the wild type and knockout mice show normal fenestrated endothelium (arrow), basal lamina (asterisk) and red blood cell (RBC)-filled lumens. Scale bar: (a-d) 500 nm; (e-j) 1 μm. n = 3, with 3 replicates. f-g. Sera were collected from vWF-KO or WT mice for chemical analysis of liver function (f) and renal function (g). ALK = alkaline phosphatase; ALT = alanine amino-transferase; AST = aspartate aminotransferase; BUN = blood urea nitrogen. Error bars: standard deviation. n = 3, with 3 replicates. Two-sided t-test was performed in all significance tests. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; n.s. p>0.05.

Supplementary
Supplementary Figure 10. Absence of vWF is associated with upregulation of Ang2 in the heart. a. Angiopoietin 2 (Ang2) protein levels in serum, heart and liver of WT and vWF KO mice (ELISA, n = 5). b. Ang2 (and VE-cadherin as a control) mRNA expression in whole heart, liver and kidney of WT and vWF KO mice (qPCR normalized to 18S RNA, represented as copy numbers per 10 6 18S; n = 3). c. Ang2 mRNA expression in cultured endothelial cells isolated from WT and vWF KO hearts (n = 3). d. Immunofluorescent co-staining for vWF and Ang2 protein in human heart microvascular endothelial cells (left, confocal; right, epifluorescence microscopy; filled arrows, vWF-negative cells with a pronounced cytoplasmic Ang2 blush; empty arrows, vWF-positive cells with punctate co-staining of vWF and Ang2, colocalized in WPBs). Sclae bar: 20μm.   hVWF-core-outer-R 5'-CCC CAA AAC CCT CAA AA-3' hVWF-core-inner-F 5'-AGG GTG GTT GGT GGA TGT-3' hVWF-core-inner-R 5'-AAA ATA TTA AAA TCA TCC CTA-3' The number of LacZ-negative ECs is increased by cell division, and decreased by cell loss and random OFF-to-ON transitions, according to: The time-dependent fraction of LacZ-negative ECs is thus given by: Consequently, this simple model predicts that the LacZ-negative population decreases exponentially with time. To verify this, we measured the fraction of cells in the ON state by counting LacZ/CD31 double positive ECs in heart and liver tissue sections from vWF-Cre-ROSA26R mice, sacrificed on days 1, 4, 7, 21 and 28 after birth (Fig. 3c-e). Linear regression on the log-normal plot in Figure 3e (time after birth approximated as ~12h and 3.5, 6.7, 20.5, 27.5 days) leads to: P Heart ON = 0.019 ± 0.003 /day/EC (5) P Liver ON = -0.0001 ± 0.0002 /day/EC = 0.
(In estimating the standard deviations of slope and intercept we assumed normally distributed residual errors).
These results allowed us to estimate the percentage of LacZ negative cells at birth: Next, we assumed that the in adult vWF LacZ where G(r, μ, σ) = exp[-(r -μ) 2 /(2σ)] / (2πσ) 0.5 represents the normal distribution. We used an

SUPPLEMENTARY NOTE 3. Ruling out alternative mechanisms of mosaic heterogeneity among isogenic neighbors that are not supported by our data.
In addition to biological noise toggling a bistable switch, additional mechanisms exist for generating phenotypic heterogeneity among isogenic cells. These include excitable circuits (monostable systems in which noise can trigger prolonged but transient "excursions" through state space) 2 , cell-autonomous oscillatory circuits 3,4 and pattern-forming regulatory mechanisms (e.g., lateral inhibition) 5 . Both excitable and oscillatory circuits can give rise to dynamic phenotypic mosaics, but neither can generate static, locked-in mosaicism. Our LacZ staining data in the aortae of one year-old VWF-Cre-ROSA26R mice, however, unequivocally demonstrate that VWF expression is a static mosaic in this vascular bed (Supplementary Fig. 3). To rule out pattern-generating mechanisms leading to differences between neighboring cells, we compared the neighborhood of vWF ON vs. OFF cells in vitro (Supplementary Figs. 6c-e), and found no relationship between the vWF-state of an EC and its neighborhood. Taken together, our results appear incompatible with excitable, oscillatory or pattern-forming regulatory mechanisms, but are consistent with a bistable switch with organ-specific barrier and noise-sensitivity. and 6b show a 5% standard error.
Moreover, in 3 of 6 HUVEC clones where the parent exon 1 region is almost entirely unmethylated, the heterogeneity of the upstream segment is nonetheless highly unlikely be an artifact of conversion and/or sequencing. -UU = 8 symmetrically un-methylated CpG dyads;  (Fig. 7c).

Supplementary Note 8. vWF expression is required for cardiac health.
One-micron Giemsastained sections of the heart from 14-week old (but not 2-week old) vWF-null mice demonstrated collapse of many capillaries in the right and left ventricles (Supplementary Fig. 8a-b show left ventricle) and patches of frothy-appearing cardiomyocytes (Supplementary Fig. 8c-d). Electron microscopy revealed profound focal ultrastructural abnormalities in the left and right ventricles of the vWF knockout mouse (Supplementary Figs. 8e-h). In these areas, capillary ECs were swollen and electron-lucent. However, the lateral borders were intact, the nuclei appeared normal and the cytoplasm contained many polyribosomes, arguing against a preparation artifact.
Blebbing was observed from the surface of ECs and the lumen was often packed with membrane bound vesicles. In neighboring cardiomyocytes, many of the mitochondria were swollen and electron-lucent. By contrast, the ultrastructure of the aorta (where there is normally a fixed mosaic of vWF expression), the kidney and liver (where vWF is not normally expressed in the glomerulus and hepatic sinusoids, respectively) was normal (Supplementary Figs. 8i-n).
However, the heart weight was similar in vWF-null compared with wild type mice ( Supplementary Fig. 9a shows body and heart weight). To determine whether these ultrastructural abnormalities were associated with functional changes, we carried out echocardiography (Supplementary Fig. 9b, 9c) and pressure-volume loop (Supplementary Fig.   9d) experiments. These studies demonstrated a significant increase in left ventricle systolic volume (LV Vol-sys) and a reduction in fractional shortening (FS), ejection fraction (EF), stroke volume (SV) and cardiac output (CO) in vWF-null mice compared with wild type littermates, suggesting a systolic cardiac dysfunction. To stress the heart, we subjected mice to transverse aortic constriction (Supplementary Fig. 9e). While perioperative mortality did not differ between genotypes, 30-day survival was significantly reduced in the vWF-null mice ( Supplementary Fig. 9e). In contrast to these cardiac readouts, functional assays for liver and kidney were unchanged in vWF knockout mice (Supplementary Figs. 9f-g). Together, these findings indicate that vWF expression is necessary for cardiac health. They raise the possibility, but do not prove, that dynamic mosaicism in heart capillaries is responsible for this effect.

Supplementary Note 9. Ang2 is upregulated in vWF-negative endothelial cells.
To determine whether loss of vWF may affect the storage and secretion of Ang2 in our model, we employed ELISA to measure Ang2 protein levels in our mice. Compared with wild type mice, vWF-null mice demonstrated significantly elevated levels of Ang2 levels in the heart, but not the liver or kidney (Supplementary Fig. 10a). Moreover, Ang2 levels were unchanged in the blood, suggesting that any pathological effect of Ang2 in the heart is related to local production in that organ. Interestingly, Ang2 mRNA expression was also selectively increased in the heart of vWFnull mice and in cultured cardiac microvascular ECs from these animals (Supplementary Figs.   10b, 10c). In cultured human cardiac MVEC, Ang2 co-localized with vWF in a subset of vWFpositive granules (consistent with WPB) (Supplementary Fig. 10d, left) 16 . In cells that lacked vWF, Ang2 staining appeared as a cytoplasmic "blush", suggesting that the vWF OFF state is associated with Ang2 release through a secretory pathway (Supplementary Fig. 10d, right).
Based on these data, we hypothesize that dynamic mosaicism of vWF expression in the capillaries of the heart is associated with randomly shifting hot spots of sustained Ang2 secretion, and that disruption of the mosaic leads to dysregulated Ang2 release and secondary microvascular damage.