Cerebral Cavernous Malformations Develop Through Clonal Expansion of Mutant Endothelial Cells

Abstract

Rationale: Vascular malformations arise in vessels throughout the entire body. Causative genetic mutations have been identified for many of these diseases; however, little is known about the mutant cell lineage within these malformations.

Objective: We utilize an inducible mouse model of cerebral cavernous malformations (CCMs) coupled with a multicolor fluorescent reporter to visualize the contribution of mutant endothelial cells (ECs) to the malformation.

Methods and results: We combined a Ccm3 mouse model with the confetti fluorescent reporter to simultaneously delete Ccm3 and label the mutant EC with 1 of 4 possible colors. We acquired Z-series confocal images from serial brain sections and created 3-dimensional reconstructions of entire CCMs to visualize mutant ECs during CCM development. We observed a pronounced pattern of CCMs lined with mutant ECs labeled with a single confetti color (n=42). The close 3-dimensional distribution, as determined by the nearest neighbor analysis, of the clonally dominant ECs within the CCM was statistically different than the background confetti labeling of ECs in non-CCM control brain slices as well as a computer simulation ( P<0.001). Many of the small (<100 μm diameter) CCMs consisted, almost exclusively, of the clonally dominant mutant ECs labeled with the same confetti color, whereas the large (>100 μm diameter) CCMs contained both the clonally dominant mutant cells and wild-type ECs. We propose of model of CCM development in which an EC acquires a second somatic mutation, undergoes clonal expansion to initiate CCM formation, and then incorporates neighboring wild-type ECs to increase the size of the malformation.

Conclusions: This is the first study to visualize, with single-cell resolution, the clonal expansion of mutant ECs within CCMs. The incorporation of wild-type ECs into the growing malformation presents another series of cellular events whose elucidation would enhance our understanding of CCMs and may provide novel therapeutic opportunities.

Keywords: cell lineage; cerebral cavernous malformations; cerebrovascular disease; genetics; mutation; vascular malformation.

Figure 1.. CCMs are composed of clonally dominant ECs labeled with a single confetti color.

(A) Experimental mice contain three transgenes: endothelial-specific, tamoxifen-inducible cre recombinase with platelet derived growth factor b promoter (PDGFb-iCreERT2), loxP-flanked and null Ccm3 alleles (Ccm3^fl/KO), and R26R-Confetti reporter allele. (B) Cre recombination of a single R26R-Confetti allele leads to fluorescent labeling with one of four different constructs: nuclear GFP, cytoplasmic YFP, cytoplasmic RFP, and membrane-bound CFP. (C) Experimental design of a single, low dose (2μg) of tamoxifen injected into neonatal pups on postnatal days 3, 4, or 5. (D) Ccm3^fl/KO, PDGFb-iCreERT2, R26R-Confetti^fl/wt experimental mice develop a modest CCM burden in the cerebellum and cerebral hemispheres (n=3, scale bar: 2mm). (E) A representative 80-μm thick brain slice with discrete CCMs dispersed throughout the cerebellum as indicated by arrows (scale bar: 1 mm). (F-I) 3D reconstructions of Z-series confocal images acquired across multiple serial brain slices to visualize the entire volume of CCMs in Ccm3^fl/KO, PDGFb-iCreERT2, R26R-Confetti^fl/wt mice. Gray translucent surfaces have been added to the images to aid in visualizing the vascular lumens of each CCM. Representative CCMs for (F) YFP, (G) nGFP, (H) mCFP, and (I) RFP (scale bars: 100μm). (J) Number of ECs expressing each confetti color in serial brains slices of non-CCM control mice (PDGFb-iCreERT2, R26R-Confetti^fl/wt, n=1) and CCM mice (Ccm3^fl/KO, PDGFb-iCreERT2, R26R-Confetti^fl/wt, n=3). (K) Number of each confetti color expressed by ECs lining individual CCMs (n=42). CCMs #40, #13, #18, and #s 5–8 are the CCMs shown in panels F, G, H, and I respectively. (L) Percentage of the clonally dominant ECs, labeled with the most common confetti color within a CCM, when considering all confetti labeled ECs within the CCM (94+9 (mean + SD), n = 42).

Figure 2.. Nearest neighbor (NN) analysis as a quantitative comparison of the 3D distribution of confetti labeled ECs in CCMs, control brain slices, and computer simulations.

(A) 3D reconstruction of the entire volume of a CCM containing clonally dominant ECs labeled with nGFP. This CCM corresponds to CCM #23 in Figure 1K. (B) Visual representation of the 3D distances (white) between one EC and all of the other ECs within the CCM. The NN algorithm calculates and ranks these distances from closest to furthest. (C) Visual representation of the 3D distances (yellow) between a second EC and all of the other ECs within the CCM superimposed on the 3D distances of the first EC (white). (D-F) The same NN analysis is applied to non-CCM control brain slices: (D) 3D reconstruction of the control brain slices, (E) NN distances (white) between one nGFP EC and all of the other nGFP ECs, (F) a second set of NN distances (yellow) for a different nGFP EC and all of the nGFP ECs within the sample. (G) All of the NN distances, two of which are visually represented on the 3D reconstructions above, are ranked from closest to furthest and plotted as distance (μm) by NN number. The green data points represent the NN distances for all of the nGFP ECs within the CCM shown in panel A. The black data points represent the NN distances for all of the nGFP ECs within the control brain slices shown in panel D. The pink data points represent the average NN distances from 100 computer simulations in which the x,y,z coordinates of each member of the analysis were assigned by a random number generator. The simulation used the same volume and number of members (ECs) as the experimental CCM sample shown in panel A (error bars represent mean + SD). All scale bars are 100μm.

Figure 3.. Divergence of the nearest neighbor (NN) distances of CCMs and non-lesion controls as NN number increases.

Given the difference in recombination frequency among the four confetti colors, the NN comparison is only made between the ECs in CCMs and control brain slices labeled with the same confetti color. (A) NN distances for all CCMs containing clonally dominant ECs labeled with RFP (red) and the NN distances of ECs labeled with RFP in the control brain slices (black) (CCM n=13). (B) NN distances for YFP ECs within CCMs (yellow) and control brain slices (black) (CCM, n=18). (C) NN distances for nGFP ECs within CCMs (green) and control brain slices (black) (CCM, n=8). (D) NN distances for mCFP ECs within CCMs (blue) and control brain slices (black) (CCM, n=3). For each of the four confetti colors, the NN distance was greater for the non-lesion controls than for the CCMs for every NN number pair, beginning with the 1^st NN (Mann-Whitney test, p<0.001). Error bars represent mean + SD.

Figure 4.. Clonal expansion of mutant ECs and recruitment of neighboring ECs drives CCM development.

(A) Small (diameter <100μm), multicavernous CCM composed, nearly entirely, of clonally dominant ECs labeled with the same confetti color (RFP) in each cavern (scale bar: 100μm). This CCM corresponds to CCM #1 in Figure 1K. (B) Large (diameter >100μm), multicavernous CCM with a clonally dominant population of ECs labeled with YFP interspersed with unlabeled (appearing as blank spaces) ECs (scale bar: 100μm). This CCM corresponds to CCM #22 in Figure 1K. (C) The mutant EC density (number of clonally dominant ECs per 10⁵ μm³ CCM volume) decreases as CCMs grow larger (D) Working model of CCM development in which a single, second somatic mutation leads to clonal expansion and subsequent recruitment of wild type ECs into growing CCMs.