A Pericyte Origin of Spinal Cord Scar Tissue

See allHide authors and affiliations

Science  08 Jul 2011:
Vol. 333, Issue 6039, pp. 238-242
DOI: 10.1126/science.1203165


There is limited regeneration of lost tissue after central nervous system injury, and the lesion is sealed with a scar. The role of the scar, which often is referred to as the glial scar because of its abundance of astrocytes, is complex and has been discussed for more than a century. Here we show that a specific pericyte subtype gives rise to scar-forming stromal cells, which outnumber astrocytes, in the injured spinal cord. Blocking the generation of progeny by this pericyte subtype results in failure to seal the injured tissue. The formation of connective tissue is common to many injuries and pathologies, and here we demonstrate a cellular origin of fibrosis.

Most studies on the scar tissue that forms at injuries in the central nervous system (CNS) have focused on astrocytes, and it is often referred to as the glial scar (15). There is also a connective tissue or stromal, nonglial, component of the scar (610), but it has received much less attention. The generation of connective tissue, with large numbers of fibroblasts depositing extracellular matrix (ECM) proteins, is a general feature of scarring and fibrosis in all organs and in diverse types of pathology (11). In spite of being a major clinical problem that has been extensively studied, the origin of scar-forming fibroblasts has been difficult to establish. Most studies have suggested that they may derive from circulating cells, proliferating resident fibroblasts, endothelial cells, or epithelial cells (1214). There are also data indicating that pericytes, perivascular cells enwrapping the endothelial cells of capillaries, may differentiate into collagen-producing cells in models of dermal scarring and in kidney fibrosis (1517).

We have explored the role of pericytes in scar formation after spinal cord injury. We found that Glast-CreER transgenic mice (18) enabled recombination of the R26R-yellow fluorescent protein (R26R-YFP) reporter allele (19) in a subset of pericytes lining blood vessels in the spinal cord parenchyma, which allowed us to stably and heritably label these cells (20) (Fig. 1 and figs. S1 to S5). The recombined cells had the typical ultrastructural features of pericytes (21), including being encased in the vascular basal lamina, which separates them from endothelial cells and astrocytes (Fig. 1, A to D). The recombined cells represent a distinct pericyte subpopulation that constitutes ~10% of all pericytes in the adult spinal cord [assessed by electron microscopy (EM)]. At positions where processes intersect, the Glast-CreER–expressing pericytes were invariably located abluminal to the other pericyte subtype (Fig. 1A and fig. S6). We refer to the pericyte subclass that is recombined in Glast-CreER mice as type A pericytes and the other subclass as type B pericytes. Most or all CNS pericytes express platelet-derived growth factor receptor (PDGFR) α and β; and CD13 (Fig. 1, E and F, and fig. S3) (2224), but the expression of some other markers is heterogeneous (25). Type B pericytes can be distinguished by the expression of desmin and/or alpha smooth muscle actin (Fig. 1G and fig. S4).

Fig. 1

Genetic labeling of type A pericytes. (A to C) Electron micrographs showing a recombined type A pericyte (arrow) on a blood vessel. (A) Pseudocolors indicate a recombined pericyte (green), nonrecombined type B pericytes (blue), endothelial cells (red), and astrocytes (cyan). The inset shows the light microscopic image of the section with the same recombined pericyte revealed by DAB reaction (arrow) before cutting ultrathin sections. Higher magnifications of the recombined pericyte from (A) show (B) that it is surrounded by basal lamina (bl, arrowheads) and (C) the plasma membrane (pm, arrowheads) of the astrocyte. (D) Two recombined pericytes encapsulating two endothelial tubes (stained for von Willebrand factor, vWF), surrounded by glial fibrillary acidic protein (GFAP)+ astrocyte processes. Type A pericytes express CD13 (E) and PDGFRβ (F) but not desmin (G). Cell nuclei are visualized with 4′,6′-diamidino-2-phenylindole (DAPI) in (D) to (G) and appear blue. Scale bars: 2 μm in (A), 0.5 μm in (C), and 10 μm in (G).

To address whether pericytes participate in scar formation after spinal cord injury, we genetically labeled type A pericytes before a dorsal funiculus incision or a dorsal hemisection and assayed the fate of recombined cells for up to 7 months after the lesion (fig. S1B). The injury was made following a 7-day clearing period without tamoxifen, which ensures that all recombination occurs before the insult, so that even if cells other than type A pericytes would start to express the Glast-CreER transgene in response to the injury, it would not result in recombination (26). The injury induced an increase in the number of recombined cells (Fig. 2, A to G). This reaction was restricted to the injured segment. After 9 days, the number of recombined cells had increased more than 25-fold in the injured segment. The number of pericyte-derived cells peaked at 2 weeks and then, as the scar condensed (9), decreased after 4 months to a level at which it remained for at least 7 months after the injury (Fig. 2G). This can be compared with the dynamics of the astrocyte population, which undergoes an approximate doubling in the first 2 weeks in the same injury paradigm and then decreases thereafter with similar kinetics (27). There are about 10 times as many astrocytes as type A pericytes in an uninjured spinal cord segment (480 ± 22 and 49 ± 3 cells in a 20-μm coronal section, respectively) (27), but 2 weeks after a lesion, there are about two times as many pericyte-derived cells as newly generated astrocytes in an injured spinal cord segment (Fig. 2H). The scar is compartmentalized, with pericyte-derived cells located in the center surrounded first by a layer of astrocytes originating from ependymal cells and then by a layer of astrocytes originating by self-duplication of resident astrocytes (fig. S7) (27).

Fig. 2

Pericytes form the core of the scar in the injured spinal cord. (A) Distribution of recombined type A pericytes (YFP) and astrocytes (GFAP) in an uninjured thoracic spinal cord segment. (B to F) Type A pericyte progeny occupy the core of the scar and are surrounded by astrocytes after a dorsal funiculus incision. (G) Number of type A pericyte–derived cells at the lesion site. (H) Net addition of type A pericyte–derived cells compared with astrocytes [data from (27)] 14 days and 4 months after injury. Cell nuclei are visualized with DAPI in (A) to (F). (A) to (E) show coronal sections and (F) a sagittal section. The quantifications show the average number of recombined cells per 20-μm coronal section. Error bars represent SD. Scale bar, 200 μm.

To gain insight into the dynamics of the pericyte injury response, we analyzed the early events of pericyte recruitment. The lesion center was nearly devoid of blood vessels on days 1 and 2 after the injury, but on days 3 to 5 after the injury, blood vessel sprouts, with an increased density of associated pericytes, appeared at the lesion (Fig. 3A). All recombined pericytes were tightly associated with endothelial cells outside the injury site, but many recombined cells had lost contact with blood vessels at the lesion (Fig. 3, B and C). Ultrastructural analysis showed an increase in the number of type A pericytes and a change in their morphology. Five days after injury, they had detached from the basal lamina encasement and developed thin processes, some of which penetrated through the basal lamina to invade the surrounding tissue (Fig. 3, D to F, and fig. S8). Type A, but not type B, pericytes deposited abundant ECM within their basal lamina encasement. Type A pericyte progeny that had invaded the tissue were also surrounded by ECM (Fig. 3, D, F, and I, and fig. S8 and S9). Five days after injury, the number of pericytes associated with blood vessels had significantly increased in number (0.016 ± 0.001 pericyte nuclear planes per 10-μm capillary surface by EM in the uninjured situation; n = 94 versus 0.151 ± 0.045 after injury; n = 82 (mean ± SEM); P < 0.001 Student’s t test). Type A pericytes increased three times more in number as compared with type B pericytes, and most important, only type A pericytes showed signs of leaving the blood vessel wall (Fig. 3, C to G, and fig. S6). The recombined cells that no longer had contact with blood vessels lost their expression of CD13 and PDGFRα, but remained positive for PDGFRβ and became positive for the fibroblast marker fibronectin and transiently expressed the myofibroblast marker smooth muscle actin (28) up to 9 days after the injury (Fig. 3, H and I). They were never positive for astrocyte or oligodendrocyte lineage markers (figs. S10 and S11). The vast majority of cells expressing markers of stromal cells at the lesion, such as fibronectin, PDGFRβ, and smooth muscle actin, were recombined, which established type A pericytes as the main source of the scar connective tissue. Thus, pericytes enter the lesion area with blood vessel sprouts, and type A pericytes give rise to cells that leave the blood vessel wall and form the stromal component of the scar tissue (fig. S12).

Fig. 3

Pericytes give rise to stromal cells and deposit ECM in the injured spinal cord. (A) Accumulation of type A pericyte–derived cells (YFP) and their detachment from the vascular wall (arrowheads) in the lesion area, 5 days after injury. (B) A blood vessel crossing the border (dashed line) between intact and injured (upper right) tissue, 5 days after injury. Type A pericytes (YFP) densely cover endothelial cells (vWF) within the intact tissue and their progeny detach from the blood vessel wall and invade the surrounding damaged tissue (arrowheads). (C) A blood vessel within the lesion with type B pericytes (visualized with antibodies against desmin) and type A pericytes (YFP), 5 days after injury, showing the expansion and detachment of the YFP-labeled cells. (D) Pseudocolored electron micrograph showing a blood vessel with three recombined type A pericytes (green) 5 days after injury. Type A pericytes detach from the surrounding basal lamina (bl), form thin processes, and deposit ECM. A type B pericyte (blue) remains tightly attached to the basal lamina (see also fig. S8). Its ultrastructure is retained, similar to that in uninjured tissues (Fig. 1A). An endothelial cell is colored red and astrocytes cyan. Boxed area shows the fibrous ECM deposited around type A pericytes. (E) Three-dimensional reconstruction of a series of electron micrographs showing a leading process (lp) of a recombined type A pericyte breaking through its basal lamina (bl) encasement (gray). (F) Electron micrograph of the lesion area 14 days after injury. Several type A pericyte–derived cells (green) have left the vascular wall and show abundant fibrous ECM (arrows) in their immediate surrounding. Boxed area shows abundant fibrous ECM. (G) Large numbers of type A pericyte–derived cells (YFP) distant to blood vessels (platelet endothelial cell adhesion molecule PECAM) 14 days after injury. Dashed line outlines the ependymal layer. (H) Type A pericyte–derived cells express smooth muscle actin (SMA) 5 days after injury. (I) The distribution of recombined cells overlaps with that of fibronectin 14 days after injury. Scale bars: 20 μm in (A) to (C) and (G) to (I), 2 μm in (D) and (E), and 5 μm in (F).

Analysis of bromodeoxyuridine (BrdU) incorporation and the mitotic marker Ki67 revealed abundant proliferation of type A pericytes during the first days after the spinal cord injury. By day 5 after the injury, 36.4 ± 3.9% of YFP+ cells were Ki67+, and 95.5 ± 1.9% had incorporated BrdU, which indicated that the large increase in the number of recombined cells is the result of proliferation of type A pericytes (Fig. 4, A and B). The absence of recombined cells in bone marrow or blood excluded a circulating source of recombined cells to the injured spinal cord (fig. S13). Furthermore, the Glast-CreER line did not recombine microglia or macrophages, and type A pericyte–derived cells were distinct from these cell types (fig. S14).

Fig. 4

Pericyte-derived cells are essential for regaining tissue integrity. (A and B) Many type A pericytes are Ki67+ 5 days after the injury and incorporate BrdU during the first 5 days after injury. (C) Schematic depiction of the strategy to block the generation of progeny by type A pericytes. (D and E) The generation of type A pericyte progeny is abrogated in Glast-Rasless mice 5 days after spinal cord injury. (F) Comparison of the scar core volume within the glial borders that is occupied by PDGFRβ+ stromal cells in vehicle- and tamoxifen-treated animals (Student’s two-tailed t test). (G) The percentage of the scar core volume occupied by PDGFRβ+ stromal cells correlates with the recombination efficacy in Glast-Rasless mice (Pearson’s correlation coefficient). (H) Correlation of the tissue defect volume to the recombination efficacy in Glast-Rasless mice. Individual animals are indicated with the same color in (G) and (H). (I and J) The injury site (indicated by dashed line) of dissected spinal cords from a vehicle-treated (I) and a tamoxifen-treated animal (J) 18 weeks after injury. Arrows point to the tissue defect in (J). (K to P) Sections of the spinal cords from (I) and (J) showing a scar with PDGFRβ+ stromal cells and fibronectin in the vehicle animal and the absence of a corresponding stroma in the tamoxifen animal, which has an open tissue defect lined by GFAP+ astrocytes. The animal in (J) is represented by a green dot in (G) and (H). Scale bars: 20 μm in (B), 50 μm in (D), 0.5 mm in (I), and 0.1 mm in (O).

To assess the role of type A pericyte–derived cells in the injured spinal cord, we devised a genetic strategy to inhibit their generation. We established mice that, in addition to carrying the Glast-CreER and R26R-YFP alleles, were homozygous for H-ras and N-ras null alleles and for floxed K-ras alleles, in which type A pericytes would lack all ras genes after induction of recombination (Fig. 4C, we refer to these mice as Glast-Rasless). ras genes are necessary for cell cycle progression and mitosis (29), and inducing recombination before spinal cord injury drastically reduced the appearance of recombined cells at the lesion site (Fig. 4, D and E). Deleting all ras genes did not result in any apparent alteration of the morphology or number of type A pericytes outside the lesion nor did it alter the number or distribution of blood vessels (fig. S15).

Adult Glast-Rasless mice in which recombination had been induced by five daily injections of tamoxifen, followed by a 7-day clearing period, were subjected to a dorsal spinal cord hemisection and analyzed 18 weeks after the injury. Mice of the identical genotype that received vehicle without tamoxifen served as controls and were indistinguishable from wild-type mice with regard to spinal cord scar formation. The scar of vehicle control animals was composed of a core of PDGFRβ-expressing cells encased in fibronectin and surrounded by astrocytes, similar to mice wild type for ras genes (Fig. 4). The tamoxifen group had significantly less PDGFRβ-positive stromal cells in the scar core compared with the vehicle group (P = 0.001, Student’s t test) (Fig. 4F). Tamoxifen-induced genetic recombination with CreER is seldom complete, and we asked if the variation in the generation of stromal cells within the tamoxifen group was related to variation in recombination efficacy. The size of the stromal component in individual animals did indeed negatively correlate to the recombination efficacy (P = 0.0015, r = –0.8857, Pearson’s correlation coefficient) (Fig. 4G).

It became obvious when analyzing the injured spinal cords that the generation of progeny by pericytes is important for sealing the injury, as 33% of the tamoxifen-treated Glast-Rasless animals had failed to close the lesion and had an open tissue defect at the site of the injury (compared with none of the vehicle controls) (Fig. 4). We found a correlation between the recombination efficacy and the failure to regain tissue integrity, with the animals showing the highest recombination efficacy having open tissue defects at the lesion site (Fig. 4H). The tamoxifen-treated Glast-Rasless mice with the highest recombination efficacy were largely devoid of a stromal cell and fibronectin scar core, which demonstrated that type A pericyte–derived cells are required to seal spinal cord lesions (Fig. 4, I to P).

We have identified pericytes as a source of scar-forming cells in the adult spinal cord. Previous studies have demonstrated altered pericyte morphology in response to traumatic brain injury and suggested that they may leave the vessel wall (30, 31), which in the light of our data indicates that scar formation by pericytes may be a general response to injuries in the CNS and potentially in other organs. It is well known that pericytes are heterogeneous on the basis of the expression of markers and morphology (25, 32). Here we demonstrate functional heterogeneity of pericyte populations, with scar formation restricted to a distinct subclass. Although the presence of stromal cells in CNS scar tissue has been long recognized (69), their role has been difficult to establish in the absence of knowledge on their origin. We conclude that the generation of progeny by pericytes is essential to regain tissue integrity after spinal cord injury.

Supporting Online Material

Materials and Methods

Figs. S1 to S15


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We thank A. Simon and the Frisén and Shupliakov groups for valuable discussions. This study was supported by the Swedish Research Council, the Swedish Cancer Society, Tobias Stiftelsen, Hjärnfonden, Knut och Alice Wallenbergs Stiftelse, the Swedish Agency for Innovation Systems, and the European Research Council (ERC-AG/250297-RAS AHEAD). D.D. was supported by the Foundation for Science and Technology from the Portuguese government (SFRH/BD/63164/2009).
View Abstract

Stay Connected to Science

Navigate This Article