Type H vessel/platelet‐derived growth factor receptor β+ perivascular cell disintegration is involved in vascular injury and bone loss in radiation‐induced bone damage

Abstract Collapse of the microvascular system is a prerequisite for radiation‐induced bone loss. Since type H vessels, a specific bone vessel subtype surrounded by platelet‐derived growth factor receptor β+ (PDGFRβ+) perivascular cells (PVCs), has been recently identified to couple angiogenesis and osteogenesis, we hypothesize that type H vessel injury initiates PDGFRβ+ PVC dysfunction, which contributes to the abnormal angiogenesis and osteogenesis after irradiation. In this study, we found that radiation led to the decrease of both type H endothelial cell (EC) and PDGFRβ+ PVC numbers. Remarkably, results from lineage tracing showed that PDGFRβ+ PVCs detached from microvessels and converted the lineage commitment from osteoblasts to adipocytes, leading to vascular injury and bone loss after irradiation. These phenotype transitions above were further verified to be associated with the decrease in hypoxia‐inducible factor‐1α (HIF‐1α)/PDGF‐BB/PDGFRβ signalling between type H ECs and PDGFRβ+ PVCs. Pharmacological blockade of HIF‐1α/PDGF‐BB/PDGFRβ signalling induced a phenotype similar to radiation‐induced bone damage, while the rescue of this signalling significantly alleviated radiation‐induced bone injury. Our findings show that the decrease in HIF‐1α/PDGF‐BB/PDGFRβ signalling between type H ECs and PDGFRβ+ PVCs after irradiation affects the homeostasis of EC‐PVC coupling and plays a part in vascular damage and bone loss, which has broad implications for effective translational therapies.


| INTRODUCTION
Due to the high metabolism and calcium content, bone tissue has a higher radiation sensitivity 1 and absorption rate 2 than other tissues, resulting in common complications such as osteoradionecrosis, radiation-induced osteoporosis and refractory pathological fractures after radiotherapy. [3][4][5] Nevertheless, the pathogenesis of radiation-induced bone injury has not been fully elucidated, and the efficacy of clinically used interventions and therapies seems to be limited. An in-depth understanding of the mechanism of radiation-induced bone injury would pave the way for novel effective targeted therapy.
Radiation can directly injure cells by damaging DNA structure through ionization or inducing the accumulation of reactive oxygen Jiayan Li and Xiaodan Chen contributed equally to this work. species. It can also exert deleterious effects indirectly by remodelling the cell microenvironment. [6][7][8] Microvessels are a key component of the bone marrow microenvironment. [9][10][11] In addition to transporting oxygen and nutrients, endothelial cells (ECs) in blood vessels release a variety of paracrine signals, known as angiocrine signals, to form a unique vascular niche that regulates the selfrenewal and differentiation of perivascular stem and progenitor cells. [11][12][13] There is growing recognition that radiation deteriorates perivascular mesenchymal stem cells (MSCs) by disrupting their vascular niche 7,14 ; however, the underlying mechanism remains largely unknown.
With further research, intraosseous microvessels have been divided into two subtypes, type H vessels (CD31 high /Emcn high ) and type L vessels (CD31 low /Emcn low ), according to their differences in surface markers, morphology and distribution. 15 Type H vessels, with high expression of Endomucin (Emcn) and CD31, are located in the metaphysis and participate in bone formation by regulating perivascular osteoprogenitors via paracrine signalling, 16 whereas type L vessels, which express Emcn and CD31 at a low level, are associated with haematopoiesis. 11,15 Different microvessel subtypes are surrounded by different types of perivascular cells (PVCs). 17 Among them, PDGFRβ + PVCs are mainly distributed around osteogenesis-related type H vessels. 15,16,18 Previous studies have revealed that pericytes in the bone marrow with osteogenic capacity are PDGFR-positive, 19 and PDGFRβ is specific for the osteolineage cluster in single-cell RNA-seq and labels skeletal stem and progenitor cells and their reparative progeny during fracture healing, 20 implying the osteogenic potential of PDGFRβ + PVCs. In soft tissues, such as the brain, retina, heart and kidney, PDGFRβ + PVCs have been reported to function widely as pericytes to maintain vascular stability physiologically. In addition, PDGFRβ + pericytes can detach from microvessels and undergo differentiation transition in pathological conditions. [21][22][23][24][25][26] However, the biological characteristics of PDGFRβ + PVCs near type H vessels in the bone marrow are not yet clear. Of note, in contrast to type L vessels, type H vessels express a relatively high levels of angiocrine signals with osteogenic properties, including platelet-derived growth factor-BB (PDGF-BB), 11,15,27,28 which is the main ligand of PDGFRβ. It is speculated that PDGF-BB secreted by type H ECs plays a role in the regulation of PDGFRβ + PVCs and orchestrate angiogenesis and osteogenesis at the molecular level in the skeletal system.
Here we hypothesize that the change in type H vessels and PDGFRβ + PVCs after irradiation contributes to abnormal angiogenesis and osteogenesis during the progression of bone injury.
To address this, we first analysed the characteristic alterations  Figure 1B). In addition, a marked increase in TRAP + cells ( Figure 1C) and a significant decrease in perivascular RUNX2 + cells ( Figure 1D) were observed in the metaphysis, indicating the loss of the balance between osteogenic and osteoclastic activity. Additionally, the microvessel density and the number of total ECs in bone marrow were reduced ( Figure 1E,F), demonstrating the impairment of the microvascular system.
Previous research has shown that the sensitivity of ECs to irradiation varies in different tissues/organs. 14 Additionally, different EC subtypes in bone marrow also revealed different responses after irradiation. The result of immunofluorescent staining revealed a relatively obvious decline in microvessels in the metaphysis ( Figure 1E), which was further confirmed to be the recession of type H vessels with double-staining of Emcn and CD31 ( Figure 1G). Although flow cytometry (FCM) showed an increase in type H ECs (Emcn high /CD31 high ) in the incipient stage after irradiation ( Figure S1A), it might be due to the upregulation of Emcn and CD31 expression in sinusoid ECs according to previous studies. 15,29 The DNA damage in the metaphysial ECs was more obvious than that in diaphyseal ECs at 6 h post-irradiation ( Figure S1B), revealing the injury of type H vessels in the metaphysis. As the injury progressed, the number of type H ECs displayed a significant decrease on day 28 after irradiation ( Figure 1H), resulting in the deficiency of type H vessels in the metaphysis ( Figure 1G). Notably, the falling range of type H ECs was larger than that of type L ECs ( Figure 1H), echoing the distinct decline of microvessels in the metaphysis ( Figure 1E). Similarly, PDGFRβ + PVCs, which were mainly located close to type H vessels in the nonirradiated tibia ( Figure 1I), ranged from increase to decrease in quantitative terms after irradiation ( Figures S1C and 1J). Moreover, PDGFRβ + PVCs detached from the metaphysial vessels, resulting in vessel sparseness ( Figure 1I). These correlated alterations in number and distribution suggested a close interconnection between PDGFRβ + PVCs and type H vessels.
Overall, irradiation led to impairment of angiogenesis and osteogenesis in the bone marrow. Among these, type H vessels and PDGFRβ + PVCs in the metaphysis reveal correlated alterations. However, the internal mechanism between the ECs and PVCs, as well as F I G U R E 1 The integration of type H endothelial cells (ECs) and platelet-derived growth factor receptor β + (PDGFRβ + ) perivascular cells is disturbed after radiation. (A) Representative micro-computed tomography images and quantitative cancellous bone parameters of the control and irradiated tibiae (n = 6). (B) Representative haematoxylin and eosin staining images of the metaphysis region and quantification of adipocytes of the control and irradiated tibiae. Scale bars, 500 μm (n = 6). (C) Representative fluorescence images of the metaphysis region in the control and irradiated tibiae stained with TRAP (green) and DAPI (blue). Quantitative analysis of the number of TRAP + cells. Scale bars, 100 μm (n = 6). (D) Representative fluorescence images of the metaphysis region in the control and irradiated tibiae stained with RUNX2 (green), Emcn (red) and DAPI (blue). Quantitative analysis of the number of RUNX2 + cells. Scale bars, 100 μm (n = 6). (E) Representative anatomic images and corresponding fluorescence images of the control and irradiated tibial sections stained with Emcn (red). Vessel density was quantified according to the fluorescent area. Scale bars, 200 μm (n = 6). (F) Flow cytometric quantitation of total ECs (CD31 + /CD45 À /Ter119 À ) from the femurs and tibiae of the control and irradiated mice (n = 8). (G) Representative fluorescence images of the tibial sections stained with CD31 (green) and Emcn (red) show type H vessels in the metaphysis. Scale bars, 100 μm. (H) Flow cytometric quantitation of type H ECs (CD31 high /Emcn high ) and type L ECs (CD31 low /Emcn low ) from the femurs and tibiae of the control and irradiated mice (n = 8). (I) Representative fluorescence images of tibial sections stained with PDGFRβ (green) and Emcn (red). The white arrows represent PDGFRβ + cells detached from microvessels. Scale bars, 100 μm. (J) Flow cytometric quantitation of PDGFRβ + cells (CD45 À /Ter119 À /CD31 À /CD140b + ) from the metaphysis region of tibiae of the control and irradiated mice (n = 8). Data are represented as the mean ± SD. The p-value was calculated by unpaired, two-tailed Student's t-test. BV/TV, trabecular bone volume fraction; Tb.N, trabecular number; Tb.Th, trabecular thickness; Tb.Sp, trabecular separation; SMI, structure model index; Conn.D, connectivity density the functional change in PDGFRβ + PVCs under the radiation microenvironment, remains to be further explored.

| Radiation weakens the vascular stability and osteogenic potential of PDGFRβ;Td + cells
Given that PDGFRβ labels the osteolineage cluster in bone marrow, 19,20 it is reasonable to assume that PDGFRβ + PVCs repre-

| Exogenous PDGF-BB relieves radiationinduced bone injury
Since the downregulation of PDGF-BB/PDGFRβ signalling has been proven to be responsible for the functional alteration of PDGFRβ;Td + cells, we then proceeded to elucidate whether it is related to the changes in bone phenotype after irradiation and tried to use exogenous recombinant PDGF-BB protein as a radio-protector of bone tissue for rescue therapy. The results of micro-CT and histological staining showed that the administration of IM, a PDGFRβ inhibitor to block PDGF-BB/PDGFRβ signalling, could lead to a decline in bone mass and an increase in adipogenesis in the bone marrow that was similar to the phenotypes that appeared in irradiated bone. In contrast, exogenous PDGF-BB effectively enhanced bone mass

| Upregulation of HIF-1α/PDGF-BB/PDGFRβ signalling between ECs and PDGFRβ + PVCs alleviates radiation-induced vascular damage and bone loss
Next, we explored the underlying mechanisms by which radiation reduced PDGF-BB expression in ECs. HIF-1α is one of the major upstream regulators of PDGF-BB. 18,34 Here, we found that the expression of HIF-1α in the metaphysis was attenuated significantly after irradiation ( Figure S8A). However, when deferoxamine mesylate (DFM), an activator of HIF-1α, 35,36 was applied, the expression of HIF-1α was upregulated ( Figure S8A,B). Unsurprisingly, the expression Microvascular lesions have been proposed as an initial factor in the development of radiation-induced bone injury, 1,7,37 which is thought to be related to the low tolerance of ECs to radiation. 24,25,38 Our data further show type H vessels, which were located in the active metaphysis, display more obvious DNA damage and have a larger percentage reduction than type L vessels in the diaphysis. This might be because metabolically active tissues and cells are more sensitive to direct ionizing damage. 6 Additionally, PDGFRβ + pericytedeficient mice demonstrated significant microvessel regression and micro-circulation failure, 26,39,40 revealing that PDGFRβ + PVCs play an important role in microvascular homeostasis. Thus, the decline in PDGFRβ + PVCs and their detachment from type H vessels after irradiation also aggravate vascular instability and collapse. In summary, the remarkable reduction in type H vessels in the irradiated tibia is associated with their low radiation tolerance, high metabolic level and homeostatic disturbance caused by the reduction and detachment of PDGFRβ + PVCs.
Although PDGFRβ + cells display a similar decreasing amplitude to type H ECs in vivo, cells undergoing high doses of direct radiation still survive and maintain proliferative capacity in vitro, indicating that direct ionizing damage is not the only reason to be blamed, and the indirect influences from the vascular niche are also responsible for the alterations of PDGFRβ + PVCs in the radiation microenvironment.
According to the close spatial and functional connection between type H vessels and PDGFRβ + PVCs, we hypothesize that this indirect effect is associated with the paracrine alteration of type H ECs.
Among the paracrine factors, PDGF-BB, which is highly expressed in type H ECs, 27,28 has been proven to be the major EC-derived factor that regulates pericyte function. 45 Here we extend the findings that the PDGF-BB decline in type H vessels acts as an important trigger for the functional alteration of PDGFRβ + PVCs after radiation, including the reduction in cell number, the abnormality in cell migration and the change in cell differentiation.
In contrast to the decrease of ECs and PDGFRβ + PVCs, we found that the number of osteoclasts in metaphysis increased after irradiation, coinciding with the enhancement of bone loss. However, in addition to the classic bone-associated osteoclasts which were the major subtype occupying ageing bones, a novel subtype of osteoclasts, vessel-associated osteoclasts (VAO), were found closely associated with type H vessels and were predominant in early developing bones. 46 Some key osteoclastogenic factors such as Csf1, Il1a and Tnfsf11a were highly expressed in type H compared with type L ECs, suggesting the considerable crosstalk between type H ECs and VAOs.
Interestingly, the number of VAOs increased in the early stage after irradiation, 46 but the role of the increased VAOs in the progression of bone injury and whether the proportion of different osteoclast subtypes would change in the later period after irradiation were not clear.
Besides, we know that bone remodelling is tightly regulated by crosstalk between bone-forming osteoblasts and bone-resorbing osteoclasts. 47 56 In adipose tissue, both upregulation of PDGF-BB in M1 macrophages after HFD and administration of exogenous PDGF-BB induced detachment of perivascular PDGFRβ + cells from the original vascular surface. 57 These observations indicated that the changes in PDGF-BB concentration and distribution have a significant impact on the migration of PDGFRβ + PVCs. However, we found that the change in PDGF-BB expression in the metaphysis after irradiation seems to be related to type H vessels rather than TRAP + preosteoclasts. Interestingly, although PDGF-BB declines in most metaphysial ECs after radiation, it can still be detected around the residual microvessels tightly below the growth plate, where certain PDGFRβ + PVCs are maintained. Briefly, these findings demonstrate that the decrease of ECderived PDGF-BB in type H ECs in the metaphysis is largely responsible for the detachment of PDGFRβ + PVCs after irradiation.
The osteogenic lineage of PDGFRβ + cells in bone marrow has been revealed in previous studies. Specifically, NG2;Td + pericytes which are PDGFRβ-positive contribute to bone fracture healing as a cellular source of osteogenic cells. 19 Single-cell sequencing revealed that PDGFRβ specifically labelled the osteolineage cluster and marked all osterix + skeletal stem and progenitor cells and their reparative progeny during fracture healing. 20 On that basis, our study further shows that perivascular PDGFRβ;Td + cells highly express classic osteogenic markers and can be found in bone surface and bone lacuna. Furthermore, we propose that radiation impairs PDGF-BB signalling in metaphysial microvessels by regulating HIF-1α. As a newly identified strong radio-sensor, HIF-1α is also the main oxygen sensor activating the transcription of genes that are involved in angiogenesis, cell survival, glucose metabolism and so forth. 63 Previous studies revealed that PDGF-BB was one of the downstream targets of HIF-1α. EC-specific Hif1a knockout (Hif1a iΔEC ) mice revealed a decline in Pdgfb in ECs. 18 Besides, HIF-1α-siRNA significantly decreased PDGF-BB expression in ECs, which was increased in HIF-1αoverexpressing ECs. 34 Although bone is a highly vascularized tissue, the oxygen tension in bone marrow remains at a relatively low level, 64,65 especially in the metaphysis, which requires a large amount of oxygen consumption due to active osteogenesis. 65 This may partially explain why the expression level of HIF-1α 15 and its downstream PDGF-BB 15,27,28 are comparatively higher in type H ECs than in type L ECs. However, weakened cellular metabolism along with widespread vascular leakage in the irradiated bone significantly reduced oxygen consumption in cells and relieved hypoxia in bone marrow. 64 These were in line with our finding that HIF-1α in the metaphysis and PDGF-BB in the metaphysial ECs were both decreased after irradiation. However, ECs dysfunction caused by other factors, such as oxidative stress, is also responsible for the decline of PDGF-BB in ECs.
Notably, Hif1a iΔEC mice revealed a reduction in type H vessels and perivascular osteoblast-related cells in the metaphysis, including PDGFRβ + PVCs, leading to a simultaneous deterioration in both angiogenesis and osteogenesis. 15,18 These appearances were similar to those occurring in the irradiated tibia in our study, implying that HIF-1α revival in ECs may help alleviate radiation-induced bone injury.
DFM is a classic stabilizer of HIF-1α, which interrupts the degradation of HIF-1α by inhibiting oxygen-dependent and iron-dependent enzymes prolyl-4-hydroxylases. 66 It can increase the accumulation and nuclear translocation of HIF-1α to upregulate downstream angiogenic factors. 67 A previous study showed that DFM treatment enhanced the expression of PDGF-BB in bone marrow ECs, increased the number of type H vessels and vessel-associated osterix + cells, and ultimately increased bone mass in mice with age-related osteoporosis. 15 In this study, we extend the therapeutic mechanisms of DFM in radiation-induced osteoporosis. In particular, we focussed on the HIF-1α/PDGF-BB/PDGFRβ axis between ECs and PVCs as well as the functional alteration of PDGFRβ + cells, revealing that DFM treatment relieves bone injury by activating specific EC-PVC crosstalk. Additionally, the alleviation of radiation-induced bone injury by DFM is partly based on the activation of PDGFRβ + PVCs. When PDGFRβ is inhibited, microvascular rarefaction and bone loss still exist.  Table S1. For lineage tracing studies, mice received tamoxifen (MCE, USA) dissolved in corn oil via intra-peritoneal injection at 4-5 weeks postnatally at a dose of 75 mg/kg every day for a total of 5 consecutive days.

| Intra-tibial injection and intra-peritoneal injection
The intra-tibial injection procedure was applied according to the bone marrow transplantation protocol described previously. 31,68 Briefly, the tibia was pushed up and slightly rotated to fix the knee at $90 . A

| Immunofluorescence staining
Frozen bone sections were obtained as described previously with slight modification. 69  Nuclei were counterstained with DAPI. Fluorescent images were acquired by laser confocal microscopy (FV3000, Olympus, Japan).

| BMS collection
BMS from tibiae were collected as previously described with slight modification 27,31 and used for cell migration assays and enzymelinked immunosorbent assay (ELISA) analysis. Briefly, dissected tibiae from WT C57BL/6 mice were cleaned to remove adherent soft tissues in ice-cold PBS, and the bone marrow was exposed by carefully cutting off the distal end. Then, specimens were placed in 1.5 ml Eppendorf tubes containing 100 μl of ice-cold PBS with the distal end oriented toward the bottom of the tube. After centrifugation at 3000 rpm at 4 C for 20 min, 70 μl of total supernatant was collected and stored at À80 C.

| Conditioned medium collection
The conditioned medium (CM) of HMEC-1 cells was collected for cell migration assays and ELISA analysis. Equivalent volumes of CM in different groups with an approximate number of cells were collected.
Supernatants were centrifuged at 3000 rpm at 4 C for 10 min and carefully transferred to clean 1.5 ml Eppendorf tubes for immediate use or stored at À80 C.

| Enzyme-linked immunosorbent assay
PDGF-BB analysis of BMS or CM was performed using a QuantiCyto ® PDGF-BB ELISA kit (EMC032.96, Neobioscience, China) based on the manufacturer's instructions.

| Cell migration assay
The chemotactic effect of BMS from tibiae and CM from HMEC-1 cells on PDGFRβ;Td + cells was evaluated by a cell migration assay. In brief, PDGFRβ;Td + cells were seeded in the upper chambers of the cell-cultured insert with 8 μm pore filters (353097, Falcon, USA) at a concentration of 1 Â 10 4 cells/well, and BMS or CM with recombinant PDGF-BB protein or anti-PDGF-BB neutralizing antibody was added to the lower chambers. In the IM treatment group, PDGFRβ; Td + cells were pre-incubated with IM for 1 h to block PDGFRβ before cell seeding. After 12 h, the migrated cells were fixed in 4% paraformaldehyde, stained with crystal violet, captured under an inverted microscope system (Axio Observer 5, Zeiss, Germany), and quantified using ImageJ software.

| Western blot
Western blot analysis was performed as previously described. 22 In particular, for evaluation of the protein translation level in whole bone tissues, dissected tibiae were crushed finely in ice-cold PBS and digested with collagenase before protein extraction. The primary antibodies used in this study included Collagen1 (COL1,

| Osteo-angiogenic affinity assay
The co-culture of PDGFRβ;Td + cells and HMEC-1 cells on Matrigel (356234, Corning, USA) was arranged to detect the recruitment effect of ECs on PDGFRβ;Td + cells as previously described with slight modification. 20 Briefly, HMEC-1 cells were labelled with DiO (green) membrane dye (C1993S, Beyotime) and seeded at a concentration of 2 Â 10 4 cells/well on Matrigel-pre-coated 96-well plates. After incubation for 1 h, 2 Â 10 3 cells/well of PDGFRβ;Td + cells (red) were added and co-cultured with endothelial tubules for 3 h. In the IM treatment group, PDGFRβ;Td + cells were preincubated with 5 μM IM for 1 h. Images were taken by an inverted microscope system and analysed using ImageJ software. The endothelium-related cells were calculated by counting Td + cells on tubules per field of view.

| Statistical analyses
All data are presented as the mean ± SD. Unpaired, two-sided Student's t-test was used for comparisons between two treatment groups, and one-way analysis of variance with Tukey's post hoc test was used for multiple comparisons. Differences were considered significant at p < 0.05. GraphPad Prism 8 (GraphPad Software, USA) was employed for statistical analysis.