Introduction

Delayed brain radiation necrosis is an adverse event widely known to be associated with therapeutic irradiation to the central nervous system (CNS) and focal injuries of the late radiation effects with focal neurologic abnormalities [1]. Late radiation injuries can occur from a few months to 10 or more years after radiotherapy [2]. Histological changes in delayed radiation effects were initially described by Fischer and Holfelder [3] in 1930. Although CNS was first thought to be highly radio-resistant, an animal study showed that CNS, particularly the white matter in the deep part of the brain was more radio-responsive than previously supposed [4]. Different conditions of data acquisition mean that the reported incidence of radiation necrosis varies from 3 to 24 % [5]. In long-term survivors with metastatic brain tumors treated with gamma knife radiosurgery, 64 % of patients developed radiation necrosis, with median survival of 32 months [6]. The predictive factors for developing radiation necrosis in patients with gliomas were radiation dose, use of chemotherapy, and length of survival [5]. Therapeutic irradiation is an important treatment modality to locally control disease in malignant CNS tumors, as there are limited effective chemotherapeutic agents that are able to pass through blood–brain barrier (BBB). Recently, high-dose radiotherapies such as intensity-modulated radiation therapy, proton beam radiotherapy, and boron neutron capture therapy have been used for treatment of malignant gliomas, with better outcomes expected than conventional radiotherapy. However, there is a high incidence of radiation necrosis or symptomatic change requiring surgical management in these high-dose radiotherapies [7–9]. To conquer CNS malignancies, it is necessary to manage radiation necrosis in addition to developing high-dose radiotherapies.

However, diagnosis and treatment of radiation necrosis remain challenging as it is usually difficult to distinguish radiation necrosis from tumor recurrence. In malignant gliomas, pseudoprogression has also become a problem, and is difficult to differentiate from early disease progression following chemoradiotherapy [10]. Pseudoprogression is treatment-related necrosis occurring within 2 months of treatment [11]. The therapeutic strategies for tumor progression and treatment-related necrosis are completely different. Radiation necrosis initially responds to corticosteroids, but progressive radiation necrosis is refractory to medical treatment. However, the vascular endothelial growth factor (VEGF) has recently been reported to play an important role in the development of radiation necrosis [12–16]. In addition, bevacizumab, a monoclonal anti-VEGF antibody, has been reported to be effective in treating radiation necrosis and reducing perilesional edema [17, 18]. We recently published a review of radiation necrosis [19], providing an overview of radiation necrosis from pathophysiology to treatment for clinicians with the aim of helping the management of patients with radiation necrosis. However, further clarifying the molecular biological pathology of radiation necrosis may lead to the discovery and development of a novel, effective therapy for radiation necrosis such as anti-VEGF therapy. Therefore, in the present review, we focus on reviewing the literature on the pathology of radiation necrosis, and highlight recent molecular pathological findings with regard to radiation necrosis.

Vascular damage as the primary cause of radiation necrosis

In the 1930s, seminal case reports demonstrated radiation necrosis as the late radiation injury to the brain, presenting as neurological deterioration after cranial irradiation with a long latent interval. Pennybacker and Russell [20] reported case series including nine patients who encountered harmful effects to the brain after radiation therapy. The authors described a pathological observation of a resected lesion in a patient without brain tumor (irradiated for a rodent ulcer of the scalp). They demonstrated architectural changes of the gyrus and vasculature. Further, loss of neurons and gliosis were observed in the cortex. Hemorrhage, thrombosis, and collagenous thickening and fibrinoid necrosis of vessels were also observed in the white matter. From their clinical and experimental studies, the authors concluded that vascular damage played an important role in the development of radiation necrosis [21]. The authors also speculated that an increase in capillary permeability preceded the lesion, but could not show direct pathological evidence. A second pathological study using electron microscope showed capillary basal lamina thickening, swelling of endothelial cells, and morphological changes in glial cells [22]. These were suggested to be secondary changes following capillary dysfunction. In a human study, Martins et al. [23] tested the hypothesis that the primary site of radiation injury was endothelial cells. Their cases showed that white matter was damaged with thickened and hyalinized small vessels. Many of the fibrinoid, necrotic, and fragmented vessels were surrounded by hemorrhage and edema. Necrosis was found in not only the cerebral parenchyma but also the white matter with gliosis. Radiation injuries to blood vessels are most likely to be initiated by breaks in double-strand deoxyribonucleic acid, leading to phenotypic changes in endothelium [24].

Angiogenesis and VEGF

Tsao et al. [12] proposed that BBB dysfunction caused by endothelial cell damage played a major role in the pathogenesis of late radiation injury. The authors demonstrated that VEGF protein and mRNA were expressed in astrocytes around areas of white matter necrosis and were temporally and spatially associated with albumin extravasation as a surrogate of blood–spinal cord barrier (BSCB) breakdown in an irradiated rat spinal cord [12]. Using this animal model, the authors showed that a hypoxic marker of the nitroimidazole group and VEGF were colocalized in cells surrounding the necrotic core, and that hypoxia-inducible factor-1α (HIF-1α) and VEGF were expressed in astrocytes around areas of necrosis [13, 14]. HIF-1α and VEGF expression were colocalized following BSCB disruption. The authors postulated that endothelial cells were initially damaged after irradiation, and subsequently, the barrier was disrupted. In a study of experimental radiosurgery, intraparenchymal edema was identified as the initial histological change after radiosurgery [25]. An increase in vascular permeability may also result in hypoxia. Astrocytes responded to HIF-1α upregulated VEGF, which further increased vascular permeability and exacerbated tissue hypoxia. Approximately 74 % of VEGF-expressing cells were astrocytes, whereas only 43 % of HIF-1α-positive cells were identified as astrocytes [13, 14].

We reported a pathological examination of radiation necrosis in specimens from 18 patients who underwent surgical removal of necrotic tissue [15]. Hematoxylin and eosin staining showed enlarged vessels with thin walls, as observed with telangiectasis, at the border zone between the normal brain and necrotic core (perinecrotic area) (Fig. 1). These telangiectatic vessels were often accompanied by microbleeding and interstitial edema and could be leaky and fragile. Similar findings occurred in radiation necrosis irrespective of the original tumor pathology and modality of radiation therapy. These telangiectatic vessels could be caused by angiogenesis and affected by VEGF. In an immunohistological analysis, reactive astrocytes in the perinecrotic area exhibited strong positive staining for VEGF expression (Fig. 2). Further, glial fibrillary acidic protein (GFAP)-positive cells expressed VEGF as assessed by double immunofluorescence staining. HIF-1α was also expressed in cells that resembled glial cells around the necrotic core [26]. Thus, reactive astrocytes are likely to play an important role in angiogenesis in radiation necrosis.

Fig. 1
figure 1

Histological findings. Micrographs showing enlarged thin-walled vessels (a black arrows), hyalinized vessels (a white arrows), and reactive astrocytes (b black arrows) in the surrounding area of the necrotic core. (a hematoxylin and eosin stain, magnification 100×; b Glial fibrillary acidic protein immunohistochemistry, magnification 200×)

Full size image
Fig. 2
figure 2

Immunohistochemical findings. Immunohistochemistry for vascular endothelial growth factor-A (VEGFA) revealed that VEGFA-expressing cells were distributed around the necrotic core (a magnification 100×). Reactive astrocytes (b, c black arrows), reactive microglia (ameboid microglia) (b white arrows), and endothelial cells (c white arrowheads) showed positive staining for VEGFA in the surrounding area of the necrotic core (b, c magnification 400×)

Full size image

Inflammation, cytokines, and angiogenesis

It was reported that the myelin sheath and oligodendrocytes were directly damaged by radiation in the irradiated injury model of the spinal cord [27]. In the brain radiation injury model, the demyelinating process preceded white matter necrosis before delayed radiation necrosis, which appeared before the onset of any vascular changes [28]. In radiosurgery using a rat glioma model, cellular edema and hypocellular architectural change without alteration in blood vessel morphology were suggested to be related to the direct cytotoxic effect of radiation [29]. Degeneration of white matter associated with variable degrees of astrogliosis and demyelination were frequently observed in a postmortem study of patients with intracranial gliomas [30]. Inflammatory reactions were seen with myelin damage, hemorrhage, and edema. Kureshi et al. [31] investigated lymphocyte and macrophage infiltration as well as cytokine expression in human specimens of radiation necrosis. T cell lymphocytes and macrophages were diffusely infiltrated in the perivascular and parenchymal spaces. Macrophages heavily infiltrated the surrounding areas of necrosis and perivascular proliferation, and strongly expressed Interleukin (IL) and tumor necrosis factor (TNF)-α. Although IL-6 and TNF-α were also identified in reactive astrocytes, the staining of IL-6 and TNF-α was weaker in reactive astrocytes than in infiltrating macrophages. IL-1α was expressed in activated macrophages or microglia. The authors suggested that these inflammatory cytokines may be released for cellular protection [32], but may also be involved in the gliosis, endothelial cell proliferation, and tissue necrosis at the late phase after irradiation. A second human specimen study revealed that many inflammatory cells, including macrophages and lymphocytes, were observed with telangiectatic vascularization in the area around the necrosis [33]. The author concluded that overexpression of inflammatory cytokines played a key role in the progression of radiation necrosis. In addition, the author suggested that coagulation necrosis was accompanied by the disorder of endothelial cells.

We presumed that inflammation and angiogenesis were major contributors to the progression of radiation necrosis and investigated the relationship between these two pathophysiologies. Although HIF-1α is a key regulator for inducing VEGF expression, VEGF is also upregulated through a HIF-1α-independent pathway [34–36]. Interestingly, chemokines stimulate not only inflammation, but also angiogenesis [37–39]. In human glioblastomas, it was reported that expression of CXCL12 and CXCR4 were colocalized in the regions of necrosis and angiogenesis [40], and furthermore, function of the CXCL12/CXCR4 axis promoted VEGFA expression in glioma cells [41, 42].

In our study, CXCL12 was detected in GFAP-positive cells, and CXCR4 was expressed in CD68-positive cells in human specimens of radiation necrosis [26]. Most of the CD68-positive cells were identical to the human glucose transporter 5-positive cells. This finding indicated that microglia expressed CXCR4. Microglia also expressed IL-1α, IL-6, and TNF-α. Therefore, chemokines could stimulate microglial cells to produce these cytokines, exacerbating inflammation. However, VEGF was mostly expressed in reactive astrocytes, not microglia. The relationship between CXCL12/CXCR4 signaling and VEGF production is currently unclear. Lymphocytes expressing CD45 also expressed CXCR4, but did not express any cytokines. We concluded that microglia played a key role in inflammation in radiation necrosis. Our hypothetical mechanism of radiation necrosis is shown in Fig. 3.

Fig. 3
figure 3

Hypothetical mechanisms of radiation necrosis. Endothelial cells of cerebral vessels are initially damaged by irradiation, resulting in tissue hypoxia. Reactive astrocytes, microglia, and macrophages respond to hypoxia by endothelial dysfunction after radiotherapy, and express hypoxia-inducible factor (HIF)-1α. HIF-1α strongly mediates up-regulation of vascular endothelial growth factor (VEGF). VEGF is a vascular permeability factor as well as an angiogenic factor. Overexpression of VEGF causes leaky and fragile angiogenesis accompanied by microscopic bleeding, despite protection from hypoxic injury. Reactive astrocytes express CXCL12, and recruiting microglia and macrophages express CXCR4 via CXCL12/CXCR4 chemotaxis. These monocytes contribute to inflammation by releasing inflammatory cytokines including NF-κB, interleukin (IL)-1α, IL-6, and tumor necrosis factor (TNF)-α. These processes of angiogenesis and inflammation exacerbate tissue hypoxia and vasogenic edema, resulting in the progression of radiation necrosis

Full size image

Platelet-derived growth factors (PDGFs) are currently of interest as an angiogenic factor [43, 44]. PDGF was identified as a constituent of whole blood serum and was subsequently purified from human platelets [45]. PDGFs consist of disulphide-linked dimers and members of the PDGF/VEGF family of growth factors. PDGFs were overexpressed in pathological conditions including malignancies, atherosclerosis, and fibroproliferative diseases.

In our study using human specimens, PDGF-C and -D were expressed at a significantly higher level than PDGF-A and -B around the necrotic core [46]. PDGF-C and -D were mainly expressed by macrophages, microglial cells, reactive astrocytes, lymphocytes, and endothelial cells. Platelet-derived growth factor receptor (PDGFR)-α was also expressed by these cells. PDGF-C is potentially an angiogenic factor, similar to VEGF and PDGF-A and -B. PDGF-D has been shown to stimulate angiogenesis [47]. However, in radiation necrosis, PDGFR-β, which PDGF-D binds to with high affinity, was only expressed in endothelial cells. Therefore, PDGF-C is the critical regulator in PDGFs. Macrophages, microglia, and reactive astrocytes significantly contribute to inflammation and angiogenesis through PDGF-C/PDGF-α autocrine and paracrine signaling in radiation necrosis. The expression of inflammatory and angiogenic molecules in our study of radiation necrosis is summarized in Table 1.

Table 1 Expression of inflammatory cytokines and growth factors in cells participating in radiation necrosis
Full size table

Treatment of radiation necrosis

Although existing treatments of radiation necrosis are theoretically effective, based on pathophysiology, these treatments are not always effective. Corticosteroids are commonly used and are mildly effective on radiation necrosis [48]. Corticosteroids suppress inflammation and stabilize BBB, resulting in a reduction in vasogenic edema. Anticoagulants could improve microcirculation and are expected to prevent radiation necrosis, by improving ischemic and hypoxic conditions. Surprisingly, anticoagulants showed only mild effectiveness on symptomatic radiation necrosis [49]. Hyperbaric oxygen therapy (HBO) could raise the partial pressure of oxygen in the irradiated tissue. However, it was difficult to resolve neurological symptoms by HBO [50]. Further, prophylactic HBO did not reduce the incidence of radiation necrosis after stereotactic radiosurgery (SRS) for brain metastasis, although white matter injury was less frequent in patients with HBO after SRS than in patients without HBO [51]. In the surgical approach, resection of necrotic tissue is a useful option to improve perilesional edema. This treatment can remove not only necrotic tissue, but also cells responsible for producing VEGF and/or inflammatory cytokines in tissue surrounding the necrosis. However, radiation therapy is typically applied to unresectable residual tumors adjacent to the eloquent area after surgical removal of tumors. Therefore, surgical removal of necrotic tissue has a high morbidity of neurological deterioration. In a review of surgically treatments, 35 of 44 patients (79.5 %) showed a good outcome; however, two patients (4.5 %) experienced deterioration after surgery [31]. Furthermore, seven patients (15.9 %) died because of surgical complications. Although this review was published in 1980s, the mortality rate was far from low. In a recent report on surgical management for radiation necrosis using neuronavigation and intraoperative magnetic resonance imaging, there was no mortality [35]. However, surgical complications and neurological deterioration were still observed in 54 % of patients [52]. As previously mentioned, bevacizumab, a monoclonal anti-VEGF antibody, has been highlighted as a new alternative treatment for radiation necrosis. Bevacizumab can efficiently reduce perilesional edema through stabilizing BBB and preventing angiogenesis via trapping of VEGF. Gonzalez et al. [53] first reported the efficacy of bevacizumab on radiation necrosis of the brain. They retrospectively reviewed 15 cases who underwent bevacizumab treatment. Of the 15 patients, eight patients were retrospectively diagnosed with radiation necrosis and responded better to bevacizumab than other patients with recurrent tumors. Subsequently, several case studies and one small, randomized controlled trial have reported clinical and radiological improvements in all patients who received bevacizumab [17, 54–60]. However, there are no further clinical trials designed to acquire on-label approval for use in the United States and Europe. Therefore, we conducted a prospective multicenter clinical trial in Japan.

Conclusions

Initial injury by irradiation is prominent in the myelin and the endothelial cells. Inflammation and angiogenesis are related to these cell injuries and are responsible for delayed brain radiation necrosis. These two processes interact in a complex manner, and it is unknown which occurs first. Chronological changes in molecular mechanisms should be elucidated to resolve this question. On the other hand, the observation that bevacizumab can improve outcomes in patients with radiation necrosis demonstrated that VEGF is one of key molecules controlling the pathology of delayed radiation injury in the brain. Bevacizumab is now expected to be utilized as an alternative therapy for intractable delayed radiation necrosis. We showed that PDGF/PDGFR contributed to radiation necrosis and predict that targeted therapy for these molecules could be a novel treatment for radiation necrosis. In the future, delayed radiation necrosis may be prevented by a novel therapy that blockades certain molecules, if the molecular mechanism preceding the progression of delayed radiation necrosis is elucidated.