5-Aminolevulinic acid (5-ALA) is a natural metabolic precursor in the heme biosynthesis pathway. Oral administration of a large volume of 5-ALA overloads the heme pathway and induces the synthesis and selective accumulation of the fluorescent protoporphyrin IX (PpIX) in tumor cells and epithelial tissue (1,2). Increased vascular permeability attributable to disruption of the blood-brain barrier (BBB) in the tumor and decreased levels of ferrochelatase activity in tumor cells contribute to this phenomenon in glioblastoma multiforme (GBM) (3–6). PpIX in GBM tumors, present at higher levels compared to normal brain tissue, emits red-violet fluorescence under blue light, visually distinguishing tumor margins and allowing intraoperative, objective assessment of tumor infiltration (7,8). Evidence demonstrating the high sensitivity and specificity of 5-ALA-induced PpIX fluorescence in GBM provides support for the diagnostic accuracy of 5-ALA. As a result, 5-ALA fluorescence guidance has come to be widely used to improve the extent of tumor resection (9).

Since a number of studies have demonstrated that the extent of resection (EOR) is correlated with improved prognosis of patients with GBM, neurosurgeons have sought to maximize the EOR, an objective constrained by the difficulties of resection arising from the invasive nature of GBM (10–15). A large, randomized, controlled, multicenter phase III trial has shown that 5-ALA fluorescence-guided surgery for malignant glioma leads to a significantly higher resection rate, resulting in prolonged progression-free survival compared with conventional microsurgery guided by white light (16). In addition, several techniques, including intraoperative neuronavigation (17), intraoperative magnetic resonance (MR) imaging (18), intraoperative ultrasound (19) and intraoperative electrical stimulation (20), have been introduced to facilitate optimal resection, thereby maximizing safe EOR and improving survival in patients with GBM. Increasing the EOR of GBM to achieve these survival benefits brings with it a greater chance of entering the ventricular system during the course of cytoreduction. In some cases of 5-ALA-guided GBM surgery, upon ventricular entry we encountered fluorescence in ventricular walls that lacked enhancement on MR images and were free of macroscopic invasion of tumor cells. However, the implications of this 5-ALA-induced fluorescence of the ventricle wall are not yet fully understood.

In this study, we obtained 25 ventricular wall tissues from 19 patients with newly diagnosed GBM that showed no enhancement on MR images, and investigated the relationship between intraoperative 5-ALA fluorescence and histopathology findings of these non-enhancing ventricular wall tissues.

Because 5-ALA has been widely used for improving surgery of malignant gliomas, it is not rare to detect 5-ALA-induced fluorescence of the ventricle wall in cases where the ventricle is opened during the operation (23,24). Although the relationships between 5-ALA-induced fluorescence and pathologic parameters in gliomas and peritumoral tissues have been extensively investigated, few studies have addressed the histopathological implications of 5-ALA-induced fluorescence in ventricular wall tissues that show no tumor involvement. In a series of 7 patients with periventricular GBMs, Hayashi et al (23) found that most ventricle wall tissues with unenhanced MR images that showed 5-ALA-induced fluorescence were positive for the presence of tumor cells, whereas all tissues showed disruption of ependymal cell layers of the ventricle wall. On the basis of this study, these authors suggested that tumor cells in the ventricular wall or environmental changes around the ventricle could lead to 5-ALA fluorescence of the ventricular wall. In a separate study, Tejada-Solís et al (24) concluded that ventricular wall fluorescence does not always indicate glioma cell invasion of the ventricular wall, based on the observation that 3 out of 8 (37.5%) cases that underwent selective biopsy of fluorescent ventricle wall exhibited an intact ependymal layer and no tumor cells. However, these studies were limited by the small number of cases and the lack of negative control specimens.

In the present study, we compared pathological findings of fluorescent ventricular walls with those of non-fluorescent ventricular walls using a relatively large number of samples in a homogeneous group of patients. To maintain the homogeneity of the patient population, we only included patients with newly diagnosed GBM. To eliminate the possibility of pseudo-positive 5-ALA-induced fluorescence in ventricular wall samples, we excluded patients diagnosed with recurring GBM that had undergone prior treatment, including surgery and radiotherapy, because areas showing infiltration of inflammatory cells associated with surgical and radiation interventions can also accumulate PpIX and show 5-ALA-induced fluorescence (25,26).

Compared with the previous studies described above, we obtained a relatively large number of samples from various domains of the lateral ventricle, including non-fluorescent ventricular wall samples as a negative control group. Most of the ventricular wall samples were excised during surgery as part of the planned margin of resection surrounding the GBM. Maximizing the extent of resection likely extends time to progression and increases survival, and supratotal resection of gliomas in non-eloquent regions can be beneficial for clinical outcomes (10,14,15,27–29). Accordingly, we tried to increase the extent of resection during neurosurgical procedures. Because we sought to achieve supratotal resection and additional lobectomy if possible during the removal of GBMs in non-eloquent areas, most ventricular wall samples were consequently included in the resected area. Therefore, we could safely harvest samples from ventricular walls. Additionally, for safety we collected intraoperative samples from ventricular walls in accordance with certain a prior criteria. First, we did not intentionally open the ventricle during the surgical procedure to check for fluorescence in the ventricular wall or to obtain ventricular wall samples. Because ventricular opening during surgical procedures can be a risk factor for postoperative hydrocephalus and leptomeningeal dissemination, which may potentially worsen the clinical condition and decrease survival (30–35), opening of the ventricular system was performed only in cases where it was needed for radical supratotal resection of the tumor. Second, ventricular wall samples in the safe area were collected carefully to prevent postoperative neurological sequelae. Functionally important areas, including the fornix, were excluded from ventricular wall sampling, because neurological sequelae can occur due to direct damage to the area. We also excluded certain areas with vascular structures, such as internal cerebral veins and thalamostriate veins, to prevent ischemic insult. Third, we tried to obtain a sufficient ventricular wall sample to clearly assess histopathology, while limiting the depth of tissue resection to <5 mm to prevent damage to structures under the surface.

In our series, ventricular wall fluorescence was detected in a majority (57.9%) of patients, but not in all patients. Considering the variety of locations in the lateral ventricle that showed fluorescence, the frequency of existence of ventricular wall fluorescence may be underestimated since we did not explore the entire lateral ventricle. Five of the 11 (45.5%) fluorescent ventricular wall samples contained glioma cells, whereas none of the 14 non-fluorescent samples showed infiltration of tumor cells (Fig. 1). In contrast to previous studies of the relationships between 5-ALA fluorescence and the pathology of tissue in GBM (9,36,37), the low positive predictive values (PPVs) and high negative predictive values (NPVs) in our series indicate that non-fluorescent ventricular walls under blue light should lack invading tumor cells. On the other hand, fluorescent ventricular walls could possibly reflect the presence of tumor cells, even though the fluorescence is not always related to the infiltration of tumor cells into the ventricular wall. Because the NPV depends greatly on the non-fluorescent tissue biopsy algorithm, the high NPV in our series may imply that ventricular wall samples remote from the tumor and not invaded by tumor cells were properly collected. The likelihood of not finding tumor cells should be higher if sampling sites are remote from the tumor than would be the case if they are close to the tumor. Our results suggest that, in cases where the ventricular wall is fluorescent, there is a chance of tumor cell infiltration into the ependymal spaces; accordingly, we recommend close follow-up with imaging studies and monitoring of the patient's clinical status after the surgical resection of GBM if 5-ALA-induced fluorescence of the ventricular wall is detected during surgery. Since fluorescence can be an indication of ependymal glioma cell invasion or leptomeningeal seeding in some cases, a biopsy should be performed to decide treatment options if feasibility and safety allow. However, postoperative radiotherapy covering the whole ventricle system based on the presence of ventricular wall fluorescence should be avoided because it may not always indicate pathological glioma cell invasion.

Six of the ventricular wall samples in our series were falsely fluorescing, showing no evidence of the presence of tumor cells. This raises questions regarding the meaning of false-positive fluorescence in the ventricular wall. False-positive fluorescence previously described by others was related to infiltration of inflammatory cells and reactive astrocytes, necrosis, prominent vasculature, or peritumoral edema (26,36,37). However, we did not observe such findings in our false-positive ventricular wall samples and found no difference between falsely fluorescing ventricular wall samples and non-fluorescent control samples from the standpoint of histopathology. The intraoperative 5-ALA-induced fluorescence in the 2 samples corresponding to low-grade glioma might be a false-positive finding because the vast majority of low-grade gliomas do not exhibit visible intraoperative fluorescence under a surgical microscope (38–40). A better understanding of 5-ALA-induced fluorescence in the ventricular wall will require further investigation, including a molecular characterization, of these falsely fluorescing samples. In a study with GBM patients by Piccirillo et al (41), the histologic analysis and genomic characterization revealed that the fluorescent subependymal zone contained tumor-initiating cells, however most samples from the fluorescent subependymal zone were truly fluorescing as they were included in the contrast enhancing lesion on MR images and pathologically confirmed to be involved by tumor. Studies with samples from 5-ALA-induced fluorescent ventricular wall tissues which show no tumor involvement are still lacking.

One limitation of this study is its subjective assessment of intraoperative 5-ALA-induced ventricular wall fluorescence. We adopted a trinary approach - non-visible, weak, or strong fluorescence - to assess fluorescence. However, this approach is subjective as fluorescence was estimated by a surgeon using only a surgical microscope with a specific filter. Overcoming this limitation would require quantitative or semiquantitative fluorescence measurements capable of objectively discriminating fluorescence intensity. These modalities, such as intraoperative spectrometry or confocal microscopy, have proven to be more sensitive for the determination of 5-ALA-induced fluorescence than the surgical microscope used here, although these modalities for quantitative determination of fluorescence still need to be validated for clinical use (37–39,42,43). Another limitation of this study is that infiltration of the tumor was judged based on enhancement on T1-weighted, contrast-enhanced MR images. GBMs have an infiltrative nature, and their presence is not always associated with a disrupted BBB; thus, contrast enhancement may not show invasive areas of GBM because it only depicts areas with a disruption in the BBB. Accordingly, active tumor tissue can exist beyond the area of contrast enhancement. T2-weighted and fluid-attenuated inversion recovery MR images may depict invasive areas of GBM; however, it is difficult to determine the extent of the non-enhancing component of the tumor using these approaches owing to peritumoral edema (44,45). On this account, there is the possibility of tumor invasion into ventricular wall samples in some cases, despite the fact that we obtained ventricular wall samples that were definitely remote from the enhanced lesion in MR images. Errors in the neuronavigation system related to image-to-patient registration error and intraoperative brain shift are also a limitation of this study. In all cases, the ventricle was opened and a considerable amount of cerebrospinal fluid was drained out during the surgery, which might subsequently degrade navigational accuracy over the course of a surgical procedure. We tried to compensate for such errors using intracranial anatomical landmarks; however, this method is subjective and could not eliminate the error completely.

In summary, our data suggest the possibility that glioma cells are present in ventricle walls exhibiting 5-ALA fluorescence despite the absence of tumor involvement in MR images. Ventricular walls lacking 5-ALA fluorescence and enhancement on MR images may be free of tumor, so a decision to resect non-fluorescent ventricular walls should be made carefully. Further investigations of non-tumor cells from tissues exhibiting 5-ALA fluorescence are needed to understand the nature of ventricular wall fluorescence.