Anatomical distribution of cancer stem cells between enhancing nodule and FLAIR hyperintensity in supratentorial glioblastoma: time to recalibrate the surgical target?

It is ge nerally accepted that glioblastoma (GBM) arise from cancer stem cells (CSC); however, there is little evidence on their anatomical distribution. We investigated the expression and distribution of SOX-2-positive and CD133-positive CSCs both in the enhancing nodule (EN) of GBM and in the FLAIR hyperintensity zones on a surgical, histopathological series of 33 GBMs. The inclusion criterion was the intraoperative sampling of different tumor regions individualized, thanks to neuronavigation and positivity to intraoperative fluorescence with the use of 5-aminolevulinic acid (5-ALA). Thirty-three patients (20 males and 13 females with a mean age at diagnosis of 56 years) met the inclusion criterion. A total of 109 histological samples were evaluated, 52 for ENs and 57 for FLAIR hyperintensity zone. Considering the quantitative distribution of levels of intensity of staining (IS), ES (extent score), and immunoreactivity score (IRS), no difference was found between ENs and FLAIR regions for both the SOX-2 biomarker (respectively, IS p = 0.851, ES p = 0.561, IRS p = 1.000) and the CD133 biomarker (IS p = 0.653, ES p = 0.409, IRS p = 0.881). This evidence suggests to recalibrate the target of surgery for FLAIRECTOMY and 5-ALA could improve the possibility to achieve this goal.


Introduction
One of the most debated neuro-oncological issues today is the use of an aggressive resection beyond enhancing nodule (EN) for glioblastoma (GBM); this question arose from the evidence that recurrences generally occur in peritumoral areas [1]. Cancer stem cells (CSCs) have been shown to play an initiating role in gliomagenesis, inducing angiogenesis, metastatic spread and resistance to conventional radiotherapic and chemotherapic treatments [2]. Few published researches have studied the anatomical distribution of CSCs in GBM. We would like to investigate the distribution of CSC both in the EN and in the peritumoral region of GBM corresponding to the FLAIR hyperintensity. These results allow to understand the basics of tumor progression and recurrence, suggesting the necessity to recalibrate the surgical and therapeutic target.
It is a matter of fact that the etiology of GBM remains largely unknown. It is generally accepted that GBM arise from CSC. Due to their biological features, especially the capability of self-renewal and migratory potential, CSCs might be the starting point of gliomagenesis [3]. It has been widely demonstrated that tumor cells are found beyond the central core of the tumor [4,5] but there is little evidence about the anatomical distribution of CSCs. Peng et al. recently described, in two autoptic cases of patients affected by GBM [6], a higher expression of CSC immunohistochemical markers at the infiltrating tumor edge with respect to other GBM areas. This evidence suggests that the edge of the tumor is the moving front for tumor progression and invasion and if this was confirmed in vivo and on a larger series, it would lay the foundations for remodulating local treatments on a different target. Based on this assumption, the eradication of CSCs may induce a stable disease-remission, having a potentially curative role on GBM. Many immunohistochemical markers of stem cell differentiation have been studied in different types of cancer [7], but the transcription factor sex-determining region Y-box 2 (SOX-2) and prominin-1, also known as cluster of differentiation 133 (CD133), are the most widely used for brain tumors [8,9]. In this paper, we investigated ex vivo the expression and distribution of SOX-2-positive and CD133-positive CSCs both in the central core of GBM (radiologically identified with the Enhancing Nodule -EN-on MRI) and in the peritumoral region (radiologically identified with the FLAIR hyperintensity zone beyond the EN) on a surgical and histopathological series of 33 GBM.

Materials and methods
We have prospectively collected tissue samples in different tumor areas (see below), and we have then retrospectively retrieved all cases with a histopathologically and molecularly proven diagnosis of WHO grade IV IDH-wild type GBM from the Pathology archive of the Department "G.F. Ingrassia" of the University of Catania, Italy, between January 2020 and December 2021. All these patients were surgically treated at the Neurosurgery Unit of the University-Hospital Policlinico "G. Rodolico-S. Marco", Catania, Italy. The study protocol was approved by the local ethics committee (CE 165/2015/PO) and all patients had signed a specific informed consent for the study before surgery. The informed consent form for study participation was also approved by the local ethics committee and all signed informed consents were archived in the study files.
The inclusion criterion was pathological proved GBM in patients with an age > 18 y.o. eligible for surgery; intraoperative sampling of different tumor regions (EN and FLAIR) identified thanks to neuronavigation and intraoperative fluorescence with the use of 5-aminolevulinic acid (5-ALA) (Fig. 1). Pre-operative Gad-T1 and FLAIR sequences were merged with [11C]-methionine Positron Emission Tomography (11[C]-MET PET/CT) on StealthStation S8 (Medtronic Navigation, Louisville, CO, USA) and we took different tissue samples targeting the EN and FLAIR hyperintense areas beyond the tumoral central core identified with Gad-T1 and [11C]-methionine uptake. Only tumor samples with clear 5-ALA fluorescence were selected for the analyses of the different tumor zones. The surgical protocol for resectable tumors was previously detailed [10,11] and the intraoperative fluorescence was evaluated using a surgical microscope with 400 nm filter (Zeiss; Kinevo®) by 3 trained neurosurgeons (G.B., F.C., R.A.). The unresectable tumors (deep/eloquent locations, low KPS, etc.) were biopsied with frameless neuronavigated system (Navigus, Medtronic) and every sample was also evaluated for 5-ALA positivity (Fig. 2).
Of the 58 patients, surgically treated for brain tumor, 33 (20 males and 13 females with a mean age at diagnosis of 56 years) met the inclusion criteria and were included in the study. Twenty-two (66.6%) underwent resection while 11 (33,3%) underwent biopsy. A total of 109 histological samples were evaluated, 52 for EN and 57 for FLAIR hyperintensity zone. The clinico-pathological features of the GBM cases are summarized in Table 1.
Hematoxylin and eosin (H&E)-stained slides and additional slides stained with numerous immunohistochemical antibodies were available for each case. All the H&E slides were reviewed by three pathologists (G.B., R.C., and G.M.) to confirm the histopathological diagnosis and the correct sampling of the EN and FLAIR regions.
Immunohistochemical analyses were performed on both EN and FLAIR samples, as previously described [12,13]. Deparaffinized sections were incubated with rabbit polyclonal anti-SOX-2 (ab97959; 1 µg/ml; Abcam, Cambridge, UK) and rabbit monoclonal anti-CD133 (ab222782; 0.5 µg/ml; Abcam, Cambridge, UK) for 15 and 30 min at room temperature, respectively. A semiquantitative assessment of the immunohistochemical staining was performed by three pathologists (G.B., R.C., and G.M.) with no access to the clinico-pathological data of the patients, as previously described [14,15]. SOX-2 and CD133 were considered positive if brown chromogen was observed within the tumor cell nuclei and cytoplasm, respectively.
The immunoreactivity score (IRS), originating from the IS multiplied by the ES, was obtained: the immunohistochemical expressions of SOX-2 and CD133 were considered low if the IRS was ≤ 6, and high if the IRS > 6.

Statistical analyses
EN and FLAIR levels of IS, ES, and IRS for each stem cell marker (SOX-2 and CD133) are described as Median with 95% confidence interval for quantitative data and with frequency and percentage for categorical data. Difference within subjects between EN and FLAIR regions were computed with Wilcoxon signed-rank test for quantitative values, while categorical differences between EN and FLAIR levels were computed with Fisher's exact test. For all analyses, a p-value < 0.05 was considered statistically significant. Analyses were performed using the statistical software R, version 4.0.3.

Results
All samples were fluorescent according to the inclusion criteria, the EN revealed a lava-like fluorescence while FLAIR samples had a faint fluorescence. Immunohistochemical results are summarized in Tables 2, and 3 and Fig. 3. All histological samples from the EN region showed a proliferation of tumor cells with ovoidal, to elongated, morphology, moderate to severe nuclear atypia, brisk mitotic activity, foci of necrosis and/or microvascular proliferation; all these features were consistent with the diagnosis of WHO grade IV GBM [16,17]. Similarly, histological samples from FLAIR regions exhibited fragments of brain white matter tissue with focal to diffuse infiltration of GBM cells; neither necrosis nor microvascular proliferation were histologically found within these samples.
IS, ES, IRS of each patient are summarized in Supplementary Table S1.

Discussion
One of the most debated neuro-oncological issues today is the use of an aggressive resection beyond EN for GBM; this question arose from the evidence that recurrences generally occur in peritumoral areas [18][19][20]. It has been demonstrated that cellular composition and molecular signatures of the GBM core compared with infiltrative margins are different and many papers described the role of cross-talk between tumoral cells and the tumoral microenvironment in the regulation of tumor growth and progression [21][22][23].
The first evidence of the key role of CSCs in tumor maintenance, growth and recurrence originated from studies conducted in hematopoietic and solid neoplasms [7]. CSCs are a subpopulation of cells with several capabilities,  including self-renewal, multi-cell lineage differentiation and induction of resistance to conventional therapies. As demonstrated in other human neoplasms, CSCs were also isolated in GBMs in 2002 [24,25]; although their origin is still largely unknown, these cells have been shown to play an initiating role in gliomagenesis, inducing angiogenesis, metastatic spread and resistance to conventional radiotherapic and chemotherapic treatments [8]. Furthermore, CSCs seem to have an increased DNA repair mechanism, through which they tend to easily overcome the cell stress induced by anti-cancer therapy, resulting in shorter resistance and disease recurrence times in GBM patients [26]. Due to the critical role played by CSCs in gliomagenesis, several studies have been carried out to better understand their genetic and immunohistochemical features, as well as their anatomic distribution, in patients affected by malignant gliomas. Based on these assumptions, it has been suggested that, eradicating CSCs, might affect stable, long-lasting remission and potentially treat cancer [27,28].
The detection of proteins that are differentially expressed by CSCs and targetable by immunohistochemistry, represents the best and easiest way to study the expression and distribution of this cell subpopulation [8]. SOX-2 and CD133 are the most used immunomarkers of CSCs in GBM.
The SOX-2 gene, located on chromosome 3q26.3-q27, is part of the SOX family of transcription factors [29]. It encodes a protein that is highly expressed during the development of the central nervous system and downregulated when neural cells start to differentiate; SOX-2 is involved in the regulation of several genes that play key roles in neurogenesis and gliogenesis [9]. In the adult unaffected brain, SOX-2 expression is almost absent and restricted to proliferating cells, neural stem cells and progenitor cells. In brain tumors, SOX-2 expression has been found in glial tumors, such as astrocytomas, oligodendrogliomas and ependymomas, but not in neuronal neoplasms; in particular, a higher histological grade is associated with a higher expression of this protein [9,30].
We first compared the concentration of CSCs in the EN and FLAIR hyperintensity zones using the aid of a multimodal intraoperative imaging approach (especially neuronavigation and a metabolic intraoperative fluorescence tracer) to select the correct samples [35]. We found that 5-ALA correlated with the presence of CSCs. The peritumoral areas, corresponding to the FLAIR hyperintense zone, are populated by CSCs in the absence of necrosis and microvascular proliferation. There are no differences in CSC concentration between EN and FLAIR, which is in agreement with our hypothesis of GBM growth parabola that we previously described with a radiological series. This evidence indeed seems to suggest that gliomatogenesis starts in some specific areas of the brain thanks to a favorable microenvironment. The neural stem cell undergoes transformation in CSC and this population replicates in the corresponding FLAIR areas with a velocity of mean diametric expansion of about 40 mm/years. The continuous accumulation of mutations determines the acquisition of a more malignant property in some areas that acquire a velocity of mean diametric expansion of about 45 mm/year. This is radiologically visible with the appearance of the EN that appears more than 1 year after the first molecular events. Therefore, the discrepancy between the metabolic needs and supply, create the central core necrosis with a deceleration of EN velocity of mean diametric expansion according to the Gompertz curve [36,37]. In this way, if we surgically treat the EN, the presence of CSCs within the FLAIR zone starts the processes that will lead to the genesis of another EN. This is in agreement with the median free progression survival of 5 months without radiotherapy [38]. At this time there is no evidence about the possibilities to individualize the CSCs with radiomic technologies. Their surgical mapping could open the way to planning local treatment (surgery and radiotherapy) on the new imaging showing the real therapeutical target. Another future perspective could be to find an intraoperative tool (intraoperative fluorescent dye?) to visualize CSCs as well as guiding and maximizing resection.

Conclusion
We have confirmed that 5-ALA can visualize a tumor beyond the classical margins of the EN, and we have histologically proved that there are no differences between the concentration of CSCs in FLAIR and EN of GBM. This evidence suggests to recalibrate the target of surgery and radiotherapy on the FLAIR hyperintensity areas in order to achieve a better local control of the disease. In this way, 5-ALA could improve the possibility of achieving a FLAIRECTOMY and, if effected within the functional boundaries, could improve patient outcome [39]. Neuronavigation could have some degree of inaccuracy due to brain shift. Therefore, the sampling of FLAIR tissue versus EN could be theoretically inaccurate. However, the neuronavigation update with intraoperative CT scan [10,11], surgical experience of trained neurosurgeons on distinguishing different tissues, the use of 5-ALA besides the number of patients who underwent frameless neuronavigated needle biopsy mitigates this limitation. CD 133 and SOX-2 alone could be disputable as stemness markers but there is a robust literature about the trustworthiness of each marker, and we have tested both to enhance their reliability.