Introduction
Spinal Cord Gliomas (SCGs) account for 4–8% of central nervous system (CNS) tumors, where high-grade lesions have a poor prognosis [
1]. Most commonly, gross total resection is not an option in such lesions due to a lack of resectable margins and the risk of worsening morbidity [
2]. Importantly, SCG is considered an orphan disease that is overlooked in comparison to its intracranial counterpart whereby human specimens are limited owing to this relative rarity. Consequently, there is a paucity of literature on SCG pathogenesis, intratumoral heterogeneity, and tumor microenvironment despite some recent small retrospective studies [
3‐
6].
While rodent models have been utilized in the preclinical setting for the study of SCG, there are known limitations especially with regard to anatomic constraints, drug delivery, and surgical translatability [
7,
8]. In the past, efforts to investigate surgical techniques, drug delivery, and de novo tumor development have depended on spontaneous canine gliomas [
9] which are rare and present with ethical and logistical issues. Moreover, SCG is extremely rare in dogs. As such, development of new therapeutic strategies for SCG has been hindered by the lack of a relevant preclinical animal model for understanding the delivery, therapeutic response, efficacy, toxicity, and tumor microenvironment of SCG.
Our group has developed a minipig model of high-grade SCG using lentiviral gene transfer of oncogenic transgenes including platelet derived growth factor beta (PDGF-B), HRAS-G12V, and small hairpin-loop RNA targeting P53 (shRNA-P53) [
10] as well a vetted methodology for inducing glioma in prior rodent studies[
11‐
13] and validated this approach in the rat spinal cord [
14]. We employed this approach given involvement of the RTK/RAS/PI3K and p14/CDKN2A and P53 pathways in a large majority of high-grade gliomas [
10,
11]. This model has histopathologic and radiologic features consistent with human disease processes in a highly penetrant manner [
10]. The advantage of utilizing a large animal model is to facilitate translation given the similarity of human's metabolic, oxidative, genetic, and immunologic features to that of the swine [
15]. However, the tumor microenvironment and heterogeneity of our model is unknown. In an effort to further characterize and advance preclinical modeling, the present study reports key microenvironmental features of the model, including oxidative stress, vascular markers, and intratumoral heterogeneity between the tumor core and edge.
Among these features is oxidative stress, a well-known contributor to glioma pathogenesis. It is exerted by reactive oxygen species (ROS) that accumulate due to an imbalance between ROS generation and elimination, which can consequently promote glioma growth and progression [
16]. In the tumor microenvironment, an interplay between tumor cells and surrounding parenchyma can enable tumor progression by sustaining malignant cells and providing adaptation to restricted oxygen supply [
17]. In addition to adaptation to a hypoxic environment, ROS generation can lead to lipid, protein, and DNA damage that can further contribute to glioma pathogenesis. Furthermore, it is well documented that the tumor core in high grade supratentorial gliomas have a hypoxic core and over-expression and stabilization of hypoxia inducible factor-1 (HIF-1α) [
18,
19]. Downstream of HIF-1α are hypoxia responsive genes, including glucose transporter 1 (GLUT-1) and CA-IX (Carbonic Anhydrase 9) which are expressed in cells undergoing anaerobic metabolism and demonstrated to be overexpressed in glioblastoma cells [
20]. In addition, this hypoxic microenvironment leads to a variety of downstream effects including a pro-angiogenic environment, metabolic reprogramming, and local immunosuppression [
21]. Consequently, understanding whether this phenomenon is present in this model is crucial to its translational relevance.
The differences in oxidative stress, region specific factors and subsequent molecular features may be found depending on the geographic location within the tumor including well-reported differences in proliferation [
22], inflammatory cytokines [
23], and microvascular structure [
24,
25]. Certainly, molecular subtypes have been identified from a transcriptomic standpoint in supratentorial high-grade glioma [
26,
27] and correlated with characteristic immunohistochemical profiles in regions of hypoxia and hyper-proliferation amongst others [
28]. Given these studies in clinical and preclinical models in the brain, it stands to reason that modeling SCG requires robust characterization of factors. This is especially crucial from a clinical standpoint in the spinal cord, whereby the unresectable edge of the tumor is a major cause for recurrence.
In our present study, we compare and contrast features of the tumor core and edge to evaluate differences in redox homeostasis, hypoxic markers, and validate the aggressive proliferation of tumor cells. We further described recapitulation of these features upon evaluation in porcine SCG-derived culture models both in vitro and on xenotransplantation. These data provide the basis for future work employing this large animal model as a platform for translational surgical studies, pharmacotherapeutic studies targeting redox homeostasis, and comparative studies with patient derived samples.
Discussion
Our present study corroborated several key features of our previously described minipig model of high-grade SCG. From a practical standpoint, we continued to report a highly penetrant model system with tumor induction following vector administration. We further described and confirmed the expression of relevant transgenes on immunofluorescence with no major differences in the tumor core compared to the tumor edge. Of note, we observed significant elevation in proliferation in the SCG tumor edge. This finding was recapitulated on establishing tissue culture from biopsies of the core and edge, as well as on xenotransplantation into nude mice. This highlights a point of clinical interest, - it is often tumor cells at the periphery which have higher proliferation rates that are surgically inaccessible given the infiltrative nature of eloquent CNS regions like SCG. The inability to perform a gross total resection given the infiltrative nature of SCGs represents one of main barriers to development of an effective treatment strategies and is the key reason for our interest in the heterogeneity between cells in the tumor core and edge.
The findings in our model were in agreement with the literature in describing increased proliferation, increased vascularity, and oxidative stress at the edge in supratentorial glioma as well as increased hypoxic markers in the tumor core [
20,
37‐
39]. It has been widely reported that the tumor edge in supratentorial high-grade glioma exhibits a more aggressive, proliferative, and invasive phenotype [
22,
39]. We examined vascular markers and found that edge regions were enriched in endothelial cells, pericytes, and overall vasculature consistent with prior reports [
24]. In non-disease states, pericytes have a multitude of roles, including maintenance of the blood–brain-barrier through interaction with astrocytes, development and maturation of vasculature, paracrine signaling, as well as exhibiting stem cell properties [
40‐
43]. It has also been reported that pericytes directly promote glioma progression in in vitro and in vivo models [
44]. Moreover, microvascular proliferation is a well-known feature of high-grade gliomas. In order to explain this finding, the increased vascularity at the tumor edge in this model might simply be due to the innate nature of microvascular proliferation secondary to rising metabolic needs locally in the more mitotically active edge region. This proliferation is conventionally attributed to the fact that hypoxia was not necessarily restricted solely to the tumor core [
37]. This concept is supported by the fact that while the tumor edge had, as expected, a significantly lower degree of staining for hypoxic markers (HIF-1a, CA-IX, GLUT-1), up to 30% of the cells in the tumor edge still stained strongly for these markers (Fig.
4G, Additional file
1: Figure S1). As such, this region may represent a site of transitional hypoxia in the setting of rapid tumor proliferation that should be examined in future studies including identifying specific cell populations and highlighting potential translational targets [
37]. This is of particular interest given that since pericytes and endothelial cells both express PDGF receptors, it is also possible that our minipig model with the PDGFB transgene have been driven to elevated levels of pericytes and microvascular proliferation [
24].
In examining select cytokines in the core and edge regions, we observed an elevation in TNF-α, IL-1β, and IL-6 in the edge compared to the tumor core. IL-6 is a cytokine that is known to trigger JAK-STAT3 signaling in glioblastoma, gliomagenesis, MGMT methylation, and is correlated with tumor grade and overall patient survival [
45‐
47]. TNF-α and IL-1β are pro-inflammatory cytokines that are found in supratentorial high-grade gliomas and are involved in neuroinflammation, gliomagenesis, and development of radio-resistance [
23,
48]. Overall, the elevated presence of these cytokines throughout the tumor was consistent with reports on supratentorial high-grade gliomas. It was of interest with utilizing this model as a platform for radio-resistance, drug resistance, or aggressive phenotype in this particularly challenging region.
For evaluation of oxidative stress, we examined markers of oxidative DNA damage (8-OH-dg), protein oxidative damage (protein carbonyl content), as well as non-enzymatic antioxidant defense through ROS scavenging (GSH/GSSG ratio) [
49]. We found that the tumor edge has a significantly elevated level of markers of ROS, including a high 8-OH-dG and low GSH/GSSG ratio, as well as oxidative damage with an elevated protein carbonyl content in Edge tumor lysates. This supports the notion of elevated oxidative stress in the setting of a highly proliferative tumor edge. On histologic evaluation, we further observed an increase in NQO1 and 4-HNE staining patterns at the tumor edge. Next steps toward further examining molecular characteristics of core and edge regions of this model will include rigorous evaluation of hypoxic markers including , PDK1 and others. In addition to redox mechanisms contribution to glioma pathophysiology [
17], the difference in oxidative stress between the core and edge in this model is of particular interest given that there are a variety of druggable targets in the realm of oxidative stress [
16,
29,
50‐
53].
The development of advanced pre-clinical disease modeling has increasingly employed pigs as biomedical models given their increased anatomic, genetic, and immunologic similarity to the human [
54]. Porcine models, including minipigs, have successfully been utilized to model numerous pathologies including colorectal cancer, osteosarcoma, cystic fibrosis, muscular dystrophy, as well as numerous others [
55,
56]. Notably, the use of TALEN mediated gene editing was utilized to generate a minipig model of neurofibromatosis type 1 that better recapitulates features and natural history of the human disease [
10,
57]. Specific to glioma models, as of 2020 there are only three published models including two orthotopic xenografts [
58,
59] using commercially available cell lines, as well as our presently reported spinal cord model using vector driven gene transfer [
60]. Further work to develop enhanced porcine models of glioma may create a platform better suited for pre-clinical translational efforts.
Conclusions
Overall, the data presented in this study demonstrated multiple distinct microenvironmental features within the pig model of high-grade SCG. Future work on the development of this model system could explore each of these individual components, including but not limited to profiling the inflammatory microenvironment, transcriptomic characterization, or further elucidating the cause for findings of intra-tumoral heterogeneity between the core, edge, perivascular niche, or other regions. However, at present this model represents a step forward for facilitating surgical translation, as well as a potential space for evaluating proof-of-concept for surgical resection followed by select druggable targets that could prove more efficient in treating the infiltrative tumor edge which is surgically inaccessible. It would be advantageous to also initially employ culture and xenograft models derived from the pig SCG with high-throughput screening methodologies to identify existing FDA-approved or novel candidates that may be able to better target the tumor edge cells.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.