Introduction
Life expectancy of glioblastoma patients (e.g. beyond average 14 to 16 months since primary diagnosis) remains very low due to tumor aggressiveness and resistance to therapy associated with high recurrence rate [
1]. Still, there are glioblastoma patients who live longer than 3 years and are hence classified as long-term survivors (LTS). Many efforts have been made to uncover concrete reasons that would define this exceptional group, however, without clear outcomes. Although select clinical characteristics as well as molecular features have been found useful in stratifying patients and predicting their fate, no single biomarker, subclinical signature or even genetic or epigenetic profile proved to be universally valid for glioblastoma LTS [
2]. It is thus of paramount importance to identify concrete features or factors whose activities ultimately define behavior of this tumor and predict survival of patients. In this context, one of the promising and currently studied signature features of glioblastoma is hypoxia [
3].
Glioblastoma represents a type of malignancy with very pronounced hypoxia, i.e., low oxygen tension, which is histopathologically characterized by several hypoxia-related features such as necrotic foci, pseudopalisades, and microvascular hyperplasia. These features are thought to originate from rapid proliferation of tumor cells that outgrow their blood supply, resulting in inadequate oxygen delivery. Additionally, highly vascularized malignant gliomas develop the abnormal vasculature often lacking in the structure and functionality necessary for efficient oxygen transport which further contributes to hypoxia within the tumor [
4,
5].
Hypoxia in glioblastoma has pleiotropic effects influencing many biological aspects of its existence. Firstly, it triggers a complex intracellular signaling milieu comprising activities of hypoxia-inducible factors (HIFs), phosphoinositide 3-kinase-AKT kinase/mammalian target of rapamycin (PI3K-AKT/mTOR), mitogen activated protein kinases (MAPK), nuclear factor kappa B (NF-κB) and signal transducer and activator of transcription 3 (STAT3) pathways that promote tumor cells survival, aggressiveness, invasion and resistance to therapy [
6,
7]. The key role here is attributed to HIFs, with HIF-1α and HIF-2α being the major HIF isoforms mediating the positive HIF-dependent signaling [
8]. Both of them participate in many processes related with glioblastoma, such as angiogenesis, hypoxia-mediated apoptosis, genetic alterations or immune evasion [
9]. Secondly, hypoxia can also induce genetic changes via HIF-1α in glioma cells, leading to the activation of genes involved in tumor growth, invasion, and resistance to therapy [
10,
11]. Moreover, hypoxia within glioblastoma can impair the immune response against tumor cells including the recruitment of immune-suppressive cells, such as regulatory T cells and myeloid-derived suppressor cells, while also reducing the activity of cytotoxic immune cells [
12]. Thirdly, hypoxia-induced changes in the tumor microenvironment can render gliomas resistant to various treatments, including radiation therapy and certain chemotherapeutic agents [
13] which correlate with predicted survival of patients [
14].
Concerning the prominent role of hypoxia in glioblastoma biology, our experimental study has been designed to investigate the expression of select hypoxia and epithelial-to-mesenchymal (EMT)-related markers in a small group of LTS glioblastoms patients repeatedly treated in our university hospital. Specifically, we aimed to study and compare expressions of these markers in a time series of resected tumor samples and cell cultures derived from them. The main rationale of such an approach was based on the recognized role of hypoxia in pathogenesis of glioblastoma as well as the missing evidence concerning the involvement of hypoxia in individually recurring tumors and their behavior.
Materials and methods
Clinical samples
Clinical tumor samples were obtained from eight patients who repeatedly underwent surgery for malignant glioma at University Hospital Hradec Kralove. The study was approved by local ethics committee (Reference No. 201,906 S26P) and patients gave their written consent. All samples included in the study differed in tumor stage and grade and represent a differing number of repeated resections. The material sampled (amount and quality) for RNA analysis of hypoxia markers as well as cell culture derivation varied based on the particular specimen.
The cell derivation procedure and further manipulation with cell cultures was described previously [
15]. In this study, three glioma cell primocultures were prepared in sufficient quality and quantity required for further analyses.
Clinical data pertinent to each case were retrospectively or prospectively reviewed including surgical reports, radiological images, histological parameters and therapeutic protocol. Tumor size (before surgery) was evaluated by neuroradiologist and radiation therapist following Magnetic Resonance Imaging (MRI). It was performed by 3 Tesla magnet using a cross-sectional 2D method with the product of the largest perpendicular diameters on T1-weighted contrast-enhanced. Other parameters such as tumor location, survival, response to radiotherapy and TMZ treatment, electrocortigraphy (ECOG), among others were evaluated and correlated too.
Immunohistochemistry and tissue microarray (TMA) construction
Paraffin-embedded specimens, cryopreserved samples and native (non-fixed) tumor tissue samples were retrieved from and handled by Fingerland Department of Pathology. In case of native tumor samples, a minor part was cryopreserved and most of tumor tissue was fixed in 4% buffered formalin solution and paraffin embedded. 1 μm thick sections were cut and stained with hematoxylin-eosin to confirm the diagnosis and select appropriate areas for additional analyses. A TMA was constructed from tumor samples, using TMA Master II system (3DHISTECH Ltd., Budapest, Hungary). For each case, 2-mm-thick sections were cut and routine H&E staining and immunohistochemical studies were performed. The list of antibodies and relevant details of immunohistochemical protocols are summarized in Supplementary Table
1. Heat induced epitope retrieval (HIER) was used for antigen retrieval, employing different pH according to the antibody. For EGFR, proteolytic pretreatment step was used. The sections staining was carried out on Benchmark Ultra stainer manufactured (Ventana/Roche, Tucson, AZ, USA) using either Ventana ultraView Universal DAB detection kit or Ventana OptiView DAB IHC detection kit: both methods use avidin-biotin complex method with horseradish peroxidase as an enzyme and DAB (3,3’-diaminobenzidine) as chromogen. Agilent/Dako Autostainer 48 or Dako Omnis (Agilent, Santa Clara, CA, USA), with EnVision Flex detection kit was used for IDH1 R132H, CD57, ZEB2, TWIST, and HIF1a. For HIF1a and HIF1b, whole sections (WS) from tissue core donor blocks were used. The slides were subsequently counterstained with hematoxylin.
Mutation analysis
DNA from glioma cells was extracted using the commercial DNA Sample Preparation Kit (Roche, Basel, Switzerland). Mutation analysis was performed by multi parallel sequencing (NGS) using the hybrid-capture-based target enrichment. A custom KAPA HyperChoice MAX Library (Roche) for enrichment of the coding and 30 bp upstream and downstream overlaps of selected panel of genes (TP53, EGFR, IDH1, IDH2, PTEN, PIK3CA) was used. Paired-end cluster generation and sequencing was performed by NGS system Illumina MiniSeq. Sequencing data analysis were performed by NextGENe software (Softgenetics) and Varsome Clinical Platform. MGMT methylation analysis was performed on DNA in FFPE using DNA Sample Preparation Kit (Roche). Bisulfite conversion of isolated DNA was performed using the EZ DNA Methylation-Gold Kit (Zymo Research). Detection and quantification of the hypermethylation status of the O (6)-methylguanine-DNA methyltransferase (MGMT) promoter was performed by the methylation-specific real-time PCR method using the CE-IVD marked geneMAP MGMT Methylation Analysis Kit (GenmarkSalgik).
Athymic nude mouse model
Female Foxn1-nu athymic immunodeficient mice weighing 27–30 g were purchased from Velaz, Czech Republic. They were given a standard sterilized diet and water ad libitum. Cells suspensions intended for implantations were obtained from cultivated cells. Prepared 100 µl of cell suspension (1mil. cells/application) was injected subcutaneously into each 6 weeks old Foxn-1nu female mice on the right and left side of the back. The administration of treatment (TMZ) began two weeks after implantation (daily from day 15th to day 28th) as based on the approved project plan (ethical committee, project number: MSMT- 18,525/2021-3 attached as supplementary Fig.
4). TMZ was dosed orally (maximum volume 100 µl) at the therapeutic range of 0.9 mg/kg. Each experimental group consisted of 2–3 mice. The size of the tumors and health condition of mice were periodically checked. On the day of last application of TMZ, twenty minutes after treatment, mice were anesthetized with isoflurane, sacrificed and tumors were weighed and preserved until further analysis (in formalin at room temperature for IHC analysis, in Trisol at -80 °C for RT-PCR analysis and in lysis buffer at -20 °C for western blot analysis). Heart, liver, brain and plasma were also collected and stored at -80 °C before MS analysis.
Cell line
Human malignant glioma cell line U87MG was purchased from ATCC (LGC Standards, Poland). Fresh cells from frozen batch were used for every set of experiments (lasted 3–9 weeks). Cultures were grown in EMEM supplemented with 10% FBS and 0.5% penicillin/streptomycin and cells were maintained in incubators with a humidified atmosphere containing 5% CO2 at 37 °C. The absence of mycoplasma contamination was periodically checked.
Crispr/Cas STAT3 knockout cell model
Glioma cells U87MG grown to 50–70% confluence were transfected with transfection mixture (gRNA vectors in Opti-MEM I, the donor DNA and Turbofectin 8.0 - the ratios of 3:1 for Turbofectin: DNA) as based on manufacturer’s protocol (STAT3 Human Gene Knockout Kit (CRISPR), CAT#: KN204922, Origene). Cultures were split every 3 days (2–4 times in total) to dilute out cells containing non-integrated donor DNA. Then, puromycin selection was performed; cells of split P5 were grown directly in the puromycin containing complete media (the range of puromycin concentrations was 1 to 10 µg/ml) and medium was changed every 2–3 days. The puromycin resistant cells (approx. P10 split cells) were analyzed for genome editing, WB was used to measure gene knockdown and genomic PCR was carried out to verify the integration of the functional cassette. With primer pair of 5 F and 3R, both alleles of donor inserted and non-edited/indel were amplified. Thus established cell line was labeled as U87MG STAT3 KO.
Proliferation
The inhibitory effect of chemotherapeutic TMZ at various concentrations on viability of U87MG and its STAT3 KO variant as well as viability of primocultures was evaluated by WST-1 assay for 48 h. Principle of this colorimetric test is based on the cleavage of the tetrazolium salt to colored formazan by mitochondrial dehydrogenases in viable cells. At the end of tested interval, cells were rinsed with PBS and WST-1 solution (diluted according to manufacturer’s recommendations) was added to each well for further 2 h. Quantification of mitochondrial enzymes activity was carried out at 450 nm with 650 nm of reference wavelength by Tecan Infinite M200 spectrophotometer (Tecan, Switzerland).
Total RNA was isolated from glioma primocultures and both variants of U87MG– STAT3 expressed and STAT3 KO variant - using Direct-zol RNA MiniPrep kit according to the manufacturer’s instructions (ZymoResearch, Irvine, CA, USA). Tumors from cryopreserved tissues and excised from mice (both variants of U87MG tumors) were homogenized using Tissue lyzer (2 cycles − 25 vibration/s; 4 °C; 1 min; Qiagen, USA) in TriReagent. RNA concentration and its purity were measured using NanoDrop 2000 (Thermo Fisher Scientific). All samples had absorption ratio A260/A280 greater than 1.8. RNA integrity number (RIN) was determined using Agilent 2100 Bioanalyzer and cell line samples with RIN higher than 9.0, resp. tissue samples with RIN higher than 8.0 were used for further analysis. First strand cDNA synthesis and qPCR analysis were performed in LightCycler® 96 Instrument (Roche Life Science) as described in [
16]. Primers were designed manually, and their sequences are attached as a supplementary file (Supplementary Tables
2 - primer sequences). Calculations were based on delta-delta Cq method [
17] and expressed as fold change of the treated groups relative to the control. Beta-2-microglobulin (B2M) or TATA box binding protein (TBP) were used as reference genes for mRNA analysis.
LC-MS analysis
Frozen tissues were homogenized in 4 volumes of cold PBS (w/v) using Fastprep-24 5G sample disruption instrument. Thawed plasma or homogenized tissues in the volume of 100 µl, was mixed with the same volume of methanol and acetonitrile, vortexed for 15 min and centrifuged at 14,000 g for 3 min. Supernatant was then filtered through 0.22 μm PTFE syringe filter into the vial and measured.
Detection of TMZ and its metabolites AIC and MTIC was performed on the Agilent 1290 Infinity II UHPLC system coupled to the Agilent 6470 QqQ mass spectrometer. Chromatographic conditions were maintained at gradient elution of 0.4 ml/min by 0.1% formic acid in water and methanol (0-0.5 95:5, 0.5-3.0 gradient to 5:95, 3.0–4.0 5:95, 4.0–5.0 95:5), thermostated autosampler set to 15 °C and column thermostat equipped with the Zorbax Eclipse plus RRHD C18 2.1 × 50 mm, 1.8 μm (PN 959757-902) column kept to 30 °C. MS source parameters were set to the following: drying gas 200 °C at 2 l/min, sheath gas 400 °C at 12 l/min, nebulizer pressure 25 psi, capillary voltage 2500 V and nozzle voltage 0 V. TMZ transitions of [M + H] + ions m/z were detected with setting of dwell time 50 ms, cell accelerator 4 V, fragmentor 88 V for 195→138 and 55 (collision energy– CE 8 and 28 V). MTIC transitions of [M + H] + ions m/z were detected with setting of dwell time 50 ms, cell accelerator 4 V, fragmentor 88 V for 169→109 and 43 (CE 20 and 40 V). AIC transitions of [M + H] + ions m/z were detected with setting of dwell time 50 ms, cell accelerator 4 V, fragmentor 88 V for 127→110, 82 and 55 (CE 20 V, 20 V and 40 V).
Morphology
The TMA was evaluated for presence of the tumor tissue. Only TMA cases with at least one representative core were included for further analyses. The immunohistochemistry results were first digitalised using Leica Aperio AT2 slide scanner (Leica Biosystems, Buffalo Grove, IL, USA) and then evaluated with Aperio ImageScope software (Leica Biosystems, Buffalo Grove, IL, USA). The assessment of immunohistochemistry was performed by an experienced neuropathologist (JS). Percentage of positive cells and most prevalent staining intensity (1– weak, 2– moderate, 3– strong) was noted. Average percentage of positive cells and modified H score (mHS = percentage * intensity) was used for the analysis as reported previously [
18]. Whole sections were analyzed using the same approach.
Fluorescence microscopy
U87MG and U87MG-STAT-KO glioblastoma cells were fixed with 2% paraformaldehyde (20 min, 25 °C), rinsed with PBS permeabilized, and blocked with 1% Triton X and 5% BSA in PBS (30 min, room temperature). The cells were incubated with a primary antibody against STAT3 (D3Z2G® Rabbit mAb, Cell Signaling) at 4 °C overnight. Then, the cells were washed three times with cold PBS (5 min, 25 °C) and were incubated for additional 1 h (room temperature) with Alexa Fluor 488-labelled anti-rabbit antibody. Thereafter, the cells were rinsed three times with PBS and labelled with DAPI (10 µg/mL). The specimens were mounted into the Prolong Gold mounting medium (Invitrogen-Molecular Probes, Inc., Carlsbad, California, CA, USA) and examined using fluorescence microscopy technique (Nikon Eclipse E 400 (Nikon Corporation, Kanagawa, Japan)). The results were analyzed using LUCIA DI Image Analysis System LIM 4.2 (Laboratory Imaging Ltd., Prague, Czech Republic). All the samples were tested in duplicates in three independent experiments.
Statistics
Data in all tests used are expressed as an average ± SD from at least two experiments. The concentration of chemotherapeutic TMZ causing a 50% decrease of cell viability (IC50 value) was determined by Graph Pad Prism 7.0. Statistical analysis of the data from RT-PCR analysis, proliferation WST-1 assay and LC-MS analysis was carried out using TWO-WAY analysis of variance (ANOVA) followed by Sidak’s multiple comparison test significant at level of p˂0.05. Other data analysis was done with GraphPad Prism 7.0.
Discussion
Despite intensive investigations into the nature of glioblastoma which resulted in identification of many important mechanisms and driving forces, we are still lacking complete understanding of pathogenesis of this tumor. Moreover, and most importantly, crucial differences between standard patients and LTS patients are missing, to the large degree due to uncovered leads and incomplete evidence. Thus, the present study aimed to find out whether expression of select hypoxia markers and related signals would better characterize glioblastoma in a cohort of LTS patients.
Our initial results demonstrated mutational heterogeneity in select analyzed glioblastoma-related genes among included patients and in some cases in their serial tumor resections too. It concerned, among others, IDH1 status where samples of two of our patients carried IDH1mt variant, which has been previously suggested as the potential explanation of LTS in glioblastoma patients [
19]. Still, this hypothesis could not have been proven in other studies [
20]. Moreover, in our case the mentioned patients with IDH1mt did not survive significantly longer than others with IDH1wt, indicating that the present IDH1 variant alone is very unlikely a main contributor to LTS. These outcomes generally confirm and correspond to other studies which reported a highly variable and individual genetic terrain among LTS glioblastoma patients [
21,
22]. Interestingly, chromosomal instability, the chief reason of genetic heterogeneity in tumors, is pervasive in high grade gliomas. Still, it is not considered to be a major contributor to the phenomenon of LTS in glioblastoma too although it has been proposed that specific chromosomal instability signatures might have some prognostic value in this type of malignancy [
23].
Our subsequent evaluation of expression of targeted hypoxia, EMT and glioma-related markers revealed several opposing trends in individual analyzed marker groups. Still, a significantly increased expressions of STAT3 and pSTAT3 were found almost universally in repeated resections of most analyzed patients, thereby suggesting a possibly pronounced role of this molecule in recurrent glioblastoma. Accordingly, it is nowadays known that STAT3 is increasingly expressed in glioblastoma as shown by us and other groups [
24,
25], and this expression has likely significant influence on several key biological characteristics of this tumor such as immunosuppressive microenvironment, hypoxia-induced angiogenesis and tumor cells migration [
26,
27]. In addition, STAT3 was shown to be a positive regulator of HIF1α, VEGF, MMP-2 and
TWIST in hypoxic glioblastoma [
28] and its suppression induced proliferation arrest and apoptosis in glioblastoma cells [
29]. To verify this observation in our analyzed patients, we have established primary glioblastoma cultures from their glioblastoma samples and compared several of their biological characteristics among them and with matched cryopreserved tumor samples. Resulting data indicated a significantly varying sensitivity of cultures to temozolomide (TMZ) and differently expressed hypoxia biomarkers including STAT3. Similar expression differences were also found between matching primocultures and cryopreserved tumor samples. Since the expression of STAT3 in primary cultures and matched cryopreserved tumor samples did not reflect the original observations, we have attempted to study it upon experimental conditions of its genetic deletion in a stabilized model glioblastoma cell line. STAT3 knockout in U87MG IDH wild type cells resulted in their lesser sensitivity to TMZ as compared to the same cells with wild type STAT3. Moreover, in vivo these cells produced larger tumors displaying lesser sensitivity to TMZ despite more significant accumulation of TMZ and lower presence of inactive TMZ metabolite. Conversely, loss of STAT3 suppressed expression of all evaluated hypoxia and EMT markers in glioblastoma cell line as well as in analyzed tumor originating from these cells with notable exception of
HIF1a and
VEGFC. Collectively, these results seem to suggest that loss of STAT3 at least in the present model does not render cells more chemosensitive to TMZ nor does it act suppressively on tumor growth in vivo. This finding is not consistent with the present understanding of the hypoxia-HIF1-STAT3-TMZ sensitivity axis and might thus relativize current efforts to develop specific STAT3 inhibitors to improve efficacy of glioblastoma management [
30]. In this context, it must be verified whether the used technology of targeted STAT3 deletion did not produce some incidental off target effects in exposed cells with resulting lesser chemosensitivity to TMZ being only temporary or permanent. Consequently, larger panels of cell lines and matching in vivo experiments would be required. All the more so since it is necessary to mention here that our used U87MG cells are not considered an optimal model for glioblastoma studies due to their unclear origin [
31,
32] which poses a certain limitation of our study. On the other hand, these cells are commonly employed in similar researches [
33,
34] and thus it may argued that results obtained with them are not entirely biased. Finally, it is known that IL-6/JAK/STAT3 pathway is very plastic with a number of feedback loops and crosstalk points with other signaling systems which may then explain why knockout of STAT3 did not improve chemosensitivity or tumor growth both in vitro and in vivo as has been discussed in case of STAT3 inhibitors [
30].
All together, data presented in this work illustrate complex expression patterns of select hypoxia and EMT-related markers in recurring glioblastoma samples as well as in cultures derived from them with no clear indefinable positive or negative correlations. It concerns for instance both HIF-1 and HIF-2 whose expression levels did not reflect a recurring nature of analyzed glioblastoma. It was also the case of STAT3 whose expression consistently increased in original tumor samples but not in derived primocultures or matched cryopreserved tumor specimens. In addition, subsequent studies on the role of STAT3 in the context of glioblastoma hypoxia demonstrated opposing effects of its deletion on cell viability as well as the expression of hypoxia and EMT markers. It is also possible that hypoxia and STAT3 operate in a larger tumor context and their individual roles are not sufficiently distinctive and standalone to universally characterize glioblastoma biology in LTSs in a similar manner to other analyzed markers, signatures or profiles. Another important factor to be taken into account is the nature and topography of hypoxia which is known to be spatially defined within a given glioblastoma but very likely heterogeneous across different tumors of individual patients. Its actual extent and distribution may significantly modulate local signaling events [
35,
36] and thus its detailed mapping might identify patients with high hypoxia burden and verify whether this parameter associates with specific signaling and LTS status. Individually varied hypoxia extent could also be responsible for a discrepancy between HIF1 and pSTAT3 expression levels detected in particular analyzed samples which contradicts the commonly acknowledged expression patterns and functional relationship between both markers. On the other hand, since both HIF1 and STAT3 expressions were more correlated in primocultures and matching cryopreserved samples, one has to consider sensitivity and specificity of used detection techniques.
Hypoxia landscapes in individual glioblastoma samples also correlate with the presence of glioma stem cells [
37] whose biological features such as non-proliferative status, specific signaling profiles and resistance likely influence glioblastoma resistance and aggressiveness and, consequently, patient survival. In our work we have not specifically focused on identifying glioma stem cells and their role in our cohort of glioblastoma LTS, however, the possible presence of such populations might represent a promising, hitherto unaddressed feature of glioblastoma which should be investigated in future.
Although results of this work did not reveal in recurrent glioblastoma samples generally unanimous and distinct factors characteristic for LTS patients, STAT3 and its expression appears to be a feasible point for evaluation of existence and roles of hypoxia in recurrent glioblastoma. It is further strengthened by the fact that in our Cox regression model STAT3 came out as a statistically significant marker of patient survival. These conclusions are nevertheless preliminary and limited by small patient/sample size. A larger confirmation of these results will be needed to further validate them.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.