Insulin‐like growth factor binding protein‐2 and glucose‐regulated protein 78 kDa: Potential biomarkers affect prognosis in IDH ‐wildtype glioblastoma patients

Abstract Background The overall survival of IDH‐wildtype glioblastoma patients is poor despite best available treatments. There is an urgent need for new biomarkers to inform more precise disease stratification. Previous studies have identified insulin‐like growth factor binding protein‐2 (IGFBP‐2) as a potential biomarker for glioblastoma diagnosis and therapeutic targeting. Other studies have indicated links between the insulin‐like growth factor (IGF) axis and tumorigenic functions of a molecular chaperone glucose related protein of 78 kDa (GRP78). We aimed to interrogate the oncogenic effects of IGFBP‐2 and GRP78 in our glioma stem cell (GSC) lines and clinical cohort. Methods Immunoblotting, reverse transcription quantitative real‐time PCR were used to quantify protein and mRNA levels derived from GSCs and non‐malignant neural stem cells (NSCs). Microarray analysis was used to compare the differences in IGFBP‐2 (IGFBP‐2) and GRP78 (HSPA5) transcript expression between NSCs, GSCs and adult human cortex samples. Immunohistochemistry was used to quantify IGFBP‐2 and GRP78 expression in IDH‐wildtype glioblastoma tissue sections (n = 92) and clinical implications assessed using survival analysis. Finally, the relationship between IGFBP‐2 and GRP78 was further explored molecularly using coimmunoprecipitation. Results Here, we demonstrate that IGFBP‐2 and HSPA5 mRNA is overexpressed in GSCs and NSCs in comparison to non‐malignant brain tissue. We also determined a relationship in which G144 and G26 GSCs expressed higher IGFBP‐2 protein and mRNA than GRP78, and this was reversed in mRNA isolated from adult human cortex samples. Clinical cohort analysis revealed that Glioblastomas with high IGFBP‐2 protein expression paired with low GRP78 protein expression were significantly associated with a much shorter survival time (Median = 4 months, p = 0.019) compared with 12‐14 months for any other combination of high/low protein expression. Conclusions Inverse levels of IGFBP‐2 and GRP78 may be adverse clinical prognostic markers in IDH‐wildtype glioblastoma. Further interrogation of the mechanistic link between IGFBP‐2 and GRP78 may be important for rationalisation of their potential as biomarkers and therapeutic targets.


| INTRODUCTION
Glioblastoma is the most prevalent and aggressive primary central nervous system (CNS) tumour with an incidence of 0.59-5 per 100,000 individuals. 2 The median overall survival is approximately 15 months, with a 5-year survival rate less than 5%. 3 Survival time is largely dependent upon patient age, treatment modality and the molecular profile of the tumour. 4,5 Consequently, brain tumour patients face the highest mean average years of life lost compared with any other cancer. 6 There is urgent need for new biomarkers to inform more precise disease stratification and to inform novel therapeutic options. 7 Glioblastomas are heterogenous diseases, which are thought to comprise both differentiated tumour bulk cells and primitive glioma 'stem-like' cells (GSCs), the latter of which represent a 'moving target' which may aid tumour recurrence and resistance to standard cytotoxic therapies. [8][9][10][11][12] GSCs are therefore an important model for studying early cellular dysregulation as well as the adaptive mechanisms evoked under cellular stress. Molecular pathways integral to normal cell functioning have been implicated in driving poor glioblastoma survival and resistance, including the insulin-like growth factor (IGF) axis and the chaperone glucose-regulated protein of 78 kDa (GRP78). [13][14][15] IGFBP-2 is the second most abundant IGFBP in the human serum (after IGFBP-3) and the most abundant in cerebral spinal fluid (CSF), here synthesised by choroid plexus epithelial cells and astroglia. 16,17 Established actions of IGFBP-2 includes the binding of IGF-I and IGF-II with high affinity, modulating bioavailability and ligand-IGF-1R interactions. 18 IGFBP-2 expression is high during foetal and embryonic brain development and is significantly reduced after birth. 19 Elevated levels are recapitulated in high grade glioma tissue biopsies and plasma samples in which poor clinical outcome potentiates harnessing IGFBP-2 as outcome predictor and biomarker. [20][21][22] IGFBP-2 has also been delineated as part of a robust 9 gene signature associated with worse outcome for highgrade glioma patients. 20 The mechanism by which IGFBPs and molecular chaperones such as GRP78 influence glioma outcome is still unclear. However, experimental and clinical data have linked GRP78 and IGFBPs with carcinogenesis in the breast. 23 GRP78 is a well-established endoplasmic reticulum (ER) molecular chaperone and member five of the heat shock protein 70 (HSP70) superfamily; also known as binding immunoglobulin protein (BiP) or heat shock 70 kDa protein 5 (HSPA5). 24 GRP78 is an important regulator of the unfolded protein response (UPR), by responding to changes in cell status via sequestration and release of signalling molecules vital for activation of anti-apoptotic pathways. 24 Its overexpression and cancer specific cell surface localisation, which have been described in glioblastoma, is believed to drive tumour growth, aggression and maintenance of the GSC mesenchymal phenotype. 13,25 In this study, we aimed to explore the expression of IGFBP-2 and GRP78 protein and mRNA (IGFBP-2 and HSPA5) in our GSC and non-malignant neural stem cell (NSC) lines and investigate potential protein biomarkers in our glioblastoma clinical biopsy cohorts with patient covariates and overall survival data. Conclusions: Inverse levels of IGFBP-2 and GRP78 may be adverse clinical prognostic markers in IDH-wildtype glioblastoma. Further interrogation of the mechanistic link between IGFBP-2 and GRP78 may be important for rationalisation of their potential as biomarkers and therapeutic targets.
Total RNA was extracted according to the manufacturers protocol for TRIzol (Ambion, 15596018). RNA was resuspended in 20 μL of DEPC-treated water (Ambion, AM9906), heated at 58°C for 10 min before measuring the RNA concentration and purity using a nanophotometer (IMPLEN, P330).

| RT-qPCR (reverse transcriptionquantitative PCR)
Reverse transcription from total RNA was carried out using the High-Capacity RNA-to-cDNA Kit (ThermoFisher Scientific, 4387406). Resulting cDNA was loaded on 96 well, transparent, Low Profile, 100 μL plates (Sarstedt Ltd, 72.1981) with SYBR Green jumpstart Taq (-20°C) ready mix (Sigma, S4438). The StepOne Plus Real-Time PCR system (Applied Biosystems, 4376600) and StepOne software V2.3 was used for absolute quantification and melt curve analysis. The primer sequences used were as follows:

| Immunohistochemistry (IHC)
All immunohistochemistry was assessed by a trained consultant neuropathologist, expert in identifying glioma cells. To exclude expression in non-neoplastic cells further, Haematoxylin and eosin-stained sections were also compared to select sections with majority glioma cells.
IHC was performed on 92 IDH-wildtype sections determined by immunohistochemistry with the R132HIDH1 antibody. This was followed up by IDH1/2 sequencing for non-canonical IDH mutations in patients 55 years and younger according to the WHO 2021 guidelines. Sections were obtained from North Bristol NHS Trust as part of BRAIN UK. 1 3-μmthick FFPE sections, collected on positively charged Apex Superior Adhesive Slides (Leica) with the primary monoclonal antibodies GRP78 (ab21685) and IGFBP-2 (ab216957). All stages of the IHC were completed using a Ventana Benchmark ULTRA (Roche) fully automated IHC staining system. The staining system deparaffinised the slides and then applied CC2, pH 6 (GRP78) and CC1, pH 9 (IGFBP-2) heat induced epitope retrieval. OptiView inhibitor (3% hydrogen peroxide) was used to quench endogenous peroxidases for 8 min at room temperature. The sections were incubated with the primary antibody at 1/200 for 32 min and then stained with the OptiView DAB IHC Detection Kit: HQ Universal Linker (secondary antibody cocktail containing HQ, a hapten covalently attached to goat antibodies) 8 min, horseradish peroxidase (HRP) multimer, 8 min, and 3,3′-diaminobenzidine (DAB), 7 min. Any DAB staining was enhanced with OptiView copper sulphate (4 min) before counterstaining with Harris haematoxylin (Diapath) 1 min.
A double-blind, semi-quantitative system was used to quantify GRP78 and IGFBP-2 immunohistochemical expression (Table 1). Overall scores were formed from addition of intensity and coverage scores, in which a score of ≥5 was used to indicate high GRP78 or IGFBP-2 protein expression. Researchers blinded to the clinical data and independent of each other (K.M.K. and H.R.B.) assessed the staining. A consensus meeting was carried out to agree on expression scores in which a discrepant score was observed (a difference in combined score of ≥3 or a final difference in high or low outcome).

| Statistical analyses
Western blots and qPCR experiments were each repeated on three biological replicates in triplicate. Data are presented as mean ± SEM, except RT-qPCR data (geometric mean ± 95% CI). Statistical tests were carried out using GraphPad (version 9.3.1 for Windows, GraphPad Software, www.graph pad.com), IBM SPSS Statistics 28.0.0.0 for Windows, and the Gliovis tool (http://gliov is.bioin fo.cnio.es/). [HG-U133_Plus_2] Affymetrix Human Genome U133 Plus 2.0 Array data from Pollard et al. was sourced from the NIH Gene Expression Omnibus (series GSE15209), https://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc=GSE15209. Data analyses were performed in collaboration with Miss Lily Andrews (Bristol Integrative Epidemiology Unit, University of Bristol). Data were downloaded and normalised in R studio (R version 4.0.3) using the caret package. 27 One-way ANOVA was used to compare the difference between three or more groups followed by the post hoc Tukey test or Dunnett test as appropriate. For comparing two groups, t tests were used. Kendall's Tau correlation coefficient was used to examine correlation of proteins in glioblastoma FFPE biopsy sections. Fisher's exact test was used to evaluate the significance of possible associations between GRP78 or IGFBP-2 expression and clinicopathological covariates. OS was defined as the interval between patient diagnosis and death dates. The Log-rank test was used to compare Kaplan-Meier survival curves. Univariate and multivariate Cox proportional hazards regression analysis were used to calculate hazard ratios and associated 95% confidence intervals. All statistical tests were two-tailed. Differences with p < 0.05 were considered statistically significant.

| IGFBP-2 is overexpressed in stem cells and upregulated in GSCs
To assess the relative basal expression of GRP78 and IGFBP-2, 3 independent passages of G26 and G144 GSCs and non-malignant U3 NSCs were analysed using western blotting and qPCR. The fold-change expression for GRP78 and IGFBP-2 in GSCs was then calculated with respect to the U3 NSCs. Staining for GRP78 was the strongest in the non-tumorigenic U3 cells with significantly lower expression in G26 and G144 cells (p = 0.023 and p = 0.004 respectively, Figure 1A). IGFBP-2 staining was strongest in the G144 cells and significantly increased compared to U3 cells (p = 0.019). In addition, there was significantly lower HSPA5 mRNA expression isolated from G26 and G144 cells (p = 0.038 and p = 0.036 respectively, Figure 1B) compared to U3 when normalised to GAPDH expression. The Expression profiling data were then analysed for differences in IGFBP-2 and HSPA5 transcript expression between NSCs, GSCs and adult human cortex samples (Gene Expression Omnibus (GEO) accession number GSE15209). IGFBP-2 transcript expression was significantly increased in GSCs and NSCs in comparison to normal brain (p < 0.0001, Figure 1C). Differences in HSPA5 transcript expression did not reach significance.
Pooled expression values in this data revealed a pattern suggesting that stem cells express higher levels of IGFBP-2 than HSPA5 and this is reversed for differentiated brain samples. Despite the significant differences in GRP78 and IGFBP-2 expression observed in this study between the U3 NSCs, G26 and G144 GSCs, microarray data suggest that F I G U R E 1 Protein and gene expression of GRP78 and IGFBP-2 in G26, G144 GSC cells, U3 control neural stem cells (NSCs) and normal brain (def). (A) Immunoblot analysis of GRP78 and IGFBP-2 protein expression in three independent passages of the control U3 NSC line and two GSC lines, n = 3. Densiometric analysis below the representative immunoblot was used to quantify protein expression levels for three technical repeats. After normalisation of protein integrated area density values relative to the loading control αtubulin values, ratiometric analysis was used to examine the GSC expression values relative to grouped U3 control protein homogenate expression. Data are presented as mean ± SEM, n = 3 biological repeats, all samples were analysed in triplicate. Additional technical repeats are shown in supporting information. (B) HSPA5 and IGFBP-2 mRNA expression for three independent passages of the control U3 NSC line and two GSC lines. Examined by RT-qPCR and normalised to the reference gene GAPDH. 2 −ΔΔCt was calculated to represent fold change. Data are presented as geomean and 95% confidence interval on a logarithmic scale (base 2), n = 3 biological repeats, all samples were analysed in triplicate. (C) Box plots depicting IGFBP-2 and HSPA5 expression from human foetal NSC lines, GSCs and normal brain. ****p < 0.0001; ***p = 0.0001-0.001; **p = 0.001-0.01; *p = 0.001 to 0.05; ns, not significant p ≥ 0.05. both IGFBP-2 and GRP78 appear to be expressed at higher levels in GSCs and NSCs. Pooled expression values reveal a pattern, suggesting that stem cells express higher levels of IGFBP-2 than HSPA5 and this is reversed for differentiated brain samples.

| GRP78 and IGFBP-2 protein expression in glioblastoma patient cohort biopsy tissue
IHC was performed on 92 glioblastoma (IDH-wildtype) FFPE biopsy sections to assess levels and association between GRP78 and IGFBP-2. The double-blind, semiquantitative criteria used to score GRP78 and IGFBP-2 staining is detailed in Table 1. Example IHC images indicating weak, moderate and strong GRP78 and IGFBP-2 staining are shown in Figure 2A. We observed that with increasing GRP78 intensity score, both the IGFBP-2 intensity and section coverage score increased. Moreover, sections which had a higher score for GRP78 staining coverage also had a larger IGFBP-2 coverage score ( Figure 2B,C).  Table 2. Using double-blind semi-quantitative criteria, GRP78 IHC scores were interpreted as 'high' in 74 (80.4%) and 'low' in 18 glioblastoma samples (19.6%). IGFBP-2 IHC scores were interpreted as 'high' in 75 (81.5%) and 'low' in 17 glioblastoma samples (18.5%). The associations between GRP78 or IGFBP-2 and the clinicopathological features of this cohort were analysed (Table 2), where the p value indicates the significance of the association. No significant associations were reported between GRP78 or IGFBP-2 IHC scores and patient sex, age, KPS, tumour location or MGMT promoter methylation status. (Table 2, p > 0.05 for all covariates).

| Patient cohort analysis of GRP78 and IGFBP-2 IHC scores, clinicopathological features and OS
Patient overall survival (OS) times were then analysed for associations with clinical covariates, including extent of surgical resection and adjuvant treatment modality (Table 3). Patient OS was calculated in months from diagnosis. In the glioblastoma (IDH-wildtype) clinical cohort (n = 92), the median overall survival (OS) for all patients was 12 months (range: 1-100 months), from the date of diagnosis. Univariate Cox's proportional hazards analysis of clinical covariates demonstrated no significant correlation between patient sex, age, KPS, tumour location, MGMT promoter methylation, GRP78, IGFBP-2 glioblastoma tissue abundance and OS (p > 0.05, Table 3). However, patient extent of surgical resection, and treatment modality were significantly correlated with the OS of glioblastoma patients (p = 0.011 and p < 0.001 respectively, Table 3, Figure 3).

| Patient survival differences with GRP78 and IGFBP-2 combinations
In our adult IDH-wildtype glioblastoma patient cohort (n = 92), no significant associations were found between GRP78 or IGFBP-2 and patient OS. We next analysed the relationship between GRP78 and IGFBP-2 using the chi-squared test. In this cohort, GRP78 and IGFBP-2 IHC scores were shown to be significantly associated (p = 0.002) and correlated (gamma =0.705, p = 0.016). Moreover, combining GRP78 and IGFBP-2 into four variables based on F I G U R E 2 IGFBP-2 and GRP78 immunohistochemistry (IHC) staining in 92 IDH-wildtype glioblastoma biopsy sections. Expression levels of IGFBP-2 and GRP78 were detected using IHC from a clinical cohort including 92 IDH1-wildtype glioblastoma biopsy sections. (A) IHC representative images showing GRP78 and IGFBP-2 weak, moderate and strong staining. Images taken at × 40 magnification. (B) GRP78 intensity score associated with IGFBP-2 intensity score or section coverage. The stacked bar charts depict increasing IGFBP-2 section coverage and staining intensity with respect to GRP78 staining intensity. (C) Section coverage percentage of GRP78 associated with section coverage percentage or intensity score of IGFBP-2. The stacked bar chart depicts increasing IGFBP-2 section coverage and staining intensity with respect to GRP78 section coverage. high and low IHC scores (IGFBP-2 low ; GRP78 high , IGFBP-2 low ; GRP78 low , IGFBP-2 high ; GRP78 high and IGFBP-2 high ; GRP78 low ), revealed that for patients with high IGFBP-2 and low GRP78 tumour abundances there was a significantly shorter median survival time when compared with to the other groups (Table 4, IGFBP-2 high ; GRP78 low median OS = 4 months, p = 0.019). Kaplan-Meier curves show the four different combinations and patient outcome ( Figure 4). Further analysis demonstrated that when GRP78 is low, survival is dependent on whether IGFBP-2 is high or low (p = 0.011, Figure 4D), and when IGFBP-2 is high there is a significant survival difference between high and low GRP78 (p = 0.006, Figure 4C).
The nine patients with high IGFBP-2 and low GRP78 tumour expression were further assessed for similarities based on patient characteristics and clinical covariates ( Table 5). Four patients were positive for MGMT promoter methylation and five were negative. Two patients had tumour biopsies, three underwent partial tumour resection, three underwent complete resection and one patient's resection status was unavailable. The locations of the tumours also showed differences, with two patients presenting with right temporal tumours, one with right parietal, one with right frontal, one with left frontal, one with left occipital, one with corpus callosum, one with bilateral parietal and one location was unavailable. The survival times of the three females in the group ranged from 1 to 4 months, whilst the 6 males ranged from 1 to 13 months.

| The association between GRP78 and IGFBP-2 in GSCs
To explore a mechanistic link between GRP78 and IGFBP-2, we carried out co-immunoprecipitation experiments. A GRP78 antibody was used to immunoprecipitate G26 cell lysates. The western blot shows the interaction between GRP78 and IGFBP-2 ( Figure 5). A IgG2a isotype control antibody was used as a comparison.

| DISCUSSION
IGFBP-2 appears to play key roles in foetal development and cognitive development. [28][29][30][31] Studies have shown that the IGFBP-2 genes promoter also contains binding sites for transcription factors involved in the maintenance of a stem cell state. 14 In our in vitro analysis of protein levels, we observed elevated IGFBP-2 expression in G26 and G144 GSC lines compared to U3 NSC lines, raising the T A B L E 2 Associations between GRP78 or IGFBP-2 IHC protein expression and clinicopathological features of IDH1-wildtype glioblastoma (n = 92). The Fisher's exact test, two-tailed p value was used to evaluate significance. Note: RT + TMZ full: 60Gy total given over 5 days/ week for 6 weeks with concurrent daily temozolomide (75 mg/m 2 /day, 7 days/week) and 6 cycles of adjuvant TMZ (150-200 mg/m 2 /day for 5 days during each 28-day cycle). RT + TMZ Partial: 60Gy total given over 5 days/ week for 6 weeks with concurrent daily temozolomide (75 mg/m 2 /day, 7 days/week) without completing 6 cycles of adjuvant TMZ; or 60Gy total given over 5 days/ week for 6 weeks without concurrent temozolomide but completing 6 cycles of adjuvant TMZ (150-200 mg/m 2 /day for 5 days during each 28-day cycle. n, number of people in each group in Table 3. Events, indicates the number of people that died in each group in Table 3. Bold indicates statistical significance (p < 0.05).
possibility that the developmental pathways mediated by IGFBP-2 were aberrantly activated in these cells. GRP78 is critical for cell fate, controlling the survival/apoptotic potential of the cell in response to stress. GRP78 overexpression has been reported in glioblastoma, correlating with aggressive and radioresistant phenotypes. 13,32 Our cell line work and microarray analysis indicated that IGFBP-2 and GRP78 appear to be expressed at higher levels in GSCs and NSCs. Stem cells also showed higher expression levels of IGFBP-2 than HSPA5 and this was reversed for differentiated brain samples. In our individual cell lines, GRP78 protein and mRNA (HSPA5) expression was lower than IGFBP-2 in G26 and G144 GSCs. However, the analysis of more cells including a range of subtypes is required to determine whether this expression pattern could define a subset of GSCs. A similar relationship has been reported with IGFBP-3 and GRP78, where the combination of low GRP78 and high IGFBP-3 was clinically significant, indicating poorer overall survival of breast cancer patients. 23 Finally, we carried out co-IP experiments in lysates collected from the G26 cell line to confirm the mechanistic association between GRP78 and IGFBP-2.
We quantified GRP78 and IGFBP-2 protein expression for 92 IDH-wildtype glioblastoma patient tumour samples. Survival analysis between the clinical covariates of these patients demonstrated that patient extent of surgical resection, and treatment modality were significantly correlated with the OS of glioblastoma patients (p = 0.011 and p < 0.001 respectively). However, MGMT promoter methylation status did not reach significance with respect to overall survival. MGMT is an important predictive biomarker and indication of a patient's sensitivity to alkylating chemotherapy. In our small cohort study, 26.7% of patients with known MGMT promoter methylation status (n = 90) had no information on treatment received. Of those patients in which treatment modality was available, more patients without MGMT promoter methylation received some kind of alkylating therapy (whether full or partial), and more patients with MGMT promoter methylation received palliative care. Further survival analysis using the clinical tumour biopsy samples demonstrated that neither IGFBP-2 nor GRP78 expression were significantly associated with survival alone (IGFBP-2, p = 0.395; GRP78, p = 0.509), despite 80.4% (74/92) patients displaying high GRP78 tumour expression and 81.5% (75/92) patients displaying high IGFBP-2 tumour expression. However, larger published studies have linked IGFBP-2 with prognosis in glioma, as a prognostic biomarker and therapeutic target both in tissue 14,15,[33][34][35][36] and plasma. 20,21,37,38 In our study, tumour tissue with high IGFBP-2 expression paired with low GRP78 expression were significantly associated with, a much shorter survival time (median = 4.0 months, p = 0.019). In other studies, higher GRP78 mRNA expression positively regulated overall and progression free survival in gastric cancer patients and was reported to be a positive marker for prognosis and chemotherapy response in breast cancer. 39,40 Moreover, drug induced GRP78 degradation in murine macrophages has been linked to prolonged ER stress and inflammation mediated by IL-6 release, driving the  tumorigenic microenvironment and cellular adaptation further exploiting adaptive mechanisms. 41 The expression, pleiotropic functions and spatiotemporal dynamics of IGFBP-2 and GRP78 appear to intricately shift in response to cellular context and adaptation. Redistribution of GRP78 via translocation to the surface of cells has been recorded in cancer cells in which a signal transducing role can be adopted. 42,43 The maintenance and radio sensitisation of mesenchymal GSCs have also been shown to be dependent upon cell surface GRP78 (csGRP78), adding to evidence that expression and function are context dependent. 25 The mechanism of F I G U R E 4 Kaplan-Meier analysis showing associations between patient overall survival (OS) and combinations of GRP78high/low and IGFBP-2high/low expression in 92 primary IDH-wildtype glioblastoma patient tumours. (A) GRP78 is not related to survival (p = 0.509). (B) IGFBP-2 is not related to survival (p = 0.395). (C) When IGFBP-2 is high there is a significant survival difference between high and low GRP78 (p = 0.006). (D) When GRP78 is Low, survival is dependent on whether IGFBP-2 is High or Low (p = 0.011).
T A B L E 5 Clinicopathological variables and covariates for 9 patients with a high IGFBP-2/low GRP78 IHC protein expression in a cohort of 92 patients with IDH1-wildtype glioblastoma. translocation is largely unknown but believed to involve additional binding partners as part of a context-dependent process. 43 There appears to be overlap between signalling pathways initiated by csGRP78 and IGFBP-2. For example, csGRP78 binding to α-2-macroglobulin (α2M) is believed to stimulate invasion and metastasis. 44 However, α2M is also believed to complex with IGFBP-2 in the circulation preventing IGFBP-2 proteolysis and modulating IGF signalling. 45 Further crossover between the pathways have revealed that surface GRP78 can mediate IGF-IR phosphorylation and activation of growth and survival pathways. 46 Moreover, a feedback loop denotes that IGF-1R mediated IGF-I signalling can upregulate GRP78 via transduction through PI3K/AKT/mTORC1 signalling. 47 IGF-I has also been shown as a direct stimulator of PI3K/Akt and MAPK mediated expression and GRP78 cell surface redistribution in hepatoma cell lines. 46 In addition, extracellular functions of IGFBP-2 have been linked to the proliferative and invasive capacity of glioma cells via integrin β1mediated activation of ERK phosphorylation and nuclear translocation. 48 Other downstream signalling cascades mediated by the integrin β1 involve ILK/NFκB signalling, FAK/ERK pathway activation and CD144 expression, as well as βcatenin nuclear and cytoplasmic regulation. [49][50][51][52] Critically, the negative regulation between IGFBP-2 and PTEN has been shown to be of clinical importance in triple negative breast cancer, glioblastoma and prostate cancer. [53][54][55] Other IGFBPs have previously been identified as important malignancy biomarkers including IGFBP3; the most abundant IGFBP in the human serum. The ability of IGFBP-3 to differentially affect cell survival and invasion, depending upon the extracellular environment, is influenced by the presence of GRP78 in human breast cancer cell lines. 23 Mechanistic intricacies underlying the relationship between IGFBP-2 and GRP78 remain to be elucidated. Future investigation will aim to determine the exogenous actions of IGFBP-2 with and without the presence of GRP78, as well as the facilitation of IGFBP-2 trafficking and oncogenic pathway activation via protein complexing, as suggested by other studies. 56

| CONCLUSION
Taken together, our experimental and clinical findings suggest that inverse levels of IGFBP-2 and GRP78 may be adverse clinical prognostic markers in IDH-wildtype glioblastoma. Further experimental studies into the mechanism of interaction and the potential for therapeutic manipulation are warranted.

DATA AVAILABILITY STATEMENT
The results in this paper are in part generated using Affymetrix Human Genome U133 Plus 2.0 Array data F I G U R E 5 Immunoblot showing a co-IP assay carried out on lysates extracted from G26 GSCs. IGFBP-2 was pulled down using a GRP78 antibody. The IgG2a antibody was used as an isotype control. Immunoblots from two additional co-IP assays are shown in supporting information.