Elevated proportions of activated NK cells at diagnosis predict a favorable prognosis in glioblastoma patients

Simple Summary: Despite multimodal treatment WHO grade IV glioblastoma remains a devastating disease with dismal prognosis. Due to the high metabolic demand aggressive tumors frequently overexpress Hsp70. Depending on its intra- or extracellular localization, Hsp70 either promotes tumor progression or stimulates the immune system. All gliomas are membrane Hsp70-positive, and high grade gliomas show a strong nuclear and cytosolic Hsp70 expression. Large necrotic tumor areas are associated with significantly increased Hsp70 serum levels. Elevated frequencies of Hsp70-activated NK cells at diagnosis are associated with a more favorable prognosis in patients with grade IV gliomas. Abstract: Despite rapid progress in the treatment of many cancers, glioblastoma remains a devastating disease with dismal prognosis. The aim of this study was to identify immune-related biomarkers that more effectively predict outcome of glioblastoma. Since heat shock protein 70 (Hsp70) and IL-2 are known to increase the expression of activatory NK cell receptors, recognizing aggressive human tumor cells that present Hsp70 on their cell surface, extracellular Hsp70 levels were determined in glioma patients together with activatory NK cell receptors. All gliomas are membrane Hsp70-positive (mHsp70+) and high grade gliomas more frequently show an overexpression of Hsp70 in the nucleus and cytosol. Significantly increased extracellular Hsp70 levels are detected predominantly in glioblastomas with large necrotic areas. Overall survival (OS) is more favorable in patients with low Hsp70 serum levels indicating that a high Hsp70 expression is associated with an unfavorable prognosis. Elevated frequencies of NK cells are associated with a more favorable outcome. Of caution, a glucocorticoid therapy reduces the prevalence of NK cells. In summary, elevated frequencies of Hsp70-reactive NK cells at diagnosis and lower Hsp70 levels predict a more favorable prognosis in glioblastoma patients.


Introduction
Heat shock protein 70 (Hsp70), the major stress-inducible member of the 70-kDa heat shock protein family is evolutionary conserved and ubiquitously expressed in nearly all subcellular compartments [1]. Under physiological conditions Hsp70 maintains protein homeostasis by preventing protein aggregation, supporting the folding of nascent polypeptides and transporting other proteins across membranes [2]. Stress such as hyperthermia, ischemia, nutrient deficiency or therapeutic interventions strongly upregulate the synthesis of Hsp70. Cytosolic Hsp70 interferes with signaling pathways that affect apoptosis, proliferation and differentiation [2] and as a result, elevated intracellular Hsp70 levels protect tumor cells from apoptosis [3]. Due to their high metabolic demand many aggressive tumor types exhibit an overexpression of Hsp70 which promotes tumorigenesis [4]. Following stress, Hsp70 rapidly translocates into the nucleus and/or interacts with lysosomal membranes in the cytosol to stabilize them [5].
We have previously shown that serum levels of Hsp70 correlate with the viable tumor mass and affect peripheral blood lymphocyte (PBL) compositions and profiles in patients with non-small lung cell cancer (NSCLC) [6]. Extracellular Hsp70 exists in the form of free protein which predominantly originates from dying tumor cells or is bound to exosomes that are actively released by viable tumor cells [7,8].
As part of the innate immune system NK cells are responsible for the first line of defense against infectious diseases and cancer. NK cells release T cell recruiting chemokines including IFN-γ, TNFα. NK cell cytotoxicity is mediated either via granzyme B/perforin, apoptosis receptor interactions and/or antibody dependent cellular cytotoxicity (ADCC) through Fc-γ receptors. Their activity is independent of a clonal T cell receptor and major histocompatibility complex antigens (MHC), but is tightly regulated by the balance between the expression density of activating receptors with a short ITAM and inhibiting receptors with a long ITIM motif [9]. To license NK cells to kill mHsp70+ tumor cells they can be activated by an incubation with full length Hsp70 protein, Hsp70-expressing exosomes [7,8] or a 14-mer Hsp70-peptide TKD (TKDNNLLGRFELSG) derived from the C-terminal substrate binding domain in combination with interleukin 2 (IL-2) [10,11]. Due to the smaller size, favorable biodistribution, higher purity and easier GMP-production the Hsp70-peptide TKD is advantageous to recombinant Hsp70 protein for stimulating of NK cells. A pilot [12] and clinical phase I study [13] demonstrated excellent safety profiles of ex vivo TKD/IL-2-activated, autologous NK cells, and patients with advanced NSCLC showed promising clinical responses in a randomized clinical phase II trial after adoptive transfer of ex vivo TKD/IL-2-activated NK cells [14]. Apart from a study in grade IV glioblastoma/astrocytoma patients [15] insight into the expression and localization of Hsp70 in different types of brain tumors and its impact on the prevalence and activity of NK cells is limited. Therefore, intracellular, extracellular and mHsp70 levels were profiled as potential prognostic biomarkers and stimuli for NK cells.
Gliomas predominantly originate in the brain parenchyma. Among the diffuse growing gliomas which are classified according to the WHO classification as grade II, III and IV gliomas, glioblastoma is the most common primary malignant tumor of the central nervous system in adults [16,17]. Glioblastoma is treated with surgery, radiation and a temozolomide-based chemotherapy [18]. However, despite this multimodal therapy these patients still have a dismal prognosis with median survival rates of approximately 15 months [18]. Tumor grading, age, isocitrate-dehydrogenase 1 (IDH1) and methylation status of the DNA repair gene O-6-methylguanin-DNA methyltransferase (MGMT) promotor, performance status and extent of resection are considered for predicting clinical outcome [19,20]. However, presently only the MGMT promotor status qualifies for a prediction of clinical responses to temozolomide.
The goal of the present study was to identify immune-related biomarkers in the circulation of glioma patients that predict outcome more reliably.

MR imaging and Hsp70 staining of brain tumors
Clinical characteristics such as WHO grade, age, diagnosis, isocitrate-dehydrogenase 1 (IDH-1) wild type (WT) of all glioma patients are summarized in Table 1. Representative MR images of different brain tumors are shown in Figure 1A.

Membrane Hsp70 (mHsp70) expression on viable brain tumor cells
In addition to the cytosolic Hsp70 content, the mHsp70 positivity was determined in isolated, single cell suspensions of freshly resected brain tumors by flow cytometry. Due to the poor viability of brain tumor cells in vitro (20-30%), the mHsp70 status could be determined only in a total of 37 tumor samples. An intact plasma membrane is key for determining the membrane status of Hsp70 because otherwise a false positive, cytosolic Hsp70 staining will be determined by flow cytometry. The mHsp70 staining intensity appeared to be slightly higher in glioblastomas compared to low grade gliomas. Overall the proportion of mHsp70+ tumor cells was above 60% in all glioma grades, but did not differ significantly between the different tumor grades (Figure 2).

Extracellular Hsp70 concentrations
Serum Hsp70 levels were determined in grade II (n=5), grade III (n=12) and grade IV (n=51) patients at first diagnosis using the R&D ELISA which predominantly detects free Hsp70 originating from dying tumor cells and by an "in-house" Hsp70 ELISA [21] that detects both, free and liposomal Hsp70. As shown in Figure 3A, the median free Hsp70 concentrations in the circulation of patients with grade IV gliomas at diagnosis were significantly higher (3.48 ng/ml) than in healthy controls (n=150, 2.60 ng/ml, Tukey; *p<0.05). Median serum Hsp70 concentrations in grade II and III glioma patients were 3.14 ng/ml and 3.34 ng/ml, respectively, but did not differ significantly to that of healthy control donors ( Figure 3A). With respect to liposomal Hsp70 which is representing the viable tumor mass in other tumor entities [21], no significant differences were observed in all tumor grades (data not shown).
According to their tumor volumes, patients with grade IV gliomas (n=44) were separated into groups of patients with small (<30 cm 3 ), medium (30-90 cm 3 ) and large (>90 cm 3 ) tumors. The median tumor volumes in patients with small tumors (n=6) was 9.2 cm 3 (range 4.8-26 cm 3 ), in those with medium tumors (n=16) was 71.7 cm 3 (range 35.3-81.8 cm 3 ), and in those with large tumors (n=22) was 140.9 cm 3 (range 99-222 cm 3 ). As shown in Figure 3B, glioblastoma patients with large tumor volumes had significantly higher median Hsp70 concentrations in the serum (3.68 ng/ml, n=22) than healthy controls (2.60 ng/ml, n=150), as calculated by the Kruskal Wallis test (*p<0.05), whereas those patients with small and medium tumor volumes did not differ significantly from the healthy control group. According to MRI, the subgroup of grade IV gliomas (n=27) was subdivided into a group with a high proportion of necrosis (>10%, n=13) and a group with a low proportion of necrosis (<10%, n=14). The concentration of free Hsp70 in the circulation (3.42 ng/ml) in glioblastoma patients with highly necrotic tumors was significantly higher compared to healthy controls (2.60 ng/ml, Wilcoxon rank Test, *p<0.05) ( Figure 3C), whereas, that of patients with low tumor necrosis did not differ significantly to healthy controls (2.87 versus 2.60 ng/ml) ( Figure 3C). As expected, liposomal Hsp70 levels did not differ significantly in patients with low and high proportions of tumor necrosis, but were elevated compared to that of healthy controls (data not shown). A Kaplan-Meier analysis of glioblastoma patients whose free Hsp70 serum levels were either below (n=18) or above (n=16) a threshold of 3.5 ng/ml which distinguishes healthy human individuals from tumor patients revealed an improved OS in patients whose Hsp70 serum levels were below 3.5 ng/ml (p=0.1) ( Figure 3D).

Discussion
In 2018 the Nobel prize in Physiology and Medicine was awarded to James P. Allison and Tasuku Honjo for their discovery of immunoregulatory mechanisms mediated by immune checkpoint inhibitors [22], but already years before the interest in immunotherapeutic clinical approaches including immune checkpoint inhibitor blockades [23] and genetically engineered chimeric antigen receptor (CAR)-T/NK cell therapies has attracted the attention of many immunologists [24][25][26][27].
However, despite promising results in a number of cancer entities, a significant proportion of patients, including those with brain tumors, do not profit from these immunotherapies [28,29].
Therefore, a better understanding of immunosuppressive and immunostimulatory effects in patients with gliomas is urgently needed. In this study, we aim to identify immune-related biomarkers for predicting of prognosis of glioblastoma by profiling the composition of PBLs and by measuring the expression of intra-and extracellular Hsp70 levels. The concepts were based on the facts that Hsp70 exerts immunostimulatory effects on NK cells or anti-apoptotic activities, depending on its extra-or intracellular localization [30]. Herein, we have shown that high cytosolic levels of Hsp70 are associated with a more aggressive glioma type, but extracellular Hsp70 levels might act as a DAMP to stimulate NK cells. This hypothesis is supported by promising data of phase I and II clinical trials using ex vivo Hsp70 peptide and IL-2 stimulated NK cells to treat patients with advanced colon and lung cancers after radiochemotherapy [13,14]. The stimulation of patient-derived, anergic NK cells with Hsp70 peptide plus IL-2, but not with IL-2 alone, significantly enhanced the cytolytic activity of NK cells against membrane Hsp70+ tumor cells [13], and favorable clinical responses in a randomized phase II clinical trial could be attributed to an increased prevalence of activated CD94+ NK cells [14] that are responsible for the interaction with Hsp70 [31]. The increased cytolytic activity of patientderived, Hsp70-activated NK cells can be blocked either by a CD94 specific blocking antibody for NK cells or by an Hsp70 specific antibody for blocking mHsp70 on tumor cells [13]. Previously, we could demonstrate that only MACS purified CD3-NK cells (>97%), but not CD3+ T or γ/δ T cells do recognize mHsp70+ tumor cells and can be stimulated with Hsp70 protein in the absence of cochaperoned immunogenic peptides or TKD and IL-2 [32]. Therefore, other effector cell populations can be excluded for the recognition of mHsp70+ tumor cells. The lysis of mHsp70+ tumor cells by NK cells could be attributed to granzyme B [33]. Following binding and uptake, granzyme B is able to efficiently kill mHsp70+ tumor cells [34]. The production of granzyme B in NK cells of patients with solid tumors [35] is frequently impaired, but can be reconstituted by a stimulation with IL-2 and Hsp70 or TKD [8][9][10]. Therefore, herein, we studied CD94+ NK cells together with intra-, extracellular and membrane-bound Hsp70 levels.
In contrast to non-malignantly transformed brain tissues (unpublished observation), nearly all highly malignant tumor cells exhibit an upregulated cytosolic Hsp70 expression and a membrane Hsp70-positivity that mediate anti-apoptotic activities and therapy resistance [36,37]. In line with these results, an upregulated nuclear and cytosolic Hsp70 expression was predominantly found in high grade gliomas [38,39], whereas low grade gliomas show a weaker and purely nuclear staining pattern. Due to the limited number of patients with lower grade gliomas, future studies with larger patient cohorts are necessary to assess more detailed the localization of Hsp70 in these tumors.
Compared to healthy controls, significantly elevated serum levels of free Hsp70 were observed in glioblastoma patients with large necrotic tumor areas, but not in low grade gliomas. From this we conclude that extracellular Hsp70 predominantly originates from dying tumor cells [21]. An intact blood-brain-barrier in patients with low grade gliomas might hinder the release of Hsp70 into the circulation of these patients. Liposomal Hsp70 levels [21] predominantly originating from exosomes [7] which are assumed to be actively released from viable tumor cells, were higher in grade IV compared to lower grade gliomas, but were not found to be increased in patients with highly necrotic tumors.
An improved OS of glioblastoma patients was associated with Hsp70 serum levels below a threshold of 3.5 ng/ml which is typically found in healthy individuals. These findings support the hypothesis that the aggressiveness of glioblastomas is associated with high Hsp70 levels. Extracellular Hsp70 either free or in exosomes [7][8][9] together with other DAMPs derived from necrotic tumor cells can stimulate the release of pro-inflammatory cytokines including IL-2. We have previously established that Hsp70 in combination with IL-2 stimulates NK cell-mediated immune responses which are mediated by activatory receptors such as CD94/NKG2C can be further enhanced by immune checkpoint inhibition [31,40]. To test the hypothesis that Hsp70-mediated immunostimulation of NK cells might also occur in patients with high grade gliomas, PBLs from healthy donors were stimulated with recombinant Hsp70 or the TKD Hsp70-peptide together with low dose IL-2. This in vitro stimulation resulted in a similar upregulation of the frequency and density of these activatory NK cell receptors such as CD94+ and CD69+ on NK cells to that observed in glioblastoma patients. An upregulated CD94 expression on NK cells correlates with an elevated cytolytic activity against mHsp70+ tumor cells in vitro and in clinical trials [8,[12][13][14]. Although Hsp70 serum levels are increased in the serum of glioblastoma patients the physiological concentrations in the circulation are too low for an efficient NK cell stimulation. Presently we cannot exclude locally increased Hsp70 levels in glioblastoma patients which might have the capacity to activate NK cells.
In line with the literature, patients with high grade gliomas in our study exhibit a decreased frequency of CD3+/CD4+ T cells [41,42]. However, the value of tumor-infiltrating NK cells as a biomarker for gliomas is not completely understood. Although one study denies an involvement, another study indicates that NK cell infiltration into the tumor microenvironment is more common in high grade than low grade gliomas [43,44]. Due to the KIR2DS400101 allele being associated with the capacity to kill glioblastoma cells, NK cell-based immunotherapies have been discussed as an additional treatment option in combination with surgery and radiochemotherapy [36,45,46]. Moreover, targeted CAR-NK cell-based immunotherapies are presently tested in preclinical and clinical studies of glioblastoma [47][48][49]. However, cancer-initiating cells (CICs) escaping immune recognition, elevated Treg, M2 macrophage and MDSC counts in the tumor microenvironment, immunosuppressive factors (i.e., IL-1, TGF-β, IL-10, arginase I), apoptosis-inducers (i.e., CD95, CD70), tumor hypoxia and immune checkpoint inhibitors are major challenges for immunotherapies [50,51,52]. Future studies are necessary to elucidate the capacity of NK and CAR-NK cells to overcome these immunosuppressive activities in glioblastoma.
A pre-surgical treatment with glucocorticoids reduces the prevalence of activated CD94+ NK cells. Since long-term glucocorticoid therapy has been shown to exert unfavorable effects on the survival of glioblastoma patients treated with RT and/or CT [53] and exert immunosuppressive functions in combination with immune checkpoint inhibitor therapies [54] their use should be considered with caution.

Patients
The study was approved by the local ethical committee of the medical faculty of the Technical University Munich (TUM, #2403/09) and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients before the start of therapy. EDTAanticoagulated blood (for flow cytometry) and non-coagulated blood (for ELISA) was collected from patients with different types of brain tumors attending in the Department of Neurosurgery. A total of 88 patients were included into the study. According to the WHO classification of Tumors of the Central Nervous System [55] and histopathological analysis (Dpt. Neuropathology), 2 patients were diagnosed as having grade II oligodendroglioma, 6 as having grade II astrocytoma, 8 as having grade III anaplastic oligodendroglioma, 6 as having grade III anaplastic astrocytoma, and 66 patients as having grade IV gliomas. The samples were reviewed for this study. All patients were treated with radiotherapy concomitant with temozolomide after surgery according to the Stupp-regimen [18]. Clinical outcome and extracellular Hsp70 levels below and above a cut-off value of 3.5 ng/ml which was determined by using the Youden-index, at diagnosis were associated with OS in grade IV glioblastoma patients (n=34) by Kaplan-Meier analysis.

Tumor volumetry by magnetic resonance imaging (MRI)
Pre-operative MRI scans were performed on a 3Tesla MRI scanner (Philips Achieva, Philips Ingenia (Philips Medical Systems, The Netherlands B.V.) or Siemens Verio (Siemens Healthcare, Erlangen, Germany) and analyzed as 3DT2-FLAIR and 3DT1 pre-and post-contrast sequences. Tumor volumes were determined by 2 independent neuroradiologists. MRI sequences (3DT2-FLAIR, 3DT1 post-contrast) were co-registered using a 3D Slicer (www.slicer.org) [56] and tumor volumes were segmented semi-automatically using a freely available software ITK-SNAP (www.itksnap.org) [57] in contrast-enhancing and FLAIR-hyperintense tumor areas.

Flow cytometric analysis of mHsp70 expression on viable brain tumor cells
Freshly aspirated tumor material (Dpt. Neurosurgery) was cut in 1 mm 3 pieces, incubated with trypsin for 8 min and forced through a sterile mesh (70 µm strainer). After resuspension of the single cell suspension in PBS/10% v/v FCS, 5x10 5 cells were incubated with the following fluorescencelabelled antibodies for 30 min on ice: tube 1: IgG1-FITC /APC (BD Biosciences), tube 2: CD45-APC (Thermo Fisher) and cmHsp70.1-FITC (multimmune GmbH), tube 3: pan-HLA class I-FITC (F5662, Sigma). After two washing steps, 7AAD (BD Biosciences) was added directly before analysis. Only viable, 7AAD-negative tumor cells which are CD45-negative (to exclude lymphocytes), were gated and analyzed using a FACSCalibur™ flow cytometer (BD Biosciences). The pan-HLA class I antibody staining served as a positive control. A minimum of 100.000 events were recorded for each measurement.

Immune phenotyping of peripheral blood lymphocytes (PBLs)
Lymphocyte subpopulations in the peripheral blood were profiled by multicolor flow cytometry on a FACSCalibur™ instrument (BD Biosciences) using the following fluorescence-

Statistics
The statistical analysis was performed using the programming language R, R studio version 3.5.2. Normal distribution was tested by the Shapiro-Wilk normality test. Parametric data were analyzed by ANOVA and post-hoc Tukey tests, non-parametric were analyzed by using the Kruskal Wallis test and Wilcoxon signed-rank test, as appropriate. A value of p<0.05 was considered as representing statistically significant differences.

Conclusions
Hsp70 is presented on the plasma membrane of all gliomas and high nuclear and cytosolic Hsp70 levels are associated with high grade gliomas. The present study provides first evidence that elevated proportions of activated CD94+/CD69+ NK cells, which are known the recognize mHsp70+ tumor cells [8], in glioblastoma patients at first diagnosis might be predictive for a more favorable clinical outcome. However, due to the relatively low number of patients the results need to be confirmed in a larger patient cohort.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to ethical reasons in a clinical study.