L-Carnitine–Mediated Tumor Cell Protection and Poor Patient Survival Associated with OCTN2 Overexpression in Glioblastoma Multiforme

Glioblastoma (GBM), classified as WHO grade IV glioma, represents the most frequent and most aggressive type of primary brain tumor in adults. Despite multimodal therapy including surgical resection followed by adjuvant radiation and chemotherapy, GBMs are characterized by rapid tumor recurrence resulting in a poor prognosis with a median survival of only 12–15 months (1). Although a number of genetic and epigenetic alterations were discovered in recent decades, many of the molecular mechanisms underlying the resistance of GBM cells remain largely elusive, leading to a lack of substantial progress in the therapeutic management of GBM. Several potential reasons exist as to how these tumors acquire treatment resistance, including enhanced expression of drug efflux transporters, alterations in drug metabolism, mutations of drug targets, and the activation of survival or inactivation of death signaling pathways (2, 3). Furthermore, studies argue for a potential role of altered redox homeostasis and energy metabolism in the development of antitumoral drug resistance (4, 5).

In recent years, evidence has emerged suggesting that the amino acid derivate L-carnitine (trimethylamine-β-hydroxybutyrate, LC), which is transported into the cells mainly by the Na⁺-dependent transporter OCTN2 (SLC22A5), contributes to cell protection by interacting with various targets inside the cell. Besides its key role in energy metabolism, several studies attest the OCTN2/LC system a broad spectrum of capabilities including scavenging free radicals, stabilizing membranes, enhancing antioxidative resources, and promoting antiapoptotic pathways (6–9). Therefore, the OCTN2/LC system has been widely discussed as a possible treatment adjunct in some neurodegenerative disorders such as Parkinson's and Alzheimer's disease to stabilize cellular integrity of compromised neuronal cells (10, 11). However, it has not yet been investigated whether GBM cells may also benefit from the OCTN2/LC system to increase their survivability, and whether OCTN2 and LC could be prognostic markers in GBM progression.

LC-mediated cytoprotection is closely linked to elevated levels of glutathione peroxidase, catalase and superoxide peroxidase (12–14), and to the induction of the NRF2 (nuclear factor erythroid 2-like) transcriptional network, one of the major pathways used by both normal and cancerous cells to counteract oxidative insults (15). Given that GBM has a high proliferation rate combined with an elevated basal metabolic turnover and formation of reactive oxygen species (ROS) as a natural byproduct (16), cancer cells may upregulate their antioxidant resources and survival pathways to prevent irreversible cellular damage (17, 18). Depending on its pleiotropic effects, accumulation of LC may strengthen tumor cell resistance as anticancer treatments rely, in part, on the destructive effect of ROS in their target cells (19, 20).

The OCTN2/LC system exists in almost all tissues, including the brain, as it is involved in essential metabolic processes (21). Well-known as a key compound in the “carnitine shuttle,” LC facilitates the transport of middle- and long-chain acids into mitochondria to subsequently undergo β-oxidation, allowing cells to metabolize fatty acids as energy resource (22). Similar to other tumors, GBMs have long been thought to rely upon glycolysis for energy production, while current studies suggest that glucose accounts for only <50% of acetyl-CoA production in gliomas (23). However, it was shown that fatty acid metabolism serves as an alternative energy source to promote GBM progression, whereas inhibition of β-oxidation led to prolonged survival in a GBM mouse model (5).

With a broad spectrum of cytoprotective effects and of metabolic relevance, the OCTN2/LC system may be a potential actor in cell resistance, but its role in GBM biology has not yet been explored. Here, we report on the expression and prognostic impact of the OCTN2/LC system in resected tumor specimens of patients with newly diagnosed and recurrent GBM. We further aimed to investigate the influence of LC on GBM cell survival and migration in vitro to verify its underlying cytoprotective capabilities. To finally assess the OCTN2/LC system as a therapeutic target, we analyzed the influence of the OCTN2/LC inhibitor meldonium on in vivo GBM growth using an orthotopic mouse model.

Following an institutional review board–approved protocol (in accordance with the ethical standards of the Helsinki Declaration of the World Medical Association), fresh human GBM tissues were obtained from patients undergoing surgical resection or biopsy within their therapeutic regimen at the Department of Neurosurgery, University Medicine Greifswald (Greifswald, Germany), in the period from April 2007 to August 2016. At our university hospital, approximately 2 patients with glioblastoma are undergoing surgery every month. Because written informed consent was mandatory for both cryopreservation of the specimens and obtainment of vital status from official population registry, a total of 121 patients finally served as cohort for this study. Baseline epidemiologic and clinical characteristics of all patients are shown in Table 1. On the basis of histologic confirmation according to the 2007 WHO Classification of Tumors of the CNS (24), resected specimens included 80 primary GBM and 41 recurrences of primary GBM. Furthermore, 24 tumor samples from patients with astrocytoma grade II or III were included in this study. Eight nonmalignant brain tissues (frontal and temporal lobes) were obtained by routine autopsies. The patients died of pneumonia, heart failure, sepsis or pancreatic cancer, but had no underlying brain disease. In addition, 2 further nonmalignant brain samples (1 frontal and 1 temporal) were obtained from the BioChain Institute Inc.. All resected tumor samples and control brain tissues were cut and frozen at −80°C.

The human GBM cell lines A-172, GaMG, HF66, U251MG, U373, LN18, U87MG (bought from the ATCC), the murine cell line GL261, and patient derived primary GBM cells were used for the qualitative protein assessment. LN18, a WHO grade IV GBM cell line, was further used as in vitro model for the mechanistic studies. Because of its similarity to human GBM growth characteristics (25), GL261 cells were selected for our mouse model of malignant glioma. Detailed information on all cell lines are provided in Supplementary Table S1. Only cells at low passages were used for the experiments. By employing a PCR-based assay, all cell lines were routinely monitored for potential Mycoplasma contamination.

Total RNA was isolated using PeqGold RNAPure (PeqLab) and reverse transcripted using the High Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Quantitative real-time PCR was performed on a 7900HT Fast Real-Time PCR system (Thermo Fisher Scientific) using the following Gene Expression Assays on Demand (Thermo Fisher Scientific): OCTN2/SLC22A5, Hs00929869_m1; CPT1A, Hs00912671_m1; CPT1C, Hs00380581_m1; and eukaryotic 18S rRNA endogenous control, 4310893E. Target gene expression was normalized to 18S rRNA and presented as box plots with the median and the 5th/95th percentiles in relation to the average expression in the control samples using ΔΔC_t method.

Preparation of primary GBM cells was done as described previously (26). OCTN2-rich placenta membrane vesicles were prepared as described elsewhere (27). Preparation of lysates from the investigated cell lines as well as the detailed steps of frozen tissue sample homogenization, protein measurement, and subsequent immunoblot analysis are described in the Supplementary Data.

Free LC was measured in tissue samples using the L-carnitine Colorimetric/Fluorometric Assay Kit (BioVision) according to the manufacturer's protocol. Samples were analyzed on a TECAN Infinite M200 (Tecan) multimode reader (excitation: 535 nm, emission: 587 nm). LC concentrations were normalized to the protein content of each sample (measured in the homogenized tissue suspended in PBS) determined by the BCA method.

To achieve hypoxic conditions, LN18 GBM cells were cultivated in a hypoxia chamber (<0.5 Vol.-% O₂; GENbox) for 24 to 96 hours using a hypoxia pad (GENbox anaer Generator, BioMerieux) at 37°C.

Cells were seeded in 96-well multiplates (5,000 cells/well) using 150-μL culture medium containing 0.05% FCS. After 24 hours, cells were incubated with LC alone or in combination with etomoxir, temozolomide, or H₂O₂ under either normoxic or hypoxic conditions as described above. Afterwards, medium was replaced by fresh medium containing 10% resazurine (PromoCell) and cells were incubated for 2 hours at 37°C, before cell viability was analyzed on multimode reader (TECAN Infinite M200, Tecan) by fluorescence measurement (excitation: 530 nm; emission: 590 nm). Viability data were calculated as percentage of solvent-treated cells.

LN18 GBM cells were cultured in 12- and 96-well plates and transfected with OCTN2 or control siRNA (sc-42560/sc-37007, Santa Cruz Biotechnology, 5 pmol/well) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. To enhance knockdown effectivity siRNA transfection was repeated after 24 hours.

The functional activity of OCTN2 in control and OCTN2-siRNA–transfected LN18 cells was analyzed by L-[N-methyl- ³H]-carnitine (³H-carnitine, 80Ci/mmol, American Radiolabeled Chemicals Inc.) uptake assay. Forty-eight hours after transfection cells were washed using prewarmed PBS and incubated with L-[³H]-carnitine in transport buffer (final concentration 1 μCi/mL, 12,5 nmol/L) for 30 minutes. LC uptake was stopped by washing thrice using ice-cold PBS before cells were lysed using 300-μL lysis buffer/well (0.2% SDS, 5 mmol/L EDTA). A total of 150 μL of the lysates were mixed with 2-mL scintillation liquid Rotiszint eco plus (Carl Roth) and analyzed on a β-Counter Tri-Carb 2810TR Low Activity Liquid Scintillation Analyzer (PerkinElmer Inc.). The uptake rates were normalized to the protein content of each sample lysate.

Cells were seeded on 12-well plates (0.1 × 10⁶ cells/well). After preincubation with LC or aqua dest, in medium containing 0.05% FCS for 6 hours, cells were treated with temozolomide, carmustine, vincristine or H₂O₂ for 48 hours. Afterwards, apoptotic activity was studied by measuring caspase-3 activity using the Caspase-3 Fluorometric Assay (R&D Systems) according to manufacturer's instructions.

Cells were seeded in 24-well plates. After reaching confluence, the cell layer was scratched with a 100-μL pipette tip and washed 3 times with PBS. From each scratch 3 randomly distributed images were taken with PALM RoboSoftware on a AxioObserver.Z1 microscope (Zeiss). Afterwards, cells were cultivated with DMEM supplemented with 0.05% FCS, and 5 mmol/L hydroxyurea to prevent cell proliferation. LC was added for 24 hours. The scratches were analyzed as described above after 24 hours at the same positions. Finally, the scratch area was calculated with ImageJ (NIH, Bethesda, MD).

To evaluate OCTN2 as a suitable therapeutic target, we used the murine GBM cell line GL261, which was stereotactically implanted into the brain of C57BL/6J mice. The animal study has been evaluated and approved according to animal welfare guidelines and German laws on animal welfare by the Landesveterinär- und Lebensmitteluntersuchungsamt Mecklenburg-Vorpommern. The detailed procedures and housing conditions are described in the Supplementary Data.

Statistical analyses were performed using SPSS 21.0 (IBM) and GraphPad Prism 6 (GraphPad Software) and considered significant at *, P < 0.05; **, P < 0.01; ***, P < 0.001 level. Data are shown as mean ± SD. The tests used for statistical comparison are indicated in the respective figure legends. Overall survival was calculated from the time of diagnosis until death at the last follow-up. Information on vital status and date of death were obtained from official population registry.

Compared with nonmalignant brain (NMB), a stepwise 2.3- to 4.4-fold increase of OCTN2 mRNA expression was measured in astrocytoma grade III (AC-III) and GBM specimens, respectively (Fig. 1A), while its expression was not elevated in astrocytoma grade II (AC-II) samples. Subdivision into primary and relapsed GBM revealed no significant differences (4.3-fold vs. 4.5-fold, Fig. 1B). In comparison with NMB, OCTN2 mRNA expression was similar in U87MG cells, while LN18 cells had 3-fold higher mRNA levels (Fig. 1B). Immunoblot analysis also revealed a significant upregulation of OCTN2 on protein level in primary GBM (34.5-fold) and relapsed tumor samples (37.2-fold; Fig. 1C and E). LN18 and U87MG had 3.5-fold and about 2-fold higher OCTN2 protein levels compared with NMB, respectively. Expression of OCTN2 was detectable in various human glioma cell lines (A-172, GaMG, HF66, LN18, U251MG, U373, U87MG), in a patient-derived culture of primary GBM cells and in the murine GL261 cell line. OCTN2 expression was validated by a positive control of high OCTN2–expressing placenta vesicles (Fig. 1D). More detailed information on the used cell lines are provided in Supplementary Table S1.

The uptake of LC in cells by OCTN2 is known to depend on sodium (27). To analyze the extent to which LC-uptake into LN18 cells is mediated by OCTN2, uptake of tritium-labeled LC was investigated both in sodium-containing and sodium-free medium as well as in the presence of the OCTN2 inhibitor meldonium (28). The results demonstrate a sodium-dependent LC uptake via OCTN2, which is sensitive to meldonium (Fig. 1F).

Subsequent correlation analyses revealed a positive association between OCTN2 mRNA and protein expression in both primary GBM samples (Spearman r = 0.5327, P < 0.001; Fig. 1G) and relapsed tumor specimens (Spearman r = 0.6108, P < 0.001; Fig. 1H). Furthermore, the tissue content of the OCTN2 substrate LC was analyzed. LC concentration was elevated in both primary (4-fold) and relapsed GBM (5-fold) compared with NMB (P < 0.001 and P = 0.001, respectively; Fig. 1I), but did not correlate with the OCTN2 expression (Fig. 2E and F).

On the basis of the median OCTN2 mRNA/protein expression, the GBM patient cohort was divided into a high and a low expression group. As demonstrated in Fig. 2A and B, a high OCTN2 mRNA or protein expression was associated with a significantly worse survival time. On mRNA level, the median survival time for the low OCTN2–expressing group were 528 days (17.4 months) compared with 268 days (8.8 months) of the high OCTN2–expressing group (Fig. 2A). The unadjusted HR for death was 2.16 (95% CI, 1.25–3.73, P = 0.005) for a high OCTN2 mRNA expression. For OCTN2 protein expression, the survival values were 254 days (8.4 months) for the high OCTN2–expressing group and 579 days (19 months) for the low OCTN2–expressing group [unadjusted HR: 2.69 (95% CI, 1.47–4.91, P < 0.001; Fig. 2B)]. Later, the HR for death was adjusted by fitting the Cox proportional-hazard models. In addition to the stratification factors (extent of surgery, therapeutic regimen), other possible confounding factors such as age and sex were included. The adjusted HR for death in the high OCTN2 protein– expressing group was 2.94 (95% CI, 1.29–6.67, P = 0.01), whereas the adjusted HR for the mRNA dataset did not reach statistical significance (P = 0.06). As a result, the 1-year and 2-year survival rates of patients with GBM also varied markedly according to the OCTN2 mRNA and protein expression levels (Fig. 2C). On the basis of the protein dataset, 1 year after diagnosis, 26% of GBM patients with a high OCTN2 tumor expression were alive compared with 64% with a low intratumoral OCTN2 content. A similar observation was also made 2 years after diagnosis with survival rates of 6% (high OCTN2 expression) versus 30% in the low OCTN2–expressing group.

The LC concentration in primary GBM was not associated with the survival time (Fig. 2D). The median survival time for patients with GBM having a high intratumoral LC level (>median) accounted for 381 days versus 408 days for patients with a lower LC content [<median; adjusted HR for death: 1.01 (95% CI, 0.5–2.04, P = 0.982)].

In addition to the Cox proportional-hazards models, ANOVA II analyses were performed to determine the extent to which OCTN2 expression is influenced by the therapeutic regimen, resection status, age, and sex. For these analyses, our patient cohort was subdivided into 4 age classes (<50 years, 50–59 years, 60–70 years, and > 70 years at diagnosis) with nearly the same group size. The ANOVA II analyses revealed no significant interactions between the OCTN2 mRNA or protein expression and age at diagnosis (Supplementary Fig. S1E and S1F). However, OCTN2 mRNA and protein expression tended to be lower in female patients. In line with the results given in the Cox regression, ANOVA II analysis revealed a significant influence of the OCTN2 expression level on survival time of patients with GBM in each of the 4 age groups [P = 0.04 (mRNA) and P = 0.01 (protein), Fig. 3A and B].

Furthermore, the patient survival time for both the RCTx and radiomonotherapy group is significantly shorter at high OCTN2 levels in the resected GBM samples [vs. low OCTN2 expression level, Fig. 3C and E, P = 0.04 (mRNA) and P = 0.001 (protein)]. A similar effect was observed for the resection status [Fig. 3D and F, P = 0.04 (mRNA) and P = 0.008 (protein)]. It is noteworthy that in patients with a total tumor resection, the effect of OCTN2 expression is somewhat more pronounced in comparison with the subtotal resection group. Regarding this aspect, specific subanalysis via Cox regression revealed an adjusted HR of 4.6 for death in the high OCTN2 protein–expressing group with a total tumor resection (95% CI, 1.63–13.12, P = 0.004), and an adjusted HR of 3.3 for patients having received RCTx (95% CI, 1.31–8.37, P = 0.01).

To investigate the impact of OCTN2 and its substrate LC in GBM cells, in vitro experiments using the LN18 GBM cell line were performed. Cells treated with LC had only a slightly increased viability after 48-hour serum deprivation (116%), which, however, was significantly elevated in a 72-hour period of starvation (142%, Fig. 4A). Furthermore, administration of temozolomide, the chemotherapeutic agent used in standard of care treatment of GBM, caused a decreased viability of LN18 cells cultured in serum-starved medium (78%). This effect was reversed in the presence of 50 μmol/L LC (109%).

Later, the influence of LC on H₂O₂-induced oxidative stress in LN18 cells was evaluated. Coadministration of LC overcompensated (up to 108%) the viability loss of H₂O₂-treated LN18 cells (89%) compared with the respective control cells (Fig. 4C). To subsequently assess whether the cytoprotective effect of LC depends on its function as a mitochondrial fatty acids shuttle and thus on metabolic effects based on β-oxidation, the carnitine palmitoyltransferase-1 (CPT-1) inhibitor etomoxir (ETO) (29) was used. Furthermore, the experiments were carried out under hypoxic conditions in parallel to examine additional hypoxic stress. As shown in Fig. 4D and E, ETO reduced the viability of LN18 cells significantly. The effect of ETO was more pronounced under hypoxic conditions. Under both normoxic and hypoxic conditions, the cell viability loss was partially reversed by administration of 50 μmol/L LC.

In addition, the impact of OCTN2 on GBM cell growth was studied after siRNA-mediated OCTN2 knockdown. siRNA treatment resulted in significantly reduced OCTN2 expression and function as shown by mRNA analysis and ³H-labeled LC uptake experiments (Fig. 4F and G). Subsequently, we determined the influence of OCTN2 knockdown on viability of LC-pretreated (50 μmol/L) LN18 cells (Fig. 4H). Under both normoxic and hypoxic conditions knockdown of OCTN2 caused a significant loss of viability of LN18 cells which was, again, more pronounced in cells confronted with hypoxic stress.

To analyze whether LC protects GBM cells from apoptosis caused either by cytostatics or oxidative stress, caspase-3 activity was used as a marker for apoptotic processes. LN18 cells treated with LC showed no significant alterations in caspase-3 activity (Fig. 5A). In contrast, doxorubicin (7-fold) and carmustine (BCNU, 1.8-fold) led to a significant increase in caspase-3 activity after 48 hours. Coadministration of LC reduced this effect in BCNU-treated cells [1.8-fold (BCNU alone) to 1.2-fold (BCNU + LC)], but not for doxorubicin. Temozolomide and H₂O₂ had no effect on caspase-3 activity. Furthermore, cell migration was analyzed using a wound closure assay since migration of GBM cells into healthy brain is a main feature of this brain tumor. Treatment with BCNU and temozolomide did not alter cell migration, whereas administration of ETO, an inhibitor of β-oxidation, significantly reduced migration of LN18 cells to about 50%. Coadministration of LC did not cause any changes in migration (Fig. 5B). After the detection of ETO-induced inhibition of LN18 cell migration, we assessed expression of 2 key enzymes of the mitochondrial β-oxidation, CPT1A (carnitine palmitoyltransferase 1A) and CPT1C (carnitine palmitoyltransferase 1C), in GBM tissue. CPT1A and CPT1C showed a similar expression pattern in NMB compared with GBM specimens (Supplementary Figs. S2A and S2B). Kaplan–Meier analyses revealed no significant influence of CPT1A or CPT1C expression on patient survival indicating a limited prognostic value in GBM (Supplementary Figs. S2C and S2D).

To finally show proof of the concept of the OCTN2/LC system as a therapeutic target for GBM treatment, we analyzed the in vivo tumor growth in an orthotopic murine GBM model using the cell line GL261, which was stereotactically injected into mice brain. Twelve days postinjection, tumor development was assessed by MRI. Afterwards, mice were treated with either 0.9% NaCl (control animals) or 250 mg/kg meldonium as an OCTN2/LC inhibitor (28) every day intraperitoneally until day 26 postinjection. To ensure normality of data, tumor volumes were log-transformed. Also, tumor growth was calculated as the difference between the transformed volumes. As seen in Fig. 5C–E at day 26 postinjection (14 days after starting treatment), a decrease in mean tumor size was observed in the meldonium-treated group (−2,366 log₁₀[cm³]) compared with the starting time point (−2,306 log₁₀[cm³]). In contrast, control animals showed significant progress in tumor size from −2,385 log₁₀[cm³] (treatment starting point) to −1,618 log₁₀[cm³] at day 26, which was significantly different from the corresponding time point of the meldonium group (P < 0.05). This significant difference was also present in the growth rate calculations of the tumors (Fig. 5E), which showed a mean tumor growth rate of 0,7671 log₁₀[cm³] in control animals and −0,06014 log₁₀[cm³] in meldonium-treated animals (P = 0.01). Three representative MRI recordings for the starting time point (12 days postinjection) and at day 26 postinjection per group (low, mid, high tumor growth) are shown in Fig. 5F.

Therapy failure caused by GBM resistance is the most common reason of tumor recurrence resulting in very poor survival rates. It has been previously shown that the OCTN2 substrate LC acts as a cytoprotector promoting cellular resistance and survivability (7, 30, 31). In this study, we demonstrate that OCTN2 expression and tissue LC concentrations are significantly higher in GBM compared with healthy brain. We observed a stepwise increase of OCTN2 expression from low-grade astrocytoma to primary and recurrent GBM. These findings are consistent with those of Bayraktar and colleagues who described a gradual increase in LC levels and some of its acyl derivates from low- to high-grade astrocytomas through to primary GBM (32).

In a recent study, Singer and colleagues described the development of therapeutic resistance in GBM stem cells by enhanced expression of an antioxidant response system referred to as Xc catalytic subunit xCT (SLC7A11) that partially operates through activation of the NRF2 transcriptional network (19). This network is also addressed by LC (15, 31) and regulates antioxidant response element (ARE)-containing genes (33) leading to attenuation of ROS-mediated cell damage, arrest of apoptosis, and resistance to radiochemotherapy (31, 34). ROS homeostasis is strictly regulated by cancer cells to promote tumorigenesis and malignant transformation (17, 35, 36). OCTN2 upregulation and LC accumulation could therefore help GBM cells to maintain conducive ROS levels both to drive mutagenesis and to mitigate unfavorable conditions evoked by ROS-generating processes, for example, high metabolic turnover and radiochemotherapy (19, 20).

These considerations are underlined by our findings that LN18 GBM cells, a well-accepted WHO grade IV GBM cell line (37), become more resistant or at least maintain their viability under hypoxic, metabolic, and cytotoxic stress when pretreated with LC. The results are consistent with several in vitro and in vivo studies on LC's antioxidant and antiapoptotic activities, which have been predominantly studied in nonmalignant compromised cells in context of neurodegenerative disorders (10, 11), but are also detectable in the malignant neuroblastoma cell line SH-SY5Y (12). However, LC enhanced the viability of the human LN18 GBM cells, but had no detectable effect on cell migration. Following administration of LC, we also observed a decreased caspase-3 activity in LN18 cells treated with BCNU and a restored viability in cells treated with temozolomide, both drugs used in the therapeutic regimen for patients with GBM. These results endorse the theory that LC strengthens the tumor cells through antiapoptotic mechanisms or perhaps through a combination of several cytoprotective effects. However, the cytoprotective LC concentrations given in literature are widely spread within a range of 9–25 μmol/L (38), 30–100 μmol/L (39), 0.1–1 mmol/L (40), and 1–10 mmol/L (41), respectively. In our current work, a LC concentration of 50 μmol/L was found to be adequate in detecting cytoprotective effects. Lower LC concentrations showed no significant benefit in our in vitro experiments while higher LC concentrations did not increase the survival-promoting effects (data not shown) even though other studies claim a concentration dependent effect of LC (6, 12, 30). Of note, the LC concentration used in our cell experiments was within the physiologic range of the LC plasma concentration, which is generally estimated to be 40–50 μmol/L (42).

Recent NMR analyses of in situ glioma specimens as well as further in vitro and in vivo studies suggest that oxidative metabolism plays an integral role of cellular maintenance and proliferation in malignant glioma cells (5). Because LC is crucial for transporting middle- and long-chain fatty acids into mitochondria to drive β-oxidation, some consideration had to be given to how LC's metabolic effects could influence the tumor cells' behavior. We used the irreversible CPT1 inhibitor etomoxir (ETO) to block β-oxidation to analyze whether GBM cells rely on fatty acid oxidation and to what extent cytoprotection is attributable to LC when uncoupled from its native function as a mitochondrial shuttle. ETO caused both a significant loss of cell viability and migration in LN18 cells, which has been also shown in primary GBM cells (5, 29). Under both normoxic and hypoxic cell culture conditions, the loss of viability, but not the ETO-induced inhibition of migration, was abolished by simultaneous LC application. These results provide 3 valuable hints: First, LN18 relies to a remarkable extent on fatty acid metabolism as bioenergetic fuel and lacks viability and migration when β-oxidation is blocked. Second, LC stabilizes cell integrity independently of its metabolic function, but rather by its abovementioned extramitochondrial cytoprotective effects (31, 34). Third, LC did not reverse or alleviate the ETO-induced inhibition of migration suggesting that LC has no perceptible influence on GBM cell migration based on metabolic effects. It is quite possible, however, that malignant GBM cells gain survival advantages on both cytoprotective and metabolic levels through OCTN2 overexpression and LC accumulation, as it is also assumed that fatty acid oxidation provides NADPH for defense against oxidative stress and cell death in GBM cells (29). Nevertheless, the molecular details of malignant transformation and resistance of tumor cells need to be further elaborated in this context.

To our knowledge, no association studies on the role of the OCTN2/LC system in GBM survival have been published to date. Here, after adjustment of the HR on therapeutic regimen, resection status, age, and sex, we found a high OCTN2 protein expression to correlate with a 2.9-fold increased risk for death in patients with primary GBM. However, it has to be mentioned that further potential confounders (e.g., ECOG Performance Status, Karnofsky Index) were not recorded. The ANOVA II analyses revealed that differences in the patient survival between the low and high OCTN2–expressing group were highly pronounced in the respective subpopulation receiving RCTx and a total tumor resection, suggesting a prognostic relevance especially in holistically treated patients.

Although OCTN2 expression has been described in various tumor cells (43–45), its precise role in malignant transformation remains elusive. Contrary to our observations, Scalise and colleagues found a downregulation of OCTN2 expression in epithelial cancer cell lines, whereas an estrogen-mediated induction of OCTN2 expression in breast cancer was described (43, 44). For GBM, however, our data of increased OCTN2 expression may provide both a prognostic factor for GBM treatment as well as an option for optimizing GBM therapy by either specific blocking of OCTN2 or drug targeting via OCTN2. Very recently, it was shown that LC-conjugated nanoparticles promote permeation across the blood–brain barrier to target glioma cells via OCTN2 resulting in improved antiglioma therapy (46). Our preclinical studies using an orthotopic GBM mouse model demonstrate significantly reduced intracerebral tumor growth through inhibition of the OCTN2/LC system and suggest it as a potential target in GBM therapy. To date, our in vivo study is the first to show the antitumoral efficacy of the OCTN2/LC inhibitor meldonium. While no major side effects of meldonium have been observed so far, meldonium has even been shown to improve cardiovascular function (47, 48).

Some limitations of our study need to be addressed. Our work represents a global expression analysis of the entire tumor mass without tumor region or cell type–specific differentiation, especially considering the heterogeneity of GBM. Taking into account the different types of GBM and their distinct genetic and epigenetic profiles (49), predictive molecular biomarkers for therapy response and outcome may vary in respective GBM subtypes. Our patient cohort was not subclassified and we therefore cannot draw any conclusion regarding differences in GBM subtypes. In our study, only vital tumor samples from the edge region were used. Within the same tumor, hypoxic, necrotic, and highly vascularized areas vary in their expression profiles and thus also in their OCTN2 expression. In addition, the use of only 1 cell line for in vitro and in vivo studies is a limitation of our investigations, arguing toward analyses in further human cell lines or primary GBM cells to validate the observed effects.

Nevertheless, this study stresses the role of OCTN2 and its substrate LC in GBM cytoprotection, suggests OCTN2 as a prognostic marker and offers a novel target that may aid in combatting tumor resistance or increasing its sensitivity to therapy. Although the precise mechanisms of the OCTN2/LC system in GBM biology remain to be completely determined, the promising results of meldonium in reducing tumor growth provide strong support for further research on the OCTN2/LC system, which is aimed at improving antiglioma therapy.

No potential conflicts of interest were disclosed.

Conception and design: M.A. Fink, H. Paland, S. Bien-Möller

Development of methodology: M.A. Fink, H. Paland

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.A. Fink, H. Paland, S. Herzog, M. Grube, S. Vogelgesang, A. Bialke, W. Hoffmann, H.W.S. Schroeder, S. Bien-Möller

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.A. Fink, H. Paland, K. Weitmann, W. Hoffmann, B.H. Rauch, S. Bien-Möller

Writing, review, and/or revision of the manuscript: M.A. Fink, H. Paland, M. Grube, S. Vogelgesang, K. Weitmann, A. Bialke, W. Hoffmann, B.H. Rauch, S. Bien-Möller

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.A. Fink, H. Paland, S. Bien-Möller

Study supervision: M.A. Fink, H.W.S. Schroeder, S. Bien-Möller

This work was supported by national funding from the Forschungsverbund Neurowissenschaften. M.A. Fink received a scholarship of the Gerhard Domagk Program, funded by the Faculty of Medicine, Greifswald, Germany, and was further supported by the Germany Scholarship, funded by the German Federal Ministry for Education and Research.

We thank Prof. N. Hosten, head of the Institute for Radiology and Neuroradiology (University Medicine Greifswald) for providing the 7 Tesla MR imaging scanner and Stefan Hadlich for the excellent technical assistance with the MR imaging.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.