Myoinositol to total choline ratio in IDH wild-type gliomas as a prognostic factor on preoperative magnetic resonance spectroscopy

Background: Isocitrate dehydrogenase (IDH) wild-type gliomas tend to be pathologically de�ned as glioblastomas. We previously reported that, unlike IDH-mutant gliomas, IDH wild-type gliomas showed signi�cantly lower ratios of myoinositol to total choline (i.e., the Ins/Cho ratio) on magnetic resonance (MR) spectroscopy. Given that IDH-mutant gliomas also have much better prognoses than IDH wild-type gliomas, we hypothesized that this lower Ins/Cho ratio is associated with malignancy in adults with supratentorial gliomas. Therefore, we calculated the Ins/Cho ratios of patients with supratentorial IDH wild-type gliomas and investigated their progression free survival (PFS) and overall survival (OS) to determine its utility as a prognostic marker. Methods: We classi�ed IDH wild-type gliomas (n = 30) into two groups based on the Ins/Cho ratios, and compared patient backgrounds, pathological �ndings, PFS, OS, and copy number aberrations. Results: Compared with the group with high Ins/Cho ratios, the group with low Ins/Cho ratios had shorter PFS (P = 0.020) and OS (P = 0.037) durations. Multivariate analysis demonstrated that the Ins/Cho ratio correlated signi�cantly with PFS (hazard ratio 0.34, P = 0.027). Conclusion: We conclude that the preoperative Ins/Cho ratio can be used as a novel prognostic factor for IDH wild-type gliomas.


Background
Gliomas, which are diagnosed based on pathological and genetic ndings, are among the most common brain tumors [1][2][3][4].Extensive, surgical resection is the preferred initial therapeutic strategy to improve prognosis [5,6]; however, such an approach also increases the risk of brain dysfunction because either the border between the tumor and normal brain tissue is unclear or because tumors in ltrate invasively into adjacent healthy brain tissue.Therefore, the ability to predict tumor malignancy preoperatively could help surgeons to plan optimal surgical strategies.

Magnetic resonance (MR) spectroscopy enables us to quantify tumor metabolites noninvasively by
analyzing their spectra, and thus, it is widely used for preoperative diagnosis of brain tumors [7].We previously examined tumor metabolites in adult supratentorial gliomas and reported that the myoinositol to total choline (Ins/Cho) ratios in isocitrate dehydrogenase (IDH) wild-type gliomas were signi cantly lower than in IDH-mutant gliomas [8].Given that total choline is thought to represent cell density [9], we presumed that lower ratios indicate myoinositol consumption in a glioma cell.Myoinositols are sugar alcohol in phospholipids of the intracellular membrane [10], and they are produced through food digestion or by hydrolysis or dephosphorylation of the intracellular membrane [10][11][12].In astrocytes, myoinositols contribute to adjusting osmotic pressure in response to changes in intracranial pressure [13].
Most supratentorial gliomas harboring wild-type IDH are pathologically classi ed as glioblastomas and tend to have dismal prognoses [1][2][3].Molecular factors that contribute to this poor prognosis have been identi ed based on genetic subtype, including TERT promoter mutations, copy number aberrations (e.g., +7, 10q and 9p21), and epigenetic changes (e.g., histone H3) [14][15][16][17][18].However, such information is only available after surgical resection, and we ideally need a method for predicting outcomes in patients suspected of having glioblastomas before surgery.Based on the observation that supratentorial gliomas with IDH mutations have better outcomes than those with wild-type IDH [1,2,3], we hypothesized that low Ins/Cho ratios in IDH wild-type gliomas can be used as a prognostic marker for this genetic subtype of glioma.
In the present study, we calculated Ins/Cho ratios by MR spectroscopy for adults with supratentorial IDH wild-type gliomas to investigate the ratio's association with prognosis, its correlation with previously reported prognostic factors, and whether it could be used as a preoperative prognostic factor.

Study design
We retrospectively investigated 30 cases of newly diagnosed IDH wild-type glioma in patients analyzed by MR spectroscopy between 2013 and 2018 at Fujita Health University hospital.Patients who had already received treatment before MR spectroscopy or who were younger than 20 years old were excluded.
The primary outcomes were the progression free survival (PFS) and the overall survival (OS).PFS was calculated from the date of rst operation to the date of con rmed recurrence.OS was calculated from the date of rst resection to the date of death.Patient information was last updated in April 2019 for follow-up purposes.All resected tissues were assessed by neuropathologists according to the WHO classi cation [1].

MR spectroscopy
A single-voxel 1 H-MR spectroscopy with point-resolved spectroscopy sequence was performed with a 3 Tesla (T) scanner (Ingenia 3T; Philips Healthcare, Best, The Netherlands) using a dS Head coil and Vantage Titan 3T (Canon Medical Systems Corporation, Otawara, Japan) using a 16 or a 32-channel coil.
In each patient, 1 H-MR spectroscopy was performed using the following parameters: repetition time (TR)/echo time (TE), 2000/144 and 35 ms; number of excitation (NEX), 128; bandwidth, 1.61 HZ/point; voxel of interest (VOI) size for metabolic measurements, 15 × 15 × 15 mm.T2 weighted images in 3 directions for setting the VOI were determined for each patient by means of the following parameters: repetition time (TR)/echo time (TE), 4250/82 ms; acquisition matrix size, 416 × 344; reconstruction matrix, 640 × 640; eld of view (FOV), 230 × 230 mm; slice thickness, 4.0 mm; slice gap, 0.8 mm; number of excitation (NEX), 1; reduction factor 1.9.The VOI was selected by a board certi ed neuroradiologist (K.M) with an experience of 10-15 years' to include the VOIs inside the lesion based on T2 weighted images in 3 directions.When the tumor contains necrotic and cystic components, these components were included within VOI to accurately evaluate tumor characteristics.In 1 H-MR spectroscopy, myoinositol and total choline were measured with short TE (TE = 35 ms) and long TE (TE = 144 ms) respectively [19,20,21] because of their relaxation time in T2, and mean concentration of MR spectroscopy values were analyzed by automatic quanti cation program (LCModel; Stephen Provencher, Oakville, Ontario, Canada) [22].

Evaluation of IDH mutation status
We evaluated IDH mutations using the Sanger method, with codon 132 in IDH1 and codon 172 in IDH 2 analyzed by polymerase chain reaction (PCR).Brie y, DNA was extracted from resected frozen tissue and formalin-xed and para n-embedded tissue (FFPE), using DNeasy blood and tissue kits (QIAGEN, Hulsterweg, Netherland) and an REPLI-g FFPE Kit (QIAGEN).The reaction mixtures for PCR comprised DNA, primers, 10PCR buffer, 10mM dNTP mix (Thermo Fisher Scienti c, Waltham, MA, USA), 50 mM MGCL , and PLATINUM TagDNA polymerase (Thermo Fisher Scienti c).After we con rmed the DNA bands of the PCR products in electrophoresis, we added BigDye Sequencing Buffer (Thermo Fisher Scienti c), Ready Reaction Mix (Thermo Fisher Scienti c), and the same primer to the PCR products, and repeated the PCR.Sequencing was performed with a BigDye Terminator version 3.1 Cycle Sequencing Kit (Thermo Fisher Scienti c) and results were analyzed on an ABI 3100 (Applied Biosystems, Waltham, MA, USA).

Next generation sequencing analysis and comparison of Copy number aberration
We analyzed samples that met our inclusion criteria and classi ed them by Ins/Cho ratio into high and low and a low ratio groups.We prepared diluted genomic DNA for subsequent experiments.After whole genome ampli cation using a SurePlex DNA Ampli cation System (Illumina, San Diego, CA, USA), library preparation was performed with a Nextera XT DNA Library Preparation Kit (Illumina).Sequencing analysis was conducted with a VeriSeq PGS Kit-MiSeq (Illumina), and results were analyzed with BlueFuse Multi Software (Illumina).In all cases, chromosomes were divided into 2500 windows of approximately 1 Mb in size.

Statistical analysis
We used Fisher's exact test and the MannWhitney U test to compare age at onset, sex, laterality, MIB-1 index, and pathology between the two groups.PFS and OS were analyzed using the KaplanMeier method and compared with the log-rank test.Cox proportional hazards models were used to determine the relationship between the Ins/Cho ratio and prognosis.Given that most patients were treated with surgery, temozolomide and radiotherapy, and bevacizumab, we selected the following as explanatory factors in the multivariate analysis: gross total resection; subtotal resection, which was de ned as >90% resection; temozolomide and radiotherapy; bevacizumab; and the Ins/Cho ratio.All statistical analyses were conducted using EZR [23].

Relation of patient and pathology characteristics with the Ins/Cho ratio
The mean and median values of the Ins/Cho ratios for the 30 patients included in this cohort were 0.75 and 0.67, respectively.We opted for 0.7 as the cutoff value and classi ed those patients into high (≥0.7)and low (<0.7)Ins/Cho ratio groups (n = 13 and 17, respectively).Tables 1 and 2 show the patient characteristics, pathologies, Ins/Cho ratios, and postoperative therapies.Case comprised 25 glioblastoma, two gliosarcomas, three astrocytomas.Two of three astrocytomas gained chromosome 7 and/or loss of chromosome 10q, but the other was not associated with them.There were no statistically signi cant differences in sex, age at onset, laterality, MIB-1 index, or grade WHO classi cation between the two groups.Overall, eight patients underwent reoperation, six patients underwent additional chemotherapy, and seven patients had no treatment after recurrence.

PFS and OS analysis
We compared prognosis by the Ins/Cho ratio.During follow-up, tumor recurrence occurred in 94% of patients with low Ins/Cho ratio (i.e., 16/17) and in 77% with high Ins/Cho ratio (i.e., 10/13).Moreover, 82% (14/17) of patients with low ratios and 38% (5/13) with high ratios died.Figures 1a and 1b show that the PFS and OS in those with high ratios were signi cantly shorter than in those with low ratios, indicating higher recurrence and mortality rates in the patients with low Ins/Cho ratios (P = 0.020 and 0.037, respectively).In the multivariate analysis, Cox proportional hazards models revealed that the Ins/Cho ratio was signi cantly associated with PFS (hazard ratio 0.34, P = 0.027), which suggested that the Ins/Cho ratio was useful outcome predictor, especially for PFS.These data are summarized in Tables 3a and 3b.3a and 3b show the results of multivariate analysis: we chose GTR or STR, BEV, TMZ and RT, and the Ins/Cho ratio as explanatory factors and analyzed their relevance to PFS and OS.Abbreviations: BEV, bevacizumab; GTR, Gross total resection; Ins/Cho, ratio of myoinositol to total choline; OS, overall survival; PFS, progression free survival; RT, radiotherapy; STR, subtotal resection; TMZ, temozolomide.

Analysis of copy number aberration by next generation sequencing
We included 20 cases in the next generation sequencing, analyzing nine samples in the group with high Ins/Cho ratios (≥0.7) and 11 samples in the group with low Ins/Cho ratios (<0.7).The overall noise value was <0.3.Analysis focused on regions where speci c genes exist that are associated with poor prognosis in glioblastoma, including 7p11.2 (EGFR), 9p21.3 (p16), 10q23.3(PTEN) [15,16].Gain of 7p11.2 was detected in seven cases with high ratios and in six cases with low ratios, loss of 9p21.3 was detected in four cases with high ratios and in three cases with low ratios, and loss of 10q23.3 was detected in four cases with high ratios and in three cases with low ratios (Table 4).There were no signi cant differences between the groups in any other window.Copy number aberrations identified by next generation sequencing are compared by the Ins/Cho ratio.We defined an effective change of gain or loss when the average value of a window was >2.5 or <1.5.
Abbreviations: Ins/Cho, ratio of myoinositol to total choline.

Discussion
In this study, we demonstrated that a high Ins/Cho ratio on MR spectroscopy was associated with a better prognosis than a low Ins/Cho ratio according to KaplanMeier curves.The Ins/Cho ratio was signi cantly associated with PFS by KaplanMeier and multivariate analysis.Of the three cases in which the Ins/Cho ratios exceeded 1.0 and that started the Stupp regimen soon after maximum safe resection, the PFS durations were 26, 17, and ≥39 months (Cases 7, 8, and 26, respectively).Given that total choline in MR spectroscopy reveals cell density [9], low myoinositol levels are considered associated with poor prognosis.Indeed, a recent cohort study of recurrent glioblastoma treated with bevacizumab reported that a high myoinositol value on MR spectroscopy was associated with a good prognosis [24].Given all these results, we considered that myoinositol affects as a second messenger and has antitumor effects.Myoinositols are located in glial cells, especially in astrocytes [25], and regulate intracranial osmotic pressure [13].Phosphatidylinositol 3-phosphate, which contains myoinositols in its structure, is produced by intracellular membrane metabolism [10] and activates the PI3K-Akt pathway as a second messenger [26].Activation of this pathway may cause a poorer prognosis in patients with a low Ins/Cho ratio by prompting cell proliferation and myoinositol consumption.As shown in a previous study, oral administration of myoinositols can inhibit malignant transformation of tumor cells in patients suffering non-small-cell lung cancer, in which the PI3K-Akt pathway is important [27].Other researchers also reported that myoinositols have antitumor effects that result from their phosphorylated metabolites, such as inositol 1,3,4,5,6-pentaphosphate and inositol hexaphosphate, which can induce tumor cell apoptosis [28,29].These previous studies suggest that malignancy in IDH wild-type gliomas with low Ins/Cho ratios is associated with a simple reduction in antitumor effects.
However, multivariate analysis only showed that the Ins/Cho ratio was signi cantly associated with PFS and not OS (Figure 1, Table 3).We assumed that the therapeutic strategies used after recurrence affected these results in the multivariate analyses.Indeed, some patients chose both surgical resection and additional chemotherapy (13%), whereas other patients opted for no additional therapy after recurrence (23%), potentially affecting outcomes.Notably, all patients younger than 50 years at diagnosis were in the group with a low Ins/Cho ratio, and 60% of these chose surgical resection and chemotherapy (Table 1).Two of these patients (cases 1 and 3) survived for approximately 2 years despite having low Ins/Cho ratios (0.57 and 0.51, respectively).By contrast, 20% of patients older than 60 years at diagnosis underwent surgical resection, and only 4% of these chose additional chemotherapy.Patients who opted for no therapy after recurrence had short survival time (e.g., cases 15 and 27) despite having relatively high Ins/Cho ratios (0.85 and 0.69, respectively).
Copy number analysis by next generation sequencing revealed no signi cant correlations between the Ins/Cho ratio and chromosomal areas, despite the association of speci c copy number aberrations with glioblastoma (Table 4).Previous studies have shown that such copy number aberrations are an early stage of tumorigenesis in glioblastoma [15,30,31], and our results suggest that low myoinositol levels are less likely to be caused by decreased myoinositol synthesis due to chromosomal change and gene expression.Rather, the low myoinositol levels can be considered to result from consumption due to tumor growth.
This study is limited by its retrospective cohort design and lack of control of therapeutic strategies (e.g., resection extent by tumor location, postoperative bevacizumab use, and radiological dosages) or postoperative complications (e.g., delayed wound healing or high-fever affected the start of adjuvant therapy).These factors could have signi cantly affected the clinical courses and outcomes.We must now investigate more cases to clarify the putative correlations between the Ins/Cho ratio and patient prognosis.

Conclusions
The noninvasive Ins/Cho ratio can serve as a novel prognostic marker for adults with supratentorial IDH wild-type gliomas, providing useful preoperative information in patients with suspected glioblastomas.
However, we can only speculate on why myoinositols are associated with patient outcomes, and questions around this will be targeted in future research.

Table 1 .
Characteristics and treatments of patientsCase Age, sex Pathology Recurrence PFS Follow-up terms Outcome Ins/Cho

Table 2 .
Comparison of patient characteristics by Ins/Cho ratio threshold

Table 3a .
Multivariate analysis of PFS

Table 4 .
Comparison of copy number aberrations by Ins/Cho ratio