Human IDH mutant 1p/19q co-deleted gliomas have low tumor acidity as evidenced by molecular MRI and PET: a retrospective study

Co-deletion of 1p/19q is a hallmark of oligodendroglioma and predicts better survival. However, little is understood about its metabolic characteristics. In this study, we aimed to explore the extracellular acidity of WHO grade II and III gliomas associated with 1p/19q co-deletion. We included 76 glioma patients who received amine chemical exchange saturation transfer (CEST) imaging at 3 T. Magnetic transfer ratio asymmetry (MTRasym) at 3.0 ppm was used as the pH-sensitive CEST biomarker, with higher MTRasym indicating lower pH. To control for the confounder factors, T2 relaxometry and l-6-18F-fluoro-3,4-dihydroxyphenylalnine (18F-FDOPA) PET data were collected in a subset of patients. We found a significantly lower MTRasym in 1p/19q co-deleted gliomas (co-deleted, 1.17% ± 0.32%; non-co-deleted, 1.72% ± 0.41%, P = 1.13 × 10−7), while FDOPA (P = 0.92) and T2 (P = 0.61) were not significantly affected. Receiver operating characteristic analysis confirmed that MTRasym could discriminate co-deletion status with an area under the curve of 0.85. In analysis of covariance, 1p/19q co-deletion status was the only significant contributor to the variability in MTRasym when controlling for age and FDOPA (P = 2.91 × 10−3) or T2 (P = 8.03 × 10−6). In conclusion, 1p/19q co-deleted gliomas were less acidic, which may be related to better prognosis. Amine CEST-MRI may serve as a non-invasive biomarker for identifying 1p/19q co-deletion status.


Background
The role of molecular markers in stratifying brain tumors has gained increasing awareness in the past decade. In 2016, the World Health Organization (WHO) has revised the classification criteria of gliomas to incorporate molecular markers into diagnosis, instead of relying solely on histological phenotypes 1 . In this updated guideline, the definition of oligodendroglioma is defined by two genotypic features: the mutation in isocitrate dehydrogenase (IDH), as well as the co-deletion of the short arm of chromosome 1 (1p) and the long arm of chromosome 19 (19q). The histopathological diagnosis of oligoastrocytoma, which suffered from high interobserver discordance, has been largely abandoned with the adoption of the more robust molecular classification. Co-deletion of 1p/19q has been reported to be present in about 60-90% of histopathologically diagnosed oligodendroglioma and 30-50% of oligoastrocytoma 2,3 . In addition to its diagnostic value, 1p/19q co-deletion is also associated with better response to radiotherapy and alkylating agent chemotherapy, and longer progression-free and overall survival 4 .

Figure 2.
Comparison of MTR asym between 1p/19q co-deleted and non-co-deleted gliomas in T 2 hyperintense lesions. (a-e) Show the comparison within all patients, WHO grade II gliomas, grade III gliomas, gliomas with oligodendroglial component, and IDH mutant gliomas, respectively. Each patient data point is represented by a dot and the violin plots represent the distribution of the patient data. In all patient cohorts, significantly lower MTR asym is observed in 1p/19q co-deleted gliomas compared to 1p/19q non-co-deleted gliomas (**, P < 0.01; ***, P < 0.001; ****, and P < 0.0001). The ROC analyses for differentiating 1p/19q co-deletion status in patients with WHO grade II and/or III gliomas using MTR asym are demonstrated in (f). The colored dots represent the optimal operating points.  Fig. 4a). Also, T 2 within tumor ROI was not significantly different (co-deleted, 0.11 s ± 0.02 s; non-co-deleted, 0.11 s ± 0.03 s, P = 0.61, Fig. 4b).
In order to further isolate the effect of 1p/19q co-deletion on MTR asym from the influence of tissue transverse relaxivity (T 2 ) and amine concentration (nFDOPA), we performed ANCOVA (Fig. 4c). It was shown that 1p/19q co-deletion status was the major contributor of the variability in MTR asym (P = 2.91 × 10 −3 ), when controlling for nFDOPA (P = 0.053) and age (P = 0.46). The same result was obtained when controlling for T 2 and age (MTR asym , P = 8.03 × 10 −6 ; T 2 , P = 0.51; and age, P = 0.24). We did not perform ANCOVA with inclusion of nFDOPA, T 2 , and age into controlling factors all together due to the small number of patients with both FDOPA and T 2 measurements (N = 10, degree of freedom for error = 6).
1p/19q co-deleted and non-co-deleted gliomas demonstrated different MR-PET characteristics. Figure 5 shows the MR images (post-contrast T 1 -weighted image, FLAIR image, MTR asym at 3.0 ppm, and T 2 map) and PET images ( 18 F-FDOPA PET) of four patients with different grades and 1p/19q co-deletion statuses as representative examples. Figure 5a shows a newly diagnosed oligodendroglioma patient, with IDH mutation and 1p/19q co-deletion. The tumor was generally homogeneous, with low MTR asym and moderate nFDOPA. The second patient (Fig. 5b) was diagnosed with recurrent anaplastic oligodendroglioma (IDH mutant, Figure 3. Comparison of median MTR asym between grades and between IDH mutant and IDH wild-type gliomas. Grade III gliomas shows significantly higher MTR asym compared to grade II gliomas (a). However, when comparing within 1p/19q non-co-deleted gliomas (b) or co-deleted gliomas (c), the difference is not significant. MTR asym is also significantly higher in IDH wild-type gliomas compared to mutant gliomas (d).
Scientific RepoRtS | (2020) 10:11922 | https://doi.org/10.1038/s41598-020-68733-5 www.nature.com/scientificreports/ 1p/19q co-deleted), with a resection cavity in the left frontal lobe visible in the anatomic MR images. The FLAIR hyperintense lesion ipsilateral to the resection cavity exhibited low MTR asym as well as low nFDOPA. Meanwhile, the lesion on the contralateral side showed high nFDOPA and moderate MTR asym , indicating the existence of active tumor tissue. The voxels in this lesion were situated on the right "tail" part of the scatter plot. Examples shown in both Figs. 5c,d are IDH mutant and 1p/19q non-co-deleted, both characterized by tumor regions with relatively normal nFDOPA but high MTR asym at 3.0 ppm. From these examples, we observed substantially different MR-PET characteristics in 1p/19q co-deleted and non-co-deleted gliomas. Hence, amine CEST and FDOPA may provide complementary information regarding the reprogrammed metabolism of gliomas.

Discussion
Results confirmed that CEST contrast measured as MTR asym at 3.0 ppm has similar dependence pattern for two different α-amino acids, glycine and phenylalanine, indicating that 18 F-FDOPA can serve as a surrogate marker of tissue amine concentration. The simulation study showed that MTR asym is dependent on amine concentration and T 2 relaxation time in addition to pH. Further investigation on patients suggested that 1p/19q co-deleted gliomas have lower acidity compared with intact gliomas, as indicated by significantly lower MTR asym at 3.0 ppm and no difference in amine concentration or T 2 relaxation rate. The lower acidity seemed to be specifically associated with the loss of heterozygosity of chromosome 1p and 19q, since the difference in MTRasym was consistently observed within grade II, grade III, histological phenotype with oligodendroglial component, and IDH mutant gliomas. These results support the increasing understanding that molecular biomarkers may define a more homogeneous patient population than histopathological features 1 . The less acidity in tumor microenvironment of 1p/19q co-deleted gliomas revealed by this study is also consistent with the better prognosis and higher sensitivity to therapies reported in patients with 1p/19q co-deleted gliomas, because acidic tumor microenvironment has been shown to correlate with tumor malignancy through a number of mechanisms 7,21 . www.nature.com/scientificreports/ However, the mechanism underlying low extracellular acidity associated with 1p/19q co-deleted oligodendroglioma is still largely unknown. Traced back to the tumor origin, oligodendrocytes were shown to release less lactate than astrocytes, despite the similar level of glucose uptake 14 . The NSCs and OPCs also exhibit differential preference of energy production pathways, with the former displaying high rate of glycolytic flux and the latter favoring a higher rate of mitochondrial metabolism 13 . To the best of our knowledge, little research has been done to understand the effect of 1p/19q co-deletion on tumor microenvironment. Blough et al. found that a pH regulator, NHE-1, is silenced in oligodendroglioma subsequently to IDH-associated DNA hypermethylation and 1p www.nature.com/scientificreports/ allelic loss, and consequently, impairs the ability of tumor cells to pump out the intracellular H + increased by the Warburg glycolytic shift 10 . Another possible factor underlying lower acidity associated with 1p/19q co-deletion is the somatic mutations in the CIC gene located on chromosome 19q13.2, which were found in approximately 70% of 1p/19q co-deleted oligodendrogliomas 22 . Chittaranjan et al. showed that mutations in CIC upregulate 2-hydroxyglutarate (2-HG) levels cooperatively with IDH1 mutation 11 , which might exaggerate metabolic changes induced by 2-HG due to IDH mutation. The effects of 2-HG include the silencing of LDHA 23 and the increased activity of EGLN 24 , both contributing to a less glycolytic phenotype. We propose that more studies are required to comprehensively understand the biological basis of the less acidic microenvironment related to 1p/19q co-deletion, which may shed light on a new metabolic source of vulnerability in gliomas and potential treatment target.
Our results also demonstrate that amine CEST provides a unique imaging contrast and may serve as a quick non-invasive imaging biomarker for identifying 1p/19q co-deleted gliomas, with high sensitivity (93%), moderate specificity (71%), and AUC of 0.85. Previous studies investigating anatomic imaging characteristics showed that 1p/19q co-deleted gliomas tend to show heterogeneous signal intensity, indistinct margin, calcification 25 , and absence of T 2 -FLAIR mismatch (sensitivity = 22-45%, specificity = 100%) 26 . Other studies used advanced MRI and metabolic imaging, showing that 1p/19q co-deleted gliomas have higher relative cerebral blood volume (AUC = 0.68) 27 and increased uptake of 18 F-fluorodeoxyglucose (sensitivity = 75%, specificity = 100%) 28 , 18 F-fluoro-ethyl-tyrosine (sensitivity = 62%, specificity = 83%) 29 , and 11 C-methionine 30 . Meanwhile, 18 F-FDOPA uptake was found to be uncorrelated with 1p/19q co-deletion status 31 , consistent with our result. Branzoli et al. found that magnetic resonance spectroscopy is able to detect accumulation of cystathionine in 1p/19q co-deleted gliomas in vivo, which is related to the lower expression of both PHGDH and CTH compared with their nonco-deleted counterparts, leading to perturbed serine-and cystathionine-metabolism 9 . Compared to the other imaging methods, amine CEST demonstrated higher sensitivity and has the advantage of revealing biological information without the need of injecting contrast agents or radioactive tracers, while having a better spatial resolution than spectroscopy-based methods.
We acknowledge that there are specific limitations to the current study. First, only 20 patients in our study cohort had available 18 F-FDOPA scan data. Our data may not have had enough statistical power to reveal potentially increased amino acid PET uptake and 1p/19q co-deletion, as suggested by the 11 C-MET and 18 F-FET studies, both of which included more than 100 patients. However, our results suggesting 1p/19q co-deleted gliomas being less acidic would still hold valid, because the simulation study showed CEST contrast and amino acid concentration to be positively correlated. Another limitation of our study is that MTR asym at 3.0 ppm may have been affected by factors other than tissue pH. Although we have controlled MTR asym for the tissue transverse relaxivity and amine acid concentration using T 2 relaxometry and 18 F-FDOPA measurements, MTR asym may still be confounded by other factors including field inhomogeneity and other labile exchanging pools. Further improvement of the CEST technique is needed to achieve a more specific measurement of pH. In the future, we would like to validate our results with a larger patient cohort and potentially in a multi-institutional setting. We are also collecting MRI-guided tissue biopsy data to perform IHC staining, in order to validate our hypothesis that the observed lower acidity in 1p/19q co-deleted gliomas is related to their unique metabolic characteristics.
In addition to be useful as a non-invasive biomarker of tumor metabolism, amine CEST imaging may also work as a prognostic biomarker, because lower acidity revealed by amine CEST was associated with co-deletion of 1p/19q, which is known to be implicated in better prognosis. However, we did not perform survival analysis because only a small fraction of patients included in our study had deceased at the time of analysis. Meanwhile, it was previously reported that median tumor MTR asym decreased significantly after bevacizumab treatment in recurrent glioblastoma patients, and the change in CEST contrast was a significant predictor of progression-free survival 32 . Further studies investigating the correlation of amine CEST contrast and treatment response and survival are warranted.
To conclude, we demonstrated that 1p/19q co-deleted gliomas are less acidic than gliomas with intact 1p/19q using a combination of pH-sensitive amine CEST-EPI, T 2 relaxometry, and 18 F-FDOPA PET. Our results suggest that amine CEST-EPI may serve as a quick non-invasive imaging biomarker for identifying 1p/19q co-deletion status. Our results also support the hypothesis that the better prognosis and higher sensitivity to treatment of 1p/19q co-deleted gliomas may be related to less acidity in tumor microenvironment.

patients.
In this study, we retrospectively included a total of 76 histologically confirmed glioma patients who received CEST-EPI scan and routine MRI scan between April 2015 and July 2019. The inclusion criteria were: (1) age > 18; (2) histologically diagnosed WHO grade II (N = 40) or grade III (N = 36) glioma; (3) with IDH status available from resected or biopsied tissue, determined by genomic sequencing analysis using the polymerase chain reaction (PCR) and/or through immunohistochemistry (IHC) as described previously 33 ; (4) have CEST images with good quality (no severe motion artifact or off-resonance artifact). The 1p/19q co-deletion status, which was determined with fluorescence in situ hybridization (FISH) method at Foundation Medicine, was available in 69 of the 76 patients. We included the patients regardless of their treatment status, in order to have a more generalizable result. Out of the 76 patients, 57 were scanned either prior to radiation therapy and/or chemotherapy including temozolomide, with (N = 12) or without (N = 39) prior tumor resection surgery or had been off treatment for more than 2 years (N = 6). The other 19 patients were either on active treatment or recently off treatment at the time of MRI scanning. Detailed patient characteristics are further outlined in Table 1.

Amine CEST-EPI and anatomic MRI acquisition.
In addition to the standardized brain tumor imaging protocol 34  www.nature.com/scientificreports/ composed of a saturation pulse train of three 100-ms Gaussian pulses, with a peak amplitude of 6 μT and an inter-pulse delay of 5-ms. The offset frequencies of the saturation pulse ranged from − 3.5 to + 3.5 ppm with an uneven distribution. A total of 29 z-spectral points was acquired, densely sampled around the amine proton resonance frequency (+ 3.5 ppm), the reference frequency (− 3.5 ppm), and the water resonance frequency (0 ppm). In addition to the z-spectrum acquisition, we performed a reference (S 0 ) scan with four averages using identical sequence parameters and no saturation pulses. For the readout, we used either a single echo EPI (CEST-EPI, N = 19) or a spin-and-gradient echo EPI (CEST-SAGE-EPI, N = 57), with acquisition parameters described in more details previously 19,35 . All MRI scans were performed on 3-T MR scanners (Trio, Prisma, or Skyra, Siemens Healthcare; Erlangen, Germany). Glycine and phenylalanine phantom. To demonstrate similar amine CEST contrast between glycine and phenylalanine, we prepared 100 mM of glycine and phenylalanine in separate phantoms that also included phosphate buffered saline, titrated to 16 different pH ranging from 5.0 to 8.0 with intervals of 0.2 unit. Phantom solutions were put in falcon tubes and subsequently immersed in tap water in a secondary container. The phantom was then scanned on a Siemens Prisma 3-T MR scanner with the CEST-SAGE-EPI sequence and post-processed as described earlier. We manually created the regions of interest (ROIs) for each sample (approximately 20 mm 3 each) and calculated the mean and standard deviation of MTR asym at 3.0 ppm.

Bloch-Mcconnell simulations.
We performed Bloch-McConnell simulation of amine CEST imaging contrast with varying pH, amine concentration, and tissue T 2 relaxation time, using previously reported methods, assuming two-compartment chemical exchange between amine protons and water protons 19 . The base-catalyzed proton exchange rate (k ex ) can be expressed as k ex = k 0 + k b * 10 pH−14 , where k 0 and k b represent baseline exchange rate and base-catalyzed rate constant, respectively. The evolution of magnetization was simulated using the Bloch-McConnell equations applied to the mean MTR asym measurements from all phantom samples. The amine proton exchange rate parameters k 0 and k b that yielded the best fit to the experimental data using least squares regression were retained and used for subsequent analyses. Specifically, we simulated the CEST signal using the same saturation parameters adopted in patient scans (3 × 100 ms Gaussian saturation pulses with peak amplitude 6 μT). Additionally, we assumed the tissue relaxation characteristics to be similar to normal white matters (T 1,water = 832 ms, T 2,water = 79.6 ms) 37 , and amine protons to have relaxation rates of T 1,amine = 0.2 s, and T 2,amine = 0.1 s. We used pH values ranging from 5 to 8 with 0.2 interval and simulated the MTR asym at 3.0 ppm for amine concentrations ranging from 5 to 50 mM, to understand the effect of amine concentration on the pH dependency of CEST contrast. We also assumed an amine concentration of 20 mM and simulated the CEST signal with water T 2 ranging from 50 and 300 ms, to understand the effect of tissue transverse relaxation rate. www.nature.com/scientificreports/ T 2 relaxometry from CEST-SAGE-EPI. For patients who received CEST-SAGE-EPI scans (N = 57), transverse relaxation rates R 2 and R 2 * were estimated using the spin and gradient echo data from the reference images (S 0 ), by solving a set of Bloch signal equations as described earlier 35,38 : where where S n is signal magnitude for the n-th echo and δ is the differences in residual signal differences caused by slice profiles matching imperfection. The inverse of R 2 was calculated for each voxel to create T 2 maps, which were then registered to the post-contrast T 1 -weighted images for subsequent analysis. 18 F-FDOPA is an amino acid analog which is transported across tumor cell membranes by l-amino acid transporters 39 . As 18 F-labeled phenylalanine derivative, 18 F-FDOPA has a similar chemical structure compared to phenylalanine and contains an α-amine group likewise other α-amino acid (Fig. 1a). The amine protons on the amino acids are the main contributors of amine CEST contrast, which makes 18 F-FDOPA an appropriate measurement for controlling the effect of amine proton concentration on CEST contrast. A subset of patients (N = 23) received 18 F-FDOPA PET within 3 months of the MRI scan, with a median separation of 8 days (interquartile range of 15 days) between the PET and MRI scans. 18 F-FDOPA PET scans were performed using a high-resolution full-ring PET system (ECAT-HR; CTI/Mimvista). 18 F-FDOPA was injected intravenously with a corrected dose of 130.8 ± 26.52 MBq for each patient. We acquired 18 F-FDOPA emission data 10 min after radiotracer injection and integrated a total of 20-min PET data to obtain static three-dimensional 18 F-FDOPA images, following expectation maximization iterative reconstruction 40 . Attenuation correction was performed using data from a CT scan prior to PET. Lastly, we normalized the uptake levels to the basal ganglia, in order to reduce intersubject variability of 18 F-FDOPA uptake, thereby creating normalized FDOPA maps (nFDOPA).
Median MTR asym at 3.0 ppm, T 2 , and nFDOPA within tumor ROI excluding necrosis [combined ROI of contrast-enhancing tumor (1) and non-enhancing FLAIR hyperintense tumor (3)] were compared between 1p/19q co-deleted and non-co-deleted gliomas, using Student t-test, or Wilcoxon rank-sum test if one or both samples were not normally distributed. The normal distribution was assessed by Shapiro-Wilk parametric hypothesis test. Median MTR asym at 3.0 ppm within tumor ROI were also compared between grade II and grade III, as well as between IDH mutant and wild-type gliomas. P values less than 0.05 were considered statistically significant. All metrics were reported as mean ± standard deviation. Receiver operating characteristic (ROC) analysis was performed to assess the ability of MTR asym at 3.0 ppm to discriminate 1p/19q co-deletion status. Area under the curve (AUC), cut-off value, sensitivity, specificity, and prediction accuracy (percentage of cases predicted correctly) were reported. Lastly, analysis of covariance (ANCOVA) with continuous variable was carried out to examine the effect of 1p/19q co-deletion on MTR asym at 3.0 ppm between groups when controlling for the effect of age, nFDOPA, and T 2 . All calculations and statistical analyses were carried out using MATLAB (Release 2017b, MathWorks, Natick, MA). ethical issue. This retrospective study was approved by the "Medical IRB #2" at the University of California Los Angeles in accordance with the Helsinki Declaration of 1964. All patients provided informed written consent to have advanced imaging and medical information included in our IRB-approved research database according to IRB#14-001261 or IRB#10-000655 approved by Medical IRB #2 at the University of California Los Angeles. Out of the 76 patients, 19 were prospectively included in study IRB#14-001261, which involved surgical validation of CEST imaging method. The other 57 patients received CEST scan as part of the brain tumor standard-of-care MRI protocol in our institute. The usage of their imaging data was approved by the retrospective study protocol IRB#10-000655.

Data availability
Datasets analyzed during this study are available from the corresponding author on request. The actual raw imaging data from our patients are completely restricted due to legal and ethical restrictions on sharing these data because of potentially identifying or sensitive patient information, imposed by federal law and the ethics committee of the University of California, Los Angeles.