CEST MRI provides amide/amine surrogate biomarkers for treatment-naïve glioma sub-typing

Purpose Accurate glioma classification affects patient management and is challenging on non- or low-enhancing gliomas. This study investigated the clinical value of different chemical exchange saturation transfer (CEST) metrics for glioma classification and assessed the diagnostic effect of the presence of abundant fluid in glioma subpopulations. Methods Forty-five treatment-naïve glioma patients with known isocitrate dehydrogenase (IDH) mutation and 1p/19q codeletion status received CEST MRI (B1rms = 2μT, Tsat = 3.5 s) at 3 T. Magnetization transfer ratio asymmetry and CEST metrics (amides: offset range 3–4 ppm, amines: 1.5–2.5 ppm, amide/amine ratio) were calculated with two models: ‘asymmetry-based’ (AB) and ‘fluid-suppressed’ (FS). The presence of T2/FLAIR mismatch was noted. Results IDH-wild type had higher amide/amine ratio than IDH-mutant_1p/19qcodel (p < 0.022). Amide/amine ratio and amine levels differentiated IDH-wild type from IDH-mutant (p < 0.0045) and from IDH-mutant_1p/19qret (p < 0.021). IDH-mutant_1p/19qret had higher amides and amines than IDH-mutant_1p/19qcodel (p < 0.035). IDH-mutant_1p/19qret with AB/FS mismatch had higher amines than IDH-mutant_1p/19qret without AB/FS mismatch ( < 0.016). In IDH-mutant_1p/19qret, the presence of AB/FS mismatch was closely related to the presence of T2/FLAIR mismatch (p = 0.014). Conclusions CEST-derived biomarkers for amides, amines, and their ratio can help with histomolecular staging in gliomas without intense contrast enhancement. T2/FLAIR mismatch is reflected in the presence of AB/FS CEST mismatch. The AB/FS CEST mismatch identifies glioma subgroups that may have prognostic and clinical relevance. Supplementary Information The online version contains supplementary material available at 10.1007/s00259-022-05676-1.


Introduction
The discovery of mutations in the IDH1 and IDH2 genes in astrocytic and oligodendroglial tumours has led to a biomarker-driven classification, forming an integrated diagnosis composed of the histological appearance and the molecular profile [1].
Currently, two types of IDH-mutant gliomas are identified. The IDH-mutant astrocytoma is defined by an additional mutation of ATRX and p53. A loss in the tumour suppressor locus CDKN2A/B is an important additional prognostic marker, typically found in IDH-mutant glioblastomas. The IDH-mutant oligodendroglioma is defined by an absent ATRX mutation, and a codeletion of chromosomal arms 1p and 19q. IDH-mutant oligodendrogliomas consistently carry a TERT promoter mutation. The IDH-wild-type gliomas comprise a wide range of tumours, including, but by far not limited to, the IDH-wild-type glioblastoma. The IDH-wild-type glioblastoma is molecularly characterized by chromosome 7 gain, chromosome 10 loss, frequent EGFR amplification, and TERT promoter mutation. However, there is a wide range of other IDH-wild-type gliomas of low and high grades, with other defining mutations, such as BRAF, histone H3 K27M, and H3 G34R.
The current evidence for the clinical utility of CEST metrics has inherent study design and methodological weaknesses. The former includes mainly the small number of patients; the unbalanced cohort composition with diagnostically straightforward, by means of conventional MRI, high-grade gliomas; and the use of the outdated WHO 2007 classification for glioma grading in the prevailing number of studies. Methodological issues are related to the investigation only of the 3.5 ppm amide frequency and negligence of the T2/FLAIR mismatch effect on the CEST maps, which carries significant diagnostic implications as CEST is sensitive to fluid signal [29]. Studies in rats and phantoms have suggested that the CEST contribution at 2 ppm is sensitive to proteins, amino acid, and pH concentration changes, and may serve as a sensitive neuroimaging biomarker for many diseases [30], while a study in vitro and on mice has suggested the amide/amine ratio to be sensitive to the tissue acidity in stroke [31].
The current study sought to address the knowledge gap on the clinical value of CEST imaging to predict the histomolecular glioma type by prospectively investigating the whole range of frequency offsets either side of the water peak from 0 to 4 ppm in a single-centre setting, large patient cohort presenting mainly with the diagnostically challenging primary gliomas with absent or weak enhancement. For the first time, we sought to evaluate the diagnostic value of tumour amide/amine ratio, which might be a promising surrogate biomarker for gliomas sub-typing. We also aimed to assess the diagnostic sensitivity and specificity of two different CEST models ('asymmetry-based' and 'fluid-suppressed') to address any diagnostic compromise caused by the presence of abundant fluid in gliomas subpopulations, i.e. those with T2/FLAIR mismatch.

Patients
Treatment-naïve patients, older than 18 years of age, with presumed glioma suitable for surgical lesion sampling or resection were selected during the weekly held multidisciplinary tumour board meeting by unanimous agreement among the clinical and neuroradiology board members. Exclusion criteria included usual contraindications to MRI, pregnancy, and incapacity to provide informed consent. Eightytwo patients have been so far successfully recruited for the study. For the objectives of this publication, an interim study analysis of 68 patients with conventional indeterminate MRI findings for glioma staging was conducted (Fig. 1). After the MRI, 48 patients underwent surgical resection and four had a biopsy. Of these, 49 patients had histologically confirmed gliomas with grade II or higher and two patients had acute inflammation and multifocal cortical ischaemia, respectively. Data from four patients were discarded due to presence of artefacts. The MRIs of nine out of 45 patients showed post-contrast enhancement, of which seven were described as faint enhancement and two as moderate enhancement. Finally, 45 patients (19 male), who had undergone surgery/ biopsy a median (range) of 1.8 (0.1-30.7) months after MRI, were included in the analysis. The clinical characteristics of the patient population are summarized in Fig. 1

MRI data acquisition
Imaging data were acquired on a 3-T whole-body MRI system (MAGNETOM Prisma; Siemens Healthcare, Erlangen, Germany) with a 64-channel Head/Neck coil. Structural T 1 -weighted (T 1 w), T 2 w, fluid-attenuated inversion recovery (FLAIR), and gadolinium-enhanced T 1 w (Gd-T 1 w) images were acquired with sagittal 3D Sampling Perfection with Application optimized Contrasts using different flip angle Evolution (SPACE) images, with an isotropic resolution of 0.9 × 0.9 × 0.9 mm 3 , 1.1 × 1.1 × 1.1 mm 3 , 1.0 × 1.0 × 1.0 mm 3 , and 0.9 × 0.9 × 0.9 mm 3 , respectively. The axial singleslice prototype 2D CEST sequence (1.6 × 1.6 × 4.0 mm 3 ) was located at the largest tumour cross-section. Three acquisitions with different saturation pulse intensities (B 1rms = 1.7μT, 2μT, 2.3μT; T sat = 3.5 s) were performed. A relative B 1 map [32] was obtained by a dual flip angle method for B 1 -correction of the CEST signal. The total acquisition time was approximately 17 min for the structural data and 8 min for the CEST data. Detailed description of the conventional MRI protocol and the CEST acquisition is provided in the Supplementary Material.

Regions of interest
Regions of interest (ROIs) were manually drawn by two senior neuroradiologists (12 and 8 years of experience in neurooncology imaging) in consensus using ITK-SNAP version 3.6 (www. itksn ap. org) [33] loading first the CEST to identify the appropriate slice position and then the co-registered FLAIR, T2w, T1w, and Gd-T1w images. Areas identified were (1) solid tumour excluding cyst and necrotic areas, (2) cyst, and (3) normal appearing white matter (NAWM) contralateral to the tumour.

CEST post-processing
CEST acquisitions were processed using prototype version software created by Olea Medical® (La Ciotat, France) for the Horizon 2020 project GLINT (number 667510). After Z-Spectra-based B 0 correction [34] of the three acquired B 1 saturation levels, the Z-Spectra at 2.0μT were B 1 -corrected through an exponential fitting [32]. Magnetization transfer ratio asymmetry (MTR asym ) spectra and proton transferweighted maps with two different models and in two offset ranges were then calculated.

MTR asymmetry spectra and ∆MTR asym spectra
MTR asym spectra were calculated at steps of 0.25 ppm from 0 to 4 ppm with a previously reported formula [35]. Changes in MTR asym (∆MTR asym ) were obtained subtracting the NAWM MTR asym from the tumour one: ∆MTR asym (∆ω) = MTR asym (∆ω, tumour) − MTR asym (∆ω, NAWM). Average and 25th and 75th percentiles of the MTR asym and of the ∆MTR asym spectra from patients in the various histological groups were calculated to generate the corresponding group spectra and group range of variation.

Asymmetry-based and fluid-suppressed APT w image processing
Non-punctual metrics [36] obtained averaging the signal in a 1-ppm range centred at 3.5 ppm and at 2 ppm were adapted to generate two amide and amine proton transfer-weighted maps (APT w (∆3.5) and APT w (∆2), respectively). These had improved detection (with respect to the punctual metrics at 3.5 ppm and 2 ppm) of the CEST contrast for B 0 -and B 1 -corrected Z-Spectra at 2.0μT. APT w maps were obtained using an asymmetry-based (AB_APT w ) model and a fluidsuppressed (FS_APT w ) model [29]. The AB_APT w considered only the asymmetry-average of the Z-Spectra while the FS_APT w also attenuated the fluid signal based on the shape of Z-spectrum, as described in the Supplementary Material. Furthermore, the difference (mismatch) between AB and FS models was evaluated in terms of AB-FS maps.
The endogenous CEST signal was explored in two offset ranges: amines (∆2, from 1.5 to 2.5 ppm) and amides (∆3.5, from 3.0 to 4.0 ppm), obtaining four maps. In total, nine normalized metrics (obtained by subtracting the average signal in the contralateral NAWM ROI from the average values in the tumour ROI) were considered: the amides, amines, and the amide/amine ratios for each of the AB, FS, and AB-FS maps.

T2/FLAIR mismatch vs AB/FS mismatch in IDH-mutant gliomas
The T2 and FLAIR images were assessed in consensus by a senior neuroradiologist (14 years of experience as consultant neuroradiologist with expertise in neuro-oncology) and an MRI physicist (15 years of experience in MRI image analysis) for identification of the presence of T2/ FLAIR mismatch. The T2/FLAIR mismatch was defined as the presence of > 50% hyperintense signal on T 2 w and relatively hypointense signal on FLAIR except for a hyperintense peripheral rim in cases of complete or nearcomplete mismatch [37,38].
Cut points were calculated (method outlined in the 'Statistical analysis' section) for the AB-FS (∆3.5) and AB-FS (∆2) metrics in order to assess if AB-FS metrics could serve as surrogate biomarkers for the presence/absence of T2/FLAIR mismatch (reference standard). The diagnostic value of AB/FS mismatch, defined as AB-FS values larger than the cut point threshold, was assessed.

Histopathology
Tumour biopsies were fixed in formalin and processed into paraffin-embedded samples. Tissue sections were routinely stained with haematoxylin and eosin, and immunostained with antibodies against IDH1 R132H, ATRX, and Ki67. IDH1 R132H-positive and ATRXnegative tumours (loss of expression) were diagnosed as IDH-mutant astrocytoma, anaplastic astrocytoma, or glioblastoma, depending on the histological features. IDH1 R132H-positive and ATRX-positive (retained) tumours were examined for the presence of a 1p/19q codeletion (1p/19q ret or 1p/19q codel ). All IDH1 R132Hnegative tumours were further sequenced for the presence of a rare IDH1 or an IDH2 mutation. These tumours were also tested for TERT promoter mutation and EGFR amplification to identify the population of IDH-wild-type glioblastoma. Tumours that had no informative molecular profile were further examined with DNA methylation arrays, followed by algorithmic classification as described in [39].

Statistical analysis
Following a Shapiro-Wilk normality test, the asymmetrybased (AB) and fluid-suppressed (FS) CEST metrics were compared using a sign test with two-tailed p-values and significance threshold set to 0.05. A Fisher's exact test assessed the relationship between the presence/absence of AB/FS mismatch and the presence/absence of T2/FLAIR mismatch.
The Mann-Whitney U test was used to analyze the statistical differences between quantitative imaging parameters for the two IDH mutation statuses. The Kruskal-Wallis test followed, when significant, by the Conover-Iman test to correct for multiple pairwise comparison (significance set to 0.05), was used to assess the statistical differences for (i) the three glioma groups IDH-wild type, IDH-mutant_1p/19q ret , and IDH-mutant_1p/19q codel ; and (ii) the four tumour subgroups IDH-wild type, IDH-mutant_1p/19q ret with (without) AB/FS mismatch, and IDH-mutant_1p/19q codel . For pairs of groups with statistically significant differences after correction for multiple comparisons, areas under the receiver operating characteristic (ROC) curve (AUC) and the nearest to (0.1) cut points were estimated. All statistical analyses were performed with Stata software (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC).

Z-Spectra, APT w maps, and MTR asym and ∆MTR asym spectra
Representative examples for four different tumour types of structural MRIs and CEST maps (shown in Fig. 2a) and Z-Spectra (shown in Fig. 2b) show a cystic component present in the IDH-wild-type with reduced signal in the FS CEST maps in the amide and amine ranges (Fig. 2a, IDH-wild-type); minor differences between AB and FS maps in the solid tumour, more evident in IDH-mutant_1p/1q ret and in the amine range (Fig. 2a). Fig. 3 shows the MTR asym (Fig. 3a-c) and the ∆MTR asym spectra (Fig. 3d-f) for the different tumour subtypes (two glioma groups in Fig. 3a and d; three glioma groups in Fig. 3b and e; four glioma groups in Fig. 3c and f). Both the MTR asym and the ∆MTR asym spectra showed different contributions from the amide and the amine regions between glioma subtypes over the frequency range analyzed. The average MTR asym spectrum in the NAWM over all patients was null at a frequency of ~ 3 ppm from the water resonance (Fig. 3a).

CEST metrics
Quantitatively, the derived CEST metrics with information about amide signal, amine signal, and their ratios enabled the differentiation of IDH-wild-type from IDH-mutant gliomas and the identification of subgroups in the IDHmutant gliomas (Tables 1, 2, Fig. 4).

AB vs FS metrics and AB/FS mismatch vs T2/FLAIR mismatch
The AB metrics were statistically significantly different from the FS metrics for almost all metrics and glioma subgroup, with the exception of IDH-wild-type in the amides range, and IDH-mutant_1p/19q ret without AB/FS mismatch in the amides range (Table 3).
A T2/FLAIR mismatch was detected only in IDH-mutant_1p/19q ret gliomas and, more specifically, in 68% (15 out of 22) of them. The difference between AB and FS metrics (AB-FS) for amides and amines (positive differences in both cases, indicating AB > FS values) allowed identifying two subgroups in IDH-mutant_1p/19q ret gliomas: those with and those without AB/FS mismatch. Cut points were calculated for the AB-FS (∆3.5) metric (cut point 0.039, sensitivity 80%, specificity 71%) and AB-FS (∆2) metric (cut point 0.18, sensitivity 80%, specificity 86%) as surrogate biomarkers to assess the presence/absence of T2/FLAIR mismatch, used as a reference standard.
The presence/absence of AB/FS mismatch was in agreement with the presence/absence of T2/FLAIR mismatch, as assessed by a Fisher's exact test p-value of 0.014.

IDH mutation status
The amines, amide/amine signal ratio, and the difference AB-FS in the amine range differentiated IDH-wild-type from IDH-mutant (Tables 1, 2, Fig. 4). Of these metrics, the amide/amine ratio had the highest AUC (0.84 for AB and 0.81 for FS) and thresholds of 1.57 for AB (1.66 for FS) achieved 78% sensitivity and 78% specificity. In other words, IDH-wild-type had at least a 57% (66% for FS) larger signal coming from the amide component of the CEST spectrum than from the amines.
The AB-FS for amines signal was positive (AB > FS) and larger in IDH-mutant than in IDH-wild-type, suggesting the presence of a larger more 'fluid' component in IDH-mutant gliomas ( Table 3, Fig. 4).

IDH mutation and 1p/19q molecular status
When considering IDH mutation and 1p/19q molecular status, significant differences were observed in three pairwise Fig. 2 Representative examples of structural MRI, CEST maps, and Z-spectra. a Representative structural MRI and CEST maps for IDH-wildtype, IDH-mutant_1p/19q ret with and without AB/FS mismatch, and IDH-mutant_1p/19q codel . AB_APT w indicates the asymmetry-based metric while FS_APT w is the fluid-suppressed metric. ∆3.5 is the 1-ppm offset range centred at the amide offset of 3.5 ppm, while ∆2 is the 1-ppm offset range centred at the amine offset of 2 ppm. b Representative Z-spectra of the cyst component and of the tumour core for the same tumour types comparisons, after correction for multiple comparisons ( Other metrics that differentiated these two subgroups were the amides and the AB-FS differences for both the amides and the amines. The AB amide metric had the highest specificity (92%), but only 55% sensitivity.

IDH mutation, 1p/19q molecular status, and AB/FS mismatch
When the IDH-mutant_1p/19q ret gliomas were further subdivided in two subgroups (with and without AB/FS mismatch), significant differences were observed in four pairwise comparisons, after correction for multiple comparisons (    No metrics were statistically significant for differentiating IDH-mutant_1p/19q ret without AB/FS mismatch from IDH-mutant_1p/19q codel .

Discussion
The current study demonstrates the potential of CEST metrics using two different models and combining the signal from amides (at 3.5 ppm) and amines (at 2 ppm) to aid in the stratification of gliomas presenting with none or faint post-gadolinium enhancement. Results suggest that (1) the combination of various CEST metrics may be useful biomarkers for differentiating IDH and 1p/19q status; (2) the comparison between the two models (asymmetry-based and fluid-suppressed) may be an objective biomarker to identify the gliomas with T2/FLAIR mismatch.
Other groups report that APT w (amides) alone can differentiate IDH-wild-type/mutant, e.g. [26]. In our study, IDH-wild-type/mutant were differentiated by amines, amide/amine signal ratio, and the difference AB-FS in the amine range, but not by the amide signal. This difference is likely due to the composition of our cohort of diagnostically challenging primary gliomas, in which the IDH-wild-type Table 2 Median values and 25 to 75 percentile ranges for the asymmetry-based (AB_APT w ) and fluid-suppressed (FS_APT w ) normalized CEST metrics, for the various glioma groups and subgroups ∆3.5 = [3-4 ppm] = 1-ppm frequency range centred on the amide frequency at 3.5 ppm; ∆2 = [1.5-2.5 ppm] = 1-ppm frequency range centred on the amine frequency at 2 ppm; AB_APT w (∆ω) asymmetry-based CEST metric in the ∆ω range (∆ω = ∆3.5 or ∆ω = ∆2) normalized to the normal appearing white matter; FS_APT w (∆ω) = fluid-suppressed CEST metric in the ∆ω range (∆ω = ∆3.5 or ∆ω = ∆2) normalized to the normal appearing white matter;. AB_APT w ratio = AB_APT w (∆3.5)/AB_APT w (∆2); FS_APT w ratio = FS_APT w (∆3.5)/FS_APT w (∆2); AB-FS = AB_APT w − FS_APT w  tumours showed no enhancement in post-contrast T 1 w images (see Supplementary Table 1). The amide/amine signal ratio was the only metric differentiating IDH-wild type (worst-prognosis) from IDH-mutant_1p/19q codel (best prognosis). The amide/amine signal ratio and levels of amine signal differentiated IDHwild-type also from the pooled IDH-mutant and from IDH-mutant_1p/19q ret . The cut points identified for these comparisons suggested that IDH-wild-type had approximately 60% more signal originating from the amide pool than from amine groups compared to the IDH-mutant groups. IDH-mutant_1p/19q ret had higher amide and amine signal levels than IDH-mutant_1p/19q codel . The relevance of tumour acidity in differentiating gliomas with different IDH status is supported by recent studies [25,40]. Amine and amide signals have been shown to have a complex dependency on protein, amino acids, temperature, and pH concentrations, as well as on CEST saturation [30], whilst the amide/amine signal ratio has been shown to have more straightforward dependency primarily on pH when the hypothesis that amide and amine groups belong to the same molecules is valid (e.g. in ischemia) [31]. The dependency of the CEST amide/amine signal ratio on protein and amino acid concentrations cannot be asserted without additional measurements and simulations. However, the IDH-mutant_1p/19q ret without AB/FS mismatch could not be differentiated from the IDH-mutant_1p/19q codel , suggesting that the presence of T2/FLAIR mismatch, usually not assessed in the literature, is an important parameter to consider when assessing the efficiency of the CEST metrics to stratify gliomas.
A complementary imaging technique is amino acid PET (positron emission tomography), an extensively evaluated radiotracer imaging methodology that is playing an increasingly important role in the diagnosis and management of brain tumours. The advantages of amino acid PET are that radiolabelled amino acids cross the blood-brain barrier and j-k Three tumour types: IDH-wild-type; IDH-mutant_1p/19q ret ; IDH-mutant_1p/19q codel . e-h and l-m Four tumour types: IDH-wildtype; IDH-mutant_1p/19q ret with (without) AB/FS mismatch; IDH-mutant_1p/19q codel and their accumulation is a function of tumour avidity for them. This differential uptake can be exploited to specifically delineate cellular mass and tumour boundaries from surrounding normal tissue also in non-enhancing gliomas [41]. A study comparing O-(2-18F-fluoroethyl)-l-tyrosine (FET) PET and APT w CEST MRI in eight high-grade glioma patients showed that FET PET and APT w CEST are spatially incongruent and suggesting different biological information [42]. A further study comparing FET PET with APT w CEST and perfusion (cerebral blood volume, CBV, from a dynamic susceptibility contrast acquisition) in 46 patients (31 IDH-wild-type and 12 IDH-mutant) observed both in IDH-wild-type glioblastomas and IDH-mutant lower grade gliomas relevant overlap between tumour areas defined by different imaging modalities, strongest not only for APT w and FET in contrast-enhancing parts of glioblastomas, but also in the FLAIR-hyperintense region of lower grade gliomas. Furthermore, the author found that both APT w and FET correlated with cellularity, as opposed to CBV which was associated with vascularity [43]. Even though only few studies directly compared amino acid PET and CEST in more than five patients, the complementary and additive values of these methods, as well as their incongruent findings, suggest that further investigation including also neuropathological validation could provide useful information to understand the biological information provided.
Our results also show that the asymmetry-based CEST metric is sensitive to regions with large fluid content (e.g. cystic regions), and that the fluid-suppressed (FS) metric significantly reduces the CEST signal in cysts, without affecting the signal from the normal appearing white matter, but affecting the amide and amine signals in tumour to different extent. We suggest that thresholds can be defined in the differences between the AB and FS models (AB-FS) in the amide and in the amine ranges, to divide the IDH-mutant_1p/19q ret gliomas in two subgroups: with and without AB/FS mismatch (thresholds: AB-FS (∆3.5) = 0.039 and AB-FS (∆2) = 0.18). In IDH-mutant_1p/19q ret glioma, the presence of AB/FS mismatch was closely related to the presence of T2/FLAIR mismatch, suggesting the presence of a more fluid-like compartment (probably micro-cysts, which are observed extracellularly in astrocytomas and intracellularly in oligodendrogliomas) but further studies are needed to explain the factors contributing to our results. As a matter of fact, the presence of AB/FS mismatch and amide/amine signal ratio may be Table 3 Sign test estimates of the differences between AB and FS CEST models for all the investigated metrics: amide signal, amine signal, signal ratio for amides/amines useful biomarkers for differentiating IDH-wild-type from IDH-mutant_1p/19q ret with AB/FS mismatch. However, before any meaningful clinical application may be sought, the biological and prognostic significance of the AB/FS mismatch remains to be validated in larger cohorts. Concerning IDH-mutant gliomas, combinations of amide and amine (but not amide/amine ratio) metrics may be also useful biomarkers for differentiating other glioma subgroups: (i) IDH-mutant_1p/19q ret from IDH-mutant_1p/19q codel and (ii) IDH-mutant_1p/19q ret with AB/FS mismatch from IDH-mutant_1p/19q ret without AB/FS mismatch.
In our data, the amine range had at times a higher sensitivity than the amide range to distinguish between tumour types. Several groups have investigated the contributions to the CEST signal at 2 ppm showing that the major contributions, at pH = 7 and at a temperature of 37 °C, come from creatine (Cr), glutamate (Glu, roughly 40% of Cr signal), phosphocreatine (PCr), ATP (both approximately 20% of Cr signal), glucose (Glc, roughly 10% of Cr signal), and proteins [14], with various weightings of these different pools depending on the CEST saturation scheme and other MRI acquisition parameters. Some have suggested that at this pH and temperature, glutamine (Gln) does not contribute to the CEST signal at 2 ppm [14], while others have found contribution from Gln, especially in tumours with a more acidic extracellular microenvironment [44]. It has also been suggested that also direct water saturation, semi-solid MT, and water longitudinal relaxation time effects, which contribute to the signal non-linearly, are likely to contribute to the CEST signal at 2 ppm [14]. It is possible that the presence of all these compounds at 2 ppm increases the ability of the CEST signal at 2 ppm to differentiate between tumour types. However, contributions from the nuclear Overhauser enhancement (NOE)-mediated effect coming from the upfield resonances complicate the interpretations of our results, and it may contribute to the low sensitivity of the CEST amide signal for distinguishing between tumour types. The MTR asym offset dependence is due to the presence of saturation peaks in a frequency range from − 1 to -5 ppm, mediated by the NOE. The NOE is therefore entangled to the chemical exchange effect in the MTR asym measure of the magnetization transfer process. A positive MTR asym reflects a predominant chemical exchange effect while a negative MTR asym reflects a predominant NOEmediated effect. Limitations of the present study include the presence of the NOE-mediated effects in the MTR asym and ∆MTR asym spectra. The NOE-mediated contribution is reduced with increasing B 1 saturation power and saturation powers of 2μT had been shown in the past to null the NOE-mediated effect at 3.5 ppm [45,46]. Our normal appearing white matter MTR asym signal nulls at 3 ppm from the water resonance (Fig. 3a), indicating that the balance between CEST and NOE effects was obtained at 3 ppm and there was a slightly predominant NOE contribution for frequencies above 3 ppm [45,46].
Our data acquisition was also limited to a single slice. Since the start of our study, 3D CEST sequences have become available at both 3T and 7T [47,48] together with more complex analysis methods based on fitting multiple Lorentzian shapes to the data allowing the quantification of the different contributions from amides, amines, NOE, MT, and water [32,48]. We plan to introduce these acquisition and analysis strategies in our future studies, together with MRS and perfusion and diffusion measurements to further characterize gliomas both pre-and post-treatments.

Conclusions
The current study suggests that CEST-derived biomarkers for amines and amides, together with their ratio, which reflects in a composite manner tissue acidity and proteins/amino acid abundance, can be used for histomolecular staging in a cohort of diagnostically challenging non-or low-enhancing gliomas. The data also suggests that the mismatch between CEST maps obtained with different models (asymmetry-based (AB) and fluid-suppressed (FS)) can be used to identify subgroups in IDH-mutant_1p/19q ret gliomas that could potentially have prognostic and clinical relevance. CEST-derived biomarkers could therefore serve as risk stratification tools to inform oncologists of recurrence risk or of optimal treatment.

Availability of data and material
The authors confirm that the data supporting the findings of this study are available within the article, in its Supplementary Material and/or from the corresponding author upon reasonable request.

Declarations
Ethics approval This prospective, on-going, investigator-initiated, single-centre, neuro-oncology imaging study received ethical and institutional review board approval (IRAS project ID: 210819; Research Ethics Committee reference: 16/LO/2081).
Consent to participate and for publication All patients, prior to participation, provided written informed consent to participate and for publication.

Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.