Precision oncology in the era of radiogenomics: the case of D-2HG as an imaging biomarker for mutant IDH gliomas

See the article by Branzoli and Di Stefano et al pp. 907–916.

Molecular dissection of tumor biology reveals the existence of a veritable cancer cycle that links genetic mutations, metabolic alterations, and epigenetic modifications, which amplify and grow more complex with every turn of the cycle. A full decade has passed since the discovery of isocitrate dehydrogenase 1 and 2 (IDH1/2) mutations in gliomas.¹ The product of mutant IDH, D-2-hydroxyglutarate (D-2HG) has been shown to be a central hub in the molecular network of mutant IDH cancer cells and is a focus of research from several directions, such as metabolism, pharmacology, and biomarker science. Because of its unique biology, D-2HG is a versatile biomarker that can be used for diagnosis, prediction, prognosis, treatment response, and pharmacodynamics of targeted therapy of mutant IDH cancers. In particular, D-2HG is an ideal imaging biomarker because it is virtually absent in wild-type IDH tissues, while it reaches very high concentrations (5–30 mM) in mutant IDH cancer cells that can be detected non-invasively by in vivo magnetic resonance spectroscopy (MRS). Inherently, this provides very high contrast to the noise ratio between mutant IDH tumor and the background, and early MRS studies have sought to investigate the feasibility of in vivo D-2HG detection.^2–4 If reliable, D-2HG detection by MRS would be a very convenient examination with high impact for patients and clinicians as treatment and prognosis are different between wild-type and mutant IDH gliomas.

The study published in the current issue by Branzoli, Di Stefano, and colleagues contributes to the validation and verification of 2HG as an imaging biomarker that can be measured by in vivo MRS in glioma patients.⁵ The authors filled the gap regarding the precision and accuracy of in vivo MRS methods for 2HG quantification, and aimed to systematically evaluate the performance of the two most used D-2HG MRS methods: the long echo modulation⁴ and J-difference spectral editing using the Mescher–Garwood (MEGA) scheme.^3,4 Their study addresses two types of questions that are important from the clinical point of view: (i) is 2HG present or not? and (ii) what is the concentration of 2HG? These 2 questions are akin to the two most important properties of any diagnostic test: precision and accuracy. The simple presence of D-2HG is diagnostic for IDH mutations, and this question needs a lower bar for precision and accuracy, and technically is easier to answer. The concentration of D-2HG is more informative as it relates to the density of mutant IDH cells, and it is useful for monitoring tumor longitudinally, but technically it requires superior performance for precision and accuracy. In a prospective preoperative group of 24 mutant IDH glioma patients, the authors performed back-to-back acquisitions of both methods and quantified in vivo 2HG levels, which were compared with postoperative gold standard mass-spectrometry 2HG measurements and mutant IDH status obtained from biopsies. The presence of 2HG and the diagnosis of IDH status can be determined by both methods; however, the sensitivity, specificity, positive predictive value, and negative predictive value are higher for spectral editing across a wider range of technical performance as measured by the Cramér-Rao lower bounds (CRLB). Long echo time is superior to spectral editing only for very stringent CRLB conditions that typically require narrow spectral linewidth to resolve spectral overlap, while spectral editing is more forgetting regarding CRLB and spectral linewidth because it deals with fewer spectral lines that are well separated. Regarding the accuracy of 2HG quantification, the concentrations obtained by spectral editing MRS measurements have a statistically significant correlation with concentrations obtained by gold standard mass spectrometry, while in the case of long echo time there is a weaker correlation which is not statistically significant. Combined, these results seem to suggest a somewhat superior precision and accuracy of spectral editing (MEGA) as a diagnostic test for detection of D-2HG and mutant IDH status in gliomas. This study provides independent verification and validation for in vivo MRS methodology that is useful in the clinical workup of patients with brain masses.

Fig. 1.

Metabolic alterations are part of the cycle that drives cancer, and in vivo magnetic resonance spectroscopy is an imaging method that can be used to quantify onco-metabolite D-2HG to diagnose and monitor mutant IDH glioma patients. Combining imaging (radiological) and genomics information results in a powerful tool that enables precision medicine in patients.

Imaging of D-2HG is a prime example for the power of combining radiological and genetic information in a single investigation, and more evidence is accumulating for its clinical utility beyond the initial diagnostics. While definitive diagnosis of glioma and mutant IDH status still requires a biopsy examination, the treatment planning and longitudinal follow-up rely heavily on imaging. In particular, longitudinal D-2HG imaging could be helpful to answer questions related to treatment response and tumor progression, or to rule out pseudoprogression and pseudoresponse, as shown in two recent studies.^6,7 Furthermore, although other imaging modalities can be employed with radiomics, machine learning, and artificial intelligence to diagnose mutant IDH status,^8,9 D-2HG has a unique value, as it can be used to study directly the function of mutant enzyme, implicitly disease mechanisms, and in a more applied perspective perform mechanistic studies to probe target modulation and pharmacodynamics of targeted therapies to accelerate their clinical translation.¹⁰ The work by Branzoli, Di Stefano, and colleagues,⁵ together with previous studies,^11–13 helps solidify the case for D-2HG MRS.

Several lines of caution need to be added: (i) MRS cannot distinguish between D-2HG and L-2HG, the former being solely the product of mutant IDH tumors, while the latter is produced by normal tissues albeit at low concentrations, or in metabolic diseases at higher concentrations¹⁴; (ii) the ability of MRS to detect D-2HG toward the lower end (1 mM) of concentration range is problematic, and this is often the case with patients who had surgery or chemoradiation; (iii) the use of D-2HG to probe targeted treatment response has to account for the possibility of down assays to be susceptible to off-target effects¹⁵; (iv) technical challenges that make in vivo MRS more susceptible to failures compared with other imaging modalities. While the first 3 aspects are common for other techniques, the last aspect is particular to MRS and requires diligence in data acquisition and processing. In particular, magnetic field inhomogeneity and localization artifacts caused by patient movement or scanner drift need to be minimized. Given the complexity of cancer, D-2HG MRS is one of the imaging biomarkers, albeit an important one, and needs to be combined with other imaging to tackle this complexity. Quantitative advances in technology such as genomics, proteomics, metabolomics, biomarkers, assays, and rational drug design are enabling qualitative improvements in personalized and precision medicine/oncology, to which imaging is an integral part.