Contact for reagent and resource sharing
Further information and requests for resources, raw data and reagents should be directed and will be fulfilled by contacting D. H. Heiland, [email protected]. A full table of all materials is given in the supplementary information.
Ethical Approval
For this study, we included 43 patients who underwent surgery at the Department of Neurosurgery of the Medical Center, University of Freiburg. The local ethics committee of the University of Freiburg approved data evaluation, imaging procedures and experimental design (protocol 100020/09 and 5565/15). The methods were carried out in accordance with the approved guidelines. Written informed consent was obtained. The studies were approved by an institutional review board.
Imaging, Tissue Collection and Histology
Tumor tissue was sampled from the meningioma core, snap-frozen in liquid nitrogen immediately after resection and processed for further metabolic analysis. Representative tissue samples of all samples were fixed using 4% phosphate buffered formaldehyde and paraffin-embedded through standard procedures. Haemotoxylin and Eosin (H&E) staining was performed on 4 µm paraffin sections using standard protocols. These stainings confirmed the correct sampling.
Metabolite Extraction and 1H-NMR Analysis
Metabolites were extracted with 400 µl ice-cold 80% methanol and 400 µl ice-cold water, homogenized by a tissue grinder (VWR, Radnor, USA) and sonicated in 1 °C, then centrifuged at 15,000 × g for 20 minutes to remove proteins. Extracts were dried by lyophilization and resuspended in 650 µl deuterated water as described in Beckonert et al. (15). 600 µl suspension was transferred to NMR tubes for further NMR procedures. 1H-NMR spectra were collected at the Institute of Physical Chemistry of the University of Freiburg with a Bruker Avance III HDX 600-MHz NMR spectrometer (Bruker, Rheinstetten, Germany), equipped with a PABBO BB/19F-1H/D Z-GRD probe head. Each individual spectrum was recorded with two dummy scans and 32 scans with 64 k points in the time domain. The sweep width was set to 16.02 ppm with an offset of 4.691 ppm. This resulted in an acquisition time of 3.4 seconds for each scan with a dwell time of 52 microseconds. The relaxation delay was set to 2 seconds for acquisition, and the water signal was suppressed by an excitation sculpting scheme (16).
Postprocessing of Metabolic Data
In order to adjust the spectra from multiple batches, spectra were manually aligned by setting the peak of L-lactate acid at 1.310 ppm. All acquisition and processing of the spectra were performed with TopSpin 3.2 patch level 6. Detailed description of the methods was given in a recent published study by Heiland et al. (17). All spectra were analyzed with the software package “batman”, an R-software-based tool for metabolite detection in complex spectra (18). The batman software fits a predefined list of metabolites by a Bayesian approach. A detailed description of the batman algorithm was given by Hao et al. (18). Normalization of the spectra was performed by pseudo-counted Quantile (pQ) Normalization algorithm integrated in the KODAMA package. Further processing of metabolic data is described in the specific method section below.
Cluster Analysis
Normalized metabolic data was processed with AutoPipe (https://github.com-/heilandd/AutoPipe), a software package for automated unsupervised clustering. First, the number of subgroups was computed by “Partitioning Around Medoids” (Cluster number k = 2–12). To identify the optimal number of clusters, we calculated the mean silhouette width of each cluster composition. Next, to identify the core samples of each cluster, samples with a negative silhouette width were removed from further analysis. We then used either the PAMR(19) algorithm, a machine-learning based method, or a generalized linear model (20) to identify characteristic up- or downregulated metabolites of each subgroup.
Weighted Correlation Network Analysis (WCNA)
WCN-Analysis is a robust tool for integrative network analysis and was used in recent studies (21–23). It is based on a scaled-topology-free based network approach and uses the topological overlapping measurement to identify corresponding modules. These modules were analyzed by their eigengene correlation to each metabolite. The correlation of the intramodule connectivity (kME) and metabolites was used as input for a “Cluster of Clusters Analysis”. This analysis integrates expression modules and metabolites, which present equal correlation values (kME and metabolite intensity values). A detailed description of WCNA is given in (24).
Functional Analysis
Metabolic data was processed by pathviewer, an R package that includes KEGG pathway maps (25). Expression data (as described above) and normalized, log2 transformed and median centered metabolic data were integrated in the pathviewer algorithm. Enrichment analysis of metabolic data was performed with DOSE package and the web-based tool MetaboAnalyst 3.0 (www.metaboanalyst.ca).