RNA-seq profiles of C. elegans with various dietary conditions and gene perturbations related to vitamin B12 metabolism

Genome-scale metabolic network models (MNMs) can be used with a mathematical modeling tool called flux balance analysis to gain insights into metabolism at a systems level. Such models represent metabolism of an organism as a whole and do not automatically reveal how the flux overall is wired, i.e., which reactions carry flux, and which do not, and measuring flux at a large scale is challenging. Enzyme levels, as proxied by mRNA levels, can be used to predict reaction fluxes. However, integrating gene expression data with MNMs only partly constrains flux models. Here, we derive how metabolism is wired in unperturbed animals using a large-scale Worm Perturb-Seq dataset. We present a highly constrained, semi-quantitative optimal flux distribution that reflects how metabolism is wired in the animal at the adult stage. We discover several features about C. elegans metabolism that were heretofore unknown, including cyclic flux through the pentose phosphate pathway, and the primary use of amino acids and bacterial RNA as an energy source, all of which we validated by isotope tracing. Our strategy for inferring metabolic wiring based solely on gene expression should be applicable to a variety of systems, including human cells. This dataset includes the RNA-seq profiles collected during the proof-of-concept studies on the metrics used for predicting flux wiring from Worm Perturb-Seq dataset. It includes three sets of data: the profiles of animals fed live versus paraformaldehyde-killed (PFA killed) bacteria; the profiles of gene knockdowns in propionate degradation metabolism under a gradient of vitamin B12 supplementation; and the profiles of gene knockdowns in Met/SAM cycle under vitamin B12 and choline supplementations. Overall design: This GEO record includes three independent datasets used for investigating different topics in this study. The first includes RNA-seq profiles of C. elegans animals fed live versus paraformaldehyde-killed (PFA-killed) HT115 bacteria collected in biological triplicates. The second involves profiles of RNAi knockdowns of two genes in propionate canonical (pccb-1 and mmcm-1) and two genes in shunt (acdh-1 and hphd-1) degradation pathways, under a gradient of vitamin B12 supplementation (0-16nM). The third involves RNAi knockdowns of genes related to Met/SAM cycle (sams-1, ahcy-1, gcst-1, mel-32, pmt-1, cbs-1, metr-1, mthf-1, gcst-1/mel-32 double knockdown, metr-1/cbs-1 double knockdown) with and without the supplementation of vitamin B12 or choline. All data were collected using the multiplexed RNA-seq method developed in Worm Perturb-seq (WPS), a massively parallel RNAi and RNA-seq technology. This method generates whole-animal, bulk, 3'end RNA-seq profiles. Specifically, samples were multiplexed to form pooled sequencing libraries with around 50 samples per library. The read 1 is used to encode barcode information and read 2 maps to the 3' end region of C. elegans transcripts. Barcode information can be found from the Supplementary Table "barcode_and_metadata_one_carbon_cycle_vignette.csv" and "barcode_and_metadata_propionate_degradation_vignette.csv". These libraries can be directly processed by the WPS data processing pipeline (https://github.com/XuhangLi/WPS) providing the barcode information attached. A custom processing of the data is feasible following the typical demultiplexing methods for single-cell RNA-seq data that are able to process CEL-seq2 library. Notably, the first dataset, profiles for animals fed live versus dead bacteria, was also generated with WPS method, but readily demultiplexed into single fastq files per sample. This is because other samples pooled in the corresponding library were not included in this study, thus were excluded to avoid confusions.