Phenotype-specific estimation of metabolic fluxes using gene expression data.

A cell's genome influences its metabolism via the expression of enzyme-related genes, but transcriptome and fluxome are not perfectly correlated as post-transcriptional mechanisms also regulate reaction's kinetics. Here, we addressed the question: given a transcriptome, how unobserved mechanisms of reaction kinetics should be systematically accounted for when inferring the fluxome? To infer the most likely and least biased fluxome, we present Pheflux, a constraint-based model maximizing Shannon's entropy of fluxes per mRNA. Benchmarked against 13C fluxes of yeast and bacteria, Pheflux accurately estimates the carbon core metabolism. We applied Pheflux to thousands of normal and tumor cell transcriptomes obtained from The Cancer Genome Atlas. Pheflux showed statistically significantly higher glucose yields on lactate in breast, kidney, and bronchus-lung tumoral cells than their normal counterparts. Results are consistent with the Warburg effect, a hallmark of cancer metabolism, suggesting that Pheflux can be efficiently used to study the metabolism of eukaryotic cells.

Kinetic model of Clostridium beijerinckii's Acetone-Butanol-Ethanol fermentation considering metabolically diverse cell types.

Clostridium beijerinckii population branches into metabolically diverse cell types in batch cultures. Here, we present a new kinetic model of C. beijerinckii's Acetone-Butanol-Ethanol fermentation that considers three cell types: producers of acids (acidogenic), consumer of acids and producers of solvents (solventogenic), and spores cells. The model accurately recapitulates batch culture data. Also, the model estimates cell type-specific kinetic parameters, which can be helpful to improve the operation of the ABE fermentation and give a framework to study acidogenic and solventogenic metabolic pathways. To exemplify the latter, we used a constraint-based model to study how the ABE pathways are used among acidogenic and solventogenic cell types. We found that among both cell types, glycolytic production of ATP and consumption of NAD+ varies widely during the fermentation, with their maximum production/consumption rates happening when acidogenic and solventogenic growth rates were at their highest. However, acidogenic cells use the ABE pathway to contribute with an extra 12.5% of the total production of ATP, whereas solventogenic cell types use the ABE pathway to supply more than 75% of the demand for NAD+, alternating between the production of lactate and butyrate, being both coupled to the production of NAD+.

Metabolic flux configuration determination using information entropy.

Constraint-based models use steady-state mass balances to define a solution space of flux configurations, which can be narrowed down by measuring as many fluxes as possible. Due to loops and redundant pathways, this process typically yields multiple alternative solutions. To address this ambiguity, flux sampling can estimate the probability distribution of each flux, or a flux configuration can be singled out by further minimizing the sum of fluxes according to the assumption that cellular metabolism favors states where enzyme-related costs are economized. However, flux sampling is susceptible to artifacts introduced by thermodynamically infeasible cycles and is it not clear if the economy of fluxes assumption (EFA) is universally valid. Here, we formulated a constraint-based approach, MaxEnt, based on the principle of maximum entropy, which in this context states that if more than one flux configuration is consistent with a set of experimentally measured fluxes, then the one with the minimum amount of unwarranted assumptions corresponds to the best estimation of the non-observed fluxes. We compared MaxEnt predictions to Escherichia coli and Saccharomyces cerevisiae publicly available flux data. We found that the mean square error (MSE) between experimental and predicted fluxes by MaxEnt and EFA-based methods are three orders of magnitude lower than the median of 1,350,000 MSE values obtained using flux sampling. However, only MaxEnt and flux sampling correctly predicted flux through E. coli's glyoxylate cycle, whereas EFA-based methods, in general, predict no flux cycles. We also tested MaxEnt predictions at increasing levels of overflow metabolism. We found that MaxEnt accuracy is not affected by overflow metabolism levels, whereas the EFA-based methods show a decreasing performance. These results suggest that MaxEnt is less sensitive than flux sampling to artifacts introduced by thermodynamically infeasible cycles and that its predictions are less susceptible to overfitting than EFA-based methods.

Rainbow-Seq: Combining Cell Lineage Tracing with Single-Cell RNA Sequencing in Preimplantation Embryos.

We developed the Rainbow-seq technology to trace cell division history and reveal single-cell transcriptomes. With distinct fluorescent protein genes as lineage markers, Rainbow-seq enables each single-cell RNA sequencing (RNA-seq) experiment to simultaneously decode the lineage marker genes and read single-cell transcriptomes. We triggered lineage tracking in each blastomere at the 2-cell stage, observed microscopically inequivalent contributions of the progeny to the two embryonic poles at the blastocyst stage, and analyzed every single cell at either 4- or 8-cell stage with deep paired-end sequencing of full-length transcripts. Although lineage difference was not marked unequivocally at a single-gene level, it became clear when the transcriptome was analyzed as a whole. Moreover, several groups of novel transcript isoforms with embedded repeat sequences exhibited lineage difference, suggesting a possible link between DNA demethylation and cell fate decision. Rainbow-seq bridged a critical gap between division history and single-cell RNA-seq assays.

SMARCAD1 Contributes to the Regulation of Naive Pluripotency by Interacting with Histone Citrullination.

Histone citrullination regulates diverse cellular processes. Here, we report that SMARCAD1 preferentially associates with H3 arginine 26 citrullination (H3R26Cit) peptides present on arrays composed of 384 histone peptides harboring distinct post-transcriptional modifications. Among ten histone modifications assayed by ChIP-seq, H3R26Cit exhibited the most extensive genomewide co-localization with SMARCAD1 binding. Increased Smarcad1 expression correlated with naive pluripotency in pre-implantation embryos. In the presence of LIF, Smarcad1 knockdown (KD) embryonic stem cells lost naive state phenotypes but remained pluripotent, as suggested by morphology, gene expression, histone modifications, alkaline phosphatase activity, energy metabolism, embryoid bodies, teratoma, and chimeras. The majority of H3R26Cit ChIP-seq peaks occupied by SMARCAD1 were associated with increased levels of H3K9me3 in Smarcad1 KD cells. Inhibition of H3Cit induced H3K9me3 at the overlapping regions of H3R26Cit peaks and SMARCAD1 peaks. These data suggest a model in which SMARCAD1 regulates naive pluripotency by interacting with H3R26Cit and suppressing heterochromatin formation.

Transcriptomic insights into the genetic basis of mammalian limb diversity.

BACKGROUND: From bat wings to whale flippers, limb diversification has been crucial to the evolutionary success of mammals. We performed the first transcriptome-wide study of limb development in multiple species to explore the hypothesis that mammalian limb diversification has proceeded through the differential expression of conserved shared genes, rather than by major changes to limb patterning. Specifically, we investigated the manner in which the expression of shared genes has evolved within and among mammalian species. RESULTS: We assembled and compared transcriptomes of bat, mouse, opossum, and pig fore- and hind limbs at the ridge, bud, and paddle stages of development. Results suggest that gene expression patterns exhibit larger variation among species during later than earlier stages of limb development, while within species results are more mixed. Consistent with the former, results also suggest that genes expressed at later developmental stages tend to have a younger evolutionary age than genes expressed at earlier stages. A suite of key limb-patterning genes was identified as being differentially expressed among the homologous limbs of all species. However, only a small subset of shared genes is differentially expressed in the fore- and hind limbs of all examined species. Similarly, a small subset of shared genes is differentially expressed within the fore- and hind limb of a single species and among the forelimbs of different species. CONCLUSIONS: Taken together, results of this study do not support the existence of a phylotypic period of limb development ending at chondrogenesis, but do support the hypothesis that the hierarchical nature of development translates into increasing variation among species as development progresses.

Systematic Mapping of RNA-Chromatin Interactions In Vivo.

RNA molecules can attach to chromatin. It remains difficult to know what RNAs are associated with chromatin and where the genomic target loci of these RNAs are. Here, we present MARGI (mapping RNA-genome interactions), a technology to massively reveal native RNA-chromatin interactions from unperturbed cells. The gist of this technology is to ligate chromatin-associated RNAs (caRNAs) with their target genomic sequences by proximity ligation, forming RNA-DNA chimeric sequences, which are converted to a sequencing library for paired-end sequencing. Using MARGI, we produced RNA-genome interaction maps for human embryonic stem cells (ESCs) and human embryonic kidney (HEK) cells. MARGI revealed hundreds of caRNAs, including previously known XIST, SNHG1, NEAT1, and MALAT1, as well as each caRNA's genomic interaction loci. Using a cross-species experiment, we estimated that approximately 2.2% of MARGI-identified interactions were false positives. In ESCs and HEK cells, the RNA ends of more than 5% of MARGI read pairs were mapped to distal or inter-chromosomal locations as compared to the locations of their corresponding DNA ends. The majority of transcription start sites are associated with distal or inter-chromosomal caRNAs. Chromatin-immunoprecipitation-sequencing (ChIP-seq)-reported H3K27ac and H3K4me3 levels are positively correlated, while H3K9me3 is negatively correlated, with MARGI-reported RNA attachment levels. The MARGI technology should facilitate revealing novel RNA functions and their genomic target regions.

The Relationship between Gene Network Structure and Expression Variation among Individuals and Species.

Variation among individuals is a prerequisite of evolution by natural selection. As such, identifying the origins of variation is a fundamental goal of biology. We investigated the link between gene interactions and variation in gene expression among individuals and species using the mammalian limb as a model system. We first built interaction networks for key genes regulating early (outgrowth; E9.5-11) and late (expansion and elongation; E11-13) limb development in mouse. This resulted in an Early (ESN) and Late (LSN) Stage Network. Computational perturbations of these networks suggest that the ESN is more robust. We then quantified levels of the same key genes among mouse individuals and found that they vary less at earlier limb stages and that variation in gene expression is heritable. Finally, we quantified variation in gene expression levels among four mammals with divergent limbs (bat, opossum, mouse and pig) and found that levels vary less among species at earlier limb stages. We also found that variation in gene expression levels among individuals and species are correlated for earlier and later limb development. In conclusion, results are consistent with the robustness of the ESN buffering among-individual variation in gene expression levels early in mammalian limb development, and constraining the evolution of early limb development among mammalian species.

Genomic analysis of hepatic farnesoid X receptor binding sites reveals altered binding in obesity and direct gene repression by farnesoid X receptor in mice.

UNLABELLED: The nuclear bile acid receptor, farnesoid X receptor (FXR), is an important transcriptional regulator of liver metabolism. Despite recent advances in understanding its functions, how FXR regulates genomic targets and whether the transcriptional regulation by FXR is altered in obesity remain largely unknown. Here, we analyzed hepatic genome-wide binding sites of FXR in healthy and dietary obese mice by chromatin immunoprecipitation sequencing (ChIP-seq) analysis. A total of 15,263 and 5,272 FXR binding sites were identified in livers of healthy and obese mice, respectively, after a short 1-hour treatment with the synthetic FXR agonist, GW4064. Of these sites, 7,440 and 2,344 were detected uniquely in healthy and obese mice. FXR-binding sites were localized mostly in intergenic and intron regions at an inverted repeat 1 motif in both groups, but also clustered within 1 kilobase of transcription start sites. FXR-binding sites were detected near previously unknown target genes with novel functions, including diverse cellular signaling pathways, apoptosis, autophagy, hypoxia, inflammation, RNA processing, metabolism of amino acids, and transcriptional regulators. Further analyses of randomly selected genes from both healthy and obese mice suggested that more FXR-binding sites are likely functionally inactive in obesity. Surprisingly, occupancies of FXR, retinoid X receptor alpha, RNA polymerase II, and epigenetic gene activation and repression histone marks, and messenger RNA levels of genes examined, suggested that direct gene repression by agonist-activated FXR is common. CONCLUSION: Comparison of genomic FXR-binding sites in healthy and obese mice suggested that FXR transcriptional signaling is altered in dietary obese mice, which may underlie aberrant metabolism and liver function in obesity.

Mapping personal functional data to personal genomes.

MOTIVATION: The sequencing of personal genomes enabled analysis of variation in transcription factor (TF) binding, chromatin structure and gene expression and indicated how they contribute to phenotypic variation. It is hypothesized that using the reference genome for mapping ChIP-seq or RNA-seq reads may introduce errors, especially at polymorphic genomic regions. RESULTS: We developed a Personal Genome Editor (perEditor) that changes the reference human genome (NCBI36/hg18) into an individual genome, taking into account single nucleotide polymorphisms (SNPs), insertions and deletions, copy number variation, and chromosomal rearrangements. perEditor outputs two alleles (maternal, paternal) of the individual genome that is ready for mapping ChIP-seq and RNA-seq reads, and enabling the analyses of allele specific binding, chromatin structure and gene expression. AVAILABILITY: perEditor is available at http://biocomp.bioen.uiuc.edu/perEditor. CONTACT: szhong@illinois.edu.