Publications

Inheritance of 5-methylcytosine (5mC) from one cell generation to the next by DNMT1 plays a key role in regulating cellular identity. While recent work has shown that the activity of DNMT1 is imprecise; it remains unclear how the fidelity of DNMT1 is tuned in different genomic and cell state contexts. Here we describe Dyad-seq, a method to quantify the genome-wide methylation status of cytosines at the resolution of individual CpG dinucleotides to find that the fidelity of DNMT1-mediated maintenance methylation is related to the local density of DNA methylation and the landscape of histone modifications. To gain deeper insights into methylation/demethylation turnover dynamics, we first extended Dyad-seq to quantify all combinations of 5mC and 5-hydroxymethylcytosine at individual CpG dyads. Next, to understand how cell state transitions impact maintenance methylation, we scaled the method down to jointly profile genome-wide methylation levels, maintenance methylation fidelity and the transcriptome from single cells (scDyad&T-seq). Using scDyad&T-seq we demonstrate that while distinct cell states can substantially impact the activity of the maintenance methylation machinery, locally there exists an intrinsic relationship between DNA methylation density, histone modifications and DNMT1-mediated maintenance methylation fidelity that is independent of cell state.

Bacterial mRNA sequencing is inefficient due to the abundance of ribosomal RNA that is challenging to deplete. While commercial kits target rRNA from common bacterial species, they are frequently inefficient when applied to divergent species, including those from environmental isolates. Similarly, other methods typically employ large probe sets that tile the entire length of rRNAs; however, such approaches are infeasible when applied to many species. Therefore, we present EMBR-seq+, which requires fewer than ten oligonucleotides per rRNA by combining rRNA blocking primers with RNase H-mediated depletion to achieve rRNA removal efficiencies of up to 99% in diverse bacterial species. Further, in more complex microbial co-cultures between F. succinogenes strain UWB7 and anerobic fungi, EMBR-seq+ depleted both bacterial and fungal rRNA, with a 4-fold improvement in bacterial rRNA depletion compared to a commercial kit, thereby demonstrating that the method can be applied to non-model microbial mixtures. Notably, for microbes with unknown rRNA sequences, EMBR-seq+ enables rapid iterations in probe design without requiring to start experiments from total RNA. Finally, efficient depletion of rRNA enabled systematic quantification of the reprogramming of the bacterial transcriptome when cultured in the presence of the anerobic fungi Anaeromyces robustus or Caecomyces churrovis. We observed that F. succinogenes strain UWB7 downregulated several lignocellulose-degrading carbohydrate-active enzymes in the presence of anerobic gut fungi, suggesting close interactions between two cellulolytic species that specialize in different aspects of biomass breakdown. Thus, EMBR-seq+ enables efficient, cost-effective and rapid quantification of the transcriptome to gain insights into non-model microbial systems.

The emergence of different cell types and the role of the epigenome in regulating transcription is a key yet understudied event during human gastrulation. Investigating these questions remain infeasible due to the lack of availability of embryos at these stages of development. Further, human gastrulation is marked by dynamic changes in cell states that are difficult to isolate at high purity, thereby making it challenging to map how epigenetic reprogramming impacts gene expression and cellular phenotypes. To overcome these limitations, we describe scMAT-seq, a high-throughput one-pot single-cell multiomics technology to simultaneously quantify DNA methylation, DNA accessibility and the transcriptome from the same cell. Applying scMAT-seq to 3D human gastruloids, we characterized the epigenetic landscape of major cell types corresponding to the germ layers and primordial germ cell-like cells (hPGCLC). As the identity of the progenitors that give rise to human PGCLCs remain unclear, we used this system to discover that the progenitors emerge from epiblast cells and show transient characteristics of both amniotic- and mesoderm-like cells, before getting specified towards hPGCLCs. Finally, as cells differentiate along different lineages during gastrulation, we surprisingly find that while changes in DNA accessibility are tightly correlated to both upregulated and downregulated genes, reorganization of gene body DNA methylation is strongly related to only genes that get downregulated, with genes that turn on displaying a lineage trajectory-dependent correlation with DNA methylation. Collectively, these results demonstrate that scMAT-seq is a high-throughput and sensitive approach to elucidate epigenetic regulation of gene expression in complex systems such as human gastrulation that are marked by rapidly transitioning cell states.

Lineage reconstruction is central to understanding tissue development and maintenance. To overcome the limitations of current techniques that typically reconstruct clonal trees using genetically encoded reporters, we report scPECLR, a probabilistic algorithm to endogenously infer lineage trees at a single cell-division resolution using 5-hydroxymethylcytosine (5hmC). When applied to 8-cell preimplantation mouse embryos, scPECLR predicts the full lineage tree with greater than 95% accuracy. In addition, we developed scH&G-seq to sequence both 5hmC and genomic DNA from the same cell. As genomic DNA sequencing yields information on both copy-number variations and single-nucleotide polymorphisms, when combined with scPECLR, it enables more accurate lineage reconstruction of larger trees. Finally, we show that scPECLR can also be used to map chromosome strand segregation patterns during cell division, thereby providing a strategy to test the “immortal strand” hypothesis. Thus, scPECLR provides a generalized method to endogenously reconstruct lineage trees at an individual cell division resolution.

RNA sequencing is a powerful approach to quantify the genome-wide distribution of mRNA molecules in a population to gain deeper understanding of cellular functions and phenotypes. However, unlike eukaryotic cells, mRNA sequencing of bacterial samples is more challenging due to the absence of a poly-A tail that typically enables efficient capture and enrichment of mRNA from the abundant rRNA molecules in a cell. Moreover, bacterial cells frequently contain 100-fold lower quantities of RNA compared to mammalian cells, which further complicates mRNA sequencing from non-cultivable and non-model bacterial species. To overcome these limitations, we report EMBR-seq (Enrichment of mRNA by Blocked rRNA), a method that efficiently depletes 5S, 16S and 23S rRNA using blocking primers to prevent their amplification. EMBR-seq results in 90% of the sequenced RNA molecules from an E. coli culture deriving from mRNA. We demonstrate that this increased efficiency provides a deeper view of the transcriptome without introducing technical amplification-induced biases. Moreover, compared to recent methods that employ a large array of oligonucleotides to deplete rRNA, EMBR-seq uses a single or a few oligonucleotides per rRNA, thereby making this new technology significantly more cost-effective, especially when applied to varied bacterial species. Finally, compared to existing commercial kits for bacterial rRNA depletion, we show that EMBR-seq can be used to successfully quantify the transcriptome from more than 500-fold lower starting total RNA. EMBR-seq provides an efficient and cost-effective approach to quantify global gene expression profiles from low input bacterial samples.

The development of an in vitro system in which human primordial germ cell-like cells (hPGCLCs) are generated from human pluripotent stem cells (hPSCs) has been invaluable to further our understanding of human primordial germ cell (hPGC) specification. However, the means to evaluate the next fundamental steps in germ cell development have not been well established. In this study we describe a two dimensional extended culture system that promotes proliferation of specified hPGCLCs, without reversion to a pluripotent state. We demonstrate that hPGCLCs in extended culture undergo partial epigenetic reprogramming, mirroring events described in hPGCs in vivo, including a genome-wide reduction in DNA methylation and maintenance of depleted H3K9me2. This extended culture system provides a new approach for expanding the number of hPGCLCs for downstream technologies, including transplantation, molecular screening, or possibly the differentiation of hPGCLCs into gametes by in vitro gametogenesis.

Lineage reconstruction is central to understanding tissue development and maintenance. While powerful tools to infer cellular relationships have been developed, these methods typically have a clonal resolution that prevent the reconstruction of lineage trees at an individual cell division resolution. Moreover, these methods require a transgene, which poses a significant barrier in the study of human tissues. To overcome these limitations, we report scPECLR, a probabilistic algorithm to endogenously infer lineage trees at a single cell-division resolution using 5-hydroxymethylcytosine. When applied to 8-cell preimplantation mouse embryos, scPECLR predicts the full lineage tree with greater than 95% accuracy. Further, scPECLR can accurately extract lineage information for a majority of cells when reconstructing larger trees. Finally, we show that scPECLR can also be used to map chromosome strand segregation patterns during cell division, thereby providing a strategy to test the “immortal strand” hypothesis in stem cell biology. Thus, scPECLR provides a generalized method to endogenously reconstruct lineage trees at an individual cell-division resolution.

The epigenetic DNA modification 5-hydroxymethylcytosine (5hmC) has crucial roles in development and gene regulation. Quantifying the abundance of this epigenetic mark at the single-cell level could enable us to understand its roles. We present a single-cell, genome-wide and strand-specific 5hmC sequencing technology, based on 5hmC glucosylation and glucosylation-dependent digestion of DNA, that reveals pronounced cell-to-cell variability in the abundance of 5hmC on the two DNA strands of a given chromosome. We develop a mathematical model that reproduces the strand bias and use this model to make two predictions. First, the variation in strand bias should decrease when 5hmC turnover increases. Second, the strand bias of two sister cells should be strongly anti-correlated. We validate these predictions experimentally, and use our model to reconstruct lineages of two- and four-cell mouse embryos, showing that single-cell 5hmC sequencing can be used as a lineage reconstruction tool.

While gene expression noise has been shown to drive dramatic phenotypic variations, the molecular basis for this variability in mammalian systems is not well understood. Gene expression has been shown to be regulated by promoter architecture and the associated chromatin environment. However, the exact contribution of these two factors in regulating expression noise has not been explored. Using a dual-reporter lentiviral model system, we deconvolved the influence of the promoter sequence to systematically study the contribution of the chromatin environment at different genomic locations in regulating expression noise. By integrating a large-scale analysis to quantify mRNA levels by smFISH and protein levels by flow cytometry in single cells, we found that mean expression and noise are uncorrelated across genomic locations. Furthermore, we showed that this independence could be explained by the orthogonal control of mean expression by the transcript burst size and noise by the burst frequency. Finally, we showed that genomic locations displaying higher expression noise are associated with more repressed chromatin, thereby indicating the contribution of the chromatin environment in regulating expression noise.

Human immunodeficiency virus 1 (HIV) latency remains a significant obstacle to curing infected patients. One promising therapeutic strategy is to purge the latent cellular reservoir by activating latent HIV with latency-reversing agents (LRAs). In some cases, co-drugging with multiple LRAs is necessary to activate latent infections, but few studies have established quantitative criteria for determining when co-drugging is required. Here we systematically quantified drug interactions between histone deacetylase inhibitors and transcriptional activators of HIV and found that the need for co-drugging is determined by the proximity of latent infections to the chromatin-regulated viral gene activation threshold at the viral promoter. Our results suggest two classes of latent viral integrations: those far from the activation threshold that benefit from co-drugging, and those close to the threshold that are efficiently activated by a single drug. Using a primary T cell model of latency, we further demonstrated that the requirement for co-drugging was donor dependent, suggesting that the host may set the level of repression of latent infections. Finally, we showed that single drug or co-drugging doses could be optimized, via repeat stimulations, to minimize unwanted side effects while main- taining robust viral activation. Our results motivate further study of patient-specific latency-reversing strategies.

Viral genomes are continually subjected to mutations, and functionally deleterious ones can be rescued by reversion or additional mutations that restore fitness. The error prone nature of HIV-1 replication has resulted in highly diverse viral sequences, and it is not clear how viral proteins such as Tat, which plays a critical role in viral gene expression and replication, retain their complex functions. Although several important amino acid positions in Tat are conserved, we hypothesized that it may also harbor functionally important residues that may not be individually conserved yet appear as correlated pairs, whose analysis could yield new mechanistic insights into Tat function and evolution. To identify such sites, we combined mutual information analysis and experimentation to identify coevolving positions and found that residues 35 and 39 are strongly correlated. Mutation of either residue of this pair into amino acids that appear in numerous viral isolates yields a defective virus; however, simultaneous introduction of both mutations into the heterologous Tat sequence restores gene expression close to wild-type Tat. Furthermore, in contrast to most coevolving protein residues that contribute to the same function, structural modeling and biochemical studies showed that these two residues contribute to two mechanistically distinct steps in gene expression: binding P-TEFb and promoting P-TEFb phosphorylation of the C-terminal domain in RNAPII. Moreover, Tat variants that mimic HIV-1 subtypes B or C at sites 35 and 39 have evolved orthogonal strengths of P-TEFb binding versus RNAPII phosphorylation, suggesting that subtypes have evolved alternate transcriptional strategies to achieve similar gene expression levels.

Because of its generality, thermodynamics is applicable to all substances, including biomacromolecules. To illustrate how thermodynamics can contribute to biotechnology, each of six examples gives a brief summary of pertinent, previously published research. Each example indicates that familiar concepts in chemical engineering thermodynamics can be applied to contribute toward solution of a practical problem. These examples are discussed here to encourage thermodynamically oriented chemical engineers to devote their talents toward helping to advance industrial biotechnology.

The title compound, C₆H₁₂O₄, exists in a chair form, with three of the four OH groups equatorially disposed. All four hydroxy groups participate in extensive intermolecular O—H...O hydrogen bonding.