cnnImpute: missing value recovery for single cell RNA sequencing data.

The advent of single-cell RNA sequencing (scRNA-seq) technology has revolutionized our ability to explore cellular diversity and unravel the complexities of intricate diseases. However, due to the inherently low signal-to-noise ratio and the presence of an excessive number of missing values, scRNA-seq data analysis encounters unique challenges. Here, we present cnnImpute, a novel convolutional neural network (CNN) based method designed to address the issue of missing data in scRNA-seq. Our approach starts by estimating missing probabilities, followed by constructing a CNN-based model to recover expression values with a high likelihood of being missing. Through comprehensive evaluations, cnnImpute demonstrates its effectiveness in accurately imputing missing values while preserving the integrity of cell clusters in scRNA-seq data analysis. It achieved superior performance in various benchmarking experiments. cnnImpute offers an accurate and scalable method for recovering missing values, providing a useful resource for scRNA-seq data analysis.

Shortcut barcoding and early pooling for scalable multiplex single-cell reduced-representation CpG methylation sequencing at single nucleotide resolution.

DNA methylation is essential for a wide variety of biological processes, yet the development of a highly efficient and robust technology remains a challenge for routine single-cell analysis. We developed a multiplex scalable single-cell reduced representation bisulfite sequencing (msRRBS) technology. It allows cell-specific barcoded DNA fragments of individual cells to be pooled before bisulfite conversion, free of enzymatic modification or physical capture of the DNA ends, and achieves read mapping rates of 62.5 ± 3.9%, covering 60.0 ± 1.4% of CpG islands and 71.6 ± 1.6% of promoters in K562 cells. Its reproducibility is shown in duplicates of bulk cells with close to perfect correlation (R = 0.97-0.99). At a low 1 Mb of clean reads, msRRBS provides highly consistent coverage of CpG islands and promoters, outperforming the conventional methods with orders of magnitude reduction in cost. Here, we use this method to characterize the distinct methylation patterns and cellular heterogeneity of six cell lines, plus leukemia and hepatocellular carcinoma models. Taking 4 h of hands-on time, msRRBS offers a unique, highly efficient approach for dissecting methylation heterogeneity in a variety of multicellular systems.

Integrating Single-Cell Transcriptome and Network Analysis to Characterize the Therapeutic Response of Chronic Myeloid Leukemia.

Chronic myeloid leukemia (CML) is a myeloproliferative disease characterized by a unique BCR-ABL fusion gene. Tyrosine kinase inhibitors (TKIs) were developed to target the BCR-ABL oncoprotein, inhibiting its abnormal kinase activity. TKI treatments have significantly improved CML patient outcomes. However, the patients can develop drug resistance and relapse after therapy discontinues largely due to intratumor heterogeneity. It is critical to understand the differences in therapeutic responses among subpopulations of cells. Single-cell RNA sequencing measures the transcriptome of individual cells, allowing us to differentiate and analyze individual cell populations. Here, we integrated a single-cell RNA sequencing profile of CML stem cells and network analysis to decipher the mechanisms of distinct TKI responses. Compared to normal hematopoietic stem cells, a set of genes that were concordantly differentially expressed in various types of stem cells of CML patients was revealed. Further transcription regulatory network analysis found that most of these genes were directly controlled by one or more transcript factors and the genes have more regulators in the cells of the patients who responded to the treatment. The molecular markers including a known drug-resistance gene and novel gene signatures for treatment response were also identified. Moreover, we combined protein-protein interaction network construction with a cancer drug database and uncovered the drugs that target the marker genes directly or indirectly via the protein interactions. The gene signatures and their interacted proteins identified by this work can be used for treatment response prediction and lead to new strategies for drug resistance monitoring and prevention. Our single-cell-based findings offered novel insights into the mechanisms underlying the therapeutic response of CML.

Sample-multiplexing approaches for single-cell sequencing.

Single-cell sequencing is widely used in biological and medical studies. However, its application with multiple samples is hindered by inefficient sample processing, high experimental costs, ambiguous identification of true single cells, and technical batch effects. Here, we introduce sample-multiplexing approaches for single-cell sequencing in transcriptomics, epigenomics, genomics, and multiomics. In single-cell transcriptomics, sample multiplexing uses variants of native or artificial features as sample markers, enabling sample pooling and decoding. Such features include: (1) natural genetic variation, (2) nucleotide-barcode anchoring on cellular or nuclear membranes, (3) nucleotide-barcode internalization to the cytoplasm or nucleus, (4) vector-based barcode expression in cells, and (5) nucleotide-barcode incorporation during library construction. Other single-cell omics methods are based on similar concepts, particularly single-cell combinatorial indexing. These methods overcome current challenges, while enabling super-loading of single cells. Finally, selection guidelines are presented that can accelerate technological application.

Spatial transcriptome profiling by MERFISH reveals fetal liver hematopoietic stem cell niche architecture.

The hematopoietic stem cell (HSC) niche has been extensively studied in bone marrow, yet a more systematic investigation into the microenvironment regulation of hematopoiesis in fetal liver is necessary. Here we investigate the spatial organization and transcriptional profile of individual cells in both wild type (WT) and Tet2-/- fetal livers, by multiplexed error robust fluorescence in situ hybridization. We find that specific pairs of fetal liver cell types are preferentially positioned next to each other. Ligand-receptor signaling molecule pairs such as Kitl and Kit are enriched in neighboring cell types. The majority of HSCs are in direct contact with endothelial cells (ECs) in both WT and Tet2-/- fetal livers. Loss of Tet2 increases the number of HSCs, and upregulates Wnt and Notch signaling genes in the HSC niche. Two subtypes of ECs, arterial ECs and sinusoidal ECs, and other cell types contribute distinct signaling molecules to the HSC niche. Collectively, this study provides a comprehensive picture and bioinformatic foundation for HSC spatial regulation in fetal liver.

Colorectal Cancer Stem Cell States Uncovered by Simultaneous Single-Cell Analysis of Transcriptome and Telomeres.

Cancer stem cells (CSCs) presumably contribute to tumor progression and drug resistance, yet their definitive features have remained elusive. Here, simultaneous measurement of telomere length and transcriptome in the same cells enables systematic assessment of CSCs in primary colorectal cancer (CRC). The in-depth transcriptome profiled by SMART-seq2 is independently validated by high-throughput scRNA-seq using 10 × Genomics. It is found that rare CSCs exist in dormant state and display plasticity toward cancer epithelial cells (EPCs) that essentially are presumptive tumor-initiating cells (TICs), while both retaining the prominent signaling pathways including WNT, TGF-ß, and HIPPO/YAP. Moreover, CSCs exhibit chromosome copy number variation (CNV) pattern resembling cancer EPCs but distinct from normal stem cells, suggesting the phylogenetic relationship between CSCs and cancer EPCs. Notably, CSCs maintain shorter telomeres and possess minimal telomerase activity consistent with their nonproliferative nature, unlike cancer EPCs. Additionally, the specific signature of CSCs particularly NOTUM, SMOC2, BAMBI, PHLDA1, and TNFRSF19 correlates with the prognosis of CRC. These findings characterize the heterogeneity of CSCs and their linkage to cancer EPCs/TICs, some of which are conventionally regarded as CSCs.

Dysregulation of BRD4 Function Underlies the Functional Abnormalities of MeCP2 Mutant Neurons.

Rett syndrome (RTT), mainly caused by mutations in methyl-CpG binding protein 2 (MeCP2), is one of the most prevalent intellectual disorders without effective therapies. Here, we used 2D and 3D human brain cultures to investigate MeCP2 function. We found that MeCP2 mutations cause severe abnormalities in human interneurons (INs). Surprisingly, treatment with a BET inhibitor, JQ1, rescued the molecular and functional phenotypes of MeCP2 mutant INs. We uncovered that abnormal increases in chromatin binding of BRD4 and enhancer-promoter interactions underlie the abnormal transcription in MeCP2 mutant INs, which were recovered to normal levels by JQ1. We revealed cell-type-specific transcriptome impairment in MeCP2 mutant region-specific human brain organoids that were rescued by JQ1. Finally, JQ1 ameliorated RTT-like phenotypes in mice. These data demonstrate that BRD4 dysregulation is a critical driver for RTT etiology and suggest that targeting BRD4 could be a potential therapeutic opportunity for RTT.

An Overview of Advances in Cell-Based Cancer Immunotherapies Based on the Multiple Immune-Cancer Cell Interactions.

Tumors have a complex ecosystem in which behavior and fate are determined by the interaction of diverse cancerous and noncancerous cells at local and systemic levels. A number of studies indicate that various immune cells participate in tumor development (Fig. 1). In this review, we will discuss interactions among T lymphocytes (T cells), B cells, natural killer (NK) cells, dendritic cells (DCs), tumor-associated macrophages (TAMs), neutrophils, and myeloid-derived suppressor cells (MDSCs). In addition, we will touch upon attempts to either use or block subsets of immune cells to target cancer.

A versatile flow-based assay for immunocyte-mediated cytotoxicity.

Cell-mediated cytotoxicity is a critical function of the immune system in mounting defense against pathogens and cancers. Current methods that allow direct evaluation of cell-mediated cytotoxicity suffer from a wide-range of drawbacks. Here, we present a novel strategy to measure cytotoxicity that is direct, sensitive, rapid, and highly adaptable. Moreover, it allows accurate measurement of viability of both target and effector cells. Target cells are fluorescently labeled with a non-toxic, cell-permeable dye that covalently binds to cell proteins, including nuclear proteins. The labeled target cells are incubated with effector cells to begin killing. Following the killing reaction, the cell mixture is incubated with another dye that specifically stains proteins of dead cells, including nuclear proteins. In the final step, cell nuclei are released by Triton X-100, and analyzed by flow cytometry. This results in four nuclear staining patterns that separate target and effector nuclei as well as nuclei of live and dead cells. Analyzing nuclei, instead of cells, greatly reduces flow cytometry errors caused by the presence of target-effector cell aggregates. Target killing time can often be reduced to 2 h and the assay can be done in a high throughput format. We have successfully validated this assay in a variety of cytotoxicity scenarios including those mediated by NK-92 cells, Chimeric Antigen Receptor (CAR)-T cells, and Tumor Infiltrating Lymphocytes (TIL). Therefore, this technique is broadly applicable, highly sensitive and easily administered, making it a powerful tool to assess immunotherapy-based, cell-mediated cytotoxicity.

Telomere Maintenance-Associated PML Is a Potential Specific Therapeutic Target of Human Colorectal Cancer.

Telomere length maintenance is essential for cell proliferation, which is particularly prominent in cancer. We validate that the primary colorectal tumors exhibit heterogeneous telomere lengths but mostly (90%) short telomeres relative to normal tissues. Intriguingly, relatively short telomeres are associated with tumor malignancy as indicated by poorly differentiated state, and these tumors contain more cancer stem-like cells (CSLCs) identified by several commonly used markers CD44, EPHB2 or LGR5. Moreover, promyelocytic leukemia (PML) and ALT-associated PML nuclear bodies (APBs) are frequently found in tumors with short telomeres and high proliferation. In contrast, distant normal tissues rarely or only minimally express PML. Inhibition of PML and APBs by an ATR inhibitor decreases proliferation of CSLCs and organoids, suggesting a potential therapeutic target to progressive colorectal tumors. Together, telomere maintenance underling tumor progression is connected with CSLCs.

hESC-Derived Thalamic Organoids Form Reciprocal Projections When Fused with Cortical Organoids.

Human brain organoid techniques have rapidly advanced to facilitate investigating human brain development and diseases. These efforts have largely focused on generating telencephalon due to its direct relevance in a variety of forebrain disorders. Despite its importance as a relay hub between cortex and peripheral tissues, the investigation of three-dimensional (3D) organoid models for the human thalamus has not been explored. Here, we describe a method to differentiate human embryonic stem cells (hESCs) to thalamic organoids (hThOs) that specifically recapitulate the development of thalamus. Single-cell RNA sequencing revealed a formation of distinct thalamic lineages, which diverge from telencephalic fate. Importantly, we developed a 3D system to create the reciprocal projections between thalamus and cortex by fusing the two distinct region-specific organoids representing the developing thalamus or cortex. Our study provides a platform for understanding human thalamic development and modeling circuit organizations and related disorders in the brain.

Correction to: MISC: missing imputation for single-cell RNA sequencing data.

It was highlighted that the original article [1] contained a typesetting error in the last name of Allon Canaan. This was incorrectly captured as Allon Canaann in the original article which has since been updated.

Local and global chromatin interactions are altered by large genomic deletions associated with human brain development.

Large copy number variants (CNVs) in the human genome are strongly associated with common neurodevelopmental, neuropsychiatric disorders such as schizophrenia and autism. Here we report on the epigenomic effects of the prominent large deletion CNVs on chromosome 22q11.2 and on chromosome 1q21.1. We use Hi-C analysis of long-range chromosome interactions, including haplotype-specific Hi-C analysis, ChIP-Seq analysis of regulatory histone marks, and RNA-Seq analysis of gene expression patterns. We observe changes on all the levels of analysis, within the deletion boundaries, in the deletion flanking regions, along chromosome 22q, and genome wide. We detect gene expression changes as well as pronounced and multilayered effects on chromatin states, chromosome folding and on the topological domains of the chromatin, that emanate from the large CNV locus. These findings suggest basic principles of how such large genomic deletions can alter nuclear organization and affect genomic molecular activity.

MISC: missing imputation for single-cell RNA sequencing data.

BACKGROUND: Single-cell RNA sequencing (scRNA-seq) technology provides an effective way to study cell heterogeneity. However, due to the low capture efficiency and stochastic gene expression, scRNA-seq data often contains a high percentage of missing values. It has been showed that the missing rate can reach approximately 30% even after noise reduction. To accurately recover missing values in scRNA-seq data, we need to know where the missing data is; how much data is missing; and what are the values of these data. METHODS: To solve these three problems, we propose a novel model with a hybrid machine learning method, namely, missing imputation for single-cell RNA-seq (MISC). To solve the first problem, we transformed it to a binary classification problem on the RNA-seq expression matrix. Then, for the second problem, we searched for the intersection of the classification results, zero-inflated model and false negative model results. Finally, we used the regression model to recover the data in the missing elements. RESULTS: We compared the raw data without imputation, the mean-smooth neighbor cell trajectory, MISC on chronic myeloid leukemia data (CML), the primary somatosensory cortex and the hippocampal CA1 region of mouse brain cells. On the CML data, MISC discovered a trajectory branch from the CP-CML to the BC-CML, which provides direct evidence of evolution from CP to BC stem cells. On the mouse brain data, MISC clearly divides the pyramidal CA1 into different branches, and it is direct evidence of pyramidal CA1 in the subpopulations. In the meantime, with MISC, the oligodendrocyte cells became an independent group with an apparent boundary. CONCLUSIONS: Our results showed that the MISC model improved the cell type classification and could be instrumental to study cellular heterogeneity. Overall, MISC is a robust missing data imputation model for single-cell RNA-seq data.

Ritornello: high fidelity control-free chromatin immunoprecipitation peak calling.

With the advent of next generation high-throughput DNA sequencing technologies, omics experiments have become the mainstay for studying diverse biological effects on a genome wide scale. Chromatin immunoprecipitation (ChIP-seq) is the omics technique that enables genome wide localization of transcription factor (TF) binding or epigenetic modification events. Since the inception of ChIP-seq in 2007, many methods have been developed to infer ChIP-target binding loci from the resultant reads after mapping them to a reference genome. However, interpreting these data has proven challenging, and as such these algorithms have several shortcomings, including susceptibility to false positives due to artifactual peaks, poor localization of binding sites and the requirement for a total DNA input control which increases the cost of performing these experiments. We present Ritornello, a new approach for finding TF-binding sites in ChIP-seq, with roots in digital signal processing that addresses all of these problems. We show that Ritornello generally performs equally or better than the peak callers tested and recommended by the ENCODE consortium, but in contrast, Ritornello does not require a matched total DNA input control to avoid false positives, effectively decreasing the sequencing cost to perform ChIP-seq. Ritornello is freely available at https://github.com/KlugerLab/Ritornello.

Fusion of Regionally Specified hPSC-Derived Organoids Models Human Brain Development and Interneuron Migration.

Organoid techniques provide unique platforms to model brain development and neurological disorders. Whereas several methods for recapitulating corticogenesis have been described, a system modeling human medial ganglionic eminence (MGE) development, a critical ventral brain domain producing cortical interneurons and related lineages, has been lacking until recently. Here, we describe the generation of MGE and cortex-specific organoids from human pluripotent stem cells that recapitulate the development of MGE and cortex domains, respectively. Population and single-cell RNA sequencing (RNA-seq) profiling combined with bulk assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq) analyses revealed transcriptional and chromatin accessibility dynamics and lineage relationships during MGE and cortical organoid development. Furthermore, MGE and cortical organoids generated physiologically functional neurons and neuronal networks. Finally, fusing region-specific organoids followed by live imaging enabled analysis of human interneuron migration and integration. Together, our study provides a platform for generating domain-specific brain organoids and modeling human interneuron migration and offers deeper insight into molecular dynamics during human brain development.

One thousand somatic SNVs per skin fibroblast cell set baseline of mosaic mutational load with patterns that suggest proliferative origin.

Few studies have been conducted to understand post-zygotic accumulation of mutations in cells of the healthy human body. We reprogrammed 32 skin fibroblast cells from families of donors into human induced pluripotent stem cell (hiPSC) lines. The clonal nature of hiPSC lines allows a high-resolution analysis of the genomes of the founder fibroblast cells without being confounded by the artifacts of single-cell whole-genome amplification. We estimate that on average a fibroblast cell in children has 1035 mostly benign mosaic SNVs. On average, 235 SNVs could be directly confirmed in the original fibroblast population by ultradeep sequencing, down to an allele frequency (AF) of 0.1%. More sensitive droplet digital PCR experiments confirmed more SNVs as mosaic with AF as low as 0.01%, suggesting that 1035 mosaic SNVs per fibroblast cell is the true average. Similar analyses in adults revealed no significant increase in the number of SNVs per cell, suggesting that a major fraction of mosaic SNVs in fibroblasts arises during development. Mosaic SNVs were distributed uniformly across the genome and were enriched in a mutational signature previously observed in cancers and in de novo variants and which, we hypothesize, is a hallmark of normal cell proliferation. Finally, AF distribution of mosaic SNVs had distinct narrow peaks, which could be a characteristic of clonal cell selection, clonal expansion, or both. These findings reveal a large degree of somatic mosaicism in healthy human tissues, link de novo and cancer mutations to somatic mosaicism, and couple somatic mosaicism with cell proliferation.

Detecting heterogeneity in single-cell RNA-Seq data by non-negative matrix factorization.

Single-cell RNA-Sequencing (scRNA-Seq) is a fast-evolving technology that enables the understanding of biological processes at an unprecedentedly high resolution. However, well-suited bioinformatics tools to analyze the data generated from this new technology are still lacking. Here we investigate the performance of non-negative matrix factorization (NMF) method to analyze a wide variety of scRNA-Seq datasets, ranging from mouse hematopoietic stem cells to human glioblastoma data. In comparison to other unsupervised clustering methods including K-means and hierarchical clustering, NMF has higher accuracy in separating similar groups in various datasets. We ranked genes by their importance scores (D-scores) in separating these groups, and discovered that NMF uniquely identifies genes expressed at intermediate levels as top-ranked genes. Finally, we show that in conjugation with the modularity detection method FEM, NMF reveals meaningful protein-protein interaction modules. In summary, we propose that NMF is a desirable method to analyze heterogeneous single-cell RNA-Seq data. The NMF based subpopulation detection package is available at: https://github.com/lanagarmire/NMFEM.