A machine learning method for the discovery of minimum marker gene combinations for cell type identification from single-cell RNA sequencing.

Single-cell genomics is rapidly advancing our knowledge of the diversity of cell phenotypes, including both cell types and cell states. Driven by single-cell/-nucleus RNA sequencing (scRNA-seq), comprehensive cell atlas projects characterizing a wide range of organisms and tissues are currently underway. As a result, it is critical that the transcriptional phenotypes discovered are defined and disseminated in a consistent and concise manner. Molecular biomarkers have historically played an important role in biological research, from defining immune cell types by surface protein expression to defining diseases by their molecular drivers. Here, we describe a machine learning-based marker gene selection algorithm, NS-Forest version 2.0, which leverages the nonlinear attributes of random forest feature selection and a binary expression scoring approach to discover the minimal marker gene expression combinations that optimally capture the cell type identity represented in complete scRNA-seq transcriptional profiles. The marker genes selected provide an expression barcode that serves as both a useful tool for downstream biological investigation and the necessary and sufficient characteristics for semantic cell type definition. The use of NS-Forest to identify marker genes for human brain middle temporal gyrus cell types reveals the importance of cell signaling and noncoding RNAs in neuronal cell type identity.

The genome and preliminary single-nuclei transcriptome of Lemna minuta reveals mechanisms of invasiveness.

The ability to trace every cell in some model organisms has led to the fundamental understanding of development and cellular function. However, in plants the complexity of cell number, organ size, and developmental time makes this a challenge even in the diminutive model plant Arabidopsis (Arabidopsis thaliana). Duckweed, basal nongrass aquatic monocots, provide an opportunity to follow every cell of an entire plant due to their small size, reduced body plan, and fast clonal growth habit. Here we present a chromosome-resolved genome for the highly invasive Lesser Duckweed (Lemna minuta) and generate a preliminary cell atlas leveraging low cell coverage single-nuclei sequencing. We resolved the 360 megabase genome into 21 chromosomes, revealing a core nonredundant gene set with only the ancient tau whole-genome duplication shared with all monocots, and paralog expansion as a result of tandem duplications related to phytoremediation. Leveraging SMARTseq2 single-nuclei sequencing, which provided higher gene coverage yet lower cell count, we profiled 269 nuclei covering 36.9% (8,457) of the L. minuta transcriptome. Since molecular validation was not possible in this nonmodel plant, we leveraged gene orthology with model organism single-cell expression datasets, gene ontology, and cell trajectory analysis to define putative cell types. We found that the tissue that we computationally defined as mesophyll expressed high levels of elemental transport genes consistent with this tissue playing a role in L. minuta wastewater detoxification. The L. minuta genome and preliminary cell map provide a paradigm to decipher developmental genes and pathways for an entire plant.

Application of machine learning models to animal health pharmacovigilance: A proof-of-concept study.

Machine learning (ML) models were applied to pharmacovigilance (PV) data in a two-component proof-of-concept study. PV data were partitioned into Training, Validation, and Holdout datasets for model training and selection. During the first component ML models were challenged to identify factors in individual case safety reports (ICSRs) involving spinosad and neurological and ocular clinical signs. The target feature for the models were these clinical signs that were disproportionately reported for spinosad. The endpoints were normalized coefficient values representing the relationship between the target feature and ICSR free text fields. The deployed model accurately identified the risk factors "demodectic," "demodicosis," and "ivomec." In the second component, the ML models were trained to identify high quality and complete ICSRs free of confounders. The deployed model was presented with an external Test dataset of six ICSRs, one that was complete, of high quality, and devoid of confounders, and five that were not. The endpoints were model-generated probabilities for the ICSRs. The deployed ML model accurately identified the ICSR of interest with a greater than 10-fold higher probability score. Although narrow in scope, the study supports further investigation and potential application of ML models to animal health PV data.

A niche-dependent myeloid transcriptome signature defines dormant myeloma cells.

The era of targeted therapies has seen significant improvements in depth of response, progression-free survival, and overall survival for patients with multiple myeloma. Despite these improvements in clinical outcome, patients inevitably relapse and require further treatment. Drug-resistant dormant myeloma cells that reside in specific niches within the skeleton are considered a basis of disease relapse but remain elusive and difficult to study. Here, we developed a method to sequence the transcriptome of individual dormant myeloma cells from the bones of tumor-bearing mice. Our analyses show that dormant myeloma cells express a distinct transcriptome signature enriched for immune genes and, unexpectedly, genes associated with myeloid cell differentiation. These genes were switched on by coculture with osteoblastic cells. Targeting AXL, a gene highly expressed by dormant cells, using small-molecule inhibitors released cells from dormancy and promoted their proliferation. Analysis of the expression of AXL and coregulated genes in human cohorts showed that healthy human controls and patients with monoclonal gammopathy of uncertain significance expressed higher levels of the dormancy signature genes than patients with multiple myeloma. Furthermore, in patients with multiple myeloma, the expression of this myeloid transcriptome signature translated into a twofold increase in overall survival, indicating that this dormancy signature may be a marker of disease progression. Thus, engagement of myeloma cells with the osteoblastic niche induces expression of a suite of myeloid genes that predicts disease progression and that comprises potential drug targets to eradicate dormant myeloma cells.

Evaluation of signal detection algorithms within the Elanco Animal Health Pharmacovigilance database.

Statistical algorithms for detecting safety signals are beginning to be applied to Animal Health Pharmacovigilance (PV) databases. How these signal detection algorithms (SDAs) perform in an animal health PV database is the subject of this report. Statistical methods and SDAs were assessed against a set of known signals in order to identify which SDAs were most appropriate for signal detection using the Elanco Animal Health PV database. A reference set of adverse events that should signal was created for 31 products across four species. Nine SDAs based on five disproportionality statistical methods were evaluated against the reference set. The performance metrics were sensitivity, precision, specificity, accuracy, and F score. For bovine and porcine products, the Observed-to-Expected (O/E) SDA was the closest in terms of geometric distance to 100% sensitivity and 100% precision. For canine and feline products, the Information Component (IC) SDA was geometrically closest to 100% sensitivity and 100% precision. Principal Component Analysis confirmed that the O/E and IC SDAs were unique performers with respect to one another and other SDAs. The performance of the SDAs was dependent on the choice of the statistical method with differences seen between animal species.

Cell type discovery using single-cell transcriptomics: implications for ontological representation.

Cells are fundamental function units of multicellular organisms, with different cell types playing distinct physiological roles in the body. The recent advent of single-cell transcriptional profiling using RNA sequencing is producing 'big data', enabling the identification of novel human cell types at an unprecedented rate. In this review, we summarize recent work characterizing cell types in the human central nervous and immune systems using single-cell and single-nuclei RNA sequencing, and discuss the implications that these discoveries are having on the representation of cell types in the reference Cell Ontology (CL). We propose a method, based on random forest machine learning, for identifying sets of necessary and sufficient marker genes, which can be used to assemble consistent and reproducible cell type definitions for incorporation into the CL. The representation of defined cell type classes and their relationships in the CL using this strategy will make the cell type classes being identified by high-throughput/high-content technologies findable, accessible, interoperable and reusable (FAIR), allowing the CL to serve as a reference knowledgebase of information about the role that distinct cellular phenotypes play in human health and disease.

Simultaneous deletion of the methylcytosine oxidases Tet1 and Tet3 increases transcriptome variability in early embryogenesis.

Dioxygenases of the TET (Ten-Eleven Translocation) family produce oxidized methylcytosines, intermediates in DNA demethylation, as well as new epigenetic marks. Here we show data suggesting that TET proteins maintain the consistency of gene transcription. Embryos lacking Tet1 and Tet3 (Tet1/3 DKO) displayed a strong loss of 5-hydroxymethylcytosine (5hmC) and a concurrent increase in 5-methylcytosine (5mC) at the eight-cell stage. Single cells from eight-cell embryos and individual embryonic day 3.5 blastocysts showed unexpectedly variable gene expression compared with controls, and this variability correlated in blastocysts with variably increased 5mC/5hmC in gene bodies and repetitive elements. Despite the variability, genes encoding regulators of cholesterol biosynthesis were reproducibly down-regulated in Tet1/3 DKO blastocysts, resulting in a characteristic phenotype of holoprosencephaly in the few embryos that survived to later stages. Thus, TET enzymes and DNA cytosine modifications could directly or indirectly modulate transcriptional noise, resulting in the selective susceptibility of certain intracellular pathways to regulation by TET proteins.

Cell type discovery and representation in the era of high-content single cell phenotyping.

BACKGROUND: A fundamental characteristic of multicellular organisms is the specialization of functional cell types through the process of differentiation. These specialized cell types not only characterize the normal functioning of different organs and tissues, they can also be used as cellular biomarkers of a variety of different disease states and therapeutic/vaccine responses. In order to serve as a reference for cell type representation, the Cell Ontology has been developed to provide a standard nomenclature of defined cell types for comparative analysis and biomarker discovery. Historically, these cell types have been defined based on unique cellular shapes and structures, anatomic locations, and marker protein expression. However, we are now experiencing a revolution in cellular characterization resulting from the application of new high-throughput, high-content cytometry and sequencing technologies. The resulting explosion in the number of distinct cell types being identified is challenging the current paradigm for cell type definition in the Cell Ontology. RESULTS: In this paper, we provide examples of state-of-the-art cellular biomarker characterization using high-content cytometry and single cell RNA sequencing, and present strategies for standardized cell type representations based on the data outputs from these cutting-edge technologies, including "context annotations" in the form of standardized experiment metadata about the specimen source analyzed and marker genes that serve as the most useful features in machine learning-based cell type classification models. We also propose a statistical strategy for comparing new experiment data to these standardized cell type representations. CONCLUSION: The advent of high-throughput/high-content single cell technologies is leading to an explosion in the number of distinct cell types being identified. It will be critical for the bioinformatics community to develop and adopt data standard conventions that will be compatible with these new technologies and support the data representation needs of the research community. The proposals enumerated here will serve as a useful starting point to address these challenges.

The metabolic potential of the single cell genomes obtained from the Challenger Deep, Mariana Trench within the candidate superphylum Parcubacteria (OD1).

Candidate phyla (CP) are broad phylogenetic clusters of organisms that lack cultured representatives. Included in this fraction is the candidate Parcubacteria superphylum. Specific characteristics that have been ascribed to the Parcubacteria include reduced genome size, limited metabolic potential and exclusive reliance on fermentation for energy acquisition. The study of new environmental niches, such as the marine versus terrestrial subsurface, often expands the understanding of the genetic potential of taxonomic groups. For this reason, we analyzed 12 Parcubacteria single amplified genomes (SAGs) from sediment samples collected within the Challenger Deep of the Mariana Trench, obtained during the Deepsea Challenge (DSC) Expedition. Many of these SAGs are closely related to environmental sequences obtained from deep-sea environments based on 16S rRNA gene similarity and BLAST matches to predicted proteins. DSC SAGs encode features not previously identified in Parcubacteria obtained from other habitats. These include adaptation to oxidative stress, polysaccharide modification and genes associated with respiratory nitrate reduction. The DSC SAGs are also distinguished by relative greater abundance of genes for nucleotide and amino acid biosynthesis, repair of alkylated DNA and the synthesis of mechanosensitive ion channels. These results present an expanded view of the Parcubacteria, among members residing in an ultra-deep hadal environment.

Genome of the pathogen Porphyromonas gingivalis recovered from a biofilm in a hospital sink using a high-throughput single-cell genomics platform.

Although biofilms have been shown to be reservoirs of pathogens, our knowledge of the microbial diversity in biofilms within critical areas, such as health care facilities, is limited. Available methods for pathogen identification and strain typing have some inherent restrictions. In particular, culturing will yield only a fraction of the species present, PCR of virulence or marker genes is mainly focused on a handful of known species, and shotgun metagenomics is limited in the ability to detect strain variations. In this study, we present a single-cell genome sequencing approach to address these limitations and demonstrate it by specifically targeting bacterial cells within a complex biofilm from a hospital bathroom sink drain. A newly developed, automated platform was used to generate genomic DNA by the multiple displacement amplification (MDA) technique from hundreds of single cells in parallel. MDA reactions were screened and classified by 16S rRNA gene PCR sequence, which revealed a broad range of bacteria covering 25 different genera representing environmental species, human commensals, and opportunistic human pathogens. Here we focus on the recovery of a nearly complete genome representing a novel strain of the periodontal pathogen Porphyromonas gingivalis (P. gingivalis JCVI SC001) using the single-cell assembly tool SPAdes. Single-cell genomics is becoming an accepted method to capture novel genomes, primarily in the marine and soil environments. Here we show for the first time that it also enables comparative genomic analysis of strain variation in a pathogen captured from complex biofilm samples in a healthcare facility.

Nearly finished genomes produced using gel microdroplet culturing reveal substantial intraspecies genomic diversity within the human microbiome.

The majority of microbial genomic diversity remains unexplored. This is largely due to our inability to culture most microorganisms in isolation, which is a prerequisite for traditional genome sequencing. Single-cell sequencing has allowed researchers to circumvent this limitation. DNA is amplified directly from a single cell using the whole-genome amplification technique of multiple displacement amplification (MDA). However, MDA from a single chromosome copy suffers from amplification bias and a large loss of specificity from even very small amounts of DNA contamination, which makes assembling a genome difficult and completely finishing a genome impossible except in extraordinary circumstances. Gel microdrop cultivation allows culturing of a diverse microbial community and provides hundreds to thousands of genetically identical cells as input for an MDA reaction. We demonstrate the utility of this approach by comparing sequencing results of gel microdroplets and single cells following MDA. Bias is reduced in the MDA reaction and genome sequencing, and assembly is greatly improved when using gel microdroplets. We acquired multiple near-complete genomes for two bacterial species from human oral and stool microbiome samples. A significant amount of genome diversity, including single nucleotide polymorphisms and genome recombination, is discovered. Gel microdroplets offer a powerful and high-throughput technology for assembling whole genomes from complex samples and for probing the pan-genome of naturally occurring populations.

Candidate phylum TM6 genome recovered from a hospital sink biofilm provides genomic insights into this uncultivated phylum.

The "dark matter of life" describes microbes and even entire divisions of bacterial phyla that have evaded cultivation and have yet to be sequenced. We present a genome from the globally distributed but elusive candidate phylum TM6 and uncover its metabolic potential. TM6 was detected in a biofilm from a sink drain within a hospital restroom by analyzing cells using a highly automated single-cell genomics platform. We developed an approach for increasing throughput and effectively improving the likelihood of sampling rare events based on forming small random pools of single-flow-sorted cells, amplifying their DNA by multiple displacement amplification and sequencing all cells in the pool, creating a "mini-metagenome." A recently developed single-cell assembler, SPAdes, in combination with contig binning methods, allowed the reconstruction of genomes from these mini-metagenomes. A total of 1.07 Mb was recovered in seven contigs for this member of TM6 (JCVI TM6SC1), estimated to represent 90% of its genome. High nucleotide identity between a total of three TM6 genome drafts generated from pools that were independently captured, amplified, and assembled provided strong confirmation of a correct genomic sequence. TM6 is likely a Gram-negative organism and possibly a symbiont of an unknown host (nonfree living) in part based on its small genome, low-GC content, and lack of biosynthesis pathways for most amino acids and vitamins. Phylogenomic analysis of conserved single-copy genes confirms that TM6SC1 is a deeply branching phylum.

RNA-sequencing from single nuclei.

It has recently been established that synthesis of double-stranded cDNA can be done from a single cell for use in DNA sequencing. Global gene expression can be quantified from the number of reads mapping to each gene, and mutations and mRNA splicing variants determined from the sequence reads. Here we demonstrate that this method of transcriptomic analysis can be done using the extremely low levels of mRNA in a single nucleus, isolated from a mouse neural progenitor cell line and from dissected hippocampal tissue. This method is characterized by excellent coverage and technical reproducibility. On average, more than 16,000 of the 24,057 mouse protein-coding genes were detected from single nuclei, and the amount of gene-expression variation was similar when measured between single nuclei and single cells. Several major advantages of the method exist: first, nuclei, compared with whole cells, have the advantage of being easily isolated from complex tissues and organs, such as those in the CNS. Second, the method can be widely applied to eukaryotic species, including those of different kingdoms. The method also provides insight into regulatory mechanisms specific to the nucleus. Finally, the method enables dissection of regulatory events at the single-cell level; pooling of 10 nuclei or 10 cells obscures some of the variability measured in transcript levels, implying that single nuclei and cells will be extremely useful in revealing the physiological state and interconnectedness of gene regulation in a manner that avoids the masking inherent to conventional transcriptomics using bulk cells or tissues.

Single cells within the Puerto Rico trench suggest hadal adaptation of microbial lineages.

Hadal ecosystems are found at a depth of 6,000 m below sea level and below, occupying less than 1% of the total area of the ocean. The microbial communities and metabolic potential in these ecosystems are largely uncharacterized. Here, we present four single amplified genomes (SAGs) obtained from 8,219 m below the sea surface within the hadal ecosystem of the Puerto Rico Trench (PRT). These SAGs are derived from members of deep-sea clades, including the Thaumarchaeota and SAR11 clade, and two are related to previously isolated piezophilic (high-pressure-adapted) microorganisms. In order to identify genes that might play a role in adaptation to deep-sea environments, comparative analyses were performed with genomes from closely related shallow-water microbes. The archaeal SAG possesses genes associated with mixotrophy, including lipoylation and the glycine cleavage pathway. The SAR11 SAG encodes glycolytic enzymes previously reported to be missing from this abundant and cosmopolitan group. The other SAGs, which are related to piezophilic isolates, possess genes that may supplement energy demands through the oxidation of hydrogen or the reduction of nitrous oxide. We found evidence for potential trench-specific gene distributions, as several SAG genes were observed only in a PRT metagenome and not in shallower deep-sea metagenomes. These results illustrate new ecotype features that might perform important roles in the adaptation of microorganisms to life in hadal environments.

Sequence analysis of a complete 1.66 Mb Prochlorococcus marinus MED4 genome cloned in yeast.

Marine cyanobacteria of the genus Prochlorococcus represent numerically dominant photoautotrophs residing throughout the euphotic zones in the open oceans and are major contributors to the global carbon cycle. Prochlorococcus has remained a genetically intractable bacterium due to slow growth rates and low transformation efficiencies using standard techniques. Our recent successes in cloning and genetically engineering the AT-rich, 1.1 Mb Mycoplasma mycoides genome in yeast encouraged us to explore similar methods with Prochlorococcus. Prochlorococcus MED4 has an AT-rich genome, with a GC content of 30.8%, similar to that of Saccharomyces cerevisiae (38%), and contains abundant yeast replication origin consensus sites (ACS) evenly distributed around its 1.66 Mb genome. Unlike Mycoplasma cells, which use the UGA codon for tryptophane, Prochlorococcus uses the standard genetic code. Despite this, we observed no toxic effects of several partial and 15 whole Prochlorococcus MED4 genome clones in S. cerevisiae. Sequencing of a Prochlorococcus genome purified from yeast identified 14 single base pair missense mutations, one frameshift, one single base substitution to a stop codon and one dinucleotide transversion compared to the donor genomic DNA. We thus provide evidence of transformation, replication and maintenance of this 1.66 Mb intact bacterial genome in S. cerevisiae.

Estimation of the deamidation rates of major deamidation sites in a Fab fragment of mouse IgG1-κ by capillary isoelectric focusing of mutated Fab fragments.

The deamidation of asparagine (Asn or N) residues in proteins is a common post-translational chemical modification. The identification of deamidation sites and determination of the degree of deamidation have been carried out by the combination of peptide mapping and mass spectrometry. However, when a peptide fragment contains multiple amides, such analysis becomes difficult and sometimes impossible. In this report, a quantitative method for estimating the deamidation rate of a specific amide in a protein is presented without using peptide mapping. Five Asn residues of a recombinant fragment antigen binding (rFab) (mouse IgG1, κ) were mutated to a serine (Ser) residue, one by one, through site-directed mutagenesis, and the single-residue deamidation rates of the original rFab and the mutants were determined using capillary isoelectric focusing. The difference of the rate between the original rFab and the mutant was assumed to be equal to the deamidation rate of the specific Asn residue, which had been mutated. Among five mutants established, three major deamidation sites-H chain Asn135, L chain Asn157, and L chain Asn161, using the Kabat numbering system-were identified, accounting for 66%, 29%, and 7% of the single-residue deamidation of the original rFab, respectively. Although the former two have been known by peptide mapping, the last one, which resides on the same tryptic peptide that carries one of the former two, previously has not been identified. For the first time, the deamidation rate constants of the three sites were estimated to be 10.5 × 10(-3) h(-1), 4.6 × 10(-3) h(-1), and 1.1 × 10(-3) h(-1) in 0.1 M phosphate buffer, pH 7.5 at 37 °C, respectively, with corresponding half-life of 2.8 days, 6.3 days, and 27 days. The method should be applicable to any recombinant proteins.

Cochlear transcriptome analysis of an outbred mouse population (CFW).

Age-related hearing loss (ARHL) is the most common cause of hearing loss and one of the most prevalent conditions affecting the elderly worldwide. Despite evidence from our lab and others about its polygenic nature, little is known about the specific genes, cell types and pathways involved in ARHL, impeding the development of therapeutic interventions. In this manuscript, we describe, for the first time, the complete cell-type specific transcriptome of the aging mouse cochlea using snRNA-seq in an outbred mouse model in relation to auditory threshold variation. Cochlear cell types were identified using unsupervised clustering and annotated via a three-tiered approach - first by linking to expression of known marker genes, then using the NS-Forest algorithm to select minimum cluster-specific marker genes and reduce dimensional feature space for statistical comparison of our clusters with existing publicly-available data sets on the gEAR website (https://umgear.org/), and finally, by validating and refining the annotations using Multiplexed Error Robust Fluorescence In Situ Hybridization (MERFISH) and the cluster-specific marker genes as probes. We report on 60 unique cell-types expanding the number of defined cochlear cell types by more than two times. Importantly, we show significant specific cell type increases and decreases associated with loss of hearing acuity implicating specific subsets of hair cell subtypes, ganglion cell subtypes, and cell subtypes withing the stria vascularis in this model of ARHL. These results provide a view into the cellular and molecular mechanisms responsible for age-related hearing loss and pathways for therapeutic targeting.

Cochlear transcriptome analysis of an outbred mouse population (CFW).

Age-related hearing loss (ARHL) is the most common cause of hearing loss and one of the most prevalent conditions affecting the elderly worldwide. Despite evidence from our lab and others about its polygenic nature, little is known about the specific genes, cell types, and pathways involved in ARHL, impeding the development of therapeutic interventions. In this manuscript, we describe, for the first time, the complete cell-type specific transcriptome of the aging mouse cochlea using snRNA-seq in an outbred mouse model in relation to auditory threshold variation. Cochlear cell types were identified using unsupervised clustering and annotated via a three-tiered approach-first by linking to expression of known marker genes, then using the NSForest algorithm to select minimum cluster-specific marker genes and reduce dimensional feature space for statistical comparison of our clusters with existing publicly-available data sets on the gEAR website, and finally, by validating and refining the annotations using Multiplexed Error Robust Fluorescence In Situ Hybridization (MERFISH) and the cluster-specific marker genes as probes. We report on 60 unique cell-types expanding the number of defined cochlear cell types by more than two times. Importantly, we show significant specific cell type increases and decreases associated with loss of hearing acuity implicating specific subsets of hair cell subtypes, ganglion cell subtypes, and cell subtypes within the stria vascularis in this model of ARHL. These results provide a view into the cellular and molecular mechanisms responsible for age-related hearing loss and pathways for therapeutic targeting.

Machine Learning-Based Single Cell and Integrative Analysis Reveals That Baseline mDC Predisposition Correlates With Hepatitis B Vaccine Antibody Response.

Vaccination to prevent infectious disease is one of the most successful public health interventions ever developed. And yet, variability in individual vaccine effectiveness suggests that a better mechanistic understanding of vaccine-induced immune responses could improve vaccine design and efficacy. We have previously shown that protective antibody levels could be elicited in a subset of recipients with only a single dose of the hepatitis B virus (HBV) vaccine and that a wide range of antibody levels were elicited after three doses. The immune mechanisms responsible for this vaccine response variability is unclear. Using single cell RNA sequencing of sorted innate immune cell subsets, we identified two distinct myeloid dendritic cell subsets (NDRG1-expressing mDC2 and CDKN1C-expressing mDC4), the ratio of which at baseline (pre-vaccination) correlated with the immune response to a single dose of HBV vaccine. Our results suggest that the participants in our vaccine study were in one of two different dendritic cell dispositional states at baseline - an NDRG2-mDC2 state in which the vaccine elicited an antibody response after a single immunization or a CDKN1C-mDC4 state in which the vaccine required two or three doses for induction of antibody responses. To explore this correlation further, genes expressed in these mDC subsets were used for feature selection prior to the construction of predictive models using supervised canonical correlation machine learning. The resulting models showed an improved correlation with serum antibody titers in response to full vaccination. Taken together, these results suggest that the propensity of circulating dendritic cells toward either activation or suppression, their "dispositional endotype" at pre-vaccination baseline, could dictate response to vaccination.

SARS-CoV-2 infection of the oral cavity and saliva.

Despite signs of infection-including taste loss, dry mouth and mucosal lesions such as ulcerations, enanthema and macules-the involvement of the oral cavity in coronavirus disease 2019 (COVID-19) is poorly understood. To address this, we generated and analyzed two single-cell RNA sequencing datasets of the human minor salivary glands and gingiva (9 samples, 13,824 cells), identifying 50 cell clusters. Using integrated cell normalization and annotation, we classified 34 unique cell subpopulations between glands and gingiva. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) viral entry factors such as ACE2 and TMPRSS members were broadly enriched in epithelial cells of the glands and oral mucosae. Using orthogonal RNA and protein expression assessments, we confirmed SARS-CoV-2 infection in the glands and mucosae. Saliva from SARS-CoV-2-infected individuals harbored epithelial cells exhibiting ACE2 and TMPRSS expression and sustained SARS-CoV-2 infection. Acellular and cellular salivary fractions from asymptomatic individuals were found to transmit SARS-CoV-2 ex vivo. Matched nasopharyngeal and saliva samples displayed distinct viral shedding dynamics, and salivary viral burden correlated with COVID-19 symptoms, including taste loss. Upon recovery, this asymptomatic cohort exhibited sustained salivary IgG antibodies against SARS-CoV-2. Collectively, these data show that the oral cavity is an important site for SARS-CoV-2 infection and implicate saliva as a potential route of SARS-CoV-2 transmission.