Single-Cell RNA Sequencing Reveals Cell-Type-Specific Mechanisms of Neurological Diseases

Single-cell transcriptomic analysis has provided an unprecedented avenue for the identification of neuronal subtypes that are affected in neurological diseases. However, this strategy is largely hindered by the difficulty of collecting fresh samples of adult human brain. Instead, postmortem brain tissue from both normal and diseased human subjects is more accessible. To reduce sample contamination from other cells or RNA degradation often encountered in whole-cell dissociation and isolation, sequencing of RNA from single nuclei has been developed to reflect whole-cell RNA levels in the human brain [1, 2]. By using single-nucleus RNA sequencing (snRNA-seq), recent studies have identified cell-type-specific molecular changes in the neocortex of patients with autism and Alzheimer’s disease [3–5].

Autism spectrum disorders (ASDs) are complex and heterogeneous conditions, with patients showing persistent deficits in social interaction and communication along with stereotyped behavior [6]. Hundreds of genomic loci have been associated with the occurrence of ASDs, including those transmitted from parents and those appearing de novo in the germline [7], either as a single penetrant mutation or rare low-risk variants with cumulative effects. Although multifactorial neurodevelopmental defects are believed to underlie ASD etiologies, the cell-type-specific pathology of ASDs is unclear. Recently, Kriegstein’s group at the University of California, San Francisco, performed snRNA-seq analysis of 41 postmortem samples of prefrontal cortex and anterior cingulate cortex from 15 ASD patients without intellectual disability and 16 normal controls ranging in age from 4 years to 20 years [8]. They used the prevailing droplet-based 10x Genomics Chromium platform for the analysis and generated 104,559 single-nuclei RNA expression profiles: 52,003 from patients and 52,556 from controls. By annotating cell clusters according to the known markers of specific cell types, they identified cell-type-specific changes in gene expression (see Fig. 1 for the workflow), and intriguingly, established the association of these changes with the clinical severity of disease.

Fig. 1.

Schematic of the main technical approach used in snRNA-seq analysis.

The top differentially-expressed genes (DEGs) were down-regulated in superficial layers 2/3 (L2/3) excitatory neurons and interneurons expressing vasoactive intestinal polypeptide, and up-regulated in astrocytes and microglia. Gene Ontology analysis showed that the dysregulated genes were in pathways involved in chemical synaptic transmission, nervous system processes, postsynaptic membrane potential, axon guidance, neuronal migration, synapse assembly, and γ-aminobutyric acid signaling, across all cell types analyzed. The genes important for synaptic function and some transcription factors critical for brain development were among the top dysregulated genes in L2/3 and 4 excitatory neurons, and notably, genes reflecting activation states were enriched in the microglia and astrocytes of ASD samples. Remarkably, DEGs in L2/3 neurons and microglia were significantly correlated with clinical severity, although the most strongly correlated genes were not among the top DEGs. These results indicate that the degree of dysregulation may not be closely correlated with disease symptoms, and instead, disturbance of gene expression networks in specific neuronal cell types represents the molecular pathology of ASD.

This study also compared the molecular changes in ASD and sporadic epilepsy, a comorbidity of ASD, and found minimal overlap of the changes in cell-type-specific gene expression. For example, synaptic transmission, axon guidance, and brain development pathways were among the DEGs detected in ASD, but not in the epilepsy samples. These results suggest that the identified dysregulated pathways are mostly ASD-specific and highlight probable intervention targets and pathways. Because most ASD patients also have intellectual disability, and in particular, various ASD patients exhibit high heterogeneity in clinical symptoms, future studies should include larger cohorts to more precisely associate genetic variants with clinical severity. Considering the male bias in ASD prevalence [9], sex differences should be taken into consideration in future studies of cell-type-specific molecular changes in ASD.

Alzheimer’s disease (AD) is one of the major challenge for elder people worldwide. However, current understanding of the molecular and cellular mechanisms of AD pathogenesis is still in debate; specific markers used for early diagnosis and approaches for diease control are still not available [10]. Using the droplet-based snRNA-seq approach, a recent study performed by Tsai’s group at Massachusetts Institute of Technology revealed cell-type-specific changes in AD while considering the degree of pathology and sex differences [4]. They analyzed 80,660 single-nucleus transcriptomes from 48 post-mortem samples of prefrontal cortex (Brodmann area 10) from AD patients with various degrees of pathology. They demonstrated cell-type-specific gene expression changes in the early stages and common changes across major cell types in the later stages. Furthermore, this study revealed sexual dimorphism in transcriptional changes at the cellular level. The resource provided in this study depicts a dynamic trajectory of AD pathology at the single-cell level and reveals heterogeneous responses among cell types.

The samples from 48 individuals were collected from two longitudinal cohort studies of ageing and dementia, the Religious Order Study and the Rush Memory and Aging Project, both of which have detailed clinical records, post-mortem pathological examination, and multiple-omics bulk-tissue profiling [11]. Based on the severity of beta-amyloid and other AD pathology, the 48 samples were classified into two groups: 24 controls with little or no detectable pathology and 24 age-matched samples with mild to severe AD pathology. The years of education (medians 18 for no pathology and 19.5 for AD pathology), ages (medians 87.1 for no-pathology and 86.7 for AD-pathology), and sexes were balanced. After quality control filtering, a dataset of 17,926 protein-coding genes from 75,060 nuclei with a cut-off value of 200 unique molecular identifiers and at least 200 detected genes in each cell was subjected to further analysis. Based on known cell-type markers, they identified major cell types from 20 transcriptionally-distinct clusters, including excitatory neurons, inhibitory neurons, astrocytes, oligodendrocytes, microglia, oligodendrocyte progenitor cells, endothelial cells, and pericytes. After comparing the gene expression levels between the groups with and without AD pathology, 1,031 DEGs were identified, and interestingly, most DEGs in neurons were down-regulated and most DEGs in oligodendrocytes, astrocytes, and microglia were up-regulated. Notably, some DEGs showed opposite directionality in different cell types, such as APOE, which was down-regulated in astrocytes but up-regulated in microglia. The heterogeneous response to AD pathology is difficult to capture using bulk RNA-seq analysis. Although the majority of DEGs were strongly cell type-specific with perturbations only in a particular neuronal or glial cell type, the top DEGs involved in related processes, such as myelination and axonal growth and regeneration, were found across cell types.

Then, the samples with AD pathology were further subdivided into two groups: early-stage with an amyloid burden but modest neurofibrillary tangles and mild cognitive impairment; and late-stage with a higher amyloid burden, more pronounced neurofibrillary tangles, and severe cognitive impairment. Pairwise comparisons of gene expression in distinct cell types were made between early/late pathology and no-pathology groups. Notably, almost all up-regulated and down-regulated DEGs in the no- and early-pathology subgroups were cell-type-specific either in neurons or a single glial cell type. By contrast, up-regulated DEGs in late- versus early-pathology were global across cell types, whereas down-regulated DEGs were cell-type-specific. These results reveal dynamic disease progression from the dysregulation of gene expression networks in specific cell types at early stages prior to severe pathological features to global changes due to stress responses at late stages. Furthermore, this study also established associations between gene expression patterns in specific cell types and major pathological traits, including ß-amyloid level, neurofibrillary tangle density, plaque number, and global cognition level, and indicated gene sets responding to AD pathology in specific cell types. For example, genes positively associated with AD pathology in microglia were enriched in pathways mediating immune and inflammatory responses and β-amyloid clearance, and genes associated with oligodendrocyte pathology were enriched in pathways for oligodendrocyte differentiation and myelination.

To further relate the cell-type heterogeneity to the pathological features of individual samples, this study identified cell subpopulations that might be associated with the pathological status of AD and identified marker genes for each subpopulation. Among these, a subpopulation of excitatory neurons marked by INGO1, RASGEF1B, and SLC26A3 and a subpopulation of oligodendrocytes marked by CADM2, QDPR, NLGN, and CRYAB were over-represented in AD pathology. The study also identified a subpopulation of microglia that was distinct from the microglial state in a mouse model and did not overlap with aged microglia without AD. Genes marking these AD pathology-associated subpopulations encode proteins involved in protein quality control, cell death, and immune responses. Notably, sex differences in the response to AD pathology were identified in some cell subpopulations. Females exhibited pronounced AD pathology-associated cell subpopulations and males exhibited more no-pathology subpopulations. Furthermore, females had higher expression of the marker genes of AD-pathology subpopulations. They also discerned the sex-specific differential responses by cell types to pathological features. The most extreme differences were found in oligodendrocytes and neurons. In males, increased AD pathology correlated with transcriptional activation in oligodendrocytes, and this did not occur in females. In females, increased pathology was correlated with marked down-regulation of gene expression in both excitatory and inhibitory neurons, whereas males showed a less pronounced response in excitatory neurons and little response in inhibitory neurons. Females also showed a significant association between white-matter lesions and decreased cognition. Although these results are inspiring, a larger sample size and large-scale gene-trait association studies, as well as experimental validation, are needed to thoroughly understand sexual dimorphism in the progression of AD.

Using a similar snRNA-seq approach, a recent study provided a single-cell atlas of entorhinal cortex from control and AD brains (6 individuals in each group with a mean age of 77.6 years) [5]. The dataset from 13,214 nuclei was analyzed, leading to the findings of cell-type-specific expression of risk genes and their contributions to disease susceptibility. Specifically, this work has also integrated gene networks and genome-wide association studies, which uncovered genomic loci for AD risk [12] at the single-cell level, revealing cell-type-specific drivers of disease state transition. Taken together, these analyses have led to novel insights into cell-type-specific changes in AD.

Although numerous studies using bulk RNA-seq analysis have revealed dysfunctions of neurons and/or innate immune responses, they are unable to entangle the heterogeneity of different disease subtypes and distinct responses across cell types. Because many neurological diseases are usually accompanied by multiple systemic comorbidities, scRNA-seq technology will help to disentangle the shared and unique neurological abnormalities among different conditions, which will assist the development of precision medicine. Furthermore, it will also help to determine the dynamic progression of neuronal pathology at the molecular, cellular, and circuit levels, which will deepen the understanding of disease mechanisms and help to develop predictors of disease trajectories. The major limitation at present is the noisy nature of the resulting data, which is primarily due to the technical limitations of working with such small amounts of RNA. This noise makes it difficult to distinguish very similar cell types and identify rare cell types that may contribute to the heterogeneity of disease entities. Another challenge is how to distinguish between neuroprotective responses and neuronal pathology. Future technological improvements will enable a more profound understanding of disease mechanisms at the cell-type level, which will be of help for diagnosis as well as intervention.