Sage Journals: Discover world-class research

Abstract

Background

Migraine is a debilitating disorder characterized by severe headaches and associated neurological symptoms. A key challenge to understanding migraine has been the cellular complexity of the human brain and the multiple cell types implicated in its pathophysiology. The present study leverages recent advances in single-cell transcriptomics to localize the specific human brain cell types in which putative migraine susceptibility genes are expressed.

Methods

The cell-type specific expression of both familial and common migraine-associated genes was determined bioinformatically using data from 2,039 individual human brain cells across two published single-cell RNA sequencing datasets. Enrichment of migraine-associated genes was determined for each brain cell type.

Results

Analysis of single-brain cell RNA sequencing data from five major subtypes of cells in the human cortex (neurons, oligodendrocytes, astrocytes, microglia, and endothelial cells) indicates that over 40% of known migraine-associated genes are enriched in the expression profiles of a specific brain cell type. Further analysis of neuronal migraine-associated genes demonstrated that approximately 70% were significantly enriched in inhibitory neurons and 30% in excitatory neurons.

Conclusions

This study takes the next step in understanding the human brain cell types in which putative migraine susceptibility genes are expressed. Both familial and common migraine may arise from dysfunction of discrete cell types within the neurovascular unit, and localization of the affected cell type(s) in an individual patient may provide insight into to their susceptibility to migraine.

Keywords

Migraine genetics single-cell RNA sequencing

Introduction

Migraine is a disorder characterized by severe headaches and associated neurological symptoms. Despite twin studies that implicate a genetic link to migraine susceptibility (1 –5), our primary understanding of migraine pathophysiology has historically derived from exceedingly rare, autosomal dominant mutations (1 –6). The mechanistic diversity of these mutations illustrates the complexity of migraine, as mutations in the neuronal ion channels, CACNA1A and SCN1A, a glial Na-K ATPase, ATP1A2, and a neurovascular gene, NOTCH3, all converge on an overlapping clinical picture (7). While these monogenetic syndromes could be modeled in animals and resulted in new insights into migraine pathophysiology, they appear to be responsible for a very small subset of migraine patients, as large-scale genotyping has failed to identify these genes even at high false positive rates (7,8). In a search for common genetic contributions to migraine, genome-wide association studies (GWAS) have recently identified 38 new genomic variants that are associated with migraine (8). These new human genetics data provide an opportunity to improve our limited understanding of migraine pathophysiology. However, unlike the highly penetrant, autosomal dominant mutations, it remains unclear which of these common variants play causal roles in migraine and warrant additional investigation. Moreover, many of the migraine susceptibly loci are in noncoding genomic regions that do not lead to mutations in the gene product itself. Rather, these noncoding loci are thought to act by disrupting enhancer elements that regulate the expression of downstream genes. Gene enhancers are difficult to identify because they are highly species- and cell type-specific, determined epigenetically, and may regulate genes long distances away. Localizing migraine candidate genes to the specific brain cell types in which they are expressed is important for better understanding their function and role in migraine pathophysiology. To that end, available gene expression databanks were used to identify the tissues (8) or brain regions (9) that are most enriched for the putative migraine susceptibility genes identified by GWAS, but this approach is unable to discriminate between the numerous cell types within these heterogeneous tissues. Over the past few years, the development of single-cell RNA sequencing (scRNAseq), a new technology that describes the entire transcriptome of individual cells within heterogeneous tissues, has enabled us to answer this question directly (10). Two recent studies have used scRNAseq to describe the gene expression profiles of thousands of human brain cells (11,12). The present study uses single-human brain cell transcriptomics to determine the specific brain cell types in which each putative migraine susceptibility gene is expressed.

Methods

The expression of known migraine genes (13 –15) and genes near migraine susceptibility loci (8) were assayed in two published independent single-brain cell RNA sequencing datasets (11,12). Tables of normalized gene counts for each brain cell were graciously provided by the authors of each study. The nearest gene was assigned to each of the index single nucleotide polymorphisms (SNPs) reported in Gormley et al. For SNPs in long intergenic regions between two genes, both genes were analyzed. The normalized RNA levels were extracted from 229 individual adult human brain cells (11) or 1,810 neurons (12). In the original studies, cell types were determined by principal component analysis, graph clustering of each cell’s transcriptome, and confirmation by in situ histochemistry. This analytical method groups cells of similar gene expression patterns into clusters that are assigned to specific brain cell types based on their expression of known marker genes. Data from Darmanis et al. (11) (male and female, 21–68 years of age) were used to determine gene expression levels in major brain cell types (neurons, astrocytes, oligodendrocytes, microglia, and endothelial cells). Data from Lake et al. (12) (female, 51 years old) were used to determine migraine-related gene expression levels across 905 randomly selected excitatory neurons and their entire set of 905 inhibitory neurons. To determine whether a gene was significantly enriched in a specific cell type in the Darmanis et al. data, the normalized expression level in each cell type was compared to the expression level in a control group of cells generated by randomly assigning equal numbers of cells from each cell type. This approach ensured that the control cells had equal contributions from each cell type despite the skewed sampling of cell types in the Darmanis et al. dataset (62 astrocytes, 20 endothelial cells, 16 microglia, 131 neurons, and 38 oligodendrocytes). The control cell type was generated by randomly sampling cells 1000 times to ensure representative sampling. Enrichment statistics were calculated by one-tailed t-test comparing a gene’s expression in a specific cell type to that of a randomly generated control cell type. The p-value was adjusted (q-value) for multiple comparisons (for each gene and pairwise cell type comparison made) by controlling the false discovery rate (FDR) to q < 0.1 using the Benjamini-Hochberg procedure. There were no genes that reached significance for more than one cell type. For the Lake et al. data, enrichment was determined by directly comparing equal numbers of excitatory and inhibitory neurons. Significance of each pairwise enrichment comparison was assessed using a one-tailed t-test and then adjusted for multiple comparisons by the Benjamini–Hochberg procedure. Statistics were calculated using R version 3.3.2.

Results

Two recent studies have published scRNAseq data from human postmortem brain tissue (11,12), enabling the identification of the brain cell types in which each putative migraine susceptibility gene is expressed (Figure 1(a)). Darmanis et al. (11) generated gene expression data on five major subtypes of cells in the human cortex: neurons, oligodendrocytes, astrocytes, microglia, and endothelial cells. The expression profile of each migraine-associated gene was then extracted from individually sequenced cells and analyzed according to their cell type. As a proof of principle, the genes that cause familial hemiplegic migraine (FHM) and cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) were found to be expressed in their expected cell types (16 –19). Specifically, CACNA1A and SCN1A were primarily expressed in neurons, ATP1A2 was expressed in astrocytes, and NOTCH3 was expressed in endothelial cells (Figure 1(b)). After establishing the ability to study cell type-specific gene expression for known migraine associated genes, data from the recently published migraine GWAS was examined to describe the specific brain cell types in which each of these genes are expressed (Figure 2(a)). Genes that were DEPICT-prioritized (www.broadinstitute.org/depict) by Gormley et al. indicate increased confidence in the association between index SNP and affected gene and are bolded in the figure. Cell type-specific expression trends can be inferred from each gene’s mean expression in the heatmap (Figure 2(a)). While approximately 50% of genes show preference for a specific cell type, only 11% of putative migraine susceptibility genes were significantly enriched in a single cell type at an FDR < 0.1. Because the expression of these migraine susceptibility genes are predominantly expressed in a single cell type, they may contribute to migraine by altering the physiology of their respective brain cell type. Determination of statistically significant enrichment was best powered for neurons and astrocytes because of the relatively small number of endothelial cells, microglia, and oligodendrocytes sampled. Limited power for the lowly sampled cell types likely explains why strong enrichment trends (e.g. NOTCH4) did not meet statistical significance. Additional single cell sequencing data from the human brain are needed to confirm the enrichment trends of less abundant cell types observed in Figure 2(a). Approximately 50% of the migraine-associated genes were expressed broadly across multiple cell types, and may contribute to migraine pathophysiology in more complex ways.

Figure 1.

(a) Single-cell RNA sequencing pipeline. Human brain tissue is dissociated into a single-cell suspension. The RNA from each cell is barcoded, sequenced, and quantified using next-generation sequencing technology. Principal component analysis and graph clustering is then used to cluster cells into groups with similar gene expression profiles. This permits cell type identification bioinformatically by determining which cell clusters express known marker genes. (b) Single-cell RNA sequencing data from human cortex (11) reveal the expression level of known familial hemiplegic migraine genes (CACNA1A, ATP1A2, SCN1A) and NOTCH3 (mutated in cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy [CADASIL]) within specific cell types. Each circle represents an individual cell. TPM = transcripts per million sequencing reads.

Figure 2.

Cell-type specific expression of migraine-associated genes. (a) The heatmap displays normalized expression values (11) for migraine susceptibility genes averaged across the cells within each cell type (62 astrocytes, 20 endothelial cells, 16 microglia, 131 neurons, and 38 oligodendrocytes). Gene names of DEPICT-prioritized genes from Gormley et al. are bolded. (b) Expression enrichment of migraine susceptibility genes were determined by comparing their expression in each cell type to a randomly generated cell type (see methods). Enrichment q values were determined by t-test after correcting for multiple comparisons using the Benjamini–Hochberg procedure.

Imbalance in neuronal excitability is a key mechanism underlying migraine pathophysiology (20,21), but it is unclear whether this stems primarily from dysfunction of excitatory or inhibitory neurotransmission. As a first step towards answering this question, human cortex scRNAseq data from 905 excitatory neurons and 905 inhibitory neurons were used to determine whether neuronal migraine associated genes are preferentially expressed in excitatory or inhibitory neurons (12) (Figure 3(a)). Approximately 70% of the neuronal migraine-associated genes were significantly enriched in inhibitory neurons (e.g. PLCE1, SLC24A3, NGF), while 30% were enriched in excitatory neurons (e.g. PRRT2, PHACTR1) (Figure 3(b),(c)). The expression of certain significantly enriched genes (e.g. SCL24A3, ARMS2) was essentially restricted to a single neuronal subtype (e.g. inhibitory neurons), which raises the possibility that genomic variation in these genes could alter the function of the cell type in which they are expressed. Many significantly enriched genes (e.g. CACNA1A, SCN1A), however, are expressed in both excitatory and inhibitory neurons at different levels, and thus require additional mechanistic studies to clarify the relative significance of these genes to each cell type.

Figure 3.

Neuronal subtype specific gene expression of migraine-associated genes. (a) Heatmap displays the log2 ratio of a gene’s expression in excitatory or inhibitory neurons vs. its expression across all neurons(12). (b) Expression enrichment of neuronal migraine susceptibility genes were determined by comparing their expression in excitatory (red) vs. inhibitory (blue) neurons. Enrichment p values were determined by t-test and corrected for multiple comparisons using the Benjamini–Hochberg procedure. (c) The normalized expression level of six neuron-specific genes (SLC24A3, SCN1A, PLCE1, PRRT2, CACNA1A, PHACTR1) in excitatory and inhibitory neurons. Each black circle represents the expression level within an individual cell. The colored bar represents the mean expression of all cells within the respective neuronal subtype. TPM: transcripts per million sequencing reads.

Certain genomic variants have been linked specifically to migraine without aura (e.g. rs11172113 in LRP1, rs1024905 near FGF6) or with aura (rs1835740 near MTDH) (7,8,22,23). This raised the possibility that susceptibility genes for migraine with or without aura might localize to specific brain cell types. However, the associated genes for migraine subtypes demonstrated similar localization patterns to the broader migraine population and were observed in neurons (TRPM8, NGF), glia (LRP1), endothelial cells (FHL5), or combinations thereof (TSPAN2, HJURP, PHACTR1, FGF6, ASTN2, MTDH). This finding was perhaps expected because the majority of migraine patients genotyped in Gormley et al. had migraine without aura. While the small number of significant susceptibility loci for migraine subtypes limits the strength of such conclusions, especially for migraine with aura in which the single associated gene is only nominally significant, the limited data to date do not provide evidence for unique brain cell types in the pathophysiology of migraine subtypes.

Conclusions

The diversity of cell types in the human brain makes it difficult to link human migraine genetics with their pathophysiological consequences. The data presented here take the next step in understanding the brain cell types in which putative migraine susceptibility genes are expressed. This approach lays the foundation for localizing the affected brain cell type(s) within an individual patient based on their genetic susceptibility.

The 38 migraine susceptibly loci were initially reported to be significantly enriched in vascular and smooth muscle tissues (8). This finding, derived from gross tissue gene expression data, led the authors to hypothesize that genetic variability in vascular tissues may be a key mechanism in migraine susceptibility. While the improved cellular resolution of human brain scRNAseq described here also suggests several susceptibility loci are localized to neurovascular cells (e.g. NOTCH4, FHL5), other loci were selectively expressed in neurons, glia or more broadly expressed across multiple cell types. For the genes that are selectively expressed in neurons, inhibitory neurons are more commonly the site of expression. The susceptibility genes linked to migraine with or without aura do not appear to localize to unique cell types, but additional genetic data are needed to draw stronger conclusions.

Together, these findings support a model in which genetic variability in vascular, neuronal, and glial cell types are associated with migraine susceptibility. The diversity of brain cell types implicated in migraine susceptibility and the close functional relationship between each of these cell types within the neurovascular unit may point to a unifying physiological mechanism that could be triggered by dysfunction of neuronal, glial, or vascular cells. Indeed, there is evidence that cortical spreading depression can be activated by dysfunction in each of these cell types (20). It follows that optimal preventative migraine treatment could be directed to the specific brain cell type(s) that is/are disrupting the neurovascular unit in an individual patient. For example, patients with mutations in vascular cells may be particularly responsive to calcium channel blockers whereas a patient with mutations in excitatory neurons may be more responsive to an anticonvulsant. It is important to note, however, that migraine is a polygenic disease, so an individual patient could have several polymorphisms associated with their susceptibility to migraine that affect the function of genes in multiple cell types (24). Combining ongoing efforts to identify genomic variants that are associated with migraine treatment responsiveness (25) and scRNAseq will provide additional mechanistic and predictive information about the pathophysiological classes of migraine patients (e.g. neuronal, glial, vascular, or combinations thereof) that are most likely to respond to specific treatments.

The present analyses should be interpreted in the context of their limitations. First, it is challenging to assign significant GWAS loci to specific candidate genes without detailed functional studies at each locus. There are many tools to aid in this process (e.g. DEPICT, see Figure 2(a)), but they remain estimates based primarily on genome annotation and tissue enrichment. Even for the simplest case of an exonic SNP that causes a missense mutation, the same SNP may also disrupt a previously unknown gene regulatory element (e.g. distal enhancer) of a distant gene. Indeed, there have been several reports of pathogenic SNPs lying within gene regulatory regions (26 –28), one of which suggests that the causal gene for the SNP in PHACTR1 may be the vascular gene, Endothelin 1 (27), rather than the PHACTR1 gene itself. It remains difficult to identify these regulatory elements, however, because they are cell type-specific and poorly annotated in the human brain. While the present study assigns each significant migraine-associated locus to the nearest gene(s), additional research involving cell type-specific gene expression and epigenetic enhancer mapping is needed to confirm these assignments in the human brain. Second, if a gene is expressed in multiple cell types, even at statistically different levels, it is difficult to state precisely where the gene is functioning based on scRNAseq data alone. Additional studies are required to identify whether a gene functions in a cell type-specific fashion with respect to migraine pathophysiology. Hence, caution is needed when interpreting cell type-specific expression of migraine susceptibility genes unless the expression is truly restricted to a specific cell type. Future studies will also benefit from sequencing larger numbers of brain cells to detect less abundant cell types (e.g. smooth muscle cells, pericytes) that were not well sampled in these initial datasets. The data analyzed in this study are best powered to detect the presence of neuronal or astrocytic enrichment, so it is important to note that the absence of statistically significant enrichment in a lowly sampled cell type (e.g. endothelial cells) does not preclude the possibility that additional scRNAseq data could reveal it. More comprehensive human scRNAseq data will be particularly important given the observation by Gormley et al. that there is a statistical enrichment of migraine-associated loci in vascular tissues. Third, it is important to recognize that scRNAseq data were generated from the human cortex. While the cortex is important in migraine susceptibility (21), other brain regions involved in migraine pathophysiology (e.g. the trigeminovascular system and hypothalamus) will also need to be analyzed as these datasets become available in the near future. Despite these limitations, the analyses described here take an important next step in migraine genetics and will continue to improve as new scRNAseq data and epigenetic maps of cell type-specific gene enhancers are generated. The integration of clinical genetics, epigenetics, and single-cell transcriptomics will help inform where and how migraine susceptibility genes predispose an individual to migraine, and ultimately may lead to more personalized, genetically-informed therapies.

Article highlights

Putative migraine susceptibility genes are enriched in multiple components of the neurovascular unit: neurons, glia, and vascular cells.

The majority of neuronal migraine associated genes are enriched in inhibitory neurons.

Localization of an individual’s migraine susceptibility gene(s) to specific brain cell types provides pathophysiological insight into their susceptibility to migraine.

Footnotes

Declaration of conflicting interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

References

Chen

Keith

Nyholt

et al.

Bayesian latent trait modeling of migraine symptom data. Hum Genet 2009; 126: 277–288.

Chen

Mengersen

Keith

et al.

Linkage and heritability analysis of migraine symptom groupings: A comparison of three different clustering methods on twin data. Hum Genet 2009; 125: 591–604.

Mulder

Van Baal

Gaist

et al.

Genetic and environmental influences on migraine: A twin study across six countries. Twin Res 2003; 6: 422–431.

Nyholt

Morley

Ferreira

et al.

Genomewide significant linkage to migrainous headache on chromosome 5q21. Am J Hum Genet 2005; 77: 500–512.

Ziegler

Hur

Bouchard

Jr et al.

Migraine in twins raised together and apart. Headache 1998; 38: 417–422.

Polderman

Benyamin

de Leeuw

et al.

Meta-analysis of the heritability of human traits based on fifty years of twin studies. Nat Genet 2015; 47: 702–709.

Ferrari

Klever

Terwindt

et al.

Migraine pathophysiology: Lessons from mouse models and human genetics. Lancet Neurol 2015; 14: 65–80.

Gormley

Anttila

Winsvold

et al.

Meta-analysis of 375,000 individuals identifies 38 susceptibility loci for migraine. Nat Genet 2016; 48: 856–866.

Eising

Huisman

Mahfouz

et al.

Gene co-expression analysis identifies brain regions and cell types involved in migraine pathophysiology: A GWAS-based study using the Allen Human Brain Atlas. Hum Genet 2016; 135: 425–439.

10.

Poulin

Tasic

Hjerling-Leffler

et al.

Disentangling neural cell diversity using single-cell transcriptomics. Nat Neurosci 2016; 19: 1131–1141.

11.

Darmanis

Sloan

Zhang

et al.

A survey of human brain transcriptome diversity at the single cell level. Proc Natl Acad Sci USA 2015; 112: 7285–7290.

12.

Lake

Kaeser

et al.

Neuronal subtypes and diversity revealed by single-nucleus RNA sequencing of the human brain. Science 2016; 352: 1586–1590.

13.

Ophoff

Terwindt

Vergouwe

et al.

Familial hemiplegic migraine and episodic ataxia type-2 are caused by mutations in the Ca2+ channel gene CACNL1A4. Cell 1996; 87: 543–552.

14.

De Fusco

Marconi

Silvestri

et al.

Haploinsufficiency of ATP1A2 encoding the Na+/K+ pump alpha2 subunit associated with familial hemiplegic migraine type 2. Nat Genet 2003; 33: 192–196.

15.

Dichgans

Freilinger

Eckstein

et al.

Mutation in the neuronal voltage-gated sodium channel SCN1A in familial hemiplegic migraine. Lancet 2005; 366: 371–377.

16.

Domenga

Fardoux

Lacombe

et al.

Notch3 is required for arterial identity and maturation of vascular smooth muscle cells. Genes Dev 2004; 18: 2730–2735.

17.

McGrail

Phillips

Sweadner

. Immunofluorescent localization of three Na,K-ATPase isozymes in the rat central nervous system: Both neurons and glia can express more than one Na,K-ATPase. J Neurosci 1991; 11: 381–391.

18.

Escayg

MacDonald

Meisler

et al.

Mutations of SCN1A, encoding a neuronal sodium channel, in two families with GEFS+2. Nat Genet 2000; 24: 343–345.

19.

Lehmann-Horn

Jurkat-Rott

. Voltage-gated ion channels and hereditary disease. Physiol Rev 1999; 79: 1317–1372.

20.

Dreier

Reiffurth

. The stroke-migraine depolarization continuum. Neuron 2015; 86: 902–922.

21.

Pietrobon

Moskowitz

. Pathophysiology of migraine. Annu Rev Physiol 2013; 75: 365–391.

22.

Chasman

Anttila

Buring

et al.

Selectivity in genetic association with sub-classified migraine in women. PLoS Genet 2014; 10: e1004366–e1004366.

23.

Zhao

Eising

de Vries

et al.

International Headache Genetics Consortium. Gene-based pleiotropy across migraine with aura and migraine without aura patient groups. Cephalalgia 2016; 36: 648–657.

24.

Visscher

Wray

Zhang

et al.

10 Years of GWAS Discovery: Biology, function, and translation. Am J Hum Genet 2017; 101: 5–22.

25.

Christensen

Esserlind

Werge

et al.

The influence of genetic constitution on migraine drug responses. Cephalalgia 2016; 36: 624–639.

26.

Claussnitzer

Dankel

Kim

et al.

FTO obesity variant circuitry and adipocyte browning in humans. N Engl J Med 2015; 373: 895–907.

27.

Gupta

Hadaya

Trehan

et al.

A genetic variant associated with five vascular diseases is a distal regulator of endothelin-1 gene expression. Cell 2017; 170: 522–533.

28.

Soldner

Stelzer

Shivalila

et al.

Parkinson-associated risk variant in distal enhancer of alpha-synuclein modulates target gene expression. Nature 2016; 533: 95–99.

Localization of migraine susceptibility genes in human brain by single-cell RNA sequencing

Abstract

Background

Methods

Results

Conclusions

Keywords

Introduction

Methods

Results

Conclusions

Article highlights

Footnotes

Declaration of conflicting interests

Funding

References