Detection of allele-specific expression in spatial transcriptomics with spASE

Allele-specific expression (ASE), or the preferential expression of one allele, can be observed in transcriptomics data from early development throughout the lifespan. However, the prevalence of spatial and cell type-specific ASE variation remains unclear. Spatial transcriptomics technologies permit the study of spatial ASE patterns genome-wide at near-single-cell resolution. However, the data are highly sparse, and confounding between cell type and spatial location present further statistical challenges. Here, we introduce spASE (https://github.com/lulizou/spase), a computational framework for detecting spatial patterns in ASE within and across cell types from spatial transcriptomics data. To tackle the challenge presented by the low signal to noise ratio due to the sparsity of the data, we implement a spatial smoothing approach that greatly improves statistical power. We generated Slide-seqV2 data from the mouse hippocampus and detected ASE in X-chromosome genes, both within and across cell type, validating our ability to recover known ASE patterns. We demonstrate that our method can also identify cell type-specific effects, which we find can explain the majority of the spatial signal for autosomal genes. The findings facilitated by our method provide new insight into the uncharacterized landscape of spatial and cell type-specific ASE in the mouse hippocampus.

Genome-Wide Analysis of Allele-Specific Expression Genes in Pediatric B-Cell Precursor Acute Lymphoblastic Leukemia

Introduction. Pediatric B-cell precursor acute lymphoblastic leukemia (BCP ALL) is the most common pediatric hematological malignancy and it remains an important cause of morbidity and mortality in children. In this study, we performed an allele-specific expression (ASE) analysis of pediatric BCP ALL with the aim to investigate the relationship between cis-regulatory mutations and gene expression patterns. Materials and methods. Twenty-two high hyperdiploid ALL, twenty ETV6/RUNX1-positive ALL, seven TCF3/PBX1-positive ALL and twenty-eight genetically unclassified BCP ALL ("B-other") were subjected to whole genome sequencing, SNP array analysis and RNA sequencing. The binomial test was applied to estimate the allelic bias of heterozygous exonic single nucleotide variants (SNVs) in the RNA sequencing data against the genomic data. Allelic ratios >2 or <0.5, and P values <0.05 were used to identify allele-specific expression protein-coding genes. Results. We identified 12,693 expressed genes, of which 9,672 (76%) had heterozygous exonic SNVs (informative genes), in multiple BCP ALL samples (n>2) in 77 of the investigated samples. Genes with ASE were distributed evenly across the autosomal chromosomes in the different subtypes with a range of 30 - 165 ASE genes per case (median number, 86). We found that 630 (6.5%) genes displayed ASE in multiple BCP ALL samples (n>2), of which only eight autosomal genes had monoallelic expression in more than two investigated samples. This suggests that ASE and monoallelic expression are relatively rare in BCP ALL. Gene enrichment analyses showed that pathways involving negative regulation of natural killer cell-mediated cytotoxicity and cell proliferation were enriched, indicating that ASE events possibly were associated with the cell proliferation and leukemia progression in BCP ALL. Furthermore, the hematopoiesis pathway was also enriched in ASE genes that showed high allelic expression bias (allelic ratios >2.5), suggesting that ASE genes might be associated with leukemia development. Somatic genomic aberrations that could cause ASE were also investigated in this study. All informative cases with TCF3/PBX1 rearrangement (n=4) showed monoallelic expression of the PBX1 gene, likely associated with the PBX1 truncation caused by the fusion. Additionally, CHP1, located in 15q15.1, displayed ASE in one case with an inversion involving that chromosome band, indicating a potential cis-acting element in the inversion region that regulated the CHP1 gene expression. Notably, PAX5 displayed various patterns of ASE in BCP ALL. One of three cases with PAX5/ZCCHC7 gene rearrangements displayed PAX5 ASE while the other two did not, indicating a potential uncovered cis-regulatory element around the PAX5/ZCCHC7 breakpoints. Furthermore, two cases with no PAX5 gene rearrangement displayed monoallelic expression of the PAX5 gene, suggesting that there are additional epigenetic alterations were also involved in the regulation of PAX5 gene expression in BCP ALL. Conclusions. In this study, we have characterized genes displaying ASE in childhood BCP ALL. Our data provide new insight into pathogenesis of BCP ALL and may be used to identify novel targets for treatment. Disclosures No relevant conflicts of interest to declare.

Transposable elements that have recently been mobile in the human genome

Background Transposable elements (TE) comprise nearly half of the human genome and their insertions have profound effects to human genetic diversification and as well as disease. Despite their abovementioned significance, there is no consensus on the TE subfamilies that remain active in the human genome. In this study, we therefore developed a novel statistical test for recently mobile subfamilies (RMSs), based on patterns of overlap with > 100,000 polymorphic indels. Results Our analysis produced a catalogue of 20 high-confidence RMSs, which excludes many false positives in public databases. Intriguingly though, it includes HERV-K, an LTR subfamily previously thought to be extinct. The RMS catalogue is strongly enriched for contributions to germline genetic disorders (P = 1.1e-10), and thus constitutes a valuable resource for diagnosing disorders of unknown aetiology using targeted TE-insertion screens. Remarkably, RMSs are also highly enriched for somatic insertions in diverse cancers (P = 2.8e-17), thus indicating strong correlations between germline and somatic TE mobility. Using CRISPR/Cas9 deletion, we show that an RMS-derived polymorphic TE insertion increased the expression of RPL17, a gene associated with lower survival in liver cancer. More broadly, polymorphic TE insertions from RMSs were enriched near genes with allele-specific expression, suggesting widespread effects on gene regulation. Conclusions By using a novel statistical test we have defined a catalogue of 20 recently mobile transposable element subfamilies. We illustrate the gene regulatory potential of RMS-derived polymorphic TE insertions, using CRISPR/Cas9 deletion in vitro on a specific candidate, as well as by genome wide analysis of allele-specific expression. Our study presents novel insights into TE mobility and regulatory potential and provides a key resource for human disease genetics and population history studies.

Cis-regulatory changes in locomotor genes are associated with the evolution of burrowing behavior

SummaryHow evolution modifies complex, innate behaviors is largely unknown. Divergence in many morphological traits has been linked, at least in part, to cis-regulatory changes in gene expression, a pattern also observed in some behaviors of recently diverged populations. Given this, we compared the gene expression in the brains of two interfertile sister species of Peromyscus mice, including allele-specific expression (ASE) of their F1 hybrids, that show large and heritable differences in burrowing behavior. Because cis-regulation may contribute to constitutive as well as activity-dependent gene expression, we also captured a molecular signature of burrowing circuit divergence by quantifying gene expression in mice shortly after burrowing. We found that several thousand genes were differentially expressed between the two sister species regardless of behavioral context, with several thousand more showing behavior-dependent differences. Allele-specific expression in F1 hybrids showed a similar pattern, suggesting that much of the differential expression is driven by cis-regulatory divergence. Genes related to locomotor coordination showed the strongest signals of lineage-specific selection on burrowing-induced cis-regulatory changes. By comparing these candidate genes to independent quantitative trait locus (QTL) mapping data, we found that the closest QTL markers to these candidate genes are associated with variation in burrow shape, demonstrating an enrichment for candidate locomotor genes near segregating causal loci. Together, our results provide insight into how cis-regulated gene expression can depend on behavioral context as well as how this dynamic regulatory divergence between species can be integrated with forward genetics to enrich our understanding of the genetic basis of behavioral evolution.

Genome and Transcriptome Sequencing of Populus × sibirica Identified Sex-Associated Allele-Specific Expression of the CLC Gene

Transcriptome sequencing of leaves, catkin axes, and flowers from male and female trees of Populus × sibirica and genome sequencing of the same plants were performed for the first time. The availability of both genome and transcriptome sequencing data enabled the identification of allele-specific expression. Such an analysis was performed for genes from the sex-determining region (SDR). P. × sibirica is an intersectional hybrid between species from sections Aigeiros (Populus nigra) and Tacamahaca (Populus laurifolia, Populus suaveolens, or Populus × moskoviensis); therefore, a significant number of heterozygous polymorphisms were identified in the SDR that allowed us to distinguish between alleles. In the SDR, both allelic variants of the TCP (T-complex protein 1 subunit gamma), CLC (Chloride channel protein CLC-c), and MET1 (DNA-methyltransferase 1) genes were expressed in females, while in males, two allelic variants were expressed for TCP and MET1 but only one allelic variant prevailed for CLC. Targeted sequencing of TCP, CLC, and MET1 regions on a representative set of trees confirmed the sex-associated allele-specific expression of the CLC gene in generative and vegetative tissues of P. × sibirica. Our study brings new knowledge on sex-associated differences in Populus species.

Allele-specific expression links diverse cis-regulatory mutations to recurrent gene dysregulation in high-risk neuroblastoma

Neuroblastoma is a pediatric malignancy with a high frequency of metastatic disease at initial diagnosis. Neuroblastoma tumors have few protein-coding mutations but contain extensive somatic copy number alterations (SCNAs) suggesting that mutations that alter gene dosage are important drivers of tumorigenesis. Here we analyze allele-specific expression (ASE) in 96 high-risk neuroblastoma tumors to discover genes with cis-acting mutations that alter dosage. We identify 1,049 genes with recurrent, neuroblastoma-specific ASE, 68% of which lie within common SCNA regions. However, many genes exhibit ASE in copy neutral samples and are enriched for mutations that cause nonsense-mediated decay, indicating that neuroblastoma tumors select for multiple types of mutations that alter gene expression. We also find 24 genes with reduced expression in stage 4 disease that have neuroblastoma-specific ASE that is independent of SCNAs. At least two of these genes have evidence for tumor suppressor activity including the transcription factor TFAP2B and the protein tyrosine phosphatase PTPRH. In summary, our ASE analysis discovers genes that are recurrently dysregulated by both large SCNAs and other cis-acting mutations in high-risk neuroblastoma.

RNA-Seq Data for Reliable SNP Detection and Genotype Calling: Interest for Coding Variant Characterization and Cis-Regulation Analysis by Allele-Specific Expression in Livestock Species

In addition to their common usages to study gene expression, RNA-seq data accumulated over the last 10 years are a yet-unexploited resource of SNPs in numerous individuals from different populations. SNP detection by RNA-seq is particularly interesting for livestock species since whole genome sequencing is expensive and exome sequencing tools are unavailable. These SNPs detected in expressed regions can be used to characterize variants affecting protein functions, and to study cis-regulated genes by analyzing allele-specific expression (ASE) in the tissue of interest. However, gene expression can be highly variable, and filters for SNP detection using the popular GATK toolkit are not yet standardized, making SNP detection and genotype calling by RNA-seq a challenging endeavor. We compared SNP calling results using GATK suggested filters, on two chicken populations for which both RNA-seq and DNA-seq data were available for the same samples of the same tissue. We showed, in expressed regions, a RNA-seq precision of 91% (SNPs detected by RNA-seq and shared by DNA-seq) and we characterized the remaining 9% of SNPs. We then studied the genotype (GT) obtained by RNA-seq and the impact of two factors (GT call-rate and read number per GT) on the concordance of GT with DNA-seq; we proposed thresholds for them leading to a 95% concordance. Applying these thresholds to 767 multi-tissue RNA-seq of 382 birds of 11 chicken populations, we found 9.5 M SNPs in total, of which ∼550,000 SNPs per tissue and population with a reliable GT (call rate ≥ 50%) and among them, ∼340,000 with a MAF ≥ 10%. We showed that such RNA-seq data from one tissue can be used to (i) detect SNPs with a strong predicted impact on proteins, despite their scarcity in each population (16,307 SIFT deleterious missenses and 590 stop-gained), (ii) study, on a large scale, cis-regulations of gene expression, with ∼81% of protein-coding and 68% of long non-coding genes (TPM ≥ 1) that can be analyzed for ASE, and with ∼29% of them that were cis-regulated, and (iii) analyze population genetic using such SNPs located in expressed regions. This work shows that RNA-seq data can be used with good confidence to detect SNPs and associated GT within various populations and used them for different analyses as GTEx studies.

Feasibility of predicting allele specific expression from DNA sequencing using machine learning

Allele specific expression (ASE) concerns divergent expression quantity of alternative alleles and is measured by RNA sequencing. Multiple studies show that ASE plays a role in hereditary diseases by modulating penetrance or phenotype severity. However, genome diagnostics is based on DNA sequencing and therefore neglects gene expression regulation such as ASE. To take advantage of ASE in absence of RNA sequencing, it must be predicted using only DNA variation. We have constructed ASE models from BIOS (n = 3432) and GTEx (n = 369) that predict ASE using DNA features. These models are highly reproducible and comprise many different feature types, highlighting the complex regulation that underlies ASE. We applied the BIOS-trained model to population variants in three genes in which ASE plays a clinically relevant role: BRCA2, RET and NF1. This resulted in predicted ASE effects for 27 variants, of which 10 were known pathogenic variants. We demonstrated that ASE can be predicted from DNA features using machine learning. Future efforts may improve sensitivity and translate these models into a new type of genome diagnostic tool that prioritizes candidate pathogenic variants or regulators thereof for follow-up validation by RNA sequencing. All used code and machine learning models are available at GitHub and Zenodo.