Fig 1.
snRNA-seq of mature pollen nuclei from five different Arabidopsis F1 hybrids.
(A) Diagram of snRNA-seq analysis pipeline. Heterozygous Arabidopsis F1 plants were generated by crossing Col-0 to five different accessions. Pollen was collected from these five F1 plants and pooled, before nuclei isolation, barcoding, and cDNA synthesis using the 10x snRNA-seq protocol. After sequencing, nuclei from the five genotypes were demultiplexed, and meiotic recombination events were predicted. Gene expression and recombination patterns were then compared to identify eQTLs. Figure created using BioRender. (B) Scatter plot with marginal kernel density estimates, showing the first two principal components of the expression data for 1,394 pollen nuclei. Nuclei form two unequally sized clusters, separated by the first principal component. (C) Violin plots showing the expression of known marker genes for sperm and vegetative nuclei. Expression of genes MGH3 and PCR11 indicates that the blue cluster corresponds to sperm nuclei, whilst expression of the marker genes VGD1 and VCK indicate that the orange cluster corresponds to vegetative nuclei. (D) Bar plot showing the number of high-quality nuclei in the dataset predicted to originate from each of the five different F1 hybrids. (E) Violin plot showing the genotyping score for high quality nuclei from each of the five different F1 hybrids. The data underlying this figure can be found in dataset 1 at https://doi.org/10.5281/zenodo.14864053.
Fig 2.
Identification of meiotic recombination patterns from snRNA-seq data.
(A) Log10-scale histogram showing the effective number of unique markers identified for each high-quality nucleus in the snRNA-seq dataset. (B) Histogram showing the distribution of genes (top panels) and markers for Col-0 and parent 2 (bottom panels, Col-0 in blue, parent 2 in orange) across chromosomes. Centromere locations are shown as grey bands. (C) Genome-wide plot showing marker positions and predicted meiotic recombination events for an example nucleus. Markers which support the Col-0 genome are shown in blue above the axis, whilst markers supporting the Parent 2 genome are shown in orange, below the axis. Regions predicted by the rHMM to originate from the Col-0 or Parent 2 are shown with blue or orange background shading, respectively. (D) Violin plots showing the predicted number of recombination events per nucleus for the five different F1 hybrids. Mean crossover numbers for the five different F1 hybrids are shown above each violin. The data underlying this figure can be found in dataset 2 at https://doi.org/10.5281/zenodo.14864053.
Fig 3.
eQTL mapping identifies cis- and trans-eQTLs in Arabidopsis pollen nuclei.
(A) Genome-wide map of eQTL positions (x axis) shown relative to the genomic position of the gene whose expression was tested (y axis). Error bars represent the confidence intervals for each eQTL determined using the 1.5 LOD-drop method, and the size of the points is directly proportional to the LOD score of the eQTL. The orange shaded diagonal indicates the region within which eQTLs were considered to be cis-eQTLs. The green shaded region indicates the hotspot of trans-eQTLs referred to as POLLEN SPERM VARIANT 1 (PSV1). (B) Histogram showing the distance between the genomic location of genes with cis-eQTLs affecting their gene expression and the mapped location of the cis-eQTL. (C) Barplots showing that genes with cis-eQTLs are enriched for rearrangements and large insertions or deletions, compared to genes expressed in pollen which do not have cis-eQTLs. The data underlying this figure can be found in dataset 3 at https://doi.org/10.5281/zenodo.14864053.
Fig 4.
Pollen snRNA-seq identifies cis-eQTLs.
(A) eQTL plot showing the haplotypes whose inheritance correlates with the expression of the gene AT1G31990. eQTL peaks are shown as vertical dashed lines with 1.5 LOD drop confidence intervals shown as grey shaded regions. The location of the AT1G31990 gene is shown as a solid vertical black line. The black line labelled “All” shows the FDR calculated from the log ratio test of all 5 parent 2 haplotypes compared to Col-0. (B) Violinplot showing the gene expression of AT1G31990 in nuclei separated by the cis-haplotype (i.e., the haplotype at AT1G31990). Nuclei that inherit the Db-1, Ms-0, Rubezhnoe-1, or Tsu-0 haplotype of AT1G31990 have significantly reduced expression of AT1G31990, compared to sister nuclei that inherit the Col-0 haplotype. (C) Synteny plot showing the conservation of the AT1G31990 locus and neighbouring loci in the 6 accessions used in this study. AT1G31990 is conserved only in Col-0 and Kar-1. (D) eQTL plot showing the haplotypes whose inheritance correlates with the expression of the gene CTF7 (AT4G31400). eQTL peaks are shown as vertical dashed lines with 1.5 LOD drop confidence intervals shown as grey shaded regions. The location of the CTF7 gene is shown as a solid vertical black line. The black line labelled “All” shows the FDR calculated from the log ratio test of all 5 parent 2 haplotypes compared to Col-0. (E) Violinplot showing the gene expression of CTF7 in nuclei separated by the cis-haplotype (i.e., the haplotype at CTF7). Nuclei that inherit the Db-1, or Tsu-0 haplotype of CTF7 have significantly increased expression of CTF7, compared to sister nuclei that inherit the Col-0 haplotype. (F) Gene track showing promoter haplotypes of the CTF7 locus between Col-0, Db-1 and Tsu-0. CTF7 is a known target of the E2F family of transcription factors that bind to a GGCGCCA motif in the CTF7 promoter (DAP-seq binding peak of E2FA from O’Malley and colleagues 2016 is shown in orange). The Db-1/Tsu-0 haplotype of CTF7 contains an extra copy of this binding motif, which may affect E2F recruitment and explain CTF7 expression changes. The data underlying this figure can be found in datasets 1, 2 and 3 at https://doi.org/10.5281/zenodo.14864053.
Fig 5.
Pollen snRNA-seq identifies a trans-eQTL hotspot on Chromosome 1.
(A) Histogram showing the distribution of parent 2 specific trans-eQTL hits across the Col-0 genome. trans-eQTL hits were filtered for those that were significant at the 5% threshold in each parent 2 contrast. A trans-eQTL hotspot was identified at Chr1:26 Mb, which was named PSV1 (green vertical shaded region). (B) eQTL plot showing the haplotypes whose inheritance correlates with the expression of the gene PAB7 (AT2G36660). eQTL peaks are shown as vertical dashed lines with 1.5 LOD drop confidence intervals shown as grey shaded regions. The location of the PAB7 gene is shown as a solid vertical black line. The black line labelled “All” shows the FDR calculated from the log ratio test of all 5 parent 2 haplotypes compared to Col-0. (C) Violinplot showing the gene expression of PAB7 in nuclei separated by the PSV1-haplotype. Nuclei that inherit the Db-1 or Rubezhnoe-1 haplotype of PSV1 have significantly higher expression of PAB7, compared to sister nuclei that inherit the Col-0 haplotype. (D) UMAP projection from Ichino and colleagues 2022, showing the expression of PAB7 throughout the developmental stages of the pollen. PAB7 is only expressed in the sperm nucleus cluster, and is absent from microspore, generative and vegetative nuclei, as well as from the soma. The data underlying this figure can be found in datasets 1, 2 and 3 at https://doi.org/10.5281/zenodo.14864053.
Fig 6.
Reproducibility of the PSV1 trans-eQTL hotspot across experiments.
(A) Genome-wide map of eQTL positions (x axis) shown relative to the genomic position of the gene whose expression was tested (y axis). Error bars represent the confidence intervals for each eQTL determined using the 1.5 LOD-drop method, and the size of the points is directly proportional to the LOD score of the eQTL. The orange shaded diagonal indicates the region within which eQTLs were considered to be cis-eQTLs. The green shaded region indicates the PSV1 hotspot of trans-eQTLs, and the yellow shaded region indicates the CPV1 trans-eQTL hotspot. (B) Volcano plot showing the modelled effect of the Db-1 PSV1 haplotype on the expression of genes with a trans-eQTL peak at PSV1. Blue dots represent genes with novel PSV1 trans-eQTLs identified only in the second dataset, whilst orange dots represent genes which also had PSV1 trans-eQTLs identified in the first dataset. (C) Venn-diagram showing the overlap of PSV1 trans-eQTL genes identified in Db-1 and Rubezhnoe-1 compared to Col-0 in the first dataset, with the PSV1 trans-eQTL genes identified in Db-1 compared to Col-0 in the second dataset. (D) Histogram showing the distribution of cell/nucleus-type specific trans-eQTLs identified on Chromosome 1 in sperm and vegetative nuclei. A broken axis is used to allow inspection of the CPV1 locus, which is broader and has fewer significant trans-eQTLs than the PSV1 locus. The data underlying this figure can be found in datasets 4 and 5 at https://doi.org/10.5281/zenodo.14864053.
Fig 7.
Fine mapping of the PSV1 locus implicates the mitotic cell cycle regulator DUO3.
(A) QTL plot showing the haplotypes whose inheritance correlates with the first principal component of the expression of genes which have a PSV1 trans-eQTL peak. Inset shows the fine-mapping of the PSV1 locus. Confidence intervals for the QTL locus (QTL CI) were identified using the 1.5 LOD drop method (grey shaded area). The candidate gene DUO3 is located at the center of the PSV1 interval (DUO3 location shown in black). (B) Dotplots showing the structure of the DUO3 promoter in the five parent 2 accessions compared to Col-0. The promoter region contains a hyper allelic tandem repeat. (C) Gene track showing the SNP sharing patterns between the five parent 2 accessions at DUO3 and the upstream gene/promoter region AT1G46563. Missense variants are denoted with asterisks. SNPs shared between Db-1 and Rubezhnoe-1 (but not other accessions) are shown in orange. Two regions of haplotype sharing are visible: one in the DUO3 promoter region, and another over the DUO3 3' UTR (note that in Araport11 the DUO3 3' UTR is incorrectly annotated as a separate gene). (D–E) Violinplots showing the gene expression of (D) UBC20 and (E) CYCB1;2 in sperm and vegetative nuclei separated by the haplotype of PSV1. Sperm nuclei that inherit the Db-1 haplotype of PSV1 have significantly increased expression of UBC20 and decreased expression of CYCB1;2, compared to sister nuclei that inherit the Col-0 haplotype. The data underlying this figure can be found in datasets 4 and 5 at https://doi.org/10.5281/zenodo.14864053.