FANCD2 binding identifies conserved fragile sites at large transcribed genes in avian cells

Abstract

Common Chromosomal Fragile Sites (CFSs) are specific genomic regions prone to form breaks on metaphase chromosomes in response to replication stress. Moreover, CFSs are mutational hotspots in cancer genomes, showing that the mutational mechanisms that operate at CFSs are highly active in cancer cells. Orthologs of human CFSs are found in a number of other mammals, but the extent of CFS conservation beyond the mammalian lineage is unclear. Characterization of CFSs from distantly related organisms can provide new insight into the biology underlying CFSs. Here, we have mapped CFSs in an avian cell line. We find that, overall the most significant CFSs coincide with extremely large conserved genes, from which very long transcripts are produced. However, no significant correlation between any sequence characteristics and CFSs is found. Moreover, we identified putative early replicating fragile sites (ERFSs), which is a distinct class of fragile sites and we developed a fluctuation analysis revealing high mutation rates at the CFS gene PARK2, with deletions as the most prevalent mutation. Finally, we show that avian homologs of the human CFS genes despite their fragility have resisted the general intron size reduction observed in birds suggesting that CFSs have a conserved biological function.

Figure 1.

Genome-wide FANCD2 binding analysis. (A) Circos plot of Gallus gallus chromosomes 3 and 4 indicating replication timing, gene position, gene density, FANCD2 ChIP-seq results, transcription levels and DNA sequence features. Replication timing obtained by (41) is indicated by blue shading where darker is later (level 9, outermost circle). Position of genes in forward or reverse orientation is shown as gray bars, above or below center, respectively (level 8). Genes larger than 500 kb are marked with a red line at the center (level 8). Gene density per megabase window is indicated with gray bars. The average gene density in the chicken genome is 16.6 genes per megabase with a standard deviation (σ) of 13.1. Regions with a gene density of more than 16.6 +2*σ are black (level 7). The ChIP-seq data is represented in 100 kb windows showing the –log(p) values from two FANCD2 ChIP-seq experiments (exp:G2/M, marked by green lines) and (exp:async, marked by yellow lines) where cells treated with APH have or have not been subjected to elutriation before ChIP of FANCD2, at levels 6 and 5, respectively. Significant peaks identified using the DROMPA2 peak finder are shaded in red (40). RNA-seq results are plotted as RNA-seq depth over exon density at level 4. Twist angle flexibility (level 3), AT content (level 2) and repeat content (level 1, innermost circle) are indicated by color codes showing deviations from mean as indicated. Positions of MACROD2, PARK2, GRID1, CCSER1 and a putative ERFS are indicated. (B) The 10 most significant peaks from exp:G2/M listed with the position in the genome, the -log(p) value and name and size of overlapping genes >200 kb.

Figure 2.

Peak regions from FANCD2 ChIP-seq form breaks on metaphase chromosomes after APH treatment. (A) Quantification of breaks or gaps at the indicated loci with or without 0.6 μM APH treatment for 16 h. BCL2 and RNF8 are control loci, which were not bound by FANCD2 in the ChIP-seq experiment. At least 170 metaphases were counted for each locus. (B) Representative image of metaphase spread from DT40 cells treated with 0.6 μM APH for 16 h. The FISH probe hybridizes to PARK2, which is broken (close up image and red arrowhead). One allele of PARK2 is not broken (grey arrowhead).

Figure 3.

FANCD2 ChIP-seq peaks from exp:G2/M form two distinct clusters based on exp:async, transcription, replication timing and gene density. Heatmap showing multivariate clustering of significant peaks from exp:G2/M using peak values from ChIP-seq exp:async, RNA-seq, replication timing and gene density. Cluster A includes putative CFSs, which are characterized by FANCD2 binding, low transcription, late replication and low gene density. Cluster B includes putative ERFSs, which are characterized by FANCD2 binding only in G2/M, high transcription, early replication and high gene density. Rows are centered; unit variance scaling is applied to rows. Red colors indicate a high relative value and blue colors indicate a low relative value (as indicated by the red/blue heatmap unit variance color scale). Rows are clustered using correlation distance and average linkage. 96 rows, 4 columns.

Figure 4.

Very long transcripts are generated at CFS genes. (A) Long genes replicate late in the cell cycle. Graph shows the replication timing of genes >200 kb. Gene size (kb) given on the X-axis is plotted against replication timing in arbitrary units (AU) on the Y-axis with highest values corresponding to early replication. (B) Fragility of long genes correlate with transcription. Plots show the absolute RNA-seq read numbers in large fragile genes. (C–E) Non-fragile genes exhibit low or partial expression. Plot shows the absolute RNA-seq read numbers from large non-fragile genes. The X-axis represents the gene including introns. Exons are indicated by grey pins on the X-axis. Position of the reads (black) are indicated relative to the start of the coding sequence, X = 0. The Y-axis shows the number of reads. The total number of mapped reads was 17 725 385.

Figure 5.

The PARK2 locus has a high mutation rate in response to APH. (A) Top. Map of the PARK2 genomic locus. The chicken PARK2 gene is indicated by the thick green arrow. Exons are indicated by red boxes. Ruler indicates position along the gene (bases). Integration site for the HyTK cassette is indicated. HyTK is not drawn to scale. Bottom. Schematic representation of the HyTK cassette, including the CMV promotor, polyA signal and the coding sequence for hygromycin phosphotransferase (green) fused to the coding sequencing for HSV1 thymidine kinase (blue). Ruler indicates position along the gene (bases). (B) Mutation rates at PARK2 and at the OVAL control locus with or without treatment with the replication inhibitor APH. Briefly, single cells were expanded for 10 days in medium with or without 0.2 μM APH. Single colonies were expanded further before dilution-plating in ganciclovir counter-selective medium. The number of colonies appearing is used to calculate mutation rates using the Ma-Sandri-Sarkar Maximum Likelihood Estimator method provided by FALCOR (44). Error bars represent 95% confidence intervals. (C) Percentage of deletions in a number of mutants derived from PARK2-HyTK and OVAL-HyTK cell lines, which were untreated or APH-treated. P-value is indicated (exact binomial test).

Figure 6.

Class specific intron length preservation in long fragile Gallus gallus genes. Intron lengths in base pairs (black points) for 10 genes and 203 vertebrates together with fitted values (colored points) for the additive model (left) and the class-gene interaction model (right). Classes and subclasses are indicated by different colors. Gallus gallus is marked by yellow. Gene names are given as the ortholog identifier with Gallus gallus gene names in parentheses. The subgrouping of Bird identified by the additive model is marked by the shaded background on the left figure.