Maize centromere structure and evolution: sequence analysis of centromeres 2 and 5 reveals dynamic Loci shaped primarily by retrotransposons

Abstract

We describe a comprehensive and general approach for mapping centromeres and present a detailed characterization of two maize centromeres. Centromeres are difficult to map and analyze because they consist primarily of repetitive DNA sequences, which in maize are the tandem satellite repeat CentC and interspersed centromeric retrotransposons of maize (CRM). Centromeres are defined epigenetically by the centromeric histone H3 variant, CENH3. Using novel markers derived from centromere repeats, we have mapped all ten centromeres onto the physical and genetic maps of maize. We were able to completely traverse centromeres 2 and 5, confirm physical maps by fluorescence in situ hybridization (FISH), and delineate their functional regions by chromatin immunoprecipitation (ChIP) with anti-CENH3 antibody followed by pyrosequencing. These two centromeres differ substantially in size, apparent CENH3 density, and arrangement of centromeric repeats; and they are larger than the rice centromeres characterized to date. Furthermore, centromere 5 consists of two distinct CENH3 domains that are separated by several megabases. Succession of centromere repeat classes is evidenced by the fact that elements belonging to the recently active recombinant subgroups of CRM1 colonize the present day centromeres, while elements of the ancestral subgroups are also found in the flanking regions. Using abundant CRM and non-CRM retrotransposons that inserted in and near these two centromeres to create a historical record of centromere location, we show that maize centromeres are fluid genomic regions whose borders are heavily influenced by the interplay of retrotransposons and epigenetic marks. Furthermore, we propose that CRMs may be involved in removal of centromeric DNA (specifically CentC), invasion of centromeres by non-CRM retrotransposons, and local repositioning of the CENH3.

Figure 1. Fiber FISH map of a 342 kb region within the approximately 7 Mb B73 centromere 5.

An oat-maize addition line for B73 chromosome 5 shows a predominantly CentC-containing region with interspersed CRMs that is flanked by a CRM1-rich region. CentC  =  blue, CRM1  =  green, CRM2/CRM3  =  red.

Figure 2. Fine-scale physical maps of centromere 2.

(A,B) Chromosomal views. (A) Moving average of 9 windows of the number of sequence reads mapped per 100 kb window using MUMmer (red line) or BLAST (purple line) . Colored boxes denote single CRM elements whose insertion was dated using the method of San Miguel et al. . Only elements that have inserted outside of the functional centromere are shown. Filled squares  =  full-length elements, empty squares  =  fragmented elements. κ  =  estimated number of nucleotide substitutions per site. (B) centromeric repeats CRM1, CRM2, CRM3, CRM4 and CentC mapped onto the reference chromosomes using competitive BLAST and graphed as number of nucleotides per 100 kb window. (C–E) Close-up of centromere region. The functional centromere plus approximately 2.3 Mb of pericentromeric region are shown. (C) CENH3 data same as (A). Retroelements include CRMs not pictured in (A) and non-CRM elements (triangles - details in Table S6); filled symbols  =  full-length elements, empty symbols  =  fragmented elements. Only two bak1 elements have κ>0.1 and are located at 91,278,432 (κ = 0.24) and 92,902,773 (κ = 0.16) and for space reasons are drawn at κ = 0.1. (D) Genetic and molecular markers used to anchor this region to chromosome 2 – see Table S7 for details. Large vertical bar denotes contig gap in reference chromosome. (E) Centromeric repeats as in (B).

Figure 3. Fine-scale physical maps of centromere 5.

Panels and legend as in Figure 2. Elements with κ>0.1 include: (A,B) a CRM1 at 45,864,292 (κ = 0.125), CRM4s at 96,757,409 (κ = 0.101), 115,640,763 (κ = 0.14), 118,276,888 (κ = 0.18), and 138,633,510 (κ = 0.172), (C,D) cinful at 107,583,697 (κ = 0.11) and 107,592,411 (κ = 0.11).

Figure 4. Successive centromere invasion by different CRM1 recombinant subgroups and CRM2 document centromere location over time and the progressive split of centromere 5.

CRM elements are graphed by chromosome coordinate and insertion time (κ) for (A) centromere 2 and (B) centromere 5. Note that the more recent insertions represented by the more recently derived recombinant CRM1 elements R4 and R5 are tightly associated with the present-day CENH3 region. Older elements indicate CENH3 location in the past. Linear regression lines were calculated for all elements of the ancestral CRM1 B/R1 subfamily and document the shift of the centromeres over time. Boxes denote approximate centromere positions at different times based on CRM elements for which the time of insertion could be calculated, except for box C′, the left and right borders of which are based on the CENH3 data and the sole CRM element, respectively. Boxes illustrate the gradual increase in size of “I” over time. Green bars denote approximate positions of CentC clusters.

Figure 5. Karyotype of maize inbred line B73 illustrating CRM1 and CRM2 distribution.

CRM1 was labeled with Texas Red and CRM2 with AlexaFlour 488 (green). Other features that permit the classification of each chromosome are 180 bp knob repeat labeled with Cascade Blue, subtelomeric probe 4-12-1 and 5S ribosomal RNA labeled with AlexaFluor 488, 5S rDNA with Texas Red (to produce a yellow composite) and the TR1 knob repeat labeled with Cy5 (pseudocolored white). The alignment of red and green labelings of the 5S cluster on chromosome 2 assures the relative alignment of CRM1 and CRM2 in the centromeric regions. Note that centromeres 2 and 8 contain relatively high amount of CRM2 relative to CRM1. The reverse is true for centromere 9. The merged image is at the top. The Texas Red signal is shown in the middle panel that includes CRM1 at the primary constriction and the lower panel green image includes the CRM2 signal at the primary constriction. Chromosomes are representative of multiple metaphase cells each observed from root tip biological replicates. Scale bar  = 5 µm.

Figure 6. Relative positioning of CRM1 and CRM2 on somatic metaphase chromosomes.

A somatic root tip chromosome spread of B73 is presented and labeled with CRM1 and CRM2 as described in Figure 5. As determined from the karyotyping features shown in Figure 5, the presence of a smaller interstitial knob on the long arm identifies the boxed chromosome as 5. The inset to the upper left illustrates the different channels from left to right, CRM1, CRM2, CRM1+CRM2 and the chromosome composite. As is generally the case with the chromosomes in the spread, CRM1 label has a more internal positioning than CRM2, which lies to the exterior of the chromosome opposed to the sites of sister chromatid cohesion although there is also obvious overlap. The metaphase spread is representative of multiple metaphase cells each observed from root tip biological replicates. Scale bar  = 5 µm.