Metagenomic reconstructions of bacterial CRISPR loci constrain population histories

Abstract

Bacterial CRISPR-Cas systems provide insight into recent population history because they rapidly incorporate, in a unidirectional manner, short fragments (spacers) from coexisting infective virus populations into host chromosomes. Immunity is achieved by sequence identity between transcripts of spacers and their targets. Here, we used metagenomics to study the stability and dynamics of the type I-E CRISPR-Cas locus of Leptospirillum group II bacteria in biofilms sampled over 5 years from an acid mine drainage (AMD) system. Despite recovery of 452,686 spacers from CRISPR amplicons and metagenomic data, rarefaction curves of spacers show no saturation. The vast repertoire of spacers is attributed to phage/plasmid population diversity and retention of old spacers, despite rapid evolution of the targeted phage/plasmid genome regions (proto-spacers). The oldest spacers (spacers found at the trailer end) are conserved for at least 5 years, and 12% of these retain perfect or near-perfect matches to proto-spacer targets. The majority of proto-spacer regions contain an AAG proto-spacer adjacent motif (PAM). Spacers throughout the locus target the same phage population (AMDV1), but there are blocks of consecutive spacers without AMDV1 target sequences. Results suggest long-term coexistence of Leptospirillum with AMDV1 and periods when AMDV1 was less dominant. Metagenomics can be applied to millions of cells in a single sample to provide an extremely large spacer inventory, allow identification of phage/plasmids and enable analysis of previous phage/plasmid exposure. Thus, this approach can provide insights into prior bacterial environment and genetic interplay between hosts and their viruses.

Figure 1

CRISPR spacer diversity in Leptospirillum group II. (a) Rarefaction curve for spacer groups recovered from the 5way March 2002 sample (black line) and UBA July 2005 sample (grey line) datasets. Note that neither curve is approaching saturation, despite deep sampling. (b) Rank abundance graph for the 5way CRISPR showing that only a few spacer groups were highly sampled (>1000 counts).

Figure 2

Reconstruction of Leptospirillum group II CRISPR loci variants from 5way, UBA and C75 datasets. CRISPRs are shown vertically from trailer to leader end, with spacers represented as wide rectangles. White rectangles represent spacers shared between at least two CRISPR loci variants while colored rectangles represent spacers unique to a specific locus. Stripped lines show spacer loss. The two black rectangles in the C75 strain denote genes (transposase and hypothetical, from top to bottom) that interrupt the CRISPR locus. Note that the 5way variants are shown split in half owing to space constraints. In the eight columns right of each reconstructed CRISPR variant, the placement of squares indicates the sample that contained the matching mobile element sequence. The eight columns represent the following samples (from left to right): 5way-Mar 2002, UBA-Jun 2005, UBA-Nov 2005, C75-Jun 2006, C75-Aug 2006, C75-Nov 2006, C75-May 2007 and C75-Aug 2007. Perfect spacer matches with a PAM are shown as black squares while perfect spacer match without a PAM and imperfect spacer match with or without a PAM are shown as grey squares.

Figure 3

Frequency of tri-nucleotide sequences in the PAM position of proto-spacers found across all samples. Only the accurate Leptospirillum group II PAM sequence (‘AAG') and imperfect PAMs (allowing for one polymorphism in any position) are shown. Relative abundances of perfect (black) and imperfect (grey) matches are shown.

Figure 4

Abundance of Leptospirillum group II spacers from 5way and UBA samples matching to non-CRISPR host genomic regions. Perfect (solid) and imperfect (striped) spacer matches to intergenic and intragenic regions in the 5way type Leptospirillum group II genome (black) and in the UBA type Leptospirillum group II genome (grey).

Figure 5

Summary of Leptospirillum group II and group III spacer matches to non-CRISPR, non-host genome reads across all datasets. Matches are separated into four categories, as listed in the legend.

Figure 6

Different types of Leptospirillum group II spacer matches to all targets in the metagenomic datasets (excluding CRISPR and the host genome). The four types of matches include: perfect spacer matches with a PAM, imperfect spacer matches with a PAM, perfect spacer matches without a PAM and imperfect spacer matches without a PAM. (a) Plot shows relative abundance of matches from all spacers. (b) Plot shows relative abundance of matches from all spacers from the trailer end, limited to the region containing shared spacers (Figure 2). (c) Plot shows relative abundance of all matches from spacers from the leader end, limited to the spacers not shown in Figure 2. (d) The ratio of matches from spacers at the leader end relative to matches from all spacers.

Figure 7

Spacers with matches to phage AMDV1 in Leptospirillum group II CRISPR loci from 5way, UBA and C75 datasets. Reconstructed loci are represented in the same manner as in Figure 2. In the column right of each reconstructed locus, the placement of black squares indicates the spacer has a perfect or imperfect match to phage AMDV1.