Sequence features of the new split-inteins. Multiple sequence alignment of N-terminal (A) and C-terminal (B) halves of full-length split-inteins. Conserved motifs of the HINT protein-splicing family are boxed and labeled, and active-site residues are marked with an asterisk. The sequence alignment was refined based on structural modeling with the cyanobacterial DnaE split-intein as a template (PDB code: 1ZD7). Sequences are named after their protein host. (C) Electrostatic characteristics of full-length split-inteins. The number of salt-bridges was calculated from the modeled tertiary structures, and is indicated together with the local charges of the two interacting beta-strands of the intein-halves. (D) Illustration of salt-bridges at the interaction interface between the N-terminal (blue) and C-terminal (red) halves, in the modeled structure of gp41-1. The two longest anti-parallel beta-strands of the intein molecule are shown in cartoon representation.

Nucleotide features of endonuclease genes. (A) RNA hairpin structures at the 5′ untranslated region of endonuclease ORFs in the gp41-1 (representing the very similar sequences of gp41-1–7), nrdA-5 and DnaE-1 gene loci. Initiator codons are marked by arrows, conserved putative T4 late promoter elements are boxed, and conserved sequence motifs (Supplementary Figure S4) are highlighted in grey. The expected values for motifs 1 and 2 are 1.7⁻¹⁰ and 9.9⁻³, respectively. RNA structures were calculated using the Vienna package (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi), and sequence motifs were identified using the MEME program. (B) Overlapping#protein coding regions of endonuclease 3′ termini and the 5′ termini of their downstream genes.

Schematic representation of fractured gene arrangements. Genomic arrangement of 27 loci, assembled from GOS reads, and grouped according to the type of the enzyme host. Protein coding regions are shown as rectangles, with the enzyme hosts in green, split-intein parts in blue, and endonucleases in red. N-terminal intein half (In) and C-terminal intein half (Ic); free-standing homing endonuclease domains: GIY-YIG (GIY), very short repeat like (Vsr). Abbreviations for other gene names are specified in the text. Possible hairpin structures are marked as vertical lines on the 5′ untranslated regions of endonuclease genes. Coding frames overlap are marked by offset of overlapping coding regions.

Sequence features of the Vsr-like putative homing endonuclease family and its similarity to Vsr repair endonucleases. (A) Conserved sequence motifs of Vsr-like putative homing endonucleases, and their sequence logos. (B) Structure based alignment of DnaE-2 locus Vsr-like putative homing endonuclease with E.coli Vsr repair endonuclease (Vsr; PDB code 1CW0). Residues that were modeled in similar positions and backbone conformations after sequence threading and energy minimization are shown in upper case; unaligned sequence regions are shown in grey lowercase. Identical residues are highlighted in red. Vsr active site residues are marked by asterisks, DNA binding residues are marked by ‘d's, and Zinc binding residues are marked by ‘z's. The secondary structure of Vsr is shown above its sequence. Conserved sequence motifs of DnaE-2 are marked as in A. The Phyre server Z-score for this alignment was 5 × 10⁻¹⁴. (C) Predicted structure of DnaE-2 locus Vsr-like protein positions 196–315, and its similarity to structure of Vsr (D) Motif to motif alignment of the Vsr-like putative homing endonucleases (top) with Vsr repair endonuclease (bottom). Alignments of the second and third (rightmost) aligned blocks are significant, expected to occur by chance 2e⁻² and <5e⁻⁷. The first aligned block has a non-significant score since it is shorter and less conserved. Nevertheless, the alignment is probably genuine since the corresponding Vsr and Vsr-like regions were also found aligned in sequence to sequence, sequence to multiple alignment, and structure threading alignments. Functional residues of the Vsr endonuclease are marked as in B.

A model for gene breaking by intein-targeted homing endonuclease invasions. Observed genomic organizations are numbered in Arabic numbers, and putative evolutionary processes are marked with arrows and Roman numbers. Protein coding genes can be invaded by an in-frame intein domain (I). Invasion of the intein domain gene by an endonuclease gene can fracture the gene in two (II), or insert the endonuclease gene in frame within the intein (IIa). Invasion of an endonuclease gene into an intein-less gene can fracture that gene (Ia). Such an invasion can also insert the endonuclease gene into a group I intron (Ib). Loss of the endonuclease gene from the split-intein locus (III) may be followed by dislocation of the two fractured gene parts into two separate loci (IV). Loss of the flanks of an endonuclease gene in a fractured gene locus could insert the endonuclease coding region in-frame into the intein, recreating a contiguous gene (IIIa).