The activity, divergence, and evolutionary degradation of modern-day homing endonucleases and behavior of their reconstructed ancestors

Juliana C Young; Abigail R Lambert; Janet M Young; Lindsey A Doyle; Miriam Silverstein; David R Edgell; Barry L Stoddard

doi:10.1093/nar/gkag924

. 2026 Sep 29;54(18):gkag924. doi: 10.1093/nar/gkag924

The activity, divergence, and evolutionary degradation of modern-day homing endonucleases and behavior of their reconstructed ancestors

Juliana C Young ^1,^c, Abigail R Lambert ^2,^c, Janet M Young ³, Lindsey A Doyle ⁴, Miriam Silverstein ⁵, David R Edgell ^6,^✉, Barry L Stoddard ^7,^✉

¹ Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States

² Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States

³ Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States

⁴ Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States

⁵ Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States

⁶ Department of Biochemistry, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON N6A 5C1, Canada

⁷ Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States

^✉

To whom correspondence should be addressed. Email: bstoddar@fredhutch.org

^✉

Correspondence may also be addressed to David R. Edgell. Email: dedgell@uwo.ca

Juliana C Young and Abigail R Lambert Co-first authors.

Roles

Juliana C Young: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft

Abigail R Lambert: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft

Janet M Young: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft

Lindsey A Doyle: Formal analysis, Investigation, Validation, Visualization, Writing - review & editing

Miriam Silverstein: Formal analysis, Investigation, Visualization, Writing - review & editing

David R Edgell: Conceptualization, Formal analysis, Investigation, Methodology, Writing - review & editing

Barry L Stoddard: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing - original draft

PMCID: PMC13619312 PMID: 42805603

Abstract

Homing endonucleases (HEs) are selfish genetic elements that drive the mobilization of their own coding sequences, often in concert with surrounding introns. Homing endonuclease genes (HEGs) usually display life cycles in which they accumulate inactivating mutations after invading a host genomic target site, leading to eventual removal from the genome. We identified several hundred novel HEGs and determined the distribution of their proteins’ behaviors and activities. Approximately 10% are expressed as well-folded functional proteins that cleave predictable DNA target sites. Another ∼20% display significant expression but little to no cleavage activity; the remainder display severely reduced expression. Despite the presence of debilitating mutations throughout most HEGs, ancestral reconstructions yielded endonucleases with improved expression and stability. One such reconstructed model, at a hypothetical node preceding highly diverged HEs that cleave unique target sites, binds (but does not cleave) their individual targets. It instead cleaves a DNA sequence that resembles a hybrid of those modern-day DNA targets, while displaying a similar specificity profile to previously characterized wild-type HEs. Its DNA-bound crystal structure adds detail to our understanding of how HE DNA contacting surfaces and residues shift and rearrange during evolution, ultimately leading to their action at new target sites.

Graphical Abstract

For image description, please refer to the figure legend and surrounding text.

Introduction

Mobile genetic elements (MGEs) are pervasive agents of genomic change and innovation that are found throughout all domains of life [1]. Such sequences are typified by DNA transposons [2], retrotransposons [3], plasmids [4], self-splicing introns and inteins [5], integrated and conjugative elements (ICEs) [6], bacteriophage-like elements [7], integrons [8], CRISPR-associated transposons [9], and many other specialized mobile elements. Collectively, they facilitate horizontal gene transfer, genomic rearrangement, and the spread of genetic novelty between species across both single-cell microbial [10–12] and multi-cellular [13] organisms. Regardless of their origin and genomic environment, MGEs often provide little or no selective advantage to their hosts and usually succumb to evolutionary degradation and inactivation due to the effect of accumulated mutations and the action of genomic defense mechanisms [14, 15]. This dynamic process reflects the balance between a potential benefit to the host (such as the delivery by the MGE of a useful new gene product such as a viral restriction or antibiotic resistance factor) versus a potential cost (e.g. insertional mutagenesis leading to inactivation of a host gene, and/or an increase in genome size leading to increased DNA replication time [16]).

Homing endonucleases (HEs), also termed “meganucleases,” are highly specific DNA-cleaving enzymes that are encoded either by stand-alone genes or by reading frames embedded within surrounding self-splicing introns or inteins (reviewed in [17]). They are ubiquitous throughout all forms of microbial life, being encoded in phage, bacteria, archaea, and the organellar genomes of single-cell eukaryotes. Homing endonuclease genes (HEGs) encode proteins that introduce DNA double strand breaks or nicks at specific genomic targets, which are then repaired through homology-driven recombination and the corresponding insertion of the entire mobile element into the site of DNA repair [18–20]. When the HEG is a free-standing reading frame, its mobility is generally limited to invasion of intergenic target sites (to avoid disruption of a surrounding host gene [21]). In contrast, when a HEG is embedded within a self-splicing intron or intein, its range of insertion sites is greatly increased, because the entire element can invade host genes without disrupting their expression and the generation of their corresponding products.

The propagation of HEGs is extremely efficient, leading to dominant, non-Mendelian inheritance in diploid genomes [18], gene transfer between phage and cellular genomes [22, 23], genetic competition between HEGs during mixed phage infections [24], and intron transfer between unrelated subcellular compartments (usually chloroplast and mitochondrial genomes) [25]. The accumulation of HEs and mobile introns in individual genomes can be extremely high. For example, over 10% of the T4 phage genome encodes a total of 12 free-standing and 3 intron-encoded HEGs (reviewed in [26]). Similarly, alignment and comparisons of microbial ribosomal DNA sequences indicate the presence of many intervening HEG sequences [27, 28].

At least six families of HEs have been identified. The most well-studied (and the focus of this study) are LAGLIDADG homing endonucleases (“LHEs”), which contain a small, highly conserved protein domain (typically ∼150 residues in size) encompassing both an extended DNA-binding surface and the enzyme’s active site. They are encoded within fungal mitochondria, algal chloroplast, and archaeal genomes. LHEs exist both as homodimeric proteins (in which two identical protein subunits, each containing a single identical copy of the endonuclease fold, collaborate to act at highly symmetric DNA target sites) and monomeric proteins (in which two non-identical domains are present within a single larger protein chain, allowing them to act at far more asymmetric DNA target sites [17]). LHEs are the most sequence-specific class of HEs, typically recognizing 22 base pair target sites, and are the focus of engineering efforts for a variety of gene editing applications [29–31].

Comparative analysis of a small clade of well-characterized mobile introns and their corresponding LAGLIDADG HEGs (all encoding closely related homologues of the I-SceI HE in yeast) indicated that such elements display an evolutionary life cycle of invasion, mutational degradation, loss from the genome, and subsequent re-invasion by a related element, with an estimated periodicity of approximately 2 million years [32, 33]. A subsequent examination of a single LAGLIDADG endonuclease (I-AniI, encoded in the fungus Aspergillus nidulans) and its corresponding DNA target site provided further details of initial mutations leading to the evolutionary degradation of LHE activity: the enzyme harbors at least two point mutations that have significantly reduced its activity [34], while its insertion target site in the host genome differs from the enzyme’s optimal DNA recognition sequence by at least one basepair [35]. The loss of form and function for any particular LHE is not inevitable, however: the acquisition of a secondary “moonlighting” function of importance to the host (e.g. their occasional participation in folding and splicing of their surrounding cognate intron, termed “maturase activity”) can result in host selective pressure to maintain protein folding and stability, which in turn can stabilize their original endonuclease function [36].

For the majority of LHEs whose sole function is to act only as an endonuclease that drives genetic mobility, one would predict that at any moment in time, a significant fraction of current-day HEGs would encode dysfunctional proteins, destined for eventual removal from the genome. One study that attempted to test that prediction examined the folding and function of a small group of 34 putative LAGLIDADG HEs (that all display significant homology to the well-characterized and highly active I-OnuI enzyme). That analysis found that approximately one-third displayed poor folding behavior, versus approximately two-thirds that were capable of cleaving target sites in their cognate host genomes [29].

To assess the same prediction more thoroughly and in a less biased manner (a goal that was motivated and facilitated by the rapidly growing number of LHE sequences that could be identified within microbial and metagenomic sequence databases), we identified a diverse subset of over 300 reading frames corresponding to putative monomeric LAGLIDADG HEs, all of which are embedded within surrounding group I introns, and then characterized their folding, expression, and DNA cleavage activity. Our analysis was focused on monomeric LHEs for three reasons: their predicted structural similarity to several monomeric LHEs that have been previously characterized [17, 37], their ability to recognize and act at completely asymmetric DNA target sites, and our own ability to predict their target sites (which correspond to a fusion of the DNA sequences that immediately flank the insertion site of the mobile intron or intein and its corresponding HEG).

Our assessment of the expression and DNA cleavage activity of those candidate enzymes revealed a broad spectrum of behaviors, ranging from well-expressed and fully active endonucleases to those that display significant disruptions in their form and function. Subsequently, we identified additional homologues to generate an even larger collection of over 600 putative LHE sequences, which enabled the creation of a more densely populated phylogenetic tree and the generation of reconstructed ancestral sequences at nodes that precede several different modern-day “extant” HEs. This allowed us to address two additional questions: (i) whether ancestral reconstruction attempts can generate active endonucleases, despite the fact that their current day descendants are not under strong selective pressure to maintain function and often contain debilitating mutations; and (ii) whether the behavior of ancestral enzymes might provide additional insight into the evolutionary divergence of their DNA target site specificity.

Materials and methods

Bioinformatic LHE sequence screening

An initial set of 371 sequences encoding putative monomeric LHEs were identified in a 2015 bioinformatic search of all fully sequenced plastid and mitochondrial genomes that were available at that time in the GenBank database [38] for all reading frames that were annotated as containing a group I intron. The exon–intron boundaries for each reading frame were identified, and the collected sequences were examined using a Psi-BLAST search [39, 40] to identify potential LHEs. All candidates were further culled based on a cutoff of 20% or higher sequence identity to any one of nine query sequences available at that time with known, validated endonuclease activity against established target sequences and corresponding DNA-bound structures (DNA-bound crystal structures: I-AabMI PDB ID 4YIT, I-CpaMI 4YIS, I-GpeMI 4YHX, I-GzeII 4Z1X, I-LtrI 3R7P, I-LtrWI 4LQ0, I-OnuI 3QQY, I-PanMI 5ESP, and I-SmaMI 5E5O). Putative ORFs <200 amino acids were assumed to exist as homodimers and were excluded.

Several years after our initial search (in 2023), we expanded our initial collection of LHE sequences via an additional search of available sequence databases, and identified an additional 246 putative LHE genes. This larger set of 625 coding sequences was used for to select subsets of sequences to use for computational ancestral sequence reconstruction (ASR) efforts.

Cloning of synthetic LHE ORFs

A small “pilot” set of 18 putative LHE ORFs from our original bioinformatic search were first ordered from Genscript (Piscataway, NJ) as synthetic genes cloned into the pETCON vector, for expression on the surface of yeast. Those constructs are denoted hereafter with identifiers corresponding to “HE##” throughout this document, corresponding figures, and Supplementary Information. The remaining 353 LHE sequences (which are denoted hereafter with identifiers corresponding to “Gen9-###”) from the 2015 dataset were ordered from Gen9 (Cambridge, MA) as synthetic genes delivered in their pG9m-2 plasmid. Those constructs were synthesized commercially via a parallel high-throughput approach at Gen9 Bio (relying on a technology previously described in [41]) and then sequence-verified. 30 sequences were not able to be synthesized from the Gen9 order, leaving a total of 341 LHE ORFs that were delivered. For all constructs, the gene design and synthesis process accounted for optimal codon usage for yeast expression.

The synthetic LHE ORFs from Gen9 were subcloned en masse into the pETCON yeast surface display vector (Addgene #41522; [42]) via homologous recombination in yeast, and later isolated as single clones. First, 100 ng of each synthetic plasmid was mixed to create a pool of PCR templates. This pool of LHE ORFs was PCR-amplified using primers designed to create 80–90 base pair overlaps with each end of a linearized pETCON yeast surface display vector (generated via digests with NdeI and XhoI restriction enzymes). The pool of PCR-amplified LHE inserts was transformed into EBY100 yeast (American Type Culture Collection, ATCC) alongside open pETCON vector using a lithium acetate transformation protocol [43]. The yeast transformation reaction was cultured overnight, and plasmids containing cloned LHE sequences were extracted from the yeast using a 2-h Zymolase (Zymo Research) digest followed by a modified plasmid extraction protocol utilizing Qiagen miniprep kit components (Qiagen). While cloning in batch format was most effective using yeast, the necessary plasmid prep and sequence verification steps were more efficiently performed in bacteria. Therefore, the plasmids from the yeast miniprep were transformed into DH5α E. coli (New England Biolabs) and plated. Individual bacterial colonies were sequenced to identify successfully cloned LHE ORFs. Plasmids were isolated from bacterial cultures containing unique LHEs, and these plasmids were transformed back into yeast for further analysis. This process was repeated until clones were isolated for 307 unique LHEs (34 LHE sequences were set aside after multiple attempts at subcloning failed).

Yeast induction of surface expression

Yeast cultures were grown from single colonies at 30°C with shaking in selective culture media (Synthetic complete or “SC” media, -Ura -Trp, Sunrise Science Products) + 4% glucose (the higher percentage of glucose over the typical 2% helps to discourage flocculation). For induction of surface expression, yeast cultures were grown from a starting density of 24 million/ml in SC + 2% raffinose + 0.1% glucose for 6 h at 30°C with shaking, washed with sterile water, and then incubated at a density of 20 million/ml in SC + 2% galactose overnight on the benchtop (room temperature).

Yeast staining for surface expression

LHEs expressed on the surface of yeast were analyzed by flow cytometry as previously described [37]. Briefly, yeast were stained for 45–60 min at 4°C in a yeast staining buffer (10 mM HEPES pH 7.5, 180 mM KCl, 10 mM NaCl, 0.1% w/v galactose, and 0.1% w/v bovine serum albumin) with an anti-Myc-FITC antibody (ICL) along with either an anti-HA-APC antibody (BioLegend) or an anti-HA-biotin/Streptavidin-phycoerythrin (SAV-PE) conjugation (BioLegend and Thermo Fisher Scientific) (Fig. 1a). Full-length expression of the LHE was indicated by the presence of FITC signal after staining of the C-terminal Myc tag. Staining of the N-terminal HA tag with a second color could be used to identify LHEs that were not expressed all the way through to the C-terminal Myc tag (presumed to be stalled, unstable, or otherwise truncated).

These experiments were performed in batches over many months, so we chose to quantify and compare levels of surface expression between LHEs using the overall percentage of cells with FITC signal rather than comparing the magnitude of the FITC signal (which varies depending on factors like expression level on a given day and voltage settings on the instrument). Flow cytometric data was analyzed with FlowJo [44] and the percentage of cells with full-length surface expression was quantified by drawing gates around populations of cells with detectable FITC signal (Fig. 1b). Fluorescent signal from staining of the N-terminal HA tag was not quantified. An average percent expression value was calculated for those constructs which were assayed multiple times (see Deposited Raw Data). The quantified % expression values were sorted in order from highest to lowest, and a value of 3% was chosen as the cutoff for designation of “full-length expression” versus “N-terminal only expression.” Some non-zero FITC signal values were the result of noise along the boundary of quantification gates or due to averaging of several zero-expression experiments alongside one with a small amount of expression. This 3% value best matches visual inspection of the flow cytometric data.

Generation of labeled DNA target site substrates

Double-stranded DNA oligonucleotide target site substrates were generated as previously described [45]. Briefly, an oligo template containing a 58 bp DNA target region flanked by two 16 bp primer binding sequences was amplified by PCR with Platinum Taq High Fidelity DNA polymerase (Invitrogen) using labeled primers which add a biotin to one end of the PCR product and an AlexaFluor647 (A647) fluorophore to the opposite end (IDT DNA). The A647 label can be added to either the 5′ or the 3′ end of the DNA target substrate. The resulting dual-labeled, double-stranded PCR products were digested for 6 h with Exonuclease I (New England Biolabs) to remove any unincorporated primers or leftover single-stranded template. The double-stranded DNA target substrates were then purified using a small G-100 Sephadex size-exclusion column (GE Healthcare) in a filter plate.

The position of the A647 label (on either the 5′ end or the 3′ end of the DNA target substrate) does not affect the output of non-tethered cleavage assays. However, we typically make and test substrates with both possible substrate labeling schemes (A647 on the 5′ end versus A647 on the 3′ end) for the tethered flow cytometric cleavage assay, since some LHEs remain bound to one end of the cleaved DNA more tightly than the other. This end-holding behavior can affect the magnitude of the drop in A647 signal which indicates successful cleavage activity in the tethered cleavage assay.

DNA cleavage assays using surface-displayed endonucleases (“tethered” DNA cleavage assays)

LHEs expressed on the surface of yeast were analyzed for DNA cleavage activity using a previously described tethered flow cytometric cleavage assay [37]. Briefly, the C-terminal Myc tag was stained with anti-Myc-FITC antibody, and double-stranded dual-labeled DNA target site substrate was physically tethered to the N-terminus of each surface-expressed LHE (2 million yeast cells per well of the experiment) using an anti-HA-biotin/Streptavidin-PE (SAV-PE) conjugation (40 nM DNA target substrate and 5 nM SAV-PE) (Fig. 2a). The stained yeast were divided into two duplicate wells and incubated in 50 µl cleavage buffer (10 mM HEPES pH 7.4, 150 mM KCl, 10 mM NaCl, 5 mM K-glutamate, and 0.05% bovine serum albumin) for 30–60 min at 37°C in the presence of 5 mM MgCl₂ (allows for cleavage to occur) or 5 mM CaCl₂ (cleavage of the tethered DNA substrate cannot occur). The yeast were analyzed by flow cytometry to measure the A647 signal from the cleaved versus uncleaved tethered DNA target substrate in the magnesium versus calcium samples. Successful target cleavage leads to a drop in A647 signal when the fluorescent tag on the DNA substrate is separated and washed away (Supplementary Fig. S1).

This DNA cleavage activity can be quantified by drawing a thin rectangular gate along the PE axis where the cell population is most dense and calculating the “Mg²⁺/Ca²⁺ ratio” within that gate, equal to the ratio of the median A647 signal from the magnesium (cleavage) sample to the median A647 signal from the calcium (no cleavage) sample (Fig. 2b). The presence of A647 signal in this experiment is due to uncleaved tethered DNA target substrate, so the Mg²⁺/Ca²⁺ ratio represents the fraction of uncleaved DNA target substrate remaining. Therefore, the value of 1 – (Mg²⁺/Ca²⁺ ratio) indicates the fraction of tethered DNA target substrate that was successfully cleaved, and this is the value chosen to represent DNA cleavage activity plotted in Fig. 2c. A sample with no cleavage of the tethered substrate (i.e. no change in A647 signal) has a Mg²⁺/Ca²⁺ ratio of 1, and the corresponding value of 1 – (Mg²⁺/Ca²⁺ ratio) equals zero. A cleavage activity value of 0.1 was chosen as the cutoff for designation of an active versus inactive enzyme. There is a small amount of noise in the quantification of the cleavage data which results in small non-zero values for some inactive enzymes, and this cutoff best matches visual inspection of the flow cytometric data. See the Deposited Raw Data and Supplementary Information for flow cytometric plots and a table of quantified cleavage values, respectively.

DNA cleavage assays using endonucleases released from the yeast surface (“untethered” DNA cleavage assays)

Approximately 5 million yeast cells with surface-expressed LHE were incubated for 30–60 min at 37°C in 40 µl cleavage buffer (described above) along with 5 mM MgCl₂, 20 nM A647-labeled DNA target site substrate, and 10 mM dithiothreitol (DTT). The DTT reduces the disulfide bonds of the AGA1–AGA2 interaction that anchors the expressed LHE to the yeast surface, leading to the release of the enzyme into solution. Depending on the strength of expression, the resulting free protein concentration is estimated to range from roughly 0.3 to 3 nM per 1 million expressing yeast cells in 50 µl of volume [46]. A colorless Ficoll loading dye was added to the reactions before loading samples onto an 11.3% acrylamide gel. The gel was run at 120 V for 60 min, and cleavage bands were visualized using the Cy5 channel on a Typhoon fluorescent scanner (Amersham/GE) (Supplementary Fig. S2a).

Determination of site and centering of DNA target cleavage via run-off DNA sequencing

For the characterization of the 29 active enzymes identified in the first part of this study, 5–10 million yeast cells with surface-expressed LHE were incubated for 1–2 h at 37°C with approximately 1 µg (∼2 nM) of linearized target site plasmid in 50 µl of cleavage buffer containing 10 mM DTT and 5 mM MgCl₂. For later experiments during the characterization of ancestral reconstructions, the digest was performed with 100 nM purified recombinant protein combined with circular target site plasmid at a similar (∼2 nM) substrate concentration. In either case, the entire cleavage reaction was loaded onto a 1% agarose gel to separate the cleaved bands. Cleavage bands (either two cleaved fragments from a linear DNA target or a single linearized band after cleaving circular plasmid DNA) were extracted from the gel and purified with the Zymoclean Gel DNA Recovery Kit (Zymo Research). Purified bands were used as templates for individual BigDye (Thermo Fisher Scientific) Sanger sequencing reactions.

The resulting run-off sequencing chromatograms revealed a single large adenine peak where the Taq polymerase came to the end of the cleaved template. This large “A” peak marked the precise position where that strand of DNA was cleaved by the LHE. In cases where two cleavage product bands were not able to be purified, a single clear run-off sequencing reaction was sufficient to determine the center of the DNA target site given the verified characteristic that LAGLIDADG LHEs cleave their targets to create 4-base 3′ overhangs. In the absence of sufficient purified cleavage products or clean run-off sequencing data, the center of some target sites could still be determined through alignments with known target sites from closely related LHEs. See the Deposited Raw Data for the chromatograms produced from run-off sequencing reactions and their corresponding analysis.

Ancestral sequence reconstructions

In order to reconstruct selected ancestral sequences in the I-OnuI (focal sequences I-OnuI and I-GpeMI) and I-PnoMI groups (focal sequences PnoMI, HE08, and Gen9-348), we first used an expanded collection of 625 HE sequences to identify members of each group as follows.

We constructed a tree of all LAGLIDADG domains in our collection of 625 sequences (Supplementary Fig. S4). Each monomeric LHE usually contains two such domains: we used the hmmsearch tool (HMMER package version 3.3.2) with the “LAGLIDADG_1” Hidden Markov Model (PFAM PF00961.22) [47] as query to extract all domain sequences in the form of an alignment to the HMM. We filtered this alignment to remove sequences with an ungapped length of <75aa (the HMM has 102aa positions), and to remove alignment positions where ≥75% of sequences contained a gap. This filtered alignment contained 1140 domain sequences and 102aa positions. We used ProtTest (version 3.4.2) [48] to infer that best fitting amino acid substitution model is the CpREV model. We then used PHYML (version 3.3.20200621) [49] to estimate a phylogeny, estimating the proportion of invariable sites, with 4 rate categories and estimating the shape of the gamma distribution (i.e. parameters “–pinv e –alpha e -f e -m CpREV”).

We analyzed the resulting LAGLIDADG domain phylogeny using R [50] with the ape [51] and ggtree [52] packages. For each sequence group of interest, for each of the two LAGLIDADG domains, we identified the MRCA (most recent common ancestor) of the group’s focal sequences, then stepped back three nodes deeper in the tree in order to obtain an extended clade that includes outgroups. We selected all descendent tips of these “MRCA-minus-3” nodes, merged the lists obtained from the two LAGLIDADG domains of all focal sequences in each group, and extracted any HEs with full-length sequences (≥280aa) for use in ancestral reconstruction.

For each group, we then aligned full-length selected sequences using MAFFT (version 7.453) [53], trimmed the alignment to remove regions outside the extent of the focal sequences, and removed any sequences whose ungapped length was now <250aa. We also generated a non-redundant alignment, because some sequences were identical to one another within the trimmed region. We also noticed that our sequence list for the I-OnuI group included three engineered sequences, which we removed from further analysis (ancestral reconstruction methods assume sequences evolved naturally). The resulting alignments contained 40 (I-PnoMI clade) sequences or 18 (I-Onu-I clade) sequences (Fig. 3).

Next, we performed ancestral reconstruction on each trimmed alignment, using the BaliPhy suite (version 3.6.0) [54]. In more detail, we used the bali-phy command to run two separate Markov chains to simultaneously estimate phylogenies and alignments. We ran each chain overnight, by which time all chains had gone through >27 000 iterations, and used the bp-analyze command to summarize chain statistics. We used trees-consensus and tree-tool to determine the consensus tree over both chains and ran the tree-tool command again to add internal node names to the consensus tree. We used the cut-range, alignment-chop-internal and alignment-max commands to extract the most likely alignment of extant sequences, skipping the first 1000 samples in each chain. We then used the summarize-ancestors command to add inferred ancestral sequences to the summary alignment.

For each clade, we inspected the resulting consensus trees and summary alignments, and selected ancestral nodes to test experimentally. For ancestors in the PnoMI clade, we tested Bali-Phy’s ancestral reconstructions without modification. For ancestors in the I-OnuI clade, we noticed that indel positions often contained amino acid predictions in the ancestral reconstructions, even when a gap seemed a more likely ancestral state according to parsimony principles. We created edited versions of the ancestral reconstructions that contain gaps at these positions, and proceeded with experimental testing.

Expression and purification of recombinant LHEs

LHEs were expressed as untagged constructs in the pET21d bacterial expression vector (Novagen) using either Escherichia coli BL21-CodonPlus (DE3)-RIL (Agilent) or Rosetta (DE3) (Novagen) competent cells. Both cell types were cultured in 1 l Luria Broth (LB) media (RPI) with 100 µg/ml of ampicillin. Each liter of LB-amp was inoculated with 10 ml of a stationary phase overnight culture and grown at 37°C with shaking until reaching an OD₆₀₀ of 0.6–0.8. Expression was induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.5 mM, and the cultures were placed in a shaking incubator at 16°C overnight. Each liter of culture was then pelleted and stored at −20°C until ready for purification.

Recombinant LHEs were purified in two stages, using heparin ion exchange chromatography followed by size-exclusion chromatography. Frozen bacterial expression pellets were thawed and resuspended in buffer (25 mM Tris–HCl pH 7.5, 200 mM NaCl) supplemented with 50 µM phenylmethylsulfonyl fluoride (PMSF) and 0.01 U/μl benzonase nuclease. The cells were broken open using sonication, and the resulting cell debris was pelleted by centrifugation for 20 min at 18 000 RPM. The supernatant was removed and filtered through a 0.45 micron syringe filter before loading onto a 5 ml Heparin HP HiTrap Column (Sigma–Aldrich). A linear salt gradient was used to elute the bound LHE, and fractions containing the correct sized protein were pooled and concentrated to a volume of 5 ml or less. The concentrated protein sample was filtered through a 0.22 micron pore size filter and loaded onto a Superdex HiLoad 16/60 prep grade size-exclusion column (Sigma–Aldrich) using a 5 ml loop size. The SEC buffer used was as follows: 25 mM Tris–HCl pH 7.5, 200 mM NaCl, and 5% glycerol. See Supplementary Fig. S5b for gels from the purification of aHE31 and Supplementary Fig. S6b for gels from the purification of aHE15.

Determination of LHE thermal stability via thermal denaturation and circular dichroism spectroscopy

Plane polarized light was measured with a J-800 CD spectrometer (JASCO). Purified LHEs were dialyzed into circular dichroism (CD) buffer (10 mM potassium phosphate, pH 7.0), diluted to 25 µM concentration, and scanned in a 1 mm path length cuvette at a wavelength of 210 nm across a range of temperatures from 20°C to 98°C. The melting temperature (T_m) was calculated for each enzyme using denaturation analysis in Spectra Manager II software provided by JASCO.

In vitro cleavage assays using purified recombinant LHEs and 500 bp target site substrates

In vitro cleavage reactions varying either digest time (5–60 min) or enzyme concentration (12.5–400 nM) were performed to compare the cleavage activity of purified recombinant ancestor aHE31 to that of related modern-day enzyme I-PnoMI. For all digest reactions, 1.5 µg of purified 500 bp substrate containing the I-PnoMI target site sequence was combined with cleavage buffer at pH 7.4 (described above), 5 mM MgCl₂, and purified recombinant enzyme in a total reaction volume of 40 μl (corresponding to a DNA substrate concentration of ∼5 nM). Each reaction was incubated at 37°C and stopped with the addition of 4 μl of 0.5 M EDTA solution. Reactions examining the effect of increasing incubation time used an enzyme concentration of 100 nM and varied the total digest time from 5 to 60 min. Reactions looking at the effect of increasing enzyme concentration used an incubation time of 30 min and varied enzyme concentration from 12.5 to 400 nM. Cleavage products were separated on a 1.7% agarose gel at 120 V for 30 min and visualized on a BioRad Gel Doc imager.

The target site substrate used for this set of reactions was generated through PCR amplification of a 500 bp region of an I-PnoMI target site plasmid using PhusionPlus polymerase (Thermo Fisher Scientific). The PCR product was purified by gel extraction from a 1% agarose gel using the Zymo Research gel extraction kit and eluted with 30 μl of water. The purified target concentration was determined by measuring the UV absorbance at 260 nm with a nanodrop spectrophotometer.

Yeast SELEX library generation and screening

In order to determine whether ancestral reconstruction aHE15 was an active endonuclease, and if so what its DNA target specificity corresponded to, we carried out a selection experiment in which (i) a random, unbiased DNA target library was panned against surface-displayed aHE15, to select those sequences that could be bound by the enzyme under moderate stringency (300 mM KCl, pH 7.4), followed by (ii) all captured DNA sequences that could be bound by the enzyme were screened for their ability to be cleaved by recombinant aHE15. For both steps, the selection process was carried out for multiple rounds to enrich the selected DNA targets for binding and then cleavage. Details are provided below.

The generation of the SELEX library substrate for binding-based selection was performed according to a previously described SELEX protocol [55, 56] with a few customized parameters noted here. In brief, aHE15 was induced on the surface of yeast (described above), and 3 million yeast cells were used for each experiment. The double-stranded DNA target substrate for each round of selection was generated by PCR amplification using HiFi Platinum Taq polymerase (Invitrogen) with the initial “SELEX0” fully randomized target created from an oligonucleotide template (IDT) containing 30xN bases flanked by two primer binding sequences (Supplementary Fig. S8a). Three thymidine bases preceding and three adenine bases following the 30xN sequence were added to discourage high-affinity binding to the constant regions of the substrate. A small amount of A647-labeled SELEX0 substrate was generated for use in a series of optimization tests which varied the concentrations of both KCl and total DNA target substrate in the reaction. For each test condition, 3 million induced yeast cells were incubated for 30 min at room temperature in a total volume of 100 μl of binding buffer (100 mM NaCl, 10 mM HEPES pH 7.5, 50 mM K-Glu, and 2 mM CaCl₂) plus a range of 200–300 mM KCl and 250–2000 nM labeled SELEX0 DNA substrate. The goal of binding approximately 5% of the initial randomized oligo substrate was achieved with a combination of 300 mM KCl and 1000 nM DNA target substrate (Supplementary Fig. S8b).

This set of parameters (3 million induced yeast cells, binding buffer with 300 mM KCl, 1000 nM DNA target substrate, 100 μl reaction volume, and 30-min room temperature incubation) was then applied to multiple rounds of SELEX binding, starting with a fully randomized, non-labeled SELEX0 DNA target substrate. After each incubation step, the yeast were washed six times and resuspended in 10 μl of buffer EB (10 mM Tris–HCl, pH 8.5). Next, the yeast were heated in a thermocycler for 10 min at 90°C to release any bound DNA substrate. This elevated incubation temperature was necessary due to the usually high T_m of the aHE15 enzyme (78°C), thus requiring a higher temperature than the original published SELEX protocol for complete denaturation of the protein to release the bound DNA substrate. After incubating at 90 degrees, the samples were spun down to pellet the yeast, and the supernatant containing the heat-released DNA was collected. This pool of bound DNA sequences was then used as the template for another set of PCR reactions to generate the double-stranded DNA target substrate for the next round of SELEX (SELEX1). This binding selection process (Fig. 6a) was repeated for a total of four rounds using the same culture of galactose-induced yeast to ensure equivalent levels of surface-expressed aHE15 for each round.

In vitro cleavage library generation and screen

The in vitro cleavage-based screen was performed largely as described previously [35]. The cleavage selection steps were performed using purified recombinant aHE15 protein instead of yeast surface-expressed enzyme, and the DNA target sites were presented in the form of a plasmid library rather than short double-stranded oligonucleotide substrates. The initial target site library was generated from the DNA target substrates collected from Round 4 of the SELEX binding screen. Primers were designed to add 22 bp Gibson overlaps to the DNA target substrates isolated from SELEX4 to allow for cloning into open pIDT-SMART vector (Genscript) using the NEBuilder HiFi DNA Assembly Kit (New England Biolabs):

SELEX_Gib_forward: CACTGCTCGATCCGCTCGCACCCAGGGATCCATGCACTGTACGTTT
SELEX_Gib_reverse: GTCGCGCTGCTCTCGTCGATCCAGGATCCGCAGTCAAGTGGTTT

For each cleavage selection step, 50 nM of both the target site library plasmid and recombinant aHE15 enzyme were mixed in a 40 μl volume with cleavage buffer pH 7.4 (described above) and 10 mM MgCl₂ to facilitate cleavage. The digest was incubated at 37°C for 2 h, and the entire reaction was loaded into a 0.5% agarose gel and run at 120 V for 40 min. Linearized plasmid was not visible on the gel until Round 3, so both uncut/supercoiled plasmid DNA and a PvuI-linearized plasmid reference were run in adjacent lanes of the gel to help locate the area where the cleaved product should be. This designated area was extracted regardless of whether or not the linearized product was visible. The cleaved/linearized plasmid was extracted and purified using a gel extraction kit (Zymo Research) and eluted with 15 μl of water. The purified linearized plasmid was then re-ligated back into circular DNA using T4 ligase (Thermo Fisher Scientific) and transformed into competent DH5α bacterial cells. Half of the transformation reaction was plated onto LB-AMP plates and half was used to start a 5 ml of LB-AMP overnight miniprep culture. Forty-eight colonies from the plates were sequenced by colony PCR (Supplementary Fig. S9), and a miniprep was performed to generate the plasmid substrate for the next round of cleavage selection. A total of four rounds of cleavage selection were performed (Fig. 6b).

In vitro cleavage assays using purified recombinant aHE15 and short A647-labeled target site substrates

Purified recombinant aHE15 was incubated at 100 nM concentration in a 40 μl reaction volume containing cleavage buffer at pH 7.4 (described above) with 20 nM of an A647-labeled double stranded DNA target site substrate (I-OnuI, I-SmaMI, 4xSeq, 17xSeq) and 5 mM divalent cation (CaCl₂ or MgCl₂). Digests were performed at 37°C for 60–120 min, and the reactions were stopped by adding 4 μl of 0.5M EDTA to 36 μl of the digest and cooling on ice. A colorless Ficoll loading buffer was added to each reaction, and samples were run on an 11.3% acrylamide gel for 60 min at 120 V. Gels were visualized on a Typhoon fluorescent scanner (Amersham/GE).

One-off specificity profile using short A647-labeled DNA target site substrates and purified recombinant aHE15

Purified recombinant aHE15 protein was assayed against a set of 66 “one-off” DNA target substrates, with each substrate containing a single base change away from the most enriched target sequence found in our cleavage screen (“17xSeq”) in order to identify the preferred DNA base at each position of the DNA target. Target site templates were designed based on the centered 22 bp 17xSeq target identified by the cleavage selection screen, synthesized by IDT, and generated as described above. Sixty-seven total reactions were run (17xSeq plus the set of 66 one-off substrates), containing 50 nM purified recombinant aHE15 enzyme, 50 nM A647-labeled DNA target substrate, and cleavage buffer at pH 7.4 (described above) supplemented with 5 mM MgCl₂ in a total volume of 40 μl. The reactions were incubated for 2 h at 37°C degrees and stopped by adding 10 μl of colorless Ficoll loading dye and moving to ice. Eleven microliters of each sample was run on an 11.3% acrylamide gel for 60 min at 120 V. Gels were then visualized on a Typhoon fluorescent scanner (Amersham/GE), and cleavage bands were quantified by densitometry using ImageJ software. Cleavage efficiency was quantified by determining the ratio of “cut”/“uncut” using the density of bands for cut and uncut substrate. This value was normalized relative to the ratio of cut/uncut product for the aHE15 enzyme against the original enriched target sequence (17xSeq). A total of six acrylamide gels were required to visualize samples from all one-off target substrates. A duplicate sample of the 17xSeq cleavage reaction was run on each gel, and the densitometry analysis was performed separately on samples within each individual gel.

Crystallographic analysis of aHE31 and aHE15

The ancestral reconstructions aHE31 and aHE15 were each purified as fully untagged constructs and purified via heparin affinity chromatography followed by size-exclusion chromatography as described above. The proteins were each concentrated to 150 μM concentration and mixed in the presence of 10 mM CaCl₂ with their respective DNA target duplex at 300 μM concentration, to achieve a protein:DNA molar ratio of 1:2. Cocrystallization samples were incubated at room temperature for at least 10 min prior to setting up crystallization drops. Crystals were grown at room temperature using hanging drop vapor diffusion in a 24-well format.

The sequences of target site DNA duplex oligos for each construct are provided below. For aHE31, crystals were grown in the presence of 28% (w/v) PEG 3350, 0.2 M ammonium sulfate, and 0.1 M HEPES pH 7.5. For aHE15, crystals were grown in the presence of 28% (w/v) PEG 3350, 150 mM NaCl, and 200 mM HEPES pH 7.5. Glycerol (25% v/v) was used as a cryoprotectant for both crystals prior to flash cooling in liquid nitrogen.

graphic file with name gkag924ufig1.jpg

Data were collected at the Advanced Light Source synchrotron facility (Berkeley, CA) at beamline 5.0.1 using an X-ray wavelength of 0.97 Å. The dataset was auto-processed with XDS (data integration and scaling) [57], POINTLESS (space group determination) [58], and AIMLESS (final merging) [59]. A signal to noise cutoff of I/sigI > 2.0 was applied (corresponding to an ultimate resolution limit of 2.37 Å), and the Phenix crystallography program suite was used [60] for structure determination. Phasing was performed by molecular replacement with Phaser [61] using an AlphaFold-3 generated model of aHE15 as a search model, subjected to AutoBuild, and then the DNA (and all necessary regions of the protein) was manually built and/or rebuilt with iterative rounds of refinement with phenix refine and rebuilding with Coot [62]. Final structures were deposited and validated by the RCSB PDB (RCSB.org) [63] with the PDB identifiers 12FK (aHE31) and 36OF(aHE15). Data and refinement statistics are provided in Table 1.

Table 1.

Crystallographic Data and Refinement statistics

	aHE31	aHE15
PDB ID	12FK	36OF
*Data collection*
Space group	P 1 21 1	P 21 21 21
Unit cell
a, b, c (Å)	58.45, 47.37.74.74	53.89, 54.04, 164.83
α, β, γ (°)	90.00, 98.65, 90.00	90.00, 90.00, 90.00
Wavelength (Å)	1.00	0.98
Resolution range (Å)	47.37–2.17 (2.24–2.17)	45.19–2.37 (2.46–2.37)
Unique reflections	78 378 (2215)	36 820 (732)
R-merge	0.258 (1.031)	0.089 (0.674)
R-means	0.280 (1.124)	0.093 (0.717)
R-pim	0.109 (0.440)	0.026 (0.232)
CC1/2	0.985 (0.689)	0.999 (0.881)
I/sigma(I)	6.2 (1.9)	19.1 (2.4)
Chi^2	1.05 (0.98)	0.93 (0.59)
Multiplicity	6.6 (6.3)	12.1 (8.4)
Completeness (%)	98.9 (96.9)	92.7 (61.1)
Wilson B-factor	21.72	41.17
*Refinement*
R-work	0.2303	0.2306
R-free	0.2855	0.2785
Number of non-hydrogen atoms	3694	3401
Macromolecules	3391	3336
Ligands	2	2
Water	301	63
Protein residues	298	295
RMS bonds (Å)	0.004	0.004
RMS angle (°)	0.69	0.6
Ramachandran favored (%)	95.61	97.61
Ramachandran allowed (%)	4.39	2.39
Ramachandran outliers (%)	0.00	0.00
Clashscore	5.8	5.2
Average B-factor	25.13	49.67
Macromolecules	24.12	45.55
Ligands	12.25	53.83
Solvent	28.89	41.59

Open in a new tab

Results

Most extant (present-day) LHEs display compromised folding and cleavage activity

We first asked what fraction of monomeric LHEs are capable of being expressed as full-length proteins. Then we then asked how many of the well-expressed proteins display cleavage activity against their predicted DNA target site. We performed these analyses using a dataset of 371 putative LHEs identified via sequence database searches carried out when this project was initiated in 2015. Eight years later, we repeated our database searches, almost doubling the size of our sequence collection of such sequences. We used this expanded set of sequences for phylogenetic analyses and ancestral reconstruction, as described in the second half of “Results” section.

Assessment of expression and folding

We were able to successfully subclone 307 of our dataset of 371 candidate LHEs into a yeast surface display vector. Each of those 307 candidate LHE ORFs was assayed for expression on the surface of yeast using flow cytometry to determine the presence or absence of full-length displayed proteins (Fig. 1a). The folding of protein chains in the eukaryotic endoplasmic reticulum (even those that are not produced via ER transit in their normal physiological context) is subject to rigorous quality control, with misfolded and unstable proteins usually subjected to proteasomal degradation [64]; therefore, the successful expression of a full-length surface-displayed protein is generally consistent with a protein chain that does not suffer from gross misfolding defects [65].

The observed expression behavior of the 307 LHEs was binned into three broad overlapping categories (Fig. 1b): full-length surface expression (staining observed for both N- and C-terminal tags), expression of stalled or truncated constructs (staining observed for only the N-terminal tag), or no surface expression (no staining from either tag). A plot of quantified values for full-length C-terminal staining (Fig. 1c) reveals a wide range of surface expression signal, with the strongest expressers staining more than two orders of magnitude higher than the weakest. We found that 119 clones (39%) displayed strong signal from both tags, indicating expression of full-length protein on the yeast surface. 142 clones (46%) displayed significantly reduced signal from the C-terminal Myc tag, and the remaining 46 clones (15%) showed no measurable signal above background signal from either the N- or the C-terminal tag.

Assessment of cleavage activity

The next question we asked was, for those LHEs that displayed measurable expression on the surface of yeast, what fraction displayed sequence-specific DNA cleavage activity. For that purpose, a total of 188 strongly expressed candidate LHEs (Figs 1c and 2c, and Supplementary Fig. S1) were assayed. To predict a candidate DNA target site for each enzyme, we relied on a well-described aspect of their behavior: because the mobile element inserts into the host gene at the site of cleavage [66], the target sequence can be reconstructed by fusing the host gene sequences that flank the inserted mobile element (i.e. the target sequence corresponds to the intron insertion site in each uninterrupted host gene). The DNA duplexes used in our cleavage assays were all 58 basepairs in length and were centered around the intron insertion site in each corresponding uninterrupted host gene.

We carried out our analysis of cleavage activity using two complementary DNA cleavage assays. In the first approach (Fig. 2a and b), DNA cleavage was measured directly on the surface of yeast using a flow cytometric cleavage assay where a fluorescently labeled, linear DNA target substrate was tethered to the N-terminal HA tag of the displayed enzyme [37]. The second approach corresponded to a more traditional “non-tethered” assay in solution, using the same linear DNA target substrates. For this non-tethered assay, the surface-expressed HEs were released from the yeast surface via treatment with dithiothreitol, and the corresponding DNA target substrate was then added to the supernatant for an in vitro cleavage digest reaction. The digest products were visualized via polyacrylamide gene electrophoresis (Supplementary Fig. S2a).

By employing both of these assays, we could identify active enzymes that might have been missed using only one approach or the other. Those that bind their DNA targets with particularly low affinity may only display cleavage activity when their substrate is physically tethered in proximity to the surface-displayed enzyme. Conversely, those that bind DNA with exceptionally high affinity (and that often do not readily release cleaved DNA products) may only display cleavage activity in a more traditional digest, where bound DNA is physically dissociated from the protein prior to electrophoretic analysis. Altogether, a total of 29 LHEs displayed measurable cleavage activity against their predicted targets. Twenty-four cleaved their DNA targets in both assays, while three (HE07, Gen9-156, and Gen9-171) displayed cleavage activity only in the tethered assay on the yeast surface, and the final two (Gen9-022 and Gen9-051) displayed cleavage activity only in non-tethered in vitro cleavage assays.

The DNA target region for each active enzyme was then subcloned into a plasmid backbone, and each plasmid substrate was subjected to a follow-up in vitro digest with its corresponding enzyme. Run-off Sanger sequencing of the linearized plasmid product was used to determine the precise location of cleavage on each strand of the DNA and thereby identify the center of each target site and the 4-base overhangs generated by each endonuclease. In the end, we were able to determine the exact center of the target and corresponding sites of DNA strand cleavage for 19 of the 29 active enzymes (Supplementary Fig. S2b). For the remaining 10 enzymes, we could not obtain sufficient amounts of digested plasmid and/or clear run-off sequencing data, so their targets are listed in Supplementary Fig. S2c as the full 58 bp DNA substrate sequence (with the site of cleavage expected to occur somewhere within that target region). Raw run-off sequencing data is provided in the Deposited Raw Data.

The results of these experiments indicated that (i) as many as 90% of the LHEs in our analysis displayed significantly compromised folding, expression, and/or cleavage activity (presumed to mostly reflect the accumulation of debilitating mutations, although in some cases an error in the predicted target region sequence would also cause a lack of observed activity), and (ii) the remaining 10% of LHEs are expressed well enough to exhibit site-specific DNA cleavage activity. We generated a multiple sequence alignment and phylogenetic tree from the collection of 371 LHEs (Fig. 3a). The newly identified, active LHEs add significant evolutionary diversity to the active LHE repertoire: the new active LHEs (red) are distributed around the tree, whereas the previously characterized enzymes (black), which were all identified from a search for close homologues to I-OnuI [37], are found in a fairly narrow subclade of the tree.

Ancestral reconstruction experiments generate active constructs with enhanced stability and the potential for novel DNA recognition specificity

Expansion of the LHE sequence collection, phylogenetic analyses, and ancestral reconstructions

We next asked two further questions. First, given that most modern HEs appear to harbor mutations that disrupt folding, stability, and/or activity, can well-established methods for generating ancestral reconstructions produce active enzymes? Second, if ancestral reconstructions do in fact yield active enzymes, might some of them reveal additional details of how DNA recognition specificity changes during their evolutionary diversification? To explore these questions, we decided to focus on two subfamilies of related enzymes, one containing I-PnoMI and two newly validated active enzymes (Gen9-348 and HE08, Fig. 3b), and the other containing the well-studied I-OnuI enzyme [67] and its close homolog I-GpeMI (Fig. 3c).

Before choosing LHE sequences to use in these ancestral reconstructions, we repeated the bioinformatic search first performed in 2015 in order to obtain a denser sampling of extant sequences. Our 2023 search yielded over 300 additional LHE sequences, so that our newer dataset comprised 622 LHE homologues (Supplementary Fig. S3). Using those sequences, we extracted the sequences of each LADLIDADG domain (two per single chain monomeric LHE) and generated a phylogenetic tree (Supplementary Fig. S4). Using that domain-based tree, we identified sequences comprising each subfamily of interest along with selected outgroups (see Materials and methods). Using all full-length subfamily sequences as input to the BaliPhy suite [54], we performed multiple sequence alignment, estimated phylogenies and generate predicted ancestral reconstructions at selected nodes, with manual adjustment at positions containing gaps.

“Shallow” ancestral reconstructions display enhanced expression, increased thermal stability, and retention of activity

The first ancestral reconstruction studied in detail, named aHE31, was generated within the clade containing I-PnoMI, Gen9-348, and HE08 (Fig. 3b). The pairwise amino acid identity between those three extant enzymes ranges from 63% to 73%; the reconstructed aHE31 has 66% to 78% identity to each of them (Fig. 4a and Supplementary Fig. S5a). The DNA target sites for I-PnoMI, Gen9-348, and HE08 differ at 5 of the 22 basepair positions (Fig. 4b).

When all four enzymes are expressed on the surface of yeast, we observed higher expression of the reconstructed ancestor aHE31 than any of its three extant enzyme relatives (Fig. 4c). We then tested the ability of each of the four enzymes to cleave each of their three closely related DNA targets and observed that aHE31 displayed robust cleavage of each (Fig. 4d). In contrast, two of the extant enzymes (I-PnoMI and HE08) display slightly lower cleavage activities against a subset of the three targets, while the third (Gen9-348) displayed cleavage activity against all three targets.

To further examine the behavior and activity of aHE31, we generated purified recombinant aHE31 (Supplementary Fig. S5b) and compared its endonuclease activity to that displayed by recombinant I-PnoM against that enzyme’s DNA target sequence. In cleavage assays conducted over a range of digest times and enzyme concentrations, aHE31 displayed roughly equivalent cleavage rates and yields of product formation as a function of both digest time and decreasing enzyme concentration (Fig. 4e).

We examined the thermal stability of aHE31 by determining the thermal denaturation temperature and unfolding behavior of aHE31 compared to modern-day enzymes I-PnoMI and Gen9-348 (Fig. 4f). A thermal denaturation curve was not obtained for HE08 because that protein could not be adequately expressed and purified from bacteria. The purified recombinant I-PnoMI protein was markedly unstable in the CD buffer, and the resulting melting curve barely allowed for the determination of an estimated T_m value of ∼29°C. The melting curve for Gen9-348 displayed a much better overall shape with cooperative unfolding and a T_m value of 42°C. In contrast, the T_m of aHE31 (60°C) was significantly increased relative to I-PnoMI and Gen9-348 (similar increases in thermal folded stability are often observed for ancestral reconstructions [68] and prove valuable for enzyme and protein engineering studies [69]). The crystal structure of aHE31, bound to the I-PnoMI target site, was determined to 2.17 Å resolution (Table 1 and Supplementary Fig. S5c) and found to be closely superimposable to the previously determined structure of I-PnoMI [70], as well as more distantly related LHE homologues such as I-OnuI [67].

A second “shallow” ancestral reconstruction (aHE11) was also examined, to assess the reproducibility of the observations described above. Similar to aHE31, the aHE11 construct corresponded to a node located a short step backwards in the phylogenetic tree, near a pair of even more closely related extant enzymes (I-OnuI and I-GpeMI) (Fig. 3c). Those two modern day LHEs are 86% identical to each other and recognize DNA target sites that differ at only two positions (Fig. 5a and b). aHE11 is 93% identical to each of those enzymes. Like aHE31, aHE11 displays a higher level of full-length expression on the surface of yeast than its descendants (Fig. 5c), cleaves both the I-OnuI and I-GpeMI targets (Fig. 5d), and displays well-behaved, cooperative unfolding corresponding to a thermal denaturation temperature (T_m = 50°C) that is slightly increased relative to the previously determined denaturation temperature for I-OnuI (T_m = 46°C) [29] (Fig. 5e).

A “deeper” ancestral reconstruction displays robust expression and folding, broad DNA-binding specificity, and a novel DNA cleavage specificity. A third and final ancestral reconstruction, aHE15, resides in the same clade as aHE11 (Fig. 3c), but is a deeper (older) ancestor that connects I-OnuI and I-GpeMI to one of the outgroup sequences used for this clade, an active LHE named I-SmaMI [71]. I-SmaMI has a more highly diverged protein sequence relative to I-OnuI and I-GpeMI and significantly different DNA target specificity (Fig. 5a and b, and Supplementary Fig. S6a). I-OnuI and I-GpeMI do not cleave the target site of I-SmaMI and vice-versa.

aHE15 displayed visibly higher expression on the surface of yeast as compared to I-OnuI and I-GpeMI, similar expression to its shallower ancestral neighbor, aHE11, and significantly higher expression than I-SmaMI (Fig. 5c). Unlike the “shallow” ancestral reconstructions described above, surface-expressed aHE15 did not demonstrate measurable cleavage of any of the DNA targets for I-OnuI, I-GpeMI, or I-SmaMI (Fig. 5d). We, therefore, examined the ability of aHE15 to bind any of those three targets, despite its lack of cleavage activity. We observed that aHE15 bound to all three targets when the DNA substrate was provided at concentrations of 25 nM or higher, with binding to the I-SmaMI target being the strongest (Supplementary Fig. S7).

We expressed and purified recombinant aHE15 (Supplementary Fig. S6b) and determined the thermal denaturation behavior and corresponding T_m values for that construct, two of its modern-day extant relatives, and the shallower ancestor aHE11 (Fig. 5e). The thermal stability of aHE15 was found to be significantly higher (T_m = 78°C) than I-OnuI (T_m = 46°C), aHE11 (T_m = 50°C), and I-SmaMI (T_m = 58°C).

Given that ancestor aHE15 did not display cleavage activity against the DNA targets of its three nearest characterized “relatives” (even though it displayed binding to each), we hypothesized that the DNA cleavage specificity for this deeper ancestral reconstruction (if it was indeed active) might differ significantly from those extant enzymes. We therefore employed an unbiased two-stage selection approach (see Materials and methods and Fig. 6) to determine the cleavage activity and specificity of aHE15. That strategy consisted of (i) four rounds of SELEX DNA target binding selection against a fully randomized 30xN bp DNA target substrate library (Supplementary Fig. S8), followed by (ii) four rounds of a DNA target cleavage selection using a plasmid library containing the target sequences previously isolated via the SELEX binding protocol. The concept of this screen was to (i) reduce the initial complexity of the 4³⁰ DNA library (spanning roughly 1.15 × 10¹⁸, or 2 micromoles, of unique DNA target sequences) to only those sequences that could be bound by the aHE15 enzyme with roughly low micromolar affinities under moderate stringency (300 mM KCl, pH 7.4.), and then (ii) pass on those bound DNA target candidates to a secondary screen to identify DNA targets that the aHE15 enzyme could also cleave.

Cleavable DNA targets were monitored after each round of the cleavage selection by sequencing recovered DNA target sites and looking for enrichment of cleavable sequences. After four rounds of cleavage selection, there were 3 cleavable target sequences that appeared at least twice, with the two most highly enriched sequences present 4 times (“4xSeq”) and 17 times (“17xSeq”) in 29 sequence reads (Supplementary Fig. S9). Purified recombinant aHE15 was used to cleave the target site plasmids for each of these two enriched sequences, and run-off sequencing was performed to determine the precise position of cleavage within each target sequence and then to infer the enzyme’s 22 bp target site within the 30 bp selected sequence (Supplementary Figs S10 and S11). Alignment of the centered 22 bp 4xSeq and 17xSeq sequences with one another revealed that they are identical at 14 of the 22 positions, demonstrating that our cleavage screen was converging on a pool of similar targets (Supplementary Fig. S12). Further, both sequences could be aligned with the target sites of I-OnuI and I-SmaI (Supplementary Fig. S12).

Next, the 22 bp 4xSeq and 17xSeq targets were generated as short linear fluorescently-labeled DNA substrates and used in cleavage digests to determine which sequence was cleaved most efficiently. The more highly enriched target site sequence (17xSeq) was confirmed to be the superior substrate. In the same experiments, we also observed that purified recombinant aHE15 enzyme, at high enzyme concentration and over an extended digest time, displayed very low cleavage activity against the I-OnuI DNA target sequence (that was not previously observed on the surface of yeast in a tethered cleavage assay), while exhibiting no measurable cleavage of the I-SmaMI target sequence (Supplementary Fig. S12).

Determination of the DNA cleavage specificity profile of the aHE15 reconstructed ancestor

Having identified a cleavable DNA substrate sequence for the aHE15 enzyme, we next asked (i) whether the corresponding specificity profile displayed by the enzyme (i.e. its sensitivity toward single basepair substitutions at each position in its target) resembles those of previously examined modern-day LHEs, and (ii) whether the selected target sequence represents an optimal target site for the enzyme (or if one or more single base substitutions would further increase the target’s cleavability). We therefore performed a “one-off” cleavage specificity assay, using the 17xSeq target sequence as the starting point and testing 66 different target substrates corresponding to the systematic incorporation of each of three alternative bases at each of the target’s 22 individual positions.

The resulting specificity profile for aHE15 (Fig. 7a) qualitatively resembles those previously established for both I-OnuI and I-SmaMI [37, 67], including strong sensitivity toward individual base pair substitutions (resulting in significantly reduced cleavage) across the central four basepairs of the target and many positions extending outwards, contrasted by reduced fidelity at the outer-most two or three base pair positions in each DNA half-site. At least four individual basepair substitutions (-4C to A, +4A to T, +4A to C, and + 10T to A) increased cleavage efficiency relative to the original 17xSeq target sequence. We therefore tested an additional seven target sequences that included all possible combinations of each of the single basepair substitutions (Fig. 7b and Supplementary Fig. S13a,b). That secondary set of assays indicated that the most efficiently cleaved DNA target sequence for aHE15 corresponded to 5′ TTTACCGATATACACCCCTAAG 3′. That target site shares 54% identity with the target of I-OnuI (12 out of 22 base pairs) and 27% identity with the target of I-SmaMI (6 out of 22 base pairs) (Fig. 7c).

Structural features of DNA recognition and its divergence

Having determined a DNA target sequence that was near-optimal for the reconstructed ancestral LHE construct aHE15, we cocrystallized the protein bound to its DNA target site and solved the structure to 2.37 Å resolution. The final refined model (Fig. 8a) revealed a single copy of the aHE15 construct bound in an unambiguous orientation to a single copy of its DNA target site, with the central four basepairs and corresponding scissile phosphate groups on each strand centered within the enzyme’s endonuclease actives sites as previously observed in the DNA-bound structures of its two previously visualized “descendants,” I-OnuI and I-SmaMI [37, 67].

Whereas the conformation of the aHE15 protein was closely superimposable to DNA-bound structures of both I-OnuI and I-SmaMI (Fig. 8b; pairwise superposition RMSD values against either of those enzymes, calculated for all comparable α-carbons, were both approximately 0.9 Å), the corresponding superimposed DNA targets displayed more significant local differences in their conformations (pairwise superposition RMSD values both approximately 2 Å), with much of the difference in DNA conformation attributable to differences in major groove dimensions along the length of the respective DNA targets.

The identity and details of individual DNA contacts formed by I-OnuI, I-SmaMI, and their common reconstructed ancestor aHE15 display a mixture of conservation and divergence from one another (Figs 8c and 9). All three enzymes rely on approximately ten residues, largely located at comparable positions in their aligned structures, to form direct side chain contacts with individual DNA bases. Those contacts are augmented by water-mediated contacts between additional protein side chains and individual DNA bases and further sequence-nonspecific contacts to the DNA backbone, collectively involving roughly an additional 20 residues in DNA contacts and binding. Those roughly 30 DNA-contacting residues in each structure are in contact with an additional 20 neighboring residues that provide structural context that reinforces the overall shape and stability of the DNA-contacting surface of the protein.

The identity of some, but not all, of the individual residues that are in contact with DNA bases in each structure are conserved—for example, lysines K225 and K227 in I-OnuI versus lysines K218 and K220 in aHE15 (Fig. 8c) or arginines R28 and R30 in I-OnuI versus arginines R26 and R28 in I-SmaMI. Although conserved, the sidechains of those residues often display differing rotameric conformations that place them in direct contact with chemically identical but differently located base pairs within each enzyme’s target site. For example, K225 and K227 in I-OnuI form contacts to guanine bases at positions + 5 and + 6 in that enzyme’s target, whereas comparable lysine residues K218 and K220 in aHE15 make similar contacts to guanine bases at positions + 4 and + 5 in its target (Figs 8c and 9). Conversely, whereas R28 and R30 in I-OnuI (and R26 and R28 in I-SmaMI) make similar direct contacts to guanine bases located at positions −7 and −8 in their respective targets, aHE15 utilizes arginines at two nearby but different positions (R21 and R35) to form similar direct contacts to guanine bases at positions −6 and −7 in its own DNA target. Such indications of structural malleability in contacts between protein side chains and DNA bases have been previously observed for various related LHEs [31, 37, 72].

Our observations of malleability in the LHE–DNA target contact map between a present-day LHE and its predicted ancestor reinforce previous studies that found similar divergence in protein–DNA contacts between naturally evolved LHE homologues [37] and between different laboratory-engineered LHE–DNA complexes [31].

Discussion

Homing endonuclease diversity and evolutionary degradation

Most MGEs provide little or no benefit to their host and are under minimal selection pressure to maintain their form and function, thereby leading to the accumulation of debilitating mutations and eventual inactivation [73]. For example, LINE-1 retrotransposons represent 17% of the human genome, but fewer than 1% are active, due to truncations, internal rearrangements, and point mutations [74]. LINE-1s have even become totally “extinct” in some mammalian lineages [75]. Similarly, the human genome harbors many fossil DNA transposon sequences, but none are now active (having become extinct over the past ∼40 to 50 million years) [2].

The presence, action, and evolution of MGEs extends to (and likely originated in) the microbial universe; the life cycle and distribution of their activities and dysfunctions have been examined in that context as well. Early studies of LHE life cycles, beginning with an analysis of the yeast I-SceI HE lineage (and their corresponding mobilized introns) indicated that they undergo an evolutionary life cycle of invasion, mutational degradation, loss from the genome, and subsequent re-invasion by a related element, with an estimated periodicity of approximately 2 million years [32]. A similar study of the I-OnuI HE examined a group of closely related homologues from ascomycete fungi and concluded that over half appeared to harbor inactivating genetic truncations and mutations [76]. Yet another analysis of a small collection of mobile introns that reside in rDNA host genes, and that are mobilized by members of the “His-Cys box” HE family, arrived at similar numbers and conclusions [77].

Those prior studies of LHEs relied on approaches in which a relatively small number of homologues were characterized, in each case closely related to least one highly active HE. To more broadly examine how frequently such elements retain both form and function, we applied a strategy in which a much larger, unbiased collection of LHEs was identified based on simple criteria that (i) they are encoded within group I introns (that often act as a surrounding vehicle for HE reading frames), and (ii) that they appear to be members of the single-chain (monomeric) LAGLIDADG homing endonuclease (LHE) family. We cloned 307 candidate enzymes, assayed each for protein expression and folding, and further assayed the well-expressed subset for DNA cleavage activity against their predicted target sequences.

We found that roughly 90% of these 307 HEGs display significant deficiencies in their expression, folding, stability, and/or DNA cleavage activity. Furthermore, while expression of full-length protein is obviously a necessary prerequisite for cleavage activity, a significant fraction of the HEGs that displayed robust expression and folding nonetheless displayed little or no endonuclease function (Fig. 2c). The 29 newly identified active endonucleases are distributed throughout the LHE phylogenetic tree, in contrast to previously identified active LHEs which are found in a more restricted subclade (Fig. 3a and Supplementary Fig. S3). Each new active enzyme cleaves a DNA target corresponding to the mobile intron insertion site within its host gene. This expanded range of DNA target sequences might provide an excellent starting point for engineering enzymes with new, useful specificities.

We note that the estimate of the fraction of putative LHEs that appear to display significant decay in their form and function (roughly nine out of ten) may be slightly inflated for several reasons. First, the conditions corresponding to in vitro expression, folding and activity via yeast surface display is quite different from conditions within the environment of a eukaryotic organelle. Second, we did not account for repeat-content or mRNA secondary-structural features that might influence the efficiency of protein translation. Finally, it is possible that the putative DNA target region, corresponding to fused exon sequences in the host gene, may be incorrectly assigned for some constructs (e.g. if the sequence of the original recognition site for a given LHE, at the time of the initial genome invasion event, has subsequently drifted evolutionarily).

Ancestral reconstructions yield LHEs with robust expression and cleavage activity

ASR methods infer the most probable sequences of extinct ancestral molecules, by combining alignments of extant homologues with phylogenetic models of sequence evolution [78]. Modern ASR methods typically use likelihood- or Bayesian-based approaches to estimate the most likely ancestral states at internal nodes of an evolutionary tree, generating hypotheses about historical protein or nucleic acid sequences that can subsequently be evaluated computationally or experimentally. Reconstruction accuracy and confidence depend on dense taxonomic sampling, high-quality multiple sequence alignments, accurate phylogenetic models, and the existence of sufficiently conserved homologous families. ASR therefore performs best when proteins being used for that purpose have evolved under functional constraint (purifying selection) because phylogenetic signal is preserved and historical states can thereby be reliably inferred [79].

In this context, successful ancestral reconstruction of mobile elements might have been expected to prove difficult. Unlike many protein families that evolve under sustained functional constraint, selfish mobile elements experience episodic bursts of positive selection during invasion and spread, followed by prolonged periods of neutral evolution after their fixation. As a consequence, many mobile elements accumulate substitutions, indels and truncations that reduce expression, stability or catalytic activity, and ultimately become inactive prior to elimination from the host genome [17]. These mutations might erode the phylogenetic signal required for accurate ancestral inference.

Despite such reservations, we noted that ASRs have previously been applied to MGEs to predict and experimentally evaluate ancestral states using collections of extant, often degraded, sequences (e.g. [80–82]). Conceptually, many ancestral nodes in a mobile element phylogeny represent transposition events that led to new element insertions, and thus by definition represent active enzymes (while other nodes may represent copies derived by speciation events, i.e. divergence of host genomes rather than of active mobile element copies). Ancestral reconstructions therefore (if accurate) should be enriched for active enzymes when compared to extant sequences. Collectively, those studies illustrate the potential of ancestral reconstructions of such elements: resurrection of ancestral molecules can reveal biochemical activities no longer represented among extant sequences, but reconstruction fidelity ultimately depends on retaining sufficient phylogenetic signal despite subsequent sequence degeneration.

In this study, we apply ancestral reconstruction to the LHE family of mobile elements, performing detailed functional characterization of one “deep” and two “shallow” ancestral models. All three enzymes are strongly expressed and fold well. This result reflects a well-known outcome common to many flavors of ancestral reconstruction efforts and that is likely a by-product of biases inherent in such calculations toward “return to consensus” mutations within the protein scaffold [68]. They also exhibit robust DNA target site cleavage activity, with one of the three ancestral reconstructions targeting and cleaving a novel DNA sequence not cleaved by its descendent enzymes. This successful “resurrection” indicates that sufficient evolutionary information remains to recover coherent ancestral states, indicating that ancestral reconstruction can provide an effective strategy for recovering latent functional diversity from degraded populations of selfish genetic elements, offering access to evolutionary intermediates and activities that have been lost during the natural life cycle of mobile endonucleases.

However, it is worth noting that while the reconstruction effort leading to the “deeper” aHE15 model was successful in generating a functional enzyme that displays a DNA recognition and cleavage specificity profile that resembles the features of such profiles for modern-day extant LHEs, that its enzymatic power (as measured by its rate of DNA cleavage) as compared to its “descendants” is slightly lower than its modern-day relatives, both on the surface of yeast and in assays using purified recombinant protein. One possibility is that the features we would consider to be optimal for engineered enzymes may not be truly optimal features for the evolutionary success of a mobile element. Maximal expression and cleavage activity may result in mobile elements that are “too efficient,” harming their hosts through excessive insertions. Lower levels of activity may be more evolutionarily successful, allowing the element to create new copies at appreciable levels without population-level damage to the host.

As well, technical considerations, including the possibility that the identification of the optimal target of aHE15 may be incomplete, may mean we have not reconstructed a true ancestor or the most accurate possible model of an ancestral reconstruction and its corresponding target site. While aHE15 clearly possesses the necessary components of a LAGLIDADG endonuclease active site, and displays DNA-binding affinity and specificity that also approximates the behavior of current day enzymes, it may be that subtle features of structure and dynamics that are thought to be critical for optimal enzymatic function are not fully recapitulated in the computations leading to the final version of such a reconstructed enzyme. Each ancestral sequence we tested is only one of many possible solutions to the true sequence at that node of the phylogeny. While at each protein position, the algorithm chooses the most likely amino acid, very few positions are predicted unambiguously, and the combinatorial uncertainty over the entire protein means that the single full-length sequence prediction we tested is highly unlikely to exactly match the true ancestral sequence. Furthermore, the deeper the node being predicted, the greater the uncertainty. It is also possible that we have not identified an optimal DNA target sequence for the enzyme, and that deeper exploration of DNA sequence space around the current “best” substrate might yield further improvements in endonuclease function.

Finally, this work lays the groundwork for future uses of ancestral reconstruction to aid in LHE engineering and to understand LHE biology. While the use of LHEs (as well as TAL effector nucleases and zinc finger nucleases) for potential gene therapy and modification has significantly diminished with the advent of CRISPR-based technologies, re-engineered LHEs are still created and used for a variety of such applications, largely due to their small size and robust activity. Two recent examples are described in [83, 84].

For example, one could reconstruct stepwise series of ancestral nodes in order to examine how these enzymes evolve to change their DNA targeting specificity, and how they acquire the ability to invade novel host genomes. Understanding the molecular patterns that govern enzyme–target recognition could help engineer enzymes with novel desired target sites. While our current work reconstructs ancestors of active descendant enzymes, we also anticipate that ancestral reconstructions would also succeed in clades that contain only “dead” enzymes, potentially increasing the sequence space of active enzymes even further. Furthermore, ancestral reconstructions of such elements can yield enzymes with new DNA targeting specificities, as well as highly expressed, thermostabilized predecessors of compromised modern-day enzymes. Given the size of the current collection of LHEs represented in this study (both modern-day enzymes at the tips of the branches of the phylogenetic tree, along with the underlying nodes that connect every branch point in that tree), it seems possible that this system, like many others of similar diversity and number, may offer considerable potential for the future computational design of a wide range of compact gene-targeting proteins with similarly diverse DNA sequence targets, as well as the potential to readily re-activate and characterize the specificity of many hundreds of such enzymes.

Supplementary Material

gkag924_Supplemental_Files

gkag924_supplemental_files.zip^{(4.7MB, zip)}

Acknowledgements

We thank Professor Harmit Malik at the Fred Hutchinson Cancer Center for advice and support, and members of his group, as well as those of the Brett Kaiser research group at Seattle University and the Stoddard group at Fred Hutchinson Cancer Center for additional advice and discussion, particularly Alex J. Kaiser for assistance with subcloning ancestral reconstructions and assistance with protein expression and purification.

Author contributions: Juliana Young (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Investigation [equal], Methodology [equal], Validation [equal], Visualization [equal], Writing – original draft [supporting]), Abigail R. Lambert (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Investigation [equal], Methodology [equal], Validation [equal], Visualization [equal], Writing – original draft [supporting]), Janet M. Young (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Investigation [equal], Methodology [equal], Validation [equal], Visualization [equal], Writing – original draft [supporting]), Lindsey Doyle (Formal analysis [supporting], Investigation [supporting], Validation [supporting], Visualization [supporting], Writing – review & editing [supporting]), Miriam Silverstein (Formal analysis [supporting], Investigation [supporting], Visualization [supporting], Writing – review & editing [supporting]), David R. Edgell (Conceptualization [equal], Formal analysis [equal], Investigation [equal], Methodology [equal], Writing – review & editing [supporting]), and Barry L. Stoddard (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Funding acquisition [equal], Investigation [equal], Methodology [equal], Project administration [equal], Supervision [equal], Validation [equal], Visualization [equal], Writing – original draft [equal])

Contributor Information

Juliana C Young, Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States.

Abigail R Lambert, Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States.

Janet M Young, Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States.

Lindsey A Doyle, Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States.

Miriam Silverstein, Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States.

David R Edgell, Department of Biochemistry, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON N6A 5C1, Canada.

Barry L Stoddard, Division of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. North, Seattle, WA 98109, United States.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

B.L.S. holds the position of Senior Executive Editor for Nucleic Acids Research and has not peer-reviewed or made any editorial decisions for this paper, nor had any access to any aspect of the peer review process beyond receiving anonymized reviewer comments. B.L.S. holds patents and patent applications related to the engineering of LHEs for various biotech applications, and has received past income distributions from that intellectual property.

Funding

Funding support was provided for this project by bluebird bio inc., the Fred Hutchinson Cancer Center, and NIGMS grant R35 GM148166 to B.L.S., NIGMS grant R01 GM074108 to J.M.Y. (PI: Harmit S. Malik), NCI T32 CA080416 to J.C.Y., and the NSERC Discover Grant RGPIN-2015-04800 and RGPIN-2022-05459 to D.R.E. Crystallographic data was collected at the Fred Hutchinson Cancer Center home source (supported in part by a shared instrumentation grant, NIH S10OD028581) and the Advanced Light Source (ALS) synchrotron facility. The ALS-ENABLE beamlines are supported in part by the National Institutes of Health,National Institute of General Medical Sciences, grant P30 GM124169-01. The Advanced Light Source is a Department of Energy Office of Science User Facility under Contract No. DE-AC02-05CH11231. Funding to pay the Open Access publication charges for this article was provided by NIH R35 GM148166.

Data availability

The crystallographic coordinates and diffraction data corresponding to the structure of aHE31 and aHE15 have been deposited in the protein database (www.rcsb.org) and are available via PDB ID codes: 12FK (https://doi.org/10.2210/pdb12FK/pdb) and 36OF https://doi.org/10.2210/pdb36OF/pdb). Raw data (sequencing output, gels, and flow cytometric data) have been deposited and made available for public accession at the Harvard Datavase Depository (https://doi.org/10.7910/DVN/Z4WCBW).

References

1. Kazazian HH. Mobile elements: drivers of genome evolution. Science. 2004;303:1626–32. 10.1126/science.1089670 [DOI] [PubMed] [Google Scholar]
2. Feschotte C, Pritham EJ. DNA transposons and the evolution of eukaryotic genomes. Annu Rev Genet. 2007;41:331–68. 10.1146/annurev.genet.40.110405.090448 [DOI] [PMC free article] [PubMed] [Google Scholar]
3. Mita P, Boeke JD. How retrotransposons shape genome regulation. Curr Opin Genet Dev. 2016;37:90–100. 10.1016/j.gde.2016.01.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
4. Smillie C, Garcillán-Barcia MP, Francia MV et al. Mobility of plasmids. Microbiol Mol Biol Rev. 2010;74:434–52. 10.1128/MMBR.00020-10 [DOI] [PMC free article] [PubMed] [Google Scholar]
5. Belfort M. Mobile self-splicing introns and inteins as environmental sensors. Curr Opin Microbiol. 2017;38:51–8. 10.1016/j.mib.2017.04.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Johnson CM, Grossman AD. Integrative and conjugative elements (ICEs): what they do and how they work. Annu Rev Genet. 2015;49:577–601. 10.1146/annurev-genet-112414-055018 [DOI] [PMC free article] [PubMed] [Google Scholar]
7. Pfeifer E, Sousa JM, Touchon M et al. When bacteria are phage playgrounds: interactions between viruses, cells, and mobile genetic elements. Curr Opin Microbiol. 2022;70:102230. 10.1016/j.mib.2022.102230 [DOI] [PubMed] [Google Scholar]
8. Cambray G, Guerout AM, Mazel D. Integrons. Annu Rev Genet. 2010;44:141–66. 10.1146/annurev-genet-102209-163504 [DOI] [PubMed] [Google Scholar]
9. Tenjo-Castaño F, Rout SS, Dey S et al. Unlocking the potential of CRISPR-associated transposons: from structural to functional insights. Trends Genet. 2025;41:660–77. 10.1016/j.tig.2025.04.005 [DOI] [PubMed] [Google Scholar]
10. Haudiquet M, de Sousa JM, Touchon M et al. Selfish, promiscuous and sometimes useful: how mobile genetic elements drive horizontal gene transfer in microbial populations. Philos Trans R Soc Lond B Biol Sci. 2022;377:20210234. 10.1098/rstb.2021.0234 [DOI] [PMC free article] [PubMed] [Google Scholar]
11. Horne T, Orr VT, Hall JP. How do interactions between mobile genetic elements affect horizontal gene transfer?. Curr Opin Microbiol. 2023;73:102282. 10.1016/j.mib.2023.102282 [DOI] [PubMed] [Google Scholar]
12. Tokuda M, Shintani M. Microbial evolution through horizontal gene transfer by mobile genetic elements. Microb Biotechnol. 2024;17:e14408. 10.1111/1751-7915.14408 [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Chuong EB, Elde NC, Feschotte C. Regulatory activities of transposable elements: from conflicts to benefits. Nat Rev Genet. 2017;18:71–86. 10.1038/nrg.2016.139 [DOI] [PMC free article] [PubMed] [Google Scholar]
14. Pritham EJ. Transposable elements and factors influencing their success in eukaryotes. J Hered. 2009;100:648–55. 10.1093/jhered/esp065 [DOI] [PMC free article] [PubMed] [Google Scholar]
15. Wells JN, Feschotte C. A field guide to eukaryotic transposable elements. Annu Rev Genet. 2020;54:539–61. 10.1146/annurev-genet-040620-022145 [DOI] [PMC free article] [PubMed] [Google Scholar]
16. Song MJ, Schaack S. Evolutionary conflict between mobile DNA and host genomes. Am Nat. 2018;192:263–73. 10.1086/698482 [DOI] [PubMed] [Google Scholar]
17. Stoddard BL. Homing endonucleases from mobile group I introns: discovery to genome engineering. Mob DNA. 2014;5:7. 10.1186/1759-8753-5-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
18. Jacquier A, Dujon B. An intron-encoded protein is active in a gene conversion process that spreads an intron into a mitochondrial gene. Cell. 1985;41:383–94. 10.1016/S0092-8674(85)80011-8 [DOI] [PubMed] [Google Scholar]
19. Macreadie IG, Scott RM, Zinn AR et al. Transposition of an intron in yeast mitochondria requires a protein encoded by that intron. Cell. 1985;41:395–402. 10.1016/S0092-8674(85)80012-X [DOI] [PubMed] [Google Scholar]
20. Zinn AR, Butow RA. Nonreciprocal exchange between alleles of the yeast mitochondrial 21S rRNA gene: kinetics and the involvement of a double-strand break. Cell. 1985;40:887–95. 10.1016/0092-8674(85)90348-4 [DOI] [PubMed] [Google Scholar]
21. Edgell DR, Belfort M, Shub DA. Barriers to intron promiscuity in bacteria. J Bacteriol. 2000;182:5281–9. 10.1128/JB.182.19.5281-5289.2000 [DOI] [PMC free article] [PubMed] [Google Scholar]
22. Michel F, Dujon B. Genetic exchanges between bacteriophage T4 and filamentous fungi. Cell. 1986;46:323–30. 10.1016/0092-8674(86)90651-3 [DOI] [PubMed] [Google Scholar]
23. Shub DA, Gott JM, Xu MQ et al. Structural conservation among three homologous introns of bacteriophage T4 and the group I introns of eukaryotes. Proc Natl Acad Sci USA. 1988;85:1151–5. 10.1073/pnas.85.4.1151 [DOI] [PMC free article] [PubMed] [Google Scholar]
24. Goodrich-Blair H, Shub DA. Beyond homing: competition between intron endonucleases confers a selective advantage on flanking genetic markers. Cell. 1996;84:211–21. 10.1016/S0092-8674(00)80976-9 [DOI] [PubMed] [Google Scholar]
25. Turmel M, Cote V, Otis C et al. Evolutionary transfer of ORF-containing group I introns between different subcellular compartments (chloroplast and mitochondrion). Mol Biol Evol. 1995;12:533–45. [DOI] [PubMed] [Google Scholar]
26. Edgell DR, Gibb EA, Belfort M. Mobile DNA elements in T4 and related phages: a review in the series on bacteriophage T4 and its relatives. Virol J. 2010;7:290–300. 10.1186/1743-422X-7-290 [DOI] [PMC free article] [PubMed] [Google Scholar]
27. Turmel M, Boulanger J, Schnare MN et al. Six group I introns and three internal transcribed spacers in the chloroplast large subunit ribosomal RNA gene of the green alga Chlamydomonas eugametos. J Mol Biol. 1991;218:293–311. 10.1016/0022-2836(91)90713-G [DOI] [PubMed] [Google Scholar]
28. Turmel M, Gutell RR, Mercier JP et al. Analysis of the chloroplast large subunit ribosomal RNA gene from 17 Chlamydomonas taxa. Three internal transcribed spacers and 12 group I intron insertion sites. J Mol Biol. 1993;232:446–67. 10.1006/jmbi.1993.1402 [DOI] [PubMed] [Google Scholar]
29. Lambert AR, Hallinan JP, Werther R et al. Optimization of protein thermostability and exploitation of recognition behavior to engineer altered protein–DNA recognition. Structure. 2020;28:760–775. 10.1016/j.str.2020.04.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
30. Stoddard BL. Homing endonucleases: from microbial genetic invaders to reagents for targeted DNA modification. Structure. 2011;19:7–15. 10.1016/j.str.2010.12.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
31. Werther R, Hallinan JP, Lambert ARea. Crystallographic analyses illustrate significant plasticity and efficient recoding of meganuclease target specificity. Nucleic Acids Res. 2017;45:8621–34. 10.1093/nar/gkx544 [DOI] [PMC free article] [PubMed] [Google Scholar]
32. Burt A, Koufopanou V. Homing endonuclease genes: the rise and fall and rise again of a selfish element. Curr Opin Genet Dev. 2004;14:609–15. 10.1016/j.gde.2004.09.010 [DOI] [PubMed] [Google Scholar]
33. Koufopanou V, Burt A. Degeneration and domestication of a selfish gene in yeast: molecular evolution versus site-directed mutagenesis. Mol Biol Evol. 2005;22:1535–8. 10.1093/molbev/msi149 [DOI] [PubMed] [Google Scholar]
34. Takeuchi R, Certo M, Caprara MG et al. Optimization of in vivo activity of a bifunctional homing endonuclease and maturase reverses evolutionary degradation. Nucleic Acids Res. 2009;37:877–90. 10.1093/nar/gkn1007 [DOI] [PMC free article] [PubMed] [Google Scholar]
35. Scalley-Kim M, McConnell-Smith A, Stoddard BL. Coevolution of a homing endonuclease and its host target sequence. J Mol Biol. 2007;372:1305–19. 10.1016/j.jmb.2007.07.052 [DOI] [PMC free article] [PubMed] [Google Scholar]
36. Ho Y, Kim SJ, Waring RB. A protein encoded by a group I intron in Aspergillus nidulans directly assists RNA splicing and is a DNA endonuclease. Proc Natl Acad Sci USA. 1997;94:8994–9. 10.1073/pnas.94.17.8994 [DOI] [PMC free article] [PubMed] [Google Scholar]
37. Lambert AR, Hallinan JP, Shen BWea. Indirect DNA sequence recognition and its impact on nuclease cleavage activity. Structure. 2016;24:862–73. 10.1016/j.str.2016.03.024 [DOI] [PMC free article] [PubMed] [Google Scholar]
38. Sayers EW, Cavanaugh M, Frisse L et al. GenBank 2025 update. Nucleic Acids Res. 2025;53:D56–61. 10.1093/nar/gkae1114 [DOI] [PMC free article] [PubMed] [Google Scholar]
39. Altschul SF, Gish W, Miller W et al. Basic local alignment search tool. J Mol Biol. 1990;215:403–10. 10.1016/S0022-2836(05)80360-2 [DOI] [PubMed] [Google Scholar]
40. Altschul SF, Madden TL, Schaffer AA et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997;25:3389–402. 10.1093/nar/25.17.3389 [DOI] [PMC free article] [PubMed] [Google Scholar]
41. Kosuri S, Eroshenko N, Leproust EM et al. Scalable gene synthesis by selective amplification of DNA pools from high-fidelity microchips. Nat Biotechnol. 2010;28:1295–9. 10.1038/nbt.1716 [DOI] [PMC free article] [PubMed] [Google Scholar]
42. Jarjour J, West-Foyle H, Certo MT et al. High-resolution profiling of homing endonuclease binding and catalytic specificity using yeast surface display. Nucleic Acids Res. 2009;37:6871–80. 10.1093/nar/gkp726 [DOI] [PMC free article] [PubMed] [Google Scholar]
43. Gietz RD, Schiestl RH. Large-scale high-efficiency yeast transformation using the LiAc/SS carrier DNA/PEG method. Nat Protoc. 2007;2:38–41. 10.1038/nprot.2007.15 [DOI] [PubMed] [Google Scholar]
44. Becton DaC. 10 ed, Ashland OR. 2023;
45. Baxter SK, Lambert AR, Scharenberg AM et al. Flow cytometric assays for interrogating LAGLIDADG homing endonuclease DNA-binding and cleavage properties. Methods Mol Biol. 2013;978:45–61. 10.1007/978-1-62703-293-3 10.1007/978-1-62703-293-3_4 [DOI] [PMC free article] [PubMed] [Google Scholar]
46. Chao G, Lau WL, Hackel BJ et al. Isolating and engineering human antibodies using yeast surface display. Nat Protoc. 2006;1:755–68. 10.1038/nprot.2006.94 [DOI] [PubMed] [Google Scholar]
47. Mistry J, Chuguransky S, Williams Lea. Pfam: the protein families database in 2021. Nucleic Acids Res. 2021;49:D412–9. 10.1093/nar/gkaa913 [DOI] [PMC free article] [PubMed] [Google Scholar]
48. Darriba D, Taboada GL, Doallo R et al. ProtTest 3: fast selection of best-fit models of protein evolution. Bioinformatics. 2011;27:1164–5. 10.1093/bioinformatics/btr088 [DOI] [PMC free article] [PubMed] [Google Scholar]
49. Guindon S, Dufayard JF, Lefort V et al. New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. Syst Biol. 2010;59:307–21. 10.1093/sysbio/syq010 [DOI] [PubMed] [Google Scholar]
50. R Core Team . R: A Language and Environment for Statistical Computing, Vienna, Austria: R Foundation for Statistical Computing; 2025.; [Google Scholar]
51. Paradis E, Schliep K. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics. 2019;35:526–8. 10.1093/bioinformatics/bty633 [DOI] [PubMed] [Google Scholar]
52. Yu G. Using ggtree to visualize data on tree-like structures. Curr Protoc Bioinformatics. 2020;69:e96. 10.1002/cpbi.96 [DOI] [PubMed] [Google Scholar]
53. Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013;30:772–80. 10.1093/molbev/mst010 [DOI] [PMC free article] [PubMed] [Google Scholar]
54. Redelings BD. BAli-Phy version 3: model-based co-estimation of alignment and phylogeny. Bioinformatics. 2021;37:3032–4. 10.1093/bioinformatics/btab129 [DOI] [PubMed] [Google Scholar]
55. Jacoby K, Scharenberg AM. Homing endonuclease target determination using SELEX adapted for yeast surface display. Methods Mol Biol. 2014;1123:165–90. 10.1007/978-1-62703-968-0 10.1007/978-1-62703-968-0_13 [DOI] [PubMed] [Google Scholar]
56. Jacoby K, Lambert AR, Scharenberg AM. Characterization of homing endonuclease binding and cleavage specificities using yeast surface display SELEX (YSD-SELEX). Nucleic Acids Res. 2017;45:e11. [DOI] [PMC free article] [PubMed] [Google Scholar]
57. Kabsch W. Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr D Biol Crystallogr. 2010;66:133–44. 10.1107/S0907444909047374 [DOI] [PMC free article] [PubMed] [Google Scholar]
58. Evans P. Scaling and assessment of data quality. Acta Crystallogr D Biol Crystallogr. 2006;62:72–82. 10.1107/S0907444905036693 [DOI] [PubMed] [Google Scholar]
59. Evans PR, Murshudov GN. How good are my data and what is the resolution?. Acta Crystallogr D Biol Crystallogr. 2013;69:1204–14. 10.1107/S0907444913000061 [DOI] [PMC free article] [PubMed] [Google Scholar]
60. Adams PD, Afonine PV, Bunkoczi Gea. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010;66:213–21. 10.1107/S0907444909052925 [DOI] [PMC free article] [PubMed] [Google Scholar]
61. McCoy AJ, Grosse-Kunstleve RW, Adams PD et al. Phaser crystallographic software. J Appl Crystallogr. 2007;40:658–74. 10.1107/S0021889807021206 [DOI] [PMC free article] [PubMed] [Google Scholar]
62. Emsley P, Lohkamp B, Scott WG et al. Features and development of Coot. Acta Crystallogr D Biol Crystallogr. 2010;66:486–501. 10.1107/S0907444910007493 [DOI] [PMC free article] [PubMed] [Google Scholar]
63. Berman HM, Westbrook J, Feng Z et al. The Protein Data Bank. Nucleic Acids Res. 2000;28:235–42. 10.1093/nar/28.1.235 [DOI] [PMC free article] [PubMed] [Google Scholar]
64. Sun Z, Brodsky JL. Protein quality control in the secretory pathway. J Cell Biol. 2019;218:3171–87. 10.1083/jcb.201906047 [DOI] [PMC free article] [PubMed] [Google Scholar]
65. Shusta EV, Kieke MC, Parke E et al. Yeast polypeptide fusion surface display levels predict thermal stability and soluble secretion efficiency. J Mol Biol. 1999;292:949–56. 10.1006/jmbi.1999.3130 [DOI] [PubMed] [Google Scholar]
66. Bonocora RP, Shub DA. A likely pathway for formation of mobile group I introns. Curr Biol. 2009;19:223–8. 10.1016/j.cub.2009.01.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
67. Takeuchi R, Lambert AR, Mak AN et al. Tapping natural reservoirs of homing endonucleases for targeted gene modification. Proc Natl Acad Sci USA. 2011;108:13077–82. 10.1073/pnas.1107719108 [DOI] [PMC free article] [PubMed] [Google Scholar]
68. Wheeler LC, Lim SA, Marqusee S et al. The thermostability and specificity of ancient proteins. Curr Opin Struct Biol. 2016;38:37–43. 10.1016/j.sbi.2016.05.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
69. Thomson RES, Carrera-Pacheco SE, Gillam EMJ. Engineering functional thermostable proteins using ancestral sequence reconstruction. J Biol Chem. 2022;298:102435. 10.1016/j.jbc.2022.102435 [DOI] [PMC free article] [PubMed] [Google Scholar]
70. Esler MA, Werther R, Doyle LA et al. Caveat emptor: predicting and modeling protein–DNA recognition and binding via machine-learning computational approaches. Nucleic Acids Res. 2026;54. 10.1093/nar/gkag608 [DOI] [PMC free article] [PubMed] [Google Scholar]
71. Shen BW, Lambert A, Walker BC et al. The structural basis of asymmetry in DNA binding and cleavage as exhibited by the I-SmaMI LAGLIDADG meganuclease. J Mol Biol. 2016;428:206–20. 10.1016/j.jmb.2015.12.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
72. Chevalier B, Turmel M, Lemieux C et al. Flexible DNA target site recognition by divergent homing endonuclease isoschizomers I-CreI and I-MsoI. J Mol Biol. 2003;329:253–69. 10.1016/S0022-2836(03)00447-9 [DOI] [PubMed] [Google Scholar]
73. Blumenstiel JP. Birth, school, work, death, and resurrection: the life stages and dynamics of transposable element proliferation. Genes. 2019;10:336. 10.3390/genes10050336 [DOI] [PMC free article] [PubMed] [Google Scholar]
74. Brouha B, Schustak J, Badge RM et al. Hot L1s account for the bulk of retrotransposition in the human population. Proc Natl Acad Sci USA. 2003;100:5280–5. 10.1073/pnas.0831042100 [DOI] [PMC free article] [PubMed] [Google Scholar]
75. Grahn RA, Rinehart TA, Cantrell MA et al. Extinction of LINE-1 activity coincident with a major mammalian radiation in rodents. Cytogenet Genome Res. 2005;110:407–15. 10.1159/000084973 [DOI] [PubMed] [Google Scholar]
76. Sethuraman J, Majer A, Friedrich NC et al. Genes within genes: multiple LAGLIDADG homing endonucleases target the ribosomal protein S3 gene encoded within an rnl group I intron of Ophiostoma and related taxa. Mol Biol Evol. 2009;26:2299–315. 10.1093/molbev/msp145 [DOI] [PubMed] [Google Scholar]
77. Haugen P, Reeb V, Lutzoni F et al. The evolution of homing endonuclease genes and group I introns in nuclear rDNA. Mol Biol Evol. 2004;21:129–40. 10.1093/molbev/msh005 [DOI] [PubMed] [Google Scholar]
78. Hochberg GKA, Thornton JW. Reconstructing ancient proteins to understand the causes of structure and function. Annu Rev Biophys. 2017;46:247–69. 10.1146/annurev-biophys-070816-033631 [DOI] [PMC free article] [PubMed] [Google Scholar]
79. Chisholm LO, Orlandi KN, Phillips SR et al. Ancestral reconstruction and the evolution of protein energy landscapes. Annu Rev Biophys. 2024;53:127–46. 10.1146/annurev-biophys-030722-125440 [DOI] [PMC free article] [PubMed] [Google Scholar]
80. Campitelli LF, Yellan I, Albu M et al. Reconstruction of full-length LINE-1 progenitors from ancestral genomes. Genetics. 2022;221:iyac074. 10.1093/genetics/iyac074 [DOI] [PMC free article] [PubMed] [Google Scholar]
81. Dewannieux M, Harper F, Richaud A et al. Identification of an infectious progenitor for the multiple-copy HERV-K human endogenous retroelements. Genome Res. 2006;16:1548–56. 10.1101/gr.5565706 [DOI] [PMC free article] [PubMed] [Google Scholar]
82. Lee YN, Bieniasz PD. Reconstitution of an infectious human endogenous retrovirus. PLoS Pathog. 2007;3:e10. 10.1371/journal.ppat.0030010 [DOI] [PMC free article] [PubMed] [Google Scholar]
83. Aubert M, Haick AK, Strongin DEea. Gene editing for latent herpes simplex virus infection reduces viral load and shedding in vivo. Nat Commun. 2024;15:4018. 10.1038/s41467-024-47940-y [DOI] [PMC free article] [PubMed] [Google Scholar]
84. Jalligampala A, Young JM, Feist J et al. RHO1-2 meganuclease gene editing targets human P23H rhodopsin-induced retinitis pigmentosa to rejuvenate rods and maintain cones. bioRxiv, 10.1101/2025.09.25.678625,. 25 September 2025, preprint: not peer reviewed. [DOI]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gkag924_Supplemental_Files

gkag924_supplemental_files.zip^{(4.7MB, zip)}

Data Availability Statement

[B1] 1. Kazazian HH. Mobile elements: drivers of genome evolution. Science. 2004;303:1626–32. 10.1126/science.1089670 [DOI] [PubMed] [Google Scholar]

[B2] 2. Feschotte C, Pritham EJ. DNA transposons and the evolution of eukaryotic genomes. Annu Rev Genet. 2007;41:331–68. 10.1146/annurev.genet.40.110405.090448 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B3] 3. Mita P, Boeke JD. How retrotransposons shape genome regulation. Curr Opin Genet Dev. 2016;37:90–100. 10.1016/j.gde.2016.01.001 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B4] 4. Smillie C, Garcillán-Barcia MP, Francia MV et al. Mobility of plasmids. Microbiol Mol Biol Rev. 2010;74:434–52. 10.1128/MMBR.00020-10 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B5] 5. Belfort M. Mobile self-splicing introns and inteins as environmental sensors. Curr Opin Microbiol. 2017;38:51–8. 10.1016/j.mib.2017.04.003 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B6] 6. Johnson CM, Grossman AD. Integrative and conjugative elements (ICEs): what they do and how they work. Annu Rev Genet. 2015;49:577–601. 10.1146/annurev-genet-112414-055018 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B7] 7. Pfeifer E, Sousa JM, Touchon M et al. When bacteria are phage playgrounds: interactions between viruses, cells, and mobile genetic elements. Curr Opin Microbiol. 2022;70:102230. 10.1016/j.mib.2022.102230 [DOI] [PubMed] [Google Scholar]

[B8] 8. Cambray G, Guerout AM, Mazel D. Integrons. Annu Rev Genet. 2010;44:141–66. 10.1146/annurev-genet-102209-163504 [DOI] [PubMed] [Google Scholar]

[B9] 9. Tenjo-Castaño F, Rout SS, Dey S et al. Unlocking the potential of CRISPR-associated transposons: from structural to functional insights. Trends Genet. 2025;41:660–77. 10.1016/j.tig.2025.04.005 [DOI] [PubMed] [Google Scholar]

[B10] 10. Haudiquet M, de Sousa JM, Touchon M et al. Selfish, promiscuous and sometimes useful: how mobile genetic elements drive horizontal gene transfer in microbial populations. Philos Trans R Soc Lond B Biol Sci. 2022;377:20210234. 10.1098/rstb.2021.0234 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B11] 11. Horne T, Orr VT, Hall JP. How do interactions between mobile genetic elements affect horizontal gene transfer?. Curr Opin Microbiol. 2023;73:102282. 10.1016/j.mib.2023.102282 [DOI] [PubMed] [Google Scholar]

[B12] 12. Tokuda M, Shintani M. Microbial evolution through horizontal gene transfer by mobile genetic elements. Microb Biotechnol. 2024;17:e14408. 10.1111/1751-7915.14408 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B13] 13. Chuong EB, Elde NC, Feschotte C. Regulatory activities of transposable elements: from conflicts to benefits. Nat Rev Genet. 2017;18:71–86. 10.1038/nrg.2016.139 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B14] 14. Pritham EJ. Transposable elements and factors influencing their success in eukaryotes. J Hered. 2009;100:648–55. 10.1093/jhered/esp065 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B15] 15. Wells JN, Feschotte C. A field guide to eukaryotic transposable elements. Annu Rev Genet. 2020;54:539–61. 10.1146/annurev-genet-040620-022145 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B16] 16. Song MJ, Schaack S. Evolutionary conflict between mobile DNA and host genomes. Am Nat. 2018;192:263–73. 10.1086/698482 [DOI] [PubMed] [Google Scholar]

[B17] 17. Stoddard BL. Homing endonucleases from mobile group I introns: discovery to genome engineering. Mob DNA. 2014;5:7. 10.1186/1759-8753-5-7 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B18] 18. Jacquier A, Dujon B. An intron-encoded protein is active in a gene conversion process that spreads an intron into a mitochondrial gene. Cell. 1985;41:383–94. 10.1016/S0092-8674(85)80011-8 [DOI] [PubMed] [Google Scholar]

[B19] 19. Macreadie IG, Scott RM, Zinn AR et al. Transposition of an intron in yeast mitochondria requires a protein encoded by that intron. Cell. 1985;41:395–402. 10.1016/S0092-8674(85)80012-X [DOI] [PubMed] [Google Scholar]

[B20] 20. Zinn AR, Butow RA. Nonreciprocal exchange between alleles of the yeast mitochondrial 21S rRNA gene: kinetics and the involvement of a double-strand break. Cell. 1985;40:887–95. 10.1016/0092-8674(85)90348-4 [DOI] [PubMed] [Google Scholar]

[B21] 21. Edgell DR, Belfort M, Shub DA. Barriers to intron promiscuity in bacteria. J Bacteriol. 2000;182:5281–9. 10.1128/JB.182.19.5281-5289.2000 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B22] 22. Michel F, Dujon B. Genetic exchanges between bacteriophage T4 and filamentous fungi. Cell. 1986;46:323–30. 10.1016/0092-8674(86)90651-3 [DOI] [PubMed] [Google Scholar]

[B23] 23. Shub DA, Gott JM, Xu MQ et al. Structural conservation among three homologous introns of bacteriophage T4 and the group I introns of eukaryotes. Proc Natl Acad Sci USA. 1988;85:1151–5. 10.1073/pnas.85.4.1151 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B24] 24. Goodrich-Blair H, Shub DA. Beyond homing: competition between intron endonucleases confers a selective advantage on flanking genetic markers. Cell. 1996;84:211–21. 10.1016/S0092-8674(00)80976-9 [DOI] [PubMed] [Google Scholar]

[B25] 25. Turmel M, Cote V, Otis C et al. Evolutionary transfer of ORF-containing group I introns between different subcellular compartments (chloroplast and mitochondrion). Mol Biol Evol. 1995;12:533–45. [DOI] [PubMed] [Google Scholar]

[B26] 26. Edgell DR, Gibb EA, Belfort M. Mobile DNA elements in T4 and related phages: a review in the series on bacteriophage T4 and its relatives. Virol J. 2010;7:290–300. 10.1186/1743-422X-7-290 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B27] 27. Turmel M, Boulanger J, Schnare MN et al. Six group I introns and three internal transcribed spacers in the chloroplast large subunit ribosomal RNA gene of the green alga Chlamydomonas eugametos. J Mol Biol. 1991;218:293–311. 10.1016/0022-2836(91)90713-G [DOI] [PubMed] [Google Scholar]

[B28] 28. Turmel M, Gutell RR, Mercier JP et al. Analysis of the chloroplast large subunit ribosomal RNA gene from 17 Chlamydomonas taxa. Three internal transcribed spacers and 12 group I intron insertion sites. J Mol Biol. 1993;232:446–67. 10.1006/jmbi.1993.1402 [DOI] [PubMed] [Google Scholar]

[B29] 29. Lambert AR, Hallinan JP, Werther R et al. Optimization of protein thermostability and exploitation of recognition behavior to engineer altered protein–DNA recognition. Structure. 2020;28:760–775. 10.1016/j.str.2020.04.009 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B30] 30. Stoddard BL. Homing endonucleases: from microbial genetic invaders to reagents for targeted DNA modification. Structure. 2011;19:7–15. 10.1016/j.str.2010.12.003 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B31] 31. Werther R, Hallinan JP, Lambert ARea. Crystallographic analyses illustrate significant plasticity and efficient recoding of meganuclease target specificity. Nucleic Acids Res. 2017;45:8621–34. 10.1093/nar/gkx544 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B32] 32. Burt A, Koufopanou V. Homing endonuclease genes: the rise and fall and rise again of a selfish element. Curr Opin Genet Dev. 2004;14:609–15. 10.1016/j.gde.2004.09.010 [DOI] [PubMed] [Google Scholar]

[B33] 33. Koufopanou V, Burt A. Degeneration and domestication of a selfish gene in yeast: molecular evolution versus site-directed mutagenesis. Mol Biol Evol. 2005;22:1535–8. 10.1093/molbev/msi149 [DOI] [PubMed] [Google Scholar]

[B34] 34. Takeuchi R, Certo M, Caprara MG et al. Optimization of in vivo activity of a bifunctional homing endonuclease and maturase reverses evolutionary degradation. Nucleic Acids Res. 2009;37:877–90. 10.1093/nar/gkn1007 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B35] 35. Scalley-Kim M, McConnell-Smith A, Stoddard BL. Coevolution of a homing endonuclease and its host target sequence. J Mol Biol. 2007;372:1305–19. 10.1016/j.jmb.2007.07.052 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B36] 36. Ho Y, Kim SJ, Waring RB. A protein encoded by a group I intron in Aspergillus nidulans directly assists RNA splicing and is a DNA endonuclease. Proc Natl Acad Sci USA. 1997;94:8994–9. 10.1073/pnas.94.17.8994 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B37] 37. Lambert AR, Hallinan JP, Shen BWea. Indirect DNA sequence recognition and its impact on nuclease cleavage activity. Structure. 2016;24:862–73. 10.1016/j.str.2016.03.024 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B38] 38. Sayers EW, Cavanaugh M, Frisse L et al. GenBank 2025 update. Nucleic Acids Res. 2025;53:D56–61. 10.1093/nar/gkae1114 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B39] 39. Altschul SF, Gish W, Miller W et al. Basic local alignment search tool. J Mol Biol. 1990;215:403–10. 10.1016/S0022-2836(05)80360-2 [DOI] [PubMed] [Google Scholar]

[B40] 40. Altschul SF, Madden TL, Schaffer AA et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997;25:3389–402. 10.1093/nar/25.17.3389 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B41] 41. Kosuri S, Eroshenko N, Leproust EM et al. Scalable gene synthesis by selective amplification of DNA pools from high-fidelity microchips. Nat Biotechnol. 2010;28:1295–9. 10.1038/nbt.1716 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B42] 42. Jarjour J, West-Foyle H, Certo MT et al. High-resolution profiling of homing endonuclease binding and catalytic specificity using yeast surface display. Nucleic Acids Res. 2009;37:6871–80. 10.1093/nar/gkp726 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B43] 43. Gietz RD, Schiestl RH. Large-scale high-efficiency yeast transformation using the LiAc/SS carrier DNA/PEG method. Nat Protoc. 2007;2:38–41. 10.1038/nprot.2007.15 [DOI] [PubMed] [Google Scholar]

[B44] 44. Becton DaC. 10 ed, Ashland OR. 2023;

[B45] 45. Baxter SK, Lambert AR, Scharenberg AM et al. Flow cytometric assays for interrogating LAGLIDADG homing endonuclease DNA-binding and cleavage properties. Methods Mol Biol. 2013;978:45–61. 10.1007/978-1-62703-293-3 10.1007/978-1-62703-293-3_4 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B46] 46. Chao G, Lau WL, Hackel BJ et al. Isolating and engineering human antibodies using yeast surface display. Nat Protoc. 2006;1:755–68. 10.1038/nprot.2006.94 [DOI] [PubMed] [Google Scholar]

[B47] 47. Mistry J, Chuguransky S, Williams Lea. Pfam: the protein families database in 2021. Nucleic Acids Res. 2021;49:D412–9. 10.1093/nar/gkaa913 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B48] 48. Darriba D, Taboada GL, Doallo R et al. ProtTest 3: fast selection of best-fit models of protein evolution. Bioinformatics. 2011;27:1164–5. 10.1093/bioinformatics/btr088 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B49] 49. Guindon S, Dufayard JF, Lefort V et al. New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. Syst Biol. 2010;59:307–21. 10.1093/sysbio/syq010 [DOI] [PubMed] [Google Scholar]

[B50] 50. R Core Team . R: A Language and Environment for Statistical Computing, Vienna, Austria: R Foundation for Statistical Computing; 2025.; [Google Scholar]

[B51] 51. Paradis E, Schliep K. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics. 2019;35:526–8. 10.1093/bioinformatics/bty633 [DOI] [PubMed] [Google Scholar]

[B52] 52. Yu G. Using ggtree to visualize data on tree-like structures. Curr Protoc Bioinformatics. 2020;69:e96. 10.1002/cpbi.96 [DOI] [PubMed] [Google Scholar]

[B53] 53. Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013;30:772–80. 10.1093/molbev/mst010 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B54] 54. Redelings BD. BAli-Phy version 3: model-based co-estimation of alignment and phylogeny. Bioinformatics. 2021;37:3032–4. 10.1093/bioinformatics/btab129 [DOI] [PubMed] [Google Scholar]

[B55] 55. Jacoby K, Scharenberg AM. Homing endonuclease target determination using SELEX adapted for yeast surface display. Methods Mol Biol. 2014;1123:165–90. 10.1007/978-1-62703-968-0 10.1007/978-1-62703-968-0_13 [DOI] [PubMed] [Google Scholar]

[B56] 56. Jacoby K, Lambert AR, Scharenberg AM. Characterization of homing endonuclease binding and cleavage specificities using yeast surface display SELEX (YSD-SELEX). Nucleic Acids Res. 2017;45:e11. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B57] 57. Kabsch W. Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr D Biol Crystallogr. 2010;66:133–44. 10.1107/S0907444909047374 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B58] 58. Evans P. Scaling and assessment of data quality. Acta Crystallogr D Biol Crystallogr. 2006;62:72–82. 10.1107/S0907444905036693 [DOI] [PubMed] [Google Scholar]

[B59] 59. Evans PR, Murshudov GN. How good are my data and what is the resolution?. Acta Crystallogr D Biol Crystallogr. 2013;69:1204–14. 10.1107/S0907444913000061 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B60] 60. Adams PD, Afonine PV, Bunkoczi Gea. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010;66:213–21. 10.1107/S0907444909052925 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B61] 61. McCoy AJ, Grosse-Kunstleve RW, Adams PD et al. Phaser crystallographic software. J Appl Crystallogr. 2007;40:658–74. 10.1107/S0021889807021206 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B62] 62. Emsley P, Lohkamp B, Scott WG et al. Features and development of Coot. Acta Crystallogr D Biol Crystallogr. 2010;66:486–501. 10.1107/S0907444910007493 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B63] 63. Berman HM, Westbrook J, Feng Z et al. The Protein Data Bank. Nucleic Acids Res. 2000;28:235–42. 10.1093/nar/28.1.235 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B64] 64. Sun Z, Brodsky JL. Protein quality control in the secretory pathway. J Cell Biol. 2019;218:3171–87. 10.1083/jcb.201906047 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B65] 65. Shusta EV, Kieke MC, Parke E et al. Yeast polypeptide fusion surface display levels predict thermal stability and soluble secretion efficiency. J Mol Biol. 1999;292:949–56. 10.1006/jmbi.1999.3130 [DOI] [PubMed] [Google Scholar]

[B66] 66. Bonocora RP, Shub DA. A likely pathway for formation of mobile group I introns. Curr Biol. 2009;19:223–8. 10.1016/j.cub.2009.01.033 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B67] 67. Takeuchi R, Lambert AR, Mak AN et al. Tapping natural reservoirs of homing endonucleases for targeted gene modification. Proc Natl Acad Sci USA. 2011;108:13077–82. 10.1073/pnas.1107719108 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B68] 68. Wheeler LC, Lim SA, Marqusee S et al. The thermostability and specificity of ancient proteins. Curr Opin Struct Biol. 2016;38:37–43. 10.1016/j.sbi.2016.05.015 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B69] 69. Thomson RES, Carrera-Pacheco SE, Gillam EMJ. Engineering functional thermostable proteins using ancestral sequence reconstruction. J Biol Chem. 2022;298:102435. 10.1016/j.jbc.2022.102435 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B70] 70. Esler MA, Werther R, Doyle LA et al. Caveat emptor: predicting and modeling protein–DNA recognition and binding via machine-learning computational approaches. Nucleic Acids Res. 2026;54. 10.1093/nar/gkag608 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B71] 71. Shen BW, Lambert A, Walker BC et al. The structural basis of asymmetry in DNA binding and cleavage as exhibited by the I-SmaMI LAGLIDADG meganuclease. J Mol Biol. 2016;428:206–20. 10.1016/j.jmb.2015.12.005 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B72] 72. Chevalier B, Turmel M, Lemieux C et al. Flexible DNA target site recognition by divergent homing endonuclease isoschizomers I-CreI and I-MsoI. J Mol Biol. 2003;329:253–69. 10.1016/S0022-2836(03)00447-9 [DOI] [PubMed] [Google Scholar]

[B73] 73. Blumenstiel JP. Birth, school, work, death, and resurrection: the life stages and dynamics of transposable element proliferation. Genes. 2019;10:336. 10.3390/genes10050336 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B74] 74. Brouha B, Schustak J, Badge RM et al. Hot L1s account for the bulk of retrotransposition in the human population. Proc Natl Acad Sci USA. 2003;100:5280–5. 10.1073/pnas.0831042100 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B75] 75. Grahn RA, Rinehart TA, Cantrell MA et al. Extinction of LINE-1 activity coincident with a major mammalian radiation in rodents. Cytogenet Genome Res. 2005;110:407–15. 10.1159/000084973 [DOI] [PubMed] [Google Scholar]

[B76] 76. Sethuraman J, Majer A, Friedrich NC et al. Genes within genes: multiple LAGLIDADG homing endonucleases target the ribosomal protein S3 gene encoded within an rnl group I intron of Ophiostoma and related taxa. Mol Biol Evol. 2009;26:2299–315. 10.1093/molbev/msp145 [DOI] [PubMed] [Google Scholar]

[B77] 77. Haugen P, Reeb V, Lutzoni F et al. The evolution of homing endonuclease genes and group I introns in nuclear rDNA. Mol Biol Evol. 2004;21:129–40. 10.1093/molbev/msh005 [DOI] [PubMed] [Google Scholar]

[B78] 78. Hochberg GKA, Thornton JW. Reconstructing ancient proteins to understand the causes of structure and function. Annu Rev Biophys. 2017;46:247–69. 10.1146/annurev-biophys-070816-033631 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B79] 79. Chisholm LO, Orlandi KN, Phillips SR et al. Ancestral reconstruction and the evolution of protein energy landscapes. Annu Rev Biophys. 2024;53:127–46. 10.1146/annurev-biophys-030722-125440 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B80] 80. Campitelli LF, Yellan I, Albu M et al. Reconstruction of full-length LINE-1 progenitors from ancestral genomes. Genetics. 2022;221:iyac074. 10.1093/genetics/iyac074 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B81] 81. Dewannieux M, Harper F, Richaud A et al. Identification of an infectious progenitor for the multiple-copy HERV-K human endogenous retroelements. Genome Res. 2006;16:1548–56. 10.1101/gr.5565706 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B82] 82. Lee YN, Bieniasz PD. Reconstitution of an infectious human endogenous retrovirus. PLoS Pathog. 2007;3:e10. 10.1371/journal.ppat.0030010 [DOI] [PMC free article] [PubMed] [Google Scholar]

[B83] 83. Aubert M, Haick AK, Strongin DEea. Gene editing for latent herpes simplex virus infection reduces viral load and shedding in vivo. Nat Commun. 2024;15:4018. 10.1038/s41467-024-47940-y [DOI] [PMC free article] [PubMed] [Google Scholar]

[B84] 84. Jalligampala A, Young JM, Feist J et al. RHO1-2 meganuclease gene editing targets human P23H rhodopsin-induced retinitis pigmentosa to rejuvenate rods and maintain cones. bioRxiv, 10.1101/2025.09.25.678625,. 25 September 2025, preprint: not peer reviewed. [DOI]

PERMALINK

The activity, divergence, and evolutionary degradation of modern-day homing endonucleases and behavior of their reconstructed ancestors

Juliana C Young

Abigail R Lambert

Janet M Young

Lindsey A Doyle

Miriam Silverstein

David R Edgell

Barry L Stoddard

Roles

Abstract

Graphical Abstract

Graphical Abstract.

Introduction

Materials and methods

Bioinformatic LHE sequence screening

Cloning of synthetic LHE ORFs

Yeast induction of surface expression

Yeast staining for surface expression

Figure 1.

Generation of labeled DNA target site substrates

DNA cleavage assays using surface-displayed endonucleases (“tethered” DNA cleavage assays)

Figure 2.

DNA cleavage assays using endonucleases released from the yeast surface (“untethered” DNA cleavage assays)

Determination of site and centering of DNA target cleavage via run-off DNA sequencing

Ancestral sequence reconstructions

Figure 3.

Expression and purification of recombinant LHEs

Determination of LHE thermal stability via thermal denaturation and circular dichroism spectroscopy

In vitro cleavage assays using purified recombinant LHEs and 500 bp target site substrates

Yeast SELEX library generation and screening

Figure 6.

In vitro cleavage library generation and screen

In vitro cleavage assays using purified recombinant aHE15 and short A647-labeled target site substrates

One-off specificity profile using short A647-labeled DNA target site substrates and purified recombinant aHE15

Crystallographic analysis of aHE31 and aHE15

Table 1.

Results

Most extant (present-day) LHEs display compromised folding and cleavage activity

Assessment of expression and folding

Assessment of cleavage activity

Ancestral reconstruction experiments generate active constructs with enhanced stability and the potential for novel DNA recognition specificity

Expansion of the LHE sequence collection, phylogenetic analyses, and ancestral reconstructions

“Shallow” ancestral reconstructions display enhanced expression, increased thermal stability, and retention of activity

Figure 4.

Figure 5.

Determination of the DNA cleavage specificity profile of the aHE15 reconstructed ancestor

Figure 7.

Structural features of DNA recognition and its divergence

Figure 8.

Figure 9.

Discussion

Homing endonuclease diversity and evolutionary degradation

Ancestral reconstructions yield LHEs with robust expression and cleavage activity

Supplementary Material

Acknowledgements

Contributor Information

Supplementary data

Conflict of interest

Funding

Data availability

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases