Why Plasmids Rearrange: Direct-Repeat Deletions
12 min read · Updated September 11, 2026
A plasmid that repeatedly comes back shorter may contain its own deletion substrate. Two sufficiently similar sequences in the same orientation can align during replication or recombination, leaving a product with one copy of the repeat and less of the DNA between them. Reusing one promoter, terminator or coding block in a multi-cassette construct is a common way to build that substrate deliberately without noticing it.
An exact-repeat scan can locate the homology and construct the sequence products the pair permits. It cannot say that the event will happen, assign a deletion rate or rule out every other cause of instability. This guide shows how to read that bounded result, redesign exact homology, propagate difficult constructs conservatively and screen for the specific shortened molecule.
First confirm that the symptom is a structural rearrangement
A smaller restriction pattern, a whole-plasmid sequence missing one cassette or colonies that repeatedly carry the same shortened map are evidence for a structural change. Slow growth, loss of fluorescence or poor plasmid yield alone are not: they can arise from toxic expression, plasmid loss, copy-number changes, selection failure or culture conditions without deleting any DNA.
Rule out a wrong starting file, sample swap, partial digest and an incomplete sequencing read before assigning a mechanism. A Sanger read that stops before the missing interval does not prove the interval is absent. A supercoiled plasmid also migrates by conformation rather than by its base-pair length, so an unexpected uncut band is not a deletion-size measurement.
When multiple independent colonies carry the same sequence junction, that junction is the useful clue. Ask whether its two sides correspond to copies of one repeated element. A direct-repeat analysis starts from the complete sequence and explains one class of structural product; it should not replace observing the product itself.
A direct repeat has sequence, orientation and spacing
A direct repeat is a pair of homologous stretches written in the same orientation on the molecule. In a simple diagram, repeat 1 points right, a spacer follows, and repeat 2 also points right. Recombination or slippage between those copies can leave one repeat at the new junction and remove the intervening arc from that product.
Promoters, terminators, polyadenylation signals, insulators, affinity tags and duplicated coding regions create this geometry when the same part is reused. The repeated part can be perfectly functional in isolation; the issue is the second homologous copy in the same construct.
An inverted repeat points in the opposite orientation. It can form secondary structures and can stimulate deletions, inversions or other rearrangements through different pathways, but it is not the same search problem. SeqBench's Plasmid Stability Checker looks only for exact, co-oriented direct repeats. A clean result says nothing about inverted pairs.
More than one molecular route can produce a deletion
Short-repeat deletions in bacteria can arise without RecA through replication slippage or sister-strand exchange, while longer homologies can also use RecA-dependent recombination. Lovett et al. demonstrated RecA-independent sister-strand exchange, and Bi and Liu showed strong effects of repeat length and spacing on deletion pathways (Genetics 1993;135:631-642, doi:10.1093/genetics/135.3.631; J Mol Biol 1994;235:414-423, doi:10.1006/jmbi.1994.1002).
The idealised crossover drawing on one circular molecule partitions its two sequence arcs into two circles, each carrying one repeat copy. Replication-based and sister-strand routes can instead produce a shortened monomer, a dimer or other intermediates. The sequence geometry tells you which junctions are possible; it does not prove that two stable free circles are the products present in your culture.
Selection then decides what is recovered. A candidate retaining a replication origin and antibiotic-resistance marker can propagate, while an excised arc lacking them may disappear. A shorter plasmid can also grow faster and overtake a culture even if the initial event was rare. Inspect the features on both candidate arcs before calling one “the deletion product.”
Repeat length and spacing matter, but they are not a universal rate
Defined bacterial assays generally find more rearrangement with longer, more identical repeats and, in many systems, with closer spacing. Sequence divergence and mismatch repair can suppress exchange. Host recombination genotype, replication origin, plasmid copy number, transcription, growth temperature, culture history and selection all change what is observed.
Published empirical models do relate repeat length and spacing to measured frequencies in particular plasmid datasets. Oliveira et al. fitted such relationships across defined repeats and hosts, and the EFM Calculator later implemented a model for engineered constructs (Plasmid 2008;60:159-165, doi:10.1016/j.plasmid.2008.06.004; ACS Synth Biol 2015;4:939-943, doi:10.1021/acssynbio.5b00068). Those calibrations do not supply a validated deletion rate for every host, replicon, sequence context and culture protocol.
SeqBench therefore does not output a probability. It sorts exact pairs by repeat length, then spacer and coordinate for deterministic review. That order is a triage order, not a claim that the first row is the product most likely to dominate your cells.
What the exact-repeat checker actually searches
The checker finds literal same-orientation matches at or above the chosen minimum, extends each seed to a maximal disjoint pair and reports 1-based inclusive coordinates. Circular mode searches through the arbitrary point where the plasmid sequence starts, so a repeat crossing the origin is not lost. Linear mode does not join the ends.
The default minimum is 20 bp, and values down to 8 bp are accepted. That is a search threshold, not a biological boundary below which recombination stops. Lowering it produces many chance matches and low-complexity hits; raising it can hide a shorter repeat that matters in a sensitive construct.
For a circular molecule, the two repeats divide the sequence into two arcs and the result returns both candidate circles. Rotating the same plasmid can swap which arc receives which label, while the unordered pair of product sequences remains the same. For a linear molecule, the shortened product is returned without inventing an excised circle.
Highly repetitive seeds are skipped to keep the search bounded, and the result reports a low-complexity warning when that happens. A hard no-repeat gate can still show PASS while this soft warning says part of the sequence was not searched exhaustively. The defensible wording is “no exact direct repeat at or above this threshold was reported in the searched regions,” not “this plasmid is stable.”
Worked example: one reused 44 bp block
The built-in synthetic example is A + T + B + T + C. A, B and C are each 204 bp, while T is the same 44 bp sequence in both positions. The complete circular construct is 700 bp. It is a compact demonstration rather than a biological vector with an origin and marker.
The checker reports the T copies at positions 205-248 and 453-496. The spacer between them is 204 bp. A junction that keeps one copy and removes the spacer plus the other copy removes 248 bp, not 204, because one of the two repeated copies also disappears from that product.
The two sequence arcs become candidate circles of 452 bp and 248 bp. In the current coordinate representation they are A + T + C and B + T. Their lengths sum to the original 700 bp, which is the useful arithmetic check on the circular construction.
Nothing in that synthetic sequence says which circle would replicate or be selected. On a real plasmid, annotate the origin, resistance marker and expression cassette on both products. The useful output is the exact candidate junction and full candidate sequence you can test, not a prediction that one named circle will be recovered.
Remove exact homology without removing function
The cleanest fix is usually to replace one repeated regulatory part with a characterised sequence-diverse alternative that performs the same role. Multi-gene architectures benefit from promoter and terminator libraries precisely because using different parts avoids long exact homology as well as offering expression tuning.
For duplicated coding sequence, synonymously recode one copy while preserving the protein, codon constraints and any required motifs. Rescan the entire assembled plasmid after recoding: moving the exact repeat or creating a new homologous block elsewhere is possible, and recoding can introduce restriction sites, extreme GC windows or other synthesis problems.
Sleight et al. showed the practical design principle in synthetic circuits: repeated terminators supported deletions, while sequence-diverse replacement improved evolutionary lifetime (J Biol Eng 2010;4:12, doi:10.1186/1754-1611-4-12). The lesson is not that any different-looking part is equivalent. Preserve measured function, orientation and host compatibility while reducing unnecessary identity.
If a repeated element is biologically required, shorten exact identity where possible and document the unavoidable junction. A sequence redesign should be validated for function as well as for the absence of one repeat pair; a stable construct that no longer expresses correctly is not a successful fix.
Propagate unavoidable repeats conservatively
A recombination-deficient or stability-oriented cloning strain can reduce some rearrangements, but recA− does not eliminate short-repeat deletion because RecA-independent pathways remain. Match the strain to the construct and follow the supplier's genotype-specific guidance rather than treating “stable” as a universal property.
For difficult plasmids, manufacturer protocols commonly recommend growth around 30 °C, fresh transformants, short culture histories and avoiding unnecessary stationary-phase propagation. These practices can reduce selection for a faster-growing deletion product in some constructs; none guarantees that the starting colony is intact.
Transform and screen multiple independent colonies, keep a verified low-passage stock and minimise the number of generations between that stock and the experiment. Maintain appropriate selection, but remember that selection protects only the DNA needed to retain the marker and origin. A deletion that keeps both can be selected perfectly well.
Screen for the specific shortened molecule
A candidate deletion sequence turns a vague concern into a diagnostic alternative. Copy it into the Diagnostic Digest Planner beside the intended plasmid and search for an enzyme whose observable band pattern differs between them. The current pages use manual copy and paste; the sequences are API-compatible but there is no automatic prefilled handoff.
Choose a digest that cuts the intended molecule and distinguishes the two complete-digest patterns without relying only on a tiny or faint fragment. Include other plausible products if the assembly can make them. A digest designed only against this one repeat deletion proves only that the screened clone does not share that particular pattern.
Junction PCR is another direct test: one primer on each side of the predicted new junction can produce a shorter amplicon from the rearranged molecule. Whole-plasmid sequencing or read-back verification is stronger when other rearrangements matter. Sequence the survivors before banking them, because a matching digest cannot exclude point mutations or an unlisted structural product.
What a clean direct-repeat result does not rule out
Near-repeats with mismatches can still recombine, and mismatch repair and divergence change their frequency. Exact matching deliberately misses them. Inverted repeats, palindromes, simple sequence repeats, mobile elements and structures formed during replication can produce different rearrangements outside this model.
The checker also does not model toxic gene expression, plasmid loss, copy-number instability, host incompatibility, metabolic burden, transcription-replication conflicts or a marker that no longer selects under the culture conditions. Any of those can make a construct disappear or enrich a variant without one exact direct-repeat deletion.
Use the result as a bounded statement about one sequence property. A reported pair is a real candidate substrate and gives testable products. No reported pair removes that substrate class above the selected threshold only where the search was complete. Stability remains an observation made across propagation and verification.
Frequently asked questions
Why can repeated promoters or terminators make a plasmid rearrange?
Two copies in the same orientation create homologous direct repeats. Replication slippage or recombination can align those copies and produce a junction retaining one repeat while removing intervening DNA from a product. The parts are not inherently bad; reusing the same sequence creates the substrate.
Does finding a direct repeat mean my plasmid will delete?
No. It means the sequence permits a particular homologous junction. Observed frequency also depends on repeat length and identity, spacing, host genotype, replicon, copy number, transcription, culture history and selection. SeqBench constructs candidate products but does not assign a deletion probability.
Can a recA-negative strain prevent direct-repeat deletions?
It can reduce RecA-dependent recombination, but it does not eliminate RecA-independent replication slippage and sister-strand exchange, especially for short repeats. Use a strain appropriate for unstable constructs and conservative propagation conditions, then verify multiple colonies rather than treating the genotype as a guarantee.
Why does a circular result contain two products?
The two repeat copies divide the original circle into two sequence arcs, so an ideal crossover representation can construct two circles, each with one copy. Those are candidate sequence products, not a promise that both persist as free plasmids. Which molecule is recovered depends on mechanism and on where the origin and selection marker lie.
What minimum repeat length should I search?
Twenty base pairs is the tool's default, not a biological threshold. Lowering it can expose shorter candidates but rapidly increases chance matches and low-complexity work; raising it reduces noise but can hide a relevant short repeat. Start at 20, inspect warnings and lower the threshold when the construct or observed junction gives a reason.
Can repeat length and spacing predict a deletion rate?
Published empirical models relate those variables to frequencies in defined bacterial plasmid datasets, but they are calibrated to particular hosts, replicons and conditions. SeqBench does not transfer one rate to an arbitrary construct. It reports exact geometry and candidate products, leaving frequency to be measured in the actual system.
Does no exact repeat mean my plasmid is stable?
No. It means no exact co-oriented repeat at or above the chosen threshold was reported in the searched regions. Near-repeats, inverted repeats, low-complexity regions skipped by the search, mobile elements, toxic expression, plasmid loss and host effects remain outside that statement.
Does the checker find inverted repeats?
No. It searches exact direct repeats in the same orientation. Inverted repeats can form secondary structures and contribute to inversions, deletions or synthesis and sequencing problems through other mechanisms, so their absence from this output is not evidence they are harmless or absent.
How should I verify a predicted deletion product?
Use its exact sequence as an alternative when designing a diagnostic digest, or design PCR across the predicted new junction. Screen several independent colonies and confirm selected clones by junction or whole-plasmid sequencing. A digest matching the intended pattern excludes only the alternatives it was designed to distinguish.
Related references
Related tools
Locate the exact direct repeats that let a construct recombine away the DNA between them, and build the shortened molecule you would actually recover.
Find the restriction digest that tells your intended construct apart from the empty vector, a flipped insert or a mis-assembly — with the bands you would actually be able to read.
Scan a coding sequence for premature stops, cryptic RBS/polyA signals, unwanted restriction sites, GC extremes and repeats.
Auto-detect promoters, tags, origins and resistance markers in a plasmid, then run a deep scan against pLannotate's full feature databases.
Align a Sanger read to a reference and get a pass / needs-review verification report.
Align many NGS/Nanopore/Sanger reads to a reference with minimap2, call multi-read consensus variants, and build a corrected consensus sequence.