Choosing a Cloning Vector: What the Backbone Decides
6 min read · Updated October 30, 2026
A plasmid backbone is four decisions that have nothing to do with each other: where it replicates and at what copy number, what selects for it, what drives the insert, and which hosts it can live in. Most cloning problems that look like bad luck are one of those four being wrong for the job.
This is what each one commits you to, with the vector families that exist for each case.
Copy number is the origin, and it is a trade
The origin of replication sets how many copies of the plasmid each cell carries. pUC-type origins give several hundred, pBR322-derived pMB1 and ColE1 origins give tens, p15A around ten, and single-copy BAC and fosmid origins exactly one or two.
High copy is what you want for a prep: more plasmid per millilitre of culture, which is why routine cloning vectors are pUC derivatives. It is also what you do not want for a toxic gene, a large insert or anything repetitive — a few hundred copies of a leaky promoter kills the cells, and high copy number multiplies the chance that two repeats find each other.
Low and single-copy vectors exist for exactly those cases, and the price is yield: a BAC prep from the same culture volume gives orders of magnitude less DNA.
The marker is chosen for the experiment around it, not for the plasmid
Ampicillin resistance is the default on cloning vectors and is the weakest selection of the common ones: beta-lactamase is secreted, so it destroys the antibiotic in the medium around a colony and satellite colonies carrying no plasmid grow in the cleared zone. Carbenicillin is more stable and reduces this. Kanamycin and chloramphenicol select cleanly; chloramphenicol is bacteriostatic, which changes what a plate looks like.
The constraint is usually the rest of the experiment: a second plasmid in the same cell needs a different marker and a compatible origin, a genomic integration cassette needs a marker the strain does not already carry, and a plasmid going into a plant or a mammalian line needs a eukaryotic selection marker in addition to the bacterial one.
In a library of 309 publicly deposited vectors, 67 carry a feature annotated AmpR and 93 carry one annotated bla — the same gene under its other name, and between them the commonest resistance marker by a long way, against 13 for kanamycin and 1 for chloramphenicol. That is a statement about the history of cloning rather than about what is best.
The promoter decides what you can ask of the construct
A cloning vector often has no promoter aimed at the insert at all, which is right when the job is to hold and propagate a sequence. An expression vector has one, and which one determines what kind of control you have: T7 for high-yield induced expression in a strain carrying T7 polymerase, tac or trc for strong IPTG-inducible expression in ordinary strains, araBAD for tunable expression across a range of arabinose concentrations, and constitutive promoters where you want no control at all.
Sixty of those 309 vectors carry an annotated T7 promoter. A T7 vector in a strain without the polymerase expresses nothing, which is the commonest version of a construct that is perfect and silent.
Host range is a property of the origin, and shuttle vectors have two
A ColE1-family origin replicates in E. coli and relatives and nowhere else. A yeast vector carries a 2-micron or ARS/CEN origin alongside its bacterial one; a plant binary vector carries a broad-host-range origin for Agrobacterium plus T-DNA borders; a mammalian vector usually does not replicate in the target cell at all and either integrates or is transient.
That is why these are the categories a vector library is organised by — E. coli cloning and expression, yeast, mammalian, plant binary, BAC/fosmid, recombineering, phage/M13 — and why the first question about an unfamiliar plasmid is which of them it belongs to.
Size, and what a public record does and does not tell you
Backbone size matters at both ends: of those 309 vectors, 17 are 3 kb or smaller — the pUC and pGEM class, where transformation is efficient and the prep is clean — and 31 are 10 kb or larger, which is where binary vectors and BACs live and where transformation efficiency starts to cost you.
What a public record shows is whatever its depositor annotated. pUC19's GenBank record, one of the most-used plasmids in biology, carries seven features and none of them is labelled: the record states spans, not names. A thinly annotated record is not a thinly featured plasmid, and the gap is filled by annotating the sequence yourself rather than by assuming the features are absent.
The other thing a record cannot tell you is whether it matches the tube in your freezer. Lab plasmids accumulate changes; the sequence that matters is the one you read back.
Frequently asked questions
What is the difference between a high-copy and a low-copy plasmid?
The origin of replication. pUC-type origins give several hundred copies per cell, pMB1 and ColE1 tens, p15A about ten, and BAC/fosmid origins one or two. High copy gives yield; low copy gives stability for toxic, large or repetitive inserts, and is the right choice when the insert is expressed at all.
Which antibiotic marker should I choose?
Whichever the rest of the experiment leaves free. Ampicillin is the default and the weakest — secreted beta-lactamase clears the antibiotic around a colony and allows satellite colonies — so use carbenicillin in its place where that is a problem, and kanamycin or chloramphenicol where you need clean selection or a second plasmid in the same cell.
Why is my T7 expression vector not expressing?
Most often because the strain carries no T7 RNA polymerase. A T7 promoter is inert in an ordinary cloning strain; it needs a DE3 lysogen or an equivalent polymerase source. The construct can be perfectly correct and completely silent, and nothing about the sequence says so.
What makes a vector a shuttle vector?
Two origins, one for each host, plus a selection marker that works in each. A yeast shuttle vector replicates in E. coli for cloning and in yeast for the experiment; the bacterial half exists so you can build and verify the plasmid in a system that is fast and cheap.
Related references
Related tools
309 publicly deposited vectors with their full GenBank feature tables, and 617 parts harvested from them.
Auto-detect promoters, tags, origins and resistance markers in a plasmid, then run a deep scan against pLannotate's full feature databases.
Which antibiotic, which host, which backbone, and what will bite you later.
Assemble fragments and design junction primers for Gibson, Golden Gate or restriction cloning.
Locate the exact direct repeats that let a construct recombine away the DNA between them, and build the shortened molecule you would actually recover.