
Molecular cloning underpins most of modern molecular biology. The principle has not changed in fifty years: you join a piece of DNA to a vector, put it into a host, and let the host make the copies. What has changed is how many ways there are to make the join, and how much this choice affects the work around it.
The method you choose decides your primer design, how you prepare your DNA, how many colonies you have to screen, and what sequence you end up with at each junction. Modern laboratories build multi-gene constructs that encode whole metabolic pathways, and for work like this the assembly method is a design decision rather than a protocol detail.
This page is the directory. Below, each method has a short description, a note on where it is awkward, and a link to a full guide with design rules and troubleshooting.
How to choose your cloning method
Start with the construct you want to build, not with the method you used last time.
| Method | Seamless | Typical fragments | Typical use |
|---|---|---|---|
| Restriction cloning | No | 1 | Simple subcloning when suitable sites are available |
| TA and blunt-end cloning | No | 1 | Cloning PCR products for sequencing or storage |
| TOPO cloning | No | 1 | Fast cloning of PCR products into prepared vectors |
| Gibson Assembly | Yes | 2 to 5 | Joining fragments through matching overlaps |
| NEBuilder HiFi | Yes | 2 to 6, more with care | Overlap assembly, including demanding multi-fragment constructs |
| Golden Gate | Yes, see note | 2 to 20, more with validated overhangs | Ordered assembly of many parts and reusable modules |
| In-Fusion | Yes | 2 to 5 | Insertion into a linearized vector without restriction sites |
| Gateway | No, att sites remain | 1 to 4 | Moving one gene between many compatible expression vectors |
Golden Gate junctions can be seamless. Standardized systems such as MoClo use defined fusion sites, so whether a junction is seamless depends on how the parts were designed.
Seamless means the method adds no junction sequence of its own. It does not mean the final sequence is whatever you imagine it to be. The overlaps and overhangs you design become part of the construct, so check every junction before you order primers.
The fragment numbers are practical ranges rather than limits. Efficiency falls as fragment count rises, and the published maximum for a method usually assumes clean DNA and a well-chosen set of junctions.
It is also worth thinking about the whole experiment rather than the reaction. DNA preparation, transformation, colony screening and sequence verification take several days, and they take the same several days whichever method you chose. What changes is how many colonies you have to screen before you find a correct clone. A method that joins your fragments in five minutes but puts the insert in either orientation will cost you more screening than a slower method that can only put it in one way round.
Part I: The fundamentals
Three things sit underneath every method: the shape of the workflow, the enzymes that cut DNA, and the PCR that generates your fragments.
The plasmid cloning workflow
A cloning experiment begins long before you pick up a pipette. In silico design, meaning design on screen before any bench work, is where you fix the exact sequence, check reading frames, and place the overlaps, overhangs or recombination sites your method needs.
Once your design is set, you prepare the insert and the vector, usually by PCR or restriction digestion.
Next, you join the fragments, with a ligase, an enzyme mix or a recombination system. Then you introduce the product into a host, normally competent E. coli, to propagate the plasmid.
Finally, you screen colonies by colony PCR or diagnostic digest, and confirm the promising ones by sequencing across the insert, the junctions and any region that was amplified by PCR. An insert of the right size is a good first sign. It is not a verified construct.
The whole process, from design to verification: The Plasmid Cloning Workflow: From DNA to Transformed E. coli.
Restriction enzymes
Restriction enzymes are the original tools of molecular cloning. Bacteria use them to defend themselves against phages. Each enzyme recognizes a specific sequence and cuts at or near it, leaving a short single-stranded overhang or a blunt end.
Even if you work almost entirely with seamless methods, these enzymes stay useful. You need them to linearize a vector, release an insert, or check a finished construct on a gel.
Most commonly used Type II enzymes cut inside their recognition site. Type IIS enzymes, a subgroup, cut at a fixed distance outside it. This is the property Golden Gate is built on, because it lets you choose the overhang sequence independently of the sequence the enzyme recognizes.
Enzyme types, double digests and star activity: Restriction Enzymes Explained: Types, Star Activity and Buffer Compatibility.
PCR fundamentals for cloning
Modern cloning relies heavily on PCR, both to generate inserts and to linearize vectors. With a custom primer you amplify your gene and add the overlap, overhang or restriction site your method needs to the end of the product, in one step.
A cloning PCR is not a diagnostic PCR. A diagnostic reaction answers whether a sequence is present. A cloning reaction has to produce a building block with the right ends and no mutations, because whatever the polymerase gets wrong ends up in your construct.
Choose the polymerase and design the primers together with the cloning method. A clean band of the right size is not enough if the ends are wrong for the joining reaction.
Polymerase choice, primer tails and cycling conditions: PCR Fundamentals for Cloning.
Part II: Restriction cloning and cloning PCR products
The oldest methods, and the ones to use when your construct is simple or your budget is limited.
Traditional restriction cloning
Traditional cloning is the cut-and-paste method most people meet first. You digest a vector and an insert to make compatible ends, then join them with T4 DNA ligase.
It is still the cheapest way to move a gene from one plasmid to another. Standard enzymes and a ligase cost a fraction of a proprietary assembly mix, which keeps the cost down on routine subcloning.
Enzyme choice is the constraint. You are limited to the sites that already exist where you need them, and they have to be unique. A site in the middle of your gene will cut your insert in two. Using two enzymes that leave incompatible ends sets the insert's orientation and reduces vector self-ligation, although not every pair of enzymes gives you incompatible ends, so check the overhangs.
Restriction cloning also leaves a few base pairs of scar at each junction. This can shift a reading frame or add unwanted amino acids to the protein you are trying to express.
How to prevent the common ligation failures: Traditional Restriction Cloning.
Cloning PCR products: TA and blunt-end
Sometimes you only need a PCR product in a stable plasmid, for sequencing or for storage. Designing restriction sites or assembly overlaps for this is more work than the job needs.
TA cloning uses a quirk of standard Taq polymerase, which tends to add a single adenine to the 3' end of its products. You mix the product with a linearized vector carrying matching single thymine overhangs, and the ends pair up for ligation.
Simple does not mean reliable. Both ends of the insert carry the same overhang, so it can go in either orientation. For storage this does not matter. If a later experiment needs the insert to face a particular way, you will have to screen clones until you find one that does.
A high-fidelity proofreading polymerase leaves blunt ends instead. You can clone these into a blunt-cut vector, or add A overhangs first and use TA cloning. Blunt ends have nothing to hold the fragments together while the ligase works, so expect a background of empty vectors, inverted inserts and concatemers, meaning several copies joined end to end.
For both routes, budget time for screening. A five-minute reaction can still leave you with a week of screening to find the right clone facing the right way.
Choosing between them, and the phosphorylation step that is easy to miss: Cloning PCR Products: TA, Blunt-End and Phosphorylation.
TOPO cloning
Thermo Fisher's TOPO system is the option to consider when you are short of time. It uses topoisomerase I, an enzyme that cuts and rejoins DNA, so there is no separate ligation step and no ATP.
The kit supplies a linearized vector with the enzyme already bound to its ends. The joining reaction can take five minutes, although DNA preparation, transformation, colony growth and verification still take as long as they always do.
TOPO kits come in TA, blunt and directional versions, and they are not interchangeable. Directional TOPO needs a CACC sequence at the 5' end of your forward primer, so you have to choose the system before you amplify.
The trade-offs are the per-reaction cost, which is high compared with other cloning methods, and the limited range of vectors.
How the enzyme works, and when the premium is worth paying: TOPO Cloning: How It Works and When to Pay the Premium.
Part III: Seamless and multi-fragment assembly
These methods join fragments through matching sequences you design, rather than through sites you happen to have. They are what made multi-part constructs practical.
Gibson Assembly
Gibson Assembly joins fragments in a single reaction at one temperature. Three enzyme activities work together: an exonuclease digests the 5' ends to expose single-stranded 3' ends, which anneal to their matching partner, a polymerase fills the gaps, and a ligase seals the nicks.
Diagrams make multi-fragment assembly look easy. In practice, every extra fragment adds a junction that has to find the right partner, so efficiency falls as the fragment count rises. The exposed single strands can also fold back on themselves, or pair with a short unintended match somewhere else in the set.
Before you change the enzyme mix, check your design and starting material. Does each overlap match only its intended partner? Have you removed template DNA, unwanted fragments and salt? Are the DNA amounts right for the number of fragments? For a difficult construct, building it in two stages is usually faster than optimizing a single reaction.
Overlap design and the failures that look like enzyme problems: Gibson Assembly: Primer Design and Troubleshooting.
NEBuilder HiFi DNA Assembly
Gibson Assembly made overlap-based cloning routine. NEBuilder HiFi uses the same strategy with a different enzyme formulation: an exonuclease processes the fragment ends, a polymerase fills the gaps, and a ligase seals the nicks.
NEB reports higher fidelity than its own Gibson mix, better performance on complex assemblies, and the ability to correct certain mismatches at fragment ends. In daily use it is the last of these that researchers notice, because a few leftover bases at a fragment end do not necessarily stop the reaction.
None of this changes your design. The overlaps still have to be unique, the fragments still have to be clean, and the amounts still have to be calculated in moles rather than microliters.
When you are choosing between Gibson and HiFi, compare the recommended conditions for the specific mix, what your lab already has experience with, and what the reagents cost you now. These methods share a mechanism, so it is tempting to carry a protocol across from one to the other. Do not. The overlap lengths and the insert-to-vector ratios differ between them, and in HiFi the ratio also changes with the number of fragments.
Design rules as NEB publishes them, and three common failures: NEBuilder HiFi Assembly: Primer Design and Troubleshooting.
Golden Gate Assembly
Golden Gate uses a Type IIS enzyme such as BsaI or BsmBI, which recognizes one sequence and cuts outside it. The cut leaves a four-base overhang, and because you choose those four bases, you control which part joins to which.
A correct assembly destroys the recognition site, so the enzyme cannot cut the finished product again. This is what lets you run digestion and ligation in the same tube while the correct product accumulates and the wrong products get cut apart and tried again.
This is good for modular work. You can prepare promoters, coding sequences, tags and terminators as reusable parts, then combine them in different arrangements with no new primer design.
The cost is design effort rather than reagents, because every junction needs its own overhang, none of them palindromic, and no part may contain an internal copy of the enzyme's recognition site.
Overhang rules and what to do when a one-pot reaction stalls: Golden Gate Assembly: Overhang Design Rules and Multi-Part Troubleshooting.
In-Fusion cloning
Takara's In-Fusion is also an overlap method, and Takara reports that it needs less homology and less time than Gibson. An exonuclease activity in the mix exposes single-stranded ends, which anneal through matching sequences, and the final bonds are made by the bacteria after transformation rather than in the tube.
You can join one or more inserts to a linearized vector without needing particular restriction sites at the junctions. The overlaps decide where each fragment goes. Takara recommends 15 bp overlaps for a single insert and 20 bp for multiple inserts, positioned at the very end of each fragment.
Pay attention to the vector. Any uncut plasmid that survives into the transformation gives you background colonies, and because the product is nicked when it enters the cell, the method can be more sensitive than others to the state of your competent cells.
Overlap design, and the elution buffer mistake that gives you nothing: In-Fusion Cloning: Primer Design, Seamless Assembly and Troubleshooting.
Part IV: Recombination-based cloning
Gateway cloning
Gateway is a recombination system adapted from lambda phage biology. It uses defined att sites and two enzyme mixes, BP and LR, to move DNA between compatible vectors without restriction enzymes or a ligase.
Its strength is reuse. Once you have a sequence-verified entry clone, you can move your gene into many compatible destination vectors in parallel, for mammalian expression, fluorescent tagging or protein purification, without designing anything new. MultiSite Gateway extends this to ordered combinations of several fragments.
This saves a lot of work when one gene has to go into several vectors. With restriction cloning, five vectors means designing primers, choosing enzymes and sequencing the result five times over. With Gateway, the design and the sequence verification happen once, in the entry clone.
The drawbacks are the cost of the Clonase mixes, the need for compatible vectors, and the att scars, about 25 bp, left flanking your gene in the final construct. If the cassette is expressed, those scars are translated. Check where they sit and what they add to the protein before you commit.
Why BP and LR only run one way, and the ccdB trap: Gateway Cloning: Understanding BP and LR Reactions.
Part V: Cloning for CRISPR experiments
CRISPR guide RNA design
CRISPR-Cas9 changed what is possible in genome editing, and a lot of plasmid cloning is now used for it. Researchers clone guide sequences into expression vectors and build plasmids that carry both the Cas protein and the guide.
The outcome of an edit depends heavily on the guide. You need a suitable target and a PAM site, meaning the short sequence next to the target that the Cas protein requires, a check for off-target sites elsewhere in the genome, and, for plasmid delivery, ends and orientation that match your vector. A guide that looks fine in a sequence viewer can fail for reasons that have nothing to do with the cloning, such as a run of four thymines, which can end transcription from a U6 promoter early.
Some workflows skip plasmid cloning and deliver the Cas protein and guide RNA together as a ribonucleoprotein complex. Guide choice, delivery method, cell type and culture conditions all affect what you get.
Guide selection, the seed region and off-target avoidance: CRISPR gRNA Design: Best Practices and Off-Target Avoidance.
Plan it first
Every method on this page has the same failure underneath: the reaction worked, the colonies grew, and the construct is not what you meant to build.
Before you prepare any DNA, check the insert orientation, the junctions, the overlaps or overhangs, and the reading frame across each join. A design problem is far cheaper to find now than after two rounds of colony screening and a sequencing bill.
A careful sequence check still cannot tell you everything. Insert toxicity, plasmid instability and host compatibility affect whether you can propagate and use a construct, and none are visible in the sequence.
In Photo51 you can lay out the vector and inserts, map the overlaps or overhangs onto them, and inspect the finished construct and its junctions before you order a single oligo. The guides linked above cover the methods and their common problems. The Photo51 how-to guides and video tutorials show you how to carry out supported design tasks in the software.
If you are still choosing a design tool, see The 7 Best SnapGene Alternatives in 2026 and our head-to-head comparison in Photo51 vs. SnapGene vs. Benchling.