Restriction enzyme list: sites, overhangs and compatible ends

Which ends ligate to which

Two sticky ends anneal when their overhangs are the same kind, both 5′ or both 3′, the same length, and complementary to each other. An overhang written 5′ to 3′ therefore pairs with its own reverse complement, which is why a palindromic overhang such as GATC pairs with itself. A 5′ overhang never pairs with a 3′ one, whatever the letters: KpnI and Acc65I cut the same site, GGTACC, and leave the same four bases, but one leaves them on the 3′ side and the other on the 5′ side, so their ends do not go together.

Below are the 15 groups in this list whose members leave ends that ligate to each other. Joining two enzymes from one group usually destroys both recognition sites, which is the point of the trick: the insert cannot fall back out. The exceptions are listed under each group, and most of them are enzymes with a short site that still finds it inside the join.

Blunt ends · 34 enzymes

Any blunt end ligates to any other blunt end, so these 34 enzymes form one group with no sequence to match: AfeI, AleI, AluI, BsaAI, BsaBI, BmgBI, BsrBI, BstZ17I, DraI, Eco53kI, EcoRV, FspI, HaeIII, HincII, HpaI, MscI, MspA1I, NaeI, NruI, PmeI, PmlI, PsiI, PvuII, RsaI, ScaI, SfoI, SmaI, SnaBI, SrfI, SspI, StuI, SwaI, XmnI, ZraI. The join is less efficient than a sticky one and the insert can go in either way round.

Overhang set by the flanking sequence · 19 enzymes

These cut outside their recognition site, so the overhang is whatever the neighbouring bases happen to be, not a property of the enzyme. That is what Golden Gate assembly is built on: you choose the overhangs. They are listed with no group for that reason: AcuI, BbsI, BccI, BceAI, BciVI, BpmI, BsaI, BseRI, BsmBI, BsmI, BspQI, BsrDI, BtgZI, BtsI, EarI, Esp3I, SapI, BfuAI, PaqCI.

Overhang depends on the site · 35 enzymes

These recognise a site with an ambiguous base inside the overhang, so the four bases left behind depend on the particular site that was cut. AvaI, for example, leaves 5′ YCGR, which is CCGA, CCGG, TCGA or TCGG depending on the sequence. Two of them ligate only when the real overhangs work out complementary: AccI 5' MK, AflIII 5' CRYG, AhdI 3' N, AlwNI 3' NNN, AvaI 5' YCGR, AvaII 5' GWC, BaeGI 3' KGCM, BanI 5' GYRC, BanII 3' RGCY, BglI 3' NNN, BlpI 5' TNA, BsiEI 3' RY, BsiHKAI 3' WGCW, BsoBI 5' YCGR, Bsp1286I 3' DGCH, Bpu10I 5' TNA, BstAPI 3' NNN, BstEII 5' GTNAC, BstXI 3' NNNN, Bsu36I 5' TNA, BtgI 5' CRYG, DraIII 3' NNN, DrdI 3' NN, EcoNI 5' N, EcoO109I 5' GNC, PasI 5' CWG, PflFI 5' N, PflMI 3' NNN, RsrII 5' GWC, SexAI 5' CCWGG, SfiI 3' NNN, SmlI 5' TYRA, StyI 5' CWWG, Tth111I 5' N, XcmI 3' N.

How this table is made

Every column is worked out from the recognition site and the two cut offsets, which is all the enzyme list holds. The site is written REBASE style, with ^ where the top strand is cut, or as GGTCTC(1/5) when the cut falls past the site. The overhang is the stretch between the two cuts, the end is 5′ when the top strand is cut first and 3′ when the bottom strand is, and the group is the overhang together with its reverse complement so that complementary ends land in the same group.

Each join shown under a group is the left half of the first site up to its cut, followed by the right half of the second. The site finder then runs on that joined sequence for both enzymes. Only a site lying wholly inside the join can be decided this way: bases further out belong to the rest of your construct, so a site that would need them is neither claimed nor ruled out here.

Dam and Dcm methylation are not taken into account here. The PlasmidLab app can mark the sites they block. Nothing on this page knows about buffers, reaction temperatures or how well an enzyme cuts near the end of a fragment, because the list does not carry that.

Recognition sites and cut positions from REBASE, the Restriction Enzyme Database (rebase.neb.com), release 611. This is the same enzyme list the PlasmidLab app uses. Everything runs in your browser; the sequence is not sent anywhere.