In traditional DNA-DNA hybridization, for example in southern blotting, two different DNA pieces are recombined together with their complementary regions. We reasoned that with more hybridization events occurring, more DNA pieces can be recombined together for construction purpose. Based on this assumption we designed our SHA method. As shown in Fig. 1 every substrate fragment (SF) is designed to have its 3′-half overlapped with the 5′-half of the next SF. As these overlaps (halves) can cover about 1/3 to 2/3 of each SF, after denaturation and renaturation, there are good chances for inter-fragment hybridizations. Then a circular plasmid can form and be ready for transformation. Nicks and gaps in it can be repaired in the bacteria cells.

To generate the two-half structure, SFs are usually constructed by joining two DNA elements together using OE-PCR. For example, in Fig. 1, kan (selectable marker), ori (replication origin) and gene1–4 are first PCR-isolated from their initial sources. Then each two neighboring ones are joined together by OE-PCR to generate SF kan-gene1, SF gene1-gene2, SF gene2-gene3, SF gene3-gene4, SF gene4-ori. SF ori-kan can be simply PCR-amplified from a vector plasmid and work as a vector backbone. In practice we also successfully got SFs by tailed primer PCR or chemical synthesis (illustrated in Fig. 2).

The core step of our method, hybridization, was carried out simply with a boiling water bath treatment (see Materials and Methods). To see the impact directly, we analyzed the DNAs by gel electrophoresis. As shown in Fig. 3 after the hybridization treatment, SFs were combined into larger structures.

As a proof of concept and to evaluate the assembly efficiency, we first constructed three plasmids, pOKCA, pOKC2μUA and pOKCA2 from four, six and eight SFs respectively. Plasmid structures, overlapped SFs and the overlap lengths are illustrated in Fig. 4A. The genetic elements assembled were: O, the ori from pET28a (originally from pBR322); K, kanamycin resistance gene kan; C, chloramphenicol resistance gene cat; A, ampicillin resistance gene amp; 2μ, 2μ replication origin; U, yeast ura3 gene. Most SFs were prepared by jointing two elements together by OE-PCR. SFs containing 50 bp overlaps were prepared by tailed primer PCR (details of SF preparations were illustrated in Fig. S1). After mixing, DpnI digestion and hybridization, 5 microliters of SFs were transformed into DH5α chemical competent cells. The results were summarized in Fig. 4 B. As shown, SHA was able to assemble up to eight SFs with overlaps down to 50 bp with sufficient efficiency. In contrast, mixture without hybridization treatment gave no or only a few colonies which were traced to undigested plasmid.

By varying overlap length we found that the longer overlap used the higher assembly efficiency was achieved, and no overlap led to no assembly (Fig. 5).

A vector is typically composed of a replication origin, a selectable marker, multiple cloning sites and some other elements which can endow it with different functional traits, such as copy number, maintenance stability, induction method, and expression intensity. Existing vectors usually can't exactly meet all the demands for a specific task. So it is very desirable if we can pick every element we need and assemble them into a vector right before using.

Here we constructed a customized vector, pJXL (Fig. 4), for B. subtilis metabolic engineering within two days. It was consisted of four DNA pieces: the Amp and ori of pBR322, the 3 kb replication region of plasmid pBS72 [4], [5], and the chloramphenicol acetyltransferase gene and spaS promoter from plasmid pNZ8901 [6]. Two SFs were chemically synthesized; the other two were prepared by OE-PCR (details were illustrated in Fig. S1).

The function of constructed pJXL was demonstrated by controlled production of enhanced green fluorescent protein reporter (Fig. S3).

Previously, a similar pathway was constructed for terpenoids production by two rounds of traditional cloning and OE-PCR [7]. Here using SHA, we finished the construction, named pTRICLow (Fig. 4A), within two days, including the time for SFs preparation. Genes encoding the last four enzymes of the MVA pathway, ERG12, ERG8, ERG19, and IDI1, were amplified from Saccharomyces cerevisiae genomic DNA individually; two vector fragments, B5′ fragment and B3′ fragment, were amplified from expression vector pTrcHis2A. From them six SFs were prepared by OE-PCR. Three RBS-containing sequences of 45 bp each were added to separate and organize all 4 genes into a single synthetic operon (details about SF preparations were illustrated in Fig. S1). The assembly efficiency was sufficient but lower than expected (Fig. 4B). To investigate the reason, we retransformed DH5α cells with the constructed pTRICLow (in covalently closed circular form). The efficiency was about two orders of magnitude lower than normal plasmids with similar sizes. Thus this plasmid was difficult to construct inherently. Similar phenomenon was also reported for traditional cloning methods [8].

pTRICLow, was then cotransformed with pACY-IspS4 (coding for a modified Populus alba isoprene synthase, constructed previously in our lab by traditional cloning, see Method S1 for details.) into E. coli BL21 (DE3). Isoprene production was confirmed by GC-MS. (see Fig. S3)

During the above constructions, we designed the SFs (by choosing the primer positions, examples were shown in Fig. 4A) in a way that after the hybridization nicks formed between SFs. It was previously reported that, like nicks, gaps (of 6-nt, in the LIC method) could also be repaired by the cells [9]. Here, we thought our method should be able to tolerate even longer gaps, since it could stick the fragments together more stably (the complementary regions were much longer than in LIC method). The benefit of this ability would be making the primer designing easier and providing more flexibility.

To test it, we constructed a synthetic acetone pathway, pAcetone, with gaps of 110, 54, 58 and 12 nt formed after the hybridization (Fig. 4A). This pathway was derived from the isopropanol production pathway first designed by Hanai et al [10]. thl and adc were amplified from Clostridium acetobutylicum ATCC 824 genomic DNA individually; atoA and atoD were amplified from E. coli K12 genomic DNA together; the recipient vector was amplified from plasmid pET28aΔlacI which was reduced from pET28a (see Method S2 for details). Then by OE-PCR, they were joined into 4 SFs. Similar to what we did in pTRICLow construction, two RBS-containing sequences were inserted between the genes (details was illustrated in Fig. S1).

In order to evaluate the impact of the long gaps, we reconstructed this plasmid in a no-gap way (Method S3). The results demonstrated that our method still held 76% of its efficiency with gaps comparing to the no-gap control.

Then pAcetone was transformed into E. coli BL21 (DE3). Acetone production was confirmed by GC-MS (Fig. S3).

We sequenced all the constructs, total of 51.9 kb. Two mutations were found to be introduced by the primers used and one mutation was introduced by DNA polymerase during strand extension (see Table S1 for details). This high fidelity was achieved by applying high-fidelity PCR mix and avoiding gel purification in which DNA might be damaged by the UV light. Hybridization itself isn't likely to generate any mutation.

In contrast to previous methods, no faulty recombination was found for SHA according to the restriction mapping (Fig. S2) and DNA sequencing results. This should owe to the special mechanism of SHA, which relies on the recognition between long DNA sequences compared with previous methods relying on the recognition between short sequences (usually no longer than 50 bp).

To compare them more directly, we attempted to construct the pAcetone using SLIC. This was only a 4-fragment assembly. Every fragment had two 50 bp homologous ends that mediate the construction. RBS together with start codon, total of 18 bp, repeat twice in the homologous areas. In the previous study, SLIC had a faulty recombination rate of 0% for 5-fragment assembly, but 83% for 10-fragment assembly [2]. Here we found that the short 18 bp repeats made the situation even worse. As shown in Fig. 6, out of ten random picked transformants only two were assembled correctly. Seven lost acoAD insert; one lost all three inserts (it was distinguished from pET28aΔlacI background by that it contained a residual insert corresponding to the first 28 bp of thl).

E. coli DH5α was used for SHA and E. coli BL21 (DE3) was used for gene expression.

We used Luria-Bertani broth for culturing all our strains. Agar (1.5%) was added to make solid medium. Antibiotics were used if necessary as follows: ampicillin (Amp, 50 µg ml⁻¹), kanamycin (Kan, 50 µg ml⁻¹), and chloramphenicol (Cm, 30 µg ml⁻¹ for E. coli and 5 µg ml⁻¹ for B. subtilis). IPTG was used to induce gene expression at 0.5 mM.

Pyrobest™ DNA polymerase from TAKARA was used for SF preparation. EasyTaq™ DNA polymerase from Transgen was used for colony PCR screening.

OE-PCR was performed by the version of Shevchuk et al. [11], except that Pyrobest™ DNA polymerase was used throughout. OE-ends had lengths of 30–50 bp. By following their two tips we can always obtain high product yield and purity on the first attempt. First, use the two-step model including an overlap extension step without primers (step A) and an exponential amplification step with primers (step B); second, use nested primers, i. e. in step B the primers (called wing primers or flanking primers) should be within the product from step A, and have a distance above 50 bp from the ends. The PCR mixtures (each 50 µl) were made according to manufacturer's instructions: 0.2 mM each of dNTP, 1×Pyrobest Buffer II, 0.025 U µl⁻¹ Pyrobest™ polymerase, 0.4 µM each of primers (for step B) or no primers (for step A), and template. For step A, two unpurified PCR-isolated elements were used as template at a final concentration of 5 ng µl⁻¹ each; for step B, 2 µl of unpurified product from step A was used as template. Cycling parameters: for step A, initial denaturation 94°C 2 min, subsequent steps 94°C15 s, annealing at 56°C 20 s, extension 72°C 2 min, 10 cycles total, hold at 4°C; for step B, initial denaturation 94°C 2 min, subsequent steps 94°C 15 s, annealing at 62°C 20 s, extension 72°C 4 min, 35 cycles total, final additional extension 72°C 3 min, hold at 4°C. PCR templates and primers are all given in Table S2; primer sequences are given in Table S3. Plasmid templates were linearized with appropriate restriction enzymes before using (Table S2.).

Tailed primer PCR was performed similarly to regular PCR, except that the primer had a 5′ end overhang. In this case (pOKCA2 construction), SFs were purified by E.Z.N.A.™ Cycle-Pure kit (OMEGA) to exclude residual primer oligonucleotides whose lengths were comparable to that of the overlaps. In other cases, purification was omitted.

DNA chemical synthesis was conducted by the Shanghai Generay Biotech Co.

Prepared SFs were checked by agarose gel electrophoresis. Roughly equal molar amounts of SFs (about 200 ng µl⁻¹ each) were mixed in a 1.5 ml polyethylene tube, treated with DpnI at 37°C for 3 h to digest the plasmid templates, then sealed with Parafilm for water-proof and submerged in a beaker of boiling water (about 97–100°C), left there, till the water cooled to room temperature (18–25°C) which usually took about 2 hours.

The reaction was transformed into E. coli competent cells with standard heat shock method [12].

Plasmids were isolated from overnight cultures using E.Z.N.A. ™ Plasmid Mini Kit from OMEGA, subjected to restriction enzyme digestion, and then loaded to 1% agarose gels to check their restriction digestion patterns by DNA electrophoresis. Restriction enzymes used for every plasmid were listed in Fig. S2.