An innovative platform for quick and flexible joining of assorted DNA fragments

Successful synthetic biology efforts rely on conceptual and experimental designs in combination with testing of multi-gene constructs. Despite recent progresses, several limitations still hinder the ability to flexibly assemble and collectively share different types of DNA segments. Here, we describe an advanced system for joining DNA fragments from a universal library that automatically maintains open reading frames (ORFs) and does not require linkers, adaptors, sequence homology, amplification or mutation (domestication) of fragments in order to work properly. This system, which is enhanced by a unique buffer formulation, provides unforeseen capabilities for testing, and sharing, complex multi-gene circuitry assembled from different DNA fragments.


Results
The framework of TNT-cloning system. We conceived and developed a cloning platform that adopts a truly universal entry vector (pSTART) to carry all DNA elements to be joined by reiterative digestion/ligation steps using two families of assembling vectors, called alpha (α ) and omega (Ω ), which are capable of defining the order and orientation of each DNA element desired in the final construct (Fig. 1). Such element organization is determined by specific signatures (1, 2, 3, 4, 1R and 2R) left by the type IIS enzymes chosen, EarI and LguI, that allow, a) an ORF compatible 3 nucleotide (nt) overhang for cloning, b) up to three elements to be combined at once per round of assembly, and c) the pSTART to be used as destination vector to make new assemblies an entry element in the library, maximizing exchangeability (Fig. 2, Supplementary Fig. 1).
First, elements are either amplified or synthesized to include signatures "1" and "2" at the borders and cloned in the pSTART vector to build the universal library (Fig. 2a). The pSTART receives and releases the desired fragments with either EarI or LguI enzyme. Once elements are cloned, they are transferred and further combined in either alpha (α) or omega (Ω) vectors, which receive elements upon cleavage with EarI/LguI and release fragments upon cleavage with LguI/EarI, respectively (Fig. 2b,c). Upon digestion of each plasmid, a set of "signatures" that were specifically arranged to direct and orient the desired fragments are exposed (Fig. 2b, Supplementary Fig.  1). The signatures "1" and "2" are always flanking the inserts released from pSTART and are always used to join the final constructs into any α or Ω member. At the same time, the signatures "3" and "4" will be used by a specific member of each family (α and Ω) to join fragments between themselves, two fragments at once (binary assembly) using the members α1A and α2 (or Ω1A and Ω2) and three fragments at once (tertiary assembly) using the members α1A, αB and αC (or Ω1A, ΩB and ΩC) (Fig. 2b, Supplementary Fig. 1). To change the fragment orientation (sense or anti-sense) simply switch the chosen α or Ω version for its respective "R" version during the cloning step, no adjustments are necessary (Fig. 2b, Supplementary Fig. 1). The enzyme location and the signatures were designed to permit a pre-established cloning setup and to allow each final construct to be used as an insert in case a following round of cloning is needed, creating a cloning loop that can be repeated over and over, alternating α and Ω members, in order to join multiple fragments into one larger construct. For example, 27 hypothetical fragments can be customized into one single insert at any combination through just 4 cloning rounds (Fig. 1d). The (a) Fragments of interest (element) can be produced by gene synthesis ("GS") or amplified by PCR using the sequence shown (plus the three nucleotide code for signatures 1 and 2) to be inserted in the pSTART vector by either restriction enzymes (EarI/LguI) or "Gibson" isothermal assembly (which requires previous linearization of pSTART). pSTART carries signatures 1 (yellow) and 2 (green) used to transfer the fragment from the library to any member of either α family (using EarI, green arrow) or Ω family (using LguI, red arrow), (see also c). (b) Exemplification of a three fragment assembly after cloning the elements Pro, CDS and term in the library (pSTART) as shown in a. Either α or Ω versions 1A, B and C are individually (➊) used as destination vectors for Pro, CDS and term, respectively, generating the constructs Pro(α/Ω1A), CDS(α/ΩB) and term(α/ΩC) (➋). Following, all three constructs are now used as entry clones and combined with a new destination vector in one single tube (➌; any Ω /any α , respectively). Signatures 3 (red) and 4 (blue) will be used to join two (only signature 3) or three (both signatures 3 and 4) fragments together. Alternatively, the destination vector can be the pSTART, in case a construct is desired as an element in the library. After two rounds of cloning, all three elements were joined without the need of adaptors/linkers or homology between sequences. Versions "R" in both families were created to allow fragment reorientation (sense or anti-sense) (e.g., ➍), no other adjustment is needed. (c) Detailed positioning of each enzyme in pSTART, α and Ω members. (d) Transfer of 27 hypothetical elements from pSTART into either α or Ω family members using the "TNT-cloning loop". Elements are first transferred to α members (in this example) and combined, three at a time, to generate 9, 3 and finally 1 single insert after only 4 cloning steps (equals to 5 days if a fast growing E. coli strain is used, e.g. Mach1 TM -T1 R or T7Express). detailed representation of each vector member and the 3 nucleotides sequence of each signature followed by an ideogram (with a timeline included) are depicted on Supplementary Fig. 1 and Supplementary Fig. 2, respectively.

Development of EarI as a useful enzyme: methylation sensitivity.
To date, all type IIS enzymes that leave a 3nt overhang and therefore are suitable for use in our TNT-cloning system recognize either 5′ CTCTTCN ▼ NNN ▲ 3′ (e.g., EarI) or 5′ GCTCTTCN ▼ NNN ▲ 3′ (e.g., LguI) sequences 21 (Supplementary Table 1), leaving the EarI recognition site nested within the LguI site. To overcome this limitation we assessed EarI sensitivity to different methyl groups added either within or nearby the 5′ GCTCTTCN ▼ NNN ▲ 3′ sequence (EarI was chosen over Eam1104I due to previous reports on methylation sensitivity 21 ). We used three methyltransferases (M), M.SacI, M.SssI (2 sites) and M.TaqI to methylate, respectively, the cytosines at the positions 2/1 (forward/ reverse strand), 7/8 or -1/1 and the adenines at the positions 9/6 ( Fig. 3a). For this purpose, we used a 6,435 bp plasmid (pET-28-M.SacI) and different 1055 bp PCR fragments carrying at least two sites for the restriction endonuclease where at least one site would not be subjected to methylation (except for M.SssI where both sites were addressed simultaneously). Two distinct methylation sites generated by M.SssI had little (M.SssI-1) to no (M.SssI-2) effect in EarI ability to cut the modified DNA (Fig. 3b,c). On the other hand, sensitivity tests showed that M.SacI and M.TaqI inhibited the enzyme activity by 83.4% (SE ± 5.4) and 99.9% (SE ± 0.03), respectively (Fig. 3b,c). Because M.TaqI was highly capable of inhibiting digestion of DNA by EarI, we adopted this modification to design the TNT-cloning system with, a) the first nucleotide of each signature that flanks the restriction site starting with an adenine, and; b) such modification present only when EarI is the first enzyme to be used (i.e., α members).
To avoid the cost and time of performing in vitro modifications of the α members before cloning, we engineered the genome of the E. coli strain T7Express (T7X) to be capable of expressing the M.TaqI gene during its regular life cycle ( Supplementary Fig. 3). Different conditions for growing the engineered strain (T7X.MT), while keeping maximum DNA methylation, were tested ( Supplementary Fig. 4) and the optimal practice is shown in Fig. 3d, where 97.1% (SE ± 0.8) of the plasmid DNA extracted from T7X.MT was unable to be cut by EarI. Our results show the use of this strain is comparable to the modification levels obtained for the in vitro methylation. Methylated DNA extracted from T7X.MT remains stable at -15 °C for at least 11 weeks without compromising EarI/Eam1104I inhibition ( Supplementary Fig. 4). There is no methylation requirement for both the Ω members and downstream cloning steps in the α members, and therefore, any construct generated using the TNT-cloning system can be transformed in the strain of choice (T7Express must be used to allow for white/blue screening). As a consequence, it is not necessary to define LguI sensitivity to methylation; however, we report a sensitivity chart for three isoschizomers in this class: BspQI, LguI and SapI ( Supplementary Fig. 5). Importantly, LguI is also sensitive to M.TaqI modification and yet this is irrelevant because M.TaqI site is not present at a critical position on Ω members and transformation of constructs carrying the joined fragment(s) in the α members is not required for T7X.MT. Consequently, the T7X.MT strain is useful for propagation of the original TNT-plasmids but problematic for downstream cloning purposes.
Our results show that we successfully engineered two distinct sites to support the assembling loop presented, with LguI recognizing and cleaving at the sequence 5′ GCTCTTCN ▼ NNN ▲ 3′ and EarI recognizing and cleaving at the sequence 5′ CTCTTCN ▼ NNN ▲ 3′ (but not at the sequence 5′ GCTCTT * CN ▼ N * NN ▲ 3′ , where T * /N * represent a methylation of the corresponding adenines). By using the engineered E. coli strain our required modification is simple to implement.
To evaluate the effect of fragment length on the efficacy of our system we took the Fused control (≈ 4 kb), the P2AF2A cluster (≈ 4 kb) and the IbpF2A cluster (≈ 4 kb) in Ω1A, ΩB and ΩC, and used a tertiary assembly to generate a ≈ 12 kb fragment in α1A (Fig. 4d). Additionally, we developed an efficient protocol along with an improved buffer system (called TNT-Buffer; 50 mM Tris-HCl pH7.5, 2 mM DTT, 10 mM MgCl 2 , 1 mM ATP and 2% PPG) that allowed EarI and LguI enzymes to work well in combination with T4 DNA ligase in a "one-pot-reaction" (Fig. 4e). When using 75 ng of entry plasmid DNA (75 ng each entry plasmid for multiple elements assembling) and 50 ng of destination vector (TNT-members α, Ω or pSTART) a variety of fragment sizes (from 36 bp to 2.7 kb) could be efficiently cloned. The average number of positive clones retrieved from 1, 2 or 3 elements cloning using the TNT-buffer were, respectively, 12.2 × 10 4 (SE ± 16.2%), 6.1 × 10 4 (SE ± 25.4%) and 3.0 × 10 4 (SE ± 35.2%), if full ligation reaction is applied. Importantly, the accuracy, which is the number of positives clones among all clones retrieved in a plate, were ≈ 100%, ≥ 83.3% and ≥ 81.2% when 1, 2 or 3 elements were being cloned, respectively. Differently, the analogous reactions performed using the regular T4 DNA ligase buffer, the LR Reaction from the Gateway system or the isothermal (Gibson) assembly for cloning 1 element retrieved, respectively, 2.6 × 10 4 (SE ± 16.3%), 4.7 × 10 4 (SE ± 22.3%) and 7.1 × 10 4 (SE ± 4.4%) positive clones, if full ligation reaction is applied. We followed the manufacturer's instructions for each method and all three showed ≈ 100% accuracy. The EarI/LguI/T4 ligase enzymes concentration were very important, especially for accuracy and a standard 10 μ l final volume TNT-reaction includes 40 U of T4 DNA ligase plus either 5 U of EarI or 0.5 U of LguI. Regardless of the buffer system, the LguI enzyme showed some promiscuity over the 5′ aaCTCTTC3′ EarI site originally included in the Ω vectors and four point mutations upstream of the biding site (from aa into tt, gt and cc) were tested and finally changed to 5′ ccCTCTTC3′ in order to achieve such efficiencies ( Supplementary Fig. 1, Supplementary  Fig. 7). These results show other cloning strategies available scored less efficient than the TNT-system and the TNT-buffer is up to 18-fold more efficient than the T4 DNA ligase buffer (Fig. 4e). The key component of our buffer is a branched polyethylene glycol (PPG) that appears to allow efficient digestion/ligation while maintaining efficient exchange of inserts between vectors. Since the isothermal (Gibson) assembly 12 also allows for multiple fragments cloning, we also compared 2 and 3 elements cloning using both methodologies --the one-pot-reaction in TNT-buffer (50 cycles of: 34 °C for 45 sec and 16 °C for 4.5 min) or the 1 h Gibson assembling reaction (at 50 °C) (Fig. 4e). Both methods performed well for 2 or 3 elements cloning and the TNT-buffer respectively retrieved Error bars are standard error between 3 and 7 biological replicates. TNT-reactions cover both ways (α→ Ω and Ω→ α) and include both aaCTCTTC and ccCTCTTC Ω's (see Supplementary Fig. 7). Positive clones are confirmed by PCR (16 < n < 32). Same lot of competent cells was used. Bonferroni, p < 0.01(letters).
Taken together, our results show that the TNT-cloning system is a powerful tool for flexible, rapid and all-in-one efficient assembling of various DNA fragments requiring no homology or linker/adaptors between fragments. The ≈ 12-kb proof-of-principle fragment noted above is an example of how 28 fragments from the library could be easily designed and joined into a single insert using 5 cloning steps. Because each construct generated is ready to be used as an entry clone for future assembling (and as an element in the library if cloned in the pSTART, Supplementary Fig. 1), such system is also remarkably versatile and convenient, requiring minimal to no re-cloning.
Overcoming the domestication step. Available type IIS restriction enzyme-based systems also hinder its application due to mandatory mutation steps necessary before DNA elements can be cloned and assembled. One solution already mentioned above is to domesticate an element by changing a 5′ CTCTTC3′ site(s) while maintaining its functionality. However, many elements cloned are not CDS and therefore this strategy cannot be applied. The unique TNT-buffer efficiently clone elements with a internal 5′ (G)CTCTTC3′ site (Fig. 5a), however, tertiary assembling involving non-domesticated inserts were complex and positive clones were not recovered. Therefore, we utilized the ability of DNA to form triplexes [27][28][29] in an effort to change the DNA-enzyme interactivity 30 and inhibit the type IIS digestion progress by masking specific 5′ (G)CTCTTC3′ sites using oligonucleotides.
To design such oligos, we adopted the Reverse-Hoogsteen orientation 27 , which allows for all four nucleotides to be part of the triple helix. Initially, we combined the ability of the intercalating dye acridine (Acr) to stabilize triple helixes with the modified oligonucleotide DNA/BNA NC (2′-O,4′-C-aminomethylene bridged nucleic acid), which has stronger binding affinity than DNA oligos (14 bp DNA/BNA NC Tm = 82.5 °C) and is more capable of forming triplexes at physiological pH (7.0-8.3) 29 . Increasing amounts of DNA/BNA NC oligo showed oligo-dependent inhibition of the digestion progress over the 675 bp PCR product template '8 m1' , suggesting inhibition of enzyme activity by a potential triplex formation ( Supplementary Fig. 8). On the other hand, this DNA/BNA NC oligo was not able to discriminate 5′ mismatches (3 in total) as observed by similar inhibition over the template '5 m2' , showing this oligo does not differentiate small mismatch changes as those found between internal and vectorial 5′ (G)CTCTTC3′ sites. Therefore, we decided to test two regular DNA oligonucleotides (26 nt and 26 nt-Acr) covering 11 nt upstream and 8 nt downstream of the 5′ (G)CTCTTC3′ site. A "digestion-progression curve" using LguI on the non-domesticated templates 8 m1 (0 mismatches) and '4 m1' (4 mismatches) in the absence or presence of 50 μ M of the 26-nt DNA oligo were performed to understand the kinetics involved in the digestion inhibition (Fig. 5b, Supplementary Fig. 8). We demonstrate the regular unmodified 26nt DNA oligo inhibited LguI activity at the desired site by 75.9% (SE ± 0.9%; 8 m1 template) and only 8.3% (SE ± 1.6%) when 4 mismatches (4 m1 template) were present, yielding a 'inhibition coefficient' of 67.6% (Fig. 5b,c). EarI had a slightly smaller inhibition coefficient of 49.9%, requiring a higher number of mismatches to discriminate specific 5′ CTCTTC3′ -containg sequences ( Supplementary Fig. 8). Expectedly, the 26-nt-Acr oligo showed stronger inhibition but intensely compromised specificity (Fig. 5c, Supplementary Fig. 8). These in vitro data were validated by cloning alternate single and multiple fragments containing up to 4 internal 5′ (G) CTCTTC3′ sites into different α and Ω vectors. Compared to the 7 m1 domesticated (Dom) element, the cloning of a 8 m1 non-domesticated (NoDom) element reduced the number of positive clones to 32.2% (SE ± 8.9%) but to only 74.0% (SE ± 7.3%) when the 26-nt oligo was previously incubated with the template plasmid, enriching the ability to recover positive clones by 231.5% (SE ± 10.9%) (Fig. 5e). Similarly, when a tertiary assembly is performed using three elements in a total of four internal 5′ (G)CTCTTC3′ sites (template 8 m1), the somehow equivalent number of positive clones had increased accuracy when previously incubated with the 26-nt oligo, going from 31.2% (SE ± 4.4) to 77.1% (SE ± 1.4), enriching the ability to recover positive clones by 240.8% (SE ± 31.1%) (Fig. 5e). Combined, these results show the oligo incubation efficiently and specifically created a manageable "blind-spot" to minimize enzyme activity over chosen 5′ (G)CTCTTC3′ sites while leaving the remaining (vectorial) 5′ (G)CTCTTC3′ sites reliably available for LguI/EarI to recognize and digest. Thus, the TNT-cloning system is excused from the domestication process and the probability of finding a fragment unsuitable for cloning are exceedingly rare.

Discussion
Building the first synthetic organism leveraged a set of tools for assembling DNA fragments both in vivo and in vitro that were effective 4,12 . These efforts, and the construction of subsequent genomes and specific gene cassettes, have utilized a linear approach that relies on sequence homology, limiting reuse, multi-combinatorial distribution and shuffling of key fragments necessary for iterative studies of genetic circuits. An alternative would be to adopt systems based on type IIS enzymes 18,31 ; however, current methods have limits in efficiency and efficacy. We associated an efficient buffer system with the placement of methyl groups in the type IIS enzymes binding site to generate two recognition sites for two distinct enzymes creating an innovative and flexible cloning platform, allowing for multiple elements (up to 3 at once) to be combined from a single universal library in a one-pot reaction with high efficiency and high fidelity (Figs 1-4). The ability to keep ORFs in frame by using cloning signatures that bear three nucleotide tag allowed us to include all cloning fragments, as CDS pieces, into a single universal library and, therefore, simplify assembling by orderly 'picking and mixing' the elements of interest. In this approach inversions were, and can be, easily performed by merely swapping the destination vector with its corresponding "R" version. Similarly, relocation of fragments was easily performed by rearranging intermediate cloning products rather than starting from the beginning of the process. Such characteristics are responsible for  (Table 1). Notably, the TNT-system rely on three nucleotides overhang that is finally incorporated in the constructs while some homology based methods, e.g. isothermal (Gibson) assembly, provide seamless joining of fragments. However, elements released from the universal library are always compatible with isothermal (Gibson) assembly due to a 5′ extension, rather than a 3′ extension, left by the EarI/LguI enzymes and therefore well-suited for scarless cloning.
The features found in the TNT-cloning system are key for establishing a easily transferable platform for quick determination of qualitative and quantitative gene fragment interactions that will have to be performed in studies involving gene sets and gene networks 8 . Currently, the validation of such networks and the reproducibility of data are limited by the inability of building various compatible multigene constructs from one flexible universal platform, requiring multiple methods to be adopted through a labor intensive pipeline 32 . The optimized TNT-cloning system and buffer, overcome such limitations providing a common platform for different elements from a single universal library to be orderly combined into 1 insert after a minimal number of cloning steps in a matter of days.
Within the context of synthetic biology, an important aspect for studies in regulatory networks and pathway engineering is the need of numerous regulatory sequences that may be incompatible with current cloning systems and/or limited in number, e.g. promoters. Here, we were able to provide a protocol that is greatly capable of cloning fragments bearing internal 5′ (G)CTCTTC3′ sites, supporting the use of such regulatory sequences (Fig. 5). Whether the assembly efficiency is compromised when large multigene constructs involving many undomesticated internal sites are involved remains to be tested. Such factor could also limit the exchangeability of the constructs; however, our approach is affordable and straightforward allowing for use of certain elements inapt for mutagenesis.
In addition, we demonstrate the programming of polycistronic mRNAs, a valuable tool for managing bistability/hysteresis 33 in genetic circuits as well as overcoming promoter shortage in multi-gene constructs (Fig. 4). We clustered different peptide 2As to overcome flaws found when only one sequence is used 24 by assuming a simple probability test should be applicable (if one copy gives 20% flaw, for example, two copies should reduce such number to 4%). We showed that such clustering corroborates our predictions, as P2AT2A and P2AF2A constructs gave almost flawless split between two CDS while their sole use show imperfect split in several cellular backgrounds 24,34 . This observation suggests that 10 genes can be grouped into one mRNA with approximately 97% [= (0.997) 9 ] efficiency of individually translated transgenes. However, the maximum number of genes capable of being practically clustered and the protein longevity due to the N-rule turnover 35 (first amino acid of the nascent peptide after efficient split is a proline) remains to be addressed. Nevertheless, the advantages intrinsic to polycistronic mRNAs further support the development of a methodology that allows an endless assembly with CDS compatibility.
In sum total, we have developed a new cloning platform, enabling gene circuits and pathway engineering and allowing for virtually any DNA fragment to be quickly, reliably and flexibly clustered and shared. Such a platform provides essential steps required for synthetic biology studies to progress faster and with high fidelity, even as DNA synthesis costs drop. Because of the ease of transferability of the developed platform, our system also contributes to the universality expected by the synthetic biology community and highlighted by several recent papers 14,19,36,37 .

Methods
Methylation tests. Type II cytosine-5 DNA methyltransferase protein sequence from Streptomyces achromogenes, which recognizes and modifies the sequence 5′ GAGCTC3′ (M.SacI; GenBank AAC97118.1), was reverse translated, synthesized (Supplementary Table 3), cloned in pET28 (pET28-M.SacI) by Gibson assembly (NcoI-SalI sites), transformed in T7Express and induced according to vector/strain suggested protocol (4 h, 0.5 mM IPTG). Expression of the ≈ 43 kDa protein was confirmed by protein gel (data not shown) and a second fraction of the same culture had the pET28-M.SacI plasmid extracted, quantified and 1 μ g was subject to incubation with BspQI, LguI, SapI or EarI in duplicates on manufacturer recommended buffer. Digestion ran for 1 h at 37 °C (except for BspQI, where 50 °C were used) using 5 U of each enzyme (except SapI, where 10 U was 8 m1 + 50 μ M oligo (dark red line; potential triplex) or duplex 4 m1 + 50 μ M oligo (pink line). The oligo specifically delays the digestion of desired 5′ (G)CTCTTC3′ site. Lines are linear trend (R 2 values shown) of ten (N.O.) and five (8 m1 and 4 m1) time points done in duplicates (see also Supplementary Fig. 8). (c) Gel image representative for b at 64.3% (SE ± 1.1%) digestion progress. Efficient oligo-dependent inhibition keeps the 675 bp PCR band intact (8 m1 inhibition, 45.4% SE ± 0.9; 4 m1 inhibition, 9.8% SE ± 0.5). 50 μ M § represents the 4 m1 template incubated with the 26nt-Acr oligo (inhibition = 48% SE ± 4.2). (d) Scheme of the "BlindSpot" protocol. Oligos were incubated with appropriate plasmid DNA carrying 8m1 (NoDom) fragments and subjected to partial digestion, ligation to linearized vector and transformation into E. coli. Colonies were then counted, confirmed for positive clones and recorded. (e) Enrichment for positive clones during both way cloning (EarI/LguI) of one (top) and multiple (bottom) non-domesticated fragments upon incubation with a 26nt DNA oligo (BlindSpot protocol). Since absolute number of clones equals number of positive clones (accuracy ≈ 100%) during one fragment cloning, data represent real increase in colonies in the plate. Number of internal 5′ (G)CTCTTC3′ sites are indicated. Ligation was performed using T4-Buffer and error bars are from three biological replicates, Bonferroni p < 0.05(*). Graph is representative of a duplicated experiment. Note that 3ins/4sites retrieved no positive clones under a regular TNT-reaction (data not shown) but only when the "BlindSpot" protocol, which represent a partial digestion, was applied (see Methods).
Scientific RepoRts | 6:19278 | DOI: 10.1038/srep19278 used) in 20 μ l reaction volume. The reactions were stopped and loaded in agarose gel. Bands were quantified by ImageJ software (area tool after plotting lanes) and organized using Excel. A non-methylated control was always included, and for M.SacI and M.TaqI (see below) sites non-subjected to methylation inside each tube, were also used to check full restriction enzyme activity. "Digestion inhibition" was a direct measurement of the digested bands divided by total band intensities (digested plus non-digested) and "Methylation efficiency" was calculated by 1 minus "Digestion inhibition". For M.SssI assays, a 1055-bp PCR product, using the pET28-M.SacI plasmid as template, was amplified (using the primers TaqI-Fw and TaqI-Rw), purified, quantified and incubated with methyltransferase as manufacturer instructions (NEB). In this case, there are 92 sites for M.SssI (5′ CG3′ ), which counts for ≈ 25 μ M of substrate in a 20 μ l reaction if 1 μ g of DNA is used. In this case, to achieve complete methylation, 1 μ l of enzyme (4 U) is recommended by the manufacturer to fully methylate 4 μ g of such template in 20 μ l reaction supplied with 640 μ M SAM for at least 2 h at 37 °C; our reactions ran for 4 h under these conditions. Methylated DNA was purified and 400 ng used for type IIS assays in duplicates and "Digestion inhibition" and "Methylation efficiency" were assessed as described above. Both sites shown in Fig. 3 are present simultaneously in the fragment and could be addressed in the same reaction by selecting the appropriate bands for quantification. For M.TaqI assays, two PCR products using the pET28-M.SacI plasmid as template were obtained (using the primers TaqI-Fw and TaqI-Rw1.1 and TaqI-Fw1.1 and TaqI-Rw), purified, quantified, diluted at least 1000 fold and mixed together in an equimolar ratio for a secondary PCR (30 cycles) using only TaqI-Fw and TaqI-Rw to generate the 1055 bp fragment with an internal M.TaqI site as shown on Fig. 3. The 1055-bp secondary product was then purified, quantified and incubated with methyltransferase as manufacturer instructions (NEB; except we increased incubation time to 4 h). Methylated DNA was purified and 400 ng used for type IIS assays in duplicates and "Digestion inhibition" and "Methylation efficiency" as described above. After the screening in duplicates, the M.TaqI results were confirmed by other 4 biological replicates for EarI only (Fig. 3). For in vivo assays, using M.Test plasmid transformed in T7X.MT, two separate colonies were tested on each condition shown on Supplementary Fig. 4 and the best condition (cultures grown on plates with 0.3 mM IPTG right after original transformation and 0.2 mM of IPTG on liquid media overnight grown at 37 °C) were reproduced for other 4 new colonies (Fig. 3). Experiment was later reproduced once, with 3 biological replicates and M.Test DNA was then kept at − 15 °C and re-accessed after 3 weeks (data not shown) and after 11 weeks, in which Eam1104I was also included.
TNT-family of vectors. All primers, GBlocks and gene cassettes were commercially synthesized and used in this study are listed on Supplementary Table 3. Nucleic acid manipulation followed the general guidelines described in 38 . DNA preparation was performed by either traditional phenol:chloroform extraction or DNA extraction kit (5PRIME #2300010). The pSTART is a pUC19-backbone vector, which carries the ampicillin/carbenicillin resistance gene, and was built domesticating EarI sites (5′ CTCTTC3′ ) by using Gibson  Table 1. Highlights between previous restriction enzyme-dependent methods and the TNT Cloning system. Limitations firstly surpassed by each method are bold and italic. a TNT-cloning loops up to three fragments at a time. b A sub-library is necessary for inverting fragments. c One-pot reactions are suggested but takes 1.5 h longer and are unsuited for sub-cloning efficient competent cells (10 7 cfu/μ g of puc19). assembly 12 to join the PCR products of primers 1) pUPD-FW1 and pUPD-RW1 (188 bp), 2) pUPD-FW2 and pUPD-RW2 (149 bp), 3) pUPD-FW3 and pUPD-RW3 (301 bp), 4) pUPD-FW4 and pUPD-RW4 (1838 bp) and 5) pUPD-FW5 and pUPD-RW5 (274 bp). The "Δ M15ω -peptide" was separately amplified from E. coli DH5α using the primers pUPD-RW3.1 and FW_adap and assembled into domesticated pSTART linearized by PCR using the primers pUPD-FW3.1 and pUPD-RW5. For the M.Test vector, used on M.TaqI assays in T7Express and T7X.MT, the pUPD-RW5-M_Test and pUPD_adap_met.test-FW were used instead of pUPD-RW5 and FW_adap, respectively (creating the M.TaqI site 5′ TCGA3′ ). The backbone of the binary vector pPZP200 39 (positions 1 to 6495 bp), plus a spectinomycin resistance cluster, were domesticated at different 5′ CTCTTC3′ sites using the primers α Ω vector-FW and EarI-RW1 (1132 bp), EarI-FW1 and EarI-RW2 (2699 bp), EarI-FW2 and EarI-RW3 (493 bp), EarI-FW3 and EarI-RW4 (2866 bp), EarI-FW4 and EarI-RW5 (234 bp) and EarI-FW5 and α Ω vector-RW (817 bp). PCR products were purified, mixed in equimolar ratio and re-amplified using the primers α Ω vector-nested-FW and α Ω vector-nested-RW (8080 bp band). The 8080-bp band was re-amplified with primers α Ω vector-FW and α Ω vector-RW to generate the α -backbone segment. The α version had the appropriate primer pairs α 1A-Fw and α 1A-Rw, α 2-Fw and α 2-Rw, α B-Fw and α B-Rw, α C-Fw and α C-Rw, α 1R-Fw and α 1R-Rw, α 2R-Fw and α 2R-Rw amplifying the reporter Δ M15ω from pSTART during a first PCR with each product followed by a secondary PCR with the primers PCR2_to_α Vector-Fw and PCR2_to_α Vector-Rw to create the 18-bp overlap needed for joining each segment by Gibson assembly 12  versions 5′ tt and 5′ gt at the 3361-3362 bp positions were also created and tested, data not shown). Importantly, this change was performed on all versions, however, only at those sites that bear two signatures side-by-side (Supplementary Fig. 1). Lastly, the versions α B-R, α C-R, Ω B-R and Ω C-R were implemented by digesting the α 1A and Ω 1A vectors at the PstI and PmeI sites and assembling the purified backbone to three GBlock fragments, having one in common (LacZw-central-gb) and the remaining specific for each vector created (alphaBR-gb left, alphaBR-gb right, alphaCR-gb left, alphaCR-gb right, omegaBR-gb left, omegaBR-gb right, omegaCR-gb left, omegaCR-gb right) by Gibson assembly. All vectors created without exceptions had the signatures confirmed by sequencing before undergoing tests. Primers pUPD-seqFW and pUPD-seqRW (for pSTART) or primers TNT-α Ω -seqFW and TNT-α Ω -seqFW (for any α and Ω members) were used to sequence inserts and diagnose constructs by colony PCR. Entry elements relevant to this work (Supplementary Table 4 Library construction (pSTART) and constructs diagnosis. Primers, to clone fragments by either restriction/digestion or Gibson assembly, were designed as 5′ ACATGCAGCTCTTCCACCN (20) 3′ , where N is the fragment of interest sequence forward (signature 1 is underlined), and as 5′ CGAGGAAGCTCTTCCATCN (20) 3′ for reverse strand (signature 2 is underlined), as long as TM of N (20) > 50 °C. Otherwise, number of base pairs was increased over 20 nt until at least 50 °C of TM was reached (https://www.idtdna.com/calc/analyzer) (see Supplementary Fig. 1). Multiple PCR products were purified and combined by Gibson assembly. All PCR reactions were performed using Phusion DNA polymerase (Thermo Scientific) according to suggested protocol (DMSO was added accordingly if amplicon was longer than 1.5 kb). Qiagen TAQ DNA polymerase diluted 10 fold was used for diagnosis through colony PCR and the remaining settings were according to suggested protocol. Briefly, colonies were picked from the agar plate and diluted in 10 μ l of water in 96 well plates and 1 μ l was used for PCR in 10 μ l final volume. TM used was always 56 °C for 20 sec and extension was always 72 °C for 1 min; always 40 cycles. Positive clones had the remaining 9 μ l (5 μ l if colony PCR was performed in parallel to culture growth) inoculated in appropriate media (LB + chemicals). Every insert in the library was sequenced. First levels of complex assemblies shown in Fig. 4 were fully sequenced. Clones also checked by restriction digestion are noted in the text.
pSTART entry clones. Supplementary Table 2 includes the entry vectors (pSTART) relevant to this work.
The insert sequences were grouped in the Supplementary

Detailed assembly steps.
Once an element is cloned in pSTART, which receives and releases the desired fragments with either enzyme, it is transferred and further combined in either alpha (α) or omega (Ω) members, which receive fragments upon cleavage with EarI/LguI and release fragments upon cleavage with LguI/EarI, respectively ( Supplementary Fig. 1). Upon digestion of each plasmid, a set of "signatures" that were specifically arranged to direct and orient the desired fragments are exposed. The signatures "1" and "2" are always flanking the inserts released from pSTART and are always used to join the final constructs into any α or Ω member. At the same time, the signatures "3" and "4" will be used by a specific member of each family (α and Ω) to join fragments between themselves, two fragments at once (binary assembly) using the members α1A and α2 (or Ω1A and Ω2) and three fragments at once (tertiary assembly) using the members α1A, αB and αC (or Ω1A, ΩB and ΩC). To change the fragment orientation (sense or anti-sense) simply switch the chosen α or Ω version for its respective "R" version during the cloning step. We first, at the α-level, had the GFP transferred from the library (pSTART) to αB and the NLS to α1A and αC. These clones were joined in a tertiary assembly in TNT-Buffer and the standard TNT-reaction. We tested several conditions for BspQI, EarI, LguI and SapI enzymes in order to tune our one-pot reaction conditions. We found the 10 mM DTT from T4 DNA ligase buffer to inhibit EarI activity and that excessive amounts of NaCl (> 50 mM) inhibited LguI. BSA in a reaction increased the number of false positives (data not shown). We found the best DNA concentration to be ≈ 75 ng (versus 100 ng and 125 ng; 75 ng each for multiple fragment assembling) insert plasmid(s) at the range of 0.25-2.5 kb and ≈ 50 ng (versus 15 ng, 25 ng and 50 ng) of TNT-members α, Ω or pSTART. Subsequently, we found PPG to increase the number of positive colonies and allowed us to reduce the incubation time for digestion/ligation while keeping higher efficiency than the T4 DNA ligase buffer. Ideal concentration of PPG is between 0.5% and 2%. Therefore, what we called the "TNT-Buffer", used in this work has the following formulation at 1×: 50 mM Tris-HCl (pH7.5), 2 mM DTT, 10 mM MgCl 2 , 1 mM ATP and 2% PPG (which was added right before reaction setup from a 20% stock in water). 5× buffer was stored at − 20 °C. The EarI/LguI/T4 ligase enzymes concentration are extremely important especially for accuracy (number of positive clones) and a standard TNT-reaction, set up on TNT-Buffer, includes 40 U of T4 DNA ligase and either 5 U of EarI or 0.5 U of LguI, followed by incubation at '34 °C for 45 sec and 16 °C for 4.5 min' for 50 cycles. If only one fragment is being cloned (or linearized destination vector is used for binary/tertiary assemblies) reaction can be performed at 34 °C for 1 h, albeit number of positive clones is reduced. All reactions were performed in 10 μ l final volume and diluted 1-10 fold or 1-50 fold when α or Ω members were used as destination vectors, respectively, before taking 1 μ l to transform electrocompetent cells (e.g. Fig. 4 where efficiency was ≈ 10 9 cfu/μ g of Puc19 plasmid). Final number of positive clones, shown in graphs, was then calculated as: total number of clones in a plate times dilution times accuracy. For general reference, GoldenBraid reactions used here followed 50 cycles of: 37 °C for 2 min and 16 °C for 5 min.
BlindSpot protocol for cloning non-domesticated fragments. For non-domesticated fragments, a regular TNT-reaction was used for single fragment cloning (we have not tested fragments that would leave signatures 1, 2, 3, 4, 1R or 2R upon cleavage of internal site). For binary and tertiary assemblies involving non-domesticated fragments, we developed the BlindSpot protocol -i.e., fragments (≈ 150 ng each rather than ≈ 75 ng each) were first incubated with 50 μ M oligo (design details below) for 1 h in each temperature 45 °C to 12 °C every 3 °C, usually overnight, in an alternate buffer (50 mM Tris-HCl pH 5.8, 75 mM NaCl, 10 mM MgCl 2 , 2 mM DTT) in 4 μ l final volume. Following, the addition of 6 μ l of a second buffer (50 mM Tris-HCl pH 6.3, 10 mM MgCl 2 , 2 mM DTT) containing either 5 U of EarI (for 5 min, ≈ 60-65% digestion progress) or 1.5 U of LguI (for 15 min, ≈ 55-65% digestion progress) completed the reaction volume to 10 µl, which was incubated at 25 °C before being directly heated at 80 °C for 20 min. After cool down, 2 μ l were used to set up a standard TNT-reaction using either T4 DNA ligase buffer (Fig. 5, NoDom) or TNT-Buffer. For the initial screening and digestion curve ( Fig. 5 and Supplementary Fig. 8), several incubation times for digestion were used and reactions were stopped with loading dye (NEB), loaded on agarose gel and analyzed similarly to what is described for the