We have shown here that complex constructs containing many transcription units (here 11, consisting of 44 individual basic modules) can be assembled by a series of three one-pot Golden Gate cloning reactions. The construction principle exemplified in this work can theoretically be repeated indefinitely to add more transcription units, until the constructs become simply too large to be transformed or propagated in standard hosts such as E. coli. As outlined in , it is necessary to create a destination vector at each level 2 cloning step for further rounds of cloning (level 2i-2 to level 2i-3, etc…). This is done by the alternating use of end-linkers providing different type IIS restriction sites (for example Esp3I or BsaI) and allowing convenient color selection from blue to red and vice versa (). The expansion, for example, of the largest construct made in this study (cL2-13, level 2-2, 33 kb) would require its reconstruction, but with an end-linker that adds two Esp3I restriction sites to the construct (end-linker pELR-4, ). One or more genes could then be added to this level 2i-2 destination vector using an Esp3I/BpiI Golden Gate cloning reaction ().
Beside the construction of large and complex constructs encoding entire pathways, the high cloning efficiency also allows the creation of construct libraries. Instead of using one specific module for each component of a transcription unit, a module library can be used instead. In case a library of promoters is used, constructs obtained would contain a coding sequence under control of different promoters. Since nearly all constructs are correct, the library can be screened directly for optimal expression level for this particular gene, or be used for the next level of cloning in which several genes or again gene libraries are assembled. This application is of particular interest for the optimization of biochemical pathways for metabolic engineering where several genes not only have to be co-expressed, but also, their expression ratios have to be balanced to obtain optimal yield of the desired product.
The advantages of using standardized modules do not lie exclusively in the ability to easily create complex constructs. Already, the simple definition of a general cloning standard will result in tremendous synergistic effects, since the validated modules or module libraries created by different scientific groups can be reused from the whole scientific community. An impressive example is the widely used standard proposed by the BioBrick foundation 
. Here, researchers from all over the world have already contributed thousands of compatible modules to a freely available module collection. In contrast to MoClo, Biobrick modules are flanked by standard type II restriction enzymes, and assembly of two BioBricks via restriction and ligation results in an idempotent new Biobrick module. However, the two modules are separated by a scar sequence, and the process is unsuitable for the assembly of multiple fragments in one step.
The principle that a huge community contributes to a standardized system requires however that the standard shows some flexibility. Although the MoClo system described here is based on five basic modules, it is very versatile since each of these modules can be subdivided in smaller modules that would still be compatible with the existing ones. For example, a terminator can be split in two modules consisting of 3′ untranslated region and actual terminator sequences by definition of a new fusion site separating both modules. The transcription unit would be assembled with the two new modules replacing the original terminator. In case of more sophisticated cloning applications, like the shuffling of an ORF consisting of several protein modules, it may be favorable to define an entire new level. These level -1 modules have to follow the same principles as all other modules: a set of compatible overhangs, where the first and the last are compatible to the next level, a specific color selection and a specific antibiotic selection marker have to be defined.
The data presented here show that all elements required for the design of a completely automated cloning system are now in place. Operations that are required for cloning using the MoClo system consist of preparation of plasmid DNA, liquid handling and incubation to perform restriction-ligation, plating of transformation on plates, picking of colonies, and digestion and analysis of plasmid DNA. The last step can even be replaced by DNA sequencing of a single colony, because the system is so efficient. A further advantage in terms of automation is that no sophisticated construction strategies are needed since the design is automatically defined by the number and the order of modules that a user wants to assemble. The cloning strategy can be easily and unambiguously determined by a simple computer program, which could also be directly linked to the automation robots that would make the construct.