Team:Open Science Global/Wetware/A-Clash-of-Overhangs
A Clash of Overhangs
Having now designed parts to be compatible and interoperate with the BaClo extension of the MoClo assembly standard, as well as parts to be compatible and interoperate with the ProClo extension of the MoClo assembly standard, we wondered whether the BaClo and ProClo overhang sets were completely compatible. To our surprise and dismay, it would seem they are not: a Golden Gate reaction with all 19 of the combined BaClo and ProClo overhangs is predicted by NEB’s Ligase Fidelity Viewer to have a fidelity of less than 80%.
Inspecting the ligase fidelity matrix revealed that a small number of overhangs, including CCAT, GTCA and TTCG from the ProClo assembly standard, are responsible for most of the predicted incorrect assembly. And indeed, removing those three overhangs from the set boosts predicted assembly back up to 98%.
This suggests that if the RBS+Loc, Tag1, Tag2 and Tag4 part types from the PET collection were re-synthesized with three different overhangs to replace CCAT, GTCA, and TTCG, the FreeGenes ProClo and BaClo assembly standards could be unified into a single high-fidelity expanded MoClo standard capable of assembling vectors in E. coli, transferring those assembled vectors, shuttling to and propagating in a variety of other cell types, constructing sophisticated multi-TU genetic devices, and gaining control of protein localization, quantification and purification through composite peptide tags.
Dropout Cassette: Betting on Kinetics
In order to sidestep the overhang clash between the BaClo and ProClo assembly standard, as well as to simplify the assembly reactions required, we designed a dropout cassette to replace the transcription unit components in our BaClo E. coli to B. subtilis shuttle vectors. At the heart of the dropout cassette is the ccdB toxin gene from pOpen_v3’s dropout, constitutively expressed and with a transcriptional terminator added on. Only E. coli cells with the ccdA antitoxin or a specific mutation in DNA gyrase survive ccdB expression, which means transforming a Golden Gate assembly in which the backbone vector had a ccdB dropout into most competent E. coli strains will result in very low, if any, false positive colonies from the undigested backbone vector.
We’re planning to assemble this dropout cassette with other parts from our refactored BaClo modification of pHT43, which will require a BsaI digestion and assembly. One way to approach this would be to make a different type IIS restriction enzyme, like BsmBI or AarI, cut out the dropout cassette. But then, that second restriction enzyme would be required every time you attempted to assemble a new transcription unit into the backbone vector you had built. Our PI Scott Pownall, who has a lot of experience with Golden Gate assembly, recommended a different design strategy: just use BsaI as the cut sites for the dropout cassette, and drop the temperature to 16 ºC for 10 minutes at the end of the assembly. At this temperature, ligation is favored over restriction digestion; so if the assemblies are cooled and transformed, we expect to get colonies with our cassette inserted, even though it still has cut sites for the restriction enzyme in the reaction. Apparently, this is how a lot of dropout cassettes are assembled into vectors. Building vectors with dropout cassettes using our and the OYC’s BaiClo parts, and then assembling RBS/secretion tags, CDSs and multifunctional protein tags into those vectors with our and the PET’s parts, is how we plan to proceed once our homology arm (and hopefully, origin of replication) parts are synthesized.
Interlude on Type Switching
Hat tip to Keoni Gandall, who had the idea for BtgZI-BsaI cut sites.
As mentioned before, we added a ribozyme insulator to the lactose/IPTG-inducible promoter part, on the hypothesis that for bioengineering purposes, cleaving off the 5’ UTR and making gene expression from that promoter less dependent on CDS mRNA structure would be beneficial. Also, all the RBSs in the RBS/SecTag pairs in our library plasmids were designed in RBScalc with the assumption that the 5’UTR upstream of the RBS is insulator RiboJ54. That being said, we also wanted the option not to change the sequence 3’ to the lac operator too much, in case it has cryptic functional importance in pHT43’s inducible promoter. To balance these two needs, we used Type Switching.
Rapid progress in comprehensive profiling of Golden Gate assembly overhangs has enabled the discovery of 3- and 4-base overhang sets that enable modular one-pot genetic assemblies of remarkable complexity. On the one hand, this progress should usher in a new era of much more sophisticated and powerful genetic assembly standards, which incorporate part types far beyond the simple 4-component transcription unit paradigm. Custom plasmid backbones, ribozyme insulators, recombination sites, multifunctional composite peptide tags: assembly standards are now possible that could easily incorporate modular libraries of these tools into highly functional genetic devices in a single standardized reaction. On the other hand, complexity for complexity’s sake is not desirable; if all one needs is to assemble a simple transcription unit into a backbone vector, creating a bunch of extra ‘slots’ in the assembly standard that must either be filled or covered with ‘spanner’ parts adds unneeded hassle, cost, and chance of failure. One potential way to address this tension between the possibilities of complexity and the benefits of simplicity is Type Switching, the design of genetic parts that can change their ‘type,’ in other words their overhangs in a Golden Gate Assembly, just by adding a second Type IIS restriction enzyme to the mix.
The simplest way this could work, that leaves the least scar sequence at the part interfaces, uses BtgZI. BtgZI is a (now off-patent) type IIS restriction enzyme. Most type IIS restriction enzymes have gaps of 1-2 bp between the end of their binding site and the start of their cut site. for instance, here’s BsaI:
5' GGTCTC N| 3' 5' |NNNN N GAGACC 3'
3' CCAGAG N NNNN| 5' 3' |N CTCTGG 5'
BtgzI, by contrast, has a 10 bp gap between where it binds and where it cuts:
5' GCGATG NNNNNNNNNN| 3' 5' |NNNN NNNNNNNNNN CATCGC 3'
3' CGCTAC NNNNNNNNNN NNNN| 5' 3' |NNNNNNNNNN GTAGCG 5'
An entire BsaI binding site, gap, and 4 bp cut site can fit in the gap between BtgzI’s binding and cut sites:
5' GCGATGGTCTCN|NNNN| 3' 5' |NNNN|NNNN NGAGACCATCGC 3'
3' CGCTACCAGAGN NNNN|NNNN| 5' 3' |NNNN|NCTCTGGTAGCG 5'
What this means is that a DNA part flanked by BtgZI-BsaI cut sites can have completely different 4bp overhangs in a Golden Gate assembly, depending on whether or not BtgZI is added to the reaction.
This same property can also be achieved in a slightly different way with BtgzI-AarI cut sites, since AarI has a 4 bp gap between its binding and cut site:
5' CACCTGC NNNN| 3' 5' |NNNN NNNN GCAGGTG 3'
3' GTGGACG NNNN NNNN| 5' 3' |NNNN CGTCCAC 5'
5' GCGATG NNNCACCTGC|NNNN 3' 5' NNNN|NNNN GCAGGTGNNN CATCGC 3'
3' CGCTAC NNNGTGGACG NNNN|NNNN| 5' 3' |CGTCCACNNN GTAGCG 5'
Particularly for non-protein-coding parts with relatively neutral/non-functional sequence at their interface, Type Switching could potentially be a bridge between complexity and simplicity: cut with BsaI, get overhangs for a simplified assembly reaction with fewer, more commonly used part types; cut with BtgZI, and get overhangs for a more powerful assembly standard with more complex part types.
--Interlude over--
Type Switching allowed us to insert the RiboJ54 insulator just downstream of the lac operator (which assembles normally with the RBS during BsaI Golden Gate), while preserving the option to swap in a different insulator part (via BtgzI addition to the Golden Gate assembly) , or to eliminate the insulator part altogether (via adding AarI to the Golden Gate assembly). Additionally, a BtgZI-BsaI digestion can create a new overhang on the 5’ end of the promoter as well, potentially opening up space on the plasmid for more genetic part types, including recombination sites. From one part, three behaviors are possible, just depending on which restriction enzymes are in the reaction.
-->BbsI>GGAG--BtgzI>BsaI>GGAGATTC--<Prom_LacI/LacO<--TACTAACA<AarI<BtgzI<--RiboJ54--BsaI<--CGCT<BbsI--
Adding an OriT and Conjugation Helper Strain to Refactored pHT43
Another modification we made to the original pHT43 in our refactored vector backbone collection was to add an OriT for conjugative transfer from E. coli to B. subtilis. We did this at the recommendation of both Dan Ziegler and Keoni Gandall. B. subtilis’ natural competence mechanism requires exogenous double stranded DNA to be cleaved into ~10 kb chunks, digested into single stranded DNA, and then imported. This works well for genomic integration, because the imported ssDNA is so recombinogenic; but for circular replicating vectors, this process makes re-assembly of the vector in the cell difficult, unless it is pre-concatemerized by replication in a RecA+ E. coli cell. Most common cloning E. coli strains are not RecA+. Moreover, Ziegler informed us that B. subtilis’s electrocompetence rate is abysmal. Taking all these advice together, we concluded the simplest way to shuttle a theta-replicating vector like our refactored pHT43 would be to include an origin of transfer that can be mobilized by a helper E. coli strain harboring a conjugative plasmid. This method is more general than propagating plasmids in RecA+ cells (indeed, the OYC also has an E. coli to S. cerevisiae oriT), and if we can obtain and share a conjugative helper strain under an Open or no MTA, Our most promising lead is E. coli strain C600/RK2 from the Coli Genetic Stock Center (CGSC), which contains the helper-capable conjugative vector RK2. We have reached out to purchase this strain and are waiting to verify that, like BGSC, it is distributed MTA-free.
B. subtilis Strains: Once Bought, Free to Share
Two of the most valuable pieces of knowledge Dan Zeigler shared with our team were (1) that the Bacillus Genetic Stock Center distributes its strains without an MTA, meaning that once we buy the strains we are free to modify and redistribute them without restriction; and (2) that he had engineered several strains of B. subtilis 168 with all seven of the main extracellular proteases knocked out. These knockout strains are KO7, KO7S (which also has sporulation knocked out) and KO7A (which has an eighth, minor protease knocked out). When we engineer one or more of these strains to make and secrete useful proteins, we will be able to share those cells with people who want to use them.