Protein engineering

Creating Variations through Evolutionary Selection

Not all possible variations can be produced and tested in a sequence. Only changes that align with "evolutionary selection" are used, which involves selecting for the desired function or property.

The rules are to "get what is selected for" and "select what is already present". The protein must be different from the wild type (WT), but not too different. There needs to be a balance between conservation and diversification.

Library members should be similar enough in sequence to the starting protein to share a similar structure and function. However, they should also be different enough in sequence to exhibit slight differences in the functional property of interest.

In the diagram:

• The black star represents the original protein.
• In the first round of mutagenesis, a set of mutants (red/white stars) is obtained.
• Some mutants below the original protein have lost the original activity.
• Some mutants increase in activity.
• Others remain at the same level.
• The upper...

star is the best mutant (red), with the best activity. Sequence space: how much the protein differs from the original one. The most active mutant was selected from a collection of mutants.

After a first round of mutation, apply a second one and start from a new, higher, level of activity. Also do a third round and select the best mutant. Diversity of the protein is increased but not just by one round (one round is not enough usually). Both activity and sequence diversification are bigger.

Directed Evolution: is a "guided search" for the best mutant that allows to select sequences with mutations that are difficult to obtain in a simple approach of mutagenesis-selection. Indeed, each successive library is built around the most active mutant of the previous library. After three rounds of selection, the best mutant is more active and more distant in sequence with respect to the one isolated with the method a.

Methods to create libraries of mutants:

1. Random mutagenesis.

strains: bacteria that lack one or more mechanisms of DNA repair resulting in higher mutation rates and thus a greater rate of accumulation of mutations compared to normal strains. XL1-Red strain.

These two methods have the disadvantage of introducing mutations in an indiscriminate manner along the entire bacterial genome (survival, expression and translation) but also the vector carrying the transgene (expression and translation, replication). Furthermore, the second method (mutator strains) is slow. The level of mutagenesis is linked to the length of time the recombinant gene spends in the mutator strain.

Use an in vitro system to introduce mutations by mutagenic PCR, to speed up the process.

a) Mutagenic PCR: the most widely used method.

• Error-Prone PCR. The mutations are introduced under conditions that cause an increase in the speed by which the DNA polymerase generates errors. Mutagenic rate of ~1 nt/kb.
• DNTP analogs. Mutagenic dNTP analogues are incorporated into the

that can do preferential changes for both position and identity of the mutation

They can be overcome by combining libraries produced by using different ways to generate error (Mn , dNTP2+ concentrations, different type of Taq-pol.)

Codon bias: some aa change more easily than others because it’s sufficient a single nucleotide substitution to convert them in several other aa (for example V, by one single nt change, can be transformed in F, L, I, A, G, D). For other aa (C, S, P, H, R, N, T, M, E, Y) it is essential to change two nts and for others (Q, W, K) even three nts to obtain different aa.

In Nature this bias in the genetic code is optimized to ensure that amino acid substitutions are predisposed towards those that are less likely to cause loss of function.

By this codon bias, in a library, we can find 100 times more V to A then V to W,

Amplification bias: some mutations can be amplified by PCR more than others because introduced during the first cycles and therefore they may be

over-represented in the library (in principle if you have introduced a mutation during the first cycle, you should have 25% of clones with that mutation). Combination of multiple amplification reaction and reducing number of cycle are effective in reducing amplification bias.

08/03 Methods to create libraries of mutants: directed methods

This approach is focused on a specific region of a protein, selected because: it’s known the 3D-structure, it’s the putative site of interest (sequence comparison of primary sequence). There are conserved amino acids by sequence alignments. It’s possible to also select regions in a random way if there are no information.

In the simplest situation a single aa is changed with all possible variants (saturation mutagenesis). With rational design, the amino acid to change is selected and also the new residue identity. In this type of approach, however, the goal is to introduce all possible amino acids in that point and see what happens in the protein.

So it is still nonrational mutagenesis. In the common situation, a subset of amino acids in a sequence is changed in either all sites or part of the sites with any possible variant. Based on the incorporation of a sequence of synthetic DNA in the coding sequence. The region of interest can be extracted from the entire sequence and perform mutagenesis on that region (PCR or degenerated oligos) and then re-introduce the region in the whole sequence.

How is randomized the DNA of interest (how to produce the synthetic DNA)? A region of 10 aa (30 nt) is selected and amplified with specific primers. To have saturation mutagenesis of all 10 aa, in each single position there must be the 20 different aa and combine them for all 10 positions (10 codons). The variations are 20 and too many clones are needed to do so.

Normally the work is on a selected aa. Even focusing on just 4 aa, a lot of clones/mutants are needed, but is anyway a smaller number. Imaging each site as a codon for this calculation,

But knowing each codon is coded by 3nt. To produce degenerate primers we have to consider the number 3 for each codon. The final number of oligos in the final mixture increases.

Synthesis of oligonucleotides: The automated synthesis proceeds from 3’ to 5’. There are 4 bottles, each for a different nt (A, T, G, and C), and the machine pours one nt at a time in the column. For degenerated oligos, all possible nt in a particular position are needed, so there’s a 5th bottle with the mixture of all 4 nt at the same %. At the end, there’s the oligo in which the particular site can have any nt: this is a mixture of all possible combinations of the oligo. In this way, there will be both all the 20 amino acids in this position (codon = NNN) but also STOP codons (UUA, UGA, and UAG). (A)

To avoid this: (B) during the process, in the first position, just T or C can be used avoiding A or G. In this way, there are no STOP codons, but 5 amino acids are not present in this combination.

• (just 15).- (C) at first position add just G or T. Covering all 20 aa but also 1 STOP codon.
• (D) T, C or G in first position → 20 aa and 1 STOP codon.
• (E) is to have all 20 aa avoiding STOP codons, introducing directly the one codon for the aa.
• The 4 nucleotides are used at each position.
• Stop codons can be avoided by using at the third codon position only T and C, but limiting the range of aa.
• The use in the final position of the codon T, C and G, or
• Use NNG / T or NNG / C limit stop codons.
• The codon can be synthesized by the direct addition of a mixture of 20 trinucleotide phosphoramidites (i.e. codons already synthesized) in one step. Possible when a limited number of aa if required by the experiment.
• A and D are the most easily affordable.
• 16· Bias problems
• The method for the synthesis must be accurate to avoid misproduction of oligos.
• Synthesis of degenerate oligos is well established. However, any synthetic process where a number of reagents are used as mixtures,

is susceptible to bias arising from greater incorporation of one reagent than another. Where synthesis is carefully optimized and highly pure reagents are utilized, this bias is small.

Also, to consider degeneration of the genetic code: some aa will be more represented than others (intrinsic bias due to the degeneration of the genetic code). Several aa will be more represented then other and stop codon can be present, could be difficult to insert codons for a subset of aa if this is desirable. This is a problem in very large libraries, or when statistical data are needed.

If using the method just 1 STOP codon and 20 amino acids are introduced, the library of mutants can be homogenized a little bit. It's important to consider it in the statistical analysis of data.

It's possible to not use all amino acids: create libraries with a low number of aa and no stop codons but in this case the aa present all possible physical and chemical characteristic. Use for this the IUPAC nucleotide

ambiguity codes: encode a minimal se