Generating barcoded H3 DMS libraries means that we will be able to analyze epistatic interactions between higher-order mutants. However, we know that the majority of single mutants are deleterious for function. We therefore want to design a targeted pool of mutagenic primers that exclude these highly deleterious mutants, such that higher-order variants are more likely to be viable.
initial_prefs_analysis contains analysis of Perth09 DMS data to identify deleterious single mutants in H3.
targeted_library_mutations generates a set of targeted mutations that excludes deleterious mutants. This incorporates natural sequence analysis from John Huddleston, found in this repository.
targeted_library_primers preps additional input files for the create_primers.py script in TargetedTilingPrimers and holds the final csv of targeted primer sequences. Spread of primer Tm and length are also analyzed.
major_epitope_primers generates primers with generic NNK codons at 21 major epitope sites, as defined by mutations that have appeared in major clades over the past 6 years. This includes a primer with a single random codon for each of these sites, as well as primers with two random codons for every combination of nearby epitopes.
The main goal in this library design is to exclude deleterious single codons, such that higher-order variants are more likely to be viable.
Lee et al. (2019) conducted DMS of Perth/2009 and calculated amino acid preferences for each single mutant. In initial_prefs_analysis, we calculate the log2effect of each amino acid mutation. The effects of stop codon mutations are calculated separately, and used as a benchmark for defining 'deleterious' mutations. Plotted below are the log2effects of all amino acid mutations (left) and all stop codon mutations (right). Lines are drawn delimiting 75%, 90%, and 95% of the most deleterious stop codons. These quantiles are transposed onto all amino acid mutations. Mutants corresponding to the 75th quantile of stop codons were removed from the library, as this includes the major peak of deleterious mutants but leaves mutations with mild negative effects.
A .csv file listing all of the deleterious mutations (i.e., those with log2effects scores below the 75th quantile cutoff) is exported for further analysis in targeted_library_mutations. These deleterious mutants were referenced against a natural sequence alignment of H3 variants circulating since 1968 (generated by John Huddleston in this repository). Any mutation identified at a frequency of at least 0.5% among naturally circulating H3 sequences was removed from the deleterious mutants list; i.e., added back into the library.
csv files of targeted library mutations were generated from WT sequences for Kansas/14/2017 (GISAID accession #WSS2413637) and HongKong/45/2019 (GISAID accession #WSS2413591), removing all deleterious mutants remaining in the list. Final chimeras (plasmids 3022 and 3023) only include the ectodomain of these WT sequences, which corresponds to amino acid sites 17-520 (sequential numbering). Furthermore, the first 2 codons of the ectodomain are synonymously mutated in these chimeras to remove a poly-A run, and will be excluded from mutagenesis. The set of targeted mutations is therefore limited to amino acid sites 19-520 in sequential numbering for these WT sequences.
In summary:
-
7,597 amino acid mutations were identified as deleterious based on Perth09 DMS data.
-
Analysis of natural sequences identified 44 of these 'deleterious' mutants that were present in circulating H3 strains.
-
This left 7,553 amino acid mutations to exclude from these H3 DMS libraries.
-
H3 sequences were mutated from amino acid sites 19 to 520 in sequential numbering. If we limit the deleterious mutants table to sites 19-520, there are 7,033 mutations to exclude.
** (502 amino acid sites to mutate X 19 possible mutations) - 7,033 deleterious mutants = 2,505 library mutations
The analysis in h3_chimeric_library_design did indeed leave 2505 mutations in each library, distributed across 371 sites. Files listing all of these targeted mutations were then exported to targeted_library_primers.
The final set of mutations for each H3 ectodomain, spanning from residues 19 to 520 in sequential numbering, can be found here:
Based on these targeted mutations, primers were generated using the create_primers.py script in TargetedTilingPrimers. All necessary input files are prepped and collected in targeted_library_primers. This includes formatting an H3N2 codon frequency table for codon optimization. Also note that targeted mutations were identified from the WT sequences, but the primers need to bind to the chimeric sequences (plasmids 3022 and 3023), which include the first 19 amino acids from WSN-HA. primer_input_data_prep.ipynb therefore includes the necessary site numbering conversion, and ensures that only sequence that is consistent between the WT and chimeric sequences is mutated.
5010 primers were generated for both the Kansas/14/2017 and HongKong/45/2019 chimeric libraries, including fwd and rev pools. Target Tm was between 60C and 61C, and targeted length was between 25 and 51 bp. Actual distributions of length and Tm are visualized here:
The final set of mutagenic primers, designed to complement the chimeric WSNHAflank_H3-ectodomain constructs, can be found here. Note: due to numbering conventions in the primer design script, sites are numbered according to the first residue to be mutated. mut1N actually corresponds to residue 19 in the WT H3 sequences, and residue 22 in the chimeric H3 sequences. More details on this numbering conversion, plus script to ensure consistency between the WT and chimeric sequence numbering, can be found in primer_input_data_prep.
Finally, we also wanted to ensure a higher frequency of mutations at key epitope sites, and include double-mutant primers for nearby epitope sites. John Huddleston analyzed mutations that have been present in multiple major clades over the past 6 years, and identified 22 key epitope sites. major_epitope_primers contains the summary of this analysis. major_epitope_primers.ipynb generates primers with random NNK codons at each of these sites. It also generates a set of double-mutant primers with random NNK codons at every combination of epitopes within a primer length of one another.
Note: this script skips over the numbering conversions described above. Primers are labeled using epitope residue numbers, which reference wildtype H3 sequential numbering.
Single and double epitope primers for both H3 sequences are output in this directory. Given a minimum of 8 nt's between the beginning of the primer and the first site to mutate, there are 10 paired epitopes in HongKong/2019 and 9 paired epitopes in Kansas/2017.