Missense variant effect maps for the human HMBS enzyme via multiplexed yeast functional complementation assays

Defects in human hydroxymethylbilane synthase (HMBS), also known as porphobilinogen deaminase (PBGD) or uroporphyrinogen I synthase, may cause Acute Intermittent Porphyria (AIP), a potentially severe acute neurological disease. Here we describe missense variant effect maps for the erythroid-specific and ubiquitous HMBS isoforms. As no convincing difference between the maps was observed in any segment of HMBS, we aggregated the scores of the two maps, calculating a weighted average score for each variant.

We implemented a previously-validated humanized yeast model (Kachroo et al, Elife, 2017) and used it to confirm that expression of human HMBS from the hORFeome collection rescued temperature-sensitive (ts) mutations in the corresponding orthologous yeast essential gene HEM3. By coupling this efficient functional complementation assay with the TileSeq framework for multiplexed assays of variant effect (Weile et al, MSB, 2017), we sought to measure the functional consequences of all possible missense HMBS variants. First, we constructed libraries of HMBS variants (for both erythroid-specific and ubiquitous isoforms) using our previously described POPCode mutagenesis method (Weile et al, MSB, 2017). The variant libraries, initially generated as a pool of amplicons, were transferred en masse into the appropriate yeast expression vector via two steps of recombinational subcloning. The resulting libraries of variant expression clones were then transformed en masse into the appropriate ts yeast strain.

Next, pools of the yeast HMBS mutant strains were grown competitively on synthetic medium at non-permissive temperature. The allele frequencies of each HMBS variant within this laboratory strain population were then determined, before and after selection, using the TileSeq framework, in which sequencing is applied to a set of amplicon tiles that collectively span the complete coding region. These tiles are sufficiently short (~150bp) to enable sequencing of both strands (duplex sequencing), with variants called only where they are detected on both strands. Each nucleotide position was captured on both strands by ~2 million reads strands.

The sequence analysis code can be found at https://github.com/jweile/tileseqMave