Drug repurposing screen in haploid PMM2-CDG yeast models
During Stage 1 of our PerlQuest partnership with Maggie’s Cure, we generated haploid PMM2-CDG yeast models of three pathogenic variants: F119L, R141H, and V231M. We expressed these variants in the yeast ortholog of PMM2, SEC53, under four different promoters, and screened our drug re-purposing library to identify compounds that rescue the growth defects shown in Figure 1. We have completed screening of pACT1-F126L, pSEC53-V238M, and pSEC53-F126L, and data analysis is underway.
Figure 1. Growth of yeast PMM2 haploid models. pACT1-F126L, pSEC53-V238M, and pSEC53-F126L have intermediate growth defects that are conducive to screening. pREV1-V238M has a growth defect that may be too severe to overcome. 1X indicates the native SEC53 promoter. 2X indicates double the native promoter strength. 0.2X indicates 20% of the native promoter strength.
Testing a pharmacological chaperone for rescue of PMM2 yeast models
During our screening, Maggie’s Cure brought to our attention a hit compound that was identified in an in vitro screen for pharmacological chaperones of PMM2 (Yuste-Checa, et al., 2017; DOI: 10.1002/humu.23138). N’-(3-chlorophenyl)-N,N-di-2-pyridinylurea increases the thermal stability of purified wildtype and mutant PMM2 proteins. The 3-dimensional structure of a protein is fundamental to its function so a compound that stabilizes the protein may also improve its enzymatic activity. We ordered this compound from Chembridge (5914327) to test whether it can rescue our yeast PMM2 models.
Figure 2. Growth of yeast in 5914327 (N’-(3-chlorophenyl)-N,N-di-2-pyridinylurea). A. Chemical structure of 5914327. B. Normalized growth of yeast PMM2 models. Shown are averages of 16 wells per strain per condition grown in 384-well plates at 30oC.
Figure 2 shows that increasing the dose of 5914327 reduces growth, which indicates that it’s toxic in yeast. This is true in both haploid and diploid (see below) cells. We also tested it at lower doses (1 µM and 5 µM) and similarly, it reduces growth but to a lower extent. 5914327 may stabilizes PMM2 proteins, but in the cellular context, we do not know what else it is targeting. The toxicity of 5914327 may be due to off-target effects of this compound and highlights the importance of complementary approaches in drug discovery.
Generating another PMM2 allele: E139K
We were excited by the genotype-to-phenotype relationship of the SEC53 alleles we characterized previously and decided to add another allele: PMM2 E139K. Existing data on this variant is limited, but it is only two residues away from the enzymatically inactive R141H variant. The R141 residue is in the substrate binding domain of PMM2 (Andreotti et al., 2014 DOI: 10.1074/jbc.M114.586362), which suggests that E139K might disrupt substrate binding as well. We sought to better understand the E139K variant by modeling it in yeast. E139K is a result of a 415G>A mutation in the DNA sequence. Vuillaumier-Barrot et al. reported that this mutation interferes with RNA splicing that lead to skipping of exon 5 to form a partially deleted protein, but it can also generate full-length proteins (PMM2 E139K) that when expressed in bacteria show 25% residual activity (Vuillaumier-Barrot et al., 1999; DOI: 10.1002/(SICI)1098-1004(199912)14:6<543::AID-HUMU17>3.0.CO;2-S). Given the reduced activity, we expect it to grow similar to the F119L allele, which is also reported to retain about 25% activity. We placed it under the same four promoters (pREV1 < pSEC53 < pACT1 < pTEF1) that we previously used to characterize this allele (E146K in yeast SEC53).
Figure 3. Growth of SEC53-E146K variant expressed under different promoters as indicated.
Figure 3 shows that the E146K variant behaves like wildtype SEC53 and is defective only at very low expression under the REV1 promoter. In contrast to the enzymatic data reported, E146K does not lead to a growth defect despite the fact that a glutamic acid (E) is conserved at this position in all animals except zebrafish, where the residue is a chemically related glutamine (Q), and even though a lysine (K) is positively charged while a glutamic acid is negatively charged. In fact, a glutamic acid is 100% conserved at this position in a comparison of dozens of yeast strains, including natural isolates and multiple lab strains. This means that there a difference between the activity of human PMM2 E139K proteins expressed in bacteria versus yeast SEC53 E146K proteins expressed in yeast cells. Perhaps the effects of the E146K mutation are masked or blunted in a eukaryotic cellular context? Perhaps the 25% residual enzymatic activity of PMM2 E139K proteins as measured in bacteria is lower than in the human cellular context for a trivial reason? It’s also possible that the pSEC53-E146K mutant strain has synthetic growth defects in combination with specific genetic backgrounds or under specific environmental stresses.
If the 415G>A mutation more often cause a splicing defect, this would not be recapitulated in yeast because splice sites are not often conserved and splicing is less frequent in yeast. One way to interpret the contribution of the 415G>A mutation to disease manifestation is that most of the transcripts produce a protein with a partial deletion that is nonfunctional and some of the transcripts produce functional PMM2 E139K proteins, but not enough for normal cellular function. We can infer that there is a threshold for the amount of PMM2 protein required for glycosylation based on the fact that a single wildtype copy of PMM2 is asymptomatic, but we can cause a growth defect by reducing wildtype protein level to ~20% of normal with the REV1 promoter.
One of the many strengths of the yeast model system is that we can easily study it in either its haploid or diploid state. Haploid cells contain a single copy of each chromosomes so we can determine the full contribution of a single allele. Diploid cells contain two copies of each chromosomes so we can determine the effect of two alleles relative to one another. Most PMM2-CDG patients have compound heterozygous mutations where both copies of PMM2 are differently mutated. For example, the most common pair is F119L in combination with R141H. From Figure 1 and our previous post, we know how the individual mutations behave and we were curious whether we can recapitulate the growth pattern in compound heterozygous diploids. To take full advantage of the yeast system, we crossed haploid cells of different mating types and containing different variants to generate compound heterozygous containing F126L-R148H and V238M-R148H pairs.
Figure 4: Growth characteristic of diploid strains.
Figure 4 shows that pACT1-F126L/pACT1-R148H has a mild growth defect, and pSEC53-V238M/pSEC53-R148H and pSEC53-F126L/pSEC53-R148H cells have a more severe defect. Not surprisingly a single copy of wildtype SEC53 is sufficient for normal growth. This is similar to growth of these alleles in haploid cells. Similarly, we can use these defects to screen for compounds that rescue growth. We will soon start screening pACT1-F126L/pACT1-R148H, pSEC53-V238M/pSEC53-R148H, and pSEC53-F126L/pSEC53-R148H compound heterozygous diploids. We’ll compare and contrast these datasets to the haploid screens.
Stay tuned for more updates coming in the coming months!
Andreotti, G., Cabeza de Vaca, I., Poziello, A., Monti, M.C., Guallar, V., and Cubellis, M.V. (2014). Conformational response to ligand binding in phosphomannomutase2: insights into inborn glycosylation disorder. J. Biol. Chem. 289, 34900–34910.
Vuillaumier-Barrot, S., Barnier, A., Cuer, M., Durand, G., Grandchamp, B., and Seta, N. (1999). Characterization of the 415G>A (E139K) PMM2 mutation in carbohydrate-deficient glycoprotein syndrome type Ia disrupting a splicing enhancer resulting in exon 5 skipping. Hum. Mutat. 14, 543–544.
Yuste-Checa, P., Brasil, S., Gámez, A., Underhaug, J., Desviat, L.R., Ugarte, M., Pérez-Cerdá, C., Martinez, A., and Pérez, B. (2017). Pharmacological Chaperoning: A Potential Treatment for PMM2-CDG. Hum. Mutat. 38, 160–168.