At PLab we’re always on the lookout for new publications describing simple animal models of rare genetic diseases. The more non-obvious the model/disease fit appears at first blush, the more exciting are any results that confirm therapeutic relevance in humans — or more accurately in human cells. Last week, a UK-based cystic fibrosis (CF) patient and rare disease advocate named Oli Rayner emailed me the link to a new PLOS Biology paper — hereafter Veit et al — on a yeast model of CF that blew my mind. The paper is entitled “Ribosomal Stalk Protein Silencing Partially Corrects the ∆F508-CFTR Functional Expression Defect.”
I don’t want to bury the lede, so here it is: the loss of gene function caused by the most common CF mutation ∆F508 (a single deletion of a phenylalanine at position 508) can be suppressed by an evolutionarily conserved suppressor gene RPL12, which was first discovered in a yeast genetic modifier screen and then validated in patient-derived bronchial epithelial cells. When half of the recently FDA approved two-drug combination therapy lumacaftor was added to those human cells, half-maximal rescue was achieved. In theory, a novel combination therapy including lumacaftor and a second small-molecule drug (or comparable gene-knockdown technology) that targets RPL12 would turn CF patients effectively into disease-free heterozygous carriers.
Before I get to the paper proper and explain why that result is so compelling, here’s a primer on recreating Mendelian (single-gene) diseases like CF in model organisms like baker’s yeast, nematodes, fruit flies or zebrafish. The first step in determining whether or not a simple animal could be useful for modeling a disease is establishing the existence of ancient disease gene homologs. Homolog discovery can be accomplished freely online with a sequence-alignment tool on any of several genome databases (here, here or here).
The second step is to establish functional conservation of ancient disease gene homologs, as sequence conservation alone is not enough. To do so requires a real-world experiment, not just an in silico search. Ideally one would do a two-way or reciprocal gene swap experiment: knock out the disease gene homolog in the simple animal and then knock in the human disease gene, and vice versa. If the homologs are functionally interchangeable, we have a winner! More certainty can be achieved by knocking in the human disease gene harboring a pathogenic mutation and showing that this defective gene variant no longer reconstitutes function.
The first two steps are binary, with case-by-case wiggle room. Either there are ancestral forms of a human disease gene, or not. Either there is homolog interchangeability, or not. The third step is the most precarious. Does the disease state manifest itself the same way at the molecular, cellular and physiological levels in both humans and simple animals? In other words, how deep does homology go? Figuring that out requires additional experiments to characterize the phenotypes that result from knocking out the function of the disease gene homolog in the simple animal and observing the same phenotypes and biomarkers occur in people sick with the disease.
Clear? Clear. Okay, back to the paper at hand, though I still need to provide more backstory on why there is good disease/model fit between cystic fibrosis and Saccharomyces cerevisiae aka baker’s yeast.
What makes Veit et al so interesting and dare I say provocative is that on the surface, yeast cells are a poor model of CF, and not just for the trivial dismissals of simple animal models like yeast are single-celled animals and so therefore don’t have CF-affected organs like lungs or pancreases. There isn’t a single incontrovertible yeast homolog of CFTR, the CF disease gene. CFTR belongs to an ancient superfamily of proteins called ABC transporters, which include functionally diverse proteins that transport disparate substrates, from ions to xenobiotics to peptides.
That lack of all-or-none homology was noticed over two decades ago by researchers hoping to model CF. In fact what emerged was a picture of piecemeal homology, with gaps but a unifying thread. In 1993, Teem et al showed that STE6, a yeast ABC transporter gene that extrudes a peptide mating factor into the environment during yeast cell foreplay, could be reengineered as a functioning chimeric protein that includes the common nucleotide-binding domain grafted from CFTR. Critically, this domain contains the ∆F508 mutation. When a chimeric STE6/CFTR protein harboring the ∆F508 mutation was expressed in yeast, function was lost due to protein instability.
The yeast genome sequence was published in 1997, which allowed for an exhaustive accounting of all yeast homologs of CFTR. In 1999, Katzmann et al focused on one such yeast gene called YOR1. Below is Figure 3 from Katzmann et al. It reveals the strong sequence conservation of the CFTR domain containing ∆F508:
Pay attention to the red box. F508 of CFTR is equivalent to F670 of YOR1. In fact a phenylalanine is present at the analogous position in many members of the ABC transporter superfamily. Amino acid conservation can be a strong predictor of functional conservation but there’s only one way to find out for sure. So Katzmann et al created several YOR1 mutants, including ∆F670. In their words: “The ΔF670 allele of Yor1p produced a transporter protein that exhibited trafficking behavior striking in its similarity to that of the ΔF508 CFTR. These experiments suggest that ΔF670 Yor1p is trapped in the ER in a fashion analogous to that of ΔF508 CFTR.”
Promising, right? YOR1 and CFTR are both transporters that must traffic from the ER to the plasma membrane. YOR1 and CFTR don’t transport the same substrate when they arrive at the cell surface, but their intracellular journeys are the same. In both cases, the small amount of mutant protein that is able to escape the ER is non-functional. And the kicker is that loss of just a single ultra-conserved phenylalanine destabilizes both proteins in the same way. Followup work on the ∆F670-YOR1 yeast model of CF was sporadic and incremental till 2012, when a group led by John Hartman IV at the University of Alabama Birmingham made a significant advance.
Louie & Guo et al built on the foundation I summarized above. They asked the following question: which yeast genes when deleted in the F670-YOR1 strain result in double mutants that exhibit normal trafficking and restored transporter activity at the cell surface? A classic genetic suppressor screen, i.e., two wrongs make a right. They identified numerous suppressors, many of which are evolutionarily conserved. One of the most famous examples of suppressor gene discovery in humans is PCSK9, a protein involved in cholesterol homeostasis that is the target of next-generation, post-statin cardiovascular disease drugs.
Finally, I can return to Veit et al, which is the sequel to Louie & Guo et al.
Veit et al focused on the strongest suppressors that were identified in the original modifier screen in yeast. They also included a critical control: a YOR1 null mutant which doesn’t express any YOR1 protein whatsoever, and so cannot be rescued by a suppressor of the protein folding defect. The strongest suppressors fell into two categories, ribosomal or ribosome-associated genes and RNA degradation genes, shown here in Table 1 of the paper:
As is quite obvious, these genes are highly conserved in all animals. Next, they used the gene knockdown technology called siRNA to reduce expression of the human homologs of the above yeast suppressors in human bronchial epithelial cells expressing ∆F508 CFTR. Then they measured under physiological conditions not only how much mutant CFTR protein reached the cell surface but also the level of transporter activity. RPL12 was one of the top suppressors, and is the focus of the rest of the paper. What’s more, adding the drug VX-809 aka lumacaftor, which acts as a pharmacological chaperone, has an additive effect with RPL12 knockdown. These data are summarized in Figure 3 of the paper:
Veit el al go on to do a number of additional experiments to understand the mechanistic basis of suppression followed by RPL12 knockdown, as well as its specificity. After all, a common theme across many rare single-gene diseases is the occurrence of mutations like ∆F508 that destabilize proteins. Would knocking down RPL12 (or genes like it) rescue protein misfolding defects generally? So they examined several other transporters harboring destabilizing mutations and found that RPL12 knockdown only appreciably rescues ∆F508 CFTR. They also perturbed global protein translation using ribosomal inhibitors and knockdown of other components of the translational machinery. Although those interventions could contribute somewhat to rescue of ∆F508, knocking down RPL12 and other subunits of the so-called ribosomal stalk that play a role in elongation (versus initiation) produced the most convincing rescue.
So how is rescue achieved at the molecular level? Veit et al argue that by reducing (but not completely eliminating because that would in itself be toxic) protein translation at ribosomes, ∆F508 CFTR has more time to fold properly. As mRNA is translated at ribosomes, the nascent polypeptide chain is simultaneously folding up into a “native” 3D conformation. It’s known for example that ribosomes can pause during translation to allow for kinetically recalcitrant folding intermediates to resolve into a thermally stable structure.
To wrap up, besides going after RPL12 as a novel drug target for CF, I see two exciting potential paths forward:
First, the ∆F670 YOR1 should be used in a yeast-based HTS campaign to identify novel compounds that rescue protein misfolding, aberrant intracellular trafficking and transporter function at the plasma membrane. We already know to look for RPL12 and other ribosomal stalk proteins are potential targets of hits that come from such a HTS effort. Second, CF patients with ∆F508 mutations, in particular patients homozygous for ∆F508, should be stratified according to the severity and onset of disease. These subgroups — early vs late onset, for example — should have their genomes sequenced to identify suppressors that explain disease heterogeneity. I wouldn’t be surprised if there are CF patients with a more mild disease progression or attenuated disease severity out there in the vast world who have hypomorphic, that is reduced but not total loss-of-function, alleles of ribosomal stalk genes like RPL12.