PIGN Cure Collective

PIGN-CDG families are funding yeast-powered drug repurposing screens and a fibroblast biobank for hit validation. We just completed a pilot repurposing screen, and an expanded screen is in progress.

---

In collaboration with

---

---

All for one, and one for all. That solidarity animates the most productive ultra-rare disease communities and allows them to stomach the bench-to-bedside research process, which is unpredictably bumpy and disappointingly slow at times. The PIGN-CDG patient base is larger than most Congenital Disorders of Glycosylation/CDGs, and so the PIGN-CDG community especially benefits from pooling their resources and project coordination. However, not all PIGN-CDG patients are alike, mutationally speaking. The challenge of this project was to devise an approach inclusive of the largest number of families with the fewest samples, operationally speaking.

With thoughtfulness, diplomacy and hope, the PIGN Cure Collective was born, galvanized by Susie Lam and Hiran Patel on behalf of their daughter Alice; and the other kiddos living with PIGN-CDG whose families contributed funds and fibroblasts to advance a disease modeling and drug repurposing project that would otherwise not happen in academia or industry.

The PIGN Cure Collective is an embodiment of the patient-led ultra-rare disease research ethos we call 1-to-N medicine. This is the first chapter of the PIGN-CDG Cure Odyssey.

---

The PIGN gene encodes a protein (ethanolamine phosphate transferase; PIG-N) that is involved in glycosylphosphatidylinositol (GPI) anchor biosynthesis. PIG-N catalyzes the transfer of ethanolamine phosphate to the first mannose residue of the GPI anchor in the endoplasmic reticulum (ER). PIGN has been shown to also mediate protein quality control within the ER by preventing protein aggregation, but there’s still much we don’t understand about its function or functions.

To wit, the PIG-N protein is massive, weighing in at over 900 amino acids. The PIG-N protein is actually two domains linked together by a series of three alpha helices, according to this rendering by AlphaFold:

Almost 70 pathogenic variants in PIGN have been reported in the literature with the missense variant Leu311Trp (L311W) being the most commonly reported recurrent variant. All 28 patients in the PIGN Cure Collective patient list have compound heterozygous PIGN variants, resulting in 32 unique variants whose locations are mapped below:

The genome of baker’s yeast (Saccharomyces cerevisiae) contains one PIGN homolog called MCD4. Using a robust variant analysis strategy that we describe in detail below, 10 variants (highlighted in yellow) were selected from this list for yeast patient avatar generation:

Following the drug repurposing screens in yeast avatars, potential drug hits will be validated in patient fibroblasts. Primary fibroblast lines will be obtained from research labs or established from patient skin biopsies. That was the plan on paper, at least. Then like all best laid plans, it made contact with reality.

---

---

We set out to generate a panel of haploid yeast avatars of the 10 selected PIGN variants based on the protocols originally developed in Lao et al., 2019.  Diploid strains of several patient genotypes containing the recurrent missense variant L311W were to be generated. These include the L311W variant in combination with either a frameshift variant (Lys183Argfs*15; predicted to disrupt splicing which may result in the production of a non-functional or truncated protein), a missense variant (G264R) in the phosphodiesterase domain, or a missense variant (L830Q) in the PigN domain.

Rescue plasmids encoding yeast MCD4 or yeast codon optimized human PIGN with and without the selected mutations were to be ordered from a gene synthesis vendor. The growth rate of PIGN mutant yeast cells was to be assessed relative to positive and negative control strains as previously performed (Fujita et al., 2006).

The following eight variant selection criteria were developed to identify 10 suitable PIGN variants for yeast patient avatar generation:

1. Availability of or the potential to obtain a patient fibroblast line with the variant

2. Location for skin biopsy is in US or Europe

3. Variants that are recurrent in the patient list

4. Prioritize variants that may result in the production of some functional PIGN protein produced (missense); select only a couple of variants resulting in no protein or a truncated protein (nonsense, frameshift)

5. Amino acid position is conserved in yeast MCD4 protein and other model organisms (ConSurf analysis)

6. Amino acid is predicted to have functional or structural importance (motifs, Consurf analysis, model insights)

7. Variants span the entire PIGN protein; representation in both the phosphodiesterase domain and PigN domain

8. Variants associated with mild or severe clinical presentations (depending on the genotype)

Using those criteria, 71% of PIGN-CDG families that provided their variants are represented by the 10 selected variants. For the remaining 29% of PIGN-CDG families, 63% have a variant that is similar to the selected variants (either to a specific variant or the type of variant). For the remaining 37%, at least one variant will be present in the fibroblast lines collected.

Despite that impressive warmup, we struck out at the plate. We initially generated a panel of PIGN yeast avatars at the end of 2022, but after the fact uncovered a technical challenge with the expression plasmid that forced us to switch to another plasmid backbone. We reengineered the variants into the new plasmids and included several additional controls. Unfortunately, despite our best attempts in multiple rounds of troubleshooting and assay optimization, we were unable to measure consistent growth defects in tests of the 10 PIGN-CDG avatars.

Thankfully, in parallel to efforts by Cure Guides Rolando Perez and Mathura Thevandavakkam in our yeast lab, CDG Cure Guides Kamsi Nwangwu and Kristin Kantautas were conducting an international collection of patient fibroblasts from over a dozen PIGN-CDG families. Recall that our plan was to have a pairing between a yeast avatar and patient fibroblast carrying the same mutation.

The first learning was that the patient fibroblasts were the patient avatars all along. The second learning was that we could no longer generate one or a small group of yeast avatars at a time. It’s too unpredictable, it’s too slow, and current technology already allows for a “one-and-done” pooled approach to generate, characterize and screen all possible yeast avatars for any given gene: variant effect mapping, a mature technique that in its infancy was called deep mutational scanning. (We hope to boot up our first variant effect mapping project later this year — stay tuned!)

---

After months of work we knew that we couldn’t end up empty-handed. At the end of 2022, we put in motion a pivot away from patient avatars to a previously published temperature-sensitive MCD4 yeast mutant. A similar pivot worked for the PIGA yeast-powered drug repurposing project, so we had confidence that it would work for PIGN too. As step one, we confirmed the identity and presence of the temperature-sensitive (ts) mutation: G227E, which corresponds to position G217 in the human PIGN gene. G217 just so happens to be the amino acid next door to one of the 10 PIGN variants, H218Q.

Fortunately, the ts mutant behaved as expected at the Z’ optimization step as shown here:

Next, we screened the Pharmakon library at the SMDC at UCSF. Like every yeast-powered drug repurposing screen we’ve ever done, we found rescue hits! The magnitude of rescue is reminiscent of Perlara 1.0’s PMM2-CDG yeast screens where we also observed “modest” rescue: there’s still a delta between the strongest rescuer hit and the wild-type positive control.

Like all inborn errors of metabolism, PIGN-CDG must also involve one or more metabolic imbalances caused by the primary enzyme deficiency. In the simplest analogy, imagine a seesaw. When an enzyme is missing, one end of the seesaw shoots up, naturally causing the other side to plummet down. Now imagine a series of interlocking seesaws. If the first one goes out of whack, the imbalance cascades down the line. Based on our PMM2-CDG experience, there’s at least one lever to pull on to restore metabolic balance. Usually that means inhibiting an enzyme that is directly upstream or downstream of the deficient enzyme.

Here’s a plot of the Z score distribution for the complete Pharmakon PIGN dataset:

The ratio of rescuers to sensitizers is about 1:1. We’re not ready to disclose compounds IDs at this time but we can share that a cohesive mechanism of action unites the rescuers. Consistent with the restoration of a metabolic balance, the rescuers hits all inhibit a compensatory biosynthetic pathway that is rebalanced when GPI anchors lack the phosphoethanolamine linkage to the first mannose. We’re excited to share specifics once we’re more confident about the mechanism of action.

Next stop for PIGN-CDG is the TargetMol library that will be screened later this month at the HTSF at Cal, followed by hit validation in patient fibroblasts.

Onward for the PIGN Cure Collective!

---

---