A major adverse effect of statins is myopathy (known since 1980).  Now genetics have shed light on this association. Two independent teams have recently discovered that recessive mutations in HMGCR cause a severe form of muscular disease in humans.

A 🧵 for genetics lovers.

HMGCR codes for HMG CoA reductase which converts HMG-CoA to mevalonate, a rate limiting step in the cholesterol synthesis. Statins act by inhibiting HMG-CoA reductase.
https://en.wikipedia.org/wiki/HMG-CoA_reductase

Statins were discovered in 1970s by a Japanese scientist, Akiro Endo. The statins are, according to WSJ, "the first in a class of medicines that today brings $25 billion a year to pharmaceutical companies."
https://en.wikipedia.org/wiki/Akira_Endo_(biochemist)

And yet, Endo derived no financial benefit from his discovery.  Michael S. Brown and Joseph Goldstein, who got Nobel prize for cholesterol related work, famously said "The millions of people whose lives will be extended through statin therapy owe it all to Akira Endo"

I strongly recommend reading this fascinating WSJ article on Endo's story: "How One Scientist Intrigued by Molds Found First Statin".  You'll be amused by the way the article ends.
https://www.wsj.com/articles/SB113677121574341250

Okay, coming back to our main topic, it turns out that the very first human, an 18 yr old female, who received statin developed severe myopathy to the point she was unable to walk. The drug was discontinued, and the woman recovered fully.

Then Endo and his colleague (Akira Yamamoto) continued testing statins (in secret!) in lower doses in other patients and achieved a 27% drop in blood cholesterol. And they even published those results (in 1980).
https://pubmed.ncbi.nlm.nih.gov/7362699/

Then as the statins were used widely across the world, it has been well recognized that myopathy is a common adverse effect, seen in ~30% of the users. But the mechanism underlying remained elusive.

More than four decades later, now two independent group of researchers    have encountered patients with a rare muscular disease that turned out to be caused by biallelic mutations in HMGCR.


The first group of scientists, from Israel, published in the PNAS their work on the genetic investigation of a consanguineous Bedouin kindred where multiple individuals suffered from a severe limb girdle muscle disease.
https://www.pnas.org/doi/10.1073/pnas.2217831120

The disease manifested in the 4th decade of life starting as a mild to moderate pain and fatigue in proximal and axial muscles, then progressing to severe weakness, often involving respiratory muscles, requiring wheel chair and assisted ventilation.

Exome sequencing pinpointed the genetic cause: a homozygous missense mutation within a highly conserved region of HMGCR gene that severely reduced the affinity of the enzyme HMG CoA reductase to its substrate HMG CoA.

Note that these biallelic mutations only reduce the function of the enzyme. Complete deficiency of this enzyme is incompatible with life (homozygous KO is embryonically lethal in mice).

Likewise, a second group of scientists, from the USA, published in the AJHG a few days ago their work on the genetic investigation of nine individuals from unrelated families who all suffered from a limb-girdle like muscular dystrophy.
https://www.cell.com/ajhg/pdfExtended/S0002-9297(23)00130-1

The patients had a very similar features as those from the Israeli study: gradually worsening proximal and axial muscle weakness leading to loss of ambulation and respiratory insufficiency, except that here disease onset was earlier ranging from childhood to adolescence.

Through exome sequencing, the authors identified compound heterozygous missense variants in all the patients except one who, the authors speculate, might be a homozygous for a missense variant due to uniparental disomy.

All of the identified missense variants were located within conserved residues clustering within protein domains critical for enzyme activity. In vitro assays confirmed defective enzyme activity at varying levels of severity.

But what is amazing about these discoveries is that scientists have not only established the precise cause of statin associated myopathy but might have also found the treatment for it.

Going back to the Israeli study, the authors didn't stop at just diagnosing the genetic cause but went one step further and ran n=1 clinical trial.

If the HMG CoA reductase deficiency is hurting muscles, then can externally supplementing mevalonic acid (which the enzyme is supposed to help make) will reverse the symptoms?

The Israeli scientists received approval for the trial from the Israeli ministry of health under the compassionate treatment procedure and treated one of their patients with weekly oral dose of mevalonalactone.

The doctors noticed improvement in muscle function as early as day 21. While previously the patient was unable to do any of these tasks, now she can raise herself sideways from sitting position, fully abduct her arm, extend her knees and

feed her grand child, stand with assistance and breathe without ventilation. A supplementary video from the paper recorded 4mon after treatment showing that the pt was able to abduct her arm fully which wasn't able to before the treatment.

The authors also modeled statin associated myopathy in mice and demonstrate that mevalonolactone supplementation reverses statin induced myopathy.

The authors are currently extending the treatment to other patients and envision that mevalonolactone might become a standard treatment for statin-induced myopathy in the future. If this happens, all those who will benefit will owe it to these rare disease patients.

You can read the unrolled version of this thread here: https://typefully.com/doctorveera/AOYaLaN

Discovery of the first human disease caused by HMGCR mutations

Great paper from the Epi25 collaborative on the rare variant associations of different types of epilepsy based on WES analysis of ~54k individuals. This paper is packed with insightful findings. I'll highlight a few 🧵
https://www.medrxiv.org/content/10.1101/2023.02.22.23286310v1


The authors studied three types of eplipesy.

One rare, severe form: developmental and epileptic encephalopathies (DEE N=~2k), which is often comorbid with developmental delay.

Two common, milder forms: genetic generalized epilepsy (GGE, N=5.5k) characterized generalized seizures and non-acquired focal epilepsy (NAFE, N=9.2k) characterized by focal seizures.

1. Common vs rare variant architecture

We should think about genetic architecture of a disease in terms of both common and rare variants. So, you can gain more insights if you read the GWAS paper alongside that looked at the common variant associations.
https://www.medrxiv.org/content/10.1101/2022.06.08.22276120v1

A general rule is that rare and severe disease get more contributions from rare and large effect variants whereas common and milder form get more contributions from common and small effect variants, which you can appreciate from the results.

GGE (n=5.5k) and NAFE (9.2k), despite having larger sample sizes than DEE (2k), had lower power to discover rare variant associations. Analysis of burden of protein truncating variants (PTVs) yielded 5 sig genes for DEE, 1 for NAFE and none for GGE.

On the other hand, GGE has stronger contributions from common variants with SNP heritability estimate nearing ~40%. This Manhattan plot displaying 25 hits is based on a GWAS of GGE in 7.4k cases (only slightly larger N than ExWAS) and ~50k controls.

The NAFE subtype is a special and interesting case. Despite being common and so having a large N (16.3k cases) they yielded no GWAS hits. Reasons suspected: phenotype heterogeneity, somatic mutations etc.

Keeping NAFE aside, there is a clear contrast in the common vs rare variant contributions between DEE and GGE.

2. Gaining power by gene sets

As I've pointed many times before, we can increased statistical power by looking at gene sets than indiv genes.
https://twitter.com/doctorveera/status/1513349069351006209?s=20

After studying for individual gene level associations, the authors looked at associations at gene set levels, particularly gene families and those encode protein complexes.

One of the strongest associations was seen between NAFE subtype and genes encoding GATOR complex proteins.

If you go back to the individual gene associations of NAFE, you'll notice that, in addition to the top hit (DEPDC5), there are two other subthreshold hits that are components of GATOR complex: NPRL3 and NPRL2.

This finding reminds us a lesson we've learned in the past: one way to identify novel genes is to look at the sub-threshold genes that are members of the pathways implicated by significant genes.
https://twitter.com/doctorveera/status/1570300260081336323?s=20

3. Converging evidence across multiple variant types

Strong associations with PTVs than missense variants suggest that the diseaase mechanism is due to haploinsufficiency rather than gain of function (which seem to be the case for GATOR complex genes)

One way to confirm that haploinsufficiency leads to disease is to look at deletion CNVs encompassing the gene of interest. The authors found many of the genes implicated by PTVs also show association with deletion CNVs.

Among which, NRPL3 showed impressive convergence between PTVs and CNVs. Combining both the evidence together rendered NRPL3 exome-wide significant.

This is an orthogonal way of replicating genetic findings. In addition to replicating the association with same genetic variant in independent samples, we shud also search for genetic associations with independent variants in the same sample (i.e. identify allelic series).

4. Phenotypic heterogeneity

This is an extremely important concept when it comes to developmental genes. Phenotypic heterogeneity arises when different mutations in a gene lead to different phenotypes.


This phenotypic heterogeneity can be milder with different mutations causing same disease with varied clinical presentations. Here is a great example.
https://twitter.com/doctorveera/status/1516638245471211522?s=20

Or can be severe, with different mutations causing strikingly different phenotypes. e.g. there are genes where missense variants cause NDD and PTVs cause schizophrenia.
https://twitter.com/doctorveera/status/1513349259403268096?s=20

This is important to know as sometimes depletion of a class of variant in the disease group can be misinterpreted as absence of pathogenicity or even worse:protective associations, where in truth those variants are hiding among a diff disease group.

In the current study the authors see strong enrichment for ultrarare missense variants but not PTVs in GABA receptor complex genes, which is also reported in earlier studies.
https://pubmed.ncbi.nlm.nih.gov/31327507/

The lack of PTV enrichment in these genes simply could be due to ascertainment bias where the cases are enriched for early onset severe cases while PTVs are hiding in milder cases and will come out only with large N
https://twitter.com/doctorveera/status/1516638278291648513?s=20

The authors also show a beautiful example of a gene, KDM6B, with strong phenotypic heterogeneity where  PTVs cause NDD but missense variants cause epilepsy.

Epilepsy is one of the most common neurological conditions and our knowledge of their genetic etiologies has been so far predominantly based on studies of isolated cases/families with highly penetrant mutations.

It's fantastic that now we are able to study the genetics of epilepsy at scale in an unbiased hypothesis free manner. Great accomplishment by the Epi25 collaborative team of researchers. 👏 👏

Rare variant architecture of epilepsy

While I appreciate the sentiment, I am not a fan of such blind rules. Let me start with a 🧵 of some major human genetic discoveries (many translated to therapeutics) made with very few samples.
https://twitter.com/ewanbirney/status/1628340273594961920?s=20


1. Association of complement factor H with AMD was discovered with mere 96 cases and 50 controls. Today many drug pipelines are in development based on this discovery.
https://www.science.org/doi/10.1126/science.1109557

2. Association of BCL11A locus with fetal hemoglobin fraction in blood was discovered based on 179 individuals. To date this is the most successful GWAS discovery to therapeutics translation.
https://www.nature.com/articles/ng2108

https://twitter.com/doctorveera/status/1335323219151163393?s=20

3. APOL1 association with kidney disease in African Americans was discovered by studying 205 cases and 180 controls. Today there are many drug programs exist for APOL1 kidney disease in African Americans.
https://www.science.org/doi/10.1126/science.1193032
https://twitter.com/doctorveera/status/1530624682847657985?s=20

4. CCR5 loss of function variant association with HIV was discovered by screening ~600 HIV exposed sero-negative individuals.
https://twitter.com/doctorveera/status/1409189802168123392?s=20

While I support discouraging underpowered genetic association studies, putting a hard threshold for minimum sample size for genetic association studies isn't fair.

Genetic discoveries in non-European ancestries are picking up only now and we have barely touched the low hanging fruits of non-European studies (from developing and underdeveloped countries), many of which are likely to show up even in small sample sizes.

Also, still there are many rare (ish) or even common diseases that we haven't GWASed yet. So, it's possible there are low hanging fruits for such unexplored diseases/phenotypes that will show up even in small sample size.
https://twitter.com/doctorveera/status/1531333656882475011?s=20

Apart from sample sizes, there are other important factors that strongly influence statistical power: effect size, proximity of the phenotype to DNA, phenotype measurement error.

When the effect sizes are extremely large, genetic discoveries can be made in small sample sizes (which were the cases in examples highlighted above).

Molecular traits (e.g. proteins, metabolites) often lie proximal to DNA and have small measurement errors and so will yield meaningful results even in very small sample size.

Some examples.
1. GWAS of brain cell composition in the brain. A neurodegeneration locus shows up beautifully in a GWAS of just 403 individuals.
https://twitter.com/doctorveera/status/1352680473198268419?s=20

2. GWAS of neural progenitor cell proliferation. Just N=80, GWAS reveals a strong locus.
https://twitter.com/doctorveera/status/1603596231070175236?s=20

3. GWAS of cellular morphological traits in N=207 reveal impressive and interpretable results.
https://twitter.com/doctorveera/status/1613363164007260160?s=20

As we make progress in molecular phenotyping, we will see more of such GWAS leading to important discoveries even in sample sizes <1000.

Instead of putting hard thresholds for sample size, I'd recommend advocating for good study designs, statistical power consideration, replication effort and better phenotyping.

Major human genetic discoveries made with small sample size

Beautiful story of how human population genetic and proteomic studies have led to the discovery of SVEP1 as the natural ligand of the receptor PEAR1. It's been a while since I read such a fantastic paper. 🧵
https://www.nature.com/articles/s41467-023-36486-0

I am big fan of GWAS of molecular traits as they often reveal  clear and interpretable cis and trans signals reiterating the known biology of the trait.

For e.g. a GWAS of a protein will reveal genetic variants that directly modify the protein expression in cis and also, genetic variants that indirectly modify the protein expression in trans.
https://twitter.com/doctorveera/status/1514381131990413321?s=20

But I've always wondered, if the cis and trans pQTL signals can reveal known molecular interactions/connections (receptor-ligand, protein-TF), they should also be able to reveal previously unknown molecular interactions.

SVEP1-PEAR1 is one such examples where a receptor-ligand discovery has been made via human proteomics (and also genetic) studies.  Just for this fact, the paper by Elenbaas et al. will serve as an important reference in proteomics literature.
https://www.nature.com/articles/s41467-023-36486-0

The background story of GWASs identifying disease associations for SVEP1 and PEAR1 in parallel without any knowledge of the secret affair between the two is fascinating.

The story begins with the discovery of an association between a missense variant in SVEP1 (p.D2702G) and coronary artery disease that appeared to be mediated not via the classical risk factor, lipids, but via something else, a novel mechanism.
https://www.nejm.org/doi/10.1056/NEJMoa1507652

SVEP1 codes for sushi, von Willebrand factor type A, EGF, and pentraxin domain-containing protein 1 (what an interesting name!), which is an extracellular matrix protein.

The authors followed up the association via functional studies to understand more about the role of SVEP1 in CAD. They found SVEP1 is expressed is vascular smooth muscle cells and acts as a proatherogenic factor and a potential target.
https://www.science.org/doi/10.1126/scitranslmed.abe0357

Meanwhile studies linked SVEP1 genetic variants with other diseases including blood pressure, type 2 diabetes, septic shock and glaucoma.
https://www.nature.com/articles/s41467-020-20851-4

As the proteomics started catching up with SVEP1 links with CAD  and other diseases (T2D) became stronger with the ability to identify stronger genetic instruments that are cis pQTLs of SVEP1.

For e.g. this Icelandic study by Emilsson et al. mapped the genetics of ~5k proteins in ~5k individuals and chose to highlight  the SVEP1 associations among others.
https://www.nature.com/articles/s41467-022-28081-6

GWAS of SVEP1 revealed two hits. One cis (near SVEP1) and one trans (near PEAR1). The authors focussed only on the cis variants and ignored the trans variants (they had no idea the trans signal was actually pointing to the very receptor that SVEP1 binds to)

While SVEP1 associations were evolving, in a parallel world, scientists were learning about PEAR1 through its strong associations with platelet aggregation response (an important biological process in relation to atherosclerotic plaque formation & rupture)

PEAR1 association with platelet response to agonists such as ADP is so strong it was found in an early study in mere 500 individuals published in 2009.

Unlike cis signals, trans signals will have small effect sizes and so will require large sample sizes. The trans signal at SVEP1 for platelet aggregation appeared only in 2021 in a pQTL study of 3.8k individuals (that too only via gene based test).
https://www.nature.com/articles/s41467-021-23470-9

In the current paper,  Elenbaas et al. connected all the dots, guessed that SVEP1 and PEAR1 are receptor-ligand pairs, and went on prove the same using a series of proteomics, genomics and functional experiments.
https://www.nature.com/articles/s41467-023-36486-0

The authors reproduce the strong cis signal at SVEP1 and trans signal at PEAR1 for SVEP1 using the pQTLs data from INTERVAL study.

Remarkably the trans association in Chr 1 with PEAR1 missense variant is seen only with SVEP1 among the 2994 proteins assayed in INTERVAL study.

Based on pQTL results from an independent study ( by deCODE) the authors show strong inverse relationship between PEAR1 and SVEP1 and also show that the causality direction is from receptor (PEAR1) to ligand (SVEP1) (i.e. increase in receptor decreases ligand and vice versa)

Through molecular assay, the authors show that indeed SVEP1 binds to PEAR1 with high affinity and this affinity drops down my many folds in the presence of PEAR1 missense variant that associates with SVEP1 levels in trans.

Then the authors go on to map the molecular pathways downstream of SVEP1-PEAR1 binding: AKT and mTOR signaling in vascular smooth muscle cells.

Given that human genetic studies suggest that SVEP1 inhibition might be cardioprotective, the authors studied SVEP1 KO mice. While loss of SVEP1 is lethal, partial loss seems to be well tolerated in mice in normal and disease backgrounds. But no evidence for cardioprotective role

Given the strong human genetic associations that increased SVEP1 is deleterious, the authors suspect that there might be "a safe therapeutic window exists to target SVEP1 and/or PEAR1 and potentially reduce their associated disease"

Though therapeutic value of the findings is still not clear, how the discovery that SVEP1 is the physiological ligand for PEAR1 unfolded was extremely fascinating. A great example of human genetic findings leading to incredible biological insight.

Human genetics led discovery of a receptor-ligand pair

New GWAS of ADHD based on 38,691 cases and >180k controls.
Congrats! @DemontisDitte, @AndersBorglum, @iPSYCHdk and collaborators.  🧵
https://www.nature.com/articles/s41588-022-01285-8

Despite a higher prevalence (~5%) than schizophrenia (~1%), ADHD GWAS lagged far behind Schizophrenia GWAS. The latest GWAS of schizophrenia included >76k cases.
https://twitter.com/doctorveera/status/1513349006750994441?s=20&t=DYhRtjXpXU92HOK-jDQcSw


Two major reasons:
1. ADHD is a childhood onset condition and most of the biobanks (both volunteers based and hospital based) are skewed towards older individuals.

2. Doctors started giving ADHD (and ASD) diagnoses only in the past two decades. So, it's rare for those with ADHD born in 1980s and before to have received a proper diagnosis. (it's important to factor this when including adults as controls)
https://link.springer.com/article/10.1007/s00228-012-1265-y

So, ADHD sample size didn't take off as quickly as other psychiatric disorders such as schizophrenia, bipolar disorder and depression.

Thanks to @iPSYCHdk, which is based on a Danish birth cohort and comprise only those born after 1980. Hence it became a major biobank source for individuals with ADHD and ASD diagnoses.
https://www.medrxiv.org/content/10.1101/2020.11.30.20237768v1.full

First GWAS of ADHD was published in 2018 based on 20k cases and 35k controls by the iPSYCH and PGC researchers with a major contribution from iPSYCH.
https://www.nature.com/articles/s41588-018-0269-7

Five years later, the authors have doubled this sample size with, again, the majority of cases coming from the iPSYCH cohort.
https://www.nature.com/articles/s41588-022-01285-8

The new larger sample size has more than doubled the number of GWAS loci, brought clarity to many secondary analysis in terms of enrichment for brain expressed genes, strong genetic correlations with other psychiatric, addiction and cognitive traits.

The polygenic score using new sample size starkly captures the negative relationship with cognitive functions like reading skills, verbal reasoning and importantly attention skills (as the name suggests).

My favorite finding though is converge of common and rare variants at some of the GWAS loci. For example, the authors highlight a locus in chromosome 10 where both common and rare variants in SORCS3 increase the risk of ADHD.

SORCS3 codes for transmembrane receptor protein expressed specifically in brain, has some important synaptic function and has some interesting links to Alzheimer's. Mice lacking SORCS3 show behavioral abnormalities including loss of fear.
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0075006

Overall the current GWAS represents a great progress in our understanding of genetic architecture of ADHD. Congrats again to all the authors, especially my PhD supervisor @DemontisDitte who has been recently promoted to Professor 👏 👏
https://biomed.au.dk/display/artikel/inaugural-lecture-by-ditte-demontis

New GWAS of ADHD captures convergence of common and rare variants in SORCS3

Great question! In fact, I've been recently in search of examples of human genetic variations that cause tissue-specific knock-out and knock-in. Few examples that I know now 🧵
https://twitter.com/SherifMorris/status/1619730369334624258?s=20&t=RBQDeFapsMSAHFVnNoAcMA

1. A Greenland specific nonsense variant (p.Arg684Ter) knocks out the gene TBC1D4 only in the skeletal muscle (the variant causes late termination and so affects only the long isoform that is expressed only in skeletal muscle)
https://www.nature.com/articles/nature13425

TBC1D4 regulates insulin mediated GLUT4 translocation to cell surface. GLUT4 is essential for glucose uptake into muscle and adipose tissue, especially after a rich carbohydrate meal.

As this variant knocks out TBC1D4 only in skeletal muscles, it increases risk of diabetes only in homozygous state by increasing postprandial blood glucose levels. This variant is a natural experiment to demonstrate the importance of skeletal muscle in glucose regulation.

And it turns out, even the homozygotes can regulate their postprandial blood glucose levels to the same level as wild types just with moderate daily exercise. So, this variant also demonstrates the importance of physical exercise in preventing diabetes.
https://cbmr.ku.dk/news/2020/arctic-inuit-with-a-faulty-gene-could-control-their-diabetes-with-an-hour-of-daily-exercise/

2. Noncoding mutations in intron 2 of HK1 results in expression of hexokinase in pancreatic beta cells and cause congenital hyperinsulinism.  HK1 is normally not expressed in beta cells and liver.
https://www.nature.com/articles/s41588-022-01204-x

These noncoding rare mutations are natural experiments demonstrating the consequence of knocking in HK1 in a human tissue where it is normally not expressed. More in this thread.
https://twitter.com/doctorveera/status/1473879664422637568?s=20&t=RBQDeFapsMSAHFVnNoAcMA

3. A non coding structural mutation results in expression of ASIP throughout the body, including brain hypothalamus, and causes morbid obesity.  ASIP is normally expressed mainly in skin.
https://www.nature.com/articles/s42255-022-00703-9

This variant is a natural experiment demonstrating the consequence of knocking in ASIP in hypothalamic neurons where it is normally not expressed. This example is special as here there is an extraordinary agreement between mice and human physiology.
https://twitter.com/doctorveera/status/1605773329532456960?s=20&t=RBQDeFapsMSAHFVnNoAcMA

I guess there are many more examples like this either already discovered or yet to be discovered (they surely exist, given the human population size and number of generations humans have passed through already)
https://twitter.com/doctorveera/status/1444449430867005441?s=20&t=RBQDeFapsMSAHFVnNoAcMA

Having said that, I agree that animal models are indispensable for drug target discoveries and certain complex experiments are easy and perhaps possible to perform only in animal models.

But there will be more success if we combine human genetics with animal models, rather than trusting only animal models.
https://www.nature.com/articles/ng.3314

Examples of human genetic variations causing tissue-specific gene knock-out/knock-in

I am always on the lookout for cool examples of human genetics informing drug adverse effects. Here is an interesting one I recently came across in this great review article 🧵
https://www.nature.com/articles/s41576-022-00572-8

Diacylglycerol acyltransferase 1 (DGAT1) inhibitor is one of the many drug candidates designed purely based on biochemistry knowledge and animal experiments but never saw the light of the day.
https://www.ahajournals.org/doi/10.1161/01.atv.0000151874.81059.ad

The final step of triglyceride synthesis is catalyzed by two structurally different but functionally similar enzymes: DGAT1 and DGAT2. The two enzymes account for nearly all the triglyceride synthesis in the body. The loss of one is likely compensated by the other.

So, drug developers found DGAT enzymes as an attractive target to treat obesity and type 2 diabetes. Scientists started exploring the consequence of partially or completely deleting DGAT1 or DGAT2 in animal models.

The two enzymes had interesting differences in their tissue distribution in humans. While DGAT1 is expressed predominantly in the intestine and fat tissues, DGAT2 is expressed highly in the liver, fat and skin.

Animal experiments revealed that DGAT2 deletion is lethal. The mice completely lacking DGAT2 died soon after birth due to lethal dehydration caused by rapid water loss through their defective skin. So, that ruled out DGAT2 as a good target.

On the other hand, the scientist found that the mice lacking DGAT1 were viable, metabolically healthy and lived longer than the wild-type mice and importantly, showed impressive resistance to diet-induced obesity and glucose intolerance.

This led to the development of DGAT1 inhibitors as a potential treatment for type 2 diabetes and obesity. For e.g. in 2012, AstraZeneca (AZ) scientists published the discovery of a small molecule AZD7687, a potent and selective inhibitor of DGAT1.
https://pubs.acs.org/doi/full/10.1021/jm301296t

In 2014, the AZ scientists reported the results from their first clinical trial. A single dose of AZD7687 resulted in dramatic reduction in postprandial triglycerides rise. But there was a problem: the participants had severe gastrointestinal side effects.
https://pubmed.ncbi.nlm.nih.gov/22950654/

The authors concluded
"The attenuating effect of AZD7687 on postprandial TAG excursion provides proof of mechanism with respect to gut DGAT1 inhibition. However, dose and diet-related gastrointestinal side effects may impact further development of DGAT1 inhibitors."

In 2015, they reported results from a follow up trial done mainly to evaluate the safety and tolerability of AZD7687. The drug was administered in multiple doses and the participants were evaluated if they tolerated GI side effects over time.
https://pubmed.ncbi.nlm.nih.gov/24118885/

Unfortunately, they didn't. With increasing doses, the diarrhea became worse and >50% of the participants discontinued treatment. The trial hammered the final nail in the coffin of the drug programme.

The authors concluded that "lack of therapeutic window owing to GI side effects of AZD7687, particularly diarrhea, makes the utility of DGAT1 inhibition as a novel treatment for diabetes and obesity questionable." and killed the drug programme.

The part of the DGAT1 story that interested me most is a case report on two children who were were knockouts for DGAT1 published in JCI in 2012.
https://www.jci.org/articles/view/64873


In this case report, the authors describe two pediatric cases of congenital diarrheal disorder (CDD) born in an Ashkenazi Jewish family and found to be full knockouts for DGAT1.

Case 1 was a girl child born normal but soon developed severe vomiting, colicky pain and watery diarrhea on the 3rd day after she was fed breast milk. She couldn't take anything orally and had to be put on total parenteral nutrition.

Puzzlingly, the child had elevated triglycerides and cholesterol blood levels (so did her mother who had partial DGAT1 deficiency) unlike what was seen in the DGAT deficient mice.

The poor child couldn't tolerate the parenteral nutrition and eventually died due to malnutrition and sepsis at 17 months of age.

Case 2 was a boy child, the girl's brother, also had a similar history: born normal but on 3rd day started having vomiting and diarrhea and had to be put under total parenteral nutrition, which he tolerated and recovered from malnutrition.

The boy too had elevated blood lipid levels and so had to be treated with cholestyramine which reduced his blood lipid levels. He made a full recovery and at 47 months of age, he was thriving on an unrestricted diet.

Exome sequencing of the family revealed that both the kids were homozygous for splice donor variant that skipped an entire exon resulting in complete loss of DGAT1 activity. The chances of observing this homozygous mutation in general population is 1 in 50-100 million.

It's amazing to see how a rare genetic disease informs about the adverse effect of a drug designed to treat common diseases. It's not clear though how DGAT1 loss (or inhibition) lead to diarrhea only in humans.

It seems, unlike in humans where only DGAT1 is expressed in intestine, in mice both DGAT1 and DGAT2 are expressed in intestine. So, it's possible that in mice DGAT2 enzyme in the intestine compensates for DGAT1 loss.

Anyways, DGAT1 adds to the list of genes where animal models failed to recapitulate human physiology and reminds us the fact that at the end of the day the ultimate animal model is human themselves.
https://twitter.com/doctorveera/status/1616283586797400065?s=20&t=sRw612pvsu_defrcURXFFg

DGAT1 inhibitors and diarrhea--an example of human genetics informing drug adverse effects

My first favourite GWAS of the year.
Here the authors derived >3k cellular morphological traits using microscopic images of iPSCs from 297 donors and studied their associations with common and rare variants. Amazing
https://www.biorxiv.org/content/10.1101/2023.01.09.522731v1
@soumya_boston @DrAnneCarpenter

In the rare variant analysis, the authors discovered three beautiful associations where pLOF burden results in big changes in cellular morphology.

For e.g.,
pLOF burden in WASF2 ( a gene critical for actin cytoskeleton formation) disrupts the shape of the cytoplasm of the iPSCs
pLOF burden in PRLR (prolactin receptor) disrupts the pattern of mitochondrial distribution around the organelles.

The advantage of identifying these associations in cell lines is that you can validate them by knocking out the genes in the same cell lines and observing the changes in the associated phenotype.

And that's what the authors did. Look at how impressively these cellular morphological phenotypes are altered when you knock out the genes.

It feels so great to see where the GWAS field is heading. As imaging and cellular assay technologies advance, these are the kind of phenotypes we will be GWASing in the coming years.
https://twitter.com/doctorveera/status/1603596231070175236?s=20&t=WE9PJlKTEfSVZiPNmGdiZg

With the ability to quantify traits at the level of individual cells derived from patients, imagine the kind of discoveries and insights we will be able to make.

Congrats to the authors for this amazing work 👏 👏

Genetic basis of cellular morphology

Better late than never. Happy and relieved to share a final published version of one of my PhD projects in which we uncovered a fascinating genetic link between language ability, psychiatric risk and creativity. 🧵
https://www.nature.com/articles/s41598-022-26845-0

Endophenotypes are at the heart of human genetics research. They not only help discover disease genes, but they can also inform about the disease's pathophysiology, signs, symptoms etc.

Unfortunately in the field of psychiatric genetics, we are not blessed with as many endophenotypes as scientists in other areas such as cardiology, metabolism and immunology are blessed with.

We try to make the best out of the minimum that we are able to observe/measure in the patients. In this project, we look at the cognitive profile of six major psychiatric disorders through the lens of school performance at around 14-17 yrs of age.

14-17 yrs of age is a critical developmental time point in humans as the brain is still developing briskly, and it is when most psychiatric disorders lying dormant within susceptible humans start showing their real face.

Thanks to iPSYCH, a Danish register-based cohort that offers a unique opportunity to link genotypes with phenotypes that span nearly the entire life course of its participants through the extensive linkage to national registers.
https://www.nature.com/articles/mp2017196

We asked a simple question. How did the iPSYCH participants (who later in their life developed or didn't develop psychiatric disorders) fare in the school?

Particularly, we were interested in how they performed in specific subjects namely language and mathematics that might require proper functioning of different cognitive domains.

I believe I don't have to convince any of you that some of us are good at math but bad at language, some of us are good at language but bad at math and a lucky few are good at everything, and these inclinations have a genetic basis.

So, what did we find? We found that everyone who had a psychiatric diagnosis (except those with anorexia) performed poorly in mathematics compared to controls (those without disorders). But this might not come as a surprise to most of us.

What was surprising though is that we found no deficits in language grades (Danish and English) in any of the groups except ADHD. Even in ADHD, the language deficits were only modest compared to the magnitude of the math deficits.

Not only we found a lack of language deficits, but in some groups namely ASD, bipolar disorder and anorexia, we also found evidence for enhanced language performance.

And this whole pattern was true even among the controls when you correlate their subject performances with their genetic risks for psychiatric disorders (measured polygenic scores)

Even individuals who never developed any psychiatric disorders (within the follow-up period) but had a high genetic risk for the disorders performed poorly in math and better in language exams.

The contrast was most striking for schizophrenia polygenic risk: individuals with high genetic risk for schizophrenia were very good in language but bad in math and those with lower schizophrenia risk showed the opposite pattern. We replicate this in an independent cohort (TEDS)


So, it appears that psychiatric disorders start disrupting cognitive developments early in life (in some, long before the first symptoms appear). But interestingly they do not disrupt all cognitive domains similarly.  Some domains (such as language skills) seem to be spared.

And based on the genetic analysis, we think that, like how all humans lie on a continuum of risk for psychiatric disorders, the impact of the psychiatric risk on cognitive functioning, too, follows a continuum.

So, some of you might wonder how the early tuning of cognitive development by psychiatric risk affects one's life trajectory, for example, in terms of their career choices.

To find out that we did a polygenic risk score analysis in an independent cohort (Million Veteran Program biobank) and uncovered something extremely fascinating.

We found that individuals with genetic tendencies to perform poorly in math but better in the language in school, more often chose a career that requires high creativity skills (writing, acting, singing, dancing etc.)

However, we cannot confidently tell that these individuals took these professions because of their enhanced language skills. It could also be that they simply chose such careers because as they suited well to their poor math skills.

Either way, we were able to show a positive link between psychiatric risk and creativity that is known for centuries and was also shown previously via genetic analysis.
https://www.nature.com/articles/nn.4040

To conclude, using an extraordinarily unique genetic resource from Denmark (the iPSYCH cohort), we discovered a fascinating genetic link connecting language ability, psychiatric risk and creativity.

Thanks to my supervisors @DemontisDitte,
@AndersBorglum and all the co-authors for helping me to bring this project to the finish line.

Genetic links between language ability, psychiatric risk and creativity

Among the 14 disease endpoints meta-analyzed in GBMI, the largest difference in SNP heritability between ancestries was observed for gout. This is due to both differences in allele frequencies and effect size between ancestries.
https://twitter.com/yingWANGyw1/status/1612143974927720448?s=20&t=YwpRYM5JY-A6OTTy9U4hyA

For e.g. the SNP-h2 of gout is ~5% in EUR but ~9% in EAS. This difference is probably driven mainly by a single variant within ABCG2 (urate transporter), one of the two top genetic determinants of serum urate levels
https://elifesciences.org/articles/58615

The ABCG2 variant is present in ~25% of East Asians explaining 8.3% of the phenotypic variance vs  7.3% of the Europeans explaining only 3% of the phenotypic variance in Gout.