Contrary to what the librarian says to you when you are part of loud conversation, SSSH! time here is referring to the Self-Selecting Safe Harbor (SSSH) tool invented by Zach Stevenson and crew in the Patrick Phillips Lab at the U of Oregon.
There is a back story here. We are proud of our people at InVivo Biosystems (IVB). Some, like me, have been hanging around with IVB for quite a long time. Others, like Zach, come and go, but still leave their mark.
Zachary joined us when we were pre-merger Knudra Transgencis. He was fairly new to genome engineering, but Zach was a quick study. He became a master of CRISPR-based transgenesis which he leveraged in his next career move – helping him get into graduate school at the University of Oregon. Zach and the team at Knudra had tasked themselves with the aim of finding better tools for detecting genome integration. We needed efficient systems that help identify only the animals that have experienced genomic integration. Better yet, the tool would be most effective if only the desired genome integrated strains could survive after exposure to a toxic compound. During Zach’s time at Kundra, the idea floated around a bit, but it never got the legs of experiment implementation to demonstrate its feasibility.
Once in graduate school Zach teamed up with Megan J. Moerdyk-Schauwecker and Brennen Jamison in the Phillips Lab to get the real world evidence that demonstrated the idea can work. Their team chose the hygR gene, to determine if a split-hygR gene could be harvested as tool to identify integration in a safe harbor locus (Stevenson et al. G3. 2020)
The principle is simple – chop the hygR gene into two parts. Put the long part into the genome of C. elegans and put the other part in your transgene plasmid. Zach did this at the MosSCI ttTi5605 safe harbor locus. This transgenic target strain contains most of the hygR gene but is missing a critical segment needed for creation of a functional hygromycin B phosphotransferase. Next, their transgene of interest was made in a plasmid that also contains the missing hygR part. The trick now is to have the same sgRNA site in the plasmid and in the edited safe harbor site. The interaction of the plasmid and the genome when injected with CRISPR reagents renders a region of overlap of about 700 bp on each end of the insertion cargo that allows homology repair to do its magic. When designed right, you only need one sgRNA to initiate the DNA cuts that trigger efficient homologous recombination repair. This technique works great in C. elegans transgenesis. Add hygromycin B to the growth plates and only the genomic-integrated animals can survive. Whether it can work in embryo injections with other organisms remains to determined.
At IVB we are building on this to use our fast and easy CRISPR-sdm technique to place a small the small split-hygR fragment at any locus of the genome. This will allow us to drive large constructs 5 to 10 KB (and perhaps even 20-100 KB) into any native locus.
Bottom line: Getting some SSSH! time with this split-hygR technique can calm the frustration of the aggravated C. elegans researcher.
Five months into the COVID-19 pandemic, the world is at 29,119,433 confirmed cases of SARS-Cov2 infection, including 925,965 deaths (6 August 2020 https://covid19.who.int/), which is over 3% of the planet’s human population. Of all the persons infected with the virus, 3.2% have died. If COVID-19 pandemic follows the trajectory of the 1918 influenza, 1/3rd of the world population will become infected , and nearly 300 million deaths will occur in the next few years. To put it in perspective, that is more than twice the number of military and civilian casualties of World Wars I and II combined. The medical community is challenged in optimizing their response due to a highly diverse array of infection severity in the human population. Some individuals are asymptomatic but test positive for SARS-CoV-2 infection, while other exposed individuals experience severe, sometimes fatal COVID-19 infection [2–4]. Disease severity heterogeneity is in part due to patients with underlying health conditions or comorbidities such as hypertension and diabetes which we believe share a common pathophysiology of renin-angiotensin system (RAS) . Estimates for the number of people with an underlying condition for increased severity risk of COVID-19 are 1.7 billion persons (22%) . As a result, there is an urgent need to understand the common molecular and cellular pathophysiologic basis of patients with a diversity of comorbidities and identify those with a rare disease that are at high risk for clinical deterioration. Therefore a rare human disease may be the perfect physiologic model to better understand the disease and generate more individualized therapeutic medical responses and positive outcomes for higher risk COVID-19 groups.
Some Rare Disease groups may be hyper susceptible to COVID-19 infection
There is emerging evidence that RD patients have higher COVID-19 infection risk among all human populations. For example, patients with deficiencies in cellular chloride transport due to CFTR variants associated with Cystic Fibrosis (CF) are more prone to viral and bacterial infection . Yet, because COVID-19 is a newly emergent disease, clear correlation of outcomes in SARS-CoV-2 infection in CF are extremely limited in infected CF patients [8,9] but the concern remains high for these patient groups . Another RD group that is likely to be negatively influenced by SARS-CoV-2 infection, are patents with CADASIL. CADASIL (Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy) is caused by genetic lesions in the extracellular domain in NOTCH3 . Like CF, CADASIL appears to be linked to accelerated disease progression via Influenza virus infection . Yet the co-morbidity of SARS-CoV-2 infection with NOTCH3 pathogenic variations is only speculated to be associated with advancing CADASIL presentation . Clear evidence is needed in order for a pathological association between COVID-19 infection and CADASIL is supported or refuted. With this knowledge, appropriate therapeutic medical responses can be identified.
The Renin-Angiotensin System (RAS) is at the center for controlling severity of COVID-19 infection.
SARS-CoV-2 utilizes binding of ACE2 to gain entry into host tissues . The resulting internalization and attenuation of ACE2 enzymatic activity is potentially a significant contributor to COVID-19 disease severity. ACE2 is a critical negative regulator of angiotensin activity. Overactive RAS signaling is implicated in higher risk for cardiovascular , renal [16,17], and neurodegenerative  disease and diabetes . Angiotensin I (Ang I) signaling peptide is produced from the angiotensinogen precursor by the proteolytic activity of renin. In the canonical RAS signaling pathway, Ang I is catalyzed into Ang II by the ACE dipeptidase. Ang II peptide can bind two GPCR receptors AT1R and AT2R with opposing effects. Ang II, a high affinity agonist of the AT1R receptor, promotes inflammation, apoptosis and vasoconstriction, which often results in hypertension of the patient. The agonist activity of Ang II at AT2R receptors has an opposite effect of being anti-inflammatory, anti-apoptotic and vasodilating, which are activities that lead to lower blood pressure. The expression levels of these receptors will have profound influence on hypertension in tissues as well as the presence of Ang II metabolites. Ang II is catalyzed into Ang III by an amino peptidase that remove the N terminal asparagine and results in a molecule with increased affinity for the AT2R receptor and thus activates anti-hypertensive activity . Additional control of the hypertensive state is achieved via the ACE2 dipeptidase which catalyzes Ang I into Ang 1-9 and Ang II into Ang 1-7. Ang 1-9 acts as agonists of the AT2R receptor. Similar to the anti-hypertensive activity of Ang 1-9, the Ang 1-7 peptide also promotes hypotension but through agonist activity on a different GPCR, the MasR receptor .
Small Vessel Disease (SVD) is a major health issue.
SVD, defined as “perforating arteries, arterioles, capillaries, and venules” is currently associated with 20% of stroke and 40% of dementia . ACE polymorphisms have been associated with stroke-associated white matter hyperintensities  and ischemic stroke . CADASIL is one of the most common single-gene disorders causing cerebral SVD . Hypertension is a risk factor for SVD  and leads to vascular remodeling . Since Ang II leads to vascular remodeling via VSMC dedifferentiation [28,29], it becomes plausible that CADASIL variants in NOTCH3 participate in generalizable VSMC remodeling of SVD via altered RAS signaling activity which may predispose these patients to a hypersensitivity to COVID-19 infection. Yet direct evidence is lacking that CADASIL-associated NOTCH3 variants have altered RAS signaling activity that leads to higher viral infectivity and/or RAS stress response.
The zebrafish animal model is well suited to modeling CVD and CADASIL.
The zebrafish model organism is one of the fastest growing animal models. Walcot and Peterson have proposed zebrafish are a good model for cardiovascular disease due to “its morphological and physiological similarity to human cerebral vasculature, its ability to be genetically manipulated, and its fecundity allowing for large-scale, phenotype-based screens” . For instance a Tg(flk1:GFP) reporter can be expressed in blood vessels and be visualized by fluorescence microscopy. Although iPSCs allow for a more native context, the ability of zebrafish to reproducibly make microvascular structures make them highly attractive for SVD modeling. Further, expertise and skill in gene editing is allowing the rapid creation of gene knock-outs and knock-ins throughout the zebrafish genome. We now have the ability to humanize either by putting in patient gene variations at the zebrafish version of the gene. Or, swapping out the entire gene for the human gene coding sequence. The end result is a well controlled system for examining effects of clinical variants on gene function.
CRISPR Engineering can be used to install variants into zebrafish
The creation of precision gene edits in zebrafish allows accurate measurement of the functional consequence of a clinical variant. The use of CRISPR (clustered-regularly-interspaced-short-palindromic-repeats) guide RNA targets cas9 nuclease to a specific genomic locus and has become a common method for targeting DNA cleavage at a genomic locus near a clinical variant target site. The CRISPR method has become quite routine for genomic locus disruption through Non-Homologous-End-Joining (NHEJ) activity, which creates gene-disrupting indels at a cleaved locus [31,32]. More challenging to achieve is the use of Homology-Directed-Repair to create precision genome editing at a target locus . Often a donor-homology DNA is used to instruct the cell’s natural DNA repair mechanisms to insert a specific sequence of DNA at the cleaved locus. Some groups have developed ways to use donor homology sequences to guide precision insertion of content into the genome of iPSCs and create precision deletions  or precision insertion of reporter genes [35,36]. Yet, interference from NHEJ-mediated indels in attempts at precision editing can pose a problem when the researcher desires to isolate a line with only precision edits. Often biallelic editing occurs in an HDR attempt that results in a complex heterozygote. One allele may report to edit as desired, but often the other sister chromatid locus has an undesirable indel. Development of methods that suppress indel formation by avoiding NHEJ activity can be useful in the creation of a biallelic conversion that creates the desired HDR-mediated edit in both chromatid loci.
A zebrafish model system for assessing COVID-19 viral uptake sensitivity is set up by first humanizing appropriate genes (NOTCH3) and then installing CADASIL-associated variants.
Humanization increases the data relevance of animal models. In this project idea, we can create a humanized zebrafish expressing the hNOTCH3 coding sequence inserted in the first exon of the zebrafish notch3 gene. In a two step procedure, we first use HDR-directed CRISPR gene editing to insert a phiC31 transposase acceptor sequence (attB sequence: CGGTGCGGGTGCCAGGGCGTGCCCTTGGGCTCCCCGGGCGCGTACTCCAC) to disrupt the notch3 locus with early stop codon insertion. Next we rescue the null with phiC31 insertion of hNOTCH3 expression cassette, which is a proven method for inserting large cargo at site specific loci in zebrafish. The cassette cargo contains a self cleaving T2A peptide prior to the human hNOTCH3 coding sequence. This enables expression of the human transgene to be driven by the promoter elements of the zebrafish notch3 gene and yet avoid chimeric protein formation with the vestigial notch3 coding fragment. To create a clinical variant model, the same plasmid used for human hNOTCH3 insertion is modified to contain a clinical variant (C174Y). Heterozygotic animals are expected to be generated (trangene/+). Both the attB containing animal (“knock-out”) and the hNOTCH3(C174Y) are expected to not isolate as homozygous due to essential nature of NOTCH3, while the hNOTCH3(wt) is hoped to remain viable as homozygote. Once these goals and expectations are met, the animals can be used to explore CADASIL associated pathologies. For instance, viral particles expressing the S-peptide receptor binding domain can be exposed to normal and variant zebrafish to monitor for different rates of viral entry.
The uses of a humanized zebrafish model range from diagnostic to drug discovery applications.
Researchers at various institutions can use the animals to determine which human pathologies are conserved. For mechanism-directed drug development, researchers can use in silico methods to discover small molecules that can be screened for specifically restoring the cysteine balance and promoting normal gene function. For pathway-directed drug development, researchers can use the animals for RNA-seq experiment to discover biomarkers consistent with disease and then harvest these genes with high expression response to create fluorescent reporters of pathogenic activity. The end result of this project funding is a platform for rapid assessment and drug discovery in CADASIL-associated disease.
1918 Pandemic (H1N1 virus). 16 Jun 2020 [cited 28 Aug 2020]. Available: https://www.cdc.gov/flu/pandemic-resources/1918-pandemic-h1n1.html
Paces J, Strizova Z, Smrz D, Cerny J. COVID-19 and the immune system. Physiol Res. 2020;69: 379–388.
Gao Z, Xu Y, Sun C, Wang X, Guo Y, Qiu S, et al. A Systematic Review of Asymptomatic Infections with COVID-19. J Microbiol Immunol Infect. 2020. doi:10.1016/j.jmii.2020.05.001
CDC COVID-19 Response Team. Preliminary Estimates of the Prevalence of Selected Underlying Health Conditions Among Patients with Coronavirus Disease 2019 – United States, February 12-March 28, 2020. MMWR Morb Mortal Wkly Rep. 2020;69: 382–386.
Onweni CL, Zhang YS, Caulfield T, Hopkins CE, Fairweather DL, Freeman WD. ACEI/ARB therapy in COVID-19: the double-edged sword of ACE2 and SARS-CoV-2 viral docking. Crit Care. 2020;24: 475.
Clark A, Jit M, Warren-Gash C, Guthrie B, Wang HHX, Mercer SW, et al. Global, regional, and national estimates of the population at increased risk of severe COVID-19 due to underlying health conditions in 2020: a modelling study. Lancet Glob Health. 2020;8: e1003–e1017.
Bucher J, Boelle P-Y, Hubert D, Lebourgeois M, Stremler N, Durieu I, et al. Lessons from a French collaborative case-control study in cystic fibrosis patients during the 2009 A/H1N1 influenza pandemy. BMC Infect Dis. 2016;16: 55.
Colombo C, Burgel P-R, Gartner S, van Koningsbruggen-Rietschel S, Naehrlich L, Sermet-Gaudelus I, et al. Impact of COVID-19 on people with cystic fibrosis. Lancet Respir Med. 2020;8: e35–e36.
Cosgriff R, Ahern S, Bell SC, Brownlee K, Burgel P-R, Byrnes C, et al. A multinational report to characterise SARS-CoV-2 infection in people with cystic fibrosis. J Cyst Fibros. 2020;19: 355–358.
Manti S, Parisi GF, Papale M, Mulè E, Aloisio D, Rotolo N, et al. Cystic Fibrosis: Fighting Together Against Coronavirus Infection. Front Med. 2020;7: 307.
Wang MM. CADASIL. Handb Clin Neurol. 2018;148: 733–743.
Mizutani K, Sakurai K, Mizuta I, Mizuno T, Yuasa H. Multiple Border-Zone Infarcts Triggered by Influenza A Virus Infection in a Patient With Cerebral Autosomal Dominant Arteriopathy Presenting With Subcortical Infarcts and Leukoencephalopathy. Journal of Stroke and Cerebrovascular Diseases. 2020. p. 104701. doi:10.1016/j.jstrokecerebrovasdis.2020.104701
Williams OH, Mohideen S, Sen A, Martinovic O, Hart J, Brex PA, et al. Multiple internal border zone infarcts in a patient with COVID-19 and CADASIL. Journal of the Neurological Sciences. 2020. p. 116980. doi:10.1016/j.jns.2020.116980
Zhang H, Penninger JM, Li Y, Zhong N, Slutsky AS. Angiotensin-converting enzyme 2 (ACE2) as a SARS-CoV-2 receptor: molecular mechanisms and potential therapeutic target. Intensive Care Med. 2020;46: 586–590.
Paz Ocaranza M, Riquelme JA, García L, Jalil JE, Chiong M, Santos RAS, et al. Counter-regulatory renin-angiotensin system in cardiovascular disease. Nat Rev Cardiol. 2020;17: 116–129.
Ichihara A, Kobori H, Nishiyama A, Navar LG. Renal renin-angiotensin system. Contrib Nephrol. 2004;143: 117–130.
Nishiyama A, Kobori H. Independent regulation of renin-angiotensin-aldosterone system in the kidney. Clin Exp Nephrol. 2018;22: 1231–1239.
Almeida-Santos AF, Kangussu LM, Campagnole-Santos MJ. The Renin-Angiotensin System and the Neurodegenerative Diseases: A Brief Review. Protein Pept Lett. 2017;24: 841–853.
Henriksen EJ, Prasannarong M. The role of the renin-angiotensin system in the development of insulin resistance in skeletal muscle. Mol Cell Endocrinol. 2013;378: 15–22.
Bouley R, Pérodin J, Plante H, Rihakova L, Bernier SG, Maletínská L, et al. N- and C-terminal structure-activity study of angiotensin II on the angiotensin AT2 receptor. Eur J Pharmacol. 1998;343: 323–331.
Cha SA, Park BM, Kim SH. Angiotensin-(1-9) ameliorates pulmonary arterial hypertension angiotensin type II receptor. Korean J Physiol Pharmacol. 2018;22: 447–456.
Coupland K, Lendahl U, Karlström H. Role of NOTCH3 Mutations in the Cerebral Small Vessel Disease Cerebral Autosomal Dominant Arteriopathy With Subcortical Infarcts and Leukoencephalopathy. Stroke. 2018;49: 2793–2800.
Paternoster L, Chen W, Sudlow CLM. Genetic determinants of white matter hyperintensities on brain scans: a systematic assessment of 19 candidate gene polymorphisms in 46 studies in 19,000 subjects. Stroke. 2009;40: 2020–2026.
Zhang Z, Xu G, Liu D, Fan X, Zhu W, Liu X. Angiotensin-converting enzyme insertion/deletion polymorphism contributes to ischemic stroke risk: a meta-analysis of 50 case-control studies. PLoS One. 2012;7: e46495.
Choi JC. Genetics of cerebral small vessel disease. J Stroke Cerebrovasc Dis. 2015;17: 7–16.
Cuadrado-Godia E, Dwivedi P, Sharma S, Ois Santiago A, Roquer Gonzalez J, Balcells M, et al. Cerebral Small Vessel Disease: A Review Focusing on Pathophysiology, Biomarkers, and Machine Learning Strategies. J Stroke Cerebrovasc Dis. 2018;20: 302–320.
Renna NF, de Las Heras N, Miatello RM. Pathophysiology of vascular remodeling in hypertension. Int J Hypertens. 2013;2013: 808353.
Xu H, Du S, Fang B, Li C, Jia X, Zheng S, et al. VSMC-specific EP4 deletion exacerbates angiotensin II-induced aortic dissection by increasing vascular inflammation and blood pressure. Proc Natl Acad Sci U S A. 2019;116: 8457–8462.
Mondaca-Ruff D, Riquelme JA, Quiroga C, Norambuena-Soto I, Sanhueza-Olivares F, Villar-Fincheira P, et al. Angiotensin II-Regulated Autophagy Is Required for Vascular Smooth Muscle Cell Hypertrophy. Front Pharmacol. 2018;9: 1553.
Deneault E, White SH, Rodrigues DC, Ross PJ, Faheem M, Zaslavsky K, et al. Complete Disruption of Autism-Susceptibility Genes by Gene Editing Predominantly Reduces Functional Connectivity of Isogenic Human Neurons. Stem Cell Reports. 2018;11: 1211–1225.
Soldner F, Laganière J, Cheng AW, Hockemeyer D, Gao Q, Alagappan R, et al. Generation of isogenic pluripotent stem cells differing exclusively at two early onset Parkinson point mutations. Cell. 2011;146: 318–331.
Deneault E, Faheem M, White SH, Rodrigues DC, Sun S, Wei W, et al. or human iPSC-derived neurons from individuals with autism develop hyperactive neuronal networks. Elife. 2019;8. doi:10.7554/eLife.40092
How do you model splice variations in your animal model systems?
Splice variation is an important consideration in genomic analysis of patient variations and it is often overlooked (PMID: 29680930). It is estimate that 15%–60% of human disease mutations are due to splicing defect ( PMID: 29304370). So, with close to 40% of disease causing variation likely being attributable to splicing defects, it becomes an important variation to be able to model in functional studies to determine if the variant is pathogenic.
But let’s first look at the process of splicing and what is known.
This complex process is managed in a complex way. Certain cell types will favor one form of splicing, while other tissues will select other forms. This is the natural splice isoforms variation that gives us more than just the number of genes in the genome to control biological output. In fact, this is part of the explanation for why a C. elegans nematode or a zebrafish, with roughly the same number of genes as a human, have such different levels of output complexity. Currently the number of functional isoforms in humans may be an order of magnitude more than what occurs in the nematode. Furthermore, the ways splice variation can take place gets bewildering quick.
How is splicing observed in the patient?
Layer on top of this the aberrant spice variations that can cause disease, and we have a tough interpretation problem. Thankfully RNAseq is providing a huge amount of diagnostic discovery for splice variation. We can compare the splicing patterns in healthy populations with a patient suspected of a genetic disease and visualize where the splicing is going wrong (PMID:28424332).
Modeling Splice Variation in Animal Models
To reduce the complexity of biology and yet bring more comparative biology relevance, often we can take a human cDNA sequence and use it to rescue the function of the animal’s version of the gene. To do this, we use CRISPR to remove the animal’s version of the gene (a gene “knock out”). Next we take a human cDNA sequence optimized for expression in the animal and either replace the deleted locus or express the sequence in trans (at a safe harbor site using the promoter that is either endogenous to the removed gene, or a promoter well established for appropriate tissue expression). In C. elegans, we have been pleasantly surprised that more than half the time for orthologs of at least 30% identity, we can get significant rescue of the loss of function seen in the knock out. In zebrafish, we have started applying the same techniques of gene replacement. The result, a set of gene humanized animals where the conservation of biology means we are looking at functional outputs that are highly similar.
Missense variations are conceptually easy to model. An amino acid change that is pathogenic (ex R235Q in STXBP1) is is installed with CRISPR using a simple donor homology that instructs the cell’s HDR to alter the DNA coding for Q (glutamine) into a code for R (Arginine) in our “wildtype” humanized locus.
But how do we mimic a splice variation?
It is actually quite simple. We create a donor homology that makes any splice form of interest. We are not interested in the mechanism to answer “if” it occurs – RNAseq already answers that. We are after functional consequence. We want to answer “does a particular splice form in question have a measurable defect compared to the normal splicing.”
Let’s look at one of the patient examples in detail.
In the red we have 4 patients with a collagen gene splice defect suspected of involvement in their diagnosis for Ullrich Congenital Muscular Dystrophy. Since all persons have two copies of the COL6A1 gene, we can see that one copy is splicing normally while the other copy is defective and its splicing brings in a pseudoexon. “The resulting inclusion of 24 amino acids occurs within the N-terminal triple-helical collagenous G-X-Y repeat region of the COL6A1 gene, the disruption of which has been well established to cause dominant-negative pathogenicity in a variety of collagen disorders” (PMID: 28424332)
Creating Knock-in for Animal Model of Disease
In regards to disease modeling of splice variations, we use a cDNA rescue approach. The variation seen in the patient is made as a plasmid coding for expression of a modified cDNA. This cDNA contains the human gene code that is suspected of creating an aberrant spice variation. Using CRISPR techniques, the segment coding for human DNA is inserted into the genome, typically at the orthologous locus of the animal.
Modeling in the C. elegans nematode.
To model the COL6A1, we would first seek to understand the phenotype from loss of function of the animal’s ortholog version of the human gene. For COL6A1, this is the C16E9.1 gene in C. elegans. This gene is not well studied in the nematode, but does show high expression in the alternative life state of dauer.
The first step is to make a gene knock-out to remove the C16E9.1 gene from the worm genome. Next, a series of functional assays are run to determine if a functional defect can be detected for the C16E9.1 knock-out as a loss-of-function allele. For essential genes, the ultimate manifestation of loss of function is lethality as a homozygote. In other genes critical genes will often manifest with functional defect after a battery of functional screens are performed. Once a defect in activity is observed, human cDNA can be introduced to see if rescue of function can be obtained. When rescue is obtained with human cDNA, we know we are looking at conserved biology for gene function between the animal and humans.
Once we have rescue of function, the fun begins. We can use CRISPR to put in the exact content that RNAseq indicates is occurring in the human gene. The pseudoexon seen in one copy of the patient’s chromosome pair can be made in the animal. Often if the patient variant is problematic from a loss of function perspective where haploinsufficiency drives disease. When a defect is made as a homozygote in the animal, the effect is usually a severe phentoype (often lethal) and is similar to what is seen in the gene knockout. Yet in the specific case from above with the pseudoexon in COL6A1, we are dealing with a dominant negative effect, so the defective splice not only disrupts this protein, it also causes the good copy to fail to function properly. Animals homozygous for the pseudoexon defect may actually have a less strong defect phenotype than when the animals are made as heterozygotes. Creation of the patient’s heterozygous condition is achieved by crossing the splice-variant-containing humanized animal model into the wild type humanized animal model and examining the cross progeny for defects in activity.
Modeling in the Zebrafish.
We can do similar modeling in zebrafish using the Tol2 system. In zebrafish there is one ortholog for the COL6A1. The col6a1 zebrafish gene has 55% sequence identity and 70% sequence similarity to humans. Like the work in the C. elegans nematode, we can remove the native gene and look for functional consequences. CRISPR techniques are used to create a knockout by inserting a stop codon early in the gene. If designed right, this results in loss of all expression for col6a1. Next we can measure the functional consequence of the gene knock-out by first trying to see if the animal can be made homozygous. If it is not lethal, the animal can be screened by a battery of assays to determine if a functional defect exists. Finding either lethality as homozygote, or observing a functional defect, allows testing for capacity of human COL6A1 cDNA to rescue function. A gene insertion approach using Tol2 is used to bring in the cDNA with an appropriate tissue-specific promoter. Rescue of function in specific tissues, for instance with the use of the 195 bp unc45b promoter for skeletal muscle expression (PMID: 27295336), will help elucidate the important roles of COL6A1 in dystrophy diseases.
The pseudoexon insertion defect seen in COL6A1 is a dominant negative variation. So, when a single copy of this gene is brought into the animal, it will have the capacity to suppress the activity of the unmodified copy of the gene. By inserting the cDNA with the patient into a safe harbor site we create a pseudo heterozygote. The dosage of the cDNA comes from two chromosomal positions while the wildtype locus provides expression of two copies of the normal gene. If the cDNA is dominant negative on its effect on the zebrafish gene, then defect of gene function will manifest.
Recap of Splice defect Modeling in Animal Models
In summary, the ability to model splice variants is done from a cDNA level. A modified cDNA rescue construct containing the human gene of interest is designed in three forms:
Positive Control (blue): The humanized wildtype cDNA provides a reference of the normal gene seen in healthy individuals.
Negative Control (red): A knockout deletion of the animal’s gene provides reference for full loss of function of the gene.
Test (yellow): A variant is tested for its functional activity. A range of activities is expected and depends on the pathogenic variant’s mechanistic role in disease pathology. It may be a dominant negative that creates a pathology worse than the loss of function allele because it binds to and causes bad behavior from the remaining good copy of the gene. Alternatively, the variant may cause loss of function. This will be either recessive and manifest as a homozygous, or it will be dominant and manifest by haploinsufficiency as a heterozygote. Finally, the variant of interest may cause a gain of function, which is typically manifest only the heterozygote.
Like many responsible citizens of the USA, I am hunkered down in self quarantine. Trying to come to grips with new uncertainty in the world and how it will have impact on my life. I find myself wanting to help. To do what I can to help stop this beast we call COVID19. But How?
Collective Behavior Modification is Part of the Answer.
What we are dealing with here mutually is a form of trauma. We are experiencing the Kuber-Ross grief cycle. Each of us are oscillating somewhere along its spectrum.
Many of us are stuck in denial. Although as of late, I seem to be bouncing from anger to bargaining to depression and back again (note, this entire piece for the blog is an elaborate form of bargaining!!!)
I want to get to acceptance and start doing constructive things.
Shedding the denial by attempting to understand the size of the problem.
If you somehow have been in a timewarp, or you have been tunneling under a rock for the past few months, COVID19 is a raging pandemic that is threatening to kill our parents/grandparents. And it is causing major economic crises throughout the world. Some of our leadership say this will pass in just a few weeks from now, but when you dive into the data, that suggestion is just plain ludicrous fantasy. It will take many months of concerted efforts by everyone to tackle the COVID19 problem. If we choose to ignore what we need to do, the ramifications are immense.
Some are calling it a war, so lets look at it from that lens.
The civil war was an extremely traumatic event for my home country of the USA. We are still dealing with its baggage to this day and its death toll was immense. World War II was a big effort, industries stopped what they were doing and channeled vast resources to the war effort. Many a dad and mum did not come home, but the concentrated effort of its population probably helped reduce loss of life. When we look at the loss of life from World War II we have a significant 290 deaths per 100,000. Everyone knew someone who did not come home. Yet, that death rate pales in comparison to what unmitigated COVID19 could do to us in the coming months if unchecked. More than twice as many dead if we do nothing.
Is All This Stuff Real? – Yes, ….This is Getting Real, …Real Quick
A recent article by a very respected group of researchers paints an alarming picture. Ferguson et al released a study March 16 that describes 3 scenarios each with different consequence on the health system measured in critical care beds occupied per 100,000 population.
Scenario 1. Do Nothing.
In this scenario (black line) we get the 2.2 million dead in the USA. The surge capacity of hospital beds is overwhelmed 25x. Many die that could have been saved, if we had the resources.
Scenario 2. Mitigation (“soft”)
In the mitigation scenario we have multiple options. We can close schools and universities. This surprisingly does not have much effect (green line). The algorithm takes into account “Household contact rates for student families increase by 50% during closure. Contacts in the community increase by 25% during closure” which offsets gains of social distancing achieved with school closure. Case isolation by itself has a stronger effect (orange line). For Case Isolation, symptomatic cases are asked to stay home for 7 days. This reduces “non household contacts by 75% for this period. Household contacts remain unchanged. Assumes 70% of households comply with the policy.” Add to Case Isolation a Home Quarantine where family sequesters for 14 days (assume 50% comply), and we get a boost that has more than halved the lethality rate. Finally add in social distancing in the greater than 70 year olds, where this group sequesters themselves away from close contact by avoiding crowd gatherings, maintaining 6 feet distance from strangers, avoiding restaurants and the like, and we get additional decrease in hospitalization (blue line). Yet even this multi-step mitigation effort is not enough. It is too soft to have the needed impact. Even with the multiple mitigation measures in place, the capacity of the health care system is still overwhelmed by more than 8x. More severe and hard measures are needed.
Scenario 3 Suppression (“hard”)
In mitigation, we are creating a decrease in the transmission number (R0 = R “naught”), which is the average number of persons infected by a person who is actively shedding the COVID19 virus. The do nothing R0 number is 2.4. This means an infectious person spreads COVID19 to an average of 2.4 persons. Mitigation decreases the R0 number, but does not drive it down to 1 or below. To get R0 below 1 where the infected are infecting less than one person, bigger steps need to be taken by the entire population. To implement effective suppression, we do the Case Isolation, as seen in mitigation, but now we add in General Social Distancing. We ask everyone to avoid getting together in groups exceeding 10 persons. Each of us should maintain 6 foot distance when talking to others. We should capture a sneeze or cough in the elbow or a tissue, and wash hands way more frequently than usual. Further, we should avoid touching our hands to face as much as possible. It is expected that “all households reduce contact outside the household, school or workplace by 75%. School contact rates are unchanged, workplace contact rates reduced by 25%. Household contact rates are assumed to increase by 25%.” Yet this his is still not enough to get to R0 down to 1, so we explore two more steps as options.
Option 1. (Case Isolation and General Social Distancing) + Household Quarantine
Option 2. (Case Isolation and General Social Distancing) + School and University Closure
We can see that adding in Household Quarantine, during the 5 months that suppression measures are in place, gets us just to the hospital bed surge capacity (red line). Yet in contrast, School and University Closure has an even more pronounced effect. During the 5 month period, we are well below the stress capacity of the medical services. We will know the success of our suppression approaches when there is no capacity problem being detected.
Herd Immunity – Resisting the Urge to Celebrate Too Soon.
Unfortunately, when the suppression measures are lifted, the COVID19 is due to come roaring back this fall. The advantage of option 1 (Household Quarantine) is it allows for better development of Herd Immunity. From wikipedia we have the definition of Herd Immunity as occurring “when a large percentage of a population has become immune to an infection, whether through previous infections or vaccination, thereby providing a measure of protection for individuals who are not immune” With an R0 of 2.4 for COVID19, roughly half of a population needs to be exposed, either by recovering from infection or use of effective vaccine, and R0 is driven down to 1 or less.
Pharmaceutical Interventions – what can we do?
Chloroquine. Currently there are no approved treatments as a vaccine or drug therapy against COVID19. Yet, if we had them in place, we might be able to more quickly get control of this infectious disease and be spared the estimated 12 to 18 months of various mitigation and suppression techniques – approaches that slowly build herd immunity at the least amount of deaths. Scientists and the pharmaceutical industry are rallying quickly get pharmaceutical interventions in place. And they have made some interesting findings. Hydroxychloroquine, less toxic than Chloroquine, has undergone a small clinical trial on COVID19 patients with very encouraging results. Gautret et al observed a dramatic 2-3 fold faster clearance of SARS-CoV-2 (the “official name” of the COVID19 virus) relative to untreated controls. The results became even more dramatic for another small group of patients that received both hydroxychloroquine and azithromycin – complete clearance in all patients was observed at day 4 of the 6 day observation window. Yet the numbers tested are tiny. So repeating this in larger populations will tell us if they are onto something. Nevertheless, it is a promising start. The current hypothesis is that this antimalarial drug blocs viral envelope fusion by altering the pH of the endosome and thereby slowing down the activity of the acid proteases present (cathepsins or possibly TMPRSS2). Yet it is important to consult a doctor first, because self medication has resulted in unnecessary death.
The action of chloroquine may be multimodal. In 2005, it was demonstrated chloroquine in a SARS-CoV infection of a cell line caused incomplete glycosylation of ACE2 and that it can have “an antiviral effect during pre- and post-infection conditions suggest that it is likely to have both prophylactic and therapeutic advantages.”
Camostat mesilate. In a drug targeting approach, Hoffmann et al monitored the classic endosomal-lysosomal entry for coronaviruses with an endosomal fusion assay. They found entry into the cytoplasm to be mediated by the activity of TMPRSS2 and cathepsin proteases. First, these authors made a comparison between SARS-CoV (the coronavirus causing the pandemic of 2012) and the SARS-CoV-2 (the coronavirus causing the current COVID19 pandemic) and found S protein In COVID19 is more prone to cleavage at the S1/S2 site. Next, they looked at inhibitors of cathepsin (E-64d) and TMPRSS2 (camostat) and found that, depending on cell type, inhibition by either protease could interfere with early fusion. When they looked at a lung cell line (Calu-3), they found that camostat could strongly inhibit early fusion. The site of cleavage for TMPRSS2 protease is the same site for furin and loss of this ability to cleave S-protein is critical for viral entry into the cell. Further Camostat is nearly established for efficacy and safety in humans for treating pancreatitis. In summary, these researchers found “SARS-CoV-2 can use TMPRSS2 for S protein priming and camostat mesylate, an inhibitor of TMPRSS2, blocks SARS-CoV-2 infection of lung cells.” So we have two very good candidate molecules for use in suppressing COVID19.
Smoking Gun – But Where is the Expression?
Although camostat can have a dramatic impact on early fusion and it appears to be acting on TMPRSS2 serine protease, what has puzzled me is the tissue specificity. If the primary mode of infection is the lung, then it stands to reason that lung tissue should have high expression of the protease. Yet when one looks at the tissue-specific expression profiles on the Human Protein Atlas (HPA), the expression of TMPRSS2 is absent in the lung.
TMPRSS2 is related to TMPRSS11D 55% sequence similarity and 40% identity). The Pharos Drug Database indicates the TMPRSS11D gene may also be involved in coronavirus fusion in the cell. Like TMPRSS2, the Pharos database indicates TMPRSS11D protease “plays a role in the proteolytic processing of ACE2.” Intriguingly, the TMPRSS11D gene has a signature for expression for being in lung tissue via Human Protein Atlas (HPA) (dark green bar). Further, examining ligands for these two genes via the Pharos Drug Database indicates both proteins share two ligands (compound 5 and its derivative CHEMBL1809251). Since these shared molecules have similar binding affinities between the proteins, it may be the topography of their active sites is similar. Although they both can cleave ACE2, it remains to be shown if they also have similar activities on the S protein. Yet if true, then we have two enzymatic targets for therapeutic development.
ACE2 expression also puzzled me. Its protein expression overview in Human Protein Atlas (HPA) shows no lung expression but expression in the gut is off-the-charts. Auguring in on two papers indicates that its expression does occur in the lung, but only in a small subset of lung cells . These references indicate many other tissues have high expression of of ACE2 (gut and throat). Further examination of tissue localization for SARS-CoV-2 indicates the following tissues exhibit high infection. We have our expected lung (alveolar epithelial cells) but we also have gut (mucosal enterocytes of the intestine, stomach, trachea/bronchus, distal convoluted renal tubule, sweat gland, parathyroid, pituitary, pancreas, adrenal gland, liver and cerebrum) and many of these tissues also express ACE2.
Using Molecular Dynamics Modeling to Dock Candidate Compounds
If an enzyme has a good crystal structure, one can do simulated docking of compounds to come up with categories of interacting molecules that can be used as hits for exploring their capacity to block viral entry into target cells. Regrettably crystal structures are lacking for both TMPRSS2 and TMPRSS11D. Not true for ACE2 gene, since the finding of its association to SARS in 2003, multiple crystal structures are available (6LZG, 6M0J, 6M17, 6VW1). In one recent structure, we can see the binding interface between CoV-2 and ACE2. One approach might be to design an interference molecule that “cloaks” the ACE2 molecule. Ideally, it would interfere with S protein binding but not block normal enzymatic activity.
Another possible target for drug development is the main protease (“Mpro”). The main protease (also called “3CLpro”) is a protease that helps process the long protein polymer made from the viral genome into functional fragments. The main protease as been derived for its structure at the molecular level. A peptidomimetic α-ketoamides as broad-spectrum inhibitors has been designed and if safety profiling indicates it has low toxicity against human proteases, screening other drug candidates for interaction could prove to be therapeutically useful.
Time is of the essence – the evolving landscape
Finding new therapeutic approaches quickly is important because as COVID19 spreads, the virus is able to explore diversity through mutation. Increasing levels of heterogeneity can be expect for a single stranded RNA virus – errors in the genome during the viral replication cycle will accumulate. A recent study was made available on the web that looks at genetic diversity among COVID19 strains. Especially disconcerting is that as the virus spreads, different strains are arising with different mutational lineages. Mutations in the S-peptide binding to ACE2 or in the S1/S2 cleavage site could render a new lineage that is more virulent. This is important because the binding affinity (Kd) of SARS-CoV-2 is nearly 5x more than Sars-CoV.
Fight Viral Diversity with Human Diversity
The natural diversity of human populations might offer some defense against viral evolution. The Gnomad database is a good resource for examining the diversity in humans. If we look at the binding interface between S-peptide and human ACE2, we cans see if there are humans with variations at the binding interface which may disrupt COVID19 infections. There are multiple EM structures to reference for examining the binding interface (6LZG, 6M0J, 6M17, 6VW1). Using the 6M17 structure, Yan et al showed the binding interface is in close proximity to many residues in ACE2 (Gln24, Asp30, His34, Tyr41, Gln42, Met82, Lys353, and Arg357).
Many gene variants seen in human populations exist at this binding interface. One intriguing residue is Lys26Arg. This genetic variation exist in 600 for every 100,000 persons. Although the change may not seem to be too drastic – a positive charged amino acid is substituted with a similarly charged amino acid. Yet we know that it is arginine, and according to a prior blog post, the arginine amino acid hold special privilege for its involvement in population variation analysis.
We can float two hypothesis regarding this Lys26Arg variant
Hypothesis A – Resistance: Persons with the Lys26Arg human variant might have an ACE2 protein with disrupted interaction to the S-protein spike. The S-protein becomes highly compromised for tricking this ACE2 protein into helping it get inside the cell.
Hypothesis B – Sensitivity: Persons with the Lys26Arg human variant might have an ACE2 protein exhibiting stronger interaction to the S-protein spike. The S-protein can use this ACE2 protein to get inside the cell more efficiently.
The Problem of Heterogeneity
We are diploid organisms which means we have two copies for every gene. In regards to hypothesis A, this mean most of the persons carrying the Lys26Arg have only one copy. The other copy is the common natural variant (“wildtype”). Because they harbor a wild type variant, these persons at best would about 50% less susceptible. Thus hypothesis A for COVID19 resistance would be a subtle effect. If on the other hand, the variant had 10x has more binding to S-protein, then persons carrying this variant could be more susceptible at greater than 2 fold effect. Systems that could measure these binding effects in diploid animal formats would be elucidating for which hypothesis dominates for a given variant.
In summary, there are promising targets for developing therapeutics and vaccines. One target is the interaction of SARS-CoV-2 with ACE2. Another target is the activity of the human proteases (TMPRSS2, cathepsins and possibly TMPRSS11D) that process and cleave the S2 fragment of the spike (S-protein) allowing it to have easier access into the cell. And finally, the third target is the main protease, the enzyme that processes the polypeptide made from the mRNA transcript of the COVID19 genome.
You know when that hunch seems to get reinforced over and over again, then your mind starts speculating it as a fact.
!Danger! Will Robinson… it’s time for a serious fact check.
My hunch was that the amino acid arginine (Aka: “Arg” or “R) seems to be showing frequent association with pathogenicity. It started with the observation that many of the established pathogenic variants in the coding sequence of STXBP1 seem to involve a preference for arginine. Extracting from ClinVar for missense that are pathogenic and likely pathogenic gives the following table:
Indeed arginine (R) is disproportionately represented. Assuming all amino acids as equals, then there should be 4.3 for each amino acid. Disproportionally low are things that make sense. Like methionine (M), only one codon (ATG) instructs for insertion of this amino acid in a sequence. Similarly tryptophan (W) also has only one codon (TGG). These two amino acids should be represented below the average. A little bit oddly, we have similar low levels from lysine (K), phenylalanine (F) and glutamate (Q) who each have two codons. If codon dosage was key to variant proportioning, then these should have been seen at least 2x more than M and W, so perhaps something more than codon dosage mediates amino acid choice in creating pathogenic variations.
Arginine has 6 codons which still could drive its outsized proportion in the graph. Yet Serine (S) and Leucine (L) also have 6 codons. But respectively they are at 7 and 3 for being involved in pathogenicity. Only mighty arginine accounts for 13 of the 43 pathogenic variants in STXBP1 (30%). Tempering my enthusiasm is the observation that for 3 amino acid positions R292, R406 and R451, we have multiple changes being called pathogenic. Yet no other amino acid in the STXBP1 pathogenics has this changling capacity, so why is it that arginine is at high proportion in the assigned pathogenics – perhaps it is just a consequence of a biased investigator focus specific to STXBP1 and they fixed their gaze onto the repeating de novo clinical variants at positions 292, 406 and 451.
Is arginine involved in fragility elsewhere in the genome?
To normalize for possible investigator bias and find a method that can be applied to other portions of the genome, I took advantage of the Ensembl database to list and rank a gene’s codon sequence variants by bioinformatics analysis. Ranking on CADD was used to list protein coding sequence variations by their severity.
Ensembl allows us to identify which variations are theoretically likely to be disruptive of protein function. The choice to rank by CADD (stand for Combined Annotation-Dependent Depletion) allows us to use a sophisticated algorithm that avoids investigator bias because it intentionally avoids using “known” pathogenicity databases when it creates it ranking. A key test is to see if CADD can independently observe the pathogenicity known to exist in STXBP1. To construct the test, we compare the top scoring CADD variants with the lowest scoring CADD variants.
With CADD, we get an independent call for possible pathogenicity that still picks up what you might expect. Nearly half the calls in the Top-30 CADD pull up known pathogenicity and no benign calls are found. In the Bottom-30 CADD we get one known benign call and no pathogenics.
Healthy population data also is consistent. STXBP1 is autosomal dominant. That means you only need one of your two chromosomal copies to be defective and disease will occur. Selection pressure has been very tight on autosomal dominant genes. Variants in healthy population cannot occur at higher than the known frequency of the disease in the population. Published frequency in STXBP1 for causing early-infantile epileptic encephalopathy is 1/90,000. The largest healthy population database is in GnomAD. At 141,456 individuals, and the fact that STXBP1 needs to distribute across at least 43 pathogenic alleles, the likeliness of even one pathogenic variant being in healthy populations is pretty close to zero. Some of our Top-30 CADD have 1x or more frequency in healthy populations. Most of these are unassigned. For these unassigned that are seen at 1x or more, the disease frequency argument strongly implicates that they are benign variants.
So the CADD is not perfect, the top scoring hits are a mix of known pathogenic and probably benign. But the bottom scoring CADD seems to be more efficient at pulling out benign. In the Bottom-30 CADD, only one variant, I271V, is labeled Likely Benign by ClinVar, yet nearly everyone of these alleles (27 of 30) is seen in healthy populations, so they too are probably benign.
At this point in the analysis, we can pinpoint an anomaly. Y264C is labeled in ClinVar as a Likely Pathogenic. But from the population frequency argument, this assignment is highly unlikely. Y264C has been observed to occur in healthy populations. So a a bare minimum, it should be downgraded to a VUS, but probably be called a Likely Benign for causing early-infantile epileptic encephalopathy.
Finding Arginine-associated Fragility Throughout the Genome
This top-30 / bottom-30 approach was applied to a large set of genes. As a form of internal control, we add isoleucine (I) in the screen. With less conviction, I have felt this amino acid was associating with benign variants. If true, it should show an enrichment in the Bottom 30 CADD scores. So in my gene set experiment, I measured 4 bins. 2 bins for how many arginine and isoleucine in the Top 30 and 2 bins for how many arginine and isoleucine in the Bottom 30.
30% of top 30 CADD scoring variants contain arginine???!!!
An assumption of even distribution of amino acids, combined with an even more absurd assumption of an average 3.05 codons per amino acid, gives us 4.3% as average amino acid fraction per each 30 (dashed line). Arginine is 7.2x more than this average number. Yet, we need to account for the fact arginine uses about 2x more than the average codon usage. A a result Arginine bias in the Top 30 is about 3.5x more than expected. For isoleucine, the enrichment in the bottom 30 appears to be about 2x more than expected.
Test dataset – 30% arginine in Top-30 CADD prevails
The noisiest data in the Top-30 CADD appears to be the Arginine data. A cumulative trending plot was used to see how many genes were need before the trend to 30% becomes apparent. After assessing 7 genes the trend starts to stabilize. A new set of 7 genes were chosen. This time the genes were chosen from the Undiagnosed Disease Network (UDN). The UDN recently listed 54 genes as in desperate need for animal modeling to provide gene function studies. A sub-selection of these were identified as having good sequence similarity to genes in the animal models which we hold dear to our heart and expertise (zebrafish and C. elegans). The Top-30 / Bottom-30 CADD selection was applied to these genes and plotted for Arg and Leu enrichment. 30% prevails for arginine – it occurs at least 3.5x more than expected for being the top CADD variants as hypersensitive to substitution.
Most notable anomaly is arginine. 6 codons are use by arginine, but the observed frequency is low at 4.2%. To illustrate how low, they calculated the expected frequency for each amino acid biasing only for the GC richness of vertebrate genomes.
The expected frequency for arginine is quite high at about 10.5% due to its GC richness in its codons. Yet the actual observed frequency is quite low at about 4%. Based on this observed frequency, we bounce back – we now assess that we are observing arginine in the top 30 at 8x more than expected. No explanation for the anomaly and it just became more pronounced!
Taking a different approach, we can ask what percentage of ALL known pathogenic and likely pathogenic variants in a gene involve arginine substitution. 7 genes analyzed and we get the same 30% for arginine. Yet the calculations are that it should be below 4%. 8x more than expected prevails.
Are your arginines special too?
This analysis has uncovered a unique phenomenon. It appear everyone’s arginines are special. Exactly why arginine has this special status is not entirely clear. It is highly likely arginine has been strongly selected against its random incorporation during evolution. As a result of this strong negative selection (much more than what is happening for all other amino acids), arginine’s frequency in all proteins is much lower than predicted. The observed pathogenic sensitivity may be a read out of this hyperselectivity of evolution. Basically, arginine’s use in any given protein is very particular. A possible driver for this is arginine’s amazing capacity to bring high order to neighboring side chains in most protein structures. When it is gone, chaos reigns. When it is introduced where it should not be, chaos still reigns.
How prevalent are Variants of Uncertain Significance?
ClinVar database for variant interpretation was analyzed for its levels of ACMG-AMP assessments. With help from the data dumps from ClinVar Miner, the yearly distribution of assessments was plotted. Since 2016 and shortly after the ACMG-AMP guidelines came out in 2015, the number of assessments assigned to the VUS category has grown rapidly. These are the variants that clinical genetics researchers have examined, but cannot decide if they are pathogenic or not.
How big will the VUS problem get?
To estimate how large the VUS problem will become, we must first understand how big is the human genome. Controversy abounds, but current estimate are there are 21,306 protein coding genes and 21,856 non-coding genes. To be conservative, and for simplicity sake, let us use 20,000 genes as the number. The next question is how many of these are disease associated. When we look to ClinVar the number of “genes with variants specific to one protein-coding gene” we get 7221 genes. More conservatively, we can look to ClinVar’s “gene_condition_source_id” which list 4242 genes as being associated with a diagnostic condition. This lower number is reinforced by OMIM in which the “Total number of genes with phenotype-causing mutation” is 4162 genes. These list have been growing rather steady at 5% per year, so in a few years the likely number of gene-disease associations will probably approach 5000 genes, or roughly 1/4 the human genome.
VUS problem may eventually approach 7 Million variants
A recent attempt to preload the human genome with pathogenicity assessment potential has been made. InterVar database applied ACMG-AMP guidelines to ~80,000,000 amino acid positions in the genome to provide a database for easier variant interpretation. Since at least 20% of these positions are likely to be in genes with known disease association, there are roughly 16,000,000 variants that will eventually occur in patient-derived genome sequencing. If the current trend of 44% VUS translates across that number, then there will be close to 7,000,000 variants in need of functional studies to resolve their pathogenicity.
A novel animal model systems for rapid variant interpretation
The team at Nemametrix just produced a wonderful set of preliminary data that we showed at the recent American Society of Human Genetics. It shows it is possible to use a training set of known benign and pathogenic alleles in a gene to “teach” a ML algorithm to determine if pathogenicity is present in a VUS. When applied to the STXBP1 gene, a set of 5 benign and 5 pathogenic was sufficient to train for segregation in an LDA plot and the Y75C was assessed as pathogenic.
Once this type of system is trained with a set of known pathogenic and benign variants, the assessment of pathogenicity can be achieved in a soon as 10 days from start of a VUS transgenesis project.
In a prior blog post, the presence of dominant alleles in my genome gave me pause when trying to interpret the data from sequencing my DNA. Dominant alleles can be the cause disease when only one pathogenic variation occurs in only one gene copy of the chromosome pair. Contrast this to a recessive allele where you must get a defect in both chromosome copies of the gene to cause disease. In the recessive condition, if you only have one defective copy, you can expect to remain healthy, but you are a carrier of a disease allele. With the lack of immediate consequence to being a carrier status, many more individuals should be walking around with variations that are recessive towards disease. In fact, the CFTR gene variation (p.Arg117His) for Cystic Fibrosis that was highlighted for me in my Veritas Genomic sequencing report is quite common. It occurs globally at 1 per 2,500 persons, and that increases close 1 per 1,000 for northern europeans, which is a dominant portion of my ancestral genomic composition. In contrast, the CACNA1S variant (p.Arg419His) that most concerns me in my genome, has a prevalence of 1 in 25,000. Thats of low enough to be Rare Disease in Europe, but still probably way to high for disease manifestation rates.
Rare domination in CACNA1S needs to be rare enough to cause Hypokalemic Periodic Paralysis.
Dominant disease causality with the Arg419His variation in CACNA1S is unlikely because it is too frequent for the 1 per 100,000 population frequency for the disease of Hypokalemic Periodic Paralysis. Yet there are two variations known to be causative in CACNA1S, Arg528His and Arg1239His. Arg528His occurs at close to 1/100,000, while Arg1239His has yet to be detected in healthy populations. Clearly the Arg11239His is low enough population frequency to be causative for Hypokalemic Periodic Paralysis. Yet for my Arg419His, the frequency is too high for it to be causative. A variant effect that is Autosomal Dominant (AD) is extremely unlikely for my lone Arg419His allele.
If dominant alleles need to be rare in the population, how frequent is dominant status for variants of a disease?
The frequency of Autosomal Dominance (AD) for any given disease gene appears to be quite high. It is estimated that there are about 7000 Rare Diseases. If we assume the On-line Mendelian Inheritance in Man (OMIM) already represents most of these genes, then rare disease variants will map to the 4346 gene entries in OMIM with published allelic variations. Next, I listed these variations in blocks of 100 to reveals the number of genes for which they are known to exclusively Autosomal Dominant (AD) or Autosomal Recessive (AR), or some kind of hybrid.
When one runs down the inheritance pattern and tabulates them per gene, the first 100 variants have about twice as many genes in the AR category when compared to the AD category.
Running thru the another 400 more variants in the 100 variant blocks shows the trend continues – Dominance of a genetic conditions occurs for about 1/3rd of the disease genome.
Axiom for the individual : “I am not very dominating, but there are lots out there who are.”
So at the individual basis, it appears the AD status of pathogenic or likely pathogenic variants in your genome is very rare. Yet, at a population level, a large proportion of Rare Disease is caused by Autosomal Dominant variation. Rare disease calculate to occur at about 1 per 15 persons. So, for about 1 in 50 (150 million persons), their disease casing variation is likely to be Autosomal Dominant.
In today genomic medicine era, it remains challenging to understand the functional consequence of a gene variant’s contribution towards disease. Guilt by association is one of the criteria upon which a new variant is judged. We can look at healthy populations data and compare it to established Pathogenic and Likely Pathogenic variants. This helps us understand if a new variant may have a propensity to cause disease. The thought is that if a new variant is occurring at a region previously established as causing pathogenicity, then the new variant may be pathogenic too (ACMG guideline: PM1 “moderate” assessment criteria).
Is my variant guilty of pathogenicity because of its proximity to a pathogenicity hotspot?
In the image above, we see that there are hotspots (red) and coldspots (blue) for pathogenicity in STXBP1. The hotspot values were generated from the known Pathogenic and Likely-Pathogenic listed in ClinVar. The coldspot values (highMAF) come from variants seen in healthy populations. In yellow we have Variants of Uncertain Significance (VUS). Intensity of the peak is a measure of both how many times different variations are seen at an amino acid position and if their nearest neighbors have the same assignment. This plot suggest there are spots in STXBP1 that can tolerate sequence diversity (blue bars) and spots where a hit leads to pathogenic behavior (red bars). Further, the VUS are landing in both red bar and blue bar regions. Perhaps we can consider VUS to be either pathogenic or benign by this association? Yet, there is a critical assumption that leads to a question: How legitimate is it that every variant in healthy populations (“highMAF”) is ASSUMED to be benign?
2,504 healthy population genomes – Calculating the rare variants in each person
To dig into the validity (or invalidity) of this assumption, we can look to a large population study and ask how many times do we see variation and what are their types. The 1000 Genomes Project Consortium shows an average person has about 4,500,000 million variations. Of these, about 100,000 are somewhat rare because they are seen in less than 1/200 persons (<0.005 MAF). The even more rare “singletons” of the study occur at a frequency of 1 per 2504 persons. This restriction gives us about 10,000 more rare variations to think about per each person. Yet, to get even more rare and be able to ask the question how many variants per person meet the 1 per 200,000 USA definition for Rare Disease frequency, the study size would need to be 100x bigger. Nevertheless, we have interesting data reported in the 1000 Genomes study on healthy population variants that are also seen as pathogenic in Human Gene Mutation Database (HGMD) and ClinVar datasets. Filtering the observed path in healthy population as frequency per individual, every person can expect to harbor 20-25 variants of established pathogenicity.
A larger study by Karczewski et al. 2019 is approaching the scale need for assessing Rare Disease. A dataset of 141,456 human genomes (125,748 exomes and 15,708 genomes) was harvested from the wildtype controls used in various disease studies. The exomes observe variation mostly in the coding sequence of a gene, while the genomes record variant information across the gene (coding + upstream/downstream/introns). The result is a deeper measure of the frequency of missense variation that approaches the 1 in 200,000 genomes needed for Rare Disease designation. Currently the National Organization for Rare Disease (NORD) list 1258 disease in their database. STXBP1 cross references to two of these (Dravet and West Syndromes). Both of these syndromes each have a support group, which are two of the 283 total family foundation groups that are listed in the NORD member list.
Yet the situation for Rare Disease is larger. In the NIH’s Genetics and Rare Disease (GARD), there are 6264 unique genetic diseases listed. This suggest there are thousands of genes for which we can expect to have gene variant issues leading to disease. ClinVar currently list 7046 is the number of “Genes with variants specific to one protein-coding gene.” Basically it appears that a third of your 20,000 protein coding genes could take a hit that increases your risk or likeliness of coming down with genetic disease symptoms.
The GARD lists an intriguing statistics that 20-25 Americans are living with Rare Disease. The USA’s current population is 327.2 Million, so roughly 1 in 15 individuals world wide are probably living with rare disease. Assuming monogenic cause, then at least 51 million pathogenic might residing in the human population. Add polygenic burden and the number may be a multiple (100, 150, 200, 250….??) for variants associated with disease currently being experienced today. Guilt by association to hotspots and coldspots might provide some answer, but functional studies are the more definitive proof, and +50 million is a lot of animal models to build!!
What genes are good candidates for alternative animal modeling?
I set out to determine which important disease genes are good candidates for creating animal models in C. elegans. The first step was to turn to a database that has a comprehensive listing of human genes and their disease association. The DisGeNet database has nearly every human gene annotated for its level of disease association (17,549 genes as of June 2019). They provide a curated list that has 8400 genes with Gene-Disease Association (GDA) score of 0.1 or higher. For the top 1000 genes the GDA scores are 0.69 or higher, which indicates they scored high for having a significant disease association. These top 1000 were selected for examination of their ortholog status in C. elegans using the Diopt database. 749 othologies were detected, of which 411 had clear reciprocal nature (back-blast gives starting gene for the ortholog as best hit). The top 100 of these genes for high homology and detectable loss-of-function consequence were selected.
Tabulation of disease-associated genes with properties favorable for C. elegans humanization
The top 100 are tabulated in gene-alphabetical format below. These 100 genes have 8360 variants as known to be as problematic (Path, Likely Path, or VUS).
Use a search tool to quickly find out if your favorite gene occurs below.
(Note: gene knock out for 58% of these genes results in lethality.)
When you get the genomic report, you have a movement of trepidation. What will it say? ….Will it have a reveal that says you should do countermeasures immediately? ….Will it say something that you can do nothing about? The latter condition occurred for me. There were findings that had strong impact on my psyche.
Two things were called out heavy. A cancer risk of melanoma. Good thing my family, first my momma, and then my spouse, have been diligent in their liberal in the application sunscreen to the family. Once I googled and pubmed searched the MC1R(R160W) locus, I found the evidence was less than compelling for a dramatic change of lifestyle. Just keep the sunscreen coming and I will likely be fine.
The carrier result was a little more of a shocker. A good personal friend has a daughter homozygous in this gene. It was discovered in utero and they have been vigilant ever since. Their daughter is now in her teens. Doing exceptionally well and acting like any normal kid – currently enthralled with dance class and other outdoor activities. Preventative medicine done right. So getting tagged with a pathogenic in this gene is giving me mixed feelings. A mix of some worry and yet, almost pride. Even though my good friends don’t share my specific genetic lesion, it still feels very personal and connecting. Furthermore, this is one of the genes where modern genomic medicine is making great progress in understanding and treatment.
Will you too be a carrier of a pathogenic variation?
Carrier status is something all of us should expect. Veritas recently publicly disclosed at the Precision Medicine World Congress that their database has 90% of customer reports as returning with carrier status for at least one pathogenic variant. Recent discussions with Robert Green at Harvard confirm this – he showed me a large dataset that gave the number as 92% of healthy populations as being carriers for known pathogenic variants. You might think that there are a lucky few (10%) who are not carriers, but think again. The average person will have close to 3 million differences from the reference genome and this may be an underestimate. Distribute that unbiased across the genome and we have coding regions with close to 30 thousand variations. Since you have close to 20 thousand genes that means every gene has approximately 1.5 variations in it. Now lots of approximating, and does not factor in selection against bad variations. Yet in that quick calculation, the main message is every gene is likely to have a variation and some genes will have multiple variations. So the original question of how many of these are pathogenic, becomes difficult to approximate. Publications suggest we may have up about 1300 suspect variations hiding in our genome. Yet definitive variants with “known” pathogenicity is likely to be much lower in your genome.
Complicating this is issue is variable penetrance – a pathogenic variant in one family may behave with monogenic behavior in that family. While in another family, that same variation may be acting more polygenic – it needs other gene mutations to have pathos in the patient. It is behaving more like a “risk factor” for disease.
Pathogenic variant frequency in Chris Hopkins’ genome
The vagueness of my carrier status “kills” me, so I wanted to know in more. I contacted a good friend at the Rady Children’s Hospital. Dr. Matthew Bainbridge is a researcher who was a key contributor to the Rady’s renowned speed at using whole genome sequencing for rapid genetic diagnosis. Matthew introduced me to some software tools he has been developing. His company Codified Genomics has developed a variant analysis software that allows exploration of one’s genomic variants. All you need is your BAM or VCF files.
What’s that? …You don’t know what is a BAM file, …or a VCF?!!!
Dont worry, lets decode the jargon. In the clinphen journey to understand my clinical predilections, predispositions, and pathos, I found myself getting immersed into the intricacy of the end-to-end solution in genomic data acquisition and interpretation. What happens when you spit in a tube and put it in the mail? A lot of stuff! I came across an amazing guide to understanding the industry space behind genomic sequencing, the Enlightenbio Report. This help me get a tightly-focused view on the process of understanding one’s DNA.
That first box is what happens after you spit in the tube. The chemicals in the tube react with the cellular material in the spit to help stabilize it and prevent its degradation. This allows one to send the sample at room temp to the lab. On the receiving, the lab initiates a protocol to isolate the DNA that comes from the mouth epidermal cells that slough off into your spit. DNA is manipulated in such a way that it can go onto a microchip slide and set of DNA sequencing chemistry reactions are used to read out the DNA in small segments of sequence. Each of the millions of sequence segment reads is recorded as a fastq file. The fastq read segments are compared and aligned to a reference genome to make a BAM file. The BAM file alignments are processed to detect where sequence variation occurs, which is recorded as a VCF file. VCF files are analyzed by comparison to databases and assessments are made of each variant’s potential for pathogenicity. The assessment data is generally provided as a report to the clinician (or the intrepid genome wanderer such as myself). This report takes the raw data and massages it into a format for easier understanding of what is the baggage of one’s genome.
1604 suspect variations in my genome
Matthew helped me upload my VCF files into the Codified program. Next, he showed me how to wander around sifting the data by various aspect such as allele frequency, dominant and recessive status. known pathogenic genes, etc. The upload to Codified indicates I have exactly 1604 suspect variations occurring at an appreciable fraction of the reads and at positions inside, or in close proximity to, the coding sequence of my genes. These variants are suspect because they may alter protein function or levels of expression for the identified genes. If we just limit the dataset to changes that alter amino acid composition (non-synomous), we get 875 gene variations. Add back potential spicing issues, indels, and aberrant start and stop codon issues, we are back up to 1440 variants as genetic differences that are highly suspect for altering gene expression and function.
316 MIM variant hits in my genome!
What happens if we limit the entire 1604 to only those genes with recognized involvement in disease. We get 316 variants occurring in genes as recognized by the Mendelian-inheritance-in-Man (MIM) database for being disease-associated genes. When we restrict this set to coding issues only, we get 281 suspect variants.
I get clean bill of health when I get a physical exam, so can I disregard these 281 suspect variants?
One easy step is to filter for carrier only status. 111 variants are clearly identifiable as only autosomal recessive (AR). I would require two hits in each of the paired chromosome copies to have these be of concern. Since, no paired hits were detected, we can dismiss these genes as in need of my immediate concern. As a result, we are now only concerned about hits in genes with known autosomal dominant (AD) issues. These are the genes where only one bad hit is needed to render them pathogenic. Bottomline, 170 gene variants in my genome are worthy of further contemplation.
How frequent is frequent in my 170?
There is good rational to only be concerned about a hit in a gene with AD propensity, if it is rare in the population. The thinking is that if a variation is deleterious by itself (AD), it cannot be tolerated at a high level in the human population. Contrast this to the recessive (AR) variants (also called “alleles” when talking about frequency). My known AR pathogenic variant in the CFTR gene is in the human population at 0.0014 minor allele frequency (MAF). This high allelic frequency is tolerated in the human population because you need two hits in each gene copy in order to have a syndromic issue. Autosomal dominant alleles must have much lower frequency. If we cull the 170 for variations that occur at 0.00001 MAF or lower, we get 53 gene-codon-altering variations to be concerned about. Examining the list manually gave me 17 genes for which I hold varying degrees of concern, of which I list the top 10:
None are in the ACMG59
In a prior blog post, I described the list of genes for which can be included in a clinical report as a secondary findings. These are allowed in a report because these 59 genes have known actions that can be taken to mitigate their negative health effect. None of my genes of concern are in this group, so the immediate actionability is absent for my findings about the baggage in my genome. In fact, the genes I am listing as genes I am concerned about, but they actually do not significantly bother me that much. I am still alive and in good health. If I had pathogenic variations in these genes the negative health consequence, they should have manifest many years ago. Nevertheless, the three for which I hold highest concern are CACNA1S, LGI1, and RTN2.
The variation in CACNA1S (p.R419H) may sound like a benign, and it is a conservative change in amino acid composition, but it occurs in a highly-conserved region. It is present as an Arginine (“R”) in humans, mice, fish, flies and worms. This invariant use of R implies protein function will be compromised when the position is substituted with a histidine. The LGI1 (p.A253T) variant is also a conserved amino acid change, but it is in a less conserved region. This lack of complete conservation indicates this position might tolerate an Arginine to Threonine change. The RTN2 is complex variant. It does two significantly alarming changes. It makes a dramatic Leucine to Arginine change in the 4th exon up from the end of the protein. It also occurs immediately adjacent to splice junction acceptor site. This alteration of splicing region suggest it could lead to improper splicing in a highly conserved region of the protein and thus create a defective protein.
It is likely that all three of these genes yield a protein of messed up function. But what is not clear is the type of mess-up. Are they leading to loss-of-function (LOF) activity? Or do they lead to dominant gain-of-function (GOF)? These variations are most likely in the LOF category. Otherwise, I would almost certainly be dealing with the disease symptoms that the GOF variant’s manifest. Yet this is just supposition – a hypothesis. We don’t yet have solid evidence for what is going on.
How could we get final answer for if these variations are these pathogenic or not?
To get precision answers, we could model all of these variants in C elegans, For the CACNA1S and the RTN2, their high conservation from human to worm would allow direct modeling in the worm’s homologous position of the worm’s native gene (“Native locus”).
Our prior work with full gene humanization indicates more congruent results occur if we first swap in a human gene for the native gene locus (“Humanized Locus”) and then install variant. The use of a humanized locus allows modeling of any variant, whether it is highly conserved or not across many species. So far, in our studies all known pathogenic variants behave with deviant behavior, but only when put into humanized systems. Contrast this to insertion in native locus – some known pathogenic alleles did not create detectable deviance of behavior!
For the 3 genes to which I am concerned, all are of favorable size that the human sequence can be easily optimized and installed for expression from the worm’s native locus (“Humanized” animal). If we can observe that the human gene can rescues loss of function, we will know we are off-to-the-races and can study variant biology in a gene-humanized system. The humanized animals will be precision proxies serving as clinical avatars of the patient condition.
CACNA1S is a drugable target. The creation of a humanized system expressing CACNA1S as gene replacement of egl-19 gene would generate a platform for drug discovery. The patient variants might be responsive to calcium channel blockers, such as benzothiazepines, phenylalkylamines and 1,4-dihydropyridines. The end result, a highly-personalized medicine approaches would be achieved that finds drug treatments specific to the patient’s genetic pre-conditions.