Application of Hypersensitivity Assays to Discovery of Therapeutics

Chemical hypersensitivity is a common method to probe gene dysfunction and can be deployed in drug screens to find new therapeutics. In this blog post, we will focus on models of Inborn Errors of Metabolism (IEM) and describe how these genetic conditions can lead to hypersensitivity to a metabolite. By using humanization techniques, we take advantage of the ancient biology between humans and other organisms to create stand-ins – patient avatars – for drug screening studies. These genetically engineered model systems enable fast and affordable phenotypic screens in whole organism format to enable researchers to find molecules that alleviate the metabolic stress occurring from an IEM deficiency. Ultimately, this report showcases how chemical hypersensitivity is generalizable to drug discovery tool applicable to many genetic disorders.

Clinical variants disrupting metabolic gene can lead to build up of toxic metabolites (Figure 1).  This phenomenon can be used to create a functional assay where the model system has hypersensitivity to metabolites upstream of the gene’s function in a metabolic pathway.

Figure 1.  Hypothetical metabolic pathway.  When ENZ 2 gene is defective, metabolite 2 can build up to toxic level that lead to paralysis and death.

Chemical Hypersensitivity Due to Metabolic Block

In the above example, the second enzyme of the metabolic pathway, “ENZ 2,” is the cause of a genetic disease  that inhibits the enzymatic conversion of metabolite 2 into metabolite 3.  For patients with this condition, exposure to metabolite 1 leads to toxic metabolite 2 build up and activation of reactive oxygen species, ultimately leading to paralysis and death.  Because of this metabolic pathway blockade, patients can experience hypersensitivity with exposure to metabolite 1.

Model Systems in the Fluidics Paradigm

A variety of model systems are possible for use in testing for metabolite hypersensitivity.  Starting with the simple model organisms, a human gene associated with disease can be installed as a gene replacement. In the case of metabolic disorders, the high degree of sequence conservation in these ancient genes often enable the human gene to rescue the function of the removed ortholog (the animal’s version of the disease gene). When we add in iPSC and then install clinical variants in the human gene locus, we get three types of model systems (C. elegans nematode, zebrafish and iPSC to model clinical variants (Figure 2). These models are advantageous for drug discovery because they fit the fluidics paradigm – The zebrafish and nematode can live their entire lifecycle in liquid and differentiated iPSC can be hooked up in biocircuits with microfluidics. By being in fluid, assessment of oral bioavailability is simply done by adding drugs to the liquid growth media. In this liquid environment, The first step is to create the wt-Avatars. In the nematode, the evolutionary distance often renders a gene to gene comparison with low homology (sequence identity under 75%). As a result only a portion of the pathogenic alleles can be modeled as amino acid substitution in the nematode’s native locus.  For work around we use Whole Gene Humanization – CRISPR gene editing is used to remove the coding sequence at the native locus and replace it with the human gene coding sequence.  When the human sequence restores normal function, we know we are looking at a high degree of conservation of biology.

In zebrafish, the orthologous gene is often at a sequence identity that is equal to or greater than 75% of the human sequence. This often renders the zebrafish’s gene sufficient for modeling the patient condition as a single amino acid substitution, yet occasionally whole-gene humanization will need to be deployed. For iPSC, we source from a healthy patient (reference line – “wt Avatar”) and make the single amino acid variation to model the patient variant condition (var-Avatar).   Bottomline, modeling a patient condition requires the creation of a wild-type humanized animal or the use of an unmodified line (wt-Avatar) and then the insertion of the missense variant that models the genetic variant of the patient (var-Avatar). In regards to IEM deficiencies, this system enables detection of phenotypic abnormalities that are often accentuated by metabolite exposure.

Figure 2. Three types of model systems are zebrafish and C. elegans nematode animal models (most relevant when used in whole gene humanized formats) and the use of iPSCs. Humanizing mutations are made to recreate the same variation seen in the patient (var-Avatar) which are then compared to the wild type control (wt-Avatar).

Detection of Hypersensitivity Phenotypes

To provide a phenotypic screen that is linked to a mechanism of action, the var-Avatar lines can be examined for their hypersensitivity to specific environmental stressors. Applied to a metabolic gene, the environmental stressor is exposure to an upstream metabolite. Using the ENZ 2 example of Figure 1, increasing dosage of metabolite 1 can be monitored for its effect on paralysis in the var-Avatar model system. To measure activity in the nematode, the 96 well format of the wMicrotracker apparatus is used to track loss of locomotion. As the animal becomes more paralyzed light beam disruption rate decreases. By exposing a nematode var-Avatar to different concentrations of metabolite, an LD50 curve can be generated (Figure 3).

Figure 3. wMicrotracker is used to generate LD50 curves for var-Avatar and wt-Avatar upon exposure to different concentrations of metabolite. As metabolite concentration increases, survival of animals decreases. The var-Avatar has an earlier drop in survival when compared to wt-Avatar. An intermediate concentration of metabolite (40mM) can be used to discriminate hypersensitivity in the var-Avatar animals.

In the var-Avatar, the loss of enzyme function with ENZ 2 pathogenic variants creates pathway blockage and allows build up of metabolite 2. When the metabolite 2 reaches toxic levels and the cell becomes oxidatively stressed, which leads to cell apoptosis and eventually animal paralysis and death. In this example, 40 mM of metabolite 1 is enough to cause paralysis and lethality in the var-Avatar of ENZ 2 deficient animals, but it is not enough to create paralysis in wt-Avatar control animals. As a result, a drug screen can be developed that uses this 40mM concentration as a cut-off tool for finding drugs that can alleviate the metabolic block and enable survival of the ENZ 2 var-Avatar (Figure 4).

Figure 4. Rescue of hypersensitivity screen applied to drug repurposing library.  A “hit” is obtained when the drug rescues survival in var-Avatar in 40 mM metabolite but does not influence survival at the wt-Avatar at its LD50 measured metabolite concentration.

Drug Screen

A drug screen can be performed on repurposing libraries to find hits for use in clinical trials. In the above example, the 40 mM of metabolite 1 is used to detect hits as a var-Avatar animal that can survive when exposed to a drug from the repurposing library. Additionally, the hit is considered valid when the drug has minimal to no effect on the wt-Avatar control. For metabolic disorders screening,the first step (Phase 1) is a Stress Test to determine if upstream metabolites can lead to a hypersensitivity.  In the next step (Phase 2), prior to the library screening, testing for Metabolic Modulation is performed by observing if changes in hypersensitivity can be achieved with downstream metabolites, cofactors, oxidants and antioxidants (Figure 5).  In the final step (Phase 3), A Library Screen is performed to find modulatory effects from FDA approved drugs.

Figure 5.  The wt-Avatar and var-Avatar animals are compared for their response to stress test compounds (upstream metabolites) to achieve optimal separation of wt-Avatar vs var-Avatar (dotted lines). Then a set of metabolic modulators (downstream metabolites, oxidants and antioxidants) are tested for their capacity to restore normal sensitivity in var-Avatar and have minor change in wt-Avatar sensitivity (dashed lines).

Phase 1 – Stress Testing to Detect Compound Hypersensitivity

In the first phase of a 3 phase process, an assessment is made for the detection of metabolite sensitivities to metabolites and ROS mediators. In this pilot screen, a small selection of chemicals are used to determine if metabolite sensitivity can be detected. First upstream metabolites are exposed to the model system. These will often lead to their build up to toxic levels which are achieved sooner in the var-Avatar line vs the wt-Avatar line.

Scope Steps: Hypersensitivity Screen (upstream metabolites)

  1. Obtain wt-Avatar and var-Avatar: Generate humanized animals for modeling a metabolic gene deficiency – created as a prerequisite project.
  2. Obtain Stress-Test Compounds: Select upstream metabolites (1 to 10 molecules)
  3. Solubilize Stress-Test Compound: Resuspend the metabolites in 100% DMSO at 100mM concentration to mimic the sample conditions of most drug screening libraries.
  4. Measure Activity: Determine suppression as retention of locomotion using wmicrotracker plate reader instrumentation.
  5. Generate LD50 curves: Expose animals at concentrations 0.01, 0.1, 1, 10, 100, and 1000 uM (1% DMSO)
  6. Select Optimal Separation: Examine LD50 curves for conditions creating a high degree of separation between wt-Avatar and var-Avatar.
  7. Repeat With Combinations (optional): If necessary, repeat with combinations of chemicals to attempt to create at least 4x separation in chemical sensitivity between wt-Avatar and var-Avatar.

Phase 2 – Metabolic Modulation Effects

Once metabolic hypersensitivity is established, a second set of molecules are examined for their ability to alleviate metabolic defect and restore normal activity. Downstream metabolites and cofactors can be added to the system to determine if they can restore normal metabolite sensitivity.  Additionally, In many metabolic disorders the buildup of metabolite intermediates lead to disruption of normal levels of Reactive Oxygen Species (ROS) (doi: 10.1155/2018/1246069). As a result, two other groups of molecules can be screened for their effect on metabolite hypersensitivity. Various antioxidants (resveratrol, glutathione, metformin) can be tested for their ability to decrease sensitivity to metabolite build up. On the other hand, some metabolic intermediates may attenuate ROS which at low levels are necessary for homeostasis signaling. Various oxidants (paraquat, juglone and AAPH (2,2′-azobis-2-methyl-propanimidamide)) may lead to restoration of normal low levels of ROS and alleviate metabolic hypersensitivity.

Scope Steps: Metabolic Modulation (downstream metabolites, cofactors, oxidants, antioxidants)

  1. Obtain metabolic-modulator Compounds: Select downstream metabolites, and oxidants and antioxidants (5 to 10 chemicals)
  2. Solubilize metabolic-modulator Compound: Resuspend the compounds in 100% DMSO at 100mM concentration to mimic the sample conditions of most drug screening libraries.
  3. Measure Activity: Determine suppression as retention of locomotion using wmicrotracker plate reader instrumentation.
  4. Generate EC50 curves: Using a upstream metabolite at a concentration that leads to a strong deficiency (ie. death), add metabolic-modulator at concentrations 0.01, 0.1, 1, 10, 100, and 1000 uM (1% DMSO)
  5. Identify Modulator Candidates: Examine the EC50 curves for the modulators effect on wt-Avatar and var-Avatar.
  6. Repeat With Combinations (optional): If necessary, repeat with combinations of chemicals to attempt to decrease the separation between wt-Avatar and var-Avatar.

Phase 3: Library Screening Approach

For speed to clinical trials, a good choice is to use repurposing libraries. These are the FDA-approved drugs that are well vetted for ADMET issues, which makes them a safe choice for therapeutic consideration. We will apply the assay developed in phase 1 to scale across a commercially-sourced drug screening library. A variety of sources for compound libraries utilizing FDA-approved drugs are available (TargetMol, ApexBio, Chembridge, Microsource, Prestwick, Seleckchem, etc… – typically about 2000-3000 mlcs). As an example, the metabolite hypersensitivity assay is applied to the var-Avatar line (patient variant animal) at 10uM concentrations of compounds from the Apexbio FDA-approved library of 2726 compounds. The top hits (up to 200) are counter screened on wt-Avatar animals and the results are scored for minimal alteration of wild-type metabolite sensitivity. A rank is developed that balances strong response in the var-Avatar line against a strong response in the wt-Avatar line. The top 20 hits are examined at a range of doses to find the EC50 values in suppressing metabolite hypersensitivity. Data package is prepared and delivered to the client.

Scope Steps: Examine Rescue of metabolite Sensitivity in var-Avatar line with 2000+ Chemical Library

  1. Obtain library: Multiple repurposing libraries are available – for example, ApexBio library of 2726 FDA-approved chemicals.
  2. Drug Exposure: Using Stress-Test chemical concentrations optimized in phase 1, test The var-Avatar line against the library of 2726 molecules at 10uM concentration for their ability to suppress metabolite hypersensitivity.
  3. Measure Activity: Determine suppression as retention of locomotion using wmicrotracker plate reader instrumentation.
  4. Perform Primary Screen: Score compounds as positive or negative for ability to suppress metabolite hypersensitivity by retaining locomotion.
  5. Identify Preliminary Hits: Select up to 200 compounds (≤ mlcs) from the positive category for use in repeat screen on wt-Avatar animals.
  6. Perform Counter Screen: Using metabolite concentration optimized in phase 1, test the ≤200 mlcs at 10uM concentration for their ability to NOT suppress metabolite sensitivity in wt-Avatar animals.
  7. Rank Hits: Select up to 20 compounds (≤20 mlcs) that are positive in the var-Avatar for suppression of metabolite hypersensitivity and negative for suppression of metabolite sensitivity in the wt-Avatar.
  8. Characterize Hits: Perform EC50 assays on var-Avatar line for the 20 compounds.
  9. Report Out: Send a report to the client.

RWE – Real World Evidence Applications

A real world example occurs in the propionate metabolic pathway.  Defects in the ECHS1 gene lead to altered propionate metabolism shunting propionate away from Acetyl-CoA production and instead only allow Succinyl-CoA production (Figure 6). 

Figure 6. Defects in ECHS1 block propionate metabolism via acetyl-CoA pathway. Instead only the Succinyl-CoA pathway remains.

Potentially hampering succinyl-CoA production is a vitamin B12 dependency of the last enzyme in the Succinyl-CoA production pathway. When both B12 is limiting and the Acetyl-CoA production is blocked by a ECHS1 defect, hypersensitivity to propionate occurs. Since the propionate pathway is highly conserved between the nematode and humans, creating a deficiency in the nematode equivalent of ECHS1 (the ech-6 gene) can create a propionate hypersensitivity (Figure 7).

Figure 7. Differential effects of propionate exposure in C. elegans nematode models. Propionate exposure to the nematode exhibits a sensitivity in wild type (N2) control that has LD50 near 90 mM.  When RNAi mediated knockdown of ech-6 gene is performed, a propionate hypersensitivity ensues and LD50 drops to near 10 mM propionate.  The result is a near 10x difference in propionate sensitivity when the ech-6 locus is blocked from expression of the nematode’s version of ECHS1.  Addition of B12 vitamin partially rescues the propionate sensitivity.  (data adapted from Watson et al. Elife. 2016 Jul 6;5:e17670)

Loss of function of ech-6 creates propionate hypersensitivity

RNAi was used to knock down expression from the ech-6 gene (Watson et al. Elife. 2016 Jul 6;5:e17670). The result was hypersensitivity to propionate – a 9 fold increase in LD50 with exposure to the metabolite. When B12 was added to the system, partial rescue of hypersensitivity occurred. From this data where the drug screen was performed at 30 mM propionate, the level of rescue of survival was 4 fold higher with B12 exposure. 40-50 mM propionate appears to be the ideal range for detecting chemicals capable of rescue of propionate hypersensitivity in ech-6 null animals (KO). Inserting the human ECHS1 as gene replacement of ech-6 and observing restoration of normal propionate sensitivity will establish that the animal model (wt-Avatar) can be used to create a background system for modeling human disease. Next, installing variants in the humanized strain will create a model system for specifically exploring a patient’s genetic condition (var-Avatar) (Figure 8). When good separation between var-Avatar and wt-Avatar is observed, a screen for rescue of propionate hypersensitivity can be performed and B12 can be used as positive control.

Figure 8. Propionate sensitivity for four types of animal models. The KO null does not express ech-6 gene and is expected to be highly sensitive to propionate.  The var-Avatar is a whole-gene humanized wt-Avatar containing a patient coding-sequence variant.  The wt-Avatar is a humanized animal created as a whole-gene replacement of the ech-6 coding sequence. The wild-type N2 is the unmodified animal model commonly used as a control animal for comparison to gene-modified animals.

B3GAT3 Drug Target for Modulating Propionate Hypersensitivity

Reassuringly, other genes involved in propionate metabolism are known to have hypersensitivity when they are made defective. For instance, the PCCA gene, in combination with PCCB, is a heteromeric enzyme responsible for converting propionyl-CoA to D-methylmalonyl-CoA, which is an intermediate step of the metabolism of propionate to succinyl-CoA. When loss of function is created in the nematode ortholog pcca-1, propionate hypersensitivity occurs and these animals do not survive at above 50 mM propionate (Watson et al. Elife. 2016 Jul 6;5:e17670).  This same group used the propionate sensitivity assay to determine that  loss-of-function mutations in glct-3 (an ortholog of B3GAT3) create a propionate resistant animal (Na et al., PLoS Genet. 2020 Aug 28;16(8):e1008984). As a result, it appears inhibition of B3GAT3 may be useful in offsetting the hypersensitivity of defects within the propionate pathway.  In silico approaches can be used to dock molecules to the B3GAT3 structure (1FGG, 1KWS, 3CU0) (Figure 9).  Hits found can then be validated by testing in a humanized animal model.

Figure 9. Molecular dynamics screens large libraries to find top hits for validation in an animal model.  

Zebrafish Model for Study of RPE65 Defects

Switching to zebrafish models, vitamin A metabolite toxicity can be used to probe defects in the RPE65 gene. First, morpholinos or crispants can be used to dramatically reduce expression of this gene in the fish and lead to loss of function phenotypes.  Next, the addition of an mRNA to the injection mix can be used to rescue the loss of function and restore normal metabolic activity.  With regards to the loss of function activity, morpholinos in zebrafish work by shutting down translation of a gene transcript.  In the alternative approach, crispants work by disrupting the gene’s coding sequence in a large proportion of cells (>90%). Often the effect with either a pproach is a 10x or more decrease in gene expression which effectively models a loss-of-function variant. It is expected loss of function of the zebrafish isoforms for RPE65 will result in metabolite hypersensitivity.

In humans, loss of the RPE65 retinoid isomerohydrolase, an enzyme of the visual cycle involved in catalytic recycling of retinyl palmitate to 11-cys-retinol, will render the REP65 deficient animals hypersensitive to retinyl palmitate and its precursors. The all-trans-retinal (atRAL) is a precursor with established toxicity in the visual system (Chen et al. J Biol Chem. 2012 Feb 10; 287(7): 5059–5069; Gao et al., J Biol Chem. 2018 Sep 14; 293(37): 14507–14519). The RPE65 enzyme is critical in the last step of the visual cycle by creating the 11-cis-retionol needed by opsins to convert light into a biochemical signal. Deficiencies in this enzyme are likely to render the animals prone to light-induced blindness. In zebrafish, there are three genes (rpe65a, rpe65b, and rpe65c) that provide support for the RPE65 role in humans (Ward et al., Front Cell Dev Biol. 2018; 6: 37). A crispant CRISPR-based somatic knockout strategy can be employed where three sgRNAs each targeting the three isoforms are used to create removal of retinoid isomerohydrolase from nearly all tissues (~95%) in a zebrafish embryo. Exposing the resulting larvae to bright light and all-trans-retinal (atRAL) is likely to create a hypersensitivity that manifests as a high rate of blindness.

To rescue the blindness, an mRNA encoding human retinoid isomerohydrolase (hRPE65) is added to the crispant mix. This allows the visual cycle to remain active and toxic levels of atRAL are avoided. To determine if a variant in hRPE65 is pathogenic, any of the 177 missense variants identified in the clinical population database of ClinVar can be made as an mRNA and then be examined for their pathogenic potential (lack of rescue). The mRNA effectively becomes a var-Avatar model for testing a variant’s hypersensitivity to atRAL.  For the variants that exhibit accelerated blindness, small molecules can be explored for their ability to restore normal sensitivity to atRAL exposure.

An iPSC Model for Study of NOTCH3 Deficiencies

As an evolving model system, induced pluripotent stem cells (iPSCs) can be derivatized into tissues that are populated into microphysiological systems (MPS). Some of the MPS use iPSC-derived tissue that models the vascular system. An important disease to model in vascular formats is cerebrovascular disorders. CADASIL is one of the vascular diseases with an established genetic cause – variations in the extracellular domain of NOTCH3 gene result in small blood vesicle defects in the brain. Although NOTCH3 variations causative for CADASIL are rare, the disease is considered an ideal model for cerebrovascular disease, which affects 25% of all stroke patients and 45% of all dementias. Further we know there is a therapeutic antibody binding site on the extracellular side of the NOTCH3 protein that modulates the disease by promoting extracellular cleavage and turnover of the NOTCH3 gene. Loss or gain of cystines are the most common pathogenic variants in NOTCH3. A reference iPSCs can be modified by CRISPR to contain the patient variant (var-Avatar) and be compared to the unmodified cells (wt-Avatar). In CADASIL patients, oxidative stress occurs (Neves et  al., JCI Insight. 2019 Dec 5; 4(23): e131344.) Specifically, CADASIL patients exhibit Nox5-induced oxidative stress as measured by Lucigenin-enhanced chemiluminescence assay of NADPH-dependent ROS production. So, although the NOTCH3 gene is not considered a metabolic gene, disruptions in its function behave like many metabolic deficiencies and lead to ROS imbalance.  For a hypersensitivity screen, it is likely oxidative stressors such as paraquat will exhibit hypersensitivity in iPSC-derived MPS models of CADASIL patients.


Stressor hypersensitivity can be used to detect favorable drug effects.  Applied to Inborn Errors of Metabolism, the use of an upstream metabolite can be a stressor that leads to hypersensitivity to the metabolite. Patients with defects in an enzyme of a metabolic pathway experience unhealthy build up of a metabolite intermediates. We can take advantage of the ancient biology between humans and other organisms for the genes involved in metabolism where we create the patient’s genetic defect in the model system, either nematode or zebrafish (or iPSC). This humanized animal (or modified iPSC) then becomes the patient avatar for use in drug screening studies. For metabolic deficiencies, the animal can be used for phenotypic screens to find molecules that alleviate the metabolic stress occurring from the deficiency.  Ultimately the approach is likely to be widely generalizable to many genetic disorders.

Leave a Reply

Your email address will not be published. Required fields are marked *