karafecho commented 2 years ago

QotM Challenge 3: Cannabidiol (CBD) and Valproic Acid

Submitting team

Exposures Provider

SME

Paul Watkins, MD, Professor of Medicine, Professor of Toxicology, and Professor of Experimental Therapeutics at the University of North Carolina; Director of the Institute for Drug Safety Sciences, UNC Eshelman School of Pharmacy

Challenge Question

What biological mechanisms might explain the observation that cannabidiol is generally safe except in patients taking valproic acid?

Background (captured during May 6 call)

EPIDIOLEX is the first and only FDA-approved prescription cannabidiol (CBD); it is approved to treat seizures associated with Lennox-Gastaut syndrome (LGS), Dravet syndrome, or tuberous sclerosis complex (TSC) in patients 1 year of age or older

EPIDIOLEX was developed initially for children and approved in 2018 by the US FDA for the treatment of refractory seizures in children ≥2 years old with Dravet syndrome or LGS (rare diseases)

Later extended to TSC, adults

GW Pharmaceuticals developed the drug and championed it through the FDA approval process; Jazz Pharmaceuticals purchased GW and now owns rights

US FDA clinical trial data demonstrated no issues with liver safety, except in children taking valproic acid

Increased ALT, AST, bilirubin secondary to hepatotoxicity (liver cells burst open) -- Note that ALT is the primary biomarker for hepatotoxicity and is cleared by macrophages

Other clinical findings provided no additional insights

Blood levels - valproic acid, CBD, and major CBD metabolites were not higher than expected, indicating that the effect was not due to pharmacokinetics
Genomics – no clear findings
Metabolomics, proteomics not conducted

Big Picture US FDA approved EPIDIOLEX, but questioned what the findings mean, in terms of implications for other drug combinations that involve CBD

What other drugs might interact with CBD?
Might there be other cases of synergistic hepatotoxicity?
What drugs or AEs should GW Pharmaceuticals be monitoring post-marketing for purposes of pharmacovigilence?

Additional Background

Quantitative Systems Toxicology (QST) modeling

Captures mechanisms of drug toxicity at the cellular and tissue levels
See DILIsym
QST modeling was applied to independently model the toxicity profiles for CBD and valproic acid and then merge the results or look for overlap -- No clear findings

Dose-dependent liver toxicity is typically explained by three AOPs

Oxidative stress (ROS generation)
Interference with mitochondrial respiration
Impaired clearance of bile acids

Summary of findings to date

Effect appears to be synergistic, not additive
Effect does not appear to reflect a traditional drug-drug interaction (e.g., impaired clearance), but rather appears to occur at the level of the liver or hepatocyte
One hypothesis is that valproic acid depolarizes liver cells
Also of note is that CBD was recently found by Klarissa Jackson to produce long-term liver toxicity in in vitro models of liver function (human hepatic 3D spheroids)

Timeline

May 6, 2022 - First Friday of the Month Standup
Check-in/update from teams on progress

May 13, 2022 - Second Friday of the Month Standup Check-in/update from teams on progress

May 19, 2022 - Third Thursday of the Month Translator QotM mini-hackathon Review answers and refine queries with SME; wrap-up challenge

May 27, 2022 - Fourth Friday of the Month Translator QotM Challenge summary published in Gazette; next month's QotM Challenge announced

sstemann commented 2 years ago

@cbizon @bill-baumgartner @webyrd i wasn't able to tag everyone in the Assignees, so just making sure you see this ticket for Friday 5/6, 5/13 and 5/20

ShervinAbd92 commented 2 years ago

PK: ae3ed660-66cb-42f8-9852-79b2aa331b89

{
  "message": {
    "query_graph": {
      "edges": {
        "e00": {
          "subject": "n00",
          "object": "n01",
            "predicates":["biolink:related_to"]
        },
        "e01": {
          "subject": "n01",
          "object": "n02",
            "predicates":["biolink:related_to"]
        }
      },
      "nodes": {
        "n00": {
          "ids": ["PUBCHEM.COMPOUND:644019"],
          "categories": ["biolink:SmallMolecule"]
        },
        "n01": {
          "categories": ["biolink:BiologicalProcessOrActivity"]
        },
        "n02": {
          "ids": ["PUBCHEM.COMPOUND:3121"],
          "categories": ["biolink:SmallMolecule"]
        }
      }
    }
  }
}

ehinderer commented 2 years ago

Are we assuming that VA is creating some altered biochemical/metabolic state which produces an undesirable effect following CBD treatment, or are we assuming that some underlying condition (for which VA is prescribed to treat) is incompatible with CBD treatment? Or perhaps this will query reveal both? Is there evidence that it is specifically VA that is the culprit? I'm curious and knowing might help create follow-up queries.

cbizon commented 2 years ago

Do we have either a way in biolink to express drug-drug interactions or a source for the same?

karafecho commented 2 years ago

Paul will be at our meeting tomorrow and will provide some background.

In the meantime, here are a few articles (from a Google search) that may be helpful: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7443220/, https://www.frontiersin.org/articles/10.3389/fneur.2020.531939/full, and https://pubmed.ncbi.nlm.nih.gov/28782097/.

We also learned a few things about valproic acid during QotM #1: https://medlineplus.gov/druginfo/meds/a682412.html.

cbizon commented 2 years ago

This sounds like a drug-drug interaction question. There are various ways that an interaction of that type can occur, but one is that drug A affects the PK of drug B somehow so drug B doesn't get cleaned up effectively. So you might imagine a query like:

(drug A)-[metabolized by]->(gene?)-[reduced by]->(drug B)

I did a few queries around this and eventually settled on encoding metabolized by as: [biolink:substrate_of,biolink:degradation_increased_by] and reduced by with [biolink:entity_negatively_regulated_by_entity, biolink:activity_reduced_by].

The other question was whether VPA is drug A and CBD is drug B or vice versa. In the first case, CBD would be knocking down the action of a gene that metabolizes VPA, and in the second VPA would be knocking down a gene that metabolizes CBD. I didn't see any particular reason to choose one or the other, so I ran both.

VPA affects metabolism of CBD: https://arax.ncats.io/?r=bc877913-def8-41cc-916b-c1d782e791ee VPA negatively regulates several P450s (3A4, 2C9, 2C19) that have CBD as a substrate.

CBD affects metabolism of VPA: https://arax.ncats.io/?source=ARS&id=62f1a07d-0b3d-4440-bda3-88f7622899a5 CBD negatively regulates 3A5, 2B6, and PTGS1, all of which metabolize VPA.

From these results of course it's not really easy to tell if these are the main metabolizers of these entities, or how much things get downregulated etc. But if you believed everything here, you could argue that there's a sort of feedback loop where VPA+CBD prevent each other from being metabolized, and the increased exposure makes something bad happen.

colleenXu commented 2 years ago

This is a re-run of the queries from my earlier post on QotM 1:

Looking for Genes related to valproic acid and CBD. I see genes that I remember from an earlier query between valproic acid and heme (CYP genes, oxidative stress genes).

https://arax.ncats.io/?r=71f1306c-b0d7-4712-bcab-c3d8234e0555

query

``` { "message": { "query_graph": { "edges": { "e00": { "subject": "n0", "object": "n1" }, "e01": { "subject": "n2", "object": "n1" } }, "nodes": { "n0": { "ids": ["PUBCHEM.COMPOUND:644019"], "categories": ["biolink:SmallMolecule"], "name": "cannabidiol" }, "n1": { "categories": ["biolink:Gene"] }, "n2": { "ids": ["PUBCHEM.COMPOUND:3121"], "categories": ["biolink:SmallMolecule"], "name": "valproic acid" } } } } } ```

Genes related to valproic acid, CBD, and heme:

https://arax.ncats.io/?r=9e70b042-7c60-4514-b69d-8b9bbfeb9e46

query

``` { "message": { "query_graph": { "edges": { "e00": { "subject": "n0", "object": "n1" }, "e01": { "subject": "n2", "object": "n1" }, "e02": { "subject": "n3", "object": "n1" } }, "nodes": { "n0": { "ids": ["PUBCHEM.COMPOUND:644019"], "categories": ["biolink:SmallMolecule"], "name": "cannabidiol" }, "n1": { "categories": ["biolink:Gene"] }, "n2": { "ids": ["PUBCHEM.COMPOUND:3121"], "categories": ["biolink:SmallMolecule"], "name": "valproic acid" }, "n3": { "ids": ["CHEBI:30413"], "categories": ["biolink:SmallMolecule"], "name": "heme" } } } } } ```

karafecho commented 2 years ago

Thanks for re-posting the queries you ran, Colleen. I had meant to do that but forgot!

Note that I had sent your results to Paul for review, but nothing 'popped out' for him.

dkoslicki commented 2 years ago

A bunch of liver associated proteins shared by both CBD and valproic acid: https://arax.ncats.io/?r=40896

brettasmi commented 2 years ago

@cbizon re: drug-drug interactions. I can't say what we have in Translator, but I tracked this down. It's pretty new and has friendly licensing.

http://ddinter.scbdd.com/

https://academic.oup.com/nar/article/50/D1/D1200/6389535

colleenXu commented 2 years ago

@brettasmi looks like their data downloads aren't the most useful though. They don't have the info in the website (only describing interactions as Major / Minor / Moderate), and they use their own ID set (DDInter) without providing an xref set to other IDs.

ehinderer commented 2 years ago

Is a query like this feasible? Also how would I specify that I'm interested in ANY of the biological processes the proteins might participate in and not requiring proteins participate in ALL of them?

query example

dkoslicki commented 2 years ago

@ehinderer You could use ARAX DSL by using the option_group_id part. When you say ANY of the biological processes, do you mean of those three you have, or literally any node with category biolink:BiologicalProcess? I can whip together the query if you can clarify

dkoslicki commented 2 years ago

@ehinderer Here's the results for the query you asked for when looking for proteins connected to any of those three particular biological processes: https://arax.ncats.io/?r=40905 In DSL, it looks like:

add_qnode(ids=CHEMBL.COMPOUND:CHEMBL109, categories=biolink:SmallMolecule, key=n0)
add_qnode(categories=biolink:Protein, key=n1)
add_qnode(ids=CHEMBL.COMPOUND:CHEMBL190461,categories=biolink:SmallMolecule, key=n2)
add_qedge(subject=n0, object=n1, key=e0)
add_qedge(subject=n1, object=n2, key=e1)
# Now for the optional nodes and edges
add_qnode(key=n3,ids=UMLS:C1155312,categories=biolink:PhysiologicalProcess,option_group_id=g1)
add_qnode(key=n4,ids=UMLS:C1156066,categories=biolink:PhysiologicalProcess,option_group_id=g2)
add_qnode(key=n5,ids=UMLS:C0232757,categories=biolink:PhysiologicalProcess,option_group_id=g3)
add_qedge(key=e2,subject=n1,object=n3,option_group_id=g1)
add_qedge(key=e3,subject=n1,object=n4,option_group_id=g2)
add_qedge(key=e4,subject=n1,object=n5,option_group_id=g3)
# expand only with RTX-KG2 for speed purposes
expand(kp=infores:rtx-kg2,prune_threshold=9999,kp_timeout=9999)
overlay(action=compute_ngd,default_value=inf)
resultify()
filter_results(action=limit_number_of_results, max_results=100)

And can be called with TRAPI via:

{
  "message": {},
  "operations": {
    "actions": [
      "add_qnode(ids=CHEMBL.COMPOUND:CHEMBL109, categories=biolink:SmallMolecule, key=n0)",
      "add_qnode(categories=biolink:Protein, key=n1)",
      "add_qnode(ids=CHEMBL.COMPOUND:CHEMBL190461,categories=biolink:SmallMolecule, key=n2)",
      "add_qedge(subject=n0, object=n1, key=e0)",
      "add_qedge(subject=n1, object=n2, key=e1)",
      "# Now for the optional nodes and edges",
      "add_qnode(key=n3,ids=UMLS:C1155312,categories=biolink:PhysiologicalProcess,option_group_id=g1)",
      "add_qnode(key=n4,ids=UMLS:C1156066,categories=biolink:PhysiologicalProcess,option_group_id=g2)",
      "add_qnode(key=n5,ids=UMLS:C0232757,categories=biolink:PhysiologicalProcess,option_group_id=g3)",
      "add_qedge(key=e2,subject=n1,object=n3,option_group_id=g1)",
      "add_qedge(key=e3,subject=n1,object=n4,option_group_id=g2)",
      "add_qedge(key=e4,subject=n1,object=n5,option_group_id=g3)",
      "# expand only with RTX-KG2 for speed purposes",
      "expand(kp=infores:rtx-kg2,prune_threshold=9999,kp_timeout=9999)",
      "overlay(action=compute_ngd,default_value=inf)",
      "resultify()",
      "filter_results(action=limit_number_of_results, max_results=100)",
      ""
    ]
  },
  "submitter": "ARAX GUI",
  "stream_progress": true,
  "query_options": {
    "kp_timeout": "30",
    "prune_threshold": "50"
  }
}

Note: since all three of those edges are optional, that means that you also get proteins connected to CBD and VA but not connected to any of those three processes

dkoslicki commented 2 years ago

@ehinderer And here's a query for proteins that are connected to at least one of those three biological processes: https://arax.ncats.io/?r=40906 DSL:

add_qnode(ids=CHEMBL.COMPOUND:CHEMBL109, categories=biolink:SmallMolecule, key=n0)
add_qnode(categories=biolink:Protein, key=n1)
add_qnode(ids=CHEMBL.COMPOUND:CHEMBL190461,categories=biolink:SmallMolecule, key=n2)
add_qedge(subject=n0, object=n1, key=e0)
add_qedge(subject=n1, object=n2, key=e1)
# Now for the optional nodes and edges
add_qnode(key=n3,ids=[UMLS:C1155312,UMLS:C1156066,UMLS:C0232757],categories=biolink:PhysiologicalProcess)
add_qedge(key=e2,subject=n1,object=n3)
# expand only with RTX-KG2 for speed purposes
expand(kp=infores:rtx-kg2,prune_threshold=9999,kp_timeout=9999)
overlay(action=compute_ngd,default_value=inf)
resultify()
filter_results(action=limit_number_of_results, max_results=100)

ehinderer commented 2 years ago

@ehinderer You could use ARAX DSL by using the option_group_id part. When you say ANY of the biological processes, do you mean of those three you have, or literally any node with category biolink:BiologicalProcess? I can whip together the query if you can clarify

I meant any of the three listed there. And thanks for putting that together! I will use those code examples to expand upon this a bit, too.

karafecho commented 2 years ago

FYI: I added background on the challenge question, as captured during today's call with Paul.

colleenXu commented 2 years ago

The DDInter database (not in Translator) has the relationship between CBD and valproic acid: http://ddinter.scbdd.com/ddinter/interact/1002248/. They describe it as a "moderate" "synergy" relationship.

If I'm interpreting the modeling correctly, I think there are predicted inhibitor-substrate and substrate-substrate interactions for the two drugs, for CYP2C19 and CYP2C9.

I saw these genes highly scored in BTE's results for the queries I posted above, and I saw these genes in the queries posted by Bizon.

dkoslicki commented 2 years ago

@colleenXu, same for those liver enzymes (CYP2C19, CYP2C9) in the ARAX results: https://arax.ncats.io/?r=40896

cbizon commented 2 years ago

The thing about the CYP queries is that Paul said that the company explicitly checked blood levels of the compounds and they were not higher than expected, so it doesn't seem as though it's a clearance issue.

cbizon commented 2 years ago

Though this article does list several other DD interactions with CBD that are driven by CYP inhibition https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5576600/

cbizon commented 2 years ago

The DDInter database (not in Translator) has the relationship between CBD and valproic acid: http://ddinter.scbdd.com/ddinter/interact/1002248/. They describe it as a "moderate" "synergy" relationship.

I poked around in this DB a little bit. It seems like the CBD/VPA interaction is not very specific. Both of them have a lot of interactions that are related to hepatotox, but they all seem to be of the "this thing is sometimes hepatotoxic, and that thing is sometimes hepatotoxic, so let's not put them together" variety rather than a known warning that this specific combination is bad.

karafecho commented 2 years ago

Resources discussed during 05/13/2022 call:

https://www.pharmgkb.org/pathway/PA165964265/overview

https://pubmed.ncbi.nlm.nih.gov/30931853/

https://rampdb.nih.gov/

MarkDWilliams commented 2 years ago

Repo for data gaps https://github.com/NCATSTranslator/DataGaps/issues/1

karafecho commented 2 years ago

Resources from Paul (received 5/6/2022, retrieved from SPAM box 5/16/2022, sorry for the delay!):

CBD healthy volunteers pub.pdf Current Opinions in Toxicology.pdf DRAFT_VPA_CBD_Poster_AASLD_FDA_DILI_Conference_2021_updated_08April2021.pdf 2020 PharmSci 360_poster_K Yang_QST Modeling of DILI and Adaptation.pdf Dongd Valproate manuscript.pdf

karafecho commented 2 years ago

I ran a one-hop that looked for chemical entities related to MRP2/ABCC2 (NCBIGene:1244), following up on the findings reported in Dongd Valproate manuscript.pdf and DRAFT_VPA_CBD_Poster_AASLD_FDA_DILI_Conference_2021_updated_08April2021.pdf.

{
    "message": {
        "query_graph": {
            "nodes": {
                "n0": {
                    "ids": ["NCBIGene:1244"],
                    "categories": [
                        "biolink:Gene"
                    ],
                    "name": "MRP2, ABCC2"
                },
                "n1": {
                    "categories": [
                        "biolink:ChemicalEntity"
                    ]
                }
            },
            "edges": {
                "e0": {
                    "subject": "n0",
                    "object": "n1",
                    "predicates": ["biolink:related_to"]
                }
            }
        }
    }
}

PK = 440183ff-7a92-4d86-b967-d9971f28592f

Valproic acid/valproate is returned by ARAX, BTE, and Explanatory. ARAGORN further identifies relationships with symptoms of liver injury/disease. CAM shows that NCBIGene:1244 - biolink:affects_activity_of - CD320.

Cannabidiol/CBD is not identified by any Translator component.

karafecho commented 2 years ago

Summary of progress to date

Dose-dependent liver toxicity is typically explained by three AOPs
1. Oxidative stress (ROS generation)
2. Interference with mitochondrial function
3. Impaired clearance of bile acids
CBD + valproic acid effect appears to be synergistic, not additive
Effect does not appear to reflect a traditional drug-drug interaction (e.g., impaired clearance), but rather appears to occur at the level of the liver or hepatocyte
Effect may involve metabolites, but Translator does not have a good source of metabolic pathway data
- NCATS' RampDB?
- Also, GW Pharmaceuticals measured blood levels of valproic acid, CBD, and major CBD metabolites and found that they were not higher than expected, suggesting that the effect was not PK-related
Effect does not appear to reflect interference with mitochondrial respiration
- Paul's group demonstrated this
One hypothesis is that valproic acid depolarizes liver cells
- Translator replicated findings by Paul's group that valproic acid-induced hepatotoxicity may involve impaired Mrp2 trafficking; however, Translator did not surface evidence relating CBD to Mrp2 trafficking
- MRP2 mappings?
Another hypothesis is that CBD produces long-term hepatotoxicity, as was recently demonstrated by Paul's colleague, Klarissa Jackson, using an in vitro model of liver function (human hepatic 3D spheroids)

Avenues to explore:

'Hepatic depolarization' or 'hepatocyte depolarization'? "GO:0051899": "membrane depolarization"? "GO:0051882": "mitochondria depolarization"? "GO:1990281": "efflux transmembrane transporter complex", "efflux pump", "efflux pump complex"?"GO:0007006": "mitochondrial membrane organization and biogenesis"? Long-term hepatotoxicity?

andrewsu commented 2 years ago

I'm not sure I see any incredibly interesting leads here, but I tried a query looking at relationships between CBD and the membrane-related GO terms that @karafecho outlined, first mediated by Genes and ChemicalEntitys:

CBD -- [Genes,ChemicalEntity] -- Membrane Depolarization GO Terms https://arax.ncats.io/?r=aef665d6-2318-47d9-98c2-27fdb75d764c

query

``` { "message": { "query_graph": { "edges": { "e00": { "subject": "n0", "object": "n1" }, "e01": { "subject": "n2", "object": "n1" } }, "nodes": { "n0": { "ids": [ "PUBCHEM.COMPOUND:644019" ], "categories": [ "biolink:SmallMolecule" ], "name": "cannabidiol" }, "n1": { "categories": [ "biolink:Gene", "biolink:ChemicalEntity" ] }, "n2": { "ids": [ "GO:0051899", "GO:0051882", "GO:1990281", "GO:0007006" ], "categories": [ "biolink:BiologicalProcessOrActivity" ] } } } } } ```

Since that completed in a reasonable amount of time, I broadened that to NamedThing in the intermediate node. BTE and ARAX retrieved a few more results (uninteresting for BTE), and Aragorn had fewer results

CBD -- [NamedThing] -- Membrane Depolarization GO Terms https://arax.ncats.io/?r=7c698e39-51a0-4200-8fe8-780203c62385

query

``` { "message": { "query_graph": { "edges": { "e00": { "subject": "n0", "object": "n1" }, "e01": { "subject": "n2", "object": "n1" } }, "nodes": { "n0": { "ids": [ "PUBCHEM.COMPOUND:644019" ], "categories": [ "biolink:SmallMolecule" ], "name": "cannabidiol" }, "n1": { "categories": [ "biolink:NamedThing" ] }, "n2": { "ids": [ "GO:0051899", "GO:0051882", "GO:1990281", "GO:0007006" ], "categories": [ "biolink:BiologicalProcessOrActivity" ] } } } } } ```

karafecho commented 2 years ago

Synergism between valproic acid and acetaminophen

https://pubmed.ncbi.nlm.nih.gov/31029706/

Finding goes against GW Pharmaceuticals' claims that CBD was the only compound to demonstrate synergistic hepatotoxicity with valproic acid

Several mechanisms hypothesized to contribute to the synergistic effect:

karafecho commented 2 years ago

Valproic acid, POLG, and Alpers-Huttenlocher syndrome

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3841971/

Valproic acid-induced hepatotoxicity

Hyperammonemia with minimal or no evidence of hepatic injury
Acute hepatocellular hepatoxicity with jaundice
Reye-like syndrome

Alpers - Huttenlocher syndrome

Rare mutations in POLG, which codes for the mitochondrial DNA polymerase γ (polγ), cause the Alpers-Huttenlocher syndrome (AHS). AHS is a neurometabolic disorder associated with an increased risk of developing fatal VPA-hepatotoxicity.

Bassa1 commented 2 years ago

Pharmacokinetic changes due to drug-drug interaction could lead to an increase in the adverse effect or cause therapeutical failure. I found this article that gives a detailed explanation of the drug interaction effect on pharmacokinetics. Effects of cannabidiol on chronic pain reliever drug metabolism.

Glucuronidation is also involved in CBD metabolism being CBD an inhibitor of UGT1A9 and UGT2B7 enzymes
If cannabinoid concentrations are high, perhaps CBD may impair valproic acid glucuronidation, and thus, valproic acid clearance may be reduced Article

karafecho commented 2 years ago

Thanks for your thoughts, @Bassa1. I think we ruled out traditional drug-drug interactions as a potential mechanism that might explain the synergistic hepatotoxicity. For instance, in primary clinical trials, blood levels of valproic acid, CBD, and major CBD metabolites were not higher than expected. Any other ideas?

karafecho commented 1 year ago

Please see summary in Translator Gazette - May 2022.

karafecho commented 8 months ago

Closing with comment in #233 ...

NCATSTranslator / testing

Question of the Month #3: Cannabidiol and Valproic Acid #196

QotM Challenge 3: Cannabidiol (CBD) and Valproic Acid

Submitting team

SME

Challenge Question

Background (captured during May 6 call)

Additional Background

Timeline

Summary of progress to date

Synergism between valproic acid and acetaminophen

Valproic acid, POLG, and Alpers-Huttenlocher syndrome