toxicity using Caenorhabditis elegans and the RTgill cell line as - - PowerPoint PPT Presentation

Molecular toxicology approaches to address the adverse outcome pathway for narcosis toxicity using Caenorhabditis elegans and the RTgill cell line as model systems

Erica K. Brockmeier1, Philipp Antczak1, Geoff Hodges2, Emma Butler2, Cecilie Rendal2, Steve Gutsell2, and Francesco Falciani1

1Institute of Integrative Biology, University of Liverpool, UK 2Safety & Environmental Assurance Centre, Unilever, UK

WHAT DO WE KNOW ABOUT NARCOSIS?

Kow-dependent and reversible

X

• Loss of reaction to stimuli
• Loss of equilibrium
• Decreased respiration rate
• Decreased metabolism
• Lethality

70,000 industrial chemicals

Vaes et al. 1998

Class 1 Class 2

WHAT DO WE KNOW ABOUT NARCOSIS?

Antczak et al., 2015

Kow-dependent transcriptional switch ~log Kow 1.8

Calcium ATPase pump inhibitor

Correlation of significance with both Kow and thapsigargin exposures Narcotics: constant ILC50 for chemicals of log Kow > 2 (Escher 2002) X

Internal Lethal Conc.

PROJECT OBJECTIVE AND SPECIFIC AIMS

Specific aims 1. Compare species and system (in vivo versus in vitro) sensitivities between C. elegans and RTgill cells after exposure to a panel of narcotic chemicals 2. Determine if class 1 (nonpolar) and class 2 (polar) narcotic compounds differ in their molecular-level responses in C. elegans exposures 3. Identify potential membrane protein targets and impactesd pathways of narcotic compounds using C. elegans transcriptional data

OBJECTIVE:

Use C. elegans and the RTgill cell line as model systems for evaluating the underlying mechanisms of narcosis and determine if narcosis can be predicted using biological response data

TOOLS FOR EVALUATING THE MECHANISMS OF NARCOSIS

Advantages of C. elegans: – Gene knock-outs – RNA interference libraries – Transgenic strains – Motility and behavioral assays – Cell fate map – Neural networks High-throughput EC50 screening with modified ‘touch test’

in vivo model: C. elegans in vitro model: Rtgill cell line

High-throughput EC50 screening with CellTox green (Promega) Advantages of RTgill cell line: – High-throughput exposures and physiological assays (respiration, metabolism) – Single cell imaging –

• O. mykiss genome available

– Connections to aquatic toxicity data and in vivo exposures

• C. ELEGANS AND RTGILL EXPOSURES

Panel of 30 narcotic chemicals, 15 from class 1 (non-polar) and 15 from class 2 (polar)

Exposures in liquid K- media (NaCl/KCl) 24h exposures Endpoint: Lethality Exposures in L-15 media (5% FBS, proliferative cells) 24h exposures Endpoint: Cytotoxicity

Nominal RTgill Free RTgill

Significant regression with log Kow in both systems, and use of the TopLine model is able to reduce regression variation based

• n nominal concentrations

Nominal C. elegans Free C. elegans

Adj R2 = 0.585 P-value = 3.23E-05 Adj R2 = 0.73 P-value = 3.93E-07 Adj R2 = 0.27 P-value = 0.0019 Adj R2 = 0.821 P-value = 3.49E-12

QC analysis, quantile normalization, remove genes below background

1/10 LC50 exposure Day 1 chemicals and controls (3) 1/10 LC50 exposure Day 2 chemicals and controls (3) 1/10 LC50 exposure Day 3 chemicals and controls (3) 1/10 LC50 exposure Day 4 chemicals and controls (3)

Panel of 30 narcotics: 15 class 1 and 15 class 2

Error model D1 Error model D2 Error model D3 Error model D4

Worms transferred to exposure chamber 12h before exposure

• C. ELEGANS MICROARRAY EXPERIMENTS

Control variance distribution

Gene average

Expression of Gene(i)

Single microarray on pooled (N=3) biological replicates Full replication for control samples

Individual chemical expression profiles

Differential gene expression and statistical modelling pipeline

PCA of KEGG pathway scores, separated by class 1 (non-polar) and class 2 (polar)

KEGG pathway NES Score Carbon metabolism 3.01063 Pyruvate metabolism 2.96546 Arginine and proline metabolism 2.763 Glyoxylate and dicarboxylate metabolism 2.74236 Fatty acid degradation 2.72804 Citrate cycle (TCA cycle) 2.50691 Valine, leucine and isoleucine degradation 2.33069 Fatty acid metabolism 2.27663 Butanoate metabolism 2.19156 Propanoate metabolism 2.08354 Biosynthesis of amino acids 1.98238 Lysine degradation 1.92407 Biosynthesis of unsaturated fatty acids 1.86369 Peroxisome 1.81518 Tryptophan metabolism 1.72816 Glycolysis / Gluconeogenesis 1.54885 2-Oxocarboxylic acid metabolism 1.4903 Alanine, aspartate and glutamate metabolism 1.29005 Cysteine and methionine metabolism 1.1557 Glycine, serine and threonine metabolism 1.09278 Neuroactive ligand-receptor interaction

• 1.12733

Metabolism of xenobiotics by cytochrome P450

• 1.45974

What KEGG pathways contribute the most to the principal component loadings? Energy and amino acid-related processes contribute to the separation of class 1 and class 2 narcotics.

EXPLORATORY DATA ANALYSIS: ARE THERE GENE SETS THAT CAN DISTINGUISH POLAR AND NON-POLAR COMPOUNDS?

Class 1 versus class 2:

479 significant genes (FDR 10%) 134 significant genes (FDR 10%)

DIFFERENTIAL GENE EXPRESSION ANALYSIS:

Significance Analysis of Microarrays: A small numbers of significant genes but are related to energy and metabolism Low versus high log Kow:

68 significant genes (FDR 10%)

Time course SAM with Kow:

Transcriptional ‘switch’ ~ Log Kow 2.10 2.10

Gene set t enri richment analy lysis is

bioinfo.cnio.es/

Are e there se sets of

• f genes whic

ich ar are e mor

• re or
• r

les ess s pr present than expected by y ch chance?

GSEA ANALYSIS: CLASS 1 VS CLASS 2

Significant KEGG pathways (1% FDR) comparing class 1 and class 2 narcotics (avg. enrichment score)

GSEA enrichment score Overlaps with exploratory PCA KEGG analysis

PREDICTION ANALYSIS: CLASS 1 VS CLASS 2

Genetic Algorithm to Optimize problems related to variable selection: Able to distinguish class 1 from class 2 narcotics using individual genes

Class 1 (non-polar) Class 2 (polar)

Forward selection model 80% accuracy for NP and 92% for P Model uses eight individual genes: C04E12.2-Unknown membrane protein F14F7.3-CYP13A T10B9.2-CYP13A5 T16G1.6-unknown growth-related protein T16G1.7-Checkpoint protein T24B8.3-hypothetical protein B0464.8-embryo development F56A4.5-Checkpoint protein We can predict different classes of narcotics using gene expression patterns, and will next expand this model to include non-narcotics for screening procedures.

Chemical correctly sorted Chemical incorrectly sorted

IDENTIFICATION OF POTENTIAL MEMBRANE PROTEIN TARGETS

Correlations between log Kow and gene expression Correlations between membrane gene expression and all other gene expression expression What membrane genes are positively or negatively enriched in relation to Log Kow and narcosis-induced gene expression?

• Determine neighborhoods of membrane genes
• Correlate at 0% FDR and correlation >= 0.7
• GSEA of genes which positively correlate with

membrane genes

• Enrichment scores for each membrane genes

Narcosis = membrane perturbation But how?

Antczak 2015: Role of CERCA pumps?

Individual chemical expression profiles

IDENTIFICATION OF POTENTIAL MEMBRANE TARGETS

Positively enriched with Kow:

159 significant genes (0% FDR) 56 significant genes (0% FDR)

Negatively enriched with Kow:

What t fu funct ctions do

• th

these genes have? What t fu funct ctions do

• th

these genes have?

While we see a large number of potential membrane targets, for validation we will focus on genes with functions relevant for narcosis-related endpoints.

NARCOTIC EXPOSURES INCREASE TIME TO PARALYSIS IN CO- EXPOSURES WITH LEVAMISOLE: IMPACTS ON ACHE SIGNALLING

Postsynaptic neuron

levamisole

Presynaptic neuron

ACh signalling Receptor over-activation

• C. elegans paralysis

2HEE p=<0.001 1-Octanol p=0.012 1- hexanol p=0.004

Class 1 narcotics

O-CRESOL p=<0.001 ANILINE p<0.01 Octylamine <0.05

Class 2 narcotics Work done by

• L. Boakes

ACh AChR How are narcotic chemicals exerting a protective effect from the post-synaptic

• ver-activation caused by levamisole?

In top 10 highest scoring positively enriched genes Ric-3 enrichment -> Changes to post-synaptic receptor turnover -> delayed paralysis

RIC-3: Resistance to Inhibitors

• f Cholinesterase
• Necessary for function of

nicotinic acetylcholine receptors GAR-3: muscarinic AchR

• Required for

membrane potential via mobilizing calcium

HYPOTHESES ON POTENTIAL NARCOTIC TARGETS WHICH AFFECT ACHE SIGNALLING

Postsynaptic neuron Presynaptic neuron ACh AChR

In top 10 highest scoring negatively enriched genes Reduced AchR -> Less receptor over-activation -> protective effect during levamisole exposures

DEG-3: subunit of nicotinic AchR

levamisole

We are ready to validate these hypotheses using gene knock-out models to determine their role in narcosis toxicity.

CONCLUSIONS AND FUTURE DIRECTIONS

– NARCOSIS IN C. ELEGANS AND RTGILL CELL LINE

• Significant correlations between LC50 and Log Kow values for both systems

– Exposure modelling able to reduce regression variation

– MOLECULAR RESPONSES TO NARCOTIC CHEMICALS

• Significant genes and pathway-level responses between class 1 and class 2 and when

comparing low to high log Kow chemicals

• We can predict class 1 and class 2 chemicals using gene expression signatures

– MEMBRANE PROTEIN TARGETS INVOLVED IN NARCOSIS

• Relevant genes can be determined using a correlation network approach

– Preliminary hypotheses on the role of AchR signalling pathways and significant genes from this dataset

FUTURE WORK

– Determine genes relevant across species to develop as screening/predictive tool

• Develop gene or other targeted read-outs

that can be used to determine likelihood

• f a chemical being a narcotic versus a

specific-acting chemical

ACKNOWLEDGMENTS

• Francesco Falciani and our lab group

– Philipp Antczak, Danilo Basili, Leigh Boakes

• Unilever and the SEAC

– Geoff Hodges, Emma Butler, Steve Gutsell, Chris Sparham, Cecilie Rendal

• Mark Cronin (LJMU)
• Mark Viant (University of Birmingham)
• University of Liverpool Physiology department/’Red Block’

THANK YOU! ANY QUESTIONS?