Download FROM PATHWAYS TO PEOPLE: ALLERGIC CONTACT DERMATITIS GAVIN MAXWELL

FROM PATHWAYS TO PEOPLE: MODELLING
ALLERGIC CONTACT DERMATITIS
GAVIN MAXWELL
SAFETY & ENVIRONMENTAL ASSURANCE CENTRE (SEAC)
UNILEVER R&D
ALLERGIC CONTACT DERMATITIS
1. Skin penetration
and haptenation:
covalent
modification of
skin protein
9. Recruitment
of antigenspecific
memory T cells
and expansion
of effector T
cells to elicit
response
2. Migration of
Langerhans cells
and dermal
dendritic cells
8. Processing
and
presentation by
skin APCs
3. Antigen
processing and
presentation by
dendritic cells
4. Presentation of
haptenated peptide by
dendritic cell to T cells
5. Proliferation and
differentiation of
specific T cells
6. Generation of
antigen-specific
memory T cell
population
Image from: Karlberg et al. Chem. Res. Toxicol. (2008), 21, 53-69.
7. Re-exposure
to chemical
CURRENT HUMAN HEALTH RISK ASSESSMENT
PARADIGM FOR CHEMICAL INGREDIENTS
NOAEL
No Observed
Adverse Effect Level
(NOAEL) ÷ 10 - 1000
NEW HUMAN HEALTH RISK ASSESSMENT
PARADIGM FOR SENSITISING INGREDIENTS?
Allergic Immune Response
dose Y
Adverse
Non-Adverse
dose X
Time
NEW HUMAN HEALTH RISK ASSESSMENT
PARADIGM FOR SENSITISING INGREDIENTS?
1. Skin
Penetration
2. Electrophilic
substance:
directly or via
auto-oxidation or
metabolism
3-4. Haptenation:
covalent modification
of epidermal proteins
5-6. Activation of
epidermal
keratinocytes &
Dendritic cells
7. Presentation of
haptenated protein by
Dendritic cell resulting in
activation & proliferation
of specific T cells
8-11. Allergic Contact
Dermatitis: Epidermal
inflammation following
re-exposure to substance
due to T cell-mediated
cell death
Modified from ‘Adverse Outcome Pathway (AOP) for Skin Sensitisation’, OECD
DEVELOP A MATHEMATICAL MODEL OF ALLERGIC
CONTACT DERMATITIS TO ENABLE RISK ASSESSMENT
DECISION-MAKING FOR NEW CHEMICALS
Vehicle
Lymphatics
Draining Lymph
Nodes
Blood/resting
lymphatics
Epidermis
LC
Dermis
dDC
LC
dDC
Key
Chemical
Protein
Dendritic cell
CD8+ T cell: N = naïve;
CM = central memory;
EM = effector memory;
E = effector
DC
N
N
DC
CM
CM
PM
EM
EM
E
E
E
Skin
KEY ASSUMPTION: ANTIGEN DRIVING T CELL
RESPONSE IS HAPTENATED PEPTIDE
Direct Acting
- haptenated residues present on pMHC initiating the response
&/or
Altered Processing
- haptenated residues disrupt normal proteaosome processing
resulting in presentation of altered self-peptides
&/or
Altered Selection
- hapten activity disrupts MHC loading resulting in altered
selection of self
Dendritic cell
T cell
PREDICTING HAPTENATION RATE OF SKIN
PROTEIN BY DI-NITROCHLOROBENZENE (DNCB)
• Modelling approach - treat proteins as mixture of nucleophilic residues
• Use experimental data to determine ‘bulk’ haptenation rate & estimate
the fraction of nucleophiles we expect to be haptenated
Solution
F
Concentration
210
205
200
195
190
185
180
175
170
0
2
4
6
8
10
12
14
16
18
20
22
24
26
Cys
Tyr
Lys
His
Arg
Met &
N-term
300
250
Concentration
215
200
150
100
50
0
0
2
4
6
8
10
Time (hours)
T
Bound in Skin
150
100
Key
Chemical
Protein
Nucleophilic
residue
Concentration
Skin
Concentration
80
100
50
0
0
2
4
6
8
10
12
14
16
18
Time (hours)
20
22
24
26
60
40
20
0
0
2
4
6
8
10
T
MODELLING SKIN BIOAVAILABILITY OF CHEMICAL
Donor
chamber
Skin
position
Receptor
chamber
Loss from
formulation
Window
Receptor
solution out
Partitioning
Diffusion
Loss from skin
Davies et al. 2011. Toxicol Sci. 119. 308-18
PREDICTING EXTENT OF SKIN PROTEIN HAPTENATION
FOLLOWING SINGLE EXPOSURE TO DNCB
Skin bioavailability model expanded to include covalent
modification of skin protein by chemical
» Amount of haptenated protein predicted over time
Solution
Free in Skin
215
300
210
Concentration
Concentration
250
205
200
195
190
185
180
200
150
100
50
175
170
0
2
4
6
8
10
12
14
16
18
20
22
24
0
26
0
2
4
6
8
10
Time (hours)
12
14
16
18
20
22
24
26
20
22
24
26
Time (hours)
Bound in Skin
Protein
150
100
Concentration
Concentration
80
100
50
0
0
2
4
6
8
10
12
14
16
18
Time (hours)
20
22
24
26
60
40
20
0
0
2
4
6
8
10
12
14
16
18
Time (hours)
» Haptenated protein and free chemical concentrations
are inputs to immune response model
Davies et al. 2011. Toxicol Sci. 119. 308-18
TRANSLATING CHEMICAL SENSITISER EXPOSURE
INTO EXTENT OF HAPTEN PRESENTATION
Formulation
Lymphatics
Epidermis
LC
Dermis
dDC
Dendritic cell
Draining Lymph
Nodes
Blood/resting
lymphatics
LC
dDC
DC
N
N
DC
CM
CM
PM
EM
EM
E
E
E
Skin
T cell
• Intracellular LC/DC protein is haptenated by free chemical
• Proteasomal processing and Class I MHC presentation
• DC-T cell synapse in draining lymph node
MODELLING PROTEASOMAL PROCESSING
& CLASS I MHC ANTIGEN PRESENTATION
Assume ‘Direct Acting’ hypothesis (unaltered proteasomal processing)
and determine properties of resulting peptides
Prediction tools
Proteasomal
cleavage (e.g.
NetChop)
average number of
pMHC generated per
protein
MHC I binding (e.g.
NetMHCpan)
average
number of
nucleophiles
per pMHC
Estimate average pMHC surface density from considerations of:
1. the fraction of nucleophiles we expect to be haptenated
2. probability that a pMHC contains a haptenated nucleophile
ILLUSTRATION FROM YEWDELL, J.W., E. REITS, AND J. NEEFJES. (2003). Making sense of mass destruction: quantitating MHC class I antigen
presentation. Nat. Rev. Immunol. 3, 952–61.
VITA, R., L. ZAREBSKI, J.A. GREENBAUM, ET AL. (2010). The immune epitope database 2.0. Nucleic Acids Res. 38, D854–62.
MODELLING DC-T CELL INTERACTIONS
IN DRAINING LYMPH NODE
Formulation
Lymphatics
Epidermis
LC
Dermis
dDC
Draining Lymph
Nodes
Blood/resting
lymphatics
LC
dDC
DC
N
N
DC
CM
CM
PM
EM
EM
E
E
E
Skin
• LC/dDC migrate from sensitiser-exposed skin to present haptenated
peptides via Class I MHC to CD8+ T cell in draining lymph node
e.g. Pickard et al, 2009
• DC/T cell movement in lymph node is described by random walk
e.g. Day & Lythe, 2012
DEVELOP A MATHEMATICAL MODEL OF ALLERGIC
CONTACT DERMATITIS TO ENABLE RISK ASSESSMENT
DECISION-MAKING FOR NEW CHEMICALS
Solution
Free in Skin
215
300
210
Concentration
Concentration
250
205
200
195
190
185
180
200
150
100
50
175
170
0
2
4
6
8
10
12
14
16
18
20
22
24
0
26
0
2
4
6
8
10
Time (hours)
Haptenated
protein
Bound in Skin
Vehicle
12
14
16
18
20
22
24
26
Time (hours)
Protein
150
100
LC
50
0
0
2
4
6
8
10
12
14
16
18
Time (hours)
Dermis
dDC
20
22
24
26
Allergic Immune ResponseConcentration
Epidermis
Concentration
80
100
60
dose Y
40
20
0
0
2
4
6
8
10
12
14
16
18
20
22
24
26
Time (hours)
Adverse
Non-Adverse
dose X
Time
‘T LYMPHOCYTES: ORCHESTRATORS OF SKIN
SENSITISATION’ WORKSHOP – MAY 2010, LONDON
Immunologists, risk assessors &
mathematical modellers – 2 day
workshop
What are the characteristics of the T
cell response that could reflect
sensitiser potency in humans?
Number of T lymphocytes
» Magnitude: What is the extent of
sensitiser-induced T cell response
(volume, kinetics & duration)?
» Quality: Within sensitiser-induced T cell
response, what is the balance between
the T cell sub-populations?
» Breadth: What proportion of the T cell
clonal repertoire has been stimulated by
a given sensitiser?
Weaker allergen
Stronger allergen
Treg
CD8+
CD8+
Treg
Kimber et al. 2012. Toxicology. 291. 18-24
Time
CD8+ T CELL RESPONSE: INITIAL MODEL SCOPE
Current model scope models
the antigen-specific CD8+ T
cell response including:
» naïve (N) - CD45RO-veCD62L+ve
or CD45RA+veCD27+ve
» central memory (CM) CD45RO+veCD62L+ve
or CD45RA-veCD27+ve
» effector memory (EM) CD45RO+veCD62L-ve
or CD45RA-veCD27-ve
» effector (E) - CD45RO-veCD62Lve or CD45RA+veCD27-ve
Human sensitiser-specific T
cell data is largely unavailable:
» Make use of literature data
» Generate sensitiser-specific,
human-relevant data
Draining Lymph
Nodes
Blood/resting
lymphatics
Key
Chemical
Protein
Dendritic cell
CD8+ T cell: N = naïve;
CM = central memory;
EM = effector memory;
E = effector
DC
N
N
DC
CM
CM
PM
EM
EM
E
E
E
Skin
PREDICTING THRESHOLD FOR T CELL ACTIVATION
Is the nature (TCR affinity) of the antigen limiting?
- what kon/koff do TCRs have for cognate hapten pMHC
Explore effect of pMHC surface density and kon/koff on probability of T-cell
triggering with the available models (Zarnitsyna & Zhu, 2012). Simulations
generated using ‘confinement time’ model of Dushek, et al, 2009.
Figures from: Huppa & Davis, 2013; Aleksic et al., 2010
PREDICTING THRESHOLD FOR T CELL ACTIVATION
Is the nature (TCR affinity) of the antigen limiting?
- what kon/koff do TCRs have for cognate hapten pMHC
Explore effect of pMHC surface density and kon/koff on probability of T-cell
triggering with the available models (Zarnitsyna & Zhu, 2012). Simulations
generated using ‘confinement time’ model of Dushek, et al, 2009.
Molecular basis of T cell recognition: how
do TcRs interact with sensitising antigens?
TcR
V-J region usage
• Thermodynamic and
kinetic parameters
• Role of MHC
• Characteristics of the
CDR3s, and framework
•
MHC
Using DeCombinatoR:
(//github.com/uclinfectioni
mmunity/Decombinator) to
assign TcR sequences - V
region usage, J region
usage, no. of V deletions,
no. of J deletions, CDR3
sequence read
Benny Chain & Theres Matjeka
Number of nucleotides added
between V and J segment
Number of V germline deletions
Number of J germline deletions
CD8+ T CELL DIFFERENTIATION:
COMPARING CURRENT HYPOTHESES
• Experiments tracking T cell
fates have generated a range of
hypotheses on T cell differentiation
• Need to select a differentiation
mechanism despite uncertainty to
predict the number of CD8+ memory
T cells following sensitizer exposure
• Currently building CD8+ T cell
models based upon both decreasingpotential (Leeds) & asymmetricdivision (Unilever) to explore the
impact of each mechanism on
predicted T cell response
Image from: Kaech and Cui, Nat. Rev. Immunol. (2012), 12, 749-761
Sheeja Krishnan, Grant Lythe & Carmen Molina-Paris
STARTING T CELL POPULATION SIZE
» Assume no antigen specific effector or memory CD8+ T cells at
the start in an unexposed individual
» Estimate number of naïve antigen specific CD8+ T cells in DLN
& blood
» Assume exposure to skin on the arm
» 25 draining lymph nodes (DLN) in axilla out of 650 in total
» Consider a single TCR
» One in 25 million naïve T cells are antigen specific
Whole of body
72.5 bn
2900
All LNs
All TCRs
Ag specific (1 TCR)
19 bn
760
DLN
Blood
0.73 bn
29
1.45 bn
58
Vrisekoop et al, 2008, PNAS 105 (16) 6115-6120; Westermann & Pabst, 1992, Clin.
Investig. 70 539-544; Arstila et al, 1999, Science 286 958
MODELLING PROGRAMMED T CELL PROLIFERATION
CD69
• Following activation, CD8+ T cell proliferation continues
independently of further antigenic stimulation
• Going through 7-19 generations (Kaech & Ahmed, 2001; Badovinac
et al, 2007) to develop effector and memory populations
• No human data available for proliferation rates
• Obtain proliferation rates from mouse models (e.g. Yoon et al, 2010:
draining lymph node response to influenza virus infection)
CFSE
Figure from: Yoon et al, 2010, PLOS One 5 (11) e15423
Draining
Lymph Nodes
DC
DC
N
Blood/resting
lymphatics
N
Key
Chemical
Protein
Dendritic cell
CD8+ T cell: N =
naïve; CM = central
memory; EM =
effector memory;
E = effector
CM
CM
PM
EM
EM
E
E
E
No. T cells
CD8+ T CELL MODEL PREDICTIONS: 5 DAY ANTIGEN
EXPOSURE IN LYMPH NODE, 1X MODEL ITERATION
Time in days
Skin
• Combine the parameters and processes together
• Simulate single exposure to chemical and track response for one month
Draining
Lymph Nodes
DC
DC
N
Blood/resting
lymphatics
N
Key
Chemical
Protein
Dendritic cell
CD8+ T cell: N =
naïve; CM = central
memory; EM =
effector memory;
E = effector
CM
CM
PM
EM
EM
E
E
E
Frequency of
prediction
CD8+ T CELL MODEL PREDICTIONS: 5 DAY ANTIGEN
EXPOSURE IN LYMPH NODE, 1000X MODEL ITERATIONS
No. T cells
Skin
• Combine the parameters and processes together
• Simulate single exposure to chemical and track response for one month
Characterising human T lymphocyte responses to
chemical allergen p-phenylenediamine (PPD)
0µg/ml PPD
0.01
0.1
Ki-67
CD8
CD4
Allergen driven proliferation of total lymphocytes and individual
T cell subsets measured by intracellular Ki-67 expression.
Rebecca Dearman,
Amy Popple, Ian Kimber &
Jason Williams
DEVELOP A MATHEMATICAL MODEL OF ALLERGIC
CONTACT DERMATITIS TO ENABLE RISK ASSESSMENT
DECISION-MAKING FOR NEW CHEMICALS
Vehicle
Lymphatics
Draining Lymph
Nodes
Blood/resting
lymphatics
Epidermis
LC
Dermis
dDC
LC
dDC
Key
Chemical
Protein
Dendritic cell
CD8+ T cell: N = naïve;
CM = central memory;
EM = effector memory;
E = effector
DC
N
N
DC
CM
CM
PM
EM
EM
E
E
E
Skin
PATHWAYS-BASED RISK ASSESSMENT FOR SKIN SENSITISATION:
APPLICATION OF MATHEMATICAL MODELLING
1. Skin
Penetration
2.Electrophilic
substance:
directly or via
auto-oxidation
or metabolism
3-4. Haptenation:
covalent
modification of
epidermal proteins
5-6. Activation of
epidermal
keratinocytes &
Dendritic cells
7. Presentation of
haptenated protein by
Dendritic cell resulting
in activation &
proliferation of specific
T cells
8-11. Allergic Contact
Dermatitis: Epidermal
inflammation following
re-exposure to substance
due to T cell-mediated
cell death
allergic immune response
No. CD8+ T cells
dose Y
Adverse
Non-Adverse
dose X
time
1.
Generate skin bioavailability & haptenation data as model input parameters
2.
Use linked mathematical models to predict human allergic immune response
3.
Apply human immune response model prediction for risk assessment decision
4.
If exposure predicted to be non-adverse, verify prediction using clinical data
NEXT STEPS: CHALLENGES AHEAD
• Broadening current model scope to include:
- CD4+ T helper & regulatory T cell responses
- sensitiser-induced inflammation in skin – induction & elicitation
- impact of varying frequency & surface area of sensitiser exposure
- impact of varying formulation (vehicle)
• Using experimental & clinical data to inform &
benchmark initial model predictions
ACKNOWLEDGEMENTS
Unilever
Richard Cubberley, Seraya Dhadra, Michael Davies,
Nikki Gellatly, Stephen Glavin, Todd Gouin, Sandrine
Jacquoilleot, Cameron MacKay, Craig Moore, Ruth
Pendlington, Juliette Pickles, Ouarda Saib, David
Sheffield, Richard Stark, Vicki Summerfield & Sam
Windebank
University of Leeds
Sheeja Krishnan, Grant Lythe & Carmen Molina-Paris
University of Manchester
Rebecca Dearman, Amy Popple & Ian Kimber
Salford Royal NHS Foundation Trust
Jason Williams
University College London
Benny Chain & Theres Matjeka