Application of a systems biology approach for skin allergy ...

381© 2008, Japanese Society for Alternatives to Animal Experiments

Application of a systems biology approach for skin allergy risk assessment

Gavin Maxwell and Cameron MacKay

Unilever – Safety & Environmental Assurance Centre (SEAC)

Corresponding Author: Gavin MaxwellUnilever – Safety & Environmental Assurance Centre (SEAC)

Colworth Science Park, Sharnbrook, Bedfordshire MK44 1LQ, UKPhone: +(44)-1234-264888, Fax: +(44)-1234-264744, [email protected]

AbstractWe have developed an in silico model of skin sensitization induction to characterise and quantify the contribution of each pathway to the overall biological process; through this analysis we have developed a rationale for the interpretation and potential integration of in vitro test data for risk assessment purposes.

The mouse local lymph node assay (LLNA) is now in widespread use for the evaluation of skin sensitization potential. Recent changes in EU legislation [i.e. 7th Amendment to the EU Cosmetics Directive] have made developing non-animal approaches to provide the data for skin sensitization risk assessment a key business need. Several in vitro assays have already been developed for the prediction of skin sensitization; however these are based on measuring a few pathways within the overall biological process and our understanding of the relative contribution of these individual pathways to skin sensitization induction is limited.

To address this knowledge gap and thereby guide further in vitro assay development, a "systems biology" approach has been used to construct a computer-based mathematical model of the induction of skin sensitization, in collaboration with Entelos Inc. Biological mechanisms underlying the induction phase of skin sensitization are represented by non-linear ordinary differential equations, defined using qualitative and quantitative information from a total of 496 published papers. From the model, we have identified knowledge gaps for future investigative research, and key factors that have a major influence on the induction of skin sensitization (e.g. TNF-α production in the epidermis); we have quantified their relative contribution to the overall process. Through providing a biologically-relevant rationale for the interpretation and potential integration of diverse types of in vitro data, the in silico model has helped us to define and evaluate a new conceptual framework for skin sensitization risk assessment.

Keywords: in silico, in vitro, risk assessment, skin sensitization, systems biology

IntroductionDecisions about the consumer safety of our

products are made on the basis of a risk assessment, in which data on the potential hazards of the ingredients are interpreted in the context of the likely exposure to the product, i.e. the concentration of the ingredients in the product and how the product is used by consumers. Traditionally, much of the hazard data on chemicals have been generated by applying technologies developed for histological and clinical chemical analyses to animal models. Both the 6th and 7th Amendments to the EU Cosmetics Directive have stimulated considerable research into developing alternative approaches to assess consumer safety (EU, 2003) with improved non-animal in vitro and in silico models and new technologies such as proteomics and bioinformatics approaches becoming increasingly available to generate and interpret new types of non-

animal data. However, developing ways to integrate and determine the relative importance of these data to enable risk-based safety decisions to be made represents a major challenge (Fentem et al., 2004;US National Research Council, 2007).

To ensure that products do not induce skin (contact) allergy in consumers, information on the concentrations of the ingredients in the product and how the product is used by consumers, together with data generated in the mouse local lymph node assay (OECD, 2002) is used to assess whether a chemical ingredient has the potential to cause skin sensitisation, and thus skin allergy, in humans. The current generation of in vitro predictive assays for the skin sensitisation endpoint are primarily focussed in two key areas, both believed to be key events in skin sensitisation induction; measurement of sensitizer haptenation events by detection of peptide

AATEX 14, Special Issue, 381-388Proc. 6th World Congress on Alternatives & Animal Use in the Life SciencesAugust 21-25, 2007, Tokyo, Japan

382


modifications in vitro (Gerberick et al., 2007) and measurement of Dendritic cell (DC) activation (assessed by detecting changes in co-stimulatory receptor expression) following sensitizer treatment (Ade et al., 2005;Sakaguchi et al., 2006;Ashikaga et al., 2006;Sakaguchi et al., 2007;Python et al., 2007). The identified strategy moving forward is that data from these in vitro assays will be integrated with other hazard information (e.g. skin bioavailability predictions) to generate a new measure of sensitizer potential/potency that can be used to inform risk assessment decisions, in the absence of new animal data. Some relatively simple scoring approaches to integrating these data have been considered (Jowsey et al., 2006) and these together with more complex statistical approaches are currently being assessed. However we were also interested to explore whether mathematical modelling approaches (also termed systems biology) could be used to (a) generate new biological understanding; (b) guide our experimental research programme; (c) focus our development of new predictive in vitro assays; and (d) inform our risk assessments. To do so, a collaboration was

initiated with Entelos® Inc. to construct a computer-based mathematical model of the induction of skin sensitisation (termed Skin Sensitisation PhysioLab® (SSP) platform) (MacKay et al., 2007).

The aim of this study was to build a transparent, robust, mechanistic and quantitative in silico model that captured our current understanding of the biological pathways, processes and mediators involved in skin sensitisation in vivo. The SSP platform has been constructed such that the biology can be interrogated computationally in an iterative, hypothesis-driven manner (Fig. 1). The publications used to build and define the biological relationships are both accessible and can be removed and replaced with other information, if/when it becomes available. The model is not intended to be an exhaustive record of all biological pathways that could potentially be implicated in skin sensitisation induction but rather it represents the pathways known to be required (i.e. quantitative data exists in the literature to support their role). Therefore, the value of the model is in quantifying the accepted theory of skin sensitisation, in order to determine the influence of each pathway

Fig. 1. Skin Sensitization PhysioLab® Platform – CD4+ T cell Life Cycle exampleCD4+ T cell Life Cycle is shown as an example of the underlying functionality of the Skin Sensitization Physiolab® Platform. In the upper panel, the node and arc architecture can be seen, while in the lower panel a reference note (top right) and comparison chart (lower left) demonstrate how relevant literature is stored within the platform and how in silico experimental results are visualised, respectively.

383

on the predictive endpoint (e.g. antigen-specific T cell proliferation in the draining lymph node).

Materials and MethodsIn Silico Modelling

The software package PhysioLab Modeller (Entelos Inc.) was used to develop the SSP platform (Fig. 1). The software enables the user to graphically construct models of large-scale biological systems in a manner akin to pathway diagrams. Cellular and molecular interactions are represented quantitatively using ordinary differential equations (ODEs) the numerical solution of which simulates the dynamic response of the biological system. The form and parameterization of the ODEs (e.g. Michaelis-Menten, first-order, second-order or more complicated kinetics) are chosen to ensure consistency of the simulation with the published theory and experimental data. In cases where multiple accepted mechanisms or hypotheses exist, an alternate parameterisation of the model can be defined for each mechanism, subject to the constraints of experimental data. The model can then be interrogated for each parameterisation in order to obtain conclusions that are robust to the uncertainty in the mechanism. In cases where the modelled mechanism cannot explain the experimental data, the biology must be revisited to determine what the likely alternate mechanism is. The modelling is thus an iterative process in which the accepted theory is modelled, compared to experimental data and either accepted or refined.

Model ScopeCreation of the Skin Sensitization PhysioLab® platform

required a detailed literature review of approximately 500 publicly available publications. This process helped to determine the key elements that should be represented in the platform and identify the appropriate data that described those elements. The model describes the induction of skin sensitization over a period of 10 days following chemical exposure and is composed of two main tissue compartments, the epidermal layer of skin and the local draining lymph node (dLN) (Fig. 1). The epidermal compartment of the model contains keratinocytes, Langerhans' cells (LC), mast cells and T cells while the lymph node compartment is composed of DC, T cells (CD4+ and CD8+) and B cells. In addition to modelling the biological events occurring during the induction of skin sensitization, a number of modules have been included to perform miscellaneous calculations. Generally, the function of these calculations is to enable experimentally observed properties to be compared with those modelled.

Use of experimental dataTo ensure model consistency with published data,

a calibration process was performed at the individual pathway level. Published data was gathered on

modelled properties such as the release of epidermal cytokines in response to sensitizer stimulus or the migration rate of LC in response to epidermal cytokines. The majority of the biological data used to develop the model was derived from murine models due to the relative absence of skin sensitization-relevant datasets derived using human tissues. Data on rate constants, dose response curves and other temporal data were directly parameterised within the model. Where the required parameters were not directly available, the values in the model were manually adjusted within a physically feasible range in order to quantitatively and qualitatively reproduce the published results. For example, in Cumberbatch 2005, (Cumberbatch et al., 2005), the LC cell density in a mouse ear was measured at 4 and 17 hours following exposure to 1% DNCB. Additionally, the number of DC appearing in the LN was measured at 24, 48 and 72 hours following an identical exposure. In order to ensure that the model reproduced this data, the parameters governing LC migration rates were adjusted until the simulation results for an exposure of 1% DNCB were in qualitative and quantitative agreement with the observed experimental results. In order to simulate this experiment, it was necessary to first use other experimental data to determine the parameters governing the release of epidermal cytokines that drive LC migration in response to a 1% DNCB exposure. If when adjusting the parameters the chosen model could not reproduce the desired experimental results, the assumptions made about the biology were re-visited and a new model proposed. In this way the various pathways of the model were calibrated until the dynamics of the system-level response to which they contribute, i.e. Ag-specific proliferation, could be simulated.

The model was evaluated by linking together each of the individually calibrated pathways in order that the full system-level response be simulated. Among the chemical-specific information utilized for evaluation were data for 2,4-dinitrochlorobenzene (DNCB), oxazolone (OX), isoeugenol, cinnamic aldehyde, hexyl cinnamic aldehyde, and hydroquinone (chemical sensitizers), as well as p-aminobenzoic acid and glycerol (non-sensit izers) . Greatest emphasis was placed on DNCB and OX data, as the majority of chemical information available for both calibration and evaluation was restricted to various concentrations of these two chemicals. Only unknown chemical-specific parameters were adjusted during the evaluation process, while underlying biological parameters defined during the calibration process remained unchanged.

Sensitivity AnalysisSensitivity analysis was used to determine the

key pathways driving the sensitization response. The analysis was performed by modulating various

384


pathways in the model and comparing simulation results for selected outputs. In these simulations a standard protocol was implemented for sensitizer exposure, parameters associated with the selected pathways were modulated around their baseline values and the relative impact on the selected outputs was assessed. The exposure protocol employed was that of 3 consecutive exposures at 0, 24 and 48 hours [i.e. that of the murine local lymph node assay (LLNA)] to a sensitizer of a prescribed potency class (OECD, 2002).

A key aspect of sensitivity analysis was the selection of outputs to be analyzed. The maximum rate of Ag-specific CD4+ and CD8+ T cell expansion was chosen as the primary output for analysis as it was decided that this was the model output giving the best measure of the extent of skin sensitization induction. Maximum Ag-specific T cell expansion was recorded by measuring the peak level of antigen-specific T cell proliferation observed over a ten-day course following exposure. Additionally, the LLNA SI (OECD, 2002) was also used as secondary output.

Using chemical specific LLNA data model settings were defined for a prototypical weak, moderate and strong sensitizer. A reference dose was then selected for the sensitivity analysis performed on each prototypical chemical. For the prototypical weak, moderate and strong sensitizers, reference doses of 1%, 1% and 0.1% were chosen respectively and corresponded to a response at the low, moderate and high end of the dose response. Thus for the weak sensitizer the reference dose was such that pathways amplifying the response would be detected, for the moderate sensitizer pathways either amplifying or dampening the response would be detected and for the strong sensitizer pathways dampening the response would be detected. The pathway sensitivity index was defined as the range (in logs) between the maximal and minimal normalized outcome value at that reference dose. Thus, for example, a sensitivity score of 1 indicates a 10-fold difference between the outcome value observed when the pathway modulation was set to the lowest value and the outcome value observed when the pathway modulation was set to the highest value.

Results Model development required extensive quantitative

data analysis at the level of individual pathways as well as whole-scale system dynamics. Calibration experiments were performed to ensure that the key biological mechanisms and dynamics of individual pathways were consistent with published data. Several evaluation experiments were performed to ensure that the individually calibrated pathways, when acting together in a system, reproduced the dynamics observed experimentally without any changes to the biological representation (see Methods). Results show

that the calibrated and evaluated Skin Sensitization PhysioLab® platform reproduced many of the key experimental benchmarks observed in several studies. Data on dLN cellularity and composition as well as Ag-specific cell proliferation were used to create reasonable ranges that defined an acceptable response. For example, given the standard application of 0.25% of the sensitizing chemical DNCB in the LLNA, predicted dLN cell response and the dLN proliferative

Fig. 2. Implementation of 'recruitment hypothesis' within Skin Sensitization PhysioLab® PlatformA) Despite adjusting the parameters for maximum T-cell expansion, the SSP platform cannot reproduce the observed fold-increase in cellularity for 0.25% DNCB. While the absolute number of proliferating CD4+ and CD8+ T cells is in agreement with observed data, the percentage of proliferating cells exceeds that seen in the same sources (inset), suggesting the need for factors that increase the non-dividing cell population to be included. [* % proliferating cells reflects the CD62LloCD44hi percent +/- SEM cited in (Gerberick et al., 1999b), ** Fold increase compared to naïve control, ***Cellularity data from (Sikorski et al., 1996;Gerberick et al., 1999b;Van Och et al., 2000;Suda et al., 2002;Soderberg et al., 2005;Goutet et al., 2005)B) With enhanced recruitment implemented in the SSP platform, model predictions (solid bars) are consistent with in vivo data for fold increase in cellularity, Ag-specific proliferation (inset), and total cellularity in the dLN after stimulation with 0.25% DNCB and other chemical sensitizers reported in the literature. [****OX data from (Dearman et al., 1999;Gerberick et al., 1999b), DNCB data from (Gerberick et al., 1999b;Van Och et al., 2000;Suda et al., 2002;Goutet et al., 2005), HCA data from (Dearman et al., 1999;Gerberick et al., 1999b;Gerberick et al., 2002), and ISO data from (Dearman et al., 1999;Gerberick et al., 1999b).]

385

responses were in line with those reported in several studies. The platform was similarly able to capture the cellular and proliferative responses to other doses of DNCB as well as other chemicals. Furthermore, the platform ultimately reproduced the key dynamics observable with the LLNA and also al lowed exploration of underlying mechanisms (often difficult to measure in vivo) giving rise to the observed phenomena.

Novel insights from model buildingDue to the quantitative nature of modelling

experimental data, it is often possible to recognize inconsistencies that would otherwise have gone unnoticed. During model building and calibration, it was found that the expansion of Ag-specific T cells alone was not sufficient to reproduce the dLN cellularity reported in the published literature. Experiments investigating application of a 0.25% dose of DNCB in the LLNA revealed that the overall cellularity of the dLN could increase by as much as 14-fold over control conditions (Dearman et al., 1999;Gerberick et al., 1999a;Van Och et al., 2000;Suda et al., 2002). Despite accurate calibration of the dynamics associated with T-cell proliferation, however, the model could not achieve such high levels of dLN cellularity (Fig. 2). In order to investigate this apparent discrepancy, we formulated a simpler model of proliferation in the dLN in which we considered the number of Ag-specific T cells able to proliferate. The number of naïve Ag-specific clones (i.e., the number of different T cell clones able to recognize Ag generated by exposure to the chemical), the clonal frequency (i.e., the frequency with which each T cell clone is likely to occur in the T cell population) and the number of cell divisions were set to established values (Gudmundsdottir et al., 1999;Kehren et al., 1999;Jelley-Gibbs et al., 2000;Kaech and Ahmed, 2001;Vocanson et al., 2005). However, even in the scenario of no cellular efflux from the dLN, our simple model showed that the experimentally observed cellularity could not be achieved within the 4 to 10 cell divisions reported in the literature (Gudmundsdottir et al., 1999;Jelley-Gibbs et al., 2000;Kaech et al., 2001). Achievement of the observed cellularity requires biologically infeasible values of naïve Ag-specific clones or clonal frequency. Interestingly, experimental evidence suggested that increased cellularity in an excised dLN could also be observed upon treatment with sodium lauryl sulfate (SLS) (a non-sensitizer) (Gerberick et al., 1999a;Lee et al., 2002). Although SLS is recognised not to be a skin sensitizer, increased migration of LCs to the dLN has been observed in mice following SLS treatment (Cumberbatch et al., 1993). However, the increase in dLN cellularity following SLS treatment cannot be accounted for

based solely on the number of immigrant LCs in the dLN. Therefore, we sought evidence for other immune processes that might result in significant increases in cellularity. While not usually considered a primary driver of the skin sensitization response, data in other immunological contexts (e.g. inflammation and virus infection) supported the phenomenon of increased non-specific recruitment of cells to the dLN following infection (Martin-Fontecha et al., 2003;Soderberg et al., 2005). These findings resulted in an exploration of alternate mechanisms to explain the increase in dLN cellularity during sensitization.

Significance of T cell recruitmentThe platform captures dynamics of increased

recruitment through enhanced influx of naïve lymphocytes in response to maturing LCs in the dLN. By including the dynamics of enhanced recruitment in the platform, we were able to reproduce key aspects of LLNA data (Fig. 2). To test the hypothesis that enhanced recruitment is vital to reproduce this data, we created a parameterization of the model without the hypothesized enhanced recruitment mechanisms. In this model version, simulation of response to a standard LLNA protocol with a strong sensitizer resulted in a mismatch with data across several of the criteria defined for validation, particularly in extremely low total T cell counts in spite of high percentages of CD4+ and CD8+ T cell proliferation. No adjustments in other parameters were able to increase total T cells numbers or decrease the number of proliferating cells to acceptable (in range) values. Therefore, our conclusion was that increased recruitment was necessary to explain existing data on LN cellularity and cellular proliferation.

Sensitivity analysis identified key pathways driving sensitization

As a f i rs t s tep in determining the re la t ive contr ibut ion of individual pathways on skin sensitization induction we conducted a sensitivity analysis on the model. Sensitivity analysis is a method of assessing key drivers of an overall response such as sensitization. Parameters associated with these pathways are selected for analysis and are modulated around their baseline values to assess the resulting impact on selected outputs. As described in the Methods, performing a sensitivity analysis enabled determination of the relative contribution of specific pathways to the overall response. The output of primary interest for the assessment of sensitizer potency was the ability to elicit CD4+ and CD8+ Ag-specific T cell proliferation. Therefore we measured the impact of several pathways on this output variable and explored sensitivities for a prototypical weak, moderate (data not shown) and strong sensitizer (Fig. 3).

386


Antigen-specific proliferation. An analysis was performed to assess the relative contribution of functional pathways in various biological categories on maximum Ag-specific proliferation. Contribution was measured by identifying the most sensitive pathways, i.e. those that, when modulated, resulted in the greatest change in increasing or decreasing maximal Ag-specific proliferation. Pathways were grouped into generalised categories and the range of modulation for each pathway was chosen to represent a reasonable operating range for that pathway. Surprisingly, pathways other than those associated with 'Antigen presentation' were among the top drivers of Ag-specific proliferation (Fig. 3a and 3b). Indeed all pathway categories were found to be strong drivers of Ag-specific T-cell proliferation. Within the 'Epidermal activation and irritation pathway' category, TNF-α proved to have the most significant contribution (data not shown). These results suggest epidermal cytokine production and its effects on LC maturation and migration are key steps in the induction of sensitization. Analyzing the Ag-specific output for both weak and strong sensitizers did not alter the conclusions.

Local Lymph node assay Stimulation Index (LLNA SI). The sensitivities of the LLNA SI output

to the same biological pathways were also assessed. Using LLNA SI as the output variable, we found a significant difference in the relative contribution of the pathway categories between weak and strong prototypic chemicals (Fig. 3c and 3d). With prototypic strong sensitizers all pathway categories appear to contribute in a comparative fashion to LLNA SI (Fig. 3, panel d). However, this profile changes substantially when prototypic weak sensitizers are analysed, with 'Epidermal activation and irritation' and 'Non-specific response' pathways constituting the majority of influence upon LLNA SI (Fig. 3, panel c). Consequently it can be summarised that the relative contribution of the pathway categories and therefore the individual pathways to the LLNA SI is likely to vary with sensitizer potency.

DiscussionThe SSP platform, was developed to (a) generate

new biological understanding; (b) guide our experimental research programme; (c) focus our development of new predictive in vitro assays; and (d) inform our risk assessments. From our experiences to date we believe this new approach has applications in each of these areas and that it will continue to generate new biological insights through combining

Fig. 3. Sensitivity of Maximum Ag-specific proliferation and Local Lymph Node Assay Stimulation Index to key pathwaysSensitivity is defined as the range (in logs) between maximal and minimal value of output attained through modulation of each pathway. Maximum Ag-specific proliferation (panels a & b) and LLNA SI (panels c & d) were selected as the outputs for the sensitivity analysis. Results for two prototypical chemicals are shown: a weak sensitizer (panels a & c) and strong sensitizer (panels b & d) are shown. Pathway category sensitivities are computed by taking an average of the associated individual pathway sensitivity.

387

previously disparate data sets and/or drawing parallels from other disease processes. The implementation of the recruitment hypothesis within the SSP to 'balance' dLN cell dynamics is a good example of where pathogen infection literature was used to identify and explain a previously underappreciated pathway within skin sensitisation induction (Fig. 2). Clearly, this hypothesis will now need to be confirmed experimentally before any firm conclusions can be drawn and other unidentified, uncharacterised sensitizer-specific effects may also be involved; however it remains debatable that this question would even have been posed without first modelling the available dLN cell data. Consequently, there exists a belief that further biological insights will be yielded through iterative modelling of experimental data moving forward, as has been shown elsewhere (Lewis et al., 2001;Rullman et al., 2005). Therefore it is our intention to incorporate data generated from our future research and the scientific literature into the model, where possible.

With respect to guiding future in vitro assay development, the sensitivity analysis of the SSP platform increased our understanding and confirmed previously discussed hypotheses in equal measure. For example, the identified role for both antigen-specific and antigen non-specific pathways in driving sensitizer-induced proliferation in the dLN was already appreciated within the field (Kimber and Dearman, 2002). However, the dissection of the relative contribution of key pathways to LLNA SI (or antigen-specific T cell proliferation) and the assessment of how these pathways vary with sensitizer potency (using prototypic chemicals) represents a unique analysis that is potentially invaluable for guiding in vitro model development. For example, the identified strong influence of epidermal inflammatory signals upon sensitizer-induced DC activation highlights that there may be value in investigating the incorporation of key inflammatory mediators within existing DC activation models. While the apparent importance of a recruitment effect for inducing a draining lymph node proliferation may suggest that traditional DC: T cell co-culture protocols e.g. (Krasteva et al., 1996) need to be viewed in context of the proportion of the immune repertoire they can effectively scan. However these predictions should be interpreted in the context of the underlying models and technologies that were used to derive the data. For example good quality, quantitative, sensitizer-specific epidermal cytokine release data was difficult to identify and the majority of the data used to develop the SSP platform was derived from murine models; consequently our future efforts will focus on filling the identified data gaps with quantitative data generated using human models and in this way the platform represents an evolving

tool that can be iteratively improved as new data become available.

In addition to fulfilling its role as a research tool, the SSP can also be used in order to assist in determining and refining the key information required in a weight-of-evidence approach to consumer safety risk assessment. Jowsey et al. (Jowsey et al., 2006) published an initial concept of a weight-of-evidence approach to integrate non-animal data inputs as surrogates for several key processes known to be important mechanistically in the induction of skin sensitisation. Whilst recognising that the list of assays or data requirements is not definitive, this conceptual approach is a pragmatic starting point given the limited availability of new types of data. As more data are generated, statistical and mathematical approaches will be used to model and interpret data in a more robust way and it is here where the understanding gained from in silico modelling can be applied. Using the SSP sensitivity analysis results in order to propose the key data required for an integrated weight-of-evidence approach will allow the retention of a biological rationale in deriving a new measure of sensitizer potential/potency. Indeed, it may be that once all the key pieces of information have been determined, modelling approaches such as the one described here, may prove themselves to be a suitable method of dispassionately integrating diverse types of increasingly complex, multi-parameter in vitro datasets.

AcknowledgementsThe research is part of Unilever's ongoing effort to

develop novel ways of delivering consumer safety. The SSP platform was developed in collaboration with Entelos Inc., US.

References1. Ade N, Teissier S, Pallardy M, and Rousset F. Contact

Sensitizers Induce Apoptosis and CD86 Expression on U937 cells in an independent manner. 2005. Ref Type: Conference Proceeding.

2. Ashikaga T, Yoshida Y, Hirota M, Yoneyama K, Itagaki H, Sakaguchi H, Miyazawa M, Ito Y, Suzuki H, and Toyoda H (2006) Development of an in vitro skin sensitization test using human cell lines: the human Cell Line Activation Test (h-CLAT). I. Optimization of the h-CLAT protocol. Toxicol In Vitro, 20, 767-773.

3. Cumberbatch M, Clelland K, Dearman RJ, and Kimber I (2005) Impact of cutaneous IL-10 on resident epidermal Langerhans' cells and the development of polarized immune responses. Journal of Immunology, 175, 43-50.

4. Cumberbatch M, Scott RC, Basketter DA, Scholes EW, Hilton J, Dearman RJ, and Kimber I (1993) Influence of Sodium Lauryl Sulfate on 2,4-Dinitrochlorobenzene-Induced Lymph-Node Activation. Toxicology, 77, 181-191.

5. Dearman RJ, Hilton J, Basketter DA, and Kimber I (1999) Cytokine endpoints for the local lymph node assay: Consideration of interferon-gamma and interleukin 12. Journal of Applied Toxicology, 19, 149-155.

388


6. EU (2003) Directive 2003/15/EC of the European Parliament and of the Council of 27 February 2003 amending Council Directive 76/768/EEC on the approximation of the laws of the Member States relating to cosmetic products. Official Journal of the European Union, L66, 26-35.

7. Fentem J, Chamberlain M, and Sangster B (2004) The feasibility of replacing animal testing for assessing consumer safety: a suggested future direction. Altern Lab Anim, 32, 617-623.

8. Gerberick GF, Cruse LW, Miller CM, and Ridder GM (1999a) Selective modulation of B-cell activation markers CD86 and I- A(k) on murine draining lymph node cells following allergen or irritant treatment. Toxicology and Applied Pharmacology, 159, 142-151.

9. Gerberick GF, Cruse LW, and Ryan CA (1999b) Local lymph node assay: Differentiating allergic and irritant responses using flow cytometry. Methods-A Companion to Methods in Enzymology, 19, 48-55.

10. Gerberick GF, Cruse LW, Ryan CA, Hulette BC, Chaney JG, Skinner RA, Dearman RJ, and Kimber I (2002) Use of a B cell marker (B220) to discriminate between allergens and irritants in the local lymph node assay. Toxicological Sciences, 68, 420-428.

11. Gerberick GF, Vassallo JD, Foertsch LM, Price BB, Chaney JG, and Lepoittevin JP (2007) Quantification of chemical peptide reactivity for screening contact allergens: A classification tree lmodel approach. Toxicological Sciences, 97, 417-427.

12. Goutet M, Pepin E, Langonne I, Huguet N, and Ban M (2005) Identification of contact and respiratory sensitizers using flow cytometry. Toxicology and Applied Pharmacology, 205, 259-270.

13. Gudmundsdottir H, Wells AD, and Turka LA (1999) Dynamics and requirements of T cell clonal expansion in vivo at the single-cell level: effector function is linked to proliferative capacity. J Immunol, 162, 5212-5223.

14. Jelley-Gibbs DM, Lepak NM, Yen M, and Swain SL (2000) Two distinct stages in the transition from naive CD4 T cells to effectors, early antigen-dependent and late cytokine-driven expansion and differentiation. J Immunol, 165, 5017-5026.

15. Jowsey IR, Basketter DA, Westmoreland C, and Kimber I (2006) A future approach to measuring relative skin sensitising potency: a proposal. J Appl Toxicol, 26, 341-350.

16. Kaech SM and Ahmed R (2001) Memory CD8+ T cell differentiation: initial antigen encounter triggers a developmental program in naive cells. Nat Immunol, 2, 415-422.

17. Kehren J, Desvignes C, Krasteva M, Ducluzeau MT, Assossou O, Horand F, Hahne M, Kagi D, Kaiserlian D, and Nicolas JF (1999) Cytotoxicity is mandatory for CD8(+) T cell-mediated contact hypersensitivity. J Exp Med, 189, 779-786.

18. Kimber I and Dearman RJ (2002) Allergic contact dermatitis: The cellular effectors. Contact Dermatitis, 46, 1-5.

19. Krasteva M, Moulon C, Peguet-Navarro J, Courtellemont P, Redziniak G, and Schmitt D (1996) In vitro sensitization of human T cells with hapten-treated Langerhans cells: a screening test for the identification of contact allergens. Curr Probl Dermatol, 25, 28-36.

20. Lee JK, Park JH, Park SH, Kim HS, and Oh HY (2002) A nonradioisotopic endpoint for measurement of lymph node cell proliferation in a murine allergic contact dermatitis model, using bromodeoxyuridine immunohistochemistry. Journal of Pharmacological and Toxicological Methods, 48, 53-61.

21. Lewis AK, Paterson T, Leong CC, Defranoux N, Holgate ST, and Stokes CL (2001) The Roles of Cells and Mediators in a Computer Model of Chronic Asthma. International Archives of Allergy Immunology, 124, 282-286.

22. MacKay C, Bajaria S, Shaver G, Kudrycki K, Ramanujan S, Paterson T, Friedrich C, Maxwell G, Jowsey I, Lockley D, Reynolds F, and Fentem J (2007) In silico modelling of skin sensitisation. Toxicology, 231, 103.

23. Martin-Fontecha A, Sebastiani S, Hopken UE, Uguccioni M, Lipp M, Lanzavecchia A, and Sallusto F (2003) Regulation of dendritic cell migration to the draining lymph node: impact on T lymphocyte traffic and priming. J Exp Med, 198, 615-621.

24. OECD (2002) OECD Guideline for the Testing of Chemicals, No. 429. Skin Sensitisation: Local Lymph Node Assay, 7pp, OECD, Paris.

25. Python F, Goebel C, and Aeby P (2007) Assessment of the U937 cell line for the detection of contact allergens. Toxicol Appl Pharmacol, 220, 113-124.

26. Rullman JAC, Struemper H, Defranoux NA, Ramanujan S, Meeuwisse CML, and van Elesen A (2005) Systems biology for battling rheumatoid arthritis: application of the Entelos Physiolab. IEE Proceeding Systems Biology, 152, 256-262.

27. Sakaguchi H, Ashikaga T, Miyazawa M, Yoshida Y, Ito Y, Yoneyama K, Hirota M, Itagaki H, Toyoda H, and Suzuki H (2006) Development of an in vitro skin sensitization test using human cell lines; human Cell Line Activation Test (h-CLAT). II. An inter-laboratory study of the h-CLAT. Toxicol In Vitro, 20, 774-784.

28. Sakaguchi H, Miyazawa M, Yoshida Y, Ito Y, and Suzuki H (2007) Prediction of preservative sensitization potential using surface marker CD86 and/or CD54 expression on human cell line, THP-1. Archives of Dermatological Research, 298, 427-437.

29. Sikorski EE, Gerberick GF, Ryan CA, Miller CM, and Ridder GM (1996) Phenotypic analysis of lymphocyte subpopulations in lymph nodes draining the ear following exposure to contact allergens and irritants. Fundamental and Applied Toxicology, 34, 25-35.

30. Soderberg KA, Payne GW, Sato A, Medzhitov R, Segal SS, and Iwasaki A (2005) Innate control of adaptive immunity via remodeling of lymph node feed arteriole. Proceedings of the National Academy of Sciences of the United States of America, 102, 16315-16320.

31. Suda A, Yamashita M, M, Taguchi K, Vohr HW, Tsutsui N, Suzuki R, Kikuchi K, Sakaguchi K, Mochizuki K, and Nakamura K (2002) Local lymph node assay with non-radioisotope alternative endpoints. J Toxicol Sci, 27, 205-218.

32. US National Research Council (2007) Toxicity Testing in the Twenty-first Century: A Vision and a Strategy. Committee on Toxicity and Assessment of Environmental Agents, National Research Council, Washington, DC.

33. Van Och FM, Slob W, De Jong WH, Vandebriel RJ, and Van Loveren H (2000) A quantitative method for assessing the sensitizing potency of low molecular weight chemicals using a local lymph node assay: employment of a regression method that includes determination of the uncertainty margins. Toxicology, 146, 49-59.

34. Vocanson M, Saint-Mezard P, Tailhardat-Cluzel M, Benetiere J, Ducluzeau MT, Chavagnac C, Tedone R, Berard F, Kaiserlian D, and Nicolas JF (2005) CD8+T cells are effector cells of contact dermatitis to common skin allergens. Journal of Investigative Dermatology, 125, A39.

Application of a systems biology approach for skin allergy ...

Documents

Transcript of Application of a systems biology approach for skin allergy ...