Introduction ·················································································

Each year millions of people worldwide use hair dye products. It is estimated that over one third of women aged 18 or over, and approximately ten percent of men aged 40 or over, in the United States and Europe, use at least one type of hair dye product (Huncharek and Kulpelnick 2005). Hair dyes can be separated into three classes: temporary, semi-permanent, and permanent. Permanent, or oxidative, hair dyes are the most widely used class of hair dyes, accounting for approximately 80% of the hair colouring product market in the US and EU (Corbett et al. 1999, Cosmetics Europe 2014). This class of hair dyes is different to the other two classes in respect to their composition: oxidative dyes require a chemical reaction between a primary intermediate and a coupler in order to generate the coloured dye on/in the hair (Nohynek et al. 2010). The primary intermediates are normally aryl diamine or aminophenol compounds substituted at either the ortho- or para- position, such as p- aminophenol. In contrast, couplers are normally aryl aminophenol or diphenols substituted at the meta- position, such as resorcinol. In the presence of a developer, such as hydrogen peroxide, the primary intermediate is oxidised and reacts with the coupler to produce a coloured aromatic dye (Nohynek et al. 2010) (Figure 5.1).

107

Figure 5.1: A suggested reaction pathway showing how the oxidative hair dye phenylenediamine (primary intermediate) results in a coloured dye in the presence of resorcinol (a coupler) and hydrogen peroxide (adapted from Nohynek et al 2010).

Typically, hair dye products contain between 0.05-2% primary intermediate, with the higher the percentage producing a darker shade of dye. In comparison, temporary and semi- permanent hair dyes are typically acidic or basic chemicals that bind to the proteins of hair and do not use developers or couplers. Typical classes of temporary and semi-permanent hair dyes include anthraquinones and nitroaminophenols respectively. The large number of people exposed to, and the reactivity of, hair dye products has led to them becoming some of the most widely studied cosmetic ingredients. A number of studies, both in vitro and in vivo, have raised concern about the carcinogenic potential of certain members of chemicals used within hair dye products (Baan et al. 2008; Freudenthal et al. 1999; Gago-Dominguez et al. 2001; IARC 2010; Skipper et al. 2010).

Previously, safety assessments for cosmetic ingredients, including hair dyes, would have been made, at least in part, using data from in vivo experimentation. However, significant changes in the European cosmetic and chemical legislations during the last decade have concentrated efforts in the development of alternative methods for safety testing purposes (EC 2003; EC 2007). The Adverse Outcome Pathway (AOP) paradigm has emerged as a

+ +H2O2

Indoaniline dye (green) Indoaniline dye (red) Leuco-dye (colourless)

108

promising approach in that it enables key events in the pathway that leads to a toxicological outcome to be identified (Ankley et al. 2010; Vinken 2013; Vinken et al. 2013a). Key amongst these events is the Molecular Initiating Event (MIE), which has been the focus for the development of in silico profilers (Przybylak and Schultz 2013). These profilers define the chemical features associated with a given MIE in terms of collections of structural alerts and are intended to be used to categorise chemicals based on a common MIE (Enoch et al. 2011a; Enoch et al. 2013b; Enoch and Roberts 2013; Przybylak and Schultz 2013; Sakuratani et al. 2013a; Sakuratani et al. 2013b; Vinken 2013; Vinken et al. 2013a) (discussed in more detail in Chapters 1, 3 and 4). The development of mechanism-based in silico profilers suitable for category formation is a time-consuming, literature-intensive process. Previous research leading to the establishment of in silico profilers for toxicological endpoints such as skin and respiratory sensitisation utilised a mechanistic hypothesis as a starting point for structural alert development (Enoch et al. 2008; Enoch et al. 2012b). However, for complex endpoints such as organ-specific toxicity for which knowledge relating to possible MIEs is lacking, a chemoinformatics approach, coupled with a posteri mechanistic rationalisation, has been shown to be successful (Hewitt et al. 2013). Given the complexity of potential mechanisms driving oral repeat dose toxicity, the current chapter employed the latter approach using the protocol described hereafter. The mechanism-based categories of chemicals that result from such AOP-derived profilers are applicable to predict hazard via read-across and, hence, assist in the filling of data gaps. In addition, these groupings also form the basis for the more in-depth analysis that is required for an overall risk assessment. In such a situation, additional testing using in vitro and/or in chemico methods to assess other key steps in the AOP is required. The ability to group chemicals into mechanism-based categories using in silico profilers enables in vitro and/or in chemico assays to be developed to enable the prioritisation of chemicals (Gutsell and Russell 2013).

In order to generate structural alerts and, thus, mechanism-based chemical categories information pertaining to the endpoint, and chemicals, of interest are required. With respect

109

to the work undertaken in this chapter, and as part of the wider goal of the COSMOS project (discussed in Chapter 1), information relating to the repeat dose toxicity of cosmetic ingredients is required. One available source of toxicological data associated with cosmetic ingredients are the ‘Opinion On’ reports published by the Scientific Committee on Consumer Safety (SCCS) and its predecessors, the Scientific Committee on Cosmetic products and Non-Food Products intended for consumers (SCCNFP) and the Scientific Committee on Consumer Products (SCCP). The reports are generated for cosmetic substances for which some concern exists with regards to human health (e.g. colourants, preservatives, UV-filters and hair dyes) and contain data for a variety of toxicological endpoints, such as; skin irritation, acute toxicity, carcinogenicity and (sub-)chronic repeat dose studies. These reports usually contain No Observable Adverse Effect level (NOAEL)- values, and Lowest Observable Adverse Effect Level (LOAEL)-values generated by the repeat dose studies. NOAEL and LOAEL values are determined upon the completion of various repeated dose toxicity studies, such as (sub-) chronic, developmental or reproductive toxicity (discussed in more detail in Chapter 1). These data, ideally the NOAEL, are used by the SCCS, within the ‘Opinion On’ reports, in order to calculate the margin of safety (Figure 2.3, Chapter 2). Clearly, such data could provide a useful starting point for developing MIEs and identifying the chemistry required for the grouping of chemicals for read-across.

In particular for hair dyes, high quality toxicological data became available as a consequence of the step-wise strategy of the European Commission to regulate all hair dyes listed as substances in cosmetic products. The trigger for this action was the major concern of the scientific community for a putative link between the use of hair dyes and the development of cancer, with a focus on leukaemia and bladder cancer (Gago-Dominguez et al. 2001, IARC, Baan 2008, Huncharek 2005, Nohynek 2004, Skipper 2010). As such, industry was required to submit safety dossiers for hair dye components and possible mixtures for evaluation by the Scientific Committee on Consumer Safety and its predecessors. Despite the requirement

110

to assess the toxicity of hair dyes, few in silico models or structural alerts for their toxic effects, or rationale for their grouping, are currently available.

Therefore, the aim of this chapter is to develop an in silico profiler from a retrospective analysis of oral repeat dose toxicity data, available for hair dyes, retrieved from the Scientific Committees ‘Opinions On’ reports published between 2000 and 2013. These data were used to group hair dyes based upon structural similarity, with subsequent mechanistic analysis being undertaken using information from the peer reviewed literature. This mechanistic information, relating these structural alerts to potential MIEs, is important as it provides evidence for the interaction between the chemical and the biological system. The profiler could, thus, be used for a variety of process including screening data sets to identify chemicals of concern or to prioritise those chemicals that should undergo in chemico/in vitro testing first.