“How to make sense of sensitization?” Autumn meeting of NVT section on Risk assessment

On 10th Ocbober, The NVT section on Risk assessment held a scientific meeting in ‘s Hertogenbosch.

In this meeting we focused on the state of affairs with regard to the implementation of alternative methods for one of the regulatory endpoints in toxicological safety testing of chemicals: sensitization. We discussed the testing strategy proposed by ECVAM, which is based on three alternative OECD-tests that cover three important key events in the OECD Adverse Outcome Pathway for sensitization: the direct peptide reactivity assay (DPRA) (OECD TG 442C), the KeratinoSensTM or LuSensTM assay (OECD TG 442D) and the human cell line activation test (h CLAT) (OECD TG 442E). What does this testing strategy look like and what is its regulatory status? Which are the important features of the proposed tests and what experience is available so far? What can be concluded regarding the validity of the results of the testing strategy?

The subject of this meeting is clearly relevant since the Dutch State Secretary for Economic Affairs, Van Dam, stated that the Netherlands should aim at becoming World leader in animal-free innovations by 2025. On April 8, 2016, he asked the Netherlands National Committee for the protection of animals used for scientific purposes (NCad) to draw up a schedule for phasing out animal procedures. According to NCad, the use of laboratory animals in regulatory safety testing of chemicals, food ingredients, pesticides and (veterinary) medicines could be phased out by 2025, whilst maintaining the existing safety level.

Three speakers introduced the subject and shared their experience with the approach.

Janine Ezendam (RIVM) presented the background of the regulatory need for non-animal approaches for the assessment of skin sensitisation hazard. She explained that allergic skin reactions are the most common form of immunotoxicity, with a prevalence of 19% in the EU. They are an important problem for worker and consumer health. Information on sensitisation is needed for GHS/CLP, REACH and cosmetics regulation. The ban on animal testing for cosmetics makes it necessary to search for alternatives; cosmetics are a common cause of contact allergy. Also potency information is required following a recent change in the REACH information requirements. Apart from legal and ethical reasons, a scientific rationale to move away from animals is that animal tests also have limitations. The well-known LLNA test generates a high number of false positives and the potency estimation is not very accurate.
For several steps (key events) in the adverse outcome pathway (AOP) for sensitisation, in vitro tests are available; several are already accepted in the OECD Test Guideline Program. The OECD is currently revising the guidelines to include new in vitro assays that address these key events. The Direct Peptide Reactivity Assay (DPRA) assesses the molecular initiating event (MIE): binding to peptide. The next key event, keratinocyte activation (KE2), can be assessed by various reporter cell lines (KeratinoSensTM, LuSensTM); another test system, SENS-IS, uses a reconstructed 3D skin model and measures gene expression. For KE3, dendritic cell activation, human cell based assays are available that are based on induction of activation markers or cytokines. Animal tests (LLNA and GPMT) are still indispensable for the final key events, T cell activation and proliferation leading to sensitisation.

Testing strategies available at the moment are based on combining information on MIE, KE2 and KE3. OECD has developed Guidance documents on such strategies (called Integrated/intelligent Testing Strategy, Defined Approach or Integrated Approaches to Testing and Assessment). OECD document no. 256 is specific for skin sensitisation; its annexes provide case studies on skin sensitisation as well as data sources. An IATA is flexible and starts with the problem formulation: what is the regulatory need? A defined approach is more fixed; it can be either a tiere approach (sequential), or integrated (multiple sources of data). Various approaches are being used, including the “2 out of 3 ITS” by BASF and the Bayesian Network ITS by P&G. An advantage of the latter is that it predicts potency (LLNA EC3) and quantifies uncertainty.

Conclusions from this presentation were that there are too many options for defined approaches, which hampers their regulatory use. Approaches that provide potency estimation are needed; in addition, it is important to develop tests further to reduce equivocal answers and to define their field of applicability. Both Cosmetics Europe as well as EURL ECVAM are working on further evaluation of the available strategies for skin sensitisation hazard assessment. EURL-ECVAM is developing performance-based test guideline. There will also be an OECD meeting in December on this subject.

Further reading: Ezendam J,· Braakhuis HM, Vandebriel · RJ. State of the art in non‑animal approaches for skin sensitization testing: from individual test methods towards testing strategies.
Arch Toxicol (2016) 90:2861–2883.

Walter Westerink (CRL) presented his experience with non-animal skin sensitization testing under REACH.
CRL uses the Direct Peptide Reactivity Assay (DPRA), KeratinoSensTM and U-SENSTM. The relation of these assays to the AOP was explained. DPRA quantifies the reactivity of test chemicals towards two model synthetic peptides containing either lysine or cysteine: SPCC and SPCL. The results for a number of proficiency chemicals are in line with the in vivo classification KeratinoSensTM assesses gene activation by Nrf2 which accumulates in the induced state of keratinocytes. Only if the induction in repeated tests is > 1.5-fold at a non-toxic concentration (< 1000 µM), the effect is considered positive. The effect is negative in case of negative results ≥ 1000 µM. Due to solubility issues a concentration of ≥ 1000 µM can often not be tested, resulting in a high amount of inconclusive results. U-SENSTM measures the activation of dendritic cells via a surface marker in different concentrations, and the result is positive if the induction > 150% at a non-toxic concentration.

The overall test strategy by CR consists of four steps:
- in silico assessment of metabolism and potency by DEREK
- KE2 assays: DPRA and KeratinoSensTM (or LuSensTM)
- depending on the previous steps: U-SENSTM
- if WoE shows that in vivo test is necessary to determine potency: LLNA

UVCBs are problematic (Chemical Substances of Unknown or Variable Composition, Complex Reaction Products or Biological Materials); they are often poorly soluble. Here, KeratinoSensTM is applied first, often followed by in vivo LLNA. In the experience of CRL, KeratinoSensTM may give false positive results due to activation of nuclear receptors. In addition, it is important to consider information on skin penetration.

In the discussion, Janine Ezendam advised to report problems with KeratinoSensTM to the OECD.

A question was put forward about the choice of the in silico method. DEREK was chosen because it was already available and seemed reliable and well-based. Other methods such as Toolbox are also available. It was remarked that the Human Cell Line Activation Test (h-CLAT), an alternative for U-SENSTM for Key Event 3, is bad in reproducibility.

In case of solubility problems, it was suggested to change the system (e.g. serum). This may be possible, but proficiency should be checked again.
It was suggested that a drop in viability could indicate an effect leading to apoptosis. This is possible, but the guideline specifies that it should not be taken into account.

David Hart (Akzo Nobel) presented his experience with in vitro skin sensitisation testing – the Benefits and limitations. The focus was on cosmetic ingredients.
He described the three approved (OECD Guideline) in vitro tests for skin sensitisation. These tests represent the three key events (KE) in the AOP, after absorption: DPRA (OECD TG 442C) for KE1; either KeratinoSensTM (OECD TG 442D) or LuSENSTM for KE2; the h-CLAT assay (OECD TG 442E) or U SENSTM (recently accepted in TG 442) for KE3. All methods have their limitations, e.g. there is no metabolic activity in DPRA and very little in KeratinoSensTM and h CLAT.

ECHA has recently published updated Guidance for sensitisation. In addition, a webinar was organised; it was concluded that for a clear negative result all three test outcomes should be negative and QSAR should indicate a lack of potential for sensitisation. The current ECHA Guidance has several limitations; in particular, it seems difficult to distinguish whether a substance is 1A or 1B (as required now) without animal testing, whereas ECHA also states that animal testing should only be the last resort.

Akzo Nobel has gained experience with the current test methods.
In the DPRA, shifting of HPLC peaks occurred for certain surfactants affecting the HPLC column, leading to an apparent depletion of lysine peptide. Depletion of peptide can also be caused by dimerization, which is not detected by HPLC. Other examples were given of situations that may to false positives. The DPRA can also be influenced by the solvent.

For a cosmetic ingredient, a variable result was found in different h-CLAT tests; the DPRA as well as the KeratinoSensTM assay were negative. How should the overall weight of evidence be interpreted? Using the two out of three method, one would conclude that the substance is not a skin sensitiser.

The EFfCI (European Cosmetic Ingredient Manufactures Trade Association) has tested 9 lipophilic ingrediënts; for 6 of these, the LLNA gave false positive responses compared to the GPMT. Substances with high LogKow (strongly lipophilic) clearly give problems. This project lead to several conclusions: In DPRA, solvents suitable for lipophilic substances may destabilise the peptides; KeratinoSensTM is simple, but does not have dose ranging; and h-CLAT gives variable results. A paper describing this work has just been published: Kreiling R et al. In chemico, in vitro and in vivo comparison of the skin sensitizing potential of eight unsaturated and one saturated lipid compounds. Regulatory Toxicology and Pharmacology 90 (2017) 262-276.

U-SENSTM may be an option to replace h-CLAT. SENS-IS is also promising, because the reconstructed skin has stratum corneum which could be suitable for substances with very low dermal penetration; however, it is not clear yet whether ECHA will accept results by this method.

It is very important to get more experience with the new test methods soon. The 2018 registrations with ECHA will include many smaller companies who will need more guidance from ECHA on how to integrate results from the available tests. Information on experience with these tests should be shared.

In the discussion, the bad reproducibility of h-CLAT was confirmed. It seems to come up often with positives. CR does not offer it any more.
It was suggested that DPRA could be improved by using mass spectrometry to avoid column problems; MS could detect dimerization and oxidation. It was agreed that improvement of DPRA would help to have more confidence about negative results.

General discussion
The speakers had put forward theses as starting point for the discussion.

One thesis was: There is no “one fits all” for Defined Approaches for skin sensitisation hazard assessment.
It was pointed out that the current in vitro methods are only indicators, or screening assays; SENS-IS seems promising if it is further developed. On the other hand, also the in vivo test is merely a model: for example, nickel is negative in the LLNA test.
As the OECD provides general descriptions of assays and industry gains practical experience, it seems important that these experiences feed back into the guidance development process.
Another thesis was: Reliable in vitro skin sensitising potency tests are required
It was agreed that most tests are still aimed at hazard classification. For potency, more advanced models will be required that include skin penetration, metabolism, etc. Organs on a chip could be a way forward.

The validation of these new test methods should not be considered finalised
It was agreed that in practice, they are often not sufficiently predictive for new classes of substances. Again, it was emphasised that sharing experience could guide the development of better methods. The question is how this could be achieved. Not all results are published, in particular proprietary information can be a problem. The dissemination of REACH dossiers may help, although not all information is publicly available. A way of sharing could be safe harbours like in pharma. Feedback to the OECD is necessary.

Producers have an interest in ensuring their customers (Unilever, P&G) that their substances are not sensitising. Dissemination of good practice in performing assays could be improved. Post-marketing surveillance also plays a role in detecting problems.

Sensitisation is first example of an effect for which an AOP can be used to define in vitro tests. Lessons should be learned from this process for other endpoints.