Hello everyone, and welcome to SenzaGen's webcast. I'm Tina Lawesson, Head of Marketing and Communications here at SenzaGen, and I will be your host today. Today's talk focuses on the recent inclusion of GARD-skin in the OECD Test Guideline 497. We will begin with a 40-minute presentation followed by a Q&A session. Joining us are SenzaGen's toxicology and GARD experts, Chief Scientist Henrik Johansson, and Scientific Liaison Andy Fodderyd. Before we start, I also wanted to share a couple of notes. This webinar will be recorded, and the recording will soon be available on SenzaGen's website after the session. Regarding the Q&A session, you can submit your questions anytime during the presentation. To do so, please use the GoToWebinar panel and enter your question in the questions chat box. It's located on the right side of your screen.
Thank you for joining us today. We are now ready to begin this webcast featuring our first speaker, Henrik Johansson. Welcome, Henrik.
Thank you, Tina, and thank you all for joining this webcast where we will discuss, as Tina mentioned, the inclusion of our skin in OECD guideline 497. We have subtitled this presentation, "Extending the Applicability of Defined Approaches for Skin Sensitization." The agenda for this presentation will be, to start, a brief introduction to skin sensitization, and also, of course, a brief summary of the GARD technology. We will then move on to discuss the inclusion in 497, and specifically what it means for its utilization and its usage for different types of applicability domains. This will be illustrated by my colleague Andy Fodderyd in a couple of case studies as the last point of the agenda. As Tina mentioned, we will conclude with a live Q&A in the end.
I will start with an introduction to skin sensitization and GARD, and give a brief overview of the key technological features or the methods for those of you who might need a refresher or have not been in contact with GARD assays before. As I believe most of you know, the immunological mechanisms of skin sensitization are today quite well understood, and they have been mapped in what is called an adverse outcome pathway by the OECD. This defines four key events in the AOP, which leads to sensitization of exposed individuals. Here specifically, key event one, two, and three are of particular interest as they have been targeted for a series of in vitro methods to predict skin sensitizers among test chemicals.
In the 442 suite of the test guidelines, we have test guideline C, D, and E describing key event one, two, and three, respectively. There are today 10 methods in these test guidelines. Specifically, GARD-skin is in 442E, representing key event three. This is, we believe, a unique and mechanistically different method to monitor key event three compared to the other tests in the guideline. I will, of course, come back to that later on. If we look at the test guideline program in a broader sense, there's also a historical perspective, which is of relevance in these discussions. Traditionally, skin sensitization has primarily been assessed by in vivo methods, such as the local lymph node assay and the guinea pig assays.
As the 442 suite of test guidelines was introduced, they were primarily based on three methods, which are the DPRA, KeratinoSens, and the h-CLAT methods. These were also then combined most recently in the TG 497, which describes the defined approach for skin sensitization. There are specifically two concepts that are central. It's the hazard assessment and the CLP potency subcategorization assessment. Of course, the news with this webcast and the central theme of this presentation is that now there are more methods available in the battery that can form the defined approaches for these defined approaches. There is, of course, also GARD-skin. If we look at GARD-skin technology, the GARD technology platform, there is a central hypothesis behind the development and the concepts of the method.
That is that there should exist genes in dendritic cells or a model of a dendritic cell that are differentially expressed depending on if the stimuli, the test chemical, is a sensitizer or a non-sensitizer. If we can identify those genes, we can also then use them as predictive tools. That is precisely what has been done with the GARD-skin method development. We were able to identify 196 genes that are all potent predictors of sensitizing mechanism within the dendritic cell in vitro model. The concept of cell cultures is, of course, not unique. You can compare it, for example, with h-CLAT, but instead of measuring phenotypically two protein markers on the cell surface, we are dealing with more informative data in that sense. If we have a look at the identities of the 196 genes, they are, of course, all published for transparency.
Even though the key point to mention is that even though GARD-skin is considered a key event 3 method, as it is based on in vitro dendritic cells, it's a key understanding that also other immunological mechanisms typically involved with, for example, key event 2 are also monitored within these cells, as we are also measuring Nrf2 pathway, for example, and various metabolite genes associated with metabolisms of the test chemicals and thereby enabling protein reactivity. From a more practical point of view, the method is also quite straightforward. We are using an in vitro model of dendritic cells, which we refer to as SensaCell cell line. This is maintained in our cell laboratories and exposed to test chemicals of interest.
In our first series of assessments, we are monitoring the cytotoxic profiles and properties of test chemicals in order to identify a suitable GARD input concentration, which is a concentration at low to non-cytotoxic levels. Having established the GARD input concentration, exposure experiments are repeated, and total RNA is extracted from cell cultures, which have been exposed to the test chemicals. This genetic material is, of course, the starting point for doing gene expression analysis of this genetic material in order to quantify the expression levels of the 196 genes. Having access to that type of expression data and gene expression profiles, we are then able to use a support vector machine prediction algorithm in order to classify each test chemical as either a sensitizer or a non-sensitizer.
If we have a look particularly on the prediction algorithm, it's central to point out and understand that the algorithm is fixed and that all genes are contributing to the prediction models, but perhaps not at the same amount, but at fixed levels for each assessment. GARD-skin was approved for adoption into test guideline 442E in 2022. Since then, we have been working hard to characterize and expand the applicability domain of the method. You can see on the right in this slide, we have been working with hydrophobic compounds, a lot of fragrances and fragrance formulations, metal and metal salts, and complex mixtures, and so on. We believe it's key to continuously publish our findings together with the partners in the field to demonstrate the applicability of the method.
One such, of course, extending applicability is central also in the topic of today, where we will discuss centrally the inclusion of GARD-skin into the Test Guideline 497. Test Guideline 497 describes two different defined approaches, one called 2(3) Defined Approach, which is used for hazard identification of skin sensitizers. There are also two versions of an Integrated Testing Strategy, which is utilized in order to subcategorize test chemicals into the CLP categories, non-sensitizers or weak sensitizers, also called 1B, or strong sensitizers, which is referred to as 1A. In order to do so, the input data sources are generating scores, if you will, assigning test chemicals with a score from zero to three. The total score is used to give the final classification or subcategorization as 1A, 1B, or non-categorized chemicals.
In order to have GARD-skin included into the 497, there has been some key development performed. We have designed a borderline range in order to fit into the existing framework of the 2(3) Defined Approach. We have, of course, defined the scoring procedure to fit with the framework of the ITS Defined Approaches. Specifically looking at the 2(3) Defined Approach, the concept is based on combining three methods representing key event one, two, and three, respectively. You can do so in any order you choose. If two methods are concordant, then, of course, that would be the final output of the framework. One important point to keep in mind, however, is that assay limitations for each individual input data source are described in respective test guidelines.
There may occur instances for which you have one data source providing a sensitizing response and another data source providing a non-sensitizing response. For the third method, it may be the case that you have an incompatible applicability of the method in order to address that specific test chemicals. These are some of the areas where we believe that GARD-skin can improve on the usability of the defined approach. Andy will talk a lot more about that later on in this presentation. If we have a similar look at the ITS for potency subcategorization, here the concept is based on combining a key event 1 method with a key event 3 method and an in silico approach. In the original iteration of the 497 guidelines, the in silico methods were either the DEREK tool or the OECD toolbox.
Here in this slide, you can see the classification thresholds, and GARD is in the center of the table. You may note that here the continuous parameter that's being used to define the classification threshold is actually the GARD input concentration. The lower concentration used to obtain a positive classification in the GARD method, the stronger, and therefore also the stronger the sensitizing potential, and therefore also the higher the scoring output from the GARD into this ITS approach. You may note that assay limitations are described in the respective test guidelines, and they also carry forward into this defined approach. With that, I would like to hand over the word to my colleague Andy to talk a little bit more on how the recent inclusion of GARD-skin may, in fact, impact the usability of these defined approaches. Please go ahead, Andy.
Thank you very much, Henrik, for setting the stage for my part of the presentation. Before I start, I would also like to welcome everyone and thank everyone for joining today's webinar. I'm really looking forward to the Q&A session towards the end as well. The aim of my part of the presentation is to present a selection of different case studies describing how the GARD-skin assay can increase the predictive performance, but also to extend the applicability domain of the current defined approaches to enable the testing of substances that have traditionally been considered as difficult to test in the conventional assays. As Henrik also mentioned several times during his presentation, the challenges that are associated with the individual test guidelines also, of course, apply to when they are being used within the context of the defined approaches.
One of the most prominent limitations that prevents the use or at least prevents the testing of several of these materials in the defined approach is associated with the hydrophobicity threshold in the h-CLAT assay. It is not possible to generate conclusive results for test substances with a log P value above 3.5. At the same time, there is a lot of published data that indicates that the GARD-skin assay is capable of providing accurate results for this type of hydrophobic test material. The aim of the first case study was to investigate the effect of basically using the GARD-skin assay as a drop-in replacement for the h-CLAT in the context of the defined approach, and with the specific testing of very hydrophobic test materials.
In addition to that, we also wanted to investigate the effect of replacing the KeratinoSens assay with the new EpiSensA assay. EpiSensA assay is based on a reconstructed human epidermis model, and at least in theory, it's proposed to be applicable also for the testing of hydrophobic test materials. This slide showed the results from the testing. It's summarized in the table to the right. This is based on a published data set that we published approximately three years ago that basically investigated the performance of the GARD-skin assay for testing of hydrophobic test materials. We were able to collect data for a total of 21 substances with a log P value above 3.5. You can see the resulting data in the table here to the right.
The first column, you see the performance of the standard conventional defined approach using the h-CLAT as the information source. In the next column, you see basically the result from the same defined approach, but where you replace the h-CLAT with the GARD-skin. Next column showed the result from the same defined approach, but here again, you replaced also the KeratinoSens assay with the EpiSensA assay. Further to the right, you see the predicted performance of the GARD-skin assay when used as a standalone assay. I think one of the most important conclusions from this slide is, first of all, replacing the h-CLAT with the GARD-skin assay in the context of the 2(3) Defined Approach significantly increased the predicted performance for testing hydrophobic test materials when used in the context of the defined approach.
We can also see that further combining the GARD-skin with the EpiSensA assay instead of the KeratinoSens assay also improves the predicted performance of the defined approach. What I think is also really important to emphasize in this slide is that actually using the GARD-skin assay as a standalone assay actually provides higher predictive performance than using any of the combined approaches. We will come back to that later on in the presentation as well, because this is kind of important information not only in a regulatory context, but also when you select your specific model for use within the R&D project. What we also did was to investigate the same effect using exactly the same subset of hydrophobic test materials for deriving the CLP potency classification.
As Henrik already mentioned, the ITS is the defined approach that you can use in order to further subcategorize your test material as 1B and 1A materials. In the figure to the right, you see the performance of the conventional, the first column here, you see the performance of the conventional defined approach, where you use the h-CLAT as one of the information sources. In the column to the right, you then see the results from the potency prediction if you basically replace the h-CLAT with the GARD-skin. Also here, we can see that for the CLP potency classification, there is a significant improvement in predictive performance if you use the GARD-skin assay instead of the h-CLAT in the context of the defined approach.
Another important feature here is that for this specific potency predictions, you cannot use the EpiSensA assay in the context of the defined approaches, because here you need to use information from key event one together with key event two, and also combine that together with the in silico approach. Based on this data, we believe that it provides experimental evidence that replacing the h-CLAT assay with the GARD-skin assay in the context of testing hydrophobic materials significantly improves the predictive performance and also extends the applicability domain of the defined approaches. The next slide here is the next case study. It provides a very similar context as the previous case study. In this specific case, the project was performed together with our collaborators at Burleson Research Technologies and NICEATM in the United States.
The aim of this project was that NICEATM asked some of their collaborators and also different federal agencies in the United States to nominate different types of substances for testing in the defined approaches. The aim was to basically extend the applicability domain, or at least extend the number and identity of the chemicals that will be tested in the conventional OECD tests. In this specific case, there was a lot of very challenging test materials that were tested, spanning from pesticides, agrochemical formulations, dermal excipients, and also other types of substances that have traditionally been considered as difficult to test in the conventional OECD test. In terms of the GARD-skin data, there was testing performed for a total of 31 substances within the context of this project. The data was basically just published a few days ago at the NICEATM homepage.
You can see the reference to this specific publication in the bottom of this slide. This slide summarized the results from the testing. If you look at the figures to the right, you can see the performance metrics for the GARD-skin assay, which is here colored in blue color. You also see the performance of the h-CLAT assay, which is colored here in the gray color. You also see the different bars here represent further to the left the sensitivity. Then you have the specificity and also the balanced accuracy further to the right. If we look at the top of this figure first, this is basically the direct head-to-head comparison of the predicted performance between the GARD-skin in blue and the h-CLAT in gray.
Here we could initially conclude that the GARD-skin assay had both a higher sensitivity, specificity, and also balanced accuracy compared to the h-CLAT for this specific subset of test material. In the bottom of the slide, you can also see the performance of the defined approaches when you basically compare the standard defined approach in gray, where you basically have the h-CLAT assay as part of the testing strategy, compared to the performance of the GARD-skin assay and the defined approach, or the defined approach basically based on the GARD-skin assay. Here you can see again, we see absolutely the same trend as we did for the previous case study. Replacing the h-CLAT with the GARD-skin significantly increased or improved all the different predictive performance values for the defined approach.
Again, it's also an interesting observation from this data set that even though replacing h-CLAT with GARD increased the predictive performance of the defined approach, once again, we saw the same trend that using the GARD-skin assay as a standalone actually provides higher predictive performance than any of the other combinations within the context of the defined approach. The final case study I would like to highlight today is a case study or a collaboration together between Cortiva Agri Sciences and SenzaGen to evaluate the performance and applicability of using the GARD-skin assay not only for testing neat materials, but also to be able to test complex formulations, in this case, agrochemical formulations.
The rationale for conducting the testing in the GARD-skin in this specific case was based on a recent publication for which you basically can see the result in the table here to the right. This is from a publication that illustrates the performance of the different in vitro assays for testing agrochemical formulations. If you specifically look at the h-CLAT column here and you look at the specificity in the column in the top, you can see that only 12 or only 3 of the 15 non-sensitizers or expected non-sensitizers were correctly classified as non-sensitizers in the h-CLAT assay. It is a very high false positive rate with a specificity of only 20%. There was definitely a need to use a more or at least evaluate a more predictive assay for key event three in the adverse outcome pathway.
In order to evaluate the performance of the GARD-skin assay, we conducted testing together with Cortiva Agri Sciences of the same subset of materials that had also been tested in the h-CLAT assay. The testing could be performed basically following the OECD Test Guideline 442E. The only minor modification we had to do was to use a nominal molecular weight of the agrochemicals because these are complex mixtures and did not have a specific molecular weight. This slide basically summarizes the results from the testing. On top of this slide, where you see the performance metrics for the liquid-based formulations, these are the subset of chemicals that were directly comparable to the subset of chemicals tested in the h-CLAT assay.
Here you can see that there was a very good predictive performance, 81% accuracy with only one false negative classification, and in this case, five false positive classifications. If you look at the bottom of this slide, where you see the overall performance value, this is when we basically also extend the test data set to not only test liquid-based formulation, but also test solid-based formulation, which is even more challenging to test because of hydrophobicity limitations. You can see even though we extend the test data set, including these very challenging test materials, you can see that the performance values basically retain as well. You can also see that we actually used the GARD potency value as well in order to assign CLP potency classifications for the positive materials. Here you can see that 82% of the potency classifications were also correct in this space.
If we look a bit further into this table as well, because I briefly mentioned as well that five of the expected non-sensitizers were in this case classified as skin sensitizers in the GARD-skin assay. One important thing to point out here is that the reference data that we used for the comparison was mainly based on Local Lymph Node Assay data, because there's no human data available for these agrochemical formulations. If we look at the bottom of this table, where you have the negative or the expected negative formulations based on the Local Lymph Node Assay data, I would like you to specifically focus at the column that says the maximum stimulation index in the Local Lymph Node Assay. Here I just basically highlighted in yellow the misclassification of the so-called false positive classifications in the GARD-skin assay.
Here you can also see that the misclassifications or the false positives in the GARD-skin assay, they also had a very high stimulation index in the local lymph node assay, basically just below the classification threshold for inducing a positive classification. The difference in the classifications between the GARD-skin and the local lymph node assay is mainly attributable to differences in sensitivity between animal models and in vitro models, where the in vitro models are generally more sensitive. Here we can also say that for the three top formulations that were false positive, they also contain skin sensitizers in the formulation, but to a slightly lower level than could be detected in the local lymph node assay.
Overall, we could conclude that the GARD-skin assay is not only applicable for testing agrochemical active ingredients, but also provides a very high predictive performance for testing agrochemical formulations and formulated products. Before we conclude today's presentation, I would also like to briefly highlight a modification of the standard GARD-skin protocol, because in the context of the defined approaches, we have now talked about the potential to use the GARD-skin assay both for hazard identification, but also for CLP potency classifications. The GARD-skin dose response assay is essentially a modification of the standard validated GARD-skin protocol to incorporate dose response measurement to enable also to derive a continuous prediction of skin sensitizing potency. This slide illustrates the concept behind the GARD-skin dose response assay.
Instead of testing only at one single concentration, you perform testing at several different concentrations to generate a dose response curve. You use exactly the same prediction model to generate the decision value. The readout, as you can see in the figure to the right in this assay, is a CDV0 concentration, which essentially corresponds to the lowest concentration required to induce a positive classification in the standard GARD-skin. What I think is really important and interesting in this aspect is that you can basically use the CDV0 concentration in order to predict either human NESIL values or local lymph node assay AC3 value. This is something we have also published in peer-reviewed scientific journals. If you're interested to have a look at that, you have the reference on the bottom of this slide.
In the next slides, I would like to highlight as well on how this can actually be done in practice, how you can use the GARD-skin dose response assay to predict the skin sensitizing potency of an unknown test material. In this specific case, I used the chemical benzosinamate as a proof of concept. The first thing you do here is essentially to do the dose response testing, plot the decision values versus the input concentration, and then identify your CDV0 concentration. The second step is essentially just to use the experimentally derived CDV0 concentration as input into a predefined linear regression model in order to translate this concentration into a predicted human NESIL value with the unit micrograms per square centimeter of the tissue. It's a very simple process.
If we switch to the next slide, I would also show because the GARD-skin dose response assay has also been validated together with industry. We have today tested more than 200 chemicals using this approach. All of this data has also been published in three different peer-reviewed scientific publications. If you look to the right in the figure to the right, you can see an example or a subset of this data set for which there exists very well annotated reference data from both the Local Lymph Node Assay and the human repeated insult patch testing. In the figure to the right here, you see the predicted human NESIL values from the GARD-skin dose response assay on the X axis. On the Y axis, you have the human NESIL values directly derived from a human repeated insult patch testing.
If we look at this data, we can see that the predicted human NESIL values from the GARD-skin dose response assay correlated very well with available human reference data, with a Pearson correlation coefficient of 0.7 in this case. Another very interesting observation here is that if we actually compare the performance of the local lymph node assay to the human data for exactly the same subset of chemicals, we can see that the GARD-skin dose response assay is actually a more potent predictor of skin sensitizing potency in human compared to what the local lymph node assay is. We know based on this data that the GARD-skin dose response provides both accurate and reproducible potency predictions with a high correlation to both human NESIL values and local lymph node assay AC3 values.
We have also demonstrated in a recent publication on how you can use the predictive NESIL values in order to establish a maximum concentration of a sensitizer that can also be included into a formulated product. I put a reference also to this publication here. This is a joint publication between SenzaGen, International Flavors and Fragrances, and the Research Institute for Fragrance Materials. With that, before we open up for the Q&A session, I would like to just conclude some of the main findings in today's presentation. Both Henrik and I have illustrated and presented that the GARD assay is now officially included into the OECD Test Guideline 497 and can be used in order to fill some of the remaining data gaps and extend the applicability domain of the defined approaches.
I think it's really important that before any testing is actually committed, it is very important to define the testing strategy and make sure to select individual components of the defined approach to reflect the physicochemical properties of the test materials. We also demonstrated that using a combination of the GARD-skin and the EpiSensA assay may be a very suitable approach for testing very hydrophobic test materials that cannot be tested in the other in vitro assays. We have also seen that the GARD-skin assay does not only contribute with information to the hazard assessment, but you can also use it for the CLP potency assessment in the context of the integrated testing strategy in OECD Test Guideline 497.
I think also an important take-home message is when we look at the reference data from the hydrophobic test materials that even though we need to use a defined approach in order to obtain regulatory approval, we can also see that using the GARD-skin assay as a standalone approach should be considered as sufficient for non-regulatory purposes because the GARD-skin assay normally also outperforms the performance of the defined approaches. Finally, the last couple of slides I highlighted the potential to use a slightly modified version of the GARD-skin protocol, not only to do hazard identification or CLP potency classification, but also to do continuous prediction of skin sensitizing potency that can actually be incorporated and used to generate a quantitative risk assessment procedure.
By that, I would like to thank everyone for your attention and leave the word over to Tina again, and looking very much forward to the Q&A session. Thank you, Andy. Thank you very much to both of you, Henrik and Andy, for sharing insights and for this really interesting presentation. Before we move into the Q&A, we'd like to ask you just two quick poll questions. It will only take a minute or two, and we will share the results live on the screen in front of you. Your input really helps us to better understand who's joining us today and how we can make sessions like this more relevant and also more valuable to you. Thanks in advance for taking a moment to participate. You should be able to see the first question in front of you now.
Please go ahead and cast your vote. Here is the question. Have you experienced inconclusive results for challenging substances using a DA with DPRA, KeratinoSens, and h-CLAT? You have answer options like A, yes, frequently; B, occasionally; C, never; D, no experience with use animal testing. Please go ahead, cast your vote. The question is, have you experienced inconclusive results for challenging substances using a DA with DPRA, KeratinoSens, and h-CLAT? The answer options are A, yes, frequently; B, occasionally; C, never; or is it D, no experience with use animal testing? I will give it a couple more seconds for you to answer the question. I will now close the poll and share the results. Here you can see that we have 25% on the yes, frequently; 38% occasionally; 10% on never; and 28% related to no experience.
Thank you so much for sharing this with us. We will do question number two. Here is question number two. How do you usually test challenging substances, for example, low solubility or complex mixtures? Here you can select all that apply. A, outsource to specialized CRO; B, use in-house in vitro methods; C, rely on animal testing; D, I'm looking for better solutions; or E, not applicable, no experience. Please go ahead, share your votes. The question is again, how do you usually test challenging substances, those with low solubility or, for example, complex mixtures? Is it A, we outsource to specialized CROs; B, we use in-house in vitro methods; C, we rely on animal testing; D, I'm looking for better solutions; or is it E, not applicable for you? Please go ahead and cast your votes. Thank you so much.
I will close, and we should be able to see results right away. Here we go. In the top one, 50%, we outsource to specialized CROs. We use in-house in vitro methods. That goes for 11%. We rely on animal testing, 20%. Looking for better solutions, 32%. Not applicable goes for 16%. Thank you so much for doing this for us and answering the poll. We are now ready to—this is much appreciated, of course. We are now going to start the Q&A session. This is together with Dr. Andy Fodderyd and Dr. Henrik Johansson. I think they are getting ready for your questions. As a reminder, if you want to submit a question, please do so by using the questions chat box. Now, hi Henrik. We have Henrik on the camera. Welcome to you.
Thank you, Tina.
Let's see if we can also get Andy up on camera. I can see that we have questions coming in already. Welcome, Andy.
Thank you.
Okay, we are ready for some questions. Let's do the first one from the audience, directed to one of you, maybe both of you. What is the optimal role of GARD-skin within an ITS framework?
Good question. It depends. I think it depends on the chemicals that are tested. I think, you know, one thing that is important to emphasize is, which I also try to cover in my part of the presentation, is that how the defined approaches are actually set up. I mean, in the two out of three approach, you can use either method, right, key event one, two, or three, irrespectively in the order and similar. If you would like or need to generate also the CLP potency classification, you need to be a bit more careful because it directly excludes key event two, right? It needs to have data from key event one and key event three. I think, you know, it depends on what type of information you need and also the physiochemical properties of your test materials.
I mean, I think in general, for hydrophobic test materials, it could be a wise strategy to start doing the GARD-skin because of the demonstrated applicability for these types of materials. That can also provide you some guidance on what would be the next potential steps. If you need then to generate data into the ITS, I mean, either we need to do the key event one, I say DPRA or similar, or we need to expand beyond the current defined approach and instead see if we can use a weight of evidence approach where we basically combine the readout between the GARD-skin and EpiSensA, for example. We have a lot of experience in-house to develop that type of testing strategy.
I mean, if someone has a specific question or would like to discuss a specific chemical and the testing strategy, I mean, we will be able to discuss that as well on a case-to-case basis.
Yes, that's perfect. If you want to ask Andy questions like this, reach out by mail, right? We have your mail up on the screen at the moment as well. Thank you so much, Andy. Okay, we move on to the next question. It's a combination of two, I think. What makes the GARD-skin plus EpiSensA combo particularly effective for challenging test substances? A follow-up question is, is SenzaGen able to conduct EpiSensA in-house?
Do you want me to have a go, Andy? Yes, that's why I'm quiet. Thank you. The reasons may, of course, differ between the two methods. Speaking for GARD-skin, there are, of course, very well-proven and demonstrated applicability domains in these domains, which, of course, motivate the inclusion of GARD-skin in such a testing strategy if those are the main concerns with the test chemicals at hand. It has also been shown through data generated in the framework of the GARD-skin dose response method that GARD-skin is quite sensitive with a low limit of detection, specifically then compared to the LLNA method, which has been used traditionally for such substances, where we can show with lots of empirical data that GARD-skin is, on average, a thousand-fold more sensitive.
Those are two important aspects to consider when thinking about whether to use GARD-skin for those specific types of chemistries. When it comes to EpiSensA, I can say, first of all, that we do have the ability to perform EpiSensA, both at the SenzaGen laboratory in Lund and also at our partner laboratory, VitroScreen in Italy. Given the novelty of the method, we, of course, do not have the same range of experience as we have for GARD-skin.
At least in theory, it can be argued that the method is quite suitable for these types of chemistries, as it is based on a tissue model where you can directly apply test chemicals or test chemical formulations directly on the tissue, regardless of the physiological state or the level, the extent to which the formulation or test chemical has been solubilized in whatever vehicle you are using, depending on your chemistry. I would say those are perhaps different arguments for each of the two methods, but we do believe that they could make a big impact when testing these types of compounds in the industry in general. Do you want to add anything to that, Andy? No, I think you did a very good job, Henrik. Thank you. Thank you, Andy. Thank you.
Thank you, Henrik. Okay, so we have a lot of questions coming in, and some of them are pretty long, but I will see if I can, in one way or the other, read them, and we'll try to answer them. If it's easier for us, maybe for a big explanation or better guidance, we will save it and contact you directly, since I see that there are some really long questions as well. Let's see how it goes. Let's do the next one. Is the statement that the GARD-skin can be used as a standard on method for non-regulatory applications or screening, is it supported by the size of chemical sets, number of chemicals, mixture, et cetera, that have been tested? Please expand.
Yeah, I think this is a good question. Can I start, Henrik? And you can follow up then. I think, so this is an important thing to consider, right? Because, I mean, thinking in general of mixtures, there's a lot of different types of mixtures and formulations, and it's very difficult to test all and make a very general conclusion. For some types of mixtures and formulated products and similar, especially cosmetic products, fragrance materials, agrochemical formulations, we have a lot of data available. We have published part of the data. The rest of the data, we are working to publish. Yeah, I believe we have, for many types of applicability domains, a significant amount of data to conclude that it's definitely reliable for deriving this type of predictions.
One thing that I always think is important, because, you know, we have been working with the animal model for so long, the local lymph node assay or the guinea pig, but no one ever asked the same question for these methods. You only assume that they are applicable for the entire chemical space. The difference here is that we are actually generating the experimental data to actually support this statement. I think that is a very important distinction to distinguish between the way of looking at the animal models and also the in vitro tests. I don't know if you would like to add anything, Henrik. I could just add that, I mean, we all agree that more data is better. There's nothing controversial about it.
Even though we have many loads of data in-house, we cannot always use them to demonstrate or exemplify the applicability within a certain chemical space, most often because we do not have published reference values to compare with. That is also something that is related to what you just said, Andy, regarding the LLNA and other historical in vivo data points, that those are perhaps not always well curated. Thankfully, we do have some very well-curated data sets, not least the recent database, the Annex 2 for the OECD Test Guideline 497. There are lots of databases. We need to always keep in mind that we can generate data, but we need something to compare to. I believe that from a literature perspective, we are close to maximizing everything that we can test in order to demonstrate the applicability and prove it in literature.
Great. Thank you so much for a really good answer on this. I'm going to do a little shorter question related to a specific substance. Is the GARD test adapted to hydrolytically unstable substances?
This is a good question. This is something that we are currently working to address because, you know, we talk about the applicability domain and how we would like to demonstrate applicability. Chemicals that are known to hydrolyze extremely rapidly in aqueous media may be a potential problematic subgroup of chemicals. We are working on modified protocols to be able to address this. The initial data we have generated shows very promising results. Relating back to the previous question, I would like to make sure that we generate more data with this protocol before we can make any solid conclusions. I think one of the things that actually help here is to use an anhydrous vehicle when you do the actual testing. You first get the substance into solution into an anhydrous vehicle.
It is also, you know, dependent on the kinetics of the hydrolysis. Some of these substances work perfectly well in the test system. If you have a substance that hydrolyzes very quickly, you may end up in a problem. That is the protocol that we are now working to modify to see if we can improve the predictions.
Great. There is another question also related to this. Have you tested nanomaterial? What are the challenges? Do you want to comment on that too?
We have started testing a bit of nanomaterials, but mainly in the context of medical devices, because we have a modified version as well of the standard GARD-skin where we do testing of medical devices. Part of this protocol is essentially to do extract of solid materials and then test the extract, basically using the standard GARD-skin. Here we have experience with testing of nanomaterials. Here again, as Henrik mentioned, one of the major challenges with testing of the nanomaterials is that we do not have any good reference data that we can use in order to basically determine if the predictions are correct or not.
Thank you. Great. Sorry. Yes.
No. It's an area we are very interested to explore further. If anyone has a lot of experience in this area, please reach out to us as well. We can discuss it further.
Great. Yes. Okay, let's go back to questions related to the regulatory framework. Do you have experience on the regulatory acceptance by ICCA when the hazard conclusion is solely based on GARD-skin results?
That's a good question. I mean, we don't always see this data, but we know that there are dossiers available for which only GARD-skin has been used as a standalone approach. I'm not sure whether those reviews or whether those dossiers have been reviewed. I guess in general, you need to submit data for more than one key event. Henrik, do you have any more information? No. I can only call to mind some different scenarios where GARD-skin data has been the only new data created together with read-across data or old data in a weight of evidence approach. Again, I'm not sure what the outcome of the dossier submissions has been.
Okay, thanks for explaining. We will do the next question, a little longer one. Here's the scenario. For substances, hydrophobic or non-hydrophobic, that can be tested at or near the maximum recommended GARD-skin concentration, is it justifiable to use a positive result for potency determination in the ITS? It goes on. As only a single input concentration is used in the GARD-skin, then there is uncertainty as to whether the result would still have been positive at lower concentrations that contribute higher scores in the ITS. Is this specifically covered in the OECD 497?
To the best of my understanding, it is not. The guideline is, however you would like to see it, it is rigid. It's a framework that has been designed in order to be easy to use in a majority of the cases. For those types of uncertainty cases, there is, to the best of my knowledge, no writings in the guideline in order to how to approach those uncertainties. For the cases of the 2(3) Defined Approach, however, the guideline does recognize that uncertainty plays a role. This is the primary reason that each data source in the defined approach has been assigned a so-called borderline range where the output of the individual data sources must be considered inconclusive. For the CLP categorization, I don't think that is the case.
Thank you, Henrik. Okay, here's another question related to hydrophobic substances. It says, in the guideline, OECD 442E, only water and DMSO are given as appropriate vehicles. Based on experience, there might be limitations for a lot of hydrophobic substances. Thus, do you know of other valid, appropriate vehicles to use?
Yes, we have a long list of different solvents that can be used in the assay. It also says in the test guideline, I mean, DMSO and water were the solvents that were initially used when performing the validation study. But it also says in the test guideline that you can use other types of solvents if scientifically justified. We have a long list of different types of solvents spanning from the polar ones, saline and ethanol and similar, to the very non-polar solvents that can be used, including hexane and xylene, for example. That also relates to what we are now doing when we do the medical device testing, where we do extracts in both polar and non-polar extraction vehicle that are then tested using the conventional GARD-skin assay.
Here we actually do testing directly from substances in olive oil or sesame oil or, I mean, extremely non-polar solvents. There is a long list of different solvents that can be applied.
Great. Thank you so much. That's valuable information. We have a couple of questions also on the GARD-skin dose response assay. I would like to see if you want to comment on them. The first one is, could the GARD-skin dose response be used as part of the 2 out of 3 for potency prediction in a regulatory context? Do you want to clarify this, Andy or Henrik?
Sure. The GARD-skin dose response is essentially GARD-skin in a titrated range of concentrations. The top concentration on the dose curve is, in fact, the GARD input concentration. As long as you're obtaining three replicate values from three independent experiments according to the writings in 442E, you are, in fact, getting the hazard response from the conventional GARD-skin assay at the top of the dose response curve. Yes, if you design your strategy accordingly, you would be able to, for example, screen your candidate molecules during your developmental phases using GARD-skin dose response and carry forward the top concentration also to a regulatory framework such as the 2 out of 3 or, for that matter, the ITS or potency subcategorization.
Great. Thank you so much. We also have someone asking a question on where is it when it comes to the medical advice status. Do you want to comment on that? This is more related to ISO and the work that they are doing. Any updates on this?
There are updates. For those listeners or attendees that are familiar with the subject, the ISO group, the ISO Technical Committee 194 Working Group 8, which deals with skin irritation and skin sensitization, are continuously working on adapting existing OECD methods, in vitro methods, in order to fit into the framework also of the testing of medical devices. This is work that has been going on for a few years. There is of great value since last year, a published technical specification, which dictates how the qualification of OECD in vitro methods should be qualified also to fit into the ISO standard.
For the case of GARD-skin medical device, which is an adopted version of the conventional GARD-skin method, we have now completed the pre-validation phase with successful results, which means that we are continuously making progress not only in terms of the amount of generated data demonstrating the functionality of the method, but also in a regulatory sense. We believe that GARD-skin medical device will, in fact, be a regulatory approved ISO method or a method recommended within the ISO standards within the next few years.
Do you also want to comment on the current view of notified bodies when it comes to GARD-skin and maybe other in vitro methods? Do you know anything about this?
I think that maybe Andy knows more, but to the best of my knowledge, GARD-skin medical device data has successfully been used in order to get a CE mark in Europe through notified bodies. I do believe that the FDA in the U.S. is not quite ready to accept the same type of data quite yet. Andy, you may be able to elaborate. Yeah. That's absolutely true, Henrik. We have case studies where some of our clients have used the GARD-skin medical device approach as a standalone information source, submitted the data to notified bodies, and based on that data, also obtained CE marking in accordance with the medical device regulation in Europe. I mean, I don't think that's very controversial.
I think notified bodies in general are more open to use of in vitro models, specifically GARD-skin medical device in this aspect because it's applicable for testing in both polar and non-polar extraction vehicle, which is also a requirement according to ISO. Most of the other in vitro assays, the OECDs, are not applicable for the non-polar fraction. In the U.S., it's a bit more complicated. The FDA is, at least according to their statement, reviewing this type of application on a case-to-case basis. I'm not sure yet whether we have any successful case studies in the U.S. or not. I mean, it's important as well. I mean, it's a topic of another webinar, I believe. Medical device is also a bit complex.
Strictly following the recommendation in the ISO 10993 series of tests, specifically in part one and part two, you should always start doing in vitro tests before you conduct any type of animal testing anyhow. I mean, it could be useful to start doing in vitro testing before doing a Guinea pig test or similar. I mean, this type of test should not only use us as tick the box, but we need to generate in vitro data as well. If we have both in vitro data and animal data, it's obviously useful to submit those data points to the US FDA as well because that's a way to start for them to get familiar with the data as well.
Thank you so much. We have reached 5:00 P.M. today, so we're sort of running out of time. I will do the last question of today. Here it is. It is for maybe both of you to comment on. How do you recommend positioning GARD-skin within a defined approach that meets two out of three requirements? That would be the end of today's webcast, the last comment, so to speak.
I think that comes back very much to the first question as well in the Q&A session. I think, you know, it depends on the physicochemical properties of the test materials. I think irrespective of what type of chemicals you have, it might be wise doing key event one and key event three together with the in silico testing, if applicable, because that also makes sure that you can do the CLP potency classifications. I mean, in terms of the physicochemical properties of the test materials, if they're not applicable for testing in key event one assays or you cannot do any in silico predictions, for example, UVCBs or natural extract, I mean, we can also adapt the testing strategy to also derive the weight of evidence classifications for CLP using a combination of GARD-skin and EpiSensA.
That will not strictly follow the 497 test guideline, but ECHA also allowed a weight of evidence where you basically collect the scientific evidence. I think in some cases, this is actually the only way forward based on the physiochemical properties of the test material. That's something we can also help to discuss around and develop a specific testing strategy.
Thank you, Andy. Great end of this webcast. This was the last question of the day, and we have reached the end. If you have any further questions, please direct them to Henrik or Andy. You find their email addresses on the screen in front of you. Thank you all for joining today's webcast. Special thanks to our speakers, Henrik and Andy, for sharing your insights. We look forward to welcome you again later this year as we host our regular webinars. In the meantime, if you have a question, yes, I've said it already, please don't hesitate to reach out to us. That's all for today. For your information, the recorded version of this webcast will soon be available on our website and also sent to you by mail. That's all for today. Thank you once again, and have a wonderful day.
Thank you.