A framework for the safety evaluation of peptides in cosmetics

Highlights

• •A framework for the safety evaluation of peptides in cosmetics is presented.
• •A weight of evidence approach includes common risk assessment approaches coupled with bioinformatics and exposure considerations to ensure safety.
• •Recommended bioinformatic tools include BLASTp, ToxinPred3.0, Peptipedia, BIOPEP-UWM, AllerCatPro 2.0, and IEDB.
• •Peptides used for proof of concept with the bioinformatics are palmitoyl hexapeptide-12, caffeoyl hexapeptide-9, palmitoyl-pentapeptide-4, conotoxin ArlB, bradykinin, and enkephaline.

Abstract

As the cosmetic industry replaces traditional animal safety studies with next generation risk assessment approaches, the approach to safety substantiation for peptides used in cosmetic products must also evolve. While the need to provide assurances of safety for local and systemic toxicity endpoints remains the same, adoption of bioinformatic tools developed in the food, agricultural biotechnology, and drug development industries may add to the weight of evidence for the safety substantiation of peptides in cosmetics. Here we review the historical development and safety evaluation of peptides utilized in the cosmetic industry and provide a new safety evaluation framework that incorporates six bioinformatic tools. To test the framework, a variety of peptides (palmitoyl hexapeptide-12, caffeoyl hexapeptide-9, palmitoyl pentapeptide-4, amanitin alpha, conotoxin ArlB, bradykinin, and enkephaline) are evaluated with NCBI BLASTp, ToxinPred3.0, Peptipedia, BIOPEP-UWM, AllerCatPro 2.0, and IEDB bioinformatic tools. The results correctly identified safety concerns (toxins) for amanitin and conotoxin peptides and the biological actions of bradykinin and enkephaline, while palmitoyl hexapeptide-12, caffeoyl hexapeptide-9, and palmitoyl pentapeptide-4 demonstrated sequence homology with extracellular matrix proteins in the skin (collagen, elastin, fibronectin) without the safety concerns of the other peptides. The incorporation of bioinformatic tools into the safety framework provides an additional means to screen for toxins and allergens as well as insights into potential biological activities when sequence homology with existing proteins and peptides occurs. Further testing of the framework by the cosmetic industry is needed to lend support and reveal opportunities for refinements that advance the safety substantiation of peptides.

1. Introduction

The interest in using peptides in cosmetic formulations dates to 1973 when Pickart proposed the synthetic peptide Gly-His-Lys (GHK) as a signal peptide enhancing collagen production in vitro and acting as a carrier peptide when complexed with copper (Pickart and Thaler, 1973). Peptides were incorporated into cosmetic creams in the late 1980s as Lintner and Peschard (2000) described the development history of small, biologically active peptides and the obstacles needed to be overcome for utilizing them to provide cosmetic benefits. The utilization of peptides in cosmetics spans many continents and cultures. Traditional Chinese Medicine has found natural sources of small peptides in small animals (skin), in plants, in marine creatures, in acupunctured human skin and more. Lintner and co-workers (2016) explored the bridge between “evidence-based cosmetics” in the West and the holistic approaches of Traditional Chinese medicine. Peptides added to cosmetics tend to contain few amino acids (typically 2–8), are applied topically as ingredients in creams or sprays, exert their effects as mediators locally within the skin or hair, and have been safely marketed for almost 30 years.

Each ingredient added to a cosmetic product must have a purpose and function (e.g., skin conditioning agent) to balance any potential risk. Many have been skeptical of the benefits of topically applied peptides given their ppm concentrations, often poor skin permeation, and possible metabolism by peptidases in the skin. Research is conducted in vitro to investigate proposed mechanisms of action while clinical testing confirms the cosmetic benefit. Lintner and Peschard (2000) conducted a series of experiments to this end. A conjugated tripeptide fragment of collagen, palmitoyl-Gly-His-Lys (pal-GHK), increased collagen production by fibroblasts in vitro and in freshly collected skin from the abdomen ex vivo. A clinical study with 23 subjects then established the effectiveness of the peptide at 4 ppm in formula applied to the skin for 4 weeks by demonstrating an increase of skin thickness by about 4%. In a separate series of experiments, a dipeptide Tyr-Arg (YR) was made more lipophilic by attaching an N-terminal acetyl group to the tyrosine amino acid and a hexadecyl ester group to the arginine amino acid (N-acetyl-Tyr-Arg-hexadecylester) to decrease pain signals from heat and chemical irritant application to the skin. A small reduction in pain to a heat stimulus was observed at 2 h (and slightly less at 4 h) after application of the peptide at 300 ppm in a panel with 21 subjects. Similarly, nine subjects demonstrating reasonable sensitivity to capsaicin experienced reduced stinging when 300 ppm of the peptide was applied to the face 30 min prior to an additional amount of capsaicin. Both landmark placebo-controlled clinical studies show that topical application of the modified dipeptide reduces skin sensitivity to external factors such as heat and chemical irritants.

Research and development of cosmetic peptides have continued through the years and have been reviewed (Gorouhi and Maibach, 2009, Errante et al., 2020), including those with proven effectiveness in controlled in vivo studies (Schagen, 2017). Topically applied peptides in cosmetics are classified as signal peptides or matrikines (bioactive peptides derived from extracellular proteins such as collagen), carrier peptides, neurotransmitter inhibitor peptides, and enzyme inhibitor peptides (Lupo and Cole, 2007, Gorouhi and Maibach, 2009).

Matrikines target an increase in skin collagen, elastin, proteoglycan, glycosaminoglycans and fibronectin to reduce pigmentation from photodamage or to reduce the appearance of fine lines and wrinkles. Improved skin elasticity yields smoother and firmer skin. A summary of the clinical data of these matrikines is provided by Schagen, 2017 and Jariwala et al., 2022. These include hexapeptides such as Val-Gly-Val-Ala-Pro-Gly (VGVAPG), a fragment of elastin, Arg-Gly-Asp-Ser (RGDS), a tetrapeptide fragment of fibronectin, and Tyr-Ile-Gly-Ser-Arg (YIGSR), a tetrapeptide fragment of laminin (Lintner, 2002, Lintner, 2002). Lys-Thr-Thr-Lys-Ser (KTTKS), a fragment of procollagen I, has been demonstrated to stimulate collagen I, III, and fibronectin synthesis in fibroblasts in culture (Katayama et al., 1993). The KTTKS peptide was conjugated to palmitic acid to improve lipophilicity to permeate across the stratum corneum and has been shown in vivo to deliver facial skin benefits (Robinson et al., 2005). For instance, a clinical study tested 5 ppm pal-KTTKS in vehicle (N = 30) compared to vehicle alone (N = 30) over a 4-month period (Mas-Chamberlin et al., 2002). Skin replicas of the pal-KTTKS treatment showed a decrease in roughness, wrinkle volume, and wrinkle depth after 2 and 4 months of treatment. A smaller study with 16 subjects demonstrated that 3 ppm pal-KTTKS had similar benefits with skin replicas as well as increased skin thickness via echography measurements (Lintner et al., 2002). Confirmation of these cosmetic benefits was observed in a 12-week clinical trial with 93 subjects tested with a placebo moisturizer versus a moisturizer with 3 ppm pal-KTTKS, demonstrating that pal-KTTKS reduced the appearance of fine lines and wrinkles (Robinson et al., 2005). Carrier peptides are used to deliver trace elements like copper and manganese into the skin. These elements are thought to be important for skin repair and endogenous enzyme activity. Neurotransmitter inhibitor peptides imitate the amino acid sequence of the synaptic protein SNAP-25 which manages vesicle docking and fusion through the formation of the SNARE (Soluble N-ethylmaleimide-sensitive factor Attachment Receptor) complex to release acetylcholine into the synapse between muscle and nerve. A reduction in neurotransmitter activity would be expected to reduce muscle contraction and thus lessen the appearance of fine lines and wrinkles (Schagen, 2017). Acetyl hexapeptide-3 (Ac-Glu-Glu-Met-Gln-Arg-Arg-NH₂; EEMQRR), trade name Argireline, is an example of a neurotransmitter peptide. Lastly, enzyme inhibitor peptides directly or indirectly inhibit endogenous enzymes in the skin. These peptides inhibit enzymes like superoxide dismutase and proteinases and stimulate hyaluronan synthase. Table 1 provides a list of examples of topical peptides by class (Zhang and Falla, 2009, Schagen, 2017, Errante et al., 2020; Ferreira et al., 2020; Pintea et al., 2025).

Table 1.

Topically used peptides by class as reported by Zhang and Falla, 2009, Schagen, 2017, Errante et al., 2020, Ferreira et al., 2020 and Pintea et al., 2025.

Peptide Class	Peptide Name
Matrikine peptides	Carnosine, Copper tripeptide, Trifluoroacetyl-tripeptide-2, Tripeptide-10 citrulline, Acetyl tetrapeptide-5, Acetyl tetrapeptide-9, Acetyl tetrapeptide-11, Tetrapeptide PKEK, GEKG, Tetrapeptide-21, Hexapeptide, Hexapeptide-11 (also called Pentamide-6), Palmitoyl pentapeptide-4, Palmitoyl tripeptide-3/5, Palmitoyl tetrapeptide-7, Palmitoyl hexapeptide-12, Palmitoyl oligopeptide, Palmitoyl tripeptide-1, Pentamide-6, Lipospondin, SA1-III, Palmitoyl Tripeptide-8, Palmitoyl Tripeptide-38, Cyclotetrapeptide-24 Aminocyclohexane Carboxylate (also called Cyclopeptide-5), Elaidoyl Tripeptide-37, Heptapeptide, Tetrapeptide-30, Peptide PP, Peptide PCP, Pentapeptide-34, Nicotiana benthamiana Hexapeptide-40 SH-Oligopeptide-1, Nicotiana benthamiana Hexapeptide-40 SH-Polypeptide-5, Nicotiana benthamiana Hexapeptide-40 SH-Polypeptide-76, Nicotiana benthamiana SH-Polypeptide-45, Nicotiana benthamiana Hexapeptide-40 SH-Polypeptide-9, Nicotiana benthamiana SH-Polypeptide-7, Nicotiana benthamiana SH-Oligopeptide-2, Nicotiana benthamiana Hexapeptide-40 SH-Polypeptide 2, Nicotiana benthamiana SH-Polypeptide-15, Tripeptide-32, Hexapeptide-10, Palmitoyl Tripeptide-5
Carrier peptides	Copper tripeptide-1, Manganese tripeptide-1, Tripeptide-1, Diaminopropionoyl Tripeptide-33, Tripeptide-9 Citrulline
Neurotransmitter-inhibiting peptides or peptide mimetics	Acetyl hexapeptide-3, Pentapeptide-18, Pentapeptide-3, Tripeptide-3, Acetyl octapeptide 1/-3, Acetyl Dipeptide-1 Cetyl Ester, Acetyl Tetrapeptide-15, Acetyl Tetrapeptide-5, Dipeptide-2, Acetyl Octapeptide-3, Palmitoyl Tripeptide-8
Enzymatic inhibitor peptides	Soybean peptide, silk peptides (Silk fibroin peptide), rice peptides (Black rice oligopeptides)
Structural protein digestion	Keratin peptide
Unknown	Nicotiana benthamiana sh-polypeptide-15 hexapeptide-40, Pimpinella anisum extract (apiacea peptides), Pentapeptide-28

While these peptides demonstrate promising cosmetic benefits, their safety evaluation requires a tailored approach that accounts for their unique properties, which differ from both small molecules and biologics. This manuscript provides a safety assessment framework that adds bioinformatic tools to traditional risk assessment approaches, aligning with the principles of Next Generation Risk Assessment (NGRA) to ensure a human-relevant and hypothesis-driven approach for these ingredients without conducting animal studies (Luijten et al., 2020, Dent et al., 2021; SCCS Notes of Guidance, 12th revision, 2023). To test the framework, seven peptides are screened through six bioinformatic tools to determine if unique information can be added to aid in the safety evaluation of peptides.

2. Historical human safety evaluation of peptides in cosmetics

The traditional safety aspects of synthetic peptides created by chemically linked amino acids, and peptides derived from the hydrolysis of proteins have been evaluated by the Expert Panel for Cosmetic Ingredient Safety as well as other publications. The Cosmetic Ingredient Review (CIR) is an independent, nonprofit scientific body, established in 1976, that assesses the safety of cosmetic ingredients in the United States through the CIR program and procedures (Boyer et al., 2017). This group has reviewed several cosmetic peptides and hydrolyzed proteins (Table 2) and, except Triticum aestivum (wheat) peptide and hydrolyzed maple sycamore protein, all have been deemed “safe in cosmetics in the present practices of use and concentration described in this safety assessment”, with or without qualifications. Triticum aestivum (wheat) peptide and hydrolyzed maple sycamore protein did not have sufficient data to evaluate safety by the Expert Panel, likely because they were not in current use at the time of the review. These safety evaluations considered local toxicity endpoints (skin and eye irritation, skin sensitization) and systemic toxicity endpoints (genotoxicity, repeated-dose organ toxicity, developmental and reproductive toxicity) but did not include the added use of bioinformatic tools that we are now recommending, as described subsequently.

Table 2.

Peptides reviewed by the Expert Panel for Cosmetic Ingredient Safety.

Peptide INCI	CAS No. Amino Acid Sequence	Safety Conclusion	Reference
Acetyl hexapeptide-8 Also known as acetyl hexapeptide-8-amide. Retired synonyms are acetyl hexapeptide-3, acetyl hexapeptide-24, and acetyl hexapeptide-24 amide.	616204–22-9 Ac-Glu-Glu-Met-Gln-Arg-Arg-NH2 (EEMQRR)	Safe in present practices and uses when concentration does not exceed 0.005%	Cosmetic Ingredient Review website, CIR Findings, Ingredients https://cir-reports.cir-safety.org/ March 12, 2021
Bis(tripeptide-1) copper acetate	130120–57-9 Cu2+ 2[HOC(O)CH3][Gly-His-Lys]2- (copper acetate salt of GHK)	Safe in present practices and usesa	Johnson et al., 2018
Copper tripeptide-1	89030–95-5 Cu2+[Gly-His-Lys]2- (copper salt of GHK)	Safe in present practices and uses	Johnson et al., 2018
Glycine max (soybean) di-/tri-peptide and polypeptide fraction		Safe in present practices and uses	Burnett et al., 2023
Hexapeptide-12	92899–39-3 Val-Gly-Val-Ala-Pro-Gly (VGVAPG)	Safe in present practices and usesa	Johnson et al., 2018
Manganese tripeptide-1	611182–15-1 Mn2+[Gly-His-Lys]2- (manganese salt of GHK)	Safe in present practices and usesa	Johnson et al., 2018
Myristoyl hexapeptide-12	2276790–00-0 Myr-Val-Gly-Val-Ala-Pro-Gly (VGVAPG)	Safe in present practices and usesa	Johnson et al., 2018
Myristoyl tripeptide-1	No CAS number Myr-Gly-His-Lys (GHK)	Safe in present practices and usesa	Johnson et al., 2018
Palmitoyl hexapeptide-12	171263–26-6, 1992017–38-5 Pal-Val-Gly-Val-Ala-Pro-Gly (VGVAPG)	Safe in present practices and uses	Johnson et al., 2018
Palmitoyl oligopeptide (palmitoyl tripeptide-1 and palmitoyl hexapeptide-12)	171263–26-6 Gly-His-Lys (GHK) 147732–56-7 Val-Gly-Val-Ala-Pro-Gly (VGVAPG)	Safe in present practices and uses	Johnson et al., 2018
Palmitoyl tetrapeptide-7	221227–05-0 Pal-Gly-Gln-Pro-Arg (GQPR)	Safe in present practices and uses	Johnson et al., 2018
Palmitoyl tripeptide-1	147732–56-7 Pal-Gly-His-Lys (GHK)	Safe in present practices and uses	Johnson et al., 2018
Tripeptide-1	1269107–24-5 Gly-His-Lys (GHK)	Safe in present practices and uses	Johnson et al., 2018
Triticum aestivum (wheat) peptide		Insufficient data	Burnett et al., 2025
Myristoyl Pentapeptide-4	1392416–25-9 Myr- Lys-Thr-Thr-Lys-Ser or Myr-Lys-Thr-Ser-Lys-Ser (KTTKS or KTSKS)	Safe in present practices of use	Cosmetic Ingredient Review https://cir-reports.cir-safety.org/view-attachment/?id=d4bf93aa-abdb-7a7a-8b6f-795c2e499c89 Oct 1st, 2024
Palmitoyl Pentapeptide-4	521091- 64–5; 214047–00-4 Pal-Lys-Thr-Thr-Lys-Ser or Pal-Lys-Thr-Ser-Lys-Ser (KTTKS or KTSKS)	Safe in present practices of use	Cosmetic Ingredient Review https://cir-reports.cir-safety.org/view-attachment/?id=d4bf93aa-abdb-7a7a-8b6f-795c2e499c89 Oct 1st, 2024
Pentapeptide-4	149128–48-3 Lys-Thr-Thr-Lys-Ser or Lys-Thr-Ser-Lys-Ser (KTTKS or KTSKS)	Safe in present practices of use	Cosmetic Ingredient Review https://cir-reports.cir-safety.org/view-attachment/?id=d4bf93aa-abdb-7a7a-8b6f-795c2e499c89 Oct 1st, 2024
Hydrolyzed oat (Avena sativa) protein	151661-87-9	Safe in present practices and use when formulated to be non-sensitizing	Becker et al., 2019
Hydrolyzed wheat (Triticum aestivum) gluten protein	100684-25-1	Safe when formulated to restrict peptides to a weight-average MW of 3500 Da or less	Burnett et al., 2018
Hydrolyzed wheat (Triticum aestivum) protein	70084–87-6, 100209-50-5, 222400-28-4	Safe when formulated to restrict peptides to a weight-average MW of 3500 Da or less	Burnett et al., 2018
Hydrolyzed Actin	73049–73-7	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Amaranth (Amaranthus spp.) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Avocado (Persea americana) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Barley (Hordeum vulgare) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Brazil Nut (Bertholletia excelsa) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Collagen	73049–73-7 [generic to animal peptones], 92113–31-0	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Collagen Extract		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Cottonseed Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Elastin	100085–10-7, 73049–73-7 [generic to animal peptones], 91080–18-1	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Extensin	73049–73-7	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Fibronectin	100085–35-6, 73049–73-7 [generic to animal peptones]	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Gelatin	68410–45-7 [specific to enzymatic digest product]	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Hazelnut (Corylus avellana) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Hemp Seed (Cannabis sativa) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Jojoba (Simmondsia chinensis) Protein	100684-35-3	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Lupine (Lupinus spp.) Protein	73049–73-7	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Maple Sycamore (Acer pseudoplatanus) Protein	73049–73-7	Insufficient data	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Pea (Pisum sativum) Protein	222400-29-5, 227024-36-4	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Potato (Solanum tuberosum) Protein	169590-59-4	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Reticulin	73049–73-7 [generic to animal peptones], 99924–37-5	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Sesame (Sesamum indicum) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Soy (Glycine max) Protein	68607–88- 5 [generic to degree of hydrolyzation]	Safe in present practices of use	Burnett et al., 2023.
Hydrolyzed Soy (Glycine max) Protein Extract		Safe in present practices of use	Burnett et al., 2023.
Hydrolyzed Soymilk (Glycine max) Protein		Safe in present practices of use	Burnett et al., 2023.
Hydrolyzed Spongin (Demospongiae)		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Sweet Almond (Prunus amygdalus) Protein	100209-19-6	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Vegetable Protein	73049–73-7, 100209-45-8	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Hydrolyzed Zein (Zea mays) Protein		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
MEA (2-aminoethanol salts)-Hydrolyzed Collagen		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Zinc Hydrolyzed Collagen		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Ammonium Hydrolyzed Collagen	68951–88-2 [generic to ammonium hydrolyzed proteins]	Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.
Calcium Hydrolyzed Collagen		Safe in present practices of use	Burnett et al., 2022, Burnett et al., 2022.

^aAt the time of publishing the report, this ingredient was not reported to be in use. If used in the future, the expectation is that they will be used in product categories and at a concentration comparable to others in the group evaluated in the report.

Regarding the safe use of hydrolyzed proteins, sufficient hydrolysis is needed as exemplified by a case report of hydrolyzed wheat protein in cosmetics. Between March 2004 and September 2010, a bar soap containing incompletely hydrolyzed Triticum aestivum (wheat) protein in a facial soap triggered a large outbreak of wheat allergy in Japan (Nakamura et al., 2014, Nakamura et al., 2016). The first case of sensitization was reported in 2009 with periorbital edema and urticaria. A Special Committee of the Japanese Society of Allergology was created to investigate. A high molecular weight hydrolyzed wheat protein of 50 kDa was identified as the allergen and was present at 0.3% in the soap. Sensitization was ultimately eliminated by further hydrolysis to below 3.3 kDa. To date, hydrolyzed wheat ingredients are commonly used in cosmetics and considered safe if the average molecular weight of the peptides is below 3.5 kDa or has a length below approximately 32 amino acids (Scientific Committee, 2014, Burnett et al., 2018). The scientific support for hydrolysis to avoid Type I allergy is provided in Section 3.6 (Immunogenicity).

When the amino acid sequence is known (typically of synthetic peptides), the maximum reported concentration of use in finished cosmetic products was between 5 ppm and 50 ppm for those reviewed in Table 2, and the number of amino acids was less than 10. The data provided in the CIR reports demonstrate that minimal if any penetration through the skin occurs, no systemic toxicity or genotoxicity occurs, irritation to skin and eyes is none to slight, no skin sensitization occurs (Type IV delayed-type hypersensitivity), and the peptides are well tolerated in clinical studies ranging from 30 days to 7 months. If the peptides were to penetrate to a small extent through the skin and into the bloodstream, they would likely be rapidly metabolized and eliminated by aminopeptidases (Vlieghe et al., 2010). The metabolites that result from peptide catabolism are individual smaller peptides and amino acids, the latter which have also been shown to have a favorable safety profile in cosmetics (Burnett et al., 2013, Burnett et al., 2014).

Finally, the nature of peptides that have homology to peptides and proteins in the skin (or commonly ingested in the diet) and their individual amino acids, indicates that they are not foreign to the human body. Even with a palmitoyl group or other conjugation, they can often be considered endogenous. This is important when considering systemic toxicity. If a peptide does enter the systemic circulation by traveling through the stratum corneum, epidermis, and dermis without being metabolized or bound to other biomolecules, then the body would likely recognize them in the same way as fragments of proteins absorbed by the small intestine after a meal. The familiarity of small peptides is therefore not biologically new which makes systemic toxicity less of a concern.

3. Methods: A framework for the safety evaluation of peptides in cosmetics

A safety assessment framework is provided that maintains traditional cosmetic risk assessment approaches (SCCS, 2023) and expands them to include the introduction of bioinformatic tools when evaluating the safety of peptides in cosmetic products. The approach taken for this framework is based on a published evaluation of protein safety in the context of agricultural biotechnology (Delaney et al., 2008) as well as the principles governing the potential allergenicity of amino acid segments used in allergen database searches (WHO, 2016). A weight-of-evidence approach to safety follows a disciplined step-by-step evaluation of relevant aspects of peptide characteristics, combined with existing methods for local and systemic toxicity endpoints, and introduces bioinformatic tools utilizing the amino acid sequence of the peptide. A unique benefit of these tools is to rule out amino acid sequence homology with known toxins, immunogens, and peptides that could have activity outside of the skin or hair. Six of these bioinformatic tools have been chosen for evaluation with seven test peptides to determine if they can identify unique aspects of safety that would benefit the safety evaluation process.

The framework is designed to be compatible with the principles of Next Generation Risk Assessment (NGRA), which emphasizes human-relevant, exposure-led, hypothesis-driven risk assessment designed to prevent harm. Key elements of NGRA such as starting with identifying the use scenario and collecting information, understanding the mode of action to enable a hypothesis-driven assessment, as well as the use of in silico tools (including bioinformatic tools) and in vitro assays are integrated throughout the workflow to reduce reliance on traditional animal testing and to provide a more mechanistic understanding of potential risks based on exposure (SCCS Notes of Guidance, 12th revision, 2023).

When possible, in vitro testing should utilize OECD methods and the SCCS guidelines to provide for reliability, consistency in testing, and regulatory acceptance. In general, Good In Vitro Methods Practices as captured in OECD GD 286, 2018 and Good Laboratory Practices laid down in 87/18/EEC (1986) should be adhered to.

The framework presented in Fig. 1 and described in the following sections begins by characterizing the peptide in terms of the amino acid sequence and physical–chemical properties that could influence, for example, dermal penetration. Next in the framework is to rule out potent biological activity often associated with toxins, immunogens or biological activities outside of the skin or hair using a bioinformatics approach. The potential for systemic exposure is then addressed, and if the peptide is proven not to penetrate through the skin into the systemic circulation (i.e., because of metabolism in the skin), systemic toxicity risk is considered negligible. If the peptide can reach the systemic circulation and is stable, then systemic toxicity endpoints need to be considered. The more traditional endpoints of genotoxicity/carcinogenicity, immunogenicity, skin and eye irritation are also considered. Clinical studies designed to evaluate skin tolerability of the peptide-containing formulation can be combined with those evaluating cosmetic benefits or conducted as safety-in-use consumer studies under the supervision of a dermatologist. Lastly, as with all cosmetic products, a post-market surveillance program is used to identify in-market adverse events and to comply with all agency reporting requirements.

Fig. 1.

A proposed framework for safety evaluation of peptides in cosmetics.

3.1. Peptide characterization

A peptide is a polymer composed of less than or equal to 40 amino acids (approximately 4.4  kDa), regardless of how it is made (Zane et al., 2020). Understanding the size of the peptide is important to distinguish it from a protein and to dimension the potential of Type I allergy. Characterization of the peptide includes understanding the amino acid sequence and the physical and chemical properties of the peptide. Letter codes may be used for the primary structure of the cosmetic peptide, the 3-letter amino acid codes for the natural amino acids (IUPAC, 1984). Single-letter amino acid codes are also acceptable. For example, the INCI name tripeptide-1 (CAS 49557-75-7) is a three amino acid peptide containing glycine, histidine, and lysine. The code for this peptide is Gly-His-Lys or GHK.

Historically, the molecular weight of cosmetic peptides has been less than 500 Da (approximately 5 amino acids), likely to improve penetration into the skin. As well, the octanol/water partition coefficient, Log P, has been between 1 and 3, the melting point below 200°C, water solubility greater than 1 mg/mL, and there have been few or no polar centers in peptides (Gorouhi and Maibach, 2009). Beyond these physicochemical measures, additional characterization includes defining any modifications to traditional L-amino acid bonds to form the peptide. In some cases, modifications are made to improve stability. D-amino acids, which are rare in nature, are substituted for L-amino acids, or glycosylation is used to reduce susceptibility to proteases (Lien and Lowman, 2003). To improve penetration of peptides into the skin, conjugation with fatty acyl groups at the N-terminus of the peptide is often utilized (Gorouhi and Maibach, 2009, Johnson et al., 2018). Impurities should be noted so they can be evaluated for safety and can include structural components of the synthetic peptide sequence, process reagents, residual solvents, elemental impurities, counter ions such as acetic acid, and other impurities (Colalto, 2024). Incorrect peptides (peptides with incomplete or incorrect amino acid sequence) should be below analytical detection limits of chromatographic and mass spectrometric techniques or evaluated independently for safety. Peptide purity and formulation stability data are also part of peptide characterization.

3.2. Biological activity and potency: Bioinformatics and in vitro testing

The biological activity of peptides is attributed to the amino acid composition, the amino acid sequence, the length of the peptide and the chemical modification with reactive groups. In vitro cell lines derived from mammalian cells such as keratinocytes and fibroblasts can be used to predict specific aspects of in vivo activity and to assess quantitatively the relative sensitivity of various species (including human) to a peptide (Harmonised Tripartite Guideline, 2011). Pharmacological aspects, such as high-affinity receptor binding and pharmacological mechanisms of action would not be suitable for peptides being considered for cosmetic formulations.

Gupta et al., (2015) describe several in vitro assays that can be utilized to assess cytotoxicity potential, including LDH leakage, MTT reduction assay, ATP production, and red blood cell hemolysis. A list of common techniques to reduce toxicity is also provided (Gupta et al., 2015) with examples that include use of D-amino acids, end modifications such as C-amidation or N-acetylation, and lipidation to increase hydrophobicity.

The addition of bioinformatic tools to traditional safety evaluation methods may prove to be useful for the assessment of peptide and protein toxicity by the cosmetic industry. Simple bioinformatic analyses screen the peptide or protein sequence to identify potential similarities with, for example, known protein allergens or toxins. After reviewing numerous approaches, the following six tools were chosen based on their ease of use and utility demonstrated in other non-cosmetic industries. These were then evaluated with a group of seven peptides to determine their utility in contributing to the weight of evidence in the safety evaluation of peptides for use in cosmetics.

NCBI BLASTp: NCBI provides tools like BLASTp (Basic Local Alignment Search Tool) (https://blast.ncbi.nlm.nih.gov/Blast.cgi), which allows users to identify the exact peptide in a protein within a selected database (Altschul et al., 1990, Altschul et al., 1994, Altschul et al., 1997). This search tool allows users to query short peptide sequences and retrieve proteins containing that exact amino acid segment. Using the UniProtKB/Swiss-Prot database, the results provide the UniProtID with a link to manually curated high quality information and the protein sequence in FASTA format. The result page also provides an overview of the organisms in which the sequence was found, and a filter can be applied to limit the results to a particular species (e.g., Homo sapiens). The UniProtID can provide information on the peptide origin (species information, synthetic origin), location of the peptide in a protein sequence, its function, related literature (https://www.uniprot.org/) and can help to assure that the peptide of interest does not exhibit high potency by acting as agonists or antagonists at receptors where peptides serve as native ligands (Hruby 2002).

ToxinPred3.0: This tool is a widely used in silico method developed to predict and discriminate toxic or non-toxic peptides based on their amino acid sequences (Gupta et al., 2013; Gupta et al., 2015; Rathore 2024). The model was initially developed using toxic peptides with 35 or fewer residues sourced from various databases, while non-toxic peptides were obtained from SwissProt and TrEMBL. It was observed that certain residues like Cys, His, Asn, and Pro are abundant as well as preferred at various positions in toxic peptides. Val, Thr, Arg, Gln, Met, Leu, Lys, Ile, Phe, and Ala were dominant in non-toxic peptides. The latest version, ToxinPred3.0, incorporates advanced machine learning and deep learning techniques to enhance prediction accuracy. It employs a hybrid approach combining motif-based analysis with machine learning models to achieve high specificity and sensitivity. The web server for ToxinPred3.0 is accessible at https://webs.iiitd.edu.in/raghava/toxinpred3/. More advanced models, such as tAMPer (Ebrahimikondori et al., 2024), integrate not only the amino acid sequence composition but also the predicted three-dimensional structure of peptides to enhance the accuracy of toxicity prediction.

BIOPEP-UWM: This comprehensive database includes bioactive peptides (https://biochemia.uwm.edu.pl/biopep-uwm/; Minkiewicz et al., 2019). While this database is leveraged for orally ingested proteins and peptides, it can also serve as a screening tool for topically applied peptides, especially any peptide that could also be ingested (e.g., lipsticks). Information provided can add to the weight of evidence for the possible activity of the peptide of interest.

Peptipedia v2.0: This tool (https://app.peptipedia.cl/) offers the most comprehensive peptide sequence database with information on a wide range of biological activities, including cosmetic and dermatological activities, molecular binding, and anti-aging properties (Quiroz et al., 2021, Cabas-Mora et al., 2024). Additional bioinformatics tools and databases that can also help the safety evaluation by predicting peptide biological activity and structural characteristics are PeptideRanker (Mooney et al., 2012), which utilizes neural networks to predict biological activity, PepCalc (PepCalc.com − Peptide calculator), which analyzes fundamental properties such as molecular weight and isoelectric point, and TopicalPdb, which summarized over 600 experimentally verified topically delivered peptides (Mathur et al., 2018).

AllerCatPro 2.0: Predicting the allergenicity of proteins is a critical aspect of safety evaluation and several bioinformatics tools have been developed in recent years to address this challenge. AllerCatPro 2.0 (https://allercatpro.bii.a-star.edu.sg/; Maurer-Stroh et al., 2019, Nguyen et al., 2022) distinguishes itself by integrating an extensive protein allergen dataset with 3D structural analysis derived from WHO/IUIS, COMPARE, FARRP, UniProtKB, and Allergome, along with added capabilities such as the identification of potential cross-reactive protein allergens, proteins associated with autoimmune disease and proteins with low allergenic potential (Nguyen et al., 2022). AllerCatPro was developed for screening proteins and protein hydrolysates. If the peptide was derived from an endogenous protein (e.g., collagen), the whole protein sequence can be evaluated for allergenicity (Krutz et al., 2023).

IEDB: The Immune Epitope Database (IEDB) provides extensive experimental data on antibody and T cell responses from humans and other species, which can be queried for information on a peptide or protein of interest (https://www.iedb.org; Vita et al., 2025). It can also be used to predict whether a peptide is a potential T cell epitope by predicting its binding affinity to a wide range of HLA class I and II alleles covering > 97% and > 99% of the general population and by using the prediction method NetMHCpan 4.1 BA for class I and NetMHCIIpan 4.0 BA for class II (Reynisson et al., 2020). For these prediction tools, peptides lengths of at least nine amino acids are required for class I, and 15 amino acids for class II. This database and its tools can provide valuable insights into the potential immunogenicity of peptides.

3.3. Bioavailability and systemic exposure

A distinguishing feature of peptides in cosmetic formulations is that they are applied topically versus many peptides used in therapeutics that are administered parenterally, intramuscularly, or subcutaneously due to poor bioavailability when administered by other routes (Zane et al., 2020). The applied concentration of synthetic peptides in cosmetic products reviewed by CIR (Table 2) had maximum reported concentration of use between 5 and 50 ppm (Johnson et al., 2018). Permeation coefficients for some topical peptides and amino acids are provided in the systematic review by Gorouhi and Maibach (2009) and range from 8.63 x 10^-10 cm/s for GSH to 2.72 x 10^-6 cm/s for histidine-Cu. The lack of significant systemic exposure from topically applied peptides is supported by in vitro skin penetration studies. For example, Lintner and Peschard (2000) examined carnosine (β-Ala-His; AH) penetration through the skin with and without conjugation to a palmitic acid group. Standard Franz diffusion cells were used to study the diffusion and penetration of peptides labeled with radioactive iodine in human skin over a 6-hour incubation period. A known amount of peptide was applied to the surface of the skin followed by analyzing the radioactivity distributed into the various skin layers (stratum corneum, epidermis/dermis) and the amount of peptide recovered in the receptor fluid of the Franz cell (transcutaneous flow). Despite its small size, this 2-amino acid peptide did not penetrate beyond the first layer of the stratum corneum. The addition of the palmitoyl group to the terminal amino group facilitated penetration into the skin but not through the skin. Neither peptide was found to have significant transcutaneous penetration. The authors emphasized the importance of this finding, that neither peptide would be expected to reach the bloodstream or lymphatic fluids, thus, there is no risk of systemic toxicity. In a similar fashion, scientists at the US FDA (Kraeling et al., 2015) determined penetration through the skin of acetyl hexapeptide-8 (Ac-EEMQRR-amide) from an oil in water cosmetic formulation using human cadaver skin. Acetyl hexapeptide-8 was applied to a 0.64 cm² area of cadaver skin at 222.4 ± 5.4 μg. After 24 h of incubation, the skin was gently washed, the stratum corneum was tape-stripped 13 times and the epidermis and dermis were heat-separated. The results of four replicates demonstrated that 99.7% of the peptide was recovered in the skin washes, with 0.22% penetrating to the stratum corneum, 0.01% penetrating to the viable epidermis, and none of the peptide was recovered in the dermis or receptor fluid.

Choi and co-workers (2014) evaluated the permeability and stability of collagen pentapeptide (Lys-Thr-Thr-Lys-Ser; KTTKS) with and without conjugation to a palmitoyl group in Franz diffusion cell chambers with hairless mouse skin. KTTKS or pal-KTTKS (100 ppm in 15% ethanol solution) was applied to the skin at a loading dose of 50 μg test solution/cm² for 24 h. While KTTKS did not penetrate the skin, pal-KTTKS was observed in the stratum corneum (4.2 μg/cm²), epidermis (2.8 μg/cm²), and dermis (0.3 μ g/cm²), with no peptide present in the receptor fluid. Both KTTKS and pal-KTTKS were rapidly degraded in the skin by proteases, with pal-KTTKS being more stable than KTTKS.

The importance of understanding dermal penetration when evaluating peptide safety cannot be overstated. Testing methods have improved significantly since the early days of peptide discovery. Evaluation of dermal penetration of new cosmetic peptides should be assessed in vitro using Franz cell chambers with human skin following OECD TG 428, 2004 test guidelines along with the SCCS “Basic Criteria” (SCCS/, 2010) as outlined in the SCCS Notes of Guidance, 12th revision to ensure reliability of the results. This includes applying relevant amounts of finished formulations to the skin based on the intended consumer cosmetic application. Freshly collected metabolically active skin is preferred over cadaver skin. Inter- and intra-individual variability is addressed in the Basic Criteria by using at least 8 skin samples from at least 4 donors. While the test system can be modified to represent occlusive cosmetic applications (e.g., deodorants), there are no standard protocols for damaged skin. Thus, testing on intact skin is recommended. When estimating dermal absorption, the amount of target chemical in the dermis, epidermis (without stratum corneum) and the receptor fluid will be considered as dermally absorbed and considered for further calculations. In cases where substances with very low dermal absorption and limited permeation occur, the epidermis may be excluded from the calculations when it is clearly demonstrated through scientifically justified methods that no movement of the chemicals from the skin reservoir to the receptor fluid occurs (Yourick et al., 2004). This is an important consideration for dermally applied peptides when considering whether systemic toxicity endpoints are relevant based on situations where the peptide does not reach the systemic circulation. 3, 3.1, 3.2, 3.3, 3.4, 3.5, 3.6, 3.7, 3.8, 3.9, 4, 5, 6, 7 of the SCCS Notes of Guidance, 12th revision (Committee, 2023) notes that in cases where a cosmetic ingredient with a very low dermal absorption, some studies can be waived since systemic exposure via dermal absorption is expected to be minimal. In such a case, the following minimum set of data should be made available to assess the safety of the cosmetic ingredients with very low bioavailability: experimentally determined physiochemical data, local toxicity, mutagenicity/genotoxicity, and high quality in vitro dermal absorption study according to the SCCS Basic Criteria.

Metabolism can also play a significant role in the bioavailability of topically applied peptides. Peptides are metabolized by endopeptidases, then further degraded to amino acids by exopeptidases. Due to the ubiquitous availability of proteases and peptidases throughout the body, proteolytic degradation of many peptides is rapid and not limited to organs typically associated with drug elimination, such as the liver (US Department of Health and Human Services, 2023). Because proteins, peptides, and amino acids are metabolized in the skin (Gibbs et al., 2007, Dorner et al., 2025), the lack of movement from the skin reservoir to the receptor fluid is possible for peptides that show very low dermal absorption and limited permeation. These metabolic pathways are understood and often do not require metabolism experiments (ICH Harmonized Tripartite Guideline, S6(R1) 2011). If the lack of systemic exposure is proven, repeated-dose toxicity studies, including developmental and reproductive toxicity studies, are not warranted unless there is a special situation based on unique biological or chemical properties that warrant further investigation.

3.4. Systemic toxicity endpoints

There are currently no validated new approach methodologies (NAMs) to replace animal testing for systemic toxicity endpoints. Thus, understanding systemic exposure (or lack thereof) and utilizing non-animal methods (including in silico tools) becomes the basis for due diligence. The threshold of toxicological concern (TTC) refers to the establishment of a human exposure threshold value for all chemicals, based on consideration of their chemical structures, below which there would be no appreciable risk to human health (Kroes et al., 2004, Kroes et al., 2007, Yang et al., 2017). Proteins were excluded from this initial approach over concerns related to food allergies, hypersensitivity and intolerances. Regardless, similar TTC principles apply to peptide safety. De minimis exposure to the systemic circulation to amino acids and peptides that are endogenous to the body and are rapidly catabolized and eliminated pose little systemic toxicity potential, including developmental and reproductive toxicity endpoints. Because of this, a large dataset of repeated-dose studies with cosmetic peptides does not exist and would not lend itself to development of TTC thresholds, as this is not deemed necessary. Food allergies, hypersensitivity and intolerances still need to be considered. The immunogenicity potential is addressed in the section that follows and information from the bioinformatic tools suggested may add significant value for evaluating this endpoint. If a peptide can reach the systemic circulation, then organ toxicity, developmental and reproductive toxicity endpoints must be assessed.

A history of safe use in the diet of the protein sources that are hydrolyzed is also considered when assessing the potential for systemic toxicity. Peptides synthesized from common amino acids with standard peptide bonds found as a normal part of mammalian systems and/or found in significant quantities in human food sources are of low concern. Peptides that have been altered for greater skin permeation or stability following traditional well-accepted approaches, some of which have been discussed already, are also of low concern from a systemic toxicity perspective. For example, there are many examples of increasing the lipophilicity of a peptide by conjugating it to a palmitoyl group. While this may enhance penetration into the skin, the examples discussed did not show significant exposure to the receptor fluid in vitro, and palmitic acid can be synthesized from carbohydrates and amino acids in the body by de novo lipogenesis.

3.5. Genotoxicity and carcinogenicity

Validated in vitro methods that assess mutagenicity, clastogenicity and aneugenicity include the Bacterial Reverse Mutation Test (OECD TG 471, 2020) and the In Vitro Micronucleus Test (OECD TG 487) using mammalian cells. The Bacterial Reverse Mutation Test (Ames assay) assesses the potential for mutagenicity. One must be cautious when considering genotoxicity testing of peptides as the presence of histidine or tryptophan resulting from the degradation of amino acid chains can result in false positives and confounding bacterial mutation assay results. Positive Ames assay results should be interrogated for the possibility of histidine or tryptophan interference (Thybaud et al., 2016, Zane et al., 2020, Colalto, 2024). For peptides that contain or can release histidine or tryptophan, a modified Ames method could be used, wherein the bacteria incubated with the peptide would be washed to remove free amino acids before plating (Thybaud et al., 2016). The In Vitro Micronucleus test assesses both clastogenicity and aneugenicity. The In Vitro Mammalian Chromosome Aberration Test (OECD TG 473) can be utilized to assess clastogenicity but is not designed to measure aneugenicity. These protocols are conducted with and without exogenous sources of metabolic activation and employ positive controls to assure the assays are performing as intended. Test conditions and interpretation of results can be found in the OECD Test Guidelines. The 3D skin COMET (Pfuhler et al., 2020) and Micronucleus (Pfuhler et al., 2021) assays are sufficiently validated to move towards individual OECD Test Guidelines, but an independent peer review of the validation study is still needed (Committee, 2023). In vivo carcinogenicity test data is not needed nor acceptable for cosmetic ingredients given current testing bans.

Testing peptides for mutagenicity, clastogenicity and aneugenicity is largely confirmatory in nature. It is not expected that peptides would interact directly with DNA or other chromosomal material. Due to their poor cellular permeability, peptides are not efficient in interacting with cytoplasmic and nuclear targets (Yang and Hinner, 2015).

3.6. Local toxicity endpoints

Local tolerance to skin and eyes should be evaluated with exposures relevant to the specific cosmetic application. Cosmetic peptides tend to be well tolerated in skin and eye irritation test methods (Johnson et al., 2018). Any irritation observed in peptide-containing products is usually caused by the solvent, vehicle, or other formulation components. Therefore, skin and eye irritation assessments should focus on finished product testing to accurately evaluate the peptide-containing product. To demonstrate the absence of irritation potential of formulations containing peptides, in vitro cytotoxicity tests (cell-based or using reconstructed human epidermis (RhE) models) and/or 48-hour human patch testing can be considered. It is important to note that there are currently no validated or regulatory accepted skin irritation methods specifically designed for peptides as isolated ingredients. Nevertheless, the Reconstructed Human Epidermis (RhE) Test Method (OECD TG 439) can be used to assess cosmetic formulations for skin irritation potential and the Bovine Cornea Opacity Permeability (BCOP) Test (OECD TG 437), Reconstructed Human Cornea-like Epithelium Test (OECD TG 492), and the Short Time Exposure Test (OECD TG 491) can be considered for assessing cosmetic formulations for eye irritation potential. Additional eye irritation methods are provided in the SCCS Notes of Guidance (Committee, 2023) as well as an Integrated Approach for Testing and Assessment for identification of serious eye damage and eye irritation (OECD GD 263) and a Defined Approach (OECD TG 467) to provide information on potential eye hazard effects on the whole range of classifications required by the UN GHS.

Historically, delayed contact hypersensitivity (Type IV allergy) was evaluated using animal studies (guinea pig maximization test and Buehler guinea pig test (OECD TG 406, 2022) and the murine local lymph node assay (OECD TG 429, 2010)). While the results of these studies can be relied upon only if completed prior to the animal testing bans, animal studies are no longer accepted for ethical reasons moving forward. Currently, there are no in silico, in chemico or in vitro methods validated with peptides. Nevertheless, palmitoyl pentapeptide-4 (link to report provided in Table 2) was tested in a Direct Peptide Reactivity Assay using the OECD 442C protocol and in an antioxidant/electrophile response element (ARE)-Nrf2 luciferase assay in accordance with OECD 442D. Both studies predicted palmitoyl pentapeptide-4 to be non-sensitizing. The non-sensitizing nature of this peptide was supported by a historical guinea pig maximization test and with favorable human repeat insult patch tests (HRIPTs). Tests with human volunteers should only be confirmatory in nature (not for hazard identification) and considered scientifically and ethically necessary, when the toxicological profiles of the components are available and no concern is raised (Committee, 2023). Predictive testing for skin sensitization is not considered ethical since the tests carry the risk of inducing a long-lasting or permanent immunological sensitization in the individual (SCCS/, 2015). In situations where a confirmatory HRIPT is scientifically and ethically necessary, the Draize or Jordan & King protocols are followed (McNamee et al., 2008). Many of the peptides in Table 2 have been supported by HRIPTs to confirm a lack of skin sensitization.

Delayed contact hypersensitivity (Type IV allergy) is unlikely to occur following topical exposure to peptides. A hallmark for Type IV allergy is the activation of naïve T cells and the subsequent differentiation to the Th1 cell subset. This is dependent on electrophilic chemicals that covalently bind to amino acid residues in skin proteins or peptides on Human Leukocyte Antigen (HLA) class I molecules to form haptens. Importantly, peptides in cosmetics are typically not electrophilic and are not inherently reactive. As a result, they do not possess the chemical properties required to bind covalently to skin proteins or other peptides to form haptens under physiological conditions. Furthermore, these peptides differ fundamentally from the endogenous peptides presented by HLA class I molecules. HLA class I molecules are characterized by a closed binding groove and therefore require an exact peptide size for successful binding and presentation of the T cell epitope to T cells. The Immune Epitope Database (IEDB) hosts HLA class I-binding prediction tools. Peptides with at least 9 amino acids are required to predict whether a peptide has the potential to bind to HLA class I molecules. However, peptides utilized in cosmetics are different from HLA class I presented peptides and are not designed to fit these strict requirements. Thus, these peptides are not expected to be presented by HLA class I molecules in a manner that would trigger an immune response.

3.7. Immunogenicity

It is well established that certain chemicals and proteins have the potential to cause immediate Type I hypersensitivity under various exposure conditions. Of relevance to cosmetics, some proteins can be Type I allergens, including those derived from foods known to be allergenic in some individuals (e.g., milk, egg, fish, crustacea, tree nuts, peanuts, wheat, soybean, sesame). While the relevant route of exposure to food allergens is oral ingestion, cosmetic Type I allergens can induce sensitization from inhalation or via skin and mucosal exposures. Therefore, knowing if the protein source is a known allergen found in food is important.

Immediate Type I hypersensitivity is an immunoglobulin E (IgE)-mediated immune response. Specific IgE antibodies are produced during the sensitization phase against a typically harmless antigen such as proteins in pollen, food or insect venom. When such a protein allergen is encountered, antigen-presenting cells process and present peptides derived from the allergen on HLA class II molecules to naïve T cells. Such peptides therefore need to be able to bind to HLA class II molecules, be recognized by T cell receptors, activate T cells to proliferate, and direct the differentiation of T cells to a Th2 cell subset. Th2 cells can activate B cells to produce antigen-specific IgE antibodies. The sensitization phase typically requires frequent, repeated exposures over months or more (Troyano et al., 2011). Then, once the individual is sensitized towards the allergen, subsequent exposures to the same (or similar) allergen can lead to an immediate elicitation of an allergic response. IgE antibodies bind to receptors on basophils and mast cells and, upon exposure to the same (or cross-reactive) antigen, cross-link IgE antibodies to trigger degranulation and the release of histamine and other mediators during the elicitation phase. Symptoms can range from mild nausea to urticaria (hives), angioedema (swelling), acute bronchospasm, rhinoconjunctivitis, tachycardia, hypotension, and in rare instances severe life-threatening anaphylactic shock (e.g., to food allergens).

In terms of T cell activation, the peptide-binding groove of HLA class II molecules is open-ended, allowing peptides of variable length between 12 and 25 amino acids (approximately 1.3–2.7 kDa). The core binding region within these HLA class II-binding peptides is often a nonamer (9 amino acids) that contains key anchor residues essential for binding to the HLA molecule and T cell receptor recognition. Thus, peptides of less than 12 amino acids can be considered rather unlikely to successfully bind to HLA class II molecules and trigger a significant T cell activation unless the sequence has been identified as immunogenic epitope (for example, by the suggested IEDB bioinformatic tool).

The tools available to the safety framework to prevent Type I allergy include: (i.) avoidance of proteins associated with known food allergens, (ii.) hydrolyzing proteins to less than 2.5 kDa or 20–30 amino acids, (iii.) conducting a quantitative risk assessment to assure any proteins larger than 2.5 kDa are below the threshold for inducing Type I allergy, and (iv.) leveraging the IEDB and AllerCatPro 2.0 screening to identify homologies with existing allergens.

For proteins to be Type I allergens, they must be of sufficient size to bridge two IgE receptors on immune cells (Huby et al., 2000, Troyano et al., 2011). Thus, if the protein molecular weight is less than 2.5 kDa or 20–30 amino acids in size, it will not pose a Type I allergy threat. This principle is best illustrated with case reports of Type I allergy that occurred with insufficient hydrolysis of wheat protein in cosmetics described earlier (Varjonene et al., 2000, Sugiura and Sugiura, 2012, Kobayashi et al., 2015, Burnett et al., 2018).

For proteins that cannot be fully hydrolyzed below 2.5 kDa (20–30 amino acids), a quantitative risk assessment approach using exposure models and potency thresholds is taken (Basketter et al., 2010, Troyano et al., 2011). One approach is to confirm analytically the extent of hydrolysis of a protein-containing ingredient. For example, SDS-PAGE methods can be used to provide information on the molecular weight distribution of the protein composition and the estimated concentration of protein fragments above a certain size threshold (e.g., 2.5 or 3.5 kDa) in the ingredient. The amount of peptide fragments above the cutoff in the ingredient of interest can be then used to calculate consumer exposure and compare the exposure against an appropriate benchmark. One typical benchmark for incidental inhalation exposure (e.g., shower exposure model) to personal care products is < 0.1 ng protein/m³ for non-enzyme proteins (Troyano et al., 2011). For risk assessment purposes, all proteins and enzymes are considered allergens unless there is evidence to support the absence of allergenic potential (e.g., complete hydrolysis or a quantitative risk assessment below an accepted threshold of exposure). Most synthetic cosmetic peptides are less than 10 amino acids (well below the 20–30 amino acid length needed to bridge two IgE antibodies), so do not pose a Type I allergy concern.

In addition to delayed-type contact hypersensitivity (Type IV) and immediate hypersensitivity (Type I), non-IgE mediated pseudo-allergic reactions can occur that lead to mast cell degranulation triggered by peptides containing basic amino acids (Kumar et al., 2021). The symptoms are identical to anaphylaxis (skin flushing, headache, edema, hypotension, urticaria, and bronchospasm) following mast cell release of inflammatory and pro-inflammatory mediators. The receptor responsible for this non-IgE mediated mast cell activation is Mas-related G-protein coupled receptor member X2 (MRGPRX2). More than 20 ligands (agonists or antagonists) have already been identified for binding to MRGPRX2 (Kumar et al., 2021). Agonists that lead to mast cell degranulation include cationic amphiphilic drugs, insect venom chemical components, antimicrobial peptides, secreted eosinophil products, neuropeptides, small compounds, and natural compounds. Peptide agonists that bind MRGPRX2 include cortistatin-14, substance P, somatostatin, vasoactive intestinal peptide, pituitary adenylate cyclase-activating polypeptide, cathelicidins, β-defensins, indolicidin, angiogenic peptide 30/5C, AMPs derived from insulin-like growth factor binding protein 5, proadrenomedullin N-terminal peptides, mastoparan, eosinophil granule proteins, proteases, and gonadotropin-releasing hormone receptor agonist and antagonists. It is worth noting that peptide MRGPRX2 agonists are rich in hydrophobic and cationic amino acids such as proline, phenylalanine, tryptophan, and arginine/lysine (Subramanian et al., 2011). Lu et al., 2021 were able to test the binding potential of small peptides in MRGPRX2 expressing HEK-293 (HEK-X2) and LAD2 human mast cells. Results showed that MRGPRX2 can be affected by peptides as small as 5 amino acids in length. Thus, peptides rich in proline, phenylalanine, tryptophan, and arginine/lysine could be screened in a similar receptor binding assay to dimension the risk of psuedoallergic reactions. We are not aware of any peptides that have been used in cosmetics that have been associated with psuedoallergic reactions. Regardless, the understanding of this condition will likely continue to develop and should be followed.

As there is no conclusive test that will predict a likely human immunoglobulin E (IgE) response to a peptide or protein applied topically, an important first step involves undertaking a comparison of the amino acid sequence with those of established allergens. The present framework follows a similar approach taken for genetically modified enzymes in food products (WHO, 2016). The objective of this is to detect both global similarities and short contiguous amino acid sequences that may represent linear IgE epitopes (FAO/WHO, 2001). The Committee evaluating the potential allergenicity of genetically modified enzymes in foods recommends that database searches should consider only eight amino acid sequences (WHO, 2016). The FDA and additional research also confirm peptide products that are less than eight amino acids are not expected to be immunogenic (Food and Drug Administration, 2023, Rudensky et al., 1991, Rammensee et al., 1995, Saha and Raghava, 2006). Peptides of eight or more amino acids that are homologous to known allergens should be avoided in cosmetic formulations. Because sequence homology searches are limited to the sequence of known allergens in publicly available databases and the scientific literature, a post-market surveillance system is included as a second line of defense to assure cosmetics formulated with peptides are well tolerated. The IEDB and AllerCatPro 2.0 bioinformatic tools are incorporated into the safety framework for adding supportive evidence on immunogenicity potential.

3.8. Clinical experience

Clinical studies designed to evaluate skin and eye tolerability of the peptide-containing formulation can be combined with those evaluating cosmetic benefits. These studies typically range from 4 to 12 weeks. Numerous clinical studies have demonstrated the favorable skin and eye compatibility of cosmetics that contain peptides when the products are used as intended. Reviews of several clinical trials with peptides are provided by Gorouhi and Maibach, 2010 and by Schagen, 2017. Safety-in-use testing in consumers that are instructed to use the finished product following market labeling instructions for 2–4 weeks with dermatologist oversight is another method to assure the product is well tolerated (e.g., minimal skin and eye irritation).

3.9. Post-market surveillance

With the passage of the Modernization of Cosmetic Regulation Act in the United States (US Congress, 2022), manufacturers are required to have a post-market surveillance system that collect all adverse events reported from marketed products and report any serious adverse events to the FDA. Cosmetics containing peptides are no different and are not expected to have an increased incidence of adverse effects relative to the wide array of cosmetic products already marketed. While the pre-market safety assessment process gives a satisfactory assurance of safety, post-market surveillance provides a second level of defense for identifying adverse events across the larger and more genetically diverse population.

4. Peptides for proof of principle and illustrative examples for bioinformatic screening

The number of databases and bioinformatic tools for peptide evaluation continues to proliferate (Minkiewicz et al., 2019), providing toxicologists with a wide palette to choose from. For practical purposes, the six most comprehensive databases and tools (NCBI BLASTp, ToxinPred3.0, Peptipedia, BIOPEP-UWM, AllerCatPro 2.0, IEDB) were queried with a set of seven peptides (Table 3). Three peptides were peptides designed for cosmetic formulations (palmitoyl hexapeptide-12, caffeoyl hexapeptide-9, and palmitoyl pentapeptide-4), two peptides were toxins (amanitin alpha and conotoxin ArlB), and two peptides were endogenous peptides used in therapeutic applications (bradykinin and enkephaline). Table 3 provides the amino acid sequences for these peptides. Only the peptide sequence was queried in the tools. Existing safety data for palmitoyl hexapeptide-12 and palmitoyl pentapeptide-4 from the CIR reviews are included in the results.

Table 3.

Summary of seven peptide examples and their results using bioinformatic tools (NCBI BLASTp, ToxiPred3.0, Peptipedia, BIOPEP-UWM, AllerCatPro 2.0, IEDB) for proof of principle and illustration.

Peptide sequences	NCBI BLASTp (database: UniProtKB/Swiss-Prot)	ToxinPred 3.0	Peptipedia	BIOPEP-UWM	AllerCatPro 2.0	IEDB
Palmitoyl hexapeptide-12 Pal-VGVAPG	Found in 76 organisms, including Homo sapiens: Elastin (UniProtID P15502.4)	BLAST Search: Non-Toxin	ID#3751: anti-aging, chemotactic	ID# 3803: Chemotactic activity, chemotactic domain of elastin	Hom s Elastin (UniProtID P15502) is associated with autoimmune disease	ID #68811: non– immunodominant peptide in elastin, but found positive in 2 B cell and 1 T cell assay related to disease exacerbation
Caffeoyl hexapeptide-9 Caff-GPQGPQ	Found in 61 organisms, including Homo sapiens: Collagen alpha-1 (UniProtID Q8IZC6.1)	BLAST Search: Non-Toxin	Similar to e.g., RGDGPQGPQ ID#3564: anti-aging	No matches	Weak similarity to Bos d alpha2l, an uncommon allergen in Bos taurus	No hits
Palmitoyl pentapeptide-4 Pal-KTTKS	Found in 75 organisms and various proteins in Homo sapiens	BLAST Search: Non-Toxin	ID# 3504: anti-aging	No matches	Not applicable	No hits
Amanitin alpha IWGIGCNP	Found in 79 organisms, but not in Homo sapiens. Best hit: Alpha-amanitin in Amanita fuliginea (UniProtID P85421.2)	BLAST Search on UniProtID P85421: Toxin	ID# 73985: nutraceutical, toxin, venom peptide	No matches	No hits for UniProt ID P85421	No hits
Conotoxin ArlB DECCSNPACR	Found in 52 organisms, but not in Homo sapiens. Best hit: Alpha-conotoxin ArIA in Conus arenatus (UniProtID P0C8R2.1)	BLAST Search on UniProtID P0C8R2: Toxin	No matches	No matches	No hits for P0C8R2	No hits
Bradykinin RPPGFSPFR	Found in 52 organisms, including Homo sapiens: Kininogen-1 (UniProtID P01042.2) but it is identical to Bradykinin in Trichonephila clavipes (UniProtID P0DM76.1)	BLAST Search: Non-Toxin	ID# 6706: antibacterial, antifungal, hormone, therapeutic, etc.	Confirmed as bradykinin (ID# 3340)	No hits for P01042.2 or P0DM76	ID# 548816: peptide has been studied for immune reactivity
Enkephaline YGGFL	Found in 50 organisms and various proteins in Homo sapiens	BLAST Search: Non-Toxin	ID# 48028: antioxidative, hormone, therapeutic, etc.	ID# 2879, 8122, 9521: opioid antagonist, antioxidant, peptidase inhibitor	Not applicable	ID# 910542: peptide has been studied for immune reactivity

5. Results

Palmitoyl hexapeptide-12 (VGVAPG) is the palmitic acid ester of hexapeptide-12 which promotes collagen, elastin and hyaluronic acid production from fibroblasts. Characterization of the chemistry, physical and chemical properties, method of manufacture, and composition and impurities are provided in Johnson et al., 2018. An NCBI BLASTp search for the sequence using the UniProtKB/Swiss-Prot database, results in many hits across 76 organisms, including Elastin (UniProtID P15502.4) in Homo sapiens, which is a key component of the extracellular matrix in skin. ToxinPred3.0 was queried and predicted the peptide to be a non-toxin. It exhibits anti-aging and chemotactic activities (Peptipedia, BIOPEP-UWM), attracting skin fibroblasts and monocytes, which are important for skin repair and regeneration. Querying the peptide in the IEDB database, the sequence was identified as motif in human elastin (UniProt ID P15502, position 530–535). Human elastin protein has been associated with autoimmune diseases based on the AllerCatPro 2.0 results. However, the peptide showed positive results in few B cell and T cell assays only related to disease exacerbation (IEDB), indicating potential immune reactivity only in certain contexts and providing further evidence that this peptide is not an immunodominant epitope in the protein elastin. Favorable toxicology data are provided in a CIR report (Johnson et al., 2018). Ocular irritation in New Zealand white rabbits concluded the peptide mixture was a nonirritant. In vitro the peptide was tested for eye irritation in the hen’s egg chorioallantoic membrane assay (practically non-irritating) and in the Statens Seruminstitut Rabbit Cornea (SIRC) fibroblastic cell line (with negligible cytotoxicity). The peptide was considered a nonirritant when tested on the skin of New Zealand white rabbits and was very well tolerated in a 48-hour human patch test. A human repeat irritation patch test was conducted on 53 adult volunteers and was found not to be irritating or sensitizing to the skin. The Panel concluded that palmitoyl hexapeptide-12 is safe in cosmetics in the present practices of use and concentrations described in the safety assessment. Concentrations up to 20 ppm were reported in leave-on cosmetics as a skin conditioning agent. Overall, the peptide can be considered a non-toxin, shows limited immune reactivity, has favorable local toxicity data, and has beneficial properties to improve skin.

Caffeoyl hexapeptide-9 represents caffeic acid conjugated to hexapeptide-9 (GPQGPQ), a synthetic collagen peptide known for its ability to stimulate collagen and hyaluronic acid production in the skin. This sequence is a repeating motif of a natural collagen protein. An NCBI BLASTp search for the sequence results in many hits across 61 organisms, including collagen alpha-1 (UniProtID Q8IZC6.1) in Homo sapiens. ToxinPred3.0 was queried and predicted the peptide to be a non-toxin. Using the FASTA sequence of UniProtID Q8IZC6.1, AllerCatPro 2.0 identified weak similarity to Bos d alpha2l, an uncommon allergen in Bos taurus. The peptide did not result in any hits in the IEDB epitope database. BIOPEP-UWM was queried for bioactive peptides and did not provide any matches. A similar peptide was found in Peptipedia with anti-aging properties (RGDGPQGPQ, ID#3564). No toxicology data was found in the public domain, but it is registered in the Inventory of Existing Cosmetic Ingredients in China. Thus, the supplier likely has proprietary safety data that would need to be considered prior to market entry. Overall, the peptide can be considered a non-toxin, shows limited immune reactivity and has beneficial properties to improve skin.

Palmitoyl pentapeptide-4 (KTTKS) is the palmitic acid ester of pentapeptide-4 which stimulates extracellular matrix synthesis specific to collagen types I and III and fibronectin (Abu Samah and Heard, 2011). Characterization of the chemistry, physical and chemical properties, method of manufacture, and composition and impurities are provided in the reference link in Table 2. Due to the short length of the peptide, many hits were found across 75 organisms, including various proteins in Homo sapiens using NCBI BLASTp search. Therefore, no specific results could be obtained. ToxinPred3.0 was queried and predicted the peptide to be a non-toxin. BIOPEP-UWM was queried for bioactive peptides and did not provide any matches. Peptipedia annotated the sequence (ID# 3504) with anti-aging properties. AllerCatPro 2.0 is not applicable to such short sequences in the absence of a related protein sequence. The peptide did not result in any hits in the IEDB epitope database. Favorable toxicology data for palmitoyl pentapeptide-4 and pentapeptide-4 are provided in a CIR report (report link in Table 2). Results from dermal penetration studies on KTTKS was provided in Section 3.3 on bioavailability. The authors concluded that neither pentapeptide-4 nor palmitoyl pentapeptide-4 could permeate through the full-thickness hairless mouse skin over the period of the study. Dermal stability of both peptides was evaluated in vitro in epidermal and dermal skin extracts and whole skin homogenates from hairless mouse skin. Palmitoyl pentapeptide-4 was more stable in skin extracts over time compared to pentapeptide-4. Previous studies found the major degradation products of KTTKS to be TTKS and TKS, which indicates KTTKS was mostly degraded by skin aminopeptidases in a stepwise manner. Palmitoyl pentapeptide-4 was not acutely toxic to rats or guinea pigs when tested orally or dermally at 100 ppm concentrations, respectively. Palmitoyl pentapeptide-4 was not genotoxic when tested in two Ames mutagenicity assays and one in vitro micronucleus assay with mammalian cells. Palmitoyl pentapeptide-4 was not irritating to the skin when tested in vitro with reconstructed human epidermal skin or when tested directly on human skin, including a 48-hour study under semi-occlusive patch conditions. The peptide was predicted to be non-sensitizing when tested in chemico by the Direct Peptide Reactivity Assay and in vitro by the KeratinoSens cell line assay. In vivo, the peptide was non-sensitizing in a guinea pig maximization test, and in two human repeat insult patch tests (one under semi-occlusive patches and one under fully occlusive patch conditions). The peptide did not absorb UV light in the 290 to 400 nm range, confirming it to be non-phototoxic. Ocular irritation in the in vitro hen’s egg chorioallantoic membrane assay resulted in slightly irritating results in one test and moderately irritating results in another. In the human corneal epithelial in vitro model, the peptide was considered not irritating. In vivo, the peptide was nonirritating to New Zealand white rabbit eyes. Clinical testing demonstrated the peptide to be well tolerated in two 8-week studies and in one 12-week study. The Panel concluded that both palmitoyl pentapeptide-4 and pentapeptide-4 are safe in cosmetics in the present practices of use and concentration described in the safety assessment. Palmitoyl pentapeptide-4 was reportedly used up to 35 ppm in hair conditioners and up to 12 ppm in face and neck preparations and in eye lotions. Overall, the peptide can be considered a non-toxin, shows limited immune reactivity, favorable toxicology studies and has beneficial properties to improve skin.

Amanitin alpha (IWGIGCNP) is a cyclic peptide of eight amino acids. It is a deadly amatoxin found in several species of the mushroom family Aminata that inhibits RNA polymerase II, an enzyme critical for protein synthesis. The NCBI BLASTp search for the sequence results in many hits across 79 organisms, but none were found in Homo sapiens. The best match is Alpha-amanitin in Amanita fuliginea (UniProtID P85421.2). ToxinPred3.0 was queried using the FASTA sequence of UniProtID P85421 and correctly predicted the protein as a toxin. AllerCatPro 2.0 predicted no evidence for allergenicity for the protein sequence and the peptide did not result in any hits in the IEDB epitope database. Peptipedia annotated the sequence (ID# 73985) as a nutraceutical, toxin and venom peptide. BIOPEP-UWM did not result in any matches. Overall, the peptide can be considered a toxin, shows limited immune reactivity and has no known properties to improve the skin so it would not be a candidate for addition to cosmetic products.

Conotoxin ArlB (DECCSNPACR) is a venom of the Conus arenatus cone snail that causes neurotoxicity by blocking acetylcholine receptors. The NCBI BLASTp search for the sequence results in many hits across 52 organisms, but none were found in Homo sapiens. The best match is Alpha-conotoxin ArIA in Conus arenatus (UniProtID P0C8R2.1). ToxinPred3.0 was queried based on the protein FASTA sequence for UniProtID P0C8R2 and correctly predicted the peptide as a toxin. AllerCatPro 2.0 predicted no evidence for allergenicity for the protein sequence and the peptide did not result in any hits in the IEDB epitope database. Peptipedia and BIOPEP-UWM were queried and did not result in any matches. Overall, the peptide can be considered a toxin, shows limited immune reactivity and has no known properties to improve the skin so it would not be a candidate for addition to cosmetic products.

Bradykinin (RPPGFSPFR) is an endogenous human peptide important for regulation of blood pressure, inflammation, and pain sensation. The NCBI BLASTp searched for the sequence results in many hits across 52 organisms, including a protein in Homo sapiens (Kininogen-1, UniProtID P01042.2). However, the peptide is identical to Bradykinin in Trichonephila clavipes (UniProtID P0DM76.1). ToxinPred3.0 was queried and predicted the peptide to be a non-toxin. AllerCatPro 2.0 predicted no evidence for allergenicity for both protein sequences and the peptide did not result in any hits in the IEDB epitope database. Peptipedia annotated the sequence (ID# 6706) as antibacterial, antifungal, hormone, and therapeutic peptide. BIOPEP-UWM search confirmed the sequence as bradykinin (ID# 3340). Overall, the peptide is not a toxin but has therapeutic properties that would exclude it as a candidate for addition to cosmetic products and shows limited immune reactivity.

Enkephaline (YGGFL) is an endogenous opioid peptide that acts as a neurotransmitter activating opioid receptors in the brain and spinal cord to modulate pain signals. An NCBI BLASTp search for the sequence results in hits across 52 organisms, including various proteins in Homo sapiens. ToxinPred3.0 was queried and predicted the peptide to be a non-toxin. AllerCatPro 2.0 is not applicable without a defined protein sequence. The IEDB database identified the peptide as ID# 910542, which has been studied for immune reactivity. BIOPEP-UWM was queried for bioactive peptides and annotated the sequence (ID# 2879, 8122, 9521) as opioid antagonist, antioxidant, peptidase inhibitor and therefore as a therapeutic peptide. Peptipedia also identified the sequence (ID# 48028) as an antioxidative, hormone, and therapeutic peptide. Overall, the peptide is not a toxin but has therapeutic properties that would exclude it as a candidate for addition to cosmetic products and shows limited immune reactivity.

6. Discussion

A safety framework is presented that adds a group of bioinformatic tools to traditional risk assessment approaches (CIR and SCCS) to evaluate the safety of peptides for consideration in cosmetic formulations. This framework focused on characterization of the peptide of interest, knowledge gathering through bioinformatics, consideration of systemic bioavailability/systemic toxicity, genotoxicity, evaluation of local toxicity endpoints and immunogenicity, followed by clinical testing experience and post-market surveillance. The addition of in silico bioinformatic tools is consistent with how genetically modified peptides and proteins are evaluated for food products. This approach has not been previously considered in the CIR safety evaluation reports published on various peptides and would add new information to the weight of evidence approach for the safety evaluation by screening for toxins, biologically potent materials, and immunogenic peptides. As with all tools, there are both benefits and limitations. While providing a supporting role within a weight of evidence approach, it is not a stand-alone tool for decision making. For example, during the early phases of peptide development these screening tools can be used to narrow down or limit peptides if adverse safety signals are identified. Alternatively, the lack of a safety signal by itself does not provide positive assurances of safety, thus necessitating the need for careful attention to the entire framework. Bioactivity and homologies with extracellular matrix proteins generated with bioinformatic tools would need to be confirmed in vitro or in vivo (Rivero-Pino et al., 2023). A limitation of the bioinformatic tools is that they are only as good as the strength of the databases used to develop the tool. Often these databases were developed for the food or pharmaceutical industries and clearly cannot represent the entire range of amino acid sequences across all peptides. Short sequences can have a low sensitivity in tools like BLAST and AllerCatPro and there can be a high rate of false-positive hits with short sequences in BLAST. When interrogating a modified or conjugated peptide, the un-conjugated and un-modified peptide sequences should be queried, and then additional understanding is needed to justify the modification and conjugation regarding their impact on safety. Continued updating of these tools and the development of new bioinformatic tools are likely to occur, so due diligence includes monitoring these activities. These in silico tools should be used in addition to, not in place of, traditional methods that adhere to the SCCS Notes of Guidance and use non-animal OECD test methods.

The present framework interrogated the amino acid sequence of 7 different peptides through a set of 6 recommended bioinformatic tools (NCBI BLASTp, ToxinPred3.0, Peptipedia, BIOPEP-UWM, AllerCatPro 2.0, IEDB). The results of these findings are summarized in Table 3. For the peptides intended for cosmetic products, two of the three were homologous with extracellular matrix proteins in the skin, all were predicted to be non-toxins, non-allergenic, and the palmitoyl hexapeptide-12 was predicted to have homology with the chemotactic domain of elastin. Thus, the three peptides intended to be formulated into cosmetic products did not have any bioinformatic signals of concern. Traditional approaches summarized by CIR indicated palmitoyl hexapeptide-12 and palmitoyl pentapeptide-4 were well characterized and had sufficient toxicology data to alleviate concerns about local toxicity, genotoxicity, and systemic toxicity, along with favorable tolerance in clinical testing. Caffeoyl hexapeptide-9 lacked similar traditional safety data in the public domain and these gaps would need to be addressed with supplier data prior to market introduction. The two toxin peptides were clearly not good candidates for cosmetic application as both were indicated as toxins by the ToxinPred tool. The two endogenous peptides were also not good candidates for cosmetic application given the bradykinin and opioid peptide biological activities noted by both the NCBI BLASTp and BIOPEP-UWM tools. In total, the bioinformatic tools accurately predicted those peptides that could be pursued as cosmetic ingredients. None of the peptides received homology matches in the AllerCatPro tool, suggesting that none of the peptides are likely to induce an allergic reaction. While the bioinformatic tools provided unique information to assist in the safety evaluation of the peptides, the number of examples used in the present work are limited. The overall robustness of this approach can only be strengthened by routine adoption of the safety framework with additional publications attesting to the value (or limitations) of the model.

Borrowing from the principles of Lintner, 2002, Lintner, 2002 and Fields et al., 2009, cosmetic peptides that are safe and provide meaningful benefits have the following requirements: (1) The peptide must have a cosmetic benefit leading to a demonstrable effect, (2) the bioactivity does not have negative consequences due to its mechanism of action, (3) the peptide does not exhibit cytotoxicity, irritation, skin allergy, or mutagenicity, (4) the peptide is capable of reaching its desired target, and (5) the peptide can be formulated to be stable and effectively delivered to the skin. The case study with pal-KTTKS exemplifies a peptide that fulfills these criteria.

7. Conclusion

Following the adoption of animal testing bans for cosmetic products and ingredients, toxicologists and regulators continue to pursue new methods to assist in providing safety substantiation for those products. To this end, the authors have provided a safety framework that newly incorporates bioinformatic tools and have tested that framework with a variety of peptides to determine their suitability for safe use in cosmetics. The six added tools provided meaningful results for the seven peptides tested in the framework. While traditional safety evaluation methods applied previously to peptides provide a foundation to build upon, the added bioinformatic tools yielded unique information related to amino acid homology with known peptides and proteins, toxins, allergens, as well as biological information on functional activity. This safety framework for peptides adds additional value to the risk assessment process and is entirely consistent with the exposure-led, hypothesis-driven NGRA approach. Future directions include the adoption of the safety framework by toxicologists and regulators in the cosmetic industry and publishing results, which would lead to strengthening or identifying needed refinements to the framework.

CRediT authorship contribution statement

Donald L. Bjerke: Conceptualization, Writing – original draft, Writing – review & editing, Methodology. Jin Li: Writing – review & editing. Yuan Gao: Writing – review & editing. Ping Hu: Conceptualization, Methodology, Visualization, Writing – original draft, Writing – review & editing. Karl Lintner: Investigation, Methodology, Writing – original draft, Writing – review & editing. Tomohiro Hakozaki: Writing – review & editing.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: [DLB, JL, YG, PH, and TH all work for The Procter & Gamble Company, that manufactures products with some of the compounds mentioned in the manuscript].