Optimizing the detection of N-nitrosamine mutagenicity in the Ames test

Abstract

Accurately determining the mutagenicity of small-molecule N-nitrosamine drug impurities and nitrosamine drug substance-related impurities (NDSRIs) is critical to identifying mutagenic and cancer hazards. In the current study we have evaluated several approaches for enhancing assay sensitivity for evaluating the mutagenicity of N-nitrosamines in the bacterial reverse mutagenicity (Ames) test. Preincubation assays were conducted using five activation conditions: no exogenous metabolic activation and metabolic activation mixes employing both 10% and 30% liver S9 from hamsters and rats pretreated with inducers of enzymatic activity. In addition, preincubations were conducted for both 60 min and 30 min. These test variables were evaluated by testing 12 small-molecule N-nitrosamines and 17 NDSRIs for mutagenicity in Salmonella typhimurium tester strains TA98, TA100, TA1535, and TA1537, and Escherichia coli strain WP2 uvrA (pKM101). Eighteen of the 29 N-nitrosamine test substances tested positive under one or more of the testing conditions and all 18 positives could be detected by using tester strains TA1535 and WP2 uvrA (pKM101), preincubations of 30 min, and S9 mixes containing 30% hamster liver S9. In general, the conditions under which NDSRIs were mutagenic were similar to those found for small-molecule N-nitrosamines.

1. Introduction

N-Nitrosamines (or nitrosamines) are a class of carcinogenic chemicals having a structure that includes a nitroso group (NO) bonded to an amine (NH₂). Several N-nitrosamines have been detected as drug impurities and many more could potentially be present as impurities in drug substances and/or drug products (FDA, 2024, 2024a; Charoo et al., 2023). N-Nitrosamine drug impurities are commonly divided into two general subtypes based on their size and structural components; both subtypes have been detected in drug products. Those that do not share obvious structural similarity to the drug (referred to as the active pharmaceutical ingredient or API), and are found in many different drug products, are generally termed ‘small-molecule N-nitrosamines’. Examples include simple aliphatic compounds such as N-nitroso-dimethylamine (NDMA) and N-nitroso-diethylamine (NDEA).

The second subcategory of N-nitrosamine drug impurities is termed ‘nitrosamine drug substance-related impurities’ or NDSRIs. These are nitroso group addition products to drugs that contain amino groups. Thus, many NDSRIs are larger and more complex than small-molecule N-nitrosamines, and, unlike small-molecule N-nitrosamines, share structural similarity to the API or a fragment of the API (FDA, 2024; FDA 2024a). Examples of NDSRIs include N-nitroso-phenmetrazine (Greenblatt et al., 1972), an N-nitroso impurity of valsartan (Glowienke et al., 2022) and N-nitroso-propranolol (Li et al., 2023). The distinction between NDSRIs and small-molecule nitrosamines is not always clear-cut: some small-molecule nitrosamine drug impurities do share structural similarity to the API that they are associated with (Horne et al., 2023; Wichitnithad et al., 2023), and some nitrosamines derived from APIs are nitrosated fragments of the API structure (e.g., N-nitroso-desvaleryl-valsartan). For the purpose of this report, however, N-nitrosamines that contain a large portion of the API structure, including compounds like N-nitroso-desvaleryl-valsartan, will be referred to as NDSRIs.

A high proportion of N-nitrosamines are mutagens and carcinogens; and some are among the most potent mutagenic carcinogens that are known (Lijinsky, 1987; Cheeseman et al., 1999; Thresher et al., 2020). Because of this, N-nitrosamines, along with impurities from a small number of other chemical classes, comprise a ‘Cohort of Concern’ for regulatory purposes, and the Threshold of Toxicological Concern (TTC) concept (ICH, 2023) is not used for setting their limits in drugs. Until the recent introduction of the Carcinogenic Potency Categorization Approach (CPCA; EMA, 2023; FDA, 2023; Kruhlak et al., 2024; see Discussion), Acceptable Intake (AI) limits for N-nitrosamine impurities have been based on compound-specific limits derived from rodent carcinogenicity data, on read-across using suitable surrogates, the limits given in FDA guidance (FDA, 2024), or a class-specific limit, e.g., the 18 ng/day limit employed by the European Medicines Agency (EMA) (EMA, 2020; Nudelman et al., 2023; Schlingemann et al., 2023).

The ideal way of dealing with the potential risks posed by N-nitrosamine drug impurities is ensuring that they do not occur in drug products (Horne et al., 2023). For instance, it may be possible to devise approaches to reduce the concentration of N-nitrosamine impurities in drugs by inhibiting their formation (Nanda et al., 2021). The FDA has issued guidance (FDA, 2024) and published an example study (Shakleya et al., 2023) on developing strategies for limiting N-nitrosamines in drug products. In practice, N-nitrosamine impurity formation in drugs is likely to be compound-specific (Moser et al., 2023), and approaches for limiting N-nitrosamine impurities in drugs may not be straightforward to implement.

Nitrosamines formed from APIs (i.e., NDSRIs) are viewed as a particular challenge since a significant proportion of APIs contain secondary or tertiary amino groups that potentially can be nitrosated to form NDSRIs (Lijinsky and Epstein, 1970; Schlingemann et al., 2023; FDA, 2024; FDA, 2024a). Thus, totally eliminating NDSRIs from drug products may be difficult (Nudelman et al., 2023). Although the potential for mutagenic and carcinogenic N-nitrosamines being generated from APIs has been recognized for some time (Lijinsky and Epstein, 1970; Lijinsky et al., 1972; Lijinsky, 1974, 1980; Pool et al., 1979; De Flora and Picciotto, 1980; IARC, 1980; Andrews et al., 1980, 1984; De Flora, 1981; De Flora et al., 1984; Ichinotsubo et al., 1981; Raisfeld--Danse and Chen, 1983; Brambilla, 1985; Brambilla and Martelli, 2007), in recent years there has been an increased concern over the formation of NDSRIs as drug impurities and the cancer risks that they pose (e.g., FDA, 2024a; Nudelman et al., 2023).

The bacterial mutagenicity assay (referred to here as the ‘Ames test’, although it is acknowledged that the current test employs additional tester strains and methods than those described originally by Ames (Ames et al., 1975; Maron and Ames, 1983) plays an important role in determining if a drug impurity is a mutagen and thus poses a potential cancer hazard (ICH, 2023). Besides providing direct experimental evidence of the impurity’s mutagenicity, Ames data have been utilized in learning sets and the development of Structure-Activity Relationship (SAR) modeling employed in the assessment of mutagenic potential as recommended by the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) guidance document M7(R2) (ICH, 2023).

There is extensive literature describing N-nitrosamine Ames test results, some dating back to the beginnings of assay development (Malling, 1971), and well before the current set of tester strains and test guidelines were established. An influential older study indicated that the mutagenicity of a series of aliphatic small-molecule N-nitrosamines in what was then the Ames test correlated poorly with their carcinogenicity (Rao et al., 1979). This perception about a lack of correspondence between mutagenicity in the Ames test and the carcinogenicity of N-nitrosamines was reinforced in comments made by Lijinsky (1987) in another widely cited paper. In fact, detecting the mutagenicity of the common reference N-nitrosamine and potent carcinogen, NDMA, is difficult using standard Ames test methods (e.g., using the plate-incorporation method, rat liver S9, and the commonly used solvent, dimethyl sulfoxide [DMSO]) (Ames et al., 1973, 1975; Bartsch et al., 1976; Yahagi et al., 1977; Guttenplan, 1979; Prival et al., 1979; Mori et al., 1985).

Modifications of the test protocol that improve the sensitivity of the assay for detecting the mutagenicity of N-nitrosamines have been developed, and recommendations have been made regarding how best to test N-nitrosamines using the Ames assay (Ames et al., 1975; Lee and Guttenplan, 1981; Guttenplan, 1987; Gatehouse et al., 1994; OECD, 2020). However, it remains a problem to accept negative Ames findings for N-nitrosamine drug impurities given that as a class, most of the nitrosamines that have been tested have been carcinogenic in animals. This uncertainty is greater for NDSRIs since very few have been tested for carcinogenicity, and very little is known about the performance of the Ames test for detecting their mutagenicity.

Thus, there is the need to reexamine the performance of the Ames test for evaluating the mutagenicity of N-nitrosamines, especially NDSRIs. Many protocols for performing the Ames test have been used, and Organisation for Economic Co-operation and Development (OECD) Test Guideline (TG) 471 contains guidance about modifying the test for better detecting the mutagenicity of different classes of compounds (OECD, 2020). Ideally, a protocol or strategy might be devised that accurately classifies mutagenic N-nitrosamines as mutagenic, thus increasing confidence in the results of the assay. Moreover, unlike the case with small--molecule nitrosamines, very little is known about the mutagenicity and carcinogenicity of NDSRIs. It is unclear how or whether Ames testing recommendations for the well-studied small--molecule N-nitrosamines should be modified for evaluating the mutagenicity of these generally larger and more complex substances.

Given the concern over the possible health risks posed by small-molecule N-nitrosamines and NDSRIs, and the potential for these concerns disrupting the human drug supply, several groups have formed to study how best to evaluate the safety of these substances (DHHS, 2023; see Burns et al., 2023; Schlingemann et al., 2023; Nudelman et al., 2023). In addition, several studies that evaluate the mutagenicity of small-molecule N-nitrosamines in the Ames test have been published recently by individual laboratories (Bringezu and Simon, 2022; Dieckhoff et al., 2024; Thomas et al., 2024). This report describes an effort started in 2021 by the U.S. Food and Drug Administration (FDA) to evaluate bacterial mutagenicity testing protocols reported to enhance detecting the mutagenicity of N-nitrosamines with the ultimate purpose of providing recommendations on conditions that most accurately and reliably determine their mutagenic potential. In our study, we have tested a total of 29 N-nitrosamines, including 12 small-molecule N-nitrosamines, some with extensive data on their mutagenicity and carcinogenicity in the literature, and 17 NDSRIs, for which little or no data are available. Table 1 presents chemical information on the test substances and Table 2 gives information from public sources on what is known of their carcinogenicity and their activity in the Ames test.

Table 1. N-Nitrosamine test substances.

ID (identifying number) used for cross-referencing to Supplementary Material.

ID	Test substance	Chemical name (IUPAC*)	CAS #/Alternate CAS #	Molecular weight (g/mol)	Chemical structure	Source (claimed purity)
Small molecule N-nitrosamine drug impurities and model substances
SM-1	N-nitroso-dimethylamine (NDMA)	N,N-Dimethylnitrous amide	62-75-9	74.08		Chem Service (99.5%)
SM-2	N-nitroso-diethylamine (NDEA)	N,N-Diethylnitrous amide	55-18-5	102.14		TCI (99.0%)
SM-3	N-nitroso-isopropylethyl-amine (NIPEA/NEIPA)	N-Ethyl-N-propan-2-ylnitrous amide	16339-04-1	116.16		Enamine (95%)
SM-4	1-methyl-4-nitrosopiperazine (MNP)	1-Methyl-4-nitrosopiperazine	16339-07-4	129.16		TRC (99.8%)
SM-5	N-nitroso-diisopropylamine (NDIPA)	N,N-Di(propan-2-yl)nitrous amide	601-77-4	130.22		Chem Service (99.1%)
SM-6	N-nitroso-methylphenyl-amine (NMPA)	N-Methyl-N-phenylnitrous amide	614-00-6	136.15		TCI (>98.0%)
SM-7	N-nitroso-N-methyl-4-aminobutyric acid (NMBA)	4-[Methyl(nitroso)amino] butanoic acid	61445-55-4	146.15		Chem Space (95%)
SM-8	N-nitroso-N-ethylaniline (NEPA)	N-Ethyl-N-phenylnitrous amide	612-64-6	150.18		Clearsynth (99.6%)
SM-9	N-nitroso-dibutylamine (NDBA)	N,N-Dibutylnitrous amide	924-16-3	158.25		TCI (≥97.0%)
SM-10	1-cyclopentyl-4-nitrosopiperazine (CPNP)	1-Cyclopentyl-4-nitrosopiperazine	61379-66-6	183.26		TRC (97.0%)
SM-11	N-nitroso-diphenylamine	N,N-Diphenylnitrous amide	86-30-6	198.22		TRC (98.0%)
SM-12	N,N-bis(2,2-diethoxyethyl) nitrous amide	N,N-Bis(2,2-diethoxyethyl)nitrous amide	67856-67-1	278.35		Clearsynth (99.6%)
Nitrosamine drug substance-related impurities (NDSRIs)
NDSRI-13	N-nitroso-phenylephrine	N-[(2R)-2-Hydroxy-2-(3-hydroxyphenyl)ethyl)-N-methylnitrous amide	78658-64-7	196.21		Clearsynth (99.1%)
NDSRI-14	N-nitroso-lorcaserin	(5R)-7-Chloro-5-methyl-3-nitroso-1,2,4,5-tetrahydro-3-benzazepine	2518136-84-8	224.69		Clearsynth (≥99%)
NDSRI-15	N-nitroso-varenicline	8-Nitroso-7, 8, 9, 10-tetrahydro-6H-6, 10-methanoazepino[4, 5-g]quinoxaline	2755871-02-2	240.26		Clearsynth (97.2%)
NDSRI-16	N-nitroso-desmethyl-diphenhydramine	N-(2-(Benzhydryloxy)ethyl)-N-methylnitrous amide	55855-43-1	270.33		Clearsynth (97.5%)
NDSRI-17	N-nitroso-propranolol	N-(2-Hydroxy-3-naphthalen-1-yloxypropyl)-N-propan-2-ylnitrous amide	84418-35-9	288.34		Clearsynth (99.6%)
NDSRI-18	N-nitroso-nortriptyline	N-Methyl-N-[3-(2-tricyclo[9.4.0.03,8]pentadeca-1 (15),3,5,7,11,13-hexaenylidene)propyl]nitrous amide	55855-42-0	292.37		Clearsynth (99.1%)
NDSRI-19	N-nitroso-diclofenac	2-[2-(2,6-Dichloro-N-nitrosoanilino)phenyl]acetic acid	66505-80-4	325.15		Clearsynth (96.1%)
NDSRI-20	N-nitroso-duloxetine	N-Methyl-N-(3-naphthalen-1-yloxy-3-thiophen-2-ylpropyl) nitrous amide	2680527-91-5	326.41		Clearsynth (99.8%)
NDSRI-21	N-nitroso-sertraline	N-(4-(3,4-Dichlorophenyl)-1,2,3,4-tetrahydronaphthalen-1-yl)-N-methylnitrous amide	3006789-98-3	335.23		Clearsynth (99.8%)
NDSRI-22	N-nitroso-fluoxetine	N-Methyl-N-[3-phenyl-3-[4-(trifluoromethyl)phenoxy] propyl]nitrous amide	150494-06-7	338.32		Clearsynth (99.7%)
NDSRI-23	N-nitroso-paroxetine	(3S, 4R)-3-(1,3-Benzodioxol-5-yloxymethyl)-4-(4-fluorophenyl)-1-nitrosopiperidine	2361294-43-9	358.4		Clearsynth (99.9%)
NDSRI-24	N-nitroso-ciprofloxacin	1-Cyclopropyl-6-fluoro-7-(4-nitrosopiperazin-1-yl)-4-oxoquinoline-3-carboxylic acid	864443-44-7	360.34		Clearsynth (98.0%)
NDSRI-25	N-nitroso-desvaleryl-valsartan	(2R)-3-Methyl-2-[nitroso-[[4-[2-(2H-tetrazol-5-yl)phenyl] phenyl]methyl]amino]butanoic acid	2254485-68-0	380.4		LKT Labs (98%)
NDSRI-26	N-nitroso-bumetanide	3-(Butyl(nitroso)amino)-4-phenoxy-5-sulfamoylbenzoic acid	2490432-02-3	393.41		LKT Labs (≥99%)
NDSRI-27	N-nitroso-desvaleryl-valsartan methyl ester	(S)-Methyl 2-(((2’-(1H-tetrazol-5-yl)-[1, 1′-biphenyl]-4-yl) methyl)(nitroso)amino)-3-methylbutanoate		394.4		LKT Labs (99.3%)
NDSRI-28	N-nitroso-folic acid (N-10 nitrosated)	2-[[4-[(2-Amino-4-oxo-3 H-pteridin-6-yl)methyl-nitrosoamino)benzoyl]amino]pentanedioic acid	26360-21-4	470.4		Clearsynth (94.5%)
NDSRI-29	N-nitroso-dabigatran etexilate	Ethyl 3-[[2-[[4-(N-hexoxycarbonylcarbamimidoyl)-N-nitrosoanilino]methyl]-1-methylbenzimidazole-5-carbonyl]-pyridin-2-ylamino]propanoate	2892260-29-4	656.73		Clearsynth (90.9%)

IUPAC, International Union of Pure and Applied Chemistry; TCI, TCI America; TRC, Toronto Research Chemicals.

Table 2. Published/posted/publicly accessible Ames test results, cancer findings, and Acceptable Intake (AI) limits for small molecule N-nitrosamines and nitrosamine drugsubstance-related impurities (NDSRIs) tested in this study.

AI limits are those published by FDA as of September 4, 2024 (FDA, 2024; FDA, 2024a) (others given in footnotes). ID (identifying number) used for cross-referencing to Supplementary Material. Some publicly available Ames responses were only positive-negative calls, with no data or experimental details.

ID	Test substance	Ames response (all S9-mediated except where noted)	Representative Ames references	Carcinogenicitya	AI limitb	Notes
Small molecule N-nitrosamine drug impurities and model substances
SM-1	N-nitrosodimethylamine (NDMA)	Positive; preincubation more sensitive	Malling (1971); McCann et al. (1975); Yahagi et al. (1977); Lijinsky and Andrews (1983)	Lhasa positive: mouse (brain, lung, liver), rat (liver, kidney, lung, testes, nasal cavity); IARC Group 2A	96 ng/day	Mutagenic in mouse and rat TGR (Mirsalis et al., 1993; Gollapudi et al., 1998; Lynch et al., 2024) and host-mediated assay (Kerklaan et al., 1981); DNA damage in mice and rats (Brambilla et al., 1981; Tsuda et al., 2000); recombinogen in Drosophila (Vogel and Nivard, 1993); mutagenic/genotoxic in CHO, V79, L5178Y/Tk+/−, HepaRG cells (Kuroki et al., 1977; Kuroki and Drevon, 1978; Langenbach et al., 1978; Amacher and Paillet, 1983; Li, 1984; Seo et al., 2023, 2024)
SM-2	N-nitroso-diethylamine (NDEA)	Positive; preincubation more sensitive	Yahagi et al. (1977); Lijinsky and Andrews (1983)	Lhasa positive: mouse (liver, lung, esophagus, stomach, forestomach, GI track, tongue), rat (forestomach, GI track, esophagus, tongue, liver, urinary tract, kidney, nasal cavity), other species; IARC Group 2A	26.5 ng/day	Mutagenic in rat and mouse TGR, error-corrected sequencing (Akagi et al., 2015; Bercu et al., 2023; Zhang et al., 2024) and host-mediated assay (Kerklaan et al., 1981); DNA damage in mice and rats (Brambilla et al., 1981; Tsuda et al., 2000; Bercu et al., 2023; Zhang et al., 2024); recombinogen in Drosophila (Vogel and Nivard, 1993); mutagenic/genotoxic in CHO, L5178Y/Tk+/−, V79, TK6, HepaRG cells (Kuroki et al., 1977; Langenbach et al., 1978; Amacher and Paillet, 1983; Li, 1984; Li et al., 2022; Seo et al., 2023)
SM-3	N-nitroso-isopropylethylamine (NIPEA/NEIPA)	No Ames data identifiedc		Lhasa positive: rat (esophagus); IARC: not listed	400 ng/day (CPCA PC3)	Genotoxic in HepaRG, TK6 cells (Li et al., 2022; Seo et al., 2023)
SM-4	1-methyl-4-nitrosopiperazine (MNP)	Positive	Bartsch et al. (1976); Rao et al. (1978); Zeiger and Sheldon (1978); Araki et al. (1984)	Lhasa positive: rat (tumor sites not specified); IARC, not listed	400 ng/day (CPCA PC3)	Mutagenic in yeast, Drosophila, V79 cells, host-mediated assay (Zeiger et al., 1972; Kuroki et al., 1977; Larimer et al., 1980; Nix et al., 1980); Ames positive with human liver S9 (Sabadie et al., 1980); recombinogen in Drosophila (Vogel and Nivard, 1993); not mutagenic in E. coli WU3610 (Elespuru and Lijinsky, 1976)
SM-5	N-nitroso-diisopropylamine (NDIPA)	Rao: negative in plate incorporation assay; Tennant: negative in plate incorporation; positive in preincubation	Rao et al. (1979); Tennant et al. (2023)	Lhasa positive: rat (multiple sites, nasal turbinate); IARC, not listed	1500 ng/day (CPCA PC5)	‘Disputable/questionable’ mutagenicity in mouse host-mediated assay (Kerklaan et al., 1981); genotoxic in λ-phage induction assay; TK6 and HepaRG cells (Rossman et al., 1991; Li et al., 2022; Seo et al., 2023); site-specific N2-isopropyl-dG adducts, but not N6-isopropyl-dA adducts mutagenic in E. coli (Upton et al., 2006)
SM-6	N-nitroso-methylphenylamine (NMPA)	Positive in preincubation with nonstandard strains or co-mutagen	Raineri et al. (1981); Wakabayashi et al. (1981); Lijinsky and Andrews (1983); Zielenska and Guttenplan (1988)	Lhasa positive: rat (multiple sites, esophagus); IARC, not listed	100 ng/day (CPCA PC2)	Mutagenic/genotoxic in yeast, TK6, HepaRG cells (Mehta and Von Borstel, 1984; Li et al., 2022; Seo et al., 2023); forms unstable adduct with adenine (Koepke et al., 1990)
SM-7	N-nitroso-N-methyl-4-aminobutyric acid (NMBA)	Negative/very weak/inconclusive with standard methods; positive with nonstandard methods	Nagao et al. (1977); Kier et al. (1986); Inami et al. (2013); Tennant et al. (2023)	Lhasa positive: rat (ureter/urinary bladder); IARC: not listed	1500 ng/day (CPCA PC4)	Mutagenic/genotoxic in yeast, TK6, HepaRG cells (Mehta and Von Borstel, 1984; Li et al., 2022; Seo et al., 2023)
SM-8	N-nitroso-N-ethylaniline (NEPA)	Positive only with comutagen	Wakabayashi et al. (1981)	Lhasa, IARC: not listed
SM-9	N-nitroso-dibutylamine (NDBA)	Positive	Yahagi et al. (1977)	Lhasa positive: mouse (multiple sites, esophagus, urinary bladder, forestomach, liver, lung), rat (forestomach, liver, lung, urinary bladder, esophagus); IARC Group 2B		DNA damage in bladder cells, and liver and bladder in rats (Robbiano et al., 2002); DNA damage in mice and rats (Brambilla et al., 1981; Tsuda et al., 2000); mutagenic/genotoxic in CHO, TK6, V79, HepaRG cells (Kuroki et al., 1977; Li, 1984; Li et al., 2022; Seo et al., 2023)
SM-10	1-cyclopentyl-4-nitrosopiperazine (CPNP)	No Ames data identified		Lhasa, IARC: not listed	400 ng/day (CPCA PC3)	Genotoxic in HepaRG cells (Seo et al., 2023)
SM-11	N-nitroso-diphenylamine	Negative with standard strains/methods; weak positive with nonstandard strains/methods	Yahagi et al. (1977); Raineri et al. (1981); Wakabayashi et al. (1981); Lijinsky and Andrews (1983); Zielenska and Guttenplan (1988); Tennant et al., 2023	Lhasa positive: mouse (tumor sites not specified), rat (urinary bladder); IARC Group 3	d	Weakest carcinogen with a calculated TD50 (Cross and Ponting, 2021); negative for in vitro chromosome aberrations (Ishidate et al., 1988), mutation in V79 cells (Kuroki et al., 1977), recombination in Drosophila (Vogel and Nivard, 1993)
SM-12	N,N-bis(2,2-diethoxyethyl) nitrous amide	Negative	Rao et al. (1979)	Lhasa: negative in rat; IARC: not listed
Nitrosamine drug substance-related impurities (NDSRIs)
NDSRI-13	N-nitroso-phenylephrine	No Ames data identified		Lhasa, IARC: not listed	100 ng/day (CPCA PC2)
NDSRI-14	N-nitroso-lorcaserin	No Ames data identified		Lhasa, IARC: not listed		Lorcaserin withdrawn from US market
NDSRI-15	N-nitroso-varenicline	No Ames data identified		Lhasa, IARC: not listed	400 ng/day (CPCA PC3)
NDSRI-16	N-nitroso-desmethyl-diphenhydramine	No Ames data identified		Lhasa, IARC: not listed	26.5 ng/day (CPCA PC1)
NDSRI-17	N-nitroso-propranolol	Positive; direct-acting in TA98	Li et al. (2023)	Lhasa, IARC: not listed	1500 ng/day (CPCA PC4)	Earlier negative Ames and V79 mutagenicity findings (Raisfeld-Danse and Chen, 1983); positive in vitro mammalian genotoxicity findings (Robbiano et al., 1991)
NDSRI-18	N-nitroso-nortriptyline	No Ames data identified		Lhasa, IARC: not listed	26.5 ng/day (CPCA PC1)
NDSRI-19	N-nitroso-diclofenac	No Ames data identified		Lhasa, IARC: not listed	1500 ng/day (CPCA PC5)
NDSRI-20	N-nitroso-duloxetine	Positive	Jolly et al. (2024)	Lhasa, IARC: not listed	100 ng/day	Interim AI limit currently in force; mutagenic in rat TGR assay (Jolly et al., 2024)
NDSRI-21	N-nitroso-sertraline	No Ames data identified		Lhasa, IARC: not listed	100 ng/day (CPCA PC2)
NDSRI-22	N-nitroso-fluoxetine	Positive	Jolly et al. (2024)	Lhasa, IARC: not listed	100 ng/day	Fluoxetine genotoxic under nitrosating conditions (Ozhan and Alpertunga, 2003); mutagenic in rat TGR assay (Jolly et al., 2024)
NDSRI-23	N-nitroso-paroxetine	No Ames data identified		Lhasa, IARC: not listed	e	Paroxetine genotoxic under nitrosating conditions (Ozhan and Alpertunga, 2003)
NDSRI-24	N-nitroso-ciprofloxacin	No Ames data identified		Lhasa, IARC: not listed	1500 ng/day (CPCA PC4)	No useful data in bacteria: extremely toxic in Ames test (this report); interim Al limit currently in force
NDSRI-25	N-nitroso-desvaleryl-valsartan	Negative	Glowienke et al. (2022) (referred to as compound 181–14)	Lhasa, IARC: not listed		Also negative in mouse TGR assay (Glowienke et al., 2022)
NDSRI-26	N-nitroso-bumetanide	No Ames data identified		Lhasa, IARC: not listed	1500 ng/day (CPCA PC4)
NDSRI-27	N-nitroso-desvaleryl-valsartan methyl-ester	No Ames data identified		Lhasa, IARC: not listed		Could be de-esterified to NDSRI-25
NDSRI-28	N-nitroso-folic acid	Positive	Purchase et al. (1978)	Lhasa: not listed; IARC (1978): ‘no evaluation … …on the basis of available data’	1500 ng/day (CPCA PC4)	Weak positive for lung tumors in neonatal mouse assay (Wogan et al., 1975)
NDSRI-29	N-nitroso-dabigatran etexilate	No Ames data identified		Lhasa, IARC: not listed	400 ng/day (CPCA PC3)

^aCarcinogenicity responses from Lhasa carcinogenicity database with cancer target tissues for mice and rats (https://carcdb.lhasalimited.org/) and classifications by the International Agency for Research against Cancer (IARC) (https://monographs.iarc.who.int/agents-classified-by-the-iarc/).

^bAI limits based on Carcinogenic Potency Categorization Approach (CPCA) are identified and Potency Category (PC) given (FDA, 2024a); other AI limits are compound-specific or based on read-across using a surrogate.

^cBased on information provided by literature search conducted by the US FDA Library.

^dEMA recognizes an AI limit of 78 μg/day.

^eEMA recognizes an AI limit of 1300 ng/day.

There are hundreds of structurally unique N-nitrosamines that could be present as impurities in human drugs (FDA, 2024a), and there are many factors that could influence the extent of N-nitrosamine mutagenicity in the Ames test. Thus, it is anticipated that this study can provide only part of the data necessary for addressing the issue of how best to evaluate the mutagenic potential of N-nitrosamine drug impurities. This is an evolving field, with potential for affecting public health; preliminary work from this project on a single NDSRI, N-nitroso-propranolol (Li et al., 2023), already has been cited in several recommendations on how best to evaluate the mutagenicity of N-nitrosamines in the Ames test (EMA, 2024; FDA, 2024a; Health Canada, 2024). As the amount of data on different N-nitrosamines increases, it is further anticipated that modeling efforts using these data will identify more refined structure-activity relationship trends, laying the foundation for improved predictive methods to assess the mutagenicity of N-nitrosamine impurities.

2. Materials and methods

Most methods used in this study were described recently in Li et al. (2023) and are given here for completeness, and where appropriate, with explanations of choices that were made in how the testing was conducted. An effort was made to follow the principles laid out in OECD TG 471 (OECD, 2020) and in reports from various International Workshop on Genotoxicity Testing committees that have considered bacterial mutagenicity testing (Gatehouse et al., 1994; Levy et al., 2019). However, as the testing laboratory at NCTR had no history of performing Ames tests using the conditions employed for this study, no positive or negative control databases were available at the beginning of the study. For negative control values, data acceptance relied upon comparisons with literature values (e.g., Table 1 in Levy et al., 2019). Note that extensive amounts of control data were generated during the course of this study and can be found in Supplementary Material.

2.1. Materials

2.1.1. N-Nitrosamine test substances

A total of 29 N-nitrosamine test substances were evaluated for mutagenicity; one test substance, N-nitroso-ciprofloxacin, did not provide useful data because of its toxicity. Table 1 lists small-molecule N-nitrosamines and NDSRIs separately, each group in the order of their molecular weight.

2.1.2. Positive controls

Positive control mutagens plated at the time of the assay were obtained from the indicated sources and used as received from the vendor without further purification or analysis: 2-aminoanthracene (2AA; Sigma-Aldrich, St. Louis, MO), 6-chloro-9-[3-(2-chloroethylamino)propylamino]-2-methoxyacridine dihydrochloride (ICR-191; Sigma-Aldrich), 2-nitrofluorene (Aldrich), sodium azide (Sigma), NDMA (Chem Service, West Chester, PA), 1-cyclopentyl-4-nitroso-piperazine (CPNP; Toronto Research Chemicals, Toronto, Canada), and methylmethanesufonate (MMS; Sigma).

2.1.3. Tester strains and S9

The tester strains used were the Salmonella typhimurium histidine auxotrophs TA98, TA100, TA1535, and TA1537, and the E. coli tryptophan auxotroph, WP2 uvrA (pKM101). Tester strains TA98 and TA1537 are reverted from histidine dependence (auxotrophy) to histidine independence (prototrophy) by frameshift mutagens; the remaining tester strains are reverted by base pair substitution at G:C (TA1535 and TA100) and A:T base pairs [WP2 uvrA (pKM101)]. All tester stains were obtained from MolTox (Boone, NC) and stored either on paper disks at 4 °C or as frozen stocks at −80 °C.

All S9s were purchased from MolTox. In the initial testing conducted for this study, liver S9s prepared from Aroclor-1254-induced male Sprague-Dawley rats and Aroclor-1254-induced male Syrian hamsters were used in the exogenous metabolic activation system. Because Aroclor-1254 became unavailable for preparing S9 during the early stages of this study, the inducing agent was switched to a combination of phenobarbital and β-naphthoflavone (PB/BNF); S9s prepared with these inducing agents were used for all experiments conducted during the later stages of the study. Upon completing the testing of all compounds, several compounds initially tested using the S9s prepared from Aroclor-pretreated rodents, were retested with S9s prepared from rats and hamsters pretreated with PB/BNF (see Discussion). The S9s were stored at −80 °C and thawed on ice just before use.

2.1.4. Reagents and medium

Acetone was from Thermo Fisher (Waltham, MA; 423245000), dimethyl sulfoxide (DMSO) from Sigma-Aldrich (D8418), reagent water (for test agent solutions and S9 mix preparation) from Thermo Fisher (Molecular Biology Grade, BP2819–1), D-glucose-6-phospate sodium salt from Sigma-Aldrich (G7879), NADP monosodium salt from EMD Millipore USA (Burlington, MA; 48191), agar from Sigma (A1296) and Nutrient Broth No. 2 from Oxoid (Basingstoke, UK).

2.2. Ames testing methods

2.2.1. Solvents used for test substances and controls

Only solutions of the test substances were used for Ames testing, i.e., no suspensions were tested (with one exception noted in the Discussion). Aqueous test substance solutions were delivered at 100 μL per 700 μL preincubation reaction (see below), while organic solvents were used at 25 μL per reaction. Water was the solvent of choice. If a 50 mg/mL solution of the test substance could not be achieved using water, organic solvents were used, but with organic solvents the target maximum concentration for the treatment solution was 200 mg/mL. The preference for organic solvents favored acetone over methanol, with DMSO the last option. If the top concentration could not be made using water or any of the organic solvents, the highest possible solution concentration became the top concentration for conducting assays. For the test substances used in this study, NDMA, NDEA, and 1-methyl-4-nitrosopiperazine were dissolved in water; N-nitroso-varenicline, N-nitroso-folic acid and N-nitroso-ciprofloxacin were dissolved in DMSO. The remaining test substances were dissolved in acetone. See Discussion regarding further information on testing N-nitroso-folic acid.

For positive controls, water was used as the solvent for NDMA; acetone for CPNP, and DMSO was used as a solvent for 2AA, sodium azide, 2-nitrofluorene, MMS, and ICR-191. The same amounts of water and organic solvents were used in positive control preincubation reactions as used for the test substances. When different solvents were used for the test substance and one or both of the positive controls, the solvent used for the test substance served as the solvent used for the solvent control.

2.2.2. Positive and negative controls

Positive and negative controls were included with each assay; controls were used for each of the activation conditions (5 conditions) and both preincubation times (30 and 60 min). For S9-mediated positive controls, N-nitrosamines were favored as they demonstrate the metabolism relevant for activating N-nitrosamines. For all but the earliest tests that were conducted (e.g., with N-nitroso-desvaleryl-valsartan), CPNP was used as the S9-mediated positive control for TA1535 (500 μg/plate) and TA100 (1 mg/plate) and NDMA (5 mg/plate) was used for WP2 uvrA (pKM101). No small-molecule N-nitrosamine or NDSRI was known that could serve as an S9-mediated positive control for TA98 or TA1537; therefore, 2AA (10 μg/plate) was used as a positive control for these strains.

Positive controls for assays performed in the absence of S9 were strain-specific mutagens: sodium azide for TA1535 and TA100 (both 1 μg/plate), 2-nitrofluorene for TA98 (3 μg/plate), ICR191 for TA1537 (1 μg/plate), and MMS for WP2 uvrA (pKM101) (1 μl/plate).

2.2.3. Test substance concentrations

Initial studies were conducted in a single tester strain, generally TA1535, and consisted of five test article concentrations, solvent and positive controls, and a no-solvent/no-S9 (negative) control. The highest concentration of the test substance was 5 mg/preincubation mix, with lower doses spaced at 2- to 5-fold increments to the lowest concentration of 25 μg/preincubation. If the high concentration or concentrations produced cytotoxicity or a precipitate, the concentrations were adjusted so that (ideally) only one concentration produced toxicity or a precipitate. Assays producing unclear or unexpected responses were repeated using closer concentration spacing and often using additional test concentrations.

2.2.4. Preincubation reactions

Literature reports (e.g., Malling, 1971; Yahagi et al., 1977; Guttenplan, 1987; Gatehouse et al., 1994), recommendations in TG 471 (OECD, 2020), and confirmatory testing at NCTR (data not shown) indicate that the preincubation version of the Ames test is more sensitive to the mutagenicity of model small-molecule N-nitrosamines, e.g., NDMA, than the plate incorporation protocol. Thus, the preincubation version of the Ames assay was used for all of the assays reported here.

Other reports indicate that liver S9s from animals pretreated with inducers of cytochromes P450 (Cyps), like Aroclor-1254 or phenobarbital (PB), enhance the mutagenicity of small-molecule N-nitrosamines; that S9 mixes containing relatively high S9 concentrations (e.g., 30%) are more effective than the more commonly used 4%–10% S9 at activating NDMA to a mutagen; and that preincubations for longer than the standard 20–30 min also enhance detecting the mutagenicity of NDMA (Bartsch et al., 1975; Yahagi et al., 1977; Guttenplan and Bliznakov, 1981; reviewed in Guttenplan, 1987). In addition, reports indicate that liver S9 from hamsters produces greater mutagenic responses with small-molecule N-nitrosamines than the more commonly used S9s prepared from rat liver (Raineri et al., 1981; Lijinsky and Andrews, 1983; Araki et al., 1984; reviewed in Guttenplan, 1987). Thus, preincubations were set up using no S9, and S9 mixes containing 10% and 30% rat liver S9, and 10% and 30% hamster liver S9 (5 conditions; all S9s from animals pretreated with Cyp inducers) so that the test substances assayed in this study were tested both under conditions reported to enhance detecting the mutagenicity of model small-molecule N-nitrosamines, as well as under more standard conditions.

To construct the preincubation reactions, sufficient volumes of each of the five S9 mixes (no S9, 10% and 30% rat liver S9, 10% and 30% hamster liver S9: 0.5 mL for each assay) were made to conduct all the assays in an experiment, with each incubation condition conducted in triplicate. The several lots of S9 used for this project contained approximately 33–39 mg/ml of S9 protein; thus S9 mixes containing 10% S9 contained approximately 3.6 mg/ml of S9 protein while S9 mixes containing 30% S9 contained approximately 10.8 mg/ml of S9 protein. The final concentrations of the components in the S9 mixes are shown in Table 3; the mixes were kept on ice until used to assemble the final preincubation reaction mixes.

Table 3. Concentrations of components in the preincubation mix used to evaluate the bacterial mutagenicity of small molecule N-nitrosamines and NDSRIs.

The first 5 components were added as concentrated stock solutions (stored at room temperature, 4 °C or −30 °C, as appropriate). Water was added to the mix to make up any needed volume before addition of the freshly thawed S9; an additional 75 μL of water (if needed); 100 μl of an overnight bacterial tester strain culture and 25 μl or 100 μl of freshly prepared test compound or vehicle were then added in order.

Component	Stock solution concentration	Final volume/concentration
S9 mix		500 μL
β-Nicotinamide-adenine dinucleotide phosphate	0.1 M	4 mM
Potassium chloride	1.65 M	33 mM
Glucose-6-phosphate	1 M	5 mM
Magnesium chloride	0.4 M	8 mM
Phosphate Buffer (pH 7.4)	0.2 M	100 mM
Rat or hamster S9 homogenate	33–39 mg protein/ml	0%, 10% or 30% (v/v)
Water (if test substance is dissolved in an organic solvent)		75 μL
Overnight culture of bacterial tester strain		100 μL
Vehicle or test compound		25 μL (if testing a substance dissolved in an organic solvent) 100 μL (if testing a substance dissolved water)

When the test substance was dissolved in an organic solvent, 0.5 mL of the S9 mix were combined in a 2.0-mL blunt-bottomed polypropylene microcentrifuge tube (EZ BioResearch, St. Louis, MO; D1009), in order, with 0.1 mL of a late-log culture of tester strain, 75 μL H₂O (this volume was typically included with the 0.5 mL of S9 mix to reduce the number of pipetting steps), and 25 μL of the vehicle or the vehicle containing the test substance. Following these additions, the contents of the tubes were mixed rapidly and incubated (see below). Note that the ‘% S9’ concentrations shown in Table 3 and referred to in this report are the concentrations in the S9 mix, not the final concentrations in the preincubation mix. Thus, 10% S9 in the S9 mix is 7.1% S9 in the preincubation mix, and 30% S9 in the S9 mix is 21% in the preincubation mix.

When the test substance was dissolved in water, 0.5 mL of the S9 mix were combined, in order, with 0.1 mL of a late-log culture of tester strain, and 0.1 mL of water or water containing the test substance.

Preincubations were conducted for both 30 and 60 min using a FINEPCR shaker-incubator (Gyeonggi-do, Korea; model confide-S202H) set at 37 °C and 300–350 rpm.

2.2.5. Plating, incubation, and colony-counting

Following preincubation for 30 or 60 min, the contents of individual tubes were transferred, one at a time using a manual pipettor, into Fisherbrand 13 × 100 mm borosilicate glass tubes containing 2.0 mL molten, 45 °C top agar [0.6% agar, 0.45% NaCl, and 0.05 mM tryptophan for WP2 uvrA (pKM101) or 0.6% agar, 0.45% NaCl, and 0.05 mM histidine/biotin for the Salmonella strains]. The suspensions were mixed immediately and poured onto 100 mm Minimal Glucose Agar plates (MolTox 21–400.5). The plates were gently swirled, allowed to solidify, and then inverted, agar-side-up and incubated for 2 days at 37 °C. Occasionally the plates were incubated for an additional day or two at room temperature to promote revertant colony growth.

The condition of the bacterial background lawn was evaluated visually for evidence of test article toxicity. Signs of cytotoxicity included: a decrease in the number of revertant colonies to 50% or less relative to the colony counts in the vehicle control, a down-turn in a positive concentration response at higher test substance concentrations, and/or the occurrence of a thinned background bacterial lawn, possibly with ‘pin-point’ colonies or no colonies at all. Precipitate was evaluated by visual examination during the construction of the preincubation mixes, during plate preparation, and after the incubation period.

Revertant colonies were enumerated with a ProtoCOL3 (version 1.0.27.0) automated plate counter (Synbiosis USA, Frederick, MD). Occasionally colonies were counted by hand to verify counts, and when it was suspected that precipitate or toxicity may have interfered with automated counting. Data from the plate counter were transferred to an Excel spreadsheet. Occasional plates in which revertant counts were confounded by toxicity, precipitate or obscured by contaminating microorganisms were noted, and generally not included in the analysis of mutagenicity. The spreadsheet also was used to record details of the assay, including S9 source, test agent, and reagents used.

2.2.6. Assay acceptance

When the assays produced toxicity in the tester strains or a precipitate, test doses were adjusted and assays repeated. Ideally, after repeat testing, only the top dose produced clear signs of toxicity or precipitation, leaving a minimum of four interpretable doses per assay.

Some of the preincubation conditions used for the study were unusual, and historical control data bases for these conditions initially were not available at NCTR. Thus, the responses in the vehicle controls were compared with literature values (e.g., Maron and Ames, 1983; Levy et al., 2019) to confirm they were reasonable for individual strains. Comparison to historical control distributions was not used as a criterion for a positive response in this study. In general, positive controls were required to be positive, using the fold-rules described below, for an assay to be considered acceptable.

2.2.7. Data evaluation

For each replicate plating, the mean and standard deviation of the number of revertants per plate were calculated and summary data are reported in Supplementary Material.

For the test substance to be evaluated positive, two criteria must have been met:

1. The test substance must have caused a concentration-related increase in the mean revertants per plate of at least one tester strain using at least one S9 activation condition over a minimum of two increasing concentrations of the test substance.

2. A positive response also was determined by applying the following fold-rules:

   1. For strains TA1535 and TA1537: Data sets were judged positive if the increase in mean revertants per plate at the peak of the concentration response was equal to or greater than three-times the concurrent mean vehicle control value.
   2. For strains TA98, TA100, and WP2 uvrA (pKM101): Data sets were judged positive if the increase in mean revertants per plate at the peak of the concentration response was equal to or greater than two-times the concurrent mean vehicle control value.

An equivocal response is a biologically relevant increase in a revertant count that partially meets the criteria for evaluation as positive. This could be a concentration-related increase that does not achieve the respective fold-increase threshold cited above or a non-concentration-related increase that is equal to or greater than the respective threshold cited. Additional assays were often performed when an equivocal response was suspected.

A response was evaluated as negative if neither of the criteria for a positive response was met.

2.2.8. Determining the effect of assay conditions on mutagenicity responses

When results under various test conditions for a given test substance were compared, the conditions in which positive responses occurred were considered more effective at mediating the mutagenicity of the test substance than conditions producing negative responses. For instance, when a test substance was positive with 30% hamster S9 but produced a negative response with 10% hamster S9, S9 mixes containing 30% hamster S9 were considered more effective at mediating the mutagenicity of the test substance. When the responses being compared were all judged to be positive using the criteria stated above, mutagenicity dose-response ranking (Guo et al., 2016; Wills et al., 2016; Mittelstaedt et al., 2021) was conducted on the data using web-based PROAST software, version 70.1 (available at https://proastweb.rivm.nl/). The covariate approach and a Critical Effect Size (CES) of 1.0 (i.e., 100% increase or two-fold relative to the vehicle control) were used on dose-response curves fit separately by the Hill and exponential models. Analyses were constructed to evaluate the effect of the type of activation and length of preincubation on the extent of mutagenicity in tests conducted with a single tester strain. Differences were based on the occurrence of non-overlapping confidence intervals for the benchmark concentrations estimated from the concentration responses.

In addition, positive responses in different tester strains were compared with the understanding that these comparisons do not always reflect the magnitude of the response since the negative control frequencies of the tester strains differ substantially. Tester strain comparisons were made only with the metabolic activation conditions producing the greatest mutagenic response (in most cases, 30% hamster S9).

The source of S9 (rat or hamster) was considered to have an effect on mutagenicity when either 10% or 30% (but not necessarily both) rat or hamster S9 was more effective at mediating mutagenicity.

2.2.9. Data storage and reports

All raw data are stored in Excel spreadsheets. PROAST analysis reports are available upon request.

3. Results

3.1. Overall responses

A summary of test responses is given in Table 4 and shown in charts in Figs. 1 and 2; more detailed test data can be found in Supplementary Material.

Table 4.

Summary of responses observed in Ames tests conducted on small molecule N-nitrosamines and NDSRIs.

ID	Test substance	Ames response	Sensitive tester strains (greatest first to least last)	Effectiveness of hamster vs. rat S9s for each positive strain	More effective %S9 in S9 mix of more effective S9 (H or R)	Effectiveness of time of preincubation for most effective S9
Small molecule N-nitrosamine drug impurities and model substances
SM-1	N-nitroso-dimethylamine (NDMA)	Positive	TA1535 =	H > R	30% > 10%	60min >30min
			TA100 ≥	H > R	30% > 10%	60min >30min
			WP2uvrA(pKM101)	H > R	30% > 10%	60min = 30min
SM-2	N-nitroso-diethylamine (NDEA)	Positive	WP2uvrA(pKM101) ≥	H > R	30% > 10%	60min = 30min
			TA1535 >	H > R	30% > 10%	60min = 30min
			TA100 ≥	H > R	30% > 10%	60min = 30min
			TA98	H > R	30% > 10%	60min = 30min
SM-3	N-nitroso-isopropylethylamine (NIPEA/ NEIPA)	Positive	TA1535 =	H > R	30% = 10%	60min = 30min
			WP2uvrA(pKM101)	H > R	30% > 10%	60min = 30min
SM-4	1-methyl-4-nitrosopiperazine (MNP)	Positivea	TA1535 >	H > R	30% = 10%	60min = 30min
			TA100	H > R	30% > 10%	60min = 30min
SM-5	N-nitroso-diisopropylamine (NDIPA)	Positive	TA1535 only	H only	30% only	60min = 30min
SM-6	N-nitroso-methylphenylamine (NMPA)	Positive	WP2uvrA(pKM101) > TA1537	H > R	30% ≥ 10%	60min = 30min
				H only	30% = 10%	60min = 30min
SM-7	N-nitroso-N-methyl-4-aminobutyric acid (NMBA)	Negative
SM-8	N-nitroso-N-ethylaniline (NEPA)	Positive	WP2uvrA(pKM101) only	H = R	30% > 10%	60min ≥ 30min
SM-9	N-nitroso-dibutylamine (NDBA)	Positive	TA1535 >	H > R	30% = 10%	60min = 30min
			WP2uvrA(pKM101) ≥	H ≥ R	30% ≥ 10%	60min ≥30min
			TA100	H ≥ R	30% ≥ 10%	30min ≥60min
SM-10	1-cyclopentyl-4-nitrosopiperazine (CPNP)	Positive	TA1535 >	H > R	30% = 10%	60min = 30min
			TA100 >	H > R	30% = 10%	30min ≥60min
			WP2uvrA(pKM101)	H only	30% ≥ 10%	60min = 30min
SM-11	N-nitroso-diphenylamine	Negative
SM-12	N,N-bis(2,2-diethoxyethyl) nitrous amide	Positive	TA1535 only	H only	30% only	60min = 30min
Nitrosamine drug substance-related impurities (NDSRIs)
NDSRI-13	N-nitroso-phenylephrine	Negative
NDSRI-14	N-nitroso-lorcaserin	Positive	TA1535 ≥	H ≥ R	30% > 10%	60min = 30min
			WP2uvrA(pKM101) ≥	H > R	30% ≥ 10%	60min = 30min
			TA100 =	H ≥ R	10% ≥ 30%	60min = 30min
			TA98	H ≥ R	30% ≥ 10%	60min = 30min
NDSRI-15	N-nitroso-varenicline	Positive	TA1535 =	H = R	30% = 10%	60min ≥30min
			TA1537 ≥	H = R	30% = 10%	60min ≥30min
			TA100 ≥	H only	30% ≥ 10%	60min = 30min
			WP2uvrA(pKM101) ≥	H > R	30% > 10%	60min = 30min
			TA98	H ≥ R	30% ≥ 10%	60min ≥30min
NDSRI-16	N-nitroso-desmethyl-diphenhydramine	Positive	TA1535 ≥	H > R	30% = 10%	60min = 30min
			WP2uvrA(pKM101) ≥	H > R	30% ≥ 10%	60min >30min
			TA100	H > R	30% ≥ 10%	60min = 30min
NDSRI-17	N-nitroso-propranolol	Positiveb	TA1535 =	H > R	30% = 10%	60min = 30min
			TA98 ≥	H = R	30% = 10%	60min = 30min
			TA100	H > R	30% = 10%	60min = 30min
NDSRI-18	N-nitroso-nortriptyline	Positive	TA1535 ≥	H > R	30% > 10%	60min = 30min
			WP2uvrA(pKM101) >	H > R	30% > 10%	60min = 30min
			TA100	H > R	30% > 10%	60min = 30min
NDSRI-19	N-nitroso-diclofenac	Negative
NDSRI-20	N-nitroso-duloxetine	Positive	TA1535 >	H > R	30% = 10%	60min = 30min
			WP2uvrA(pKM101) >	H ≥ R	30% ≥ 10%	60min = 30min
			TA100	H > R	30% > 10%	30min >60min
NDSRI-21	N-nitroso-sertraline	Positive	WP2uvrA(pKM101) only	H only	30% = 10%	30min >60min
NDSRI-22	N-nitroso-fluoxetine	Positive	TA1535 =	H only	30% only	60min = 30min
			WP2uvrA(pKM101) ≥	H ≥ R	30% ≥ 10%	30min >60min
			TA100 >	H only	30% only	30min only
			TA98	H only	30% only	30min only
NDSRI-23	N-nitroso-paroxetine	Negative
NDSRI-24	N-nitroso-ciprofloxacin	No testc
NDSRI-25	N-nitroso-desvaleryl-valsartan	Negative
NDSRI-26	N-nitroso-bumetanide	Negative
NDSRI-27	N-nitroso-desvaleryl-valsartan methyl ester	Negative
NDSRI-28	N-nitroso-folic acid	Negatived
NDSRI-29	N-nitroso-dabigatran etexilate	Negative

H = Hamster liver S9; R = rat liver S9; 10% = 10% S9 in the S9 mix; 30% = 30% S9 in the S9 mix.

^aEquivocal direct-acting mutagen in TA1535.

^bDirect acting mutagen in TA98.

^cHighly toxic to tester strains, poorly soluble in all vehicles.

^dWeakly positive/equivocal when tested as an aqueous suspension: see text and Supplementary Material.

Fig. 1. Performance of Ames tester strains in detecting the mutagenicity of N-nitrosamines.

(A) Heatmap shows the number of test conditions that detected positive responses with each tester strain. Small-molecule nitrosamines and NDSRIs are listed separately in the order of their molecular weight. A total of 10 Ames test conditions (5 S9 conditions 2 preincubation time) were evaluated for each strain. (B) Summary of the total number of positive small-molecule nitrosamines and NDSRIs detected by each tester strain.

Fig. 2. Performance of different S9 conditions in detecting the mutagenicity of N-nitrosamines.

(A) Heatmap showing the number of test conditions that detected positive response with each S9 condition. Small-molecule nitrosamines and NDSRIs are listed separately in the order of their molecular weight. A total of 10 Ames test conditions (5 tester strains × 2 preincubation time) were evaluated for each S9 activation condition. (B) Summary of the total number of Ames-positive small-molecule nitrosamines and NDSRIs detected by each S9 activation condition.

Ten of 12 small molecule nitrosamines (83%) and eight of 16 ‘testable’ NDSRIs (50%) were positive in the Ames test under one or more of the 50 combinations of tester strain, activation condition, and preincubation times that were used. One NDSRI (N-nitroso-ciprofloxacin) was difficult to test because of its bacterial toxicity and is not included in the NDSRIs that have test data.

3.2. Effect of protocol variables

3.2.1. Tester strain

Our choice of tester stains was based on the recommendations in TG 471 (OECD, 2020) and included four tester stains originally recommended by Maron and Ames (1983) that detect basepair substitutions and frameshifts targeting G:C basepairs plus the A:T-specific strain WP2 uvrA (pKM101). Three of the strains [TA98, TA100 and WP2 uvA (pKM101)] possess the error-prone by-pass repair function encoded in pKM101, while two of the strains (TA1537 and TA1535) lack this repair function. WP2 uvrA (pKM101) was chosen based on the findings of Zielenska and Guttenplan (1988) for reasons given in the Discussion.

In all but three cases, test substances that were positive in the Ames test were positive in TA1535 (Fig. 1B). In most cases, TA1535 also was the most sensitive tester strain, or among the most sensitive tester strains for detecting the mutagenicity of the test substances evaluated in this study (Table 4). For two of the small-molecule nitrosamines, N-nitroso-diisopropylamine (NDIPA) and N,N-bis(2,2-diethoxyethyl)nitrous amide, TA1535 was the only positive strain. The three Ames-positive test substances that were negative in TA1535 were N-nitroso-methylphenylamine [NMPA; positive in WP2 uvrA (pKM101) and TA1537], and N-nitroso-N-ethylaniline (NEPA) and N-nitroso-sertraline [both positive only in WP2 uvrA (pKM101)]. Only one test substance, the NDSRI N-nitroso-varenicline, produced positive responses with all five tester strains used for the study.

The frequency of strains mutated by Ames-positive test substances was: TA1535, 15 out of 18 (eight of 10 positive small-molecule N-nitrosamines and seven of eight positive NDSRIs); WP2 uvrA (pKM101), 14 out of 18 (eight out of 10 and six out of eight); TA100, 11 out of 18 (four out of 10 and seven out of eight); TA98, five out of 18 (one out of 10 and four out of eight); and TA1537, two out of 18 (one out of 10 and one out of eight). None of the positive small-molecule N-nitrosamines or NDSRIs were positive only in TA100, TA98 or TA1537. The frequency of positive responses for mutagenic NDSRIs increased from 50% (eight out of 16) for strains without the pKM101 by-pass repair function to 71% (17 out of 24) in tester strains having the pKM101 plasmid. In comparison, the frequency of positive responses was similar for small-molecule N-nitrosamines in strains with and without the pKM101 plasmid (45%, 9 of 20 for strains without pKM101 compared with 43%, 13 of 30, for strains with pKM101). Note that the only strain with a target detecting mutations at A:T base pairs also had the pKM101 plasmid [WP2 uvrA (pKM101)], which could have contributed to the mutagenicity associated with plasmid-bearing strains for NDSRIs.

All tester strains in the set were UV-excision-repair deficient; however, data can be analyzed relative to the G:C or A:T target of the strains. Many of the nitrosamines induced mutations in both strains with G:C targets and the one strain with an A:T target. However, four test substances were positive only in G:C detecting strains [1-methyl-4-nitrosopiperazine, N-nitroso-diisopropylamine (NDIPA), N,N-bis(2,2-dietyoxyethyl)nitrous amide and N-nitroso-propranolol], while two were positive in only the A:T-detecting strain (N-nitroso-N-ethylaniline and N-nitroso-sertraline).

3.2.2. Type of S9

One NDSRI, N-nitroso-propranolol, was a direct acting mutagen in strain TA98; N-nitroso-propranolol was also mutagenic in TA1535 and TA100, but with S9 activation. One small-molecule nitrosamine, 1-methyl-4-nitrosopiperizine, was an equivocal direct-acting mutagen in TA1535, but strongly positive in TA1535 and TA100 with S9 activation. All other NDSRIs and small-molecule N-nitrosamines required S9 activation to produce a positive response.

In general, hamster liver S9 produced positive mutagenicity responses that were greater than those produced by rat liver S9 (Table 4). With two small-molecule N-nitrosamines, NDIPA and N,N-bis(2,2-diethoxyethyl)nitrous amide, positive responses were found only with hamster liver S9 (Fig. 2). Rat liver S9 did not produce greater responses for any of the test substances that were positive, and in no case was a test substance positive only with rat liver S9, even considering responses for a single tester strain (but see Discussion on N-nitroso-folic acid).

3.2.3. Concentration of S9

In the vast majority of cases, preincubations with 30% S9 produced greater mutagenic responses than preincubations with 10% S9. In two cases, with NDIPA and N,N-bis(2,2-diethoxyethyl)nitrous amide, the positive mutagenic responses were dependent upon preincubation with 30% hamster S9.

3.2.4. Length of preincubation

Although 30-min preincubations were sufficient to detect positive responses for all the positive test substances, the positive responses for NDMA, N-nitroso-ethylaniline, N-nitroso-varenicline and possibly N-nitroso-desmethyl-diphenhydramine benefited from a preincubation of 60 min compared with 30 min. In several cases, however, at least some of the assays for N-nitroso-sertraline and N-nitroso-fluoxetine produced greater mutagenic responses with 30-min preincubations than with 60-min preincubations. Our findings indicate that none of the positive responses were totally dependent upon a 30- or 60-min preincubation.

4. Discussion

The design of our current study emphasized the evaluation of several aspects of the Ames test in a detailed, systematic manner. The emphasis in our study was placed on the source and concentration of S9 used for activation, the length of the preincubation step, and the most effective tester strains for testing small-molecule N-nitrosamines and NDSRIs. We used only the preincubation version of the test, based on the findings of Araki et al. (1984) and the discussions in Gatehouse et al. (1994) related to testing nitrosamines. Also, we used the preincubation rather than plate-incorporation test due to well-known effect of the plate-incorporation test on limiting the mutagenicity of NDMA (e.g., McCann et al., 1975; Andrews et al., 1984), and the presumption that the preincubation assay should be at least as effective as the plate-incorporation test, potentially exposing the tester strains to higher concentrations of the test substance and its metabolites (Prival et al., 1979; Guttenplan, 1987; Gatehouse et al., 1994). It is recognized, however, that this higher concentration of test substance occurs at the expense of exposing the tester strains and metabolic activation system to higher concentrations of the organic solvents used as test compound vehicles for a longer period of time [commented on by Maron et al. (1981)].

Because of the increased potential for solvent toxicity (e.g., Ames et al., 1975), we used no more than 3.6% of organic solvent (acetone or DMSO, but usually acetone) in the preincubation mix (i.e., 25 μL solvent in a 700 μL preincubation mix). As advised by Maron et al. (1981), the solvent was added last to the preincubation mix. Preliminary studies on the mutagenicity of NDMA (a water-soluble N-nitrosamine) found that 25 μL of acetone had minimal effects on its mutagenicity (data not presented). This observation is consistent with those of Shibata et al. (2020), who report that 25 μL of acetone have little effect on both the viability of Ames tester stains and the mutagenicity of a variety of different chemical mutagens in the preincubation assay, including one N-nitrosamine, N-nitroso-pyrrolidine. Thus, solutions of test substances were delivered in 25 μL of organic solvents as a practical compromise between retaining bacterial viability and metabolic activity in preincubation mixes and accurately delivering the test compound using manual pipetting.

In addition, for all but a few compounds, we avoided using DMSO as a solvent, because of its well-known effect on the mutagenicity of NDMA (Yahagi et al., 1977). It is recognized, however, that the inhibition of mutation responses by DMSO has been found only for a small number of N-nitrosamines so far. For instance, we found that delivering N-nitroso-propranolol, N-nitroso-fluoxetine, NDBA, N-nitroso-desmethyl-diphenhydramine and 1-methyl-4-nitrosopiperazine in 25 μL of DMSO, methanol, or acetone resulted in similar mutagenic responses for each test substance (data not presented).

Prival et al. (1979), Raineri et al. (1981), Lijinsky and Andrews (1983) and Araki et al. (1984) found that hamster liver S9 can be much more effective than rat liver S9 in mediating the mutagenicity of N-nitrosamines. Therefore, we compared the effectiveness of hamster liver S9 with the commonly employed rat liver S9. In addition, early studies indicate that the extent of N-nitrosamine mutagenicity in the Ames test is directly proportional to the amount of S9 used for the assays (Bartsch et al., 1975; Prival et al., 1979). Thus, we used two concentrations of S9 in the S9 mixes used for the preincubations, the commonly used 10% and less commonly used 30%. Finally, the liver S9s used in these studies were prepared from rats and hamsters pretreated to induce liver enzyme expression. In our initial studies, the enzyme inducer was Aroclor 1254. During the course of our testing, S9 from rodents treated with Aroclor 1254 became unavailable, and we changed to S9 prepared from rodents pretreated with a combination of phenobarbital and β-naphthoflavone (PB/BNF).

Besides having a lower environmental impact, evidence indicates that PB/BNF induces a similar spectrum of liver enzymes as Aroclor 1254 and that S9s prepared from the livers of rodents treated with PB/BNF have a similar ability to mediate the mutagenicity of test compounds in the Ames test as liver S9s prepared from rodents treated with Aroclor 1254 (Callander et al., 1995). We confirmed this for several of the test substances tested early in the study with Aroclor-1254-induced S9s (i.e., NDMA, 1-methyl-4-nitrosopiperizine, CPNP, N-nitroso-varenicline, N-nitroso-lorcaserin, and N-nitroso-fluoxetine). Whereas quantitative differences were found for individual tester strain/activation combinations (see individual assay data files and summary table in Supplementary Material), S9-mediated test results for these six small-molecule nitrosamines and NDSRIs in terms of positive or negative responses in the different tester strains were qualitatively similar. These comparisons should be interpreted carefully since they were made using data from independent assays, in most cases run several years apart, and not in side-by-side comparisons. Overall, however, it appeared as if the PB/BNF-induced S9s were at least as competent at mediating positive mutagenic responses in the Ames assay as were the Aroclor-induced S9s. Of the 105 combinations of tester strain, type of S9 and S9 concentration that were compared (no-S9 responses not included; see table in Supplemental Material), 92 (88%) had the same qualitative response, whereas for 3 (3%), Aroclor-induced S9 produced a greater response, and for 10 (10%), PB/BNF produced a greater response. Nearly half of the differences came in the testing of one NDSRI, N-nitroso-varenicline, an unusual test substance in that it produced positive mutagenic responses in all five tester strains employed in the study (Supplementary Material; Table 4).

In general, we found that hamster S9 and S9 mixes containing 30% rather than 10% rat or hamster S9 were more effective at mediating the mutagenicity of N-nitrosamines, including NDSRIs, in the Ames test (Table 4, Fig. 2). In addition, of the 28 small-molecule nitrosamines and NDSRIs for which test data were generated, two small-molecule nitrosamines, NDIPA and N,N-bis(2,2-diethoxyethyl)nitrous amide, and one NDSRI, N-nitroso-sertraline, were exclusively mutagenic with hamster liver S9. It is true that from a hazard ID, positive-negative perspective, rat liver S9 identified nearly as many N-nitrosamines as mutagenic as did hamster liver S9 using the assay conditions that we employed (Fig. 2); however, in most cases hamster liver S9 produced stronger responses than rat liver S9, resulting in higher mutant yields overall and positive responses at lower test substance concentrations (Table 4; Supplementary Material).

Of the small-molecule nitrosamines and NDSRIs that were assayed, only N-nitroso-propranolol, and possibly 1-methyl-4-nitrosopiperizine, were direct-acting mutagens. The direct-acting mutagenicity of N-nitroso-propranolol was limited to TA98, a frameshift detecting strain, and generally not a strain that is highly sensitive to the mutagenicity of N-nitrosamines (Yahagi et al., 1977; see Fig. 1). N-Nitroso-propranolol, however, was also mutagenic with S9 activation in tester strains that are mutated more commonly by N-nitrosamines, i.e., TA1535 and TA100, and typical of other N-nitrosamines, hamster liver S9 was more effective than rat liver S9 in mediating these responses. In addition, mutagenicity studies with mammalian cell assays indicate that N-nitroso-propranolol is genotoxic in TK6 cells but only in the presence of S9 or endogenous metabolic activation provided by TK6 cells transduced with human CYPs, indicating a requirement for CYP activation to produce a mutagenic response (Li et al., 2023). Thus, it is conceivable that the mutagenicity of N-nitroso-propranolol in the Ames test occurs via two separate pathways - one resulting in direct-acting mutagenicity (e.g., either through non-enzymatic degradation to a reactive substance or through activation by bacterial metabolism) and the other mediated by liver enzyme activation though a mechanism more typical of N-nitrosamines (e.g., reviewed by Guttenplan, 1987; Li and Hecht, 2022; Snodin et al., 2024). Further discussion on the mutagenicity of N-nitroso-propranolol and detailed mutagenicity data can be found in a separate publication (Li et al., 2023).

1-Methyl-4-nitrosopiperizine was a weakly positive direct-acting mutagen in two of four assays conducted using TA1535, and in the other two assays in TA1535, there was evidence of a concentration-dependent increase in mutagenicity (Table 4; see data in Supplementary Material). Based on these observations, 1-methyl-4-nitrosopiperizine was classified as ‘equivocal’ as a direct-acting mutagen. 1-Methyl-4-nitrosopiperizine was, however, a strong S9-mediated mutagen in both TA1535 and TA100, and thus its overall classification for mutagenicity was ‘positive’ (Table 4). Although only summary results were provided, 1-methyl-4-nitrospiperizine does not appear to be a direct-acting mutagen in either yeast or V79 cells (Kuroki et al., 1977; Larimer et al., 1980; Table 2), suggesting that, as appears to be the case for N-nitroso-propranolol, the direct-acting mutagenicity observed in the Ames test may not be relevant for mammalian cells.

Based on studies indicating that the mutagenicity of N-nitrosamines was dependent on the length of the preincubation step (Bartsch et al., 1975, 1976; Yahagi et al., 1977; Guttenplan and Bliznakov, 1981), we performed preincubations for both 30 and 60 min. The mutagenicity of NDMA benefited from the longer incubation time; however, overall, few differences were found for preincubations of 30 and 60 min. In fact, the mutagenicity of N-nitroso-fluoxetine appeared to be stronger with the shorter preincubation time (Table 4). This result with N-nitroso-fluoxetine may be due to enzyme stabilization by the agar matrix and/or the increased oxidative conditions that occur when the reaction mix is plated. Bartsch et al. (1976) indicate that conditions on the plate tend to stabilize the activity of S9, making it metabolically active longer than occurs with liquid preincubation reactions.

Early studies indicated that most small-molecule N-nitrosamines are mutagenic in Salmonella tester strains having the hisG46 target, including TA1535 (e.g., Rao et al., 1977, 1978, 1979; Guttenplan, 1987), which detects basepair substitutions targeted to G:C basepairs (Maron and Ames, 1983; Hartman et al., 1986). All but two of the 10 small-molecule nitrosamines that tested positive in our studies were positive in this tester strain. TA100 has the same G46 target as TA1535, but it also has a higher spontaneous mutant frequency, which may account for its lower sensitivity for detecting the mutagenicity of N-nitrosamines (Fig. 1). It is clear, however, that even though mutation at A: T basepairs was not commonly evaluated in early studies, that N-nitrosamines also can be mutagenic in tester strains detecting mutations at A: T basepairs (Elespuru and Lijinsky, 1976, 1978; Zielenska and Guttenplan, 1988; Fig. 1).

In our study, two mutagenic small-molecule nitrosamines that were not positive in TA1535, NMPA and NEPA, were both mutagenic in E. coli WP2 uvrA (pKM101), a strain detecting mutation at an A:T basepair. Zielenska and Guttenplan (1988) observed that the mutagenicity of NMPA was detected most efficiently by strains that detect basepair substitution at A:T, are deficient in excision repair, and possess the pKM101 plasmid. They demonstrated this using TA104, an A:T-specfic Salmonella typhimurium tester strain that is DNA-repair deficient (uvrB) and possesses the pKM101 plasmid. Although WP2 uvrA (pKM101) is an E. coli tester strain, it has similar properties as TA104 in that it also detects basepair substitutions at A:T basepairs, is excision repair deficient, and has the error-prone lesion by-pass functions provided by the pKM101 plasmid. NMPA was only weakly mutagenic in TA100, which also is DNA-repair deficient and has the pKM101 plasmid, suggesting that it is the A:T basepair substitution mutation target, rather than the G: C-specific target in TA100, that is critical for NMPA mutagenesis. For our study, we used the version of WP2 uvrA that contains the pKM101 plasmid, based mainly on the observations of Zielenska and Guttenplan (1988). By the same logic used above for explaining the differential sensitivity of TA1535 vs. TA100, it is conceivable that there could be merit in using WP2 uvrA, which has the same target for detecting mutation as WP2 uvrA (pKM101), but has a lower background mutant frequency, in addition to (or in replacement of) the plasmid-containing strain. This possibility, however, was not pursued in this study.

In addition to its mutagenicity in WP2 uvrA (pKM101), we found that NMPA induced an exceedingly weak, positive response in TA1537, one of the few N-nitrosamines that was positive in this tester strain. It is possible that the structure of NMPA was responsible for its mutagenicity in TA1537, which is sensitive to mutagenesis at repeating base sequences. In addition, Snodin et al. (2024) have discussed an activation pathway by which NMPA might be mutagenic in TA1537.

Of the eight mutagenic NDSRIs, the only mutagenic NDSRI that was not positive in TA1535 was N-nitroso-sertraline, which, like NMPA and NEPA, was mutagenic in WP2 uvrA (pKM101). As indicated above, unlike TA1535, WP2 uvrA (pKM101) is sensitive to mutation at A:T basepairs.

Comparison of Ames assay responses to cancer findings was not a major goal for our study; however, cancer bioassay data were available for many of the small-molecule N-nitrosamines (Table 2). Most of small-molecule nitrosamines with positive cancer findings, also were Ames-positive in the current study, using one condition or another. The exceptions were N-nitroso-N-methyl-4-aminobutyric acid (NMBA) and N-nitroso-diphenylamine. N-Nitroso-diphenylamine has no α-hydrogens, making it impossible to activate N-nitroso-diphenylamine by α-carbon hydroxylation, the primary activation pathway for the most potent small-molecule N-nitrosamines (Keefer et al., 1973; Elespuru, 1978; Hecker et al., 1979; Lijinsky, 1986; Cross and Ponting, 2021; Li and Hecht, 2022; Snodin et al., 2024). Also, N-nitroso-diphenylamine has the weakest carcinogenic potency (TD₅₀) of all carcinogenic N-nitrosamines for which a potency has been calculated (Cross and Ponting, 2021). In addition, the International Agency for Research on Cancer (IARC) places N-nitroso-diphenylamine in Group 3 in the IARC classification groupings, meaning it is unclassifiable as to its risk for human cancer. Both NMBA and N-nitroso-diphenylamine have tested positive for mutagenicity in Salmonella, but only when using extraordinary methods (see Table 2). NMBA may be metabolized to potentially mutagenic and carcinogenic derivatives either by enzymic pathways that are not present in liver S9 fractions (Janzowski et al., 1994) or possibly because cofactors to support those pathways were not part of the S9 mix. This may be one of the reasons why NMBA tests positive for genotoxicity in assays having endogenous metabolic activity (see Notes in Table 2).

Thus, it is possible that further modifications to the testing protocol might increase the sensitivity of the Ames assay for N-nitrosamines like N-nitroso-diphenylamine and NMBA. Many potential modifications to the basic Ames test remain to be evaluated. For instance, it has been reported that employing a slightly acidic pH for the preincubation mix can increase the sensitivity of the assay for detecting the mutagenicity of some small-molecule N-nitrosamines (Guttenplan, 1979, 1980; Negishi and Hayatsu, 1980; reviewed in Guttenplan, 1987). Other studies have reported on nitrosamine mutagenicity using bacterial strains expressing individual Cyps and deficient in additional DNA repair functions (Kamataki et al., 1999; Kushida et al., 2000; Cooper and Porter, 2000, 2001; Fujita and Kamataki, 2001), and using a chemical activating system rather than S9 to mediate mutagenicity (Inami et al., 2013; Okochi et al., 1995). These modifications are not generally used for conducting the Ames test and were not evaluated in the current study.

In addition, our Ames test findings for NDSRIs are consistent with categorization of N-nitrosamines using the Carcinogenic Potency Categorization Approach (CPCA), a 5-category system that ranks N-nitrosamines based on their chemical structure (EMA, 2023; FDA, 2023; Health Canada, 2024; Kruhlak et al., 2024). Although the number of tested NDSRIs falling into each category are small, the two NDSRIs predicted to be among the most potent carcinogens by CPCA categorization (Potency Category 1 or PC1), were both Ames positive in our studies (Table 5). In contrast, N-nitroso-diclofenac was the only NDSRI falling into the CPCA category containing the least potent carcinogens (PC5) and it was Ames negative. Also, the overall proportion of Ames positive N-nitrosamines decreased as the CPCA predicted carcinogenic potency deceased (Table 5). Only six small-molecule N-nitrosamines that we tested had CPCA PCs listed in the FDA database (FDA, 2024a). Five of the six were positive in the Ames assays that we conducted (Table 5), making it difficult to discern a trend between PCs and Ames results.

Table 5. Ames responses for small-molecule N-nitrosamines and NDSRIs listed in Table 4 as a function of Carcinogenic Potency Categorization Approach (CPCA) potency category (PC; FDA, 2024a; Kruhlak et al., 2024) as listed in Table 2.

CPCA PCs are those published by the FDA as of September 4, 2024 (FDA, 2024a).

CPCA PC	Number NDSRIs tested with PCs	% Ames-positive NDSRIs	Number of small-molecule N-nitrosamines tested with PCs	% Ames-positive small-molecule N-nitrosamines
1	2	100%	0	NA
2	2	50%	1	100%
3	2	50%	3	100%
4	3	33%	1	0%
5	1	0%	1	100%

NA, not applicable.

Another question is whether positive responses in the Ames assay that occurred under very limited test conditions are truly indicative of a cancer risk. The answer to that question remains uncertain. The positive responses for N,N-bis(2,2-diethoxyethyl) nitrous amide and NDIPA fall into this category as positives were seen only with one strain (TA1535) and under one activation condition (preincubation with 30% hamster S9 mixes). Although N,N-bis(2,2-diethoxyethyl) nitrous amide is listed in the Lhasa Carcinogenicity Database as being noncarcinogenic in the rat (Table 2), this response is indicated as having questionable reliability (https://carcdb.lhasalimited.org, CAS 67856–67-1). Rao et al. (1979) indicate that N,N-bis(2,2-diethoxyethyl) nitrous amide (Compound 17 in Table 1 of the Rao et al. paper) is both Ames-negative and noncarcinogenic in rats (Table 2). Perhaps importantly, Rao et al. did not use hamster liver S9 in their Ames assays and the cancer data were only from rats and may be the same data used in the Lhasa evaluation. N,N-Bis(2, 2-diethoxyethyl) nitrous amide may be an example of a β-oxidized N-nitrosamine which Lijinsky (1987) indicated were activated better by hamster liver microsomes than rat liver microsomes. As for NDIPA, it is listed as a rat carcinogen in the Lhasa Carcinogenicity Database (Table 2). It also may be significant that NDIPA produced DNA damage in HepaRG cell spheroids, albeit at relatively high concentrations (Seo et al., 2023). So in this case, the positive Ames response seems to be indicative of a cancer risk.

High proportions of both subtypes of N-nitrosamines were positive in the Ames test conducted in this study: ten of the 12 small-molecule nitrosamines (83%) and eight of the 16 ‘testable’ NDSRIs (50%) were positive. These percentages should be interpreted with care, as no effort was made to randomize the choice of N-nitrosamines for testing. However, 83% of small-molecule N-nitrosamines being positive in the Ames test is similar to the percentage of Ames-positive N-nitrosamines found in recent data-mining studies evaluating the universe of N-nitrosamines with Ames test data (Thresher et al., 2020; Cross and Ponting, 2021; Tennant et al., 2023).

Although these retrospective studies show a general similarity between our data set and historical findings, there are also important differences. A data-mining study conducted by Trejo-Martin et al. (2022) concluded that several possible test variables, including using rat vs. hamster liver S9, have little overall effect on the ability of OECD TG 471-compliant assays to identify carcinogenic N-nitrosamines as mutagens. In another evaluation of existing data, Tennant et al. (2023) also found a generally limited impact of test protocol variables on N-nitrosamine Ames test results. These authors, however, indicated the difficulty in drawing firm conclusions when comparing data from studies employing different test protocols. In contrast, our study, which was conducted using a consistent protocol design and which considered the magnitude of the responses as well as whether or not the responses were positive or negative, found that the assay protocol did affect the responses of the test substances. The findings reported here identify enhancements to the basic Ames assay that positively impact the ability of the test to identify mutagenic N-nitrosamines. Our philosophy is that if Ames testing is to be conducted on a particular class of compounds, the assays should be designed using class-specific methods.

For instance, although rat liver S9 was generally effective at identifying mutagenic N-nitrosamines, many responses were stronger using hamster liver S9 and three (out of 18) Ames-positive N-nitrosamines were only positive with hamster liver S9. The same was true of the concentration of S9 used in the preincubation mix: many responses were stronger using the higher S9 concentration, and the two small molecule N-nitrosamines that were positive only with hamster liver S9, were positive only with 30% hamster liver S9. Although their study did not evaluate different concentrations of S9s in the assay, Thomas et al. (2024) concluded that, of all the variables considered, the use of hamster liver S9 was the most important variable affecting the mutagenicity of NDMA and NDEA in the Ames test.

In addition, all but two (out of 18) Ames-positive N-nitrosamines were positive in TA1535 (NMPA and N-nitroso-sertraline being the exceptions), and these two compounds were positive in E. coli WP2 uvrA (pKM101). Our data indicate that TA1535 and WP2 uvrA (pKM101) are clearly important for testing the mutagenicity of N-nitrosamines. Although Williams et al. (2019) suggested that TA1535 could be eliminated from the core group of Ames tester strains without loss of assay sensitivity, our data demonstrate that it may be wise to retain this strain when testing N-nitrosamines. None of the compounds we tested were uniquely positive in TA100, TA98, or TA1537. Thus, when reviewing historical N-nitrosamine Ames data, it is important that at a minimum, data from TA1535 and WP2 uvrA (pKM101) (or its equivalent, plus or minus pKM101) be available for forming conclusions on the mutagenicity of a test compound.

A curious observation made during the course of the study was that N-nitroso-folic acid tested negative in our Ames screen when it was delivered to the preincubation reaction in 25 μL of DMSO. DMSO was used as a solvent for this test substance because N-nitroso-folic acid was not sufficiently soluble in water, methanol, or acetone to test the maximum dose of 5 mg/plate. Serendipitously, however, this compound also was tested as an aqueous suspension in a 60 min preincubation assay using TA1535, where it tested equivocal or weakly positive in several assays (see summary results in Supplementary Material). It was further observed that the test substance suspension seemed to solubilize during the 60-min preincubation so that no precipitate was observed at the time of plating.

In addition, and unlike the other N-nitrosamines that were tested, the maximum mutagenic response with N-nitroso-folic acid was detected most consistently in assays conducted with 10% rat liver S9. This response clearly does not fit the pattern of results with our other test compounds, and because this equivocal/positive response was generated by violating our a priori rule of testing only solutions of test substances, we have classified N-nitroso-folic acid as Ames-negative (Table 4; Figs. 1 and 2). It should be noted, however, that N-nitroso-folic acid was reported as Ames-positive in an early study reported by Purchase et al. (1976, 1978; see Table 2) where the testing protocol and results were described only in general terms. Also, although the 1978 IARC report did not evaluate its carcinogenicity ‘on the basis of available data’, Wogan et al. (1975) reported that N-nitroso-folic acid induced lung tumors in the neonatal mouse cancer bioassay (see Table 2). Our curious Ames result, in some ways inconsistent with findings for other N-nitrosamines tested in our study, has not been investigated further.

To summarize the major findings of this study:

1. TA1535 and WP2 uvrA (pKM101) were the most informative tester strains for evaluating the mutagenicity of the set of 29 small-molecule nitrosamines and NDSRIs. These two tester strains, by themselves, were able to identify all the mutagenic test substances as mutagenic.

2. Although there was evidence of direct-acting mutagenicity for two of the test substances, all of the mutagenic test substances were activated to mutagens using liver S9, with hamster liver S9 generally producing greater mutagenic responses than rat liver S9.

3. For both rat liver and hamster liver S9, preincubations conducted with S9 mixes containing 30% S9 generally produced greater mutagenic responses than preincubations conducted with S9 mixes containing 10% S9.

4. The length of the preincubation had minimal effects on identifying mutagenic small-molecule nitrosamines and NDSRIs. All mutagenic test substances could be identified as mutagens with a preincubation of 30 min.

5. Liver S9s prepared from rats and hamster pretreated with either Aroclor-1254 or with PN/BNF were qualitatively similar in their ability to activate mutagenic test substances to mutagens in the Ames test.

6. Although many of the NDSRIs differed from small-molecule nitrosamines in terms of greater molecular size and complexity, the metabolic activation conditions and tester strain specificity that influenced the extent of their mutagenic responses in the Ames test were similar (Table 1; Table 4).

While is anticipated that no simple set of conditions will be ideal for detecting the mutagenicity of any class of compounds, including N-nitrosamines, recommendations that are consistent with the observations in this current study have been incorporated into the Enhanced Ames Test (EAT) that was recently introduced by several regulatory agencies (EMA, 2024; FDA, 2024a; Health Canada, 2024). It is also anticipated that refinements to the EAT recommendations are likely as further information on evaluating the mutagenicity of nitrosamines is gathered as a result of the several projects currently being conducted.

Supplementary Material

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.yrtph.2024.105709.

Data availability

Data will be made available on request.