Biopharma Consulting

Potentially mutagenic impurities

Potentially Mutagenic Impurities (PMIs) 



ICH M7 (R1):


· Acrylonitrile

· Allyl bromide

· 1-Bromobutane

· 2-Butenal (crotonaldehyde)

· 4′-(Chloroacetyl)-acetanilide

· Chloroacetyl chloride

· 2-Chloroaniline

· 4-Chloroaniline

· 4-Chloroaniline

· Chloroethane

· Chloromethane

· 2-Chloropropane

· Dimethylaminopropyl chloride


· Ethyl mesilate

· Formaldehyde

· Isopropyl mesilate

· Methyl mesilate

· o-Phenylenediamine

· Phenylhydrazine

· 2,6-Xylidine

Structural Alerts for Potential DNA Reactivity

“Structural alert” is mentioned but not defined in ICH M7, although FDA cites Ashby & Paton, 1993:

At its most basic a “structural alert” for mutagenic potential is an electrophilic moiety in a chemical structure, or a functional group that can be converted to an electrophilic moiety. Alerting groups have been identified by multiple researchers including Ashby & Tennant (1988, 1991) and Tennant et al. (1990). The following list of structural alerts for DNA reactivity it taken from WHO Food Additive Series 40.

a)  alkyl esters of phosphonic or sulfonic acids

b)  aromatic nitro-groups

c)  aromatic azo-groups (reduction to amine)

d)  aromatic ring N-oxides

e)  aromatic mono- and di-alkyl amino groups

f)  alkyl hydrazines

g)  alkyl aldehydes

h)  N-methylol derivatives

i)  monohaloalkanes

j)  N and S mustards, beta-haloethyl-

k)  N-chloramines

l)  propiolactones and propiosulfones

m)  aromatic and aliphatic aziridinyl derivatives

n)  aromatic and aliphatic substituted primary alkyl halides

o)  urethane derivatives (carbamates)

p)  alkyl N-nitrosamines

q)  aromatic amines and N-hydroxy derivatives

r)  aliphatic epoxides and aromatic oxides

s)  center of Michael reactivity

t)  halogenated methanes (C(X)4)

u)  aliphatic nitro groups

Additional information on structural alerts can be found in the following publications.

Sawatari et al, 2001:

Narayan (2010):

Bossa: Structural alert list:

RIVM Report on Structure-Activity Relationships:

Robinson, DL: OPRD, 2010:

Some of the structures shown above are extremely misleading in that:

· Pyridine N-oxide is non-mutagenic

· Only vinyl carbamate is mutagenic in the Ames’ assay (by activation to the epoxide), although ethyl carbamate (urethane) in converted to vinyl carbamate in vivo

· Virtually all aliphatic and alicyclic aldehydes are non-mutagens (except for formaldehyde which is a special case being an endogenous intermediate and non-carcinogenic by the oral route).

· Michael acceptors soft electrophiles and react preferentially with soft nucleophiles such as glutathione (GSH); hence they are generally Ames’-negative provided that GSH reserves are not depleted.

Commonly Misattributed Structural Alerts

These include:

  • Alkyl and aryl sulfonic acids or sulfonate anions; only the alkyl esters are alerting
  • Aromatic aldehydes; only alkyl aldehydes are alerting
  • Amines in general; only aromatic amines are alerting.
  • N-Oxides in general; only aromatic ring N-oxides are alerting
  • Only alkyl esters of phosphonic acids are alerting

Classification of Structural Alerts for Mutagenicity (probability of a correlation with bacterial mutagenicity)


Cohort of Concern*

High Probability

Medium Probability

Low Probability



Aromatic mono-amines


N-Nitroso compounds

Epoxides (except symmetrical and highly substituted epoxides)

N-Alkyl aromatic mono-amines

Aromatic N-oxides




Carbamates (except ethyl and vinyl carbamates)


Nitro compounds

Acetylated aromatic amines

Michael acceptors (eg α, β-unsaturated carbonyls)








Carboxylic acid halides


N- or S-Mustards

Boronic acids and derivatives

**Sulfonic acid halides


Azo compounds


Alkyl sulfonates or phosphonates


Aromatic polyamines




Purines, pyrimidines and intercalators


*Extremely high probability; **Except methanesulfonyl chloride

The probabilities mentioned above represent a personal evaluation of literature data. Others may suggest slightly different classifications for some alerts. For example, Galloway et al (2013) surveyed Ames’-test data provided by a number of pharmaceutical companies and concluded that alkylating agents (alkyl halides and alkyl esters of sulfonic and phosphonic acids), aromatic nitros, hydrazines and boronic acid derivatives all had probabilities of > 50% of producing positive results. On the other hand, Hansen et al, 2015 ( tested 44 commercially available boronic-acid derivatives and found that only 17 (39%) were Ames’-positive.

Acid chlorides are now considered to be falsely classified as mutagenic owing to the fact that positive results are obtained only if DMSO is used as a solvent in the Ames’ assay (with minor exceptions such as methanesulfonyl chloride). Acid chlorides react with DMSO to form chloromethylmethylsulfide which causes the positive response (Amberg et al, 2015:

In-Silico Prediction of Mutagenic Potential

We have access to the Leadscope ICH M7 suite ( for predicting mutagenic potential using two complementary systems (rule-based and statistical); we offer this service at a pass-through cost of $250 per structure.

A significant degree of caution is required when interpreting in-silico predictions and an expert review of the output is essential. Points to consider include:

· Outdated, compromised or inadequately documented literature information;

· Unconfirmed hypotheses in the published literature;

· Literature artefacts, for example by interaction with the solvent used for Ames’ testing; the default solvent is DMSO which has significant reactivity for some compounds such as acid halides[i] and esters of dichloroacetic acid[ii];

· Deciding whether an initial in-cerebro assessment is sufficient to determine that structural alerts are absent;

· Out-of-domain structures: the consensus position is that predictions for structures that are out-of-domain in respect of those in the supporting dataset are unreliable. Follow-up via use of data on appropriate reference compounds (if available) and expert assessment is generally recommended;

· Conflicting predictions: this situation arises occasionally and it is normally the statistical-based system that predicts potential mutagenicity while the rule-based system gives a negative prediction. The problem can often be resolved by evaluating the reference compounds supporting the positive prediction (which may contain structures that contain a strongly alerting element remote from the structural moiety of interest) and by manually searching for additional appropriate reference compounds;

· Incorrect data on reference compounds: although the providers of in-silico systems endeavour to ensure the reliability of their datasets, almost inevitably some inaccuracies will slip through. For example, the antimicrobial dimethoxane (2,6-dimethyl-1,3-dioxan-4-yl acetate; CAS no 828-00-2) contains no structural alerts for mutagenicity but is reported to be mutagenic in S. typhimurium TA100 +S9[iii] based on studies undertaken in the 1980s on commercial material. It is this result that is cited frequently in in-silico systems. Subsequent evaluations (on purer test material) gave negative results[iv].

· More generally, if a structure contains no alerts but is predicted to be mutagenic or a reference compound without alerts is classified as a mutagen, then alarm bells should ring. Databases of mutagenic compounds rely on test results which can be skewed by the presence of solvent interactions, impurities, etc. But in-silico systems are predicated on structures; and so just one incorrect attribution (such as that for dimethoxane) can produce a significant “ripple” effect that impacts adversely on predictions involving compounds containing similar structural elements;

· Expert assessment: Amberg et al[v] and Barber et al[vi], [vii] list a number of considerations for refuting a positive statistical-based prediction, including: the presence of coincidental features or mitigating structural features (eg steric hindrance for highly substituted epoxides), limited number of relevant examples in training dataset, irrelevant examples in the training set, small contributions from unrelated or poorly related reference compounds, incorrect or inadequate underlying Ames’ data (see above for examples).


[i] Amberg et al. 2015 OPRD, 19, 1495-1506.


[ii] Watanabe et al. 1996. Mutat Res. 361. 143-155


[iii] NTP studies on commercial dimethoxane:


[v] Amberg et al. 2016. Regul Toxicol Pharmacol. 77. 13-24.

[vi] Barber C et al. 2015. Establishing best practise in the application of expert review of mutagenicity under ICH M7. Regul Toxicol Pharmacol. 73(1):367-77


[vii] Barber C et al. 2016. Evaluation of a statistics-based Ames mutagenicity QSAR model and interpretation of the results obtained. Regul Toxicol Pharmacol. 76:7-20.