Pharmacometabonomic identification of a significant
host-microbiome metabolic interaction affecting
human drug metabolism
T. Andrew Claytona, David Bakerb, John C. Lindona, Jeremy R. Everettc, and Jeremy K. Nicholsona,1
aBiomolecular Medicine, SORA Division, Faculty of Medicine, Sir Alexander Fleming Building, Imperial College London, South Kensington, London SW7 2AZ,
United Kingdom;bPfizer Inc., 50 Pequot Avenue, New London, CT 06320; andcPfizer Global Research and Development, Ramsgate Road, Sandwich, Kent
CT13 9NJ, United Kingdom
Communicated by Burton H. Singer, Princeton University, Princeton, NJ, April 29, 2009 (received for review December 8, 2008)
We provide a demonstration in humans of the principle of pharmaco-
metabonomics by showing a clear connection between an individual’s
metabolic phenotype, in the form of a predose urinary metabolite
profile, and the metabolic fate of a standard dose of the widely used
analgesic acetaminophen. Predose and postdose urinary metabolite
were statistically analyzed in relation to drug metabolite excretion to
detect predose biomarkers of drug fate and a human-gut microbiome
cometabolite predictor was identified. Thus, we found that individuals
having high predose urinary levels of p-cresol sulfate had low postdose
We conclude that, in individuals with high bacterially mediated p-cresol
generation, competitive O-sulfonation of p-cresol reduces the effective
systemic capacity to sulfonate acetaminophen. Given that acetamino-
phen is such a widely used and seemingly well-understood drug, this
finding provides a clear demonstration of the immense potential and
power of the pharmacometabonomic approach. However, we expect
drugs and xenobiotics. We propose that assessing the effects of micro-
and of personalized health care. Furthermore, we envisage that gut
bacterial populations might be deliberately manipulated to improve
drug efficacy and to reduce adverse drug reactions.
acetaminophen ? p-cresol ? bacteria ? sulfate ? glucuronide
a potential means of personalizing human drug treatments to increase
drug efficacy and to decrease adverse reactions (1–6). However,
age, disease, and other drug use) are also important determinants of
individual metabolic phenotypes, which modulate drug metabolism,
efficacy, and toxicity. Such environmental complications, which may
of drug-induced responses that are based only on genomic differences
(7, 8). For instance, for many classes of compound, enzyme induction
state, which is environmentally determined, influences drug metabo-
personalized drug treatment has recently been proposed wherein
responses to potential drug interventions (9). This ‘‘pharmacometabo-
are sensitive to both genomic and environmental influences affecting
metabolism. A further crucial advantage and feature of this metabo-
nomic approach is its openness to finding unexpected biomarkers and
biomarker combinations, as multiple analytes are quantified simulta-
he effects of drug treatments can vary greatly between different
individuals, and pharmacogenomics has been widely advocated as
The only factors limiting which analytes are detected are the nature of
the sample that is analyzed and the analytical platform used. Thus,
pharmacometabonomic modeling need not be limited by prior under-
for the concept (9, 11–13), there has, until now, to the best of our
knowledge, been no convincing pharmacometabonomic demonstra-
tion in humans.
To test the feasibility of applying the pharmacometabonomic
approach to man, we chose as our example the well-known anal-
gesic and antipyretic drug acetaminophen (N-acetyl-p-aminophe-
nol; known as paracetamol in Europe). Acetaminophen is one of
toxicology and metabolism have been extensively investigated over
many years (14–20). However, we will show here that, even for this
most familiar drug, pharmacometabonomic analysis will yield sig-
nificantly increased understanding of its metabolic behavior in
humans. These findings have considerable implications for person-
alized drug treatment in general and lead to new and testable
hypotheses for a number of diseases.
Acetaminophen was chosen to exemplify the pharmacometabo-
usage and its low toxicity at therapeutic doses, which was necessary
to establish an ethically approved clinical trial. It is also predomi-
nantly and relatively rapidly eliminated in the urine (15–19). Thus,
by collecting postdose urine samples we could study the manner of
its excretion by each subject, such excretion being known to show
considerable intersubject variation (20). Precisely how a particular
drug is metabolized and excreted by each individual can have a
major influence on its safety and efficacy. Thus, for instance,
to influence the extent of an adverse effect, whereas the rate of
removal of the pharmacologically active compound might be ex-
pected to influence the extent and duration of the desired phar-
macological action. With this in mind, the aim of the present study
urine would allow prediction of some aspect of acetaminophen
provide a proof-of-principle for the feasibility of pharmacometa-
bonomics in humans.
Author contributions: T.A.C., J.C.L., J.R.E., and J.K.N. designed research; T.A.C. performed
research; T.A.C. and D.B. analyzed data; and T.A.C., D.B., J.C.L., J.R.E., and J.K.N. wrote the
Conflict of interest statement: T.A.C., J.C.L., J.R.E., and J.K.N. are inventors on a relevant
patent application from which financial gain might be derived. T.A.C., J.C.L., and J.K.N.
might also benefit financially from the future placement of related analytical or research
Freely available online through the PNAS open access option.
See Commentary on page 14187.
1To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/cgi/content/full/
August 25, 2009 ?
vol. 106 ?
Our ethically approved study (SI Text) was based on 99 healthy
male volunteers who were all nonsmokers between 18 and 64 years
old with a condition of their participation being that they had not
taken any drugs in the preceding week. Each participant provided
a predose urine sample and then, after taking a standard thera-
peutic dose of acetaminophen (two 500-mg tablets), each was
requested to collect all of his postdose urine over 2 consecutive 3-h
periods (0–3 h and 3–6 h after dosing). All of the samples were
analyzed by 600 MHz1H NMR spectroscopy with the predose
spectra providing profiles of the detectable, naturally occurring
(endogenous) metabolites and the postdose spectra providing
profiles of the acetaminophen-related compounds superimposed
on an endogenous metabolite ‘‘background.’’ From these NMR
spectra and the known urine volumes, the urinary acetaminophen-
related compounds excreted by each subject over each postdose
collection period were quantified as acetaminophen sulfate (S),
acetaminophen glucuronide (G), and ‘‘other,’’ with the ‘‘other’’
components expected to be mainly the parent, cysteine conjugate,
and N-acetylcysteine conjugate (17–19). We then searched for
components of intersubject variation in the predose spectra that
would correlate with intersubject variation in the postdose data.
However, from the outset, prediction of the S/G ratio was of
particular interest because this ratio is known to show extensive
interindividual variation and is assumed to be indicative of the
relative extent to which acetaminophen is metabolized via 2 major
phase 2 conjugative processes (O-sulfonation and glucuronidation)
that impact on the metabolism of many different drugs (20, 21).
Additionally, we expected that the S/G ratio would be less suscep-
tible to any sample collection errors than the absolute amounts of
Results and Discussion
The Urinary Excretion of Acetaminophen and Its Metabolites. We
found that ?30% of the 1-g dose was recovered (as any acetamin-
periods (0–3 h and 3–6 h), this being consistent with the findings
of an earlier report (15) where the urinary excretion of acetamin-
Furthermore, we found that, in terms of moles, the combined
excretion of S and G typically accounted for ?85% of the total
amount of acetaminophen-related compounds recovered in each
collection period, this being consistent with the 24-h urinary
recoveries from 111 Caucasians given a 1.5-g dose (20). Represen-
tative predose and postdose spectra are shown in Fig. 1.
The average S/G ratios for the 2 postdose collections were found
to be 0.71 (0–3 h) and 0.53 (3–6 h), respectively, with the S/G value
than in the 0–3 h collection, and this change in the S/G ratios being
consistent with what has been reported for the early excretion of a
20-mg/kg dose (17). Furthermore, there was a clear correlation
between the S/G 0–3 h and S/G 3–6 h data (r ? 0.927). However,
a plot of these data indicated an outlier and it was judged, from this
and from an unusually low excretion of acetaminophen-related
metabolites, that this subject had not fully collected his 0–3 h
sample. With this subject excluded, the correlation between S/G
0–3 h and S/G 3–6 h was slightly improved (r ? 0.948). A moderate
(17%) decrease in average S excretion was observed between the
2 collection periods along with a largely compensating increase in
average G excretion. Having excluded the 0–3 h data for the 1
the correlation coefficients between the observed S/G ratios and
the amount of S excreted were found to be 0.510 (0–3 h) and 0.835
S/G and the amount of G excreted were found to be ?0.738 (0–3
h) and ?0.803 (3–6 h), respectively. Precautionary checks showed
that the amounts of S and G excreted, and the S/G ratio, were not
related to the age of the subjects or to their body mass or to the
order in which the samples were analyzed (Table S1).
Examination of the Predose Spectral Profiles. Our initial way of
by means of the PLS (Projection to Latent Structure)-based pattern
recognition methods that are typically used in metabonomic studies
(22) and which had proved highly effective in our earlier animal-based
work (9). However, in the present case, the PLS-based approach was
this revised and relatively simple approach, we found 2 potentially
discriminatory predose metabolites, later identified as the microbial
cometabolites p-cresol sulfate (PCS) and phenylacetylglutamine
spectrum for a subject whose urine contained a relatively high level of p-cresol
sulfate. (B) The corresponding 0–3 h postdose urine spectrum, which shows a
spectrum, which shows a relatively high ratio of acetaminophen sulfate to
acetaminophen glucuronide. To facilitate their comparison, all these spectra
were processed in the same way, without resolution enhancement and with a
be observed at ?? 4.7. Furthermore, each spectrum has been scaled so that the
creatinine methylene peak at ?? 4.06 is just on scale (with the result that
Insets, which are expansions of selected spectral regions, are scaled to fill
the available space. Key to numbered peaks: 1, creatinine; 2, hippurate; 3,
from acetaminophen-related compounds; 7, acetaminophen sulfate; 8, acet-
aminophen glucuronide; 9, other acetaminophen-related compounds.
Representative1H NMR spectra of urine samples provided before and
Clayton et al.PNAS ?
August 25, 2009 ?
vol. 106 ?
no. 34 ?
principal-components analyses (PCA) focused on the aromatic region
(? 9.1–6.9) of the predose NMR spectra (SI Text and Fig. S1), and no
inspection, the predose urinary levels of PCS and PAG were found to
because there is a significant degree of commonality in their origins
before the final sulfate and glutamine conjugations; thus, p-cresol and
phenylacetic acid are known to be produced from tyrosine and phe-
our further analysis, using integrated predose spectral band intensities
the individual spectral components, only PCS was likely to provide
statistically significant discrimination with respect to S/G (SI Text).
Further Analyses Focused on PCS. To get the best possible measure of
the level of PCS relative to creatinine, which, for this study, was a
suitable internal reference compound for the urinary quantitation (SI
after local baseline correction. The peaks chosen for integration were
at ?? 4.06 and the relevant integral ratios [designated I.R., where I.R.
collections. From inspection of Fig. 4, it is readily apparent that a high
predose level of PCS (I.R. ? 0.06) is associated with a low S/G ratio
postdose and use of the Mann–Whitney U test in conjunction with an
appropriate Bonferroni correction (100) (SI Text) confirmed the sta-
tistical significance of the distribution of the high PCS group (25
subjects) with respect to the S/G ratios obtained in each postdose
collection. The Bonferroni correction was applied to counter the
multiple hypothesis testing that results from the multivariate nature of
metabonomic data (24). With a Bonferroni correction of 100, the P
value for 95% confidence becomes 0.05/100 ? 5 ? 10?4and the P
0–3 h) and 1.2 ? 10?4(for S/G 3–6 h).
In the preceding analysis of the full data set (all subjects
included), we found a high predose level of PCS to be associated
be affected by variation in the amounts of both S and G excreted.
the conversion of acetaminophen to S (Fig. 3), the observed
connection to predose PCS strongly suggests that the relevant
controlling factor is the amount of S excreted. Furthermore, with
lower S/G ratios being observed in the 3–6 h collection than in the
0–3 h collection, it appears that 1 g of acetaminophen represents a
substantial challenge to the sulfonation capacity of the subjects
studied. The extent to which any compound undergoes sulfonation
can potentially be limited both by the availability of the sulfonate
characteristics and availability of the relevant sulfotransferase en-
p-cresol and acetaminophen are both substrates for the same
human cytosolic sulfotransferase, SULT1A1 (26), and can, there-
fore, compete for enzyme binding sites as well as for PAPS.
Furthermore, in contrast to what has been reported for rats (27),
recent literature suggests that p-cresol is almost entirely converted
to PCS in humans (28–30). Thus, we envisaged that an individual’s
samples with color-coding according to postdose behavior. All of the plots were
produced in MATLAB with each individual NMR spectrum being normalized to
superimposed on the individual spectra for the 25 subjects giving the lowest
coding, over the region of ? 7.18–7.32, which contains the PCS aromatic signals
(the pair of pseudo ‘‘doublets’’ centered at ?? 7.21 and at ?? 7.29). In all plots
‘‘a.u.’’ designates arbitrary units. In plots A and B, some spectra are obscured by
the subsequently superimposed spectra.
Selected regions of1H NMR spectra obtained from predose urine
CH CH3 3
OSO OSO3 3 H
C CO H O H
C CO HO H
C CNHNH2 2
C CO H O H
C CO O
OSOOSO3 3H H
O OC H O C H O
6 6 9 9 6 6
C CO O
C CO HO H
C CNHNH2 2
C CO H O H
C CO O
HNCH(HNCH(C CO H) O H)
CHCH2 2 C CH H2 2C CO O NH NH2 2
C CO H O H
phen (1) may be sulfonated to produce acetaminophen sulfate (2) or glucu-
p-cresol sulfate (8) from tyrosine (4) (23). The green box highlights the highly
analogous and potentially competitive sulfonation of acetaminophen and p-
cresol (25, 26). (C) Stepwise production of phenylacetylglutamine (12) from
phenylalanine (9) (23) with the yellow box highlighting similarities with the
sulfate; 3, acetaminophen glucuronide; 4, tyrosine; 5, 4-hydroxyphenylpyruvic
acid; 6, 4-hydroxyphenylacetic acid; 7, p-cresol; 8, p-cresol sulfate; 9, phenylala-
the body, compounds 2 and 8 would normally be expected to exist as ROSO3
rather than as ROSO3H, where R designates the remainder of each molecule.
Relevant metabolic pathways. (A) The hydroxyl group of acetamino-
www.pnas.org?cgi?doi?10.1073?pnas.0904489106Clayton et al.
capacity to sulfonate acetaminophen will be reduced by ongoing
presentation of endogenous p-cresol, and the potential competitive
significance of the p-cresol challenge was confirmed by calculation
data set, the postdose excretion of both S and G and found that, in
the 3–6 h collection, the high predose PCS subjects (I.R. ? 0.06)
were clearly associated with lower S excretion (P ? 2.3 ? 10?5with
95% confidence at P ? 5 ? 10?4after Bonferroni correction).
Furthermore, a similar and statistically significant (P ? 1.6 ? 10?4)
relationship was found when the S excretion values for this collec-
tion period were first corrected to unit body mass. Thus, these data
are fully consistent with the hypothesis that substantial predose
production of endogenous p-cresol can reduce an individual’s
ability to sulfonate acetaminophen by acting as a competitive
the colonic mucosa is known to have significant sulfonation capac-
ity (25, 31) and could potentially convert colonically produced
p-cresol to PCS. However, with acetaminophen being rapidly
absorbed from the small intestine and with the liver being regarded as
the principal site of its metabolism (17), we envisage that some
colonically produced p-cresol may escape further colonic modification
and be sulfonated in the liver rather than in the gastrointestinal tract.
Influence of Experimental Variables. To check for experimental
variables that might be influencing these data, the distribution of
with respect to analytical run order and with respect to the data
obtained for subject age, height, body mass, and body mass index
(BMI), but no association was found. However, it was noticeable
that the 10 subjects showing the highest predose PCS levels (I.R. ?
0.09) tended to be older (P ? 0.01) and shorter (P ? 0.02)
individuals who would, in general, be expected to have less muscle
mass and to excrete less creatinine. However, the first of these
findings also suggests that aging might lead to increased p-cresol
production, and this could potentially be caused by age-related
changes in the nature of the gut bacteria (32). Furthermore,
whereas, in the present study, we found no clear evidence of a
in acetaminophen sulfonation has been observed in male rats (33).
To further check the basis of our findings we also reexamined the
data after first excluding all those subjects where there was any
known or suspected noncompliance with the study protocol [e.g.,
where a sensitive analysis of a subject’s predose urine sample
suggested some prior use of acetaminophen or where there was
evidence of recent alcohol consumption (SI Text)]. With the
remaining 78 subjects, the graphical relationship between predose
PCS and the S/G ratio was maintained [the P values for the
distribution of the high PCS subjects (I.R. ? 0.06) with respect to
S/G 0–3 h and S/G 3–6 being 7.4 ? 10?4and 9.9 ? 10?4,
respectively] and statistical significance was still achieved for the
distribution of the high predose PCS subjects with respect to the
absolute amount of S excreted during the 3–6 h collection (P ?
3.4 ? 10?4). We, therefore, conclude that the perceived predose to
postdose connection is real.
Potential Biomedical Significance. Although the potential signifi-
drug-induced reactions is becoming increasingly well recognized
(34–41), we have not seen any previous report of the present
finding, which could be of considerable importance if it can be
proven to hold for the wider human population or for specific
subsets of that population.
We envisage that, by depleting hepatic sulfonation capacity,
liver more vulnerable to acetaminophen-induced damage and that
markedly increased p-cresol production could potentially explain
the reported association between fasting and an increased likeli-
hood of hepatotoxicity from acetaminophen (42, 43). However, in
could also potentially increase acetaminophen hepatotoxicity by
other means, such as by enzyme induction or glutathione depletion
(44–46), and preliminary data (SI Text) suggest that high p-cresol
exposure might lead to a more generalized impairment of sulfur-
possibly leading to depletion of both taurine and glutathione.
significance for adverse reactions to acetaminophen. Instead, the
wider and more obvious significance of our finding lies in its
potential consequences for sulfonation reactions in general and in
Many different compounds are substrates for sulfotransferase-
a major role in modifying the physical properties of both small and
large molecules. Thus, sulfonation facilitates the excretion of many
compounds and is crucial to the structure and properties of
macromolecules such as chondroitin sulfate (a component of
cartilage). Notably, many drugs and/or their hydroxylated metab-
olites are phase II conjugated via sulfonation. Among several other
important functions, sulfonation is also known to have a role in
modulating the action of hormones and neurotransmitters and
0.020.06 0.10.14 0.18
Pre-dose PCS/creatinine integral ratio (I.R.)
Pre-dose PCS/creatinine integral ratio (I.R.)
S/G 0-3 h post-dose
S/G 3-6 h post-dose
sulfate (PCS) to creatinine and the postdose urinary ratio of the major acetamino-
phen metabolites: acetaminophen sulfate (S) and acetaminophen glucuronide (G).
of the peaks at ?? 2.35 and at ?? 4.06 in the1H NMR spectrum recorded from the
The observed relationship between the predose urinary ratio of p-cresol
Clayton et al.PNAS ?
August 25, 2009 ?
vol. 106 ?
no. 34 ?
appears to be especially important during early human develop-
(both cytosolic and membrane-bound), but human cytosolic sulfo-
transferase SULT1A1, which acts on acetaminophen, has a broad
substrate range and is one of the most important sulfotransferases
for xenobiotic sulfonation as well as acting on several endogenous
substrates (26, 51). Additionally, a key feature of sulfonation in
sulfonate donor. Thus, we might reasonably expect that, by com-
peting for PAPS or for one or more sulfotransferases, the flux of
p-cresol through the system will affect the sulfonation of a wide
range of drugs and endogenous compounds, thereby influencing
toxicity. However, given what is known about gut bacterial produc-
tion of p-cresol from protein residues (23, 43, 52–54) (Fig. 3), with
Clostridium difficile being one of a number of p-cresol producers
(53, 54), our present results show that environmental factors can
exert a dominant influence on the extent to which a compound
by altering the amount of p-cresol produced, variation in either the
diet or the gut bacteria could potentially exert a major influence on
drug-induced responses or diseases where sulfonation has an im-
of a drug where sulfonation is considered to be important in
producing the desired pharmacological effect (55). However, sul-
fonation is not always beneficial, and tamoxifen, which is used in
treating breast cancer, provides an example of a drug where
of an associated adverse reaction, namely an increased incidence of
endometrial cancer. Thus, it has been suggested that tamoxifen-
DNA adducts are formed via O-sulfonation (56). As a further
example, sulfonation phenotype could potentially influence both
the efficacy and side-effects of apomorphine, for which sulfonation
is the major metabolic pathway in humans (57). As regards known
associations with disease, hyperactivity in children provides one
example of a condition that has been associated with increased
p-cresol levels and where the involvement of dietary factors is also
suspected (58). Furthermore, an increased urinary level of PCS has
been associated with the progression of multiple sclerosis (59, 60).
Additionally, various other diseases (Parkinson disease, motor
neuron disease, rheumatoid arthritis, and childhood autism) have
been associated with a reduction in the S/G ratios obtained after
acetaminophen dosing (61–63), which leads us to tentatively sug-
p-cresol production might also have some relevance to their etiol-
ogy, with further circumstantial evidence coming from additional
associations with gastrointestinal abnormalities (64–67). However,
prove a causal role for p-cresol in respect of these diseases and also
that p-cresol can exert a variety of effects (27–30, 43, 45, 46, 68–73)
such as blocking the conversion of the neurotransmitter dopamine
to noradrenaline (68). Therefore, as far as we are aware, any
involvement of p-cresol in respect to these diseases remains to be
proven as well as the exact nature of any such involvement.
However, one general hypothesis would be that where the diet or
the profile of the gut bacteria is altered in favor of p-cresol
production, impaired sulfonation and other effects can result such
development, a variety of consequences would be expected.
Conclusions and Future Prospects. In the population studied, in this,
our first pharmacometabonomic study in humans, we have found a
clear association between an individual’s predose urinary metab-
olite profile and the postdose urinary fate of acetaminophen.
Further investigation will be required to determine the extent to
which this association holds for the wider human population, but it
is encouraging that, in hindsight, it makes such clear biochemical
sense. Thus, our findings strongly suggest that a person’s capacity
for acetaminophen sulfonation can be significantly reduced by
competitive p-cresol sulfonation, with p-cresol known to be pro-
duced from protein-derived tyrosine in reactions involving gut
bacteria. Given the range of substances for which sulfonation is
important, this finding suggests a means by which the gut bacteria
might influence both drug-induced responses and disease develop-
has been extensively studied over many years, our findings provide
a remarkable demonstration of the power and potential of the
pharmacometabonomic approach, which we hope will eventually
be used to improve drug treatment outcomes. With a view to the
future practicality of this exciting approach, it is also encouraging
Furthermore, we envisage that rapidly growing recognition of the
multiple metabolic interactions between humans and their gut
symbionts, and the potential significance of the latter in regard to
disease, drug efficacy, and adverse drug reactions, will lead to a
revolution in the way that drugs are developed. We also envisage
action (41), and that in some other cases, gut bacteria will be
manipulated by some prior or accompanying treatment in order to
improve drug treatment outcomes.
Materials and Methods
Full details are provided in the SI Text. The study volunteers were nominally
analysis by mixing 440 ?L of urine with 220 ?L of phosphate buffer (pH ?7.4 to
which sodium azide had been added as an antibacterial preservative), and the
mixture was centrifuged to remove suspended particles. 550 ?L of ‘‘clear’’ buff-
ered urine was transferred to a sample vial and 55 ?L of a TSP/D2O solution was
added to give a final TSP concentration of 1 mM. TSP (sodium 3-trimethylsilyl-
[2,2,3,3-2H4]-1-propionate) is a chemical shift reference compound used in the
NMR experiment and the D2O provided a field/frequency lock for the NMR
and operated by means of the Xwinnmr software (all from Bruker Biospin). The
‘‘noesypresat’’ pulse sequence was used to suppress the water signal and to
acquire the data, and the spectral acquisitions were automated by using the
data were processed using Xwinnmr. 1-Hz line-broadening was applied to the
predose1H NMR spectra by means of an exponential multiplication of the free
to reduce the size of the residual water signal. The resulting spectra were
manually phased to give an even baseline around the NMR signals, and the
baseline of each spectrum was manually adjusted to zero intensity by using a
straight-line baseline correction algorithm. The chemical shift scale was set by
The postdose1H NMR spectra were processed similarly, using 0.3-Hz line broad-
ening, and subsequently with resolution enhancement, and a measure approx-
imating to the mole ratio of acetaminophen sulfate (S) to acetaminophen gluc-
uronide (G) was determined by integration of the respective resolution-
enhanced N-acetyl peaks. For each subject and collection period, relative
to the added TSP and the mass of urine collected. Except where stated, these
excretion values were not adjusted to excretion per unit of body mass. Subse-
and normalized to a constant integral for the ? 4.07–4.05 region, which encom-
passes the creatinine methylene singlet. Average group spectra were then cal-
culated in MATLAB and various spectral plots were made and examined. After
and then normalizing these integrals to a constant value for the integral for the
? 4.07–4.05 region, principal-components analysis of the integrals for the ?
9.1–6.9 region was performed in Pirouette 3.11 (from Infometrix) using mean-
centered variable scaling. The PCA was also repeated after normalization to a
constant integral for the ? 3.07–3.03 region, which encompasses the creatinine
methyl singlet (Fig. S1). Local baseline correction and integration of selected
peaks in the predose spectra was performed in Xwinnmr. Statistical significance
of abstracted data was assessed by using the Mann–Whitney U test. See also
www.pnas.org?cgi?doi?10.1073?pnas.0904489106 Clayton et al.
ACKNOWLEDGMENTS. We thank Prof. R. L. Smith of Imperial College London Download full-text
Centre, Canterbury, Kent, U.K., for conducting the dosing and sample collec-
tion phase; the study volunteers for their participation; Dr. C. Legido-Quigley
of King’s College London for conducting supporting UPLC-MS analyses; Dr.
Bernard North (Imperial College Statistical Advisory Service) for statistical
advice and analysis; Dr. O. Cloarec (Royal Holloway, University of London) for
the use of MATLAB routines that he developed while at Imperial College
London and for associated guidance; Mr. J. T. M. Pearce of Imperial College
London for MATLAB support; Prof. H. Tang (now of Wuhan University, Peo-
ple’s Republic of China) and Dr. O. Beckonert of Imperial College London for
NMR support; and Dr. D. A. Parker of Imperial College London for providing
data that assisted in the identification of PAG. We acknowledge the prior
contributions of Dr. H. Antti (University of Umea ¨, Sweden) and Ms. R. Walley
(Pfizer, U.K.) in identifying the potential relationship between predose cre-
atinine and the total excretion of acetaminophen-related compounds over
the first postdose collection period (SI Text). The authors thank Pfizer Global
R & D for funding this work and T.A.C. thanks Pfizer for personal financial
support. J. K. N. is a member of the Imperial College MRC-HPA Center for
Environment and Health.
rational therapeutics. Science 286:487–491.
2. Evans WE, Johnson JA (2001) Pharmacogenomics: The inherited basis for interindi-
vidual differences in drug response. Annu Rev Genomics Hum Genet 2:9–39.
3. Spear BB, Heath-Chiozzi M, Huff J (2001) Clinical application of pharmacogenetics.
Trends Mol Med 7:201–204.
4. Pagliarulo V, Datar RH, Cote RJ (2002) Role of genetic and expression profiling in
pharmacogenomics: The changing face of patient management. Curr Issues Mol Biol
5. Evans WE, Relling MV (2004) Moving towards individualized medicine with pharma-
cogenomics. Nature 429:464–468.
6. Marsh S, McLeod HL (2006) Pharmacogenomics: From bedside to clinical practice. Hum
Mol Genet 15:R89–R93.
7. Nebert DW, Jorge-Nebert L, Vesell ES (2003) Pharmacogenomics and ‘‘individualized
drug therapy’’: High expectations and disappointing achievements. Am J Pharmacog-
8. Nebert DW, Zhang G, Vesell ES (2008) From human genetics and genomics to phar-
9. Clayton TA, et al. (2006) Pharmaco-metabonomic phenotyping and personalized drug
treatment. Nature 440:1073–1077.
10. Lindon JC, Holmes E, Bollard ME, Stanley EG, Nicholson JK (2004) Metabonomic
technologies and their applications in physiological monitoring, drug safety assess-
ment and disease diagnosis. Biomarkers 9:1–31.
at pharmaco-metabonomics. Trends Pharmacol Sci 27:580–586.
12. Li H, et al. (2007) Pharmacometabonomic phenotyping reveals different responses to
xenobiotic intervention in rats. J Proteome Res 6:1364–1370.
13. Ala-Korpela M (2007) Potential role of body fluid1H NMR metabonomics as a prog-
nostic and diagnostic tool. Exp Rev Mol Diagn 7:761–773.
14. Newson RB, Shaheen SO, Chinn S, Burney PGJ (2000) Paracetamol sales and atopic
disease in children and adults: An ecological analysis. Eur Respir J 16:817–823.
15. Cummings AJ, King ML, Martin BK (1967) A kinetic study of drug elimination: The
excretion of paracetamol and its metabolites in man. Br J Pharmacol Chemother
16. Clements JA, Heading RC, Nimmo WS, Prescott LF (1978) Kinetics of acetaminophen
absorption and gastric emptying in man. Clin Pharmacol Ther 24:420–431.
17. Prescott LF (1980) Kinetics and metabolism of paracetamol and phenacetin. Br J Clin
18. Forrest JAH, Clements JA, Prescott LF (1982) Clinical pharmacokinetics of paracetamol.
Clin Pharmacokinet 7:93–107.
19. Lau GSN, Critchley JAJH (1994) The estimation of paracetamol and its major metabo-
lites in both plasma and urine by a single high-performance liquid chromatography
assay. J Pharm Biomed Anal 12:1563–1572.
20. Critchley JAJH, Nimmo GR, Gregson CA, Woolhouse NM, Prescott LF (1986) Inter-subject
and ethnic differences in paracetamol metabolism. Br J Clin Pharmacol 22:649–657.
21. Burchell B, Coughtrie MWH (1997) Genetic and environmental factors associated with
22. Trygg J, Holmes E, Lundstedt T (2007) Chemometrics in metabonomics. J Proteome Res
23. Smith EA, Macfarlane GT (1997) Formation of phenolic and indolic compounds by
anaerobic bacteria in the human large intestine. Microb Ecol 33:180–188.
24. Broadhurst DI, Kell DB (2006) Statistical strategies for avoiding false discoveries in
metabolomics and related experiments. Metabolomics 2:171–196.
25. Klaassen CD, Boles JW (1997) The importance of 3?-phosphoadenosine 5?-phosphosul-
fate (PAPS) in the regulation of sulfation. FASEB J 11:404–418.
26. Gamage N, et al. (2006) Human sulfotransferases and their role in chemical metabo-
lism. Toxicol Sci 90:5–22.
27. Morinaga Y, Fuke C, Arao T, Miyazaki T (2004) Quantitative analysis of cresol and its
metabolites in biological materials and distribution in rats after oral administration.
Legal Med 6:32–40.
28. Schepers E, et al. (2007) p-Cresylsulphate, the main in vivo metabolite of p-cresol,
activates leucocyte free radical production. Nephrol Dial Transplant 22:592–596.
29. Martinez AW, Recht NS, Hostetter TH, Meyer TW (2005) Removal of p-cresol sulphate
by hemodialysis. J Am Soc Nephrol 16:3430–3436.
30. de Loor H, Bammens B, Evenepoel P, De Preter V, Verbeke K (2005) Gas-chromato-
graphic-mass spectrometric analysis for measurement of p-cresol and its conjugated
metabolites in uremic and normal serum. Clin Chem 51:1535–1538.
31. Ramakrishna BS, et al. (1989) Estimation of phenolic conjugation by colonic mucosa.
J Clin Pathol 42:620–623.
32. Hopkins MJ, Sharp R, Macfarlane GT (2002) Variation in human intestinal microbiota
with age. Dig Liver Dis 34(Suppl 2):S12–S18.
33. Galinsky RE, Corcoran GB (1986) Influence of advanced age on the formation and
elimination of acetaminophen metabolites by male rats. Pharmacology 32:313–320.
34. Rowland I (1981) The influence of the gut microflora on food toxicity. Proc Nutr Soc
35. Rowland IR (1986) Reduction by the gut microflora of animals and man. Biochem
37. Mikov M (1994) The metabolism of drugs by the gut flora. Eur J Drug Metab Pharma-
and the continuum of metabolism. Nat Rev Drug Discov 2:668–676.
and personalized health care. Nat Rev Microbiol 3:431–438.
40. Ley RE, Turnbaugh PJ, Klein S, Gordon JI (2006) Human gut microbes associated with
obesity. Nature 444:1022–1023.
41. Jia W, Li H, Zhao L, Nicholson JK (2008) Gut microbiota: A potential new territory for
drug targeting. Nat Rev Drug Discov 7:123–129.
42. Whitcomb DC, Block GD (1994) Association of acetaminophen hepatotoxicity with
fasting and ethanol use. J Am Med Assoc 272:1845–1850.
43. Kawakami K, Kojima K, Makino I, Kato I, Onoue M (2007) Fasting enhances p-cresol
production in the rat intestinal tract. Exp Anim 56:301–307.
44. Tanaka E, Yamazaki K, Misawa S (2000) Update: The clinical importance of acetamin-
ophen hepatotoxicity in non-alcoholic and alcoholic subjects. J Clin Pharm Ther
45. Thompson DC, Perera K, Fisher R, Brendel K (1994) Cresol isomers: Comparison of toxic
potency in rat liver slices. Toxicol Appl Pharmacol 125:51–58.
46. Thompson DC, Perera K, London R (1995) Quinone methide formation from para
isomers of methylphenol (cresol), ethylphenol, and isopropylphenol: Relationship to
toxicity. Chem Res Toxicol 8:55–60.
47. Strott CA (2002) Sulfonation and molecular action. Endocr Rev 23:703–732.
48. Coughtrie MWH (2002) Sulfation through the looking glass—recent advances in
sulfotransferase research for the curious. Pharmacogenomics J 2:297–308.
49. Kauffman FC (2004) Sulfonation in pharmacology and toxicology. Drug Metab Rev
50. Van Der Marel C, Anderson BJ, Tibboel D, Holford NHG (2002) Modelling paracetamol
urine metabolites. Paediatr Anaesth 12:827.
SULT1A1. Int J Biochem Cell Biol 39:685–689.
52. Ling WH, Ha ¨nninen O (1992) Shifting from a conventional diet to an uncooked vegan
diet reversibly alters fecal hydrolytic activities in humans. J Nutr 122:924–930.
53. Sivsammye G, Sims HV (1990) Presumptive identification of Clostridium difficile by
detection of p-cresol in prepared peptone yeast glucose broth supplemented with
p-hydroxyphenylacetic acid. J Clin Microbiol 28:1851–1853.
54. Bone E, Tamm A, Hill M (1976) The production of urinary phenols by gut bacteria and
55. Johnson GA, Barsuhn KJ, McCall JM (1982) Sulfation of minoxidil by liver sulfotrans-
ferase. Biochem Pharmacol 31:2949–2954.
56. Kim SY, et al. (2005) Formation of tamoxifen-DNA adducts via O-sulfonation, not
O-acetylation, of ?-hydroxytamoxifen in rat and human livers. Drug Metab Dispos
57. Thomas NL, Coughtrie MWH (2003) Sulfation of apomorphine by human sulfotrans-
ferases: Evidence of a major role for the polymorphic phenol sulfotransferase,
SULT1A1. Xenobiotica 33:1139–1148.
58. Adams RF, Murray KE, Earl JW (1985) High levels of faecal p-cresol in a group of
hyperactive children. Lancet 326:1313.
59. Cao L, Kirk MC, Coward LU, Jackson P, Whitaker JN (2000) p-Cresol sulphate is the
dominant component of urinary myelin basic protein like material. Arch Biochem
60. Jackson PL, Cao L, Blalock JE, Whitaker JN (2003) The requirement of ammonium or
other cations linked with p-cresol sulfate for cross-reactivity with a peptide of myelin
basic protein. Arch Biochem Biophys 418:119–124.
61. Steventon GB, et al. (1990) Metabolism of low-dose paracetamol in patients with
chronic neurological disease. Xenobiotica 20:117–122.
62. Bradley H, Waring RH, Emery P, Arthur V (1991) Metabolism of low-dose paracetamol
in patients with rheumatoid arthritis. Xenobiotica 21:689–693.
63. Alberti A, Pirrone P, Elia M, Waring RH, Romano C (1999) Sulfation deficit in ‘‘low-
functioning’’ autistic children: A pilot study. Biol Psychiatry 46:420–424.
64. Horvath K, Perman JA (2002) Autism and gastrointestinal symptoms. Curr Gastroen-
terol Rep 4:251–258.
65. Song Y, Liu C, Finegold SM (2004) Real-time PCR quantitation of clostridia in feces of
autistic children. Appl Environ Microbiol 70:6459–6465.
66. Parracho HMRT, Bingham MO, Gibson GR, McCartney AL (2005) Differences between
the gut microflora of children with autistic spectrum disorders and that of healthy
children. J Med Microbiol 54:987–991.
67. Strang RR (1965) The association of gastro-duodenal ulceration and Parkinson’s dis-
ease. Med J Aust 1:842–843.
68. Goodhart PJ, DeWolf WE, Kruse LI (1987) Mechanism-based inactivation of dopamine
beta-hydroxylase by p-cresol and related alkylphenols. Biochem 26:2576–2583.
69. Elliott AA, Elliot JR (1997) Voltage-dependent inhibition of RCK1 K?channels by
phenol, p-cresol, and benzyl alcohol. Mol Pharmacol 51:475–483.
70. Vanholder R, De Smet R, Lesaffer G (1999) p-Cresol: A toxin revealing many neglected
but relevant aspects of uraemic toxicity. Nephrol Dial Transplant 14:2813–2815.
71. Gaikwad NW, Bodell WJ (2001) Formation of DNA adducts by microsomal and perox-
idase activation of p-cresol: Role of quinone methide in DNA adduct formation. Chem
Biol Interact 138:217–229.
72. Caldero ´n-Guzma ´nD,etal.(2005)EffectoftolueneandcresolsonNa?,K?-ATPase,and
serotonin in rat brain. Regul Toxicol Pharmacol 41:1–5.
aromatic oxidation in human liver microsomes. Drug Metab Dispos 33:1867–1876.
Clayton et al. PNAS ?
August 25, 2009 ?
vol. 106 ?
no. 34 ?