Chemosphe e 351 (2024) 141221
A ailable online 13 Janua y 2024
0045-6535/© 2024 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY-NC license (h p://c ea i ecommons.o g/licenses/by-
nc/4.0/).
De elopmen and e alua ion o a comp ehensi e wo k low o suspec
sc eening o exposome- ela ed xenobio ics and phase II me aboli es in
di e se human bio luids
Mikel Musa adi
a
,
b
,
*
, In´
es Bacie o-He n´
andez
a
,
b
, Aile e P ie o
a
,
b
, Mai ane Oli a es
a
,
b
,
Nes o E xeba ia
a
,
b
, Ola z Zuloaga
a
,
b
a
Depa men o Analy ical Chemis y, Uni e si y o he Basque Coun y (UPV/EHU), 48940, Leioa, Basque Coun y, Spain
b
Resea ch Cen e o Expe imen al Ma ine Biology and Bio echnology, Uni e si y o he Basque Coun y (UPV/EHU), 48620, Plen zia, Basque Coun y, Spain
HIGHLIGHTS GRAPHICAL ABSTRACT
•A comp ehensi e suspec sc eening
wo k low was de eloped o 4 human
bio luids.
•The “peak a ing” pa ame e disca ded
bad peaks a oiding peak a ea
h esholds.
•Type I e o s we e no ed o bisphenol Z
and benzyl pa aben.
•Type II e o s occu ed o 14 xenobi-
o ics in s anda ds and 29 in spiked
bio luids.
•An inclusion mass lis o endogens was
selec ed o e an exclusion mass lis .
ARTICLE INFO
Handling Edi o : De ek Mui
Keywo ds:
Suspec and non- a ge sc eening (SNTS)
Human bio luids
“Peak a ing”
Endogenous molecules
Inclusion and exclusion suspec lis s
Type I and II e o s
ABSTRACT
Suspec and non- a ge sc eening (SNTS) me hods a e being p omo ed in o de o decode he human exposome
since a wide chemical space can be analysed in a di e si y o human bio luids. Howe e , SNTS app oaches in he
exposomics ield a e in a-s udied in compa ison o en i onmen al o ood moni o ing s udies. In his wo k, a
comp ehensi e suspec sc eening wo k low was de eloped o anno a e exposome- ela ed xenobio ics and phase
II me aboli es in di e se human bio luids. P ecisely, human u ine, b eas milk, sali a and o a ian ollicula luid
we e employed as samples and analysed by means o ul a-high pe o mance liquid ch oma og aphy coupled
wi h high esolu ion andem mass spec ome y (UHPLC-HRMS/MS). To au oma e he wo k low, he “peak
a ing” pa ame e implemen ed in Compound Disco e e 3.3.2 was op imized o a oid ime-consuming manual
e ision o ch oma og aphic peaks. In addi ion, he p esence o endogenous molecules ha migh in e e e wi h
he anno a ion o xenobio ics was ca e ully s udied as he employmen o inclusion and exclusion suspec lis s. To
* Co esponding au ho . Depa men o Analy ical Chemis y, Uni e si y o he Basque Coun y (UPV/EHU), 48940, Leioa, Basque Coun y, Spain.
E-mail add ess: [email p o ec ed] (M. Musa adi).
Con en s lis s a ailable a ScienceDi ec
Chemosphe e
jou nal homepage: www.else ie .com/loca e/chemosphe e
h ps://doi.o g/10.1016/j.chemosphe e.2024.141221
Recei ed 30 Sep embe 2023; Recei ed in e ised o m 7 Decembe 2023; Accep ed 12 Janua y 2024
Chemosphe e 351 (2024) 141221
2
e alua e he wo k low, limi s o iden i ica ion (LOIs) and ype I and II e o s (i.e., alse posi i es and nega i es,
espec i ely) we e calcula ed in bo h s anda d solu ions and spiked bio luids using 161 xenobio ics and 22
me aboli es. Fo 80.3 % o he suspec s, LOIs below 15 ng/mL we e achie ed. In e ms o ype I e o s, only wo
cases we e iden i ied in s anda ds and spiked samples. Rega ding ype II e o s, he 7.7 % e o s accoun ed in
s anda ds inc eased o 17.4 % in eal samples. Las ly, he use o an inclusion lis o endogens was a ou ed since
i a oided 18.7 % o po en ial ype I e o s, while he exclusion lis caused 7.2 % o ype II e o s despi e making
he anno a ion wo k low less ime-consuming.
1. In oduc ion
In he yea 2005, Ch is ophe Paul Wild aised he idea ha he
human genome should be complemen ed wi h he “exposome” (i.e., li e-
cou se exposu es om he p ena al pe iod onwa ds) o iden i y he
o igin o ch onic diseases (Wild, 2005). In ha con ex , human bio-
moni o ing eme ged as a ool o unde s and he isk o chemicals
h ough hei analysis in human bio luids (Ganzleben e al., 2017).
Suspec and non- a ge sc eening (SNTS) me hods based on high-
esolu ion mass spec ome y (HRMS) p o ide an added alue in he
ough ask o unde s anding he exposome since a wide ange o
chemicals (including me aboli es and/o ans o ma ion p oduc s) can
be sc eened (Pou che e al., 2020). Ne e heless, he abundan and
complex da a ob ained in SNTS app oaches by da a-dependen o in-
dependen acquisi ion modes (DDA and DIA, espec i ely) equi es
au oma ion o da a p ocessing (Chen e al., 2022; Samanipou e al.,
2018). To sol e ha issue, bo h comme cial and open-access so wa e
ha au oma ically pe o m peak de ec ion, alignmen , g ouping (i.e.,
gene a ion o unique “ ea u es”) and in eg a ion be o e s uc u e
elucida ion a e a ailable (Chen e al., 2022; Hollende e al., 2023).
Al hough a wide a ie y o ools a e use ul o s uc u al elucida ion,
such as spec al lib a ies (e.g., mzCloud, METLIN, MassBank) and in-
silico ools (e.g., Mass F on ie , Me F ag), p e ious ea u e il e ing o
p io i iza ion is c ucial. Some s a egies include common ac ions, such
as mass e o s, iso opic pa e n i s, peak a ea/in ensi y h esholds, o
signal- o-noise a ios (S/N). Howe e , mo e speci ic in o ma ion can be
used depending on he used ins umen a ion (e.g., collision c oss sec ion
(CCS) alues in ion-mobili y) o cha ac e is ics o he suspec s (e.g.,
neu al losses, mass de ec , ele an agmen s) (Gonz´
alez-Gaya e al.,
2021a; Hollende e al., 2023). Ne e heless, peak picking and he
assu ance o Lo en zian peak-shaped ea u es a e s ill unsol ed and
edious bo lenecks wi hin SNTS wo k lows. Besides, when comme cial
so wa e such as Compound Disco e e (The mo Fishe Scien i ic) is
used, some s eps like in-silico agmen a ion a e manually applied o
each candida e o e e y ea u e o be anno a ed. In ha sense, peak
e ision is necessa y o disca d unaccep able peaks ha should no be
submi ed o MS2 in e p e a ion (Gonz´
alez-Gaya e al., 2021a). Conse-
quen ly, pa ame e s ha disca d low-quali y peaks ha e been ecen ly
implemen ed and a e p omising o speed up ime-consuming manual
wo k bu a e s ill in a-s udied (Musa adi e al., 2023).
Mo eo e , ega ding he anno a ion wo k lows, some speci ic chal-
lenges o SNTS in bio luids ha e been poo ly add essed in he li e a u e
since he sc eening o xenobio ics in human samples has been less
s udied in compa ison o en i onmen al and ood ma ices (Hajeb e al.,
2022). On he one hand, comp ehensi e wo k lows a e equi ed since
se e al human bio luids can be employed o exposome e alua ion, such
as, u ine (Hube e al., 2022; Tkalec e al., 2022b), blood/plasma
(Bandow e al., 2020; Sunye -Caldú e al., 2023), b eas milk (Alcala and
Phillips, 2017; Baduel e al., 2015), sali a (Moscoso-Ruiz e al., 2022;
Mullangi e al., 2009) o e en o a ian ollicula luid (Hallbe g e al.,
2021; Li e al., 2023). On he o he hand, endogenous subs ances, which
a e a highe concen a ions in such ma ices, could in e e e wi h he
iden i ica ion o xenobio ics and need o be ca e ully conside ed.
Ne e heless, ew wo ks ake in o accoun he p esence o endogenous
subs ances using di e en s a egies, o ins ance, inclusion o exclusion
lis s and e en manual elimina ion o endogenous candida es (Musa adi
e al., 2023; Plassmann e al., 2015; Roggeman e al., 2022). Fu he -
mo e, o ganic xenobio ics unde go me abolism in he human body and
a e mo e p ominen ly p esen as me aboli es. Those me aboliza ion e-
ac ions a e phase II (i.e., oxida ion, dealkyla ion) and phase II (conju-
ga ion eac ion o ob ain glucu onides, sulpha es and glycine) eac ions
so he p esence o hose me aboli es in human samples should be also
add essed (Meije e al., 2021).
A ha monised p ocedu e is ha d o implemen in he da a p e-
p ocessing s ep bu he es ablishmen o common quali y con ol/qual-
i y assu ance (QC/QA) measu es in he SNTS anno a ion wo k lows is
necessa y (Pou che e al., 2020). Besides he abo e-men ioned p io i-
iza ion c i e ia, ype I and II e o s ( alse posi i es and nega i es,
espec i ely) and he limi s o iden i ica ion (LOIs) should be also e-
po ed in SNTS me hods besides he anspa ency in communica ing
anno a ion con idence (Hollende e al., 2023).
All in all, his wo k aimed o de elop a comp ehensi e suspec
sc eening wo k low o sc een exposome- ela ed chemicals (including
phase II me aboli es) in di e se human bio luids, such as u ine, sali a,
b eas milk and ollicula luid. To ul il ha objec i e, he op imiza ion
o se e al pa ame e s ha acili a e compound anno a ion, he calcu-
la ion o LOIs, ype I and II e o s, and he s udy o he bes s a egy o
conside endogenous subs ances we e se as sub-objec i es.
2. Ma e ials and me hods
2.1. Sample p epa a ion and UHPLC-HRMS/MS analysis
Human u ine, sali a, ollicula luid, and b eas milk we e employed
as bio luids. In o med consen was ob ained om each subjec be o e
conduc ing he expe imen s and he samples we e handled acco ding o
he indica ions o (i) The Commi ee on E hics o Resea ch on Biological
Agen s and Gene ically Modi ied O ganisms (CEIAB-UPV/EHU, BOPV
32, Feb ua y 17, 2014, M30-2021-158, M30-2022-311, M30-2022-327,
M30-2023-136) and (ii) The Commi ee on E hics o Resea ch In ol ing
Human Subjec s (CEISH-UPV/EHU, BOPV 32, Feb ua y 17, 2014, M10-
2020-230, M10-2021-124, M10-2023-135) o he Uni e si y o he
Basque Coun y (UPV/EHU).
All he speci ic de ails ega ding sample p epa a ion and analysis by
ul a-high pe o mance liquid ch oma og aphy coupled o high esolu-
ion andem mass spec ome y (UHPLC-HRMS/MS), including quali y
con ol/quali y assu ance (QC/QA), a e de ailed elsewhe e (Bacie -
o-He n´
andez e al., 2024). B ie ly, u ine and sali a samples we e ea ed
wi h solid phase ex ac ion (SPE) using Oasis HLB ca idges, while milk
and ollicula luid samples we e ex ac ed wi h sal -assis ed liquid--
liquid ex ac ion (SALLE) ollowed by p o ein p ecipi a ion a low
empe a u es. In addi ion, ollicula luid ex ac s we e cen i uged o
ensu e maximum clogging o p o eins, while an addi ional clean-up s ep
using Cap i a EMR-Lipid il e s was pe o med wi h he milk ex ac s.
No hyd olysis was pe o med in o de o sc een possible phase II me-
aboli es in he samples. Fo he UHPLC-HRMS/MS analysis, a Dionex
Ul ima e 3000 UHPLC coupled o a high-pe o mance Q Exac i e Focus
O bi ap (qO bi ap, The mo Fishe Scien i ic) mass analyse wi h a
hea ed elec osp ay ioniza ion (HESI) sou ce was used. The analyses
we e pe o med a he Full Scan – da a dependan MS
2
(Full MS –
ddMS
2
) disco e y acquisi ion mode a posi i e and nega i e ioniza ion,
so wo uns we e pe o med pe sample using an ACE Ul aCo e 2.5
M. Musa adi e al.
Chemosphe e 351 (2024) 141221
3
Supe C18 column (2.1 mm ×100 mm, 2.5
μ
m) column a acidic (pH =
2.5) and basic (pH =10.5) mobile phase condi ions, espec i ely.
2.2. Suspec sc eening wo k low
The suspec sc eening wo k low was designed using he Compound
Disco e e 3.3.2 so wa e (The mo Fishe Scien i ic) based on he
esea ch g oup’s expe ience (Gonz´
alez-Gaya e al., 2021b; Musa adi
e al., 2022) wi h adap a ions o simul aneously sc een xenobio ics and
phase II me aboli es in human samples. Fi s , he so wa e p e-p ocessed
he UHPLC-HRMS/MS da a and consis ed o e en ion ime alignmen ,
compound de ec ion, compound g ouping (e.g., composi ion p edic ion,
pa e n sco ing, neu al loss sea ch, lib a y sea ch, mass lis sea ch), gap
illing and backg ound sub ac ion. Those pa ame e s a e de ailed in
Table S1 o he Supplemen a y In o ma ion (SI).
Then, wi h he da a e ie ed by he so wa e, he suspec sc eening
wo k low was applied. Fo compound anno a ion, an in-house mass lis
o 7055 exposome- ela ed xenobio ics ob ained om NORMAN (h
ps://www.no man-ne wo k.com/?q=suspec -lis -exchange) and 7670
phase II me aboli es in-silico simula ed by BioT ans o me 3.0 (h p://
bio ans o me .ca/) we e employed. The ull mass lis s con aining he
name, molecula o mula, exac mass and chemical s uc u e (i.e.,
Simpli ied Molecula -Inpu Line-En y Sys em (SMILES)) o each sus-
pec can be ound in Table S2 in he SI. In addi ion, he in o ma ion
abou he phase II eac ion and he p ecu so is indica ed o he me-
aboli es. Rega ding phase I me aboli es, only phase I me aboli es o
ph hala es we e included (hyd oly ic monoes e s) since he es o sus-
pec s would likely unde go phase II me aboliza ion due o hei pola
unc ional g oups. Mo eo e , ano he in-house mass lis o 707 human
endogenous subs ances ob ained om he Human Me abolome Da abase
(HMDB, h ps://hmdb.ca/me aboli es) and 1739 phase II me aboli es o
hose endogens simula ed by BioT ans o me 3.0 was also employed
(see Table S3 in he SI). The mass lis o a o al o 2446 endogenous
molecules was used ei he as an inclusion lis o as an exclusion lis (see
sec ion 3.4).
F om he ini ial lis o ea u es (each wi h one o mo e po en ial
candida es om he mass lis s) e ie ed by Compound Disco e e a e
he ini ial p e-p ocessing, he peak a ing pa ame e and he minimum
peak a ea we e op imized o educe he numbe o ea u es ha should
be manually e ised o submi o MS2 e alua ion. Addi ionally, only
ea u es ha con ained a he e oa om in hei molecula o mula and a
mass e o lowe han 5 ppm we e conside ed. The peak a ing il e
allows he emo al o low-quali y ch oma og aphic peaks wi hou
se ing an o e ly es ic i e minimum peak in ensi y h eshold. To ha
end, he peak a ing pa ame e conside s peak quali y ac o s (Musa adi
e al., 2023), ela i e peak a ea, and he ela i e s anda d de ia ion
(RSD). Fo mo e de ailed in o ma ion, see he Compound Disco e e
Use Guide o LC S udies (So wa e Ve sion 3.3 SP2).
The agmen a ion spec a we e e alua ed using he mzCloud mass
spec al da abase (h ps://www.mzcloud.o g/), se ing a 70 % ma ch
h eshold be ween he expe imen al MS2 spec a and he ones collec ed
in he lib a y. I he MS2 spec um was no a ailable o a candida e, in-
silico agmen a ion was simula ed using Mass F on ie 8.0 so wa e
implemen ed in Compound Disco e e 3.3.2. Fo a sa is ac o y ma ch,
70 % o he majo agmen s should be explained when h ee o mo e
agmen s we e p esen in he MS2. When only one o wo majo ag-
men s we e p esen in he MS2, one o hem should be explained (Hol-
lende e al., 2023). In he case o glucu onides, he glucu onide loss
should be explained, and i he molecule con ained Cl, B o S a oms, he
MS1 should be explained. Fo he la e , he SFi pa ame e (i.e., he
spec al simila i y sco e be ween de ined and measu ed iso ope pa e n)
alue highe han 50 % and he pa e n co e age (i.e., ma ched in ensi y
pe cen age o he heo e ical pa e n) highe han 70 % we e
es ablished.
Las ly, he e en ion ime (RT) was conside ed. I possible, pu e
s anda ds we e used and a ±0.2 min e o was accep ed. I no , he
Re en ion Time Indices (RTI) pla o m (h p:// i.chem.uoa.g /) was
employed. In ha case, he measu ed RT and he expe imen al RT
p o ided by he RTI model buil wi h RTI calib an s (see Table S4 in he
SI) should be compa able (box1 and box2 le els we e accep ed) (Aali-
zadeh e al., 2021). Wi h all he in o ma ion men ioned abo e, he
anno a ion con idence le els we e de ined based on he Schymanski
scale (Schymanski e al., 2014).
2.3. Quali y con ol/quali y assu ance (QC/QA)
To e alua e he wo k low, ype I e o s (i.e., alse posi i es) and ype
II e o s (i.e., alse nega i es) we e calcula ed in bo h s anda d solu ions
and spiked samples. In he case o he s anda d solu ions, 15 concen-
a ion poin s be ween 0.1 and 200 ng/mL we e p epa ed o 183
exposome- ela ed xenobio ics (see Table S4 in he SI). Type I e o s we e
de ined as he pe cen age o xenobio ics w ong ully iden i ied using he
wo k low, while he pe cen age o xenobio ics ha could no be anno-
a ed was de ined as ype II e o s.
Concen a ion le els below 50 ng/mL we e injec ed in iplica e, and
limi s o iden i ica ion (LOIs) we e es ablished om he lowes con-
cen a ion poin in which he compound could be co ec ly anno a ed in
a leas wo eplica es wi h he wo k low (Ve geyns e al., 2015). Fo
compa ison, limi s o de ec ion (LODs) we e also calcula ed in he
s anda d solu ions as he lowes concen a ion poin injec ed in ipli-
ca e wi h an accep able peak shape and a ela i e s anda d de ia ion
(RSD) below 30 % using T aceFinde 5.1 so wa e (The mo Fische
Scien i ic) o peak in eg a ion (Musa adi e al., 2023).
Type I and II e o s we e also es ima ed in he 4 bio luids o conside
po en ial ex ac ion de iciencies and ma ix e ec s. Fo each luid, a
pooled sample (n =5) was spiked wi h he 183 xenobio ics in o de o
ob ain a 100 ng/mL concen a ion in he inal ex ac , and p ocessed
wi h hei espec i e sample p epa a ion p ocedu e (Bacie o-He n´
andez
e al., 2024).
In o de o s udy he in luence o endogenous me aboli es on he
iden i ica ion, he use o he in-house gene a ed endogenous suspec lis
(including phase II me aboli es) was conside ed in wo app oaches in he
da a pos -p ocessing: (i) inclusion lis and (ii) exclusion lis (see sec ion
3.4 o de ails). The s udy was no limi ed o he spiked compounds bu
o he compounds p esen in he samples we e conside ed o ge a wide
iew o he in e e ences in he anno a ion wo k low caused by
endogens.
Las ly, a signal- o-noise (S/N) a io highe han 10 was es ablished as
compulso y QC/QA c i e ia o a oid he iden i ica ion o a e ac s. In he
s anda d solu ions, ch oma og aphic a eas o he solu ions con aining
he xenobio ics (S) and a eas o pu e injec ion sol en (N) we e
conside ed, while in he bio luids, he a eas o spiked samples (S) and
non-spiked syn he ic bio luids (N) we e used. In he spiked samples, (i )
a RSD lowe han 30 % wi hin he h ee eplica es o each bio luid was
also equi ed.
3. Resul s and discussion
3.1. Op imiza ion o he suspec sc eening wo k low’s pa ame e s
The peak a ing pa ame e in oduced in Compound Disco e e 3.3.2
aims o il e ou low-quali y ch oma og aphic peaks since an app o-
p ia e in ensi y o peak a ea h eshold is di icul o es ima e o
un a ge ed expe imen s. In ha sense, he pa ame e was op imized
es ing alues in he 2–8 ange, wi h highe alues meaning s ic e
c i e ia. The con ibu ion o each pa ame e included in he “peak a -
ing”, indica ed as he p opo ion o he con ibu ion compa ed o o he
con ibu o s, was selec ed acco ding o he endo speci ica ion and is
de ailed in Table S1.
To op imize he peak a ing pa ame e , spiked human bio luids we e
used ins ead o pu e s anda d solu ions o conside he e ec ha he
ma ices migh ha e on he peak shapes. In ha sense, se e al measu es
M. Musa adi e al.
Chemosphe e 351 (2024) 141221
4
and s eps explained he eina e we e de ined. Wi h he ini ial p e-
p ocessing (Table S1), a o al o 25,801 and 27,483 molecula ea u es
we e ob ained in he posi i e and nega i e modes, espec i ely.
Then, ea u es wi hou MS2 and he e oa oms, S/N a ios lowe han
10, and mass e o s highe han 5 ppm we e disca ded. As a s a ing
poin , he ch oma og aphic peak a eas we e se o alues ha p o ided a
iable numbe o ea u es o be s udied indi idually. Consequen ly, he
a ea h eshold was se o 10
7
(posi i e mode) and 10
6
(nega i e mode)
and he numbe o ea u es was educed o 2026 and 1597 in he posi i e
and nega i e modes, espec i ely. Then, all he ea u es we e indi idu-
ally e alua ed and classi ied as “accep able” o “non-accep able” based
on hei ch oma og aphic peak shape, which should ideally be
Lo en zian-shaped (Caballe o e al., 2002). Conside ing bo h modes, a
o al o 2343 peaks (64.6 %) we e conside ed “non-sa is ac o y”. To
a oid indi idual ime-consuming e ision, peak a ing alues (2–8)
we e s udied o maximize he il e ing o ea u es wi h non-accep able
peak shapes while minimizing he loss o ea u es wi h accep able Lo -
en zian peak shapes. The esul s o bo h ioniza ion modes a e
summa ized in Fig. 1(a–b).
In he posi i e mode, he op imum alue was se o ‘5’ since i p e-
se ed 91 % o ea u es wi h accep able peak shapes while 50 % o he
ea u es wi h non-accep able peak shapes we e disca ded (601 ea u es).
Al hough a peak a ing alue o ‘6’ in he posi i e ioniza ion mode
would inc ease he elimina ion o undesi able ea u es (up o 68 %), he
loss o Lo en zian peak-shaped ea u es would inc ease o 21 % so i was
no selec ed. In he nega i e mode, a peak a ing alue o 4 u ned ou o
be he bes alue in he consensus o he p ese a ion o accep able
peaks (84 % p ese ed) and he elimina ion o non-accep able peaks
(646 ea u es).
Wi h he op imum peak a ing alues, he ch oma og aphic a eas
we e minimized so as no o disca d ea u es wi h Lo en zian shapes bu
low abundance. In he posi i e mode, a maximum o 28 ea u es we e
eco e ed minimizing he a ea om 10
7
o 10
6
. Howe e , no mo e
ea u es we e eco e ed se ing he a ea below 10
6
. In he nega i e
mode, 12 addi ional ea u es we e gained lowe ing he a ea om 10
6
o
5⋅10
5
. Al hough hose a eas seem o be ela i ely high, he peak a ing
Fig. 1. “Accep able” and “non-accep able” peak shapes a se e al “peak a ing” alues (2–8) a he (a) posi i e and (b) nega i e mode.
M. Musa adi e al.
Chemosphe e 351 (2024) 141221
5
pa ame e showed ha no mo e accep able ea u es in e ms o peak
shape we e e ie ed below hose o de s o magni ude. In ac , simila
peak a ea h eshold alues (10
5
–10
7
) a e es ablished in o he SNTS
wo ks using qO bi ap de ec ion (L´
opez e al., 2016; Musa adi e al.,
2021). Consequen ly, he peak a ea/in ensi y h eshold c i e ia can be
emo ed by he implemen a ion o he adequa e “peak a ing” alue
a oiding he emo al o low in ensi y peaks ha could be in e es ing
om he exposomic poin o iew.
3.2. E alua ion o ype I and ype II e o s
Type I e o s o alse posi i es, co espond o w ong ully anno a ed
subs ances ha a e no p esen in he sample. Ne e heless, hose e o s
a e di icul o assess and hey a e no epo ed in SNTS wo ks in he
li e a u e, especially using human luids. Type II e o s o alse nega i es
ins ead, a e subs ances p esen in he sample ha a e no epo ed in he
inal lis . Ne e heless, hose losses can occu a di e en poin s in he
analy ical p ocess, including he sample p epa a ion s ep (e.g., ex ac-
ion de iciencies), LC-HRMS/MS analysis (e.g., co-elu ion, poo ioniza-
ion, ma ix e ec s a de ec ion) o da a p ocessing (e.g., il e s, he
inadequacy o in-silico ools, limi ed anno a ion lis s, selec ion o MS2
da abases) (Hollende e al., 2023). In his s udy, ype I and II e o s
we e e alua ed in he (i) calib a ion solu ions and (ii) spiked eal bio-
luids using he suspec sc eening wo k low as i he compounds spiked
we e unknowns. All he esul s a e ga he ed in Table S5 in he SI.
In he calib a ion solu ions, 167 analy es we e success ully anno-
a ed om he o al lis o 183 (91.3 %). In mos cases, he candida es
we e included in he mzCloud lib a y and o he es in-silico ag-
men a ion was simula ed. I should be men ioned ha mos phase II
me aboli es a e no included in spec al lib a ies (Baduel e al., 2015) as
i happened wi h he me aboli es included in his wo k (i.e., bisphenol A
glucu onide, bisphenol A sul a e, 4-me hylumbelli e yl glucu onide,
1-hyd oxypy ene glucu onide, and 8:2 luo o elome alcohol glucu o-
nide) ha we e sc eened ia in-silico agmen a ion. The e o e, he
gene a ion o mass lis s including he s uc u es o me aboli es is c i ical
o hei anno a ion. In he cases o phase II me aboli es, he neu al loss
o he conjuga e g oup can be used o u he anno a ion c i e ia. The
neu al loss co esponding o he glucu onide (C
6
H
8
O
6
, 176.03209 Da)
is implemen ed in he so wa e, while he neu al loss co esponding o
he sul a e loss (as SO
3
, 79.9568 Da) needs o be manually checked and
should be in oduced in upcoming e sions.
Howe e , ype I e o s we e iden i ied o benzyl pa aben and
bisphenol Z (BZP) in he calib a ion solu ions. Benzyl pa aben was an-
no a ed as 2-hyd oxy-4-me hoxybenzophenone o 4-(benzyloxy)benzoic
acid, which bo h ha e he same molecula o mula (C
14
H
12
O
3
,
227.07094 Da) as benzyl pa aben bu he s uc u es a e di e en . 4-
(benzyloxy)benzoic acid is included in he mzCloud lib a y and he MS2
ma ch agains i was 83.9 % (Fig. 2a). In he cases o 2-hyd oxy-4-
me hoxybenzophenone and benzyl pa aben, hey a e no included in
he lib a y so in-silico agmen a ion was simula ed (Fig. 2b and c) wi h
Mass F on ie explaining 4 and 3 agmen s ou 5, espec i ely (80 %
and 60 % MS2 ma ch). Consequen ly, benzyl pa aben should be dis-
ca ded acco ding o he es ablished c i e ia and a w ong name would be
p o ided o ha ea u e.
In he case o BZP, h ee suspec s we e possible: BZP, die hyls il-
bes ol and equilin, which ha e he same exac mass (268.14632 Da)
and molecula o mula (C
18
H
20
O
2
). Al hough BZP was spiked, he
ea u e was anno a ed as die hyls ilbes ol. F om he p oposed 3 can-
dida es, only die hyls ilbes ol was included in mzCloud so in-silico
agmen a ion was simula ed o BZP and equilin. Al hough he ag-
men a ion was poo , die hyls ilbes ol passed he mzCloud ma ch
c i e ia (83.7 % ma ch), while in-silico agmen a ion was no able o
explain he MS2 spec a i he s uc u es o BPZ and equilin we e
conside ed (see Fig. 3). A p e ious wo k o he esea ch g oup also
showed weaknesses o he in-silico agmen a ion ool, especially o
compounds ionized in he nega i e mode due o poo MS2 spec a
(Musa adi e al., 2022).
In he speci ic cases o ace amip id-N-desme hyl, ca bo u an phenol,
e hyl 4-dime hylaminobenzoa e, nap oxen and amadol, o he possible
candida es ha we e no s uc u al isome s passed he c i e ia besides
he spiked a ge . Howe e , o hose candida es, he RT was con i med
wi h he RTI model, which has an unce ain y (R
2
=0.965 o he pos-
i i e mode and R
2
=0.921 o he nega i e mode). The e o e, hey could
be disca ded agains he candida e con i med wi h he pu e s anda d bu
in i s absence, hey would become ype I e o s as well.
In addi ion o he ype I e o s, some suspec s we e no de ec ed in
he s anda d solu ions ( ype II e o s) due o di e en scena ios. Fo
dime hyl phospha e, 6-chlo onico inic acid, bendioca b, 3-(2-chlo o-
3,3,3- i luo op op-1-en-1-yl)-2,2-dime hylcyclop opaneca boxylic
acid, 2,6-di- e -bu yl-4-(dime hylaminome hyl)phenol, 2-e hylhexyl 4-
(dime hylamino)benzoa e, pendime halin, bisphenol P and ip onil
desul inyl, only he molecula o mula could be elucida ed since he
MS2 ma ch was lowe han 70 %. Rega ding e hylpa aben, mecop op,
2,4-die hyl-9H- hioxan hen-9-one, chlo oxu on and diphenhyd amine-
N-glucu onide, ins ead, no e en he molecula o mula could be
elucida ed e en a he mos concen a ed calib a ion solu ion (200 ng/
mL). The e o e, conside ing he analy es wi h bad MS1 (5) and MS2 (9)
as unsa is ac o y, 7.7 % o ype II e o s we e iden i ied in he s anda d
solu ions.
In spiked and ex ac ed samples, he same ype I e o s occu ed o
benzyl pa aben and BZP. Howe e , no o he ype I e o was iden i ied
so i can be concluded ha he sample ea men p ocedu es o he
di e en ma ices did no ha e a signi ican impac in e ms o alse
posi i es (only 2 ou o 183). On he con a y, ype II e o s we e
inc emen ed in he spiked bio luids. F om he 167 analy es ha we e
co ec ly sc eened in he s anda d solu ions, 29 could no be sc eened in
he ex ac ed bio luids (17.4 %). Fo 14 spiked analy es (ace amino-
phen, gabapen in, p opamoca b, 2-(4-chlo ophenyl)-3-me hylbu y ic
acid, hyd oxypy ene, a enolol, alachlo , so alol, iclosan, deslo -
a adine, ani idine, chlo py i os-me hyl, amb oxol, bisphenol A glucu-
onide), poo peak shapes we e ob ained so hey we e disca ded a e
applying he peak a ing pa ame e . In he cases o ace amip id and
genis in, al hough hey we e no disca ded wi h he peak a ing
pa ame e , he quali y o he MS1 spec a was no good enough o
elucida e he molecula o mula.
In he pa icula cases o pe luo ooc anesul onamide (PFOSA) and
pe luo ooc anesul onic acid (PFOS), which a e pe luo ina ed mole-
cules, hey could be p ope ly anno a ed by op imizing he wo k low’s
hyd ogen- o-ca bon a om a io (H/C) pa ame e . In he suspec
sc eening wo k low employed in his wo k, he H/C a io was es ab-
lished in he 0.1–3.5 ange (de aul condi ions). To elucida e molecula
o mulas o pe - and poly-subs i u ed chemicals, he a io should be
dec eased since he so wa e p io i izes molecula o mulas wi h highe
a ios. Howe e , ha migh esul in mo e possible molecula o mulas
o each ea u e and, he e o e, an inc ease in ype I e o s. Conse-
quen ly, ha pa ame e should be ca e ully e alua ed in u u e wo ks
aimed o sc een pe - and poly-subs i u ed chemicals alongside o he
s a egies like he mass de ec (O ´
o-Nolla e al., 2023; Rehns am e al.,
2023).
Fo 4-hyd oxybenzophenone, benzophenone-1, dichlo os,
benzophenone-3, me hyl 3,5-di- e -bu yl-4-hyd oxybenzoa e, en hion
and iadimenol, he MS2 ma ch was lowe han he es ablished
h eshold. Fo me o min, ins ead, no MS2 spec um was acqui ed due o
i s non- e en ion in he non-pola column and co-elu ion wi h pola
molecules. Las ly, me hylpa aben, ie hyl phospha e and dodemo ph
we e also ype II e o s because he S/N a io was lowe han 10 showing
blank p oblems.
All in all, ex ac ion de iciencies and/o ma ix e ec s alongside he
ype II e o s in s anda d solu ions summed up o 23.5 % o ype II e o s
o alse nega i es (43 suspec s in o al). Rega ding he di e ences
obse ed among he bio luids, 95 % (131 ou o 138) o spiked suspec s
passed all he c i e ia in he ollicula luid, while he pe cen age
M. Musa adi e al.
Chemosphe e 351 (2024) 141221
6
Fig. 2. MS2 ma ch o possible candida es o he ea u e C
14
H
12
O
3
([M −H]
-
=227.07094 Da, RT =7.3 min): (a) 4-(benzyloxy)benzoic acid using mzCloud ( he
spec um abo e is he expe imen al one while he spec um below is he one included in he lib a y), (b) 2-hyd oxy-4-me hoxybenzophenone in-silico simula ion and
(c) benzyl pa aben in-silico simula ion. G een do s indica e agmen s explained by in-silico simula ion.
M. Musa adi e al.
Chemosphe e 351 (2024) 141221
7
dec eased o 78 % and 71 % in b eas milk and u ine, espec i ely.
Las ly, only 60 % o he xenobio ics we e sc eened in he pooled sali a
sample, sugges ing ha sample p epa a ion should be s udied in de ail
in he u u e o ha ma ix. No u he discussion was made ega ding
he pa icula i ies o each ma ix and hei ole in decoding he expo-
some, since he p esen wo k was ocused on he da a p ocessing pa
a he han sample p epa a ion and/o exposome- ela ed conclusions.
Wi h espec o he SNTS s udies in he li e a u e, he e a e no
de ined accep ance c i e ia o ype I and II e o s as o o he QC/QA
pa ame e s (Caballe o-Case o e al., 2021) and should be mo e
equen ly calcula ed o assess he pe o mance o non- a ge ed
me hods. Due o he labo ious SNTS wo k lows, a h eshold o 25–30
% o e o s could be conside ed as sa is ac o y. In addi ion, in e nal
s anda ds (IS) could be used o indi idually spiked all samples and
de ec e o s in la ge moni o iza ion wo ks aiming o analyse a
conside able numbe o samples since hey a e no na u ally p esen in
he samples. In ac , he use o IS is p omo ed in he SNTS wo ks aiming
o homogenise QC/QA pa ame e s (Hollende e al., 2023). Howe e ,
o he pu pose o his wo k, 183 subs ances ha migh be p esen in eal
samples we e employed o ge a wide iew o ype I and II e o s.
3.3. Limi s o iden i ica ion (LOIs)
LOIs we e de ined om he lowes concen a ion calib a ion solu ion
whe e he suspec could be anno a ed. In he cases whe e he analy e
was anno a ed a bo h ioniza ion modes (e.g., 4-hyd oxybenzophenone,
p opanil, alsa an), he lowe LOI alue was p o ided.
In he li e a u e, ew wo ks ha mos ly ocus on en i onmen al
samples de ine LOIs (Dasenaki e al., 2015; Liu e al., 2019; Segu a e al.,
2019). In some cases, he limi s o de ec ion (LODs) a e p o ided (Ge -
zinge and Fe guson, 2021; Pica do e al., 2020; Tkalec e al., 2022a)
al hough hey do no i he pu pose. As a consequence, he lack o ho-
mogenei y in QC/QA measu es o SNTS is enhanced. As a compa a i e,
ins umen al LODs we e also calcula ed (see sec ion 2.3) o he analy es
included in his wo k o con as wi h hei espec i e LOIs. All he
speci ic alues a e included in Table S5 in he SI o each suspec .
As can be seen in Fig. 4, 164 analy es (89.1 %) p o ided LOD alues
below 5 ng/mL. P ecisely, he a e age alue was 1.9 ng/mL. In he case
o LOIs, only 75 compounds (41.0 %) p o ided alues below 5 ng/mL,
while a simila pe cen age (39.3 %) was in he 5–15 ng/mL ange. Only
in he cases o benzophenone-2 and mono-benzyl ph hala e, he LOD
alues u ned ou o be highe han he espec i e LOI. I should be aken
in o accoun ha he LOD de ini ion is s ic e in e ms o p ecision.
Las ly, o 16 analy es LOIs we e no calcula ed since hey we e ei he
ype I (2) o ype II (14) e o s (sec ion 3.2).
Consequen ly, he usage o LODs could lead o misleading ou comes
abou he pe o mance o he me hod, and he e o e, an impo an s ep
owa ds he homogenisa ion o SNTS me hods migh be he usage o
Fig. 3. A poo MS2 ma ch using in-silico agmen a ion o he ea u e C
18
H
20
O
2
([M −H]
-
=267.13916 Da, RT =12.3 min) o BPZ and equilin as candida es. G een
do s indica e agmen s explained by in-silico simula ion.
Fig. 4. Ins umen al limi s o de ec ion (LODs) and iden i ica ion (LOIs) o he 183 xenobio ics.
M. Musa adi e al.
Chemosphe e 351 (2024) 141221
8
LOIs. Howe e , i he ul ima e aim is o quan i y a e con i ming he
p esence o a suspec , a ully alida ed a ge me hod should be
employed in which eliable limi s o quan i ica ion need o be de ined
(Gonz´
alez e al., 2014).
3.4. Endogenous subs ances
The in luence o endogenous subs ances ha a e ound in human
bio luids should be ca e ully conside ed in SNTS wo ks o a oid ype I
e o s (i.e., anno a ing an exogenous subs ance ha could also be an
exogenous chemical) o ype II e o s (i.e., no anno a ing he xenobio ic
due o he p esence o endogenous molecule). To add ess ha issue, a
mass lis o endogenous subs ances and hei espec i e phase II me-
aboli es (sec ion 2.2) was used in his s udy bo h as an inclusion and
exclusion lis bu always a e da a acquisi ion. One o he nodes in
Compound Disco e e 3.3.2 allows including suspec lis s o compounds
ha can be used du ing he pos -p ocessing o he da a. In ha sense,
apa om he suspec lis o exogenous and hei phase II me aboli es, a
lis o endogenous me aboli es was included. In his app oach, du ing
he pos p ocessing o he da a, he so wa e e ie ed only he ea u es
wi h a ma ch in he exogenous suspec lis . Howe e , we could know
whe he he e was also a candida e in he endogenous lis apa om he
exogenous candida e. Unlike in he p e ious sec ions, all he ea u es
de ec ed in he bio luids besides he spiked suspec s we e conside ed.
3.4.1. Inclusion lis
Using he mass lis o exogenous subs ances as an inclusion lis does
no disca d any ea u e un il he inal suspec lis is p o ided. In ha
app oach, all endogenous subs ances we e conside ed as possible can-
dida es o he ea u e and, hus, hei MS2 and RT we e also e alua ed
when applicable. I an endogenous me aboli e candida e passed all he
es ablished c i e ia o a gi en ea u e, he ea u e was emo ed in he
inal s ep e en i a xenobio ic compound passed he c i e ia as well
a oiding a po en ial ype I e o .
All o he ea u es ha ul illed he ini ial c i e ia o bo h posi i e
and nega i e modes we e submi ed o he MS2 e alua ion o all hei
en a i e candida es, no ma e whe he hey we e xenobio ics o
endogenous (see Fig. 5 o he whole anno a ion p ocedu e).
In he posi i e mode, 627 ea u es we e manually disca ded and se
o MS2 e alua ion, om which 380 ea u es passed he c i e ia. The
disca ded ea u es we e di ec ly se o le els 4 o 5. Finally, a o al o 274
suspec s we e anno a ed a le els 1–3 wi h a sa is ac o y RT e alua ion.
In he nega i e mode, 495 we e manually emo ed, 120 passed he MS2
ma ch and inally, a o al o 84 suspec s we e anno a ed a le els equal o
o abo e 3 due o sa is ac o y RT e alua ion.
In o al, 358 suspec s we e anno a ed a le els 1–3 wi hou consid-
e ing endogenous subs ances. Bu when he inclusion lis was included,
he p esence o he endogenous candida e(s) in he inal lis allowed he
emo al o 60 suspec s in he posi i e and 7 in he nega i e mode. The
ull lis s a e included in Tables S6 and S7 in he SI o bo h ioniza ion
modes. The e o e, 291 suspec s we e inally anno a ed conside ing bo h
ioniza ion modes. Despi e being a ela i ely la ge numbe o suspec s, i
should be bea in mind ha 183 analy es we e spiked o he samples and
he o al numbe o unique ea u es we e accoun ed ga he ing he 4
human bio luids.
Conside ing he abo emen ioned, he inclusion lis allowed he dis-
ca ding o 18.7 % le el 2a/2b–3 suspec s (67 ou o 358). In he absence
o pu e s anda ds o all he pu a i e exogenous and endogenous mol-
ecules, i canno be assu ed ha hose compounds elimina ed we e
uni ocally endogenous, bu he lack o his inclusion lis o endogenous
suspec s could lead o po en ial ype I e o s. Las ly, i was obse ed ha
phase II me aboli es o endogenous subs ances we e also sc eened in he
samples so hey ha e o be conside ed in u u e SNTS wo ks using
human samples.
Conside ing he inal lis o 291 xenobio ics anno a ed a le els 1–3
using he inclusion lis app oach, 202 suspec s we e sc eened in he
pooled u ine sample, ollowed by he ollicula luid (185). In he cases
o b eas milk and human sali a, ewe compounds we e anno a ed (145
and 116, espec i ely). Wi hou conside ing he spiked analy es, 146
unique suspec s we e sc eened om which 103 we e p esen in u ine. In
he es o he bio luids, conside ably ewe suspec s we e de ec ed,
since 49 we e iden i ied in ollicula luid, 36 in b eas milk and 32 in
sali a. Ne e heless, each bio luid can be ele an o exposome analysis
depending on he aim despi e ewe suspec s being de ec ed. As epo ed
in he li e a u e, he p esence o exogenous chemicals in ollicula luid
o b eas milk o ins ance, can be ela ed o dec eased e iliza ion a e
(Pe o e al., 2012) o b eas cance (Koual e al., 2020), espec i ely.
3.4.2. Exclusion lis
In his app oach, he endogenous suspec lis was used o il e any
ea u e wi h a leas one candida e in he lis . The e o e, he use o he
exclusion lis au oma ically disca ded a ea u e i a leas one o he
possible candida es was included in he endogenous mass lis . Tha ac-
ion in ol es a conside able educ ion in he numbe o ea u es ha a e
submi ed o manual e ision making he p ocess much mo e
s aigh o wa d.
In he posi i e ioniza ion mode, om he 1397 ea u es ha passed
he compulso y c i e ia (see sec ion 3.1), he inclusion lis educed he
numbe o 1001 ea u es (see Fig. 5). Those ea u es we e submi ed o
MS2 and u he RT s udy o e en ually anno a e 197 suspec s a le els
1–3. In he nega i e mode, he ini ial lis o 891 ea u es was educed o
644 be o e MS2 and RT e alua ion, om which e en ually 73 we e
anno a ed a con idence le els abo e 3. Tha means ha 643 ea u es
we e disca ded in o al be o e s udying hei agmen a ion spec a due
o hei p esence in he exclusion lis . In ha sense, he exclusion lis
made he da a p ocessing less ime-consuming.
On he nega i e side, some o he compounds anno a ed using he
inclusion lis app oach we e missed when using he exclusion lis . F om
he 291 suspec s anno a ed a le els 1–3 wi h he inclusion lis o
endogens, 21 we e missed using he exclusion lis . As an example, o he
ea u e wi h 177.10222 Da exac mass de ec ed in he posi i e mode, he
so wa e calcula ed he molecula o mula C
10
H
12
N
2
O, and wo candi-
da es we e possible, (i) co inine as an exogenous compound o being
he me aboli e o he psychoac i e subs ance nico ine, and (ii) se o onin
as an endogen. Wi h he exclusion lis , he ea u e was di ec ly disca ded
wi hou u he conside a ion. Howe e , wi h he inclusion lis , he MS2
spec a o bo h candida es we e e alua ed. In bo h cases, he mzCloud
lib a y con ained he spec a and he expe imen al MS2 acqui ed in he
Fig. 5. The anno a ion wo k low ollowed by conside ing he endogenous subs ances as (i) inclusion lis o (ii) exclusion lis .
M. Musa adi e al.
Chemosphe e 351 (2024) 141221
9
spiked samples and he ones a ailable in he lib a y ma ched 95.7 % and
41.5 % o co inine and se o onin, espec i ely (see Fig. 6). Conse-
quen ly, se o onin was disca ded and co inine was e en ually con i med
wi h he pu e s anda d.
All in all, despi e he exclusion lis elimina ing a conside able num-
be o ea u es (643) be o e hei MS2 and RT e alua ion, 7.2 % o ype II
e o s (21 ou o 291) we e iden i ied in he inal suspec lis in com-
pa ison o he inclusion lis (see co inine’s case). Consequen ly, he in-
clusion lis o endogenous subs ances seems a mo e sui able app oach
o disca ding na u al compounds and p ese ing ele an xenobio ics
as exposome bioma ke s in SNTS o xenobio ics in human samples.
4. Conclusions
A comp ehensi e wo k low was designed and applied o sc een
exposome- ela ed xenobio ics in a a ie y o human bio luids (u ine,
sali a, b eas milk and o a ian ollicula luid). The “peak a ing”
pa ame e implemen ed in Compound Disco e e so wa e, allowed
emo ing au oma ically a conside able numbe (1247) o no -accep able
ch oma og aphic peaks ha would ha e o be manually e ised o he -
wise. In addi ion, he minimum peak a ea h eshold could be a oided,
which is a con o e sial pa ame e in SNTS wo k lows. Mo eo e , ype I
and II e o s we e calcula ed in s anda d solu ions and spiked human
bio luids using 183 di e se xenobio ics. BPZ and benzyl pa aben we e
iden i ied as ype I e o s in bo h cases since he name o o he suspec s
was w ong ully p o ided o hose spiked analy es (i.e., alse posi i es).
In e ms o ype II e o s (i.e., alse nega i es), 7.7 % and 17.4 % o
spiked suspec s could no be sc eened in he s anda d solu ions and
spiked bio luids due o se e al easons (e.g., peak shape, MS1 o MS2
e o s), espec i ely. Las ly, a c i ical aspec o he SNTS me hods in
bio luids was s udied, which is he p esence o endogenous subs ances.
Using an exclusion lis o endogens in he anno a ion wo k low, 643
ea u es we e di ec ly disca ded bu ype II e o s we e iden i ied since
21 ea u es we e e oneously emo ed. In he case o he inclusion lis ,
al hough he anno a ion p ocess was a bi mo e ime-consuming, no ype
II e o s we e commi ed. Mo eo e , i ensu ed a oiding ype I e o s
since in he absence o he inclusion lis o endogens, 67 p obable
endogens would be anno a ed as xenobio ics.
CRediT au ho ship con ibu ion s a emen
Mikel Musa adi: Fo mal analysis, In es iga ion, Me hodology,
So wa e, Valida ion, Visualiza ion, W i ing – o iginal d a . In´
es
Bacie o-He n´
andez: Fo mal analysis, In es iga ion, Me hodology,
So wa e, Valida ion, W i ing – e iew & edi ing. Aile e P ie o:
Concep ualiza ion, Resou ces, W i ing – e iew & edi ing. Mai ane
Oli a es: Concep ualiza ion, Da a cu a ion, W i ing – e iew & edi ing.
Nes o E xeba ia: Funding acquisi ion, P ojec adminis a ion, Re-
sou ces. Ola z Zuloaga: Concep ualiza ion, Funding acquisi ion, P ojec
adminis a ion, Supe ision, W i ing – e iew & edi ing.
Decla a ion o compe ing in e es
The au ho s decla e ha hey ha e no known compe ing inancial
in e es s o pe sonal ela ionships ha could ha e appea ed o in luence
he wo k epo ed in his pape .
Fig. 6. MS2 ma ch o possible candida es o he ea u e C
10
H
12
N
2
O
3
([M+H]
+
=177.10222 Da, RT =1.0 min) using mzCloud lib a y: (a) sa is ac o y (95.7 %)
ma ch o co inine and (b) unsa is ac o y (41.5 %) ma ch o se o onin.
M. Musa adi e al.
