Analytica Chimica Acta 1000 (2018) 163e171

Contents lists available at ScienceDirect

Analytica Chimica Acta

journal homepage: www.elsevier.com/locate/aca

The differences in matrix effect between supercritical fluid

chromatography and reversed phase liquid chromatography coupled
to ESI/MS
Alfred Svan a, *, Mikael Hedeland a, b, Torbjo
€ rn Arvidsson a, c, Curt E. Pettersson a
a
Division of Analytical Pharmaceutical Chemistry, Uppsala University, BMC, Box 574, SE-751 23 Uppsala, Sweden
b
National Veterinary Institute (SVA), Dept. of Chemistry, Environment and Feed Hygiene, SE-751 89 Uppsala, Sweden
c
Medical Products Agency, Box 26, SE-751 03 Uppsala, Sweden

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Matrix effects were compared using

screening methods with SFC/ESI-MS
and RPLC/ESI-MS.

Blood plasma, horse urine and
influent/effluent wastewater were
investigated.

Through the use of post-column in-
fusions, matrix effect profiles were
generated.

Quantitative and qualitative infor-
mation was compared, interferences
tentatively identified.

Ion suppressions were generally
more common for SFC, and en-
hancements for LC.

a r t i c l e i n f o a b s t r a c t

Article history: For many sample matrices, matrix effects are a troublesome phenomenon using the electrospray ioni-
Received 8 August 2017 zation source. The increasing use of supercritical fluid chromatography with CO2 in combination with the
Received in revised form electrospray ionization source for MS detection is therefore raising questions: is the matrix effect
3 October 2017
behaving differently using SFC in comparison with reversed phase LC? This was investigated using urine,
Accepted 18 October 2017
Available online 23 October 2017
plasma, influent- and effluent-wastewater as sample matrices. The matrix effect was evaluated using the
post-extraction addition method and through post-column infusions. Matrix effect profiles generated
from the post-column infusions in combination with time of flight-MS detection provided the most
Keywords:
Matrix effects
valuable information for the study. The combination of both qualitative and semi-quantitative infor-
Supercritical fluid chromatography mation with the ability to use HRMS-data for identifying interfering compounds from the same exper-
Electrospray ionization iment was very useful, and has to the authors' knowledge not been used this way before. The results
Liquid chromatography showed that both LC and SFC are affected by matrix effects, however differently depending on sample
Ion enhancement matrix. Generally, both suppressions and enhancements were seen, with a higher amount of enhance-
Ion suppression ments for LC, where 65% of all compounds and all sample matrices were enhanced, compared to only 7%
for SFC. Several interferences were tentatively identified, with phospholipids, creatinine, and metal ion
clusters as examples of important interferences, with different impact depending on chromatographic

Abbreviations: E-WW, effluent wastewater; I-WW, influent wastewater; PCI,

post-column infusion; PCIMP, post-column infusion matrix profiles; PEA, post-
extraction addition.
* Corresponding author.
E-mail address: alfred.svan@farmkemi.uu.se (A. Svan).

https://doi.org/10.1016/j.aca.2017.10.014
0003-2670/© 2017 Elsevier B.V. All rights reserved.
164 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171
technique. SFC needs a different strategy for limiting matrix interferences, owing to its almost reverse
retention order compared to RPLC.
1. Introduction
Matrix effects (ME) constitute a well-known phenomenon,
mainly when using electrospray ionization (ESI) connecting liquid
chromatography (LC) to mass spectrometry (MS). ME has been
called the Achilles heel of LC/ESI-MS, and is generally described as
an alteration of the ionization efficiency by co-eluting molecules
[1]. The term matrix effects includes both enhancement and sup-
pression of the signal and although the underlying mechanisms are
not fully understood, several theories exist [2]. These signal alter-
nations can lead to incorrect quantifications and dramatically
higher LODs, since e.g. a severe signal suppression might leave only
a few percent of the signal in comparison with that of a neat
standard [3]. The most common method to handle ME is through
normalization with isotopically labelled internal standards or
reduction of the matrix interferences through sample preparation,
however with an increase in cost and time consumption. Despite
the use of matrix matched calibrators a quantification might be
problematic without a suitable internal standard, due to the relative
matrix effect: i.e., the signal difference between samples containing
the same matrix but of different origin, e.g. blood plasma from
different individuals [4,5]. An alternation of the chromatographic
method to example avoid co-elution of interferences with the an-
alyte could also be an alternative. In multiresidue methods com-
mon in e.g. environmental analysis, this might however be
problematic due to the high number of compounds, which also may
limit the amount of internal standards.
Several sample materials analyzed in e.g. bioanalytical and
environmental analysis are known to contain components which
often give rise to ME, such as blood plasma, urine, wastewater
(WW) and extracts from food, soil and tissues [2,6]. Even if the ME
may be decreased with sample preparation, such as extraction or
sometimes simply by dilution, its impact on the analytical results
has to be tested. This is generally done by an estimation of the
matrix effect influence, as recommended when validating analyt-
ical methods according to e.g. EMA or FDA guidelines [7,8].
Generally two different methods are used for ME determination,
where quantification of ME through post-extraction addition (PEA)
is the most common, providing the ME% for a specific compound at
its retention time [4]. The second method, with post-column
infusion (PCI) of the analyte to determine the effect on signal in-
tensity during the whole chromatographic run, is mainly qualita-
tive [9]. A combination of these two methods, forming the so-called
matrix effect profile, has also been described using PCI experiments
with standard and matrix injections to determine the ME% for each
data point in the chromatogram [10].
The new generation of SFC instruments with wider application
possibilities through the use of CO2 and polar organic solvents and
additives has created a new interest for this technique [11]. The
electrospray ion source is generally used to combine SFC with MS
[12], which also is the interface where ME has its highest impact.
Few studies have however been published discussing the matrix
effects using SFC/ESI-MS. Most applications have only included ME
as a validation parameter, evaluated through an estimation of the
ME using PEA [13e15] or PCI [16,17]. To best of the authors'
knowledge, two studies comparing ME in SFC and LC-MS have been
presented [18,19], developed for doping control of urine samples. In
© 2017 Elsevier B.V. All rights reserved.
both studies ME was quantified using SRM with a tandem quad-
rupole mass spectrometer, finding generally lower MEs for SFC-MS
than for LC-MS.
The aim of this study was to investigate how the matrix effects
in ESI/MS differ between SFC and reversed phase LC, using drug
substances, four complex sample matrices and commonly used
screening conditions for both techniques.
2. Material and methods
Since the aim was not to develop and validate methods for
screening and/or compound quantification, the conditions used
were gathered from the literature. For LC, a short UHPLC C18-
column was chosen, often used for gradient elution in modern
screening methods, with eluents of water and methanol containing
0.1% (v/v) formic acid as primary mobile phase. For SFC, a 2-picolyl-
amine column was chosen, using CO2 and an 8e45% gradient of the
modifier containing methanol and ammonium formate.
The set of drug compounds chosen for this study was selected to
include different chemical properties and to ensure a spread in
retention times over the chromatographic time scale. Some sub-
stances were also included owing to previously observed severe
matrix effects in one of the study matrices, e.g. gemfibrozil [20],
acetazolamide and miconazole [18].
2.1. Chemicals and reagents
Ammonia solution (4.0 M in methanol), ammonium acetate
(Bio-extra 98%), ammonium formate (Fluka, LC-MS, 99%), formic
acid (Fluka, LC-MS grade), methanol (Fluka Chromasolv, LC-MS
grade), sodium chloride (99.5), acetazolamide (>99%), amiloride
hydrochloride-hydrate (98%), atenolol (98%), carbamazepine
(>99%), diclofenac sodium (>99), enalapril maleate (98%), fluox-
etine hydrochloride (99.9%, Riedel-de Hae €n), gemfibrozil
(98.5%), hydrochlorthiazide (99%), mefenamic acid (98%),
metoprolol tartrate (99%), miconazole (99%), propranolol hy-
drochloride (99%), were purchased from Sigma-Aldrich (St. Louis,
MO, USA). Acetonitrile (LC-MS grade) was purchased from Fischer
scientific (Pittsburgh, PA, USA). Water was obtained from a Milli-Q
Q-POD-system from Millipore (Billerica, MA, USA). Carbon dioxide
(purity 99.99%) was obtained from Air Liquide (Paris, France).
2.2. Sample preparation
2.2.1. Plasma samples
Pooled human plasma with Na-EDTA (3H Biomedical, Uppsala,
Sweden) kept at 80  C during storage, was thawed in room
temperature. The proteins were precipitated by mixing 500 mL
plasma with 1000 mL ice cold acetonitrile for 15 s, and centrifuga-
tion for 5 min (12 100 g). A portion of the supernatant (1200 mL)
was removed and evaporated at 40  C under a N2 gas stream. The
samples used for SFC analysis were reconstituted in 500 mL aceto-
nitrile:water 75:25 (v/v) and the LC samples in 500 mL water with
0.1% formic acid (v/v) and transferred to vials.
2.2.2. Urine samples
Surplus blank horse urine was obtained from National
 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171
Veterinary Institute (SVA, Uppsala, Sweden) and stored in a freezer
(26  C) prior to analysis. After thawing in water bath, 100 mL were
transferred to Eppendorf tubes and diluted 1:9 (v/v) with the in-
jection medium 900 mL acetonitrile:water 75:25 (v/v) for SFC and
900 mL water þ 0.1% formic acid (v/v) for LC. The tubes were mixed
using vortex-mixer and centrifuged 10 min (12 100 g). The super-
natants (300 mL) were transferred to vials for analysis.
2.2.3. Influent and effluent wastewater samples
The influent (raw, I-WW) and effluent (treated, E-WW) waste-
water samples were collected as grab-samples in amber glass
bottles at Kungsa€ngsverket wastewater treatment plant (Uppsala,
Sweden). The water was filtered (0.7 mm glass fiber filter, Millipore)
and pH-adjusted to pH 4 using formic acid. The SPE was performed
using Oasis® HLB columns (6 cm3 200 mg, Waters Corporation,
Milford, MA, USA) and a general protocol, using 300 mL of effluent
or 100 mL influent WW. After sample application and 5 min drying
with vacuum, 5.0 mL water þ5% (v/v) methanol was used as wash,
prior to a second drying (10 min) and the elution using 8.0 mL
methanol. The methanol was evaporated at 40  C and under a
stream of N2 gas. When dry, the samples were reconstituted as the
plasma samples (see above).
All samples were prepared in triplicate and analytical blanks
were made for each matrix and chromatographic method, con-
sisting of a corresponding volume of pure water treated as the
samples. When performing the matrix effect semi-quantification
using the PEA-method, a compound mixture containing 500 nM
of the study compounds were prepared in water þ0.1% formic acid
for LC and acetonitrile:water 75:25 (v/v) for SFC. The mixture was
used as reference and as reconstitution solvent for the WW and
plasma samples and as dilution solvent for the urine samples, when
semi-quantifying ME. To prepare the urine sample reference, 100 mL
purified water was added instead of urine.
2.3. Instrumentation and analysis
The LC/ESI-Q-ToF MS system consisted of an I-class Acquity
UPLC system connected to a Synapt G2-S Q-ToF mass spectrometer,
from Waters. A Kinetex C18 1.3 mm, 50  2.1 mm column (Phe-
nomenex, Torrance, CA, USA) with a guard column with the same
stationary phase was used and kept at 50  C. A 10 min gradient with
A: 0.1% (v/v) formic acid in water and B: 0.1% (v/v) formic acid in
methanol was used, at a flow rate of 0.3 mL min1. The gradient
consisted of following steps: 4% B in 1 min, 1e3 min to 32% B lin-
early, holding 32% B in 1 min, 4e8 min to 85% B with curvature 8,
8e9 min holding 85% B, and equilibrating 9e10 min with 4% B, the
starting condition. The injection volume was 7.5 mL. As sample
manager wash, 90:10 methanol:water was used for 10 s pre- and
12 s post-injection. Water:methanol 90:10 was used as purge-
solvent.
Two mobile phases were used for both techniques, a primary for
all measurements and one alternative using the same gradient, for
validating that the observed effect was not dependent on the mo-
bile phase. The alternative mobile phase used for LC consisted of A:
5.0 mM ammonium acetate adjusted with ammonia to pH 8.5, and
B: methanol.
The LC/ESI-QqQ MS system consisted of an Acquity UPLC con-
nected to a Quattro Micro tandem quadrupole mass spectrometer
(Waters). The same settings and conditions were used as for the LC
connected to the Q-ToF system, with the exception of the sample
manager: the original Acquity UPLC system used a sample loop
instead of a “flow through needle”-system, as the I-class system.
The same SFC instrument (Acquity UPC2, Waters) was used for
both mass spectrometers. An external column block heater (Jones
Chromatography, Wales, UK) was used to hold the Torus 2-PIC (2-
165
picolyl-amine) 1.7 mm, 100  3.0 mm column (Waters) at 50  C. A
515 HPLC-pump (Waters) added a post column flow of methanol at
a flowrate of 0.45 mL min1 through a T-connection, prior a second
T-connection splitting the flow between the back-pressure regu-
lator and the MS (for dimensions and schedule, see Fig. S-1). The
mobile phase consisted of CO2 and a modifier gradient: methanol
with 20 mM ammonium formate and 2% (v/v) water, made from a
stock solution of 1.0 M ammonium formate in water. The alterna-
tive modifier contained methanol and 20 mM NH3. The 10 min
gradient started with 0e6 min from 8% modifier to 45% linearly,
hold at 45% between 6 and 8.5 min, before equilibration at 8%
modifier between 8.5 and 10 min. The flow rate was 1 mL min1,
the back-pressure 150 bar, and the injection volume 4 mL. As weak
wash, 1200 mL 2-propanol was used, with 800 mL methanol as
strong wash.
The MS detection was performed using the same detectors and
settings for both SFC and LC.
The Q-ToF/ESI MS system was used in both positive and negative
ionization mode, scanning from m/z 50 to 1200 using MSE. Using
this scan-mode, the scans were altered between the use of low
collision energy and every second scan with high collision energy
(20 eV in the transfer-cell together with a ramp of 20e40 eV in the
trap-cell) for the Q-ToF. Scan time was set to 0.2 s, data were
collected in resolution mode, and stored as centroid data. In ESIþ,
the capillary voltage was set to 0.8 kV, the cone voltage to 40 V, the
source temperature to 120  C, and the nitrogen desolvation gas
flow to 800 L h1 with a temperature of 650  C. In ESI-, a capillary
voltage of 2.0 kV was used, with remaining settings as for ESIþ.
The QqQ/ESI MS system was run in the selected reaction
monitoring (SRM) mode using one precursor and one product ion
for each compound (Tables Se1). In ESIþ, the capillary voltage was
3.0 kV, source temperature 120  C, desolvation gas flow 900 L h1
with a temperature of 450  C. In ESI-, a capillary voltage of 2.5 kV
was used, with remaining settings as for ESIþ. When post-column
infusions were performed using QqQ/ESI MS, one single compound
was monitored per analysis, using a dwell time of 0.5 s. Both MS-
instruments were equipped with ion sources using Z-spray
geometry.
Instrument control and data evaluation for all instruments were
performed using the Mass-Lynx software (version 4.1, from
Waters).
2.4. Matrix effect determination
The matrix effects (ME) were investigated using three methods.
With the QqQ-system, the post extraction addition (PEA) method
was applied, semi-quantifying ME by comparing the response from
injected compound mixtures dissolved in pure solvent with the
same mix dissolved in sample matrix, added after sample pre-
treatment. ME% was calculated in accordance with Matuszewski
et al. [4]: ME ¼ B/A 100, where B corresponds to the peak area of a
compound dissolved in a sample matrix, and A the corresponding
area from the compound in neat solvent.
The tandem quadrupole and ToF MS were both used for post-
column infusion (PCI) experiments as described by Bonfiglio et al.
[9]: for LC, a 600 nM in H2O þ 0.1% formic acid mixture containing
three of the study compounds was infused using a syringe pump at
5 mL min1, into the LC flow through a T-coupling, prior entrance to
the ESI probe. The chosen concentration was tested using calibra-
tion curves, to ensure linear response (Fig. S-2). For SFC, the com-
pound mixture was instead added to the post column make-up
solution (6.7 nM). For both techniques, the signal intensity of the
infused compounds were then monitored during injections of
sample matrix and neat solvent. In addition to the generated
infusion profiles, the signals from the PCI experiments were
166 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171
processed as first described by Stahnke et al. [10] to generate the
post column infusion matrix profiles (PCIMP): Initially the signal
was smoothed by using the mean intensity of five scans for every
scan (SI(M5)). Thereafter, the means of each scan were calculated
from the triplicates of the blank matrix injections and divided by
the mean of the corresponding scans from the analytical blank (see
Eq. (1)). The analytical blank was analyzed immediately prior to and
after the samples in the sample list. By using the PCIMP, the
approximated ME% can be illustrated for every scan in the chro-
matogram, and is therefore a hybrid of the traditional PCI and PEA
methods to measure ME [10].
 
SIM5 ðsample extractÞ
MESI ð%Þ ¼  100  100
SIM5 ðanalytical blankÞ
When using ToF and PCI data, the signal of a compound was
measured as the exact m/z with a window of 0.05 Da. In a few cases,
the signal was obviously interfered by isobaric compounds, and the
m/z window was then narrowed to 0.02 Da.
To identify background ions suspected to be involved or
responsible for ME, their retention time were compared with the
retention time window affected of ME. If matching, their accurate
m/z and product ions from the high energy function were used and
compared with literature and freely available databases (MassBank,
Human Metabolome Database, and METLIN). The identification was
therefore only tentative, since their structures were generally not
confirmed using standards.
3. Results and discussion
The sample preparation and chromatographic methods were
chosen to be representative for general screening methods of drugs
in complex matrices often causing matrix effects. For LC, the
method chosen was containing column and mobile phases
commonly used in screening methods [21e23], with a gradient
intended to spread out the compounds over the chromatographic
time scale (Fig. 1). The injection solvent was chosen to be less
eluting than the mobile phase, and similar to the starting compo-
sition of the LC gradient.
For SFC, the situation is more complicated, since the number of
published screening methods using MS-detection is so far only a
fraction of the number of LC-methods available. There is yet no such
thing as a standard column, and method development for SFC fo-
cuses to a higher degree on column selection compared to LC [24].
The chemistry of the chosen 2-picolyl-amine (2-PIC) column is very
similar to that of the established 2-ethylpyridine-column. However,
the 2-PIC provides better peak shapes [25], and it has been evalu-
ated using pharmaceutical-like compounds for screening [26,27].
The injection solvent (75:25 acetonitrile:water) was chosen to in-
crease the solubility of polar compounds, but still maintain an
acceptable peak shape [18]. A single column was chosen for SFC,
although the common practice in SFC method development is to
scan several columns. The reason to this was to keep focus on the
composition reaching the ESI source, and limit the scope to this
aspect, as column selectivity in SFC previously been investigated
[24]. The choice of modifier and additive (20 mM ammonium
formate with 2% water in methanol (v/v)) was based on recently
published screening methods using SFC-MS [26,28,29]. The SFC
method spread out the compounds in almost reverse order
compared to LC (Fig. 1). The alternative mobile phases used for both
LC and SFC were generally showing the same results as the primary,
with differences in retention time. Therefore, results from the
alternative mobile phases are only mentioned when significant
differences in ME between the mobile phases were observed.
Both ESIþ and ESI- were used in the experiments, although
positive ionization provided a higher overall signal intensity and
was more sensitive to ME [2]. The following results and discussion
are mainly based on ESIþ, as the signal intensity achieved for ESI-
when affected by the sample matrices were too low to enable ac-
curate measurements.
3.1. Plasma
The semi-quantification of ME deriving from PEA-experiments
using plasma LC/ESIþ QqQ MS/MS showed a strong enhancement
(>50%) for 2 compounds, slight enhancement (15e50%) for 4
compounds, whereas the rest were unaffected by ME (Table 1). The
signal of atenolol was enhanced by 30% (Fig. S-6), in accordance
(1)
with the PCIMP using ToF MS, where an enhancement was
observed at the retention time for atenolol (Fig. 1). Fig. 1 also shows
that the different infused compounds are affected similarly by the
interferences, although compound related differences exist, one
example being the aforementioned enhancement for atenolol be-
tween 1 and 2.5 min which is not seen for the other compounds.
This demonstrates that surrogate internal standards in multi-
residue methods have to be experimentally evaluated, otherwise
poor standardization and large errors might be the result.
The semi-quantification of ME from plasma using SFC gave
generally small effects, with 4 compounds showing slight sup-
pression (-15 - -50%) and one compound (amiloride) which was
highly suppressed (<-50%). The overall semi-quantification results
for plasma were very much in line with the summary of all matrices
(Table 1), with generally more suppression for SFC and enhance-
ment for LC.
The PCIMP generated from SFC/ESIþ ToF MS (Fig. 1) confirmed a
suppression at the retention time of atenolol, and also showed
several other areas with severe ion suppression. The strongly
suppressed amiloride with only 3% signal in the PEA-experiment
(Fig. S-6), has generally a similar ME% profile as the other com-
pounds in the post-infusion experiments. Its retention time
(6.04 min) is however in an area where all measured compounds
have close to 100% suppression (5.75e6.10 min).
Fig. 2 summarizes the proposed major signal suppression
sources from plasma samples, tentatively identified using online
database and literature searches of accurate m/z values for pro-
tonated compounds and product ions. Although ToF MS has pre-
viously been used to create matrix effect profiles [30], this
integrated way of studying ME and identifying interferences has, to
author's knowledge, not been described before and provides a
useful tool for ME studies.
For LC (Fig. 2) high levels of polyethylene glycol (PEG) is
responsible for most of the suppression, due to the ability of the LC
method to efficiently spread out these polymers over the chro-
matographic time scale. PEG is most likely a contamination origi-
nating from the plastic plasma collection tubes [31]. The
anticoagulant used in the plasma (Na-EDTA), together with endo-
genic polar compounds created a major suppression zone at t0,
however it did not affect the first eluting compound acetazolamide
at 1.60 min. The hydrophobic phospholipids eluted together at the
highest % of methanol in the gradient, and were therefore affecting
gemfibrozil and mefenamic acid, although all hydrophobic com-
pounds eluting late would suffer from suppression according to the
ME% profiles (Fig. 1). Even though the drug compounds and sample
matrices analyzed are the same for LC and SFC, the presumed cause
of ME was rather different (Fig. 2). The PEGs, causing a scattered ion
suppression zone in LC, were not well separated using the SFC
method and therefore only caused ion suppression in a single area
of the chromatogram. The corresponding but reversed effect was
observed for the phospholipids in SFC, eluting together in LC, but
scattered in SFC (Fig. 2). The highest level of suppression, co-eluting
 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171 167

Fig. 1. The PCIMP from injections of pre-treated blank plasma samples, using ToF/ESI-MS with LC and SFC. The infusion profiles show ME in % after the injection of pre-treated blank
plasma, with ME ¼ 0 indicating no effect. The dotted lines illustrate the gradients used (B%). The chromatographic peaks are added from injections of standard mixtures, arbitrarily
set to 100% intensity.

Table 1 in comparison with SFC, where suppression was more common.

Summary of the PEA semi-quantification of ME for all four matrices, using QqQ MS. This was in consistency with our study, as seen in the summary of
Enhancement No ME Suppression all matrices (Table 1 and Fig. S-5).
Strong Slight Slight Strong
In LC, the PCIMP shows high suppression in conjunction with
the elution of the void, as might be expected from urine: polar
LC: Plasma 2 4 5 0 0
compounds as e.g. uric acid and carnitine eluted here (Fig. 3). A
LC: Urine 8 1 0 2 0
LC: I-WW 6 0 1 2 2 strong enhancement is seen for the signal of enalapril between 1
LC: E-WW 8 0 1 2 0 and 2 min in the PCIMP using the acidic mobile phase which is an
LC: Summary 24 5 7 6 2
exception among the study compounds, confirming that ME in
several cases is compound specific. This enhancement is not seen
SFC: Plasma 0 0 6 4 1
using the alternative alkali mobile phase, which generally created
SFC: Urine 1 1 3 2 4
SFC: I-WW 0 0 1 4 6 more suppression than the acidic mobile phase for urine samples
SFC: E-WW 1 0 1 6 3 using LC (Fig. 4).
SFC: Summary 2 1 11 16 14
For SFC, the majority of the interferences from urine eluted later
in comparison with those from the other matrices, reflecting the
n ¼ number of compounds, of totally 11. Strong enhancement: >50%, slight
hydrophilic character of the urinary constituents (Fig. 3). Fewer
enhancement: 15e50%, no ME: ±15%, slight suppression: 15- -50% and strong
suppression: <-50%. interferences could be tentatively identified in urine than in
plasma, but the alkali-metal clusters detected in plasma were
creating severe suppression also in the urine analysis, now joined
with e.g. amiloride was however caused by two clusters of different by an Mg2þ cluster. The metal ion content in urine is naturally
adducts involving sodium and potassium ions, naturally abundant fluctuating depending on homeostasis, and will therefore
in blood plasma. These highly suppressing clusters were unique for contribute to the sample-to-sample variation of this matrix.
the SFC experiments, but not for plasma. Urine samples gave rise to considerably more carry-over con-
In summary, both techniques suffered from ME, SFC mainly from taminants (comparing total ion chromatogram from PCI experi-
ion suppression and LC from both enhancement and suppression. ments) in SFC in comparison with other matrices and LC. This is
The interfering compounds causing these effects were however most likely also a reflection of the polar content of urine,
different and related to the chromatographic selectivity of the demanding a more intense autosampler washing procedure for the
techniques. normal phase-like SFC than for reversed phase LC.
Overall, the polar nature of urine created late-eluting in-
3.2. Urine terferences for SFC in comparison with LC. The suppression
observed in SFC was often severe, especially at the elution times of
Urine as a sample matrix is varying depending on source, when the different metal ion clusters (Fig. 3). SFC-MS analyses of urine
collected and what the subject had been eating and drinking. Due would therefore strongly benefit from a sample extraction prior
to the polar content of urine, LC-MS is the current technique of analysis, to remove salts and other very polar components.
choice for e.g. doping screening of urine samples. SFC has histori-
cally been used for more hydrophobic compounds, but was recently
evaluated and compared with LC for the analysis of doping com- 3.3. Effluent and influent wastewater
pounds in human urine [18]. In the study, enhancement was
observed to a higher degree for the urine samples analyzed with LC The PEA method for ME investigations is biased with
168 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171

Fig. 2. The blue solid line is the PCIMP for atenolol (extracted m/z 267.1703) in plasma, from experiments using LC/ESIþ ToF MS, and the green line shows corresponding data for
SFC/ESIþ ToF MS. ME% ¼ 0 means no effect on signal. Through investigation of background-ions, the origins of some of the major suppression zones are propose. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3. The solid lines are the PCIMP for atenolol, amiloride and enalapril during an injection of diluted urine, using ToF/ESI-MS with LC and SFC. ME% ¼ 0 means no effect on signal.
Through investigation of background-ions, the cause of some areas with major suppression is proposed.

enhancements in both effluent and influent WW, due to the
occurrence of several compounds used in the society served by the
WW-treatment plant (Figs. S3eS4). The occurrence of analytes in
the WW does however not substantially affect the PCI-
experiments, more than through the peaks observed at the ex-
pected retention time of the monitored compound.
Although both the influent and effluent WW samples are
extracted using SPE, the high sample volumes (100 and 300 mL for
I-WW and E-WW respectively) and the complexity of the matrices
forms sample extracts with severe matrix effects. This gives a
generally higher signal intensity with more abundant peaks
compared to the other matrices, despite the SPE (data not shown).
Some differences could be noted between the matrices, e.g. the
total ion chromatogram (TIC) of the PCI using ToF MS shows that
the untreated wastewater (I-WW) contains more hydrophobic
compounds (early eluting in SFC and late in LC) than the E-WW. The
high hydrophobic content is reflected in the PCIMP for SFC, where I-
WW gives more ion suppression during the first minutes of the
chromatogram in comparison with E-WW (Fig. 5). The same
pattern is seen for LC (Fig. 5), with an overall suppressed signal for
I-WW, and slightly less for E-WW, with the largest difference in the
end of the chromatogram. Some compounds responsible for ion
suppression in urine and plasma (e.g. carnitine) were detected at
low intensities in I-WW, however not correlated to a matching loss
in sensitivity. The overall high suppression affecting the signal in
the I-WW samples was instead caused by other, unidentified,
 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171
Fig. 4. PCIMP for enalapril (extracted m/z 377.2071) in urine samples, using LC/
ESIþ ToF. Comparison between the two different mobile phases, the original (blue line)
containing A: H2Oþ0.1% FA B: MeOHþ0.1% FA and the alternative phase (purple): A:
5.0 mM AcNH4 pH 8.5 B: MeOH. This is an exception: both the enhancement seen for
enalapril in the original mobile phase between 1 and 2.5 min, and the difference be-
tween the mobile phases, that normally showed similar ME, with variations in
retention. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
interferences.
The chromatographic orthogonality of reversed phase LC and
SFC techniques is also creating a similar but reversely directed ef-
fect in comparison with the polar urine: the hydrophobic I-WW
shows generally less suppression using SFC compared to LC. This is
probably due to the strong elution of hydrophobic compounds early
in the gradient, washing these interferences away, while these
hydrophobic contaminants are more difficult to remove using LC.
This was confirmed by comparison of the TIC received from the
solvent blanks analyzed prior to and immediately after the I-WW
samples. In SFC no carry-over was seen for I-WW, while high in-
tensity carry-over peaks were observed in the end of the LC-
gradient during several blank injections (data not shown).
To investigate potential instrumental differences between the
ToF MS and the QqQ MS/MS used in this study, a PCI experiment
was performed using QqQ with SRM detection for the I-WW
samples. The SFC analysis gave a high degree of similarity of the
PCIMP between the QqQ and the ToF instruments, however the LC
runs did not (Fig. S-7). Instead of the overall suppression received
analyzing I-WW with ToF MS (Fig. 5), an enhancement was noted in
parts of the chromatogram for QqQ MS. The SRM matrix effect
profile for I-WW had similarities to the profiles received from the
urine and plasma-experiments with LC-ToF MS (Figs. 2 and 3)
showing areas of enhancement around 1e2.5 and 5e7 min. An
additional experiment using the quadrupole integrated in the ToF-
MS instrument to monitor a single nominal precursor m/z at a time
was performed, ensuring that the difference seen did not originate
from the different detection (SRM for QqQ MS or MS-scan using ToF
MS) of the ions, but from the ionization source. Older generation
ESI-sources like the source on the QqQ MS has previously proven to
benefit more from a higher organic solvent content in the mobile
phase than newer generation ESI sources, as the one on the ToF MS
system [32]. MS instruments with an older generation ESI has also
showed a generally higher gain in S/N-ratio using the highly vola-
tile CO2 in SFC than the water based reversed phase-LC, a difference
not seen using a newer ESI source [18,28]. Except from direct
interference with the ion formation in the liquid phase, it is
believed that interfering compounds might affect the droplet for-
mation in ESI [33], possibly through modifications of boiling point,
viscosity and surface tension. The high content of organic com-
pounds in the I-WW could hypothetically have a beneficial effect on
the ionization process, by e.g. lowering the surface tension. This
would explain why the enhancement is only seen for LC, as the
mobile phase in SFC is already highly volatile due to the CO2. It
169
Fig. 5. A comparison of PCIMP generated from influent and effluent WW, using ami-
loride (extracted m/z 230.0557) and LC/ESIþ ToF MS (dark and light blue) and SFC/
ESIþ ToF MS (dark and light green). ME% ¼ 0 means no effect on signal. The retention
time of amiloride from an injection of a standard is marked by the red dashed line. (For
interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)
would also explain why the enhancement mainly is observed early
in the LC chromatogram: the methanol content is lower here and
organic compounds from the sample matrix will therefore have a
higher impact here. Finally, it would also explain why the
enhancement is seen more often with the QqQ MS equipped with
the older generation ESI source than for the newer ESI source on the
ToF mass spectrometer (Fig. S-7). This theory, if valid, might also
add important information on the mechanism behind
enhancement.
In summary, both techniques suffered from ion suppression
when WW-samples were analyzed. The hydrophobic content in
especially I-WW could however provide some benefits for SFC over
LC, due to the ability of SFC to remove these interferences more
efficiently, providing less suppression in the end of chromatogram
and less memory-effects. The difference noted between the QqQ-
and ToF-MS when performing PCI-experiments is most likely
related to the different generations of the ESI sources.
4. Conclusions
When comparing the matrix effects from SFC and LC in combi-
nation with ESI-MS, the PCIMP using ToF-MS and full-scan acqui-
sition was found very useful. Through this technique, both
qualitative and semi-quantitative data could be extracted, with the
ability to identify the interfering compounds from the same data.
This made it possible to not only see differences between the
chromatographic techniques, but also to help understanding what
caused them.
In our study, we could see that a sample matrix containing
mainly polar interferences, such as urine, will mainly affect the
eluting compounds in the first part of the chromatogram in
reversed phase LC, while the first minutes were relatively unaf-
fected for SFC. The chromatographic selectivity for interferences
will have a major impact on the matrix effect. For example, in
samples containing PEG contamination (e.g. plasma) the LC method
170 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171
was able to separate different PEGs, causing suppression over a
wider time scale in comparison with SFC, where all PEGs eluted as
one peak. The use of a different SFC column will however alter the
selectivity, and thereby also the retention time for the areas
affected by ME. The overall tendency of enhancement in LC and
suppression in SFC, also observed in a previous study [18], could
also be confirmed. Hypothetically, this enhancement in LC/ESI-MS
can be connected to an improved droplet formation, as this is
probably not improved by organic compounds using SFC where no
enhancement was seen. Although, this needs a more targeted
investigation to be confirmed.
Using an unselective sample preparation, there is probably no
quick-fix for matrix effects when analyzing compounds with
varying chemical properties using the ESI source. For compounds
with more similar properties, there might sometimes be some
tricks. For example, when analyzing relatively hydrophobic com-
pounds found in a polar matrix as urine, SFC would probably pro-
vide both a short runtime and small impact from matrix. This is as
long as the compounds are eluting prior to the interferences, and
the interferences are successfully washed away between analyses.
SFC in combination with MS is still something relatively new in
comparison to LC-MS, and future applications of SFC-MS will
contain both advantages compared to existing techniques, and
difficulties, sometimes unexpected. The application to more polar
analytes from water-based samples might for example bring to
some unexpected effects, as the metal cluster formations observed
in this study when analyzing plasma and urine samples, created
heavily suppressed areas in the chromatogram.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.aca.2017.10.014.
References
[1] P.J. Taylor, Matrix effects: the Achilles heel of quantitative high-performance
liquid chromatographyeelectrosprayetandem mass spectrometry, Clin. Bio-
chem. 38 (2005) 328e334, https://doi.org/10.1016/j.clinbiochem.2004.11.007.
[2] A. Furey, M. Moriarty, V. Bane, B. Kinsella, M. Lehane, Ion suppression; a
critical review on causes, evaluation, prevention and applications, Talanta 115
(2013) 104e122, https://doi.org/10.1016/j.talanta.2013.03.048.
[3] J.-P. Antignac, K. de Wasch, F. Monteau, H. De Brabander, F. Andre, B. Le Bizec,
The ion suppression phenomenon in liquid chromatographyemass spec-
trometry and its consequences in the field of residue analysis, Anal. Chim. Acta
529 (2005) 129e136, https://doi.org/10.1016/j.aca.2004.08.055.
[4] B.K. Matuszewski, M.L. Constanzer, C.M. Chavez-Eng, Strategies for the
assessment of matrix effect in quantitative bioanalytical methods based on
HPLCMS/MS, Anal. Chem. 75 (2003) 3019e3030, https://doi.org/10.1021/
ac020361s.
[5] A. Van Eeckhaut, K. Lanckmans, S. Sarre, I. Smolders, Y. Michotte, Validation of
bioanalytical LC-MS/MS assays: evaluation of matrix effects, J. Chromatogr. B
Anal. Technol. Biomed. Life Sci. 877 (2009) 2198e2207, https://doi.org/
10.1016/j.jchromb.2009.01.003.
[6] H. Trufelli, P. Palma, G. Famiglini, A. Cappiello, An overview of matrix effects in
liquid chromatography-mass spectrometry, Mass Spectrom. Rev. 30 (2011)
491e509, https://doi.org/10.1002/mas.20298.
[7] EMA, Guideline on bioanalytical method validation, EMEA, Comm. Med. Prod.
Hum. Use 44 (2012) 1e23 doi:EMEA/CHMP/EWP/192217/2009.
[8] Guidance for industry: bioanalytical method validation, FDA, 2001. http://
www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/
Guidances/UCM070107.pdf. Accessed 4 April 2017.
[9] R. Bonfiglio, R. King, T. Olah, K. Merkle, The effects of sample preparation
methods on the variability of the electrospray ionization response for model
drug compounds., Rapid Commun, Mass Spectrom. 13 (1999) 1175e1185,
https://doi.org/10.1002/(SICI)1097-0231(19990630)13:12<1175::AID-
RCM639>3.0.CO;2e0.
[10] H. Stahnke, T. Reemtsma, L. Alder, Compensation of matrix effects by post-
column infusion of a monitor substance in multiresidue analysis with LCMS/
MS, Anal. Chem. 81 (2009) 2185e2192, https://doi.org/10.1021/ac802362s.
[11] A. Tarafder, Metamorphosis of supercritical fluid chromatography to SFC: an
Overview, TrAC - Trends Anal. Chem. 81 (2016) 3e10, https://doi.org/10.1016/
j.trac.2016.01.002.
[12] D. Wolrab, P. Frühauf, C. Gerner, Direct coupling of supercritical fluid
chromatography with tandem mass spectrometry for the analysis of amino
acids and related compounds: comparing electrospray ionization and atmo-
spheric pressure chemical ionization, Anal. Chim. Acta 981 (2017) 106e115,
https://doi.org/10.1016/j.aca.2017.05.005.
[13] R.A. Coe, J.O. Rathe, J.W. Lee, Supercritical fluid chromatography-tandem mass
spectrometry for fast bioanalysis of R/S-warfarin in human plasma, J. Pharm.
Biomed. Anal. 42 (2006) 573e580, https://doi.org/10.1016/j.jpba.2006.05.025.
[14] Y. Tao, F. Dong, J. Xu, X. Liu, Y. Cheng, N. Liu, Z. Chen, Y. Zheng, Green and
sensitive supercritical fluid chromatographic-tandem mass spectrometric
method for the separation and determination of flutriafol enantiomers in
vegetables, fruits, and soil, J. Agric. Food Chem. 62 (2014) 11457e11464,
https://doi.org/10.1021/jf504324t.
[15] T. Berg, L. Kaur, A. Risnes, S.M. Havig, R. Karinen, Determination of a selection
of synthetic cannabinoids and metabolites in urine by UHPSFC-MS/MS and by
UHPLC-MS/MS, Drug Test. Anal. (2015), https://doi.org/10.1002/dta.1844 n/a-
n/a.
[16] A. Svan, M. Hedeland, T. Arvidsson, J.T. Jasper, D.L. Sedlak, C.E. Pettersson,
Rapid chiral separation of atenolol, metoprolol, propranolol and the zwitter-
ionic metoprolol acid using supercritical fluid chromatographyetandem mass
spectrometry e application to wetland microcosms, J. Chromatogr. A 1409
(2015) 251e258, https://doi.org/10.1016/j.chroma.2015.07.075.
[17] Y. Hsieh, F. Li, C.J.G. Duncan, Supercritical fluid chromatography and high-
performance liquid chromatography/tandem mass spectrometric methods
for the determination of cytarabine in mouse plasma, Anal. Chem. 79 (2007)
3856e3861, https://doi.org/10.1021/ac062441s.
[18] L. Novakova , M. Rentsch, A. Grand-Guillaume Perrenoud, R. Nicoli, M. Saugy,
J. Veuthey, D. Guillarme, Ultra high performance supercritical fluid chroma-
tography coupled with tandem mass spectrometry for screening of doping
agents. II: analysis of biological samples, Anal. Chim. Acta 853 (2015)
647e659, https://doi.org/10.1016/j.aca.2014.10.007.
[19] V. Desfontaine, L. Nova kova
, F. Ponzetto, R. Nicoli, M. Saugy, J.L. Veuthey,
D. Guillarme, Liquid chromatography and supercritical fluid chromatography
as alternative techniques to gas chromatography for the rapid screening of
anabolic agents in urine, J. Chromatogr. A 1451 (2016) 145e155, https://
doi.org/10.1016/j.chroma.2016.05.004.
[20] M.J.M. Bueno, A. Agüera, M.J. Go mez, M.D. Hernando, J.F. García-Reyes,
A.R. Fernandez-Alba, Application of liquid chromatography/quadrupole-linear
Ion trap mass spectrometry and time-of-flight mass spectrometry to the
determination of pharmaceuticals and related contaminants in wastewater,
Anal. Chem. 79 (2007) 9372e9384, https://doi.org/10.1021/ac0715672.
[21] H.H. Maurer, Current role of liquid chromatographyemass spectrometry in
clinical and forensic toxicology, Anal. Bioanal. Chem. 388 (2007) 1315e1325,
https://doi.org/10.1007/s00216-007-1248-5.
[22] R. Díaz, M. Ib ~ ez, J.V. Sancho, F. Herna
an ndez, Building an empirical mass
spectra library for screening of organic pollutants by ultra-high-pressure
liquid chromatography/hybrid quadrupole time-of-flight mass spectrometry,
Rapid Commun. Mass Spectrom. 25 (2011) 355e369, https://doi.org/10.1002/
rcm.4860.
[23] C. Bueschl, B. Kluger, M. Lemmens, G. Adam, G. Wiesenberger, V. Maschietto,
A. Marocco, J. Strauss, S. Bo € di, G.G. Thallinger, R. Krska, R. Schuhmacher,
A novel stable isotope labelling assisted workflow for improved untargeted
LC-HRMS based metabolomics research, Metabolomics 10 (2014) 754e769,
https://doi.org/10.1007/s11306-013-0611-0.
[24] E. Lesellier, C. West, The many faces of packed column supercritical fluid
chromatography e a critical review, J. Chromatogr. A 1382 (2015) 2e46,
https://doi.org/10.1016/j.chroma.2014.12.083.
[25] V. Desfontaine, J.-L. Veuthey, D. Guillarme, Evaluation of innovative stationary
phase ligand chemistries and analytical conditions for the analysis of basic
drugs by supercritical fluid chromatography, J. Chromatogr. A 1438 (2016)
244e253, https://doi.org/10.1016/j.chroma.2016.02.029.
[26] A. Dispas, P. Lebrun, P.-Y. Sacre , P. Hubert, Screening study of SFC critical
method parameters for the determination of pharmaceutical compounds,
J. Pharm. Biomed. Anal. 125 (2016) 339e354, https://doi.org/10.1016/
j.jpba.2016.04.005.
[27] E. Lemasson, S. Bertin, P. Hennig, H. Boiteux, E. Lesellier, C. West, Development
of an achiral supercritical fluid chromatography method with ultraviolet
absorbance and mass spectrometric detection for impurity profiling of drug
candidates. Part II. Selection of an orthogonal set of stationary phases,
J. Chromatogr. A 1408 (2015) 227e235, https://doi.org/10.1016/
j.chroma.2015.07.035.
[28] L. Novakov a, A. Grand-Guillaume Perrenoud, R. Nicoli, M. Saugy, J. Veuthey,
D. Guillarme, Ultra high performance supercritical fluid chromatography
coupled with tandem mass spectrometry for screening of doping agents. I:
investigation of mobile phase and MS conditions, Anal. Chim. Acta 853 (2015)
637e646, https://doi.org/10.1016/j.aca.2014.10.004.
[29] M. Ishibashi, Y. Izumi, M. Sakai, T. Ando, E. Fukusaki, T. Bamba, High-
throughput simultaneous analysis of pesticides by supercritical fluid chro-
matography coupled with high-resolution mass spectrometry, J. Agric. Food
Chem. 63 (2015) 4457e4463, https://doi.org/10.1021/jf5056248.
[30] O. Gonza lez, M. van Vliet, C.W.N. Damen, F.M. van der Kloet, R.J. Vreeken,
T. Hankemeier, Matrix effect compensation in small-molecule profiling for an
LCeTOF platform using multicomponent postcolumn infusion, Anal. Chem. 87
(2015) 5921e5929, https://doi.org/10.1021/ac504268y.
[31] R. Weaver, R.J. Riley, Identification and reduction of ion suppression effects on
pharmacokinetic parameters by polyethylene glycol 400, Rapid Commun.
 A. Svan et al. / Analytica Chimica Acta 1000 (2018) 163e171 171

Mass Spectrom. 20 (2006) 2559e2564, https://doi.org/10.1002/rcm.2629. https://doi.org/10.1016/j.chroma.2014.06.066.

[32] A. Periat, I. Kohler, A. Bugey, S. Bieri, F. Versace, C. Staub, D. Guillarme, Hy- [33] R. King, R. Bonfiglio, C. Fernandez-Metzler, C. Miller-Stein, T. Olah, Mecha-
drophilic interaction chromatography versus reversed phase liquid chroma- nistic investigation of ionization supression in electrospray ionization, J. Am.
tography coupled to mass spectrometry: effect of electrospray ionization Soc. Mass Spectrom. 11 (2000) 942e950, https://doi.org/10.1016/S1044-
source geometry on sensitivity, J. Chromatogr. A 1356 (2014) 211e220, 0305(00)00163-X.