chapter 5: development and optimization of a system for ... 5.pdflc × lc-uv/ms for the separation...
TRANSCRIPT
LC × LC-UV/MS for the separation of complex samples
109
Chapter 5: Development and optimization of a system
for comprehensive two-dimensional liquid
chromatography with UV and mass spectrometric
detection for the separation of complex samples by
multi-step gradient elution
Journal of Chromatography A, 1188 (2008) 216–226
Chapter 5
110
Abstract
Comprehensive two-dimensional liquid chromatography (LC × LC) is a powerful tool for
the separation of complex biological samples. This technique offers the advantage of
simplified automation and greater reproducibility in a shorter analysis time than off-line
two-dimensional separation systems. In the present study, an LC × LC system is
developed enabling simultaneous UV and MS detection, and which can be easily
converted to a conventional reversed-phase LC-UV/MS system. In LC × LC, a 60-min
reversed-phase LC separation with a linear solvent gradient in the first dimension is
coupled to a second-dimension separation on a mixed-mode cation-
exchange/reversed-phase column with a modulation time of 60 s. The isocratic
separation in the second-dimension column is optimized by the use of a multi-step
gradient where the organic and the ionic modifier are varied independently. Intraday (n
= 3) and interday (n = 4) variability of the retention times were evaluated with the
complete system and found to be 0.5% and 0.7%, respectively. Good linearity was
observed in calibration curves for three different compounds varying in polarity.
LC × LC-UV/MS for the separation of complex samples
111
Introduction
Comprehensive separation of complex mixtures, e.g., urine samples for metabolomics
or biomarker studies, is a difficult challenge due to the presence of thousands of
components that vary from polar to non-polar and from picomolar to millimolar
concentrations, and that show diverse physico-chemical properties, e.g., acid–base
properties, stability, solubility, detectability. The potential of conventional separation
techniques such as liquid chromatography (LC) or gas chromatography (GC) and
detection approaches like UV–vis or mass spectrometric (MS) detection are limited. In
the last decade, comprehensive multidimensional separation techniques such as GC ×
GC [1,2], LC × GC [3,4], LC × capillary electrophoresis (CE) [5,6], and LC × LC [7–
9,31] have been developed and reported. These multidimensional techniques offer an
enormous separation power and are therefore ideally suited for the analysis of very
complex mixtures. Due to its compatibility with biological matrixes, HPLC is most
widely applied in biological applications. Therefore, the use of comprehensive two-
dimensional liquid chromatography (LC × LC) would be a logical choice for the
separation of complex biological samples where conventional LC does not provide
sufficient separation power. In contrast to multidimensional heart-cutting separation
techniques (LC–LC), where a particular fraction of the first-dimension separation is
transferred and re-separated on the second-dimension column [10], in LC × LC the
entire first dimension is analyzed in the second-dimension separation. Comprehensive
LC × LC offers various advantages over both multidimensional off-line and
conventional separation techniques, especially with respect to enhanced peak capacity,
automation potential, and reproducibility and shorter analysis time. Given the
numerous liquid separation modes, there are numerous possible combinations of first-
and second-dimension separation in LC × LC.
In many of the recent papers on LC × LC, compounds are separated by two
independent separation techniques featuring the same separation mechanism,
especially reversed-phase LC × reversed-phase LC systems (RPLC × RPLC) [11–
15,31]. In these cases, the two-dimensional separation is based on the use of different
organic modifiers and/or RPLC columns with different properties. Nevertheless, due to
the fact that the retention mechanisms in the first and second dimension are largely
the same, components with a low retention factor in the first dimension will still have a
low capacity factor in the second dimension. As a result, the high theoretical peak
capacities can never be attained in such a setup.
Chapter 5
112
Therefore, the use of two different separation mechanisms in the first and second
dimension is generally preferred. A variety of LC × LC separation approaches has been
described, including RPLC × size-exclusion chromatography (SEC) [16–18], normal-
phase (NP) LC × RPLC [7,19], SEC × RPLC [20,21], ion-exchange chromatography
(IEC) × SEC [22], IEC × RPLC [23–25]. IEC × RPLC would be a logical column
combination for the separation of biological samples because of the presence of many
ionic components that can be separated by IEC. Most LC × LC systems based on IEC ×
RPLC are operated with IEC in the first dimension and RPLC in the second dimension.
Perhaps the most widely applied comprehensive LC × LC approach, i.e., the
multidimensional protein identification technology (MudPIT) developed by the group of
Yates [26] and two-column approaches based on this design [27,28], is based on IEC
× RPLC, but in general peptide separation can be considered to be significantly
different from the small-molecule separations as under development in the present
study. Nevertheless, IEC × RPLC would be attractive because RPLC on C18-materials is
the most frequently applied LC mode. RPLC operating conditions are readily compatible
with important LC detectors, UV, fluorescence and especially MS. Due to its selectivity
and sensitivity, and its ability to confirm compound identity or even elucidate unknown
structures, MS detection is nowadays essential in the majority of the LC applications.
From this point of view, the alternative approach of RPLC × IEC, although attractive for
its comprehensive two-dimensionality, would be less favorable, because IEC generally
requires relatively high salt concentrations and is therefore less compatible with MS
detection.
An interesting alternative would be the use of mixed-mode columns featuring both
RPLC and IEC separation mechanisms in the second dimension, as these can eliminate
some of the shortcomings of both RPLC and IEC.
In the present paper, we proposed a novel approach to overcome these limitations by
implementing a mixed-mode RPLC/cation-IEC column in the second dimension in
combination with step-gradient elution. Using a test-set of 32 compounds covering a
very wide range of polarities, we describe method development and optimization
separately for each separation dimension. Optimization of the second-dimension
separation indicated the necessity of the use of a step gradient. The fully automated LC
× LC system has been coupled to a mass spectrometer.
LC × LC-UV/MS for the separation of complex samples
113
Experimental
Chemicals and reagents
Alanine, creatinine, dopamine, caffeine, theophylline, paracetamol,
desmethyldiazepam, midazolam, aspirin, aniline, 2,6-dimethylaniline, ibuprofen,
vanillin, thiobenzamide, nadolol, hydrocortisone, naproxen, diazepam, oxazepam,
sulfamethoxypyridazine, sulfaguanidine, 3,4-hydroxybenzoic acid, sulfamic acid,
imipramine, tryptophan, tyrosine, nitrazepam, carbamazepine, benzothiophen,
metamitron, prometryn, and desethylatrazine were all purchased from Sigma,
Zwijndrecht, The Netherlands. Acetonitrile (ACN), formic acid (FA), acetic acid, and
trifluoroacetic acid (TFA) were purchased from Biosolve, Valkenswaard, The
Netherlands.
The test-set was diluted in 1% (v/v) of ACN in water with 0.1% (v/v) FA at
concentrations of 125 µM for all compounds. The urine sample was collected from a
volunteer; the sample was centrifuged at 13 200 rpm with an eppendorf centrifuge
(VWR International, Amsterdam, The Netherlands) directly after collection and stored
at −20 ◦C.
Instrumentation for LC × LC
A schematic diagram of the setup of the comprehensive LC × LC system is illustrated in
Figure 5.1. The system consists of two parts which can also be used independently as
two separate conventional LC systems.
The first-dimension separation system consisted of a Shimadzu HPLC system (‘s
Hertogenbosch, The Netherlands) comprising of two LC-10AD-VP pumps, a SCL-
10ADvp system controller, a DGU-14A degasser, a SIL-10ADvp autosampler, a CTO-
ACvp column oven. For UV detection, a Kratos absorbance detector spectroflow 757
(Nieuwerkerk ad IJssel, The Netherlands) and/or an Agilent Technologies HP1050
photodiode array detector (DAD) system (Amstelveen, The Netherlands) were used.
The RPLC column was a Luna C18(2) (50 mm × 1 mm I.D.; 3 µm particle size) with a
RP-18 (2 mm × 2 mm) pre-column, purchased from Bester (Amstelveen, The
Netherlands). Pumps, autosampler, column are connected using 0.125 mm I.D.
polyetheretherketone (PEEK) tubing. Post-column, the flow-rate is split in a 1:1 ratio
by means of a T-piece and two 40 cm lengths of 0.064 mm I.D. PEEK tubing. Gradient
elution was performed mixing solvent A (from pump P1) consisting of 1% (v/v) of ACN
in water with 0.1% (v/v) FA and solvent B (from P2) consisting of 90% (v/v) of ACN in
water with 0.1% (v/v) FA at a total flow-rate of 50 µL/min. The final gradient program
Chapter 5
114
was: from 0 to 12 min isocratic at 100% A, from 12 to 50 min a linear gradient 100%
A to 20% A, and from 51 to 60 min isocratic at 100% A.
Figure 5.1: Schematic diagram of the setup of the LC × LC-UV/MS system. P1, P2: first-dimension gradient pumps. P3, P4, P5: second-dimension gradient pumps. AS1, AS2: autosampler. S: a ten-port switching valve with two sample loops of 25 µL (L1, L2). The first-dimension column is a Luna C18(2) (50 mm
× 1.0 mm), the second-dimension column is a primesep A (30 mm × 4.0 mm). Detectors: UV-detector at 220 nm for the first dimension, DAD for the second dimension, and a Q-TOF2 for APCI-MS.
The injection volume was either 5 or 25 µL, depending on the experiment performed.
The column temperature was kept constant at 40 °C. The UV detector was set on
wavelengths of 220 and 254 nm. LC–MS was performed in atmospheric-pressure
chemical ionization (APCI) mode on a Waters/Micromass QTOF2 mass spectrometer
equipped with an APCI interface probe (Waters, Etten-Leur, The Netherlands). When
applied in combination with the first-dimension column, the probe temperature was set
on 300 °C and the ion source temperature on 100 °C, the corona current and the cone
voltage were set at 3.0 µA and 25 V, respectively. The second-dimension separation
system consisted of a Shimadzu HPLC system (‘s Hertogenbosch, The Netherlands)
comprising of a LC-10AD-VPi and two LC-10ADvp pumps, CTO-ACvp column oven, and
an Agilent Technologies HP1050 DAD detector (Amstelveen, The Netherlands). The
mixed-mode column was packed in-house (30 mm × 4 mm I.D.) packed with Primesep
LC × LC-UV/MS for the separation of complex samples
115
A material (5 µm particle size), purchased from Aurora Borealis Control (Schoonebeek,
The Netherlands). In the step gradient at the second-dimension column, three solvents
were applied: Solvent C (from P3) consisting of 1% (v/v) of ACN in water with 0.01%
(v/v) TFA, solvent D (from P4) consisting of 1% (v/v) of ACN in water with 0.5% (v/v)
TFA, and solvent E (from P5) consisting of 10% (v/v) of water in ACN with 0.01% (v/v)
TFA (see below for details). The flow-rate at the second-dimension column was 1.5
mL/min. The column temperature was kept constant at 40 °C. LC–MS was performed
on the same Q-TOF2 mass spectrometer in APCI mode (probe temperature 450 °C
instead of 300 °C).
LC × LC methodology
In the LC × LC setup (see Figure 5.1), the flow-rate from the first-dimension column
was split 1:1, with 25 µL/min to the UV detector and 25 µL/min being transferred to
the second-dimension column. The columns were connected by an electronically
controlled two-position ten-port Rheodyne valve (S in Figure 5.1) which was
pneumatically actuated using compressed air at 7 bar. The valve contained two sample
loops (L1 and L2 in Figure 5.1) of 12.73 cm × 0.50 mm I.D. PEEK tubing with an
internal volume of 25 µL each. The two sample loops were used simultaneously and
alternately: while one loop (loading position) collected a 25-µL fraction of the first-
dimension RPLC column eluate, the other (injection position) released and injected the
previously collected fraction onto the second-dimension Primesep A column. Switching
between the two sample loops was performed every 60 s (duration of the modulation
cycle). This alternating process is continuously applied throughout the entire LC × LC
run.
During LC × LC experiments, chromatographic data acquisition was performed using
Agilent Technologies Chemstation software version 1.04. After acquisition, the data
were exported to a comma-separated value file and converted by an in-house written
two-dimensional (2D) converter program which converts the data into a matrix with
rows corresponding to a 60 s duration and data columns covering all successive
second-dimension chromatograms. Contour representations of the 2D chromatograms
were visualized with Transform version 3.4 which is part of Neosys (CREASO,
Apeldoorn, The Netherlands).
Experiments were performed to optimize the mobile-phase composition of the second-
dimension separation by means of a step gradient, where the ACN and TFA
concentrations were varied independently. In total, 12 different conditions (as detailed
below) were compared with an ACN concentration of 20, 40 or 60% and a TFA
concentration of 0.05, 0.10, 0.15 or 0.20%.
Chapter 5
116
The intraday (n = 3) and interday (n = 4) variability of the retention time of the
complete LC × LC system was determined. In addition, calibration curves were
measured with three components with different polarities: caffeine, metamitron and
hydrocortisone. The achievable detection limits were determined for the same three
compounds. In these experiments, the LC × LC system was operated with the usual
linear gradient in the first-dimension separation and fixed isocratic ACN and TFA
concentrations of 40% and 0.05%, respectively, in the second-dimension separation.
Results and discussion
The development and optimization of a comprehensive two-dimensional LC × LC
system requires a number of steps and considerations, e.g., concerning the choice of
stationary and mobile phases, column dimensions, injection volumes, sample transfer
between the two columns, and optimization of the experimental conditions like mobile-
phase composition, gradients and flow-rates. The final aim of our study is the
development of an LC × LC system for the comprehensive analysis of endogenous
compounds in complex biological samples like urine. Urine contains many both polar,
medium polar, and relatively non-polar components, both ionic and neutral. In order to
facilitate the development of the system, a set of 32 model compounds was chosen
(Table 5.1), greatly varying in dissociation constant pKa and polarity, indicated by the
log of octanol–water partition coefficient (log P) and the log of distribution coefficient
(log D). The difference with log P is that log D is pH dependent.
The optimization of the separation of these 32 model compounds by means of LC × LC
is reported in this paper.
The need for two-dimensional separation
Obviously, the need for a two-dimensional separation system has to be demonstrated
first and its advantages proven. This is best done by performing a separation of the
test-set of 32 model compounds on a 50 mm × 1 mm I.D. reversed-phase C18 column
(3 µm particles) and an optimized flow-rate of 50 µL/min. The resulting UV
chromatogram (220 nm) is shown in 2. As can be seen from Figure 5.2 as well as from
the retention times tR and calculated resolution Rs shown in Table 5.1, several
compounds are poorly resolved or even co-eluting in the chromatogram, even
compounds that have apparently widely different log P values.
LC × LC-UV/MS for the separation of complex samples
117
Table 5.1: Retention times and resolutions of 32 compounds with log P and log D at pH 3 separated with a conventional LC-UV system at 220 nm
Number Compound tR (min) Rs Log P(*) Log D (**)
at pH=3
Molecular
Weight
1 Creatinine 3.83 < 0.5 -1.69 -3.3 113.12
2 Alanine 4.07 < 0.5 -3.04 -3.2 75.07
3 Dopamine 4.18 < 0.5 -0.22 -2.8 153.18
4 Aniline 4.18 < 0.5 0.94 -0.6 93.13
5 Sulfamic acid 4.21 < 0.5 -1.42 -3.4 173.19
6 Tyrosine 4.22 < 0.5 0.38 -2.2 181.21
7 Sulfaguanidine 4.53 < 0.5 -1.22 -1.3 214.24
8 3,4-Hydroxybenzoic acid 8.60 2.0 1.16 1.1 154.12
9 Tryptophan 9.63 < 0.5 1.04 -1.5 204.22
10 Paracetamol 10.02 < 0.5 NC NC 151.16
11 2,6-Dimethylanilin 15.01 2.0 NC NC 121.86
12 Theophyline 20.46 2.8 -0.18 -0.2 180.16
13 Nadolol 21.07 2.1 1.29 -1.8 309.45
14 Caffeine 22.61 1.7 -0.13 -0.1 194.19
15 Vanillin 23.48 1.8 1.19 -1.2 152.15
16 Sulfamethoxypyridazine 24.09 1.0 NC NC 280.34
17 Thiobenzamide 24.62 1.0 NC NC 137.21
18 Metamitron 24.62 < 0.5 0.83 0.8 202.21
19 Benzoic acid 24.95 < 0.5 1.87 1.9 122.12
20 Atrazin-desethyl 26.10 0.9 1.50 1.4 187.63
21 Aspirin 26.31 1.3 1.23 1.1 180.16
22 Midazolam 26.83 3.4 1.53 1.7 325.93
23 Imipramine 27.47 3.3 4.80 1.7 316.88
24 Hydrocortison 29.39 1.8 1.43 1.4 362.46
25 Carbamazepine 31.82 0.9 2.67 2.7 236.33
26 Nitrazepam 32.92 1.0 2.18 1.8 281.27
27 Prometryn 33.72 0.5 3.44 2.4 241.35
28 Oxazepam 34.21 4.8 2.31 2.3 286.71
29 Desmethyldiazepam 34.63 0.5 3.15 2.6 270.72
30 Naproxen 37.46 3.2 3.18 3.0 230.26
31 Diazepam 37.88 < 0.5 2.96 2.4 284.74
32 Ibuprofen 42.78 2.7 3.72 3.7 206.32
(*) LogP was calculated with ACDlabs 8.02
(**) LogD was calculated with ACDlabs 8.05
NC: Not Calculated by ACDlabs
Chapter 5
118
However, log P can only be accurately calculated for non-ionic compounds. As several
compounds in our test-set contain one or more ionogenic groups, they may exist in
different (ionic) forms in solution, depending on the pH. Therefore, the distribution
coefficient log D would give a more appropriate description of the complex partitioning
equilibriums. From Table 1, it may be concluded that a better correlation between
retention time and log D is achieved than between retention time and log P.
Figure 5.2: Chromatogram of the 32-component test-set on the first-
dimension column only. Gradient: 0–12 min at 100% A, 12–50 min linear gradient from 100–20% A. 5_L injection and UV-detection at 220 nm.
General design of the LC × LC system
From these experiments, it was concluded that a two-dimensional LC × LC system
would be more appropriate, although a good separation could also be achieved with a
longer column and a shallower gradient program, but only at the expense of far longer
analysis times. As the composition of urine, the sample type finally aimed at in this
research project, is far more complex than that of our test-set mixture, the
development of an LC × LC seems more successful and appropriate in the long term.
A two-dimensional LC × LC system may consist of two similar columns, operated with
the same or different mobile phases, but obviously the best performance in terms of
resolution and peak capacity can be achieved if two different types of stationary phases
and mobile phases are used. Different stationary phases and different orders of these
LC × LC-UV/MS for the separation of complex samples
119
columns were tested, including RPLC, hydrophilic interaction chromatography (HILIC),
cation and anion ion-exchange chromatography (CEX and AEX), and mixed-mode ion-
exchange/reversed-phase chromatography.
Both, the RPLC × RPLC and HILIC × RPLC experiments (data not shown) did not
provide additional separation in comparison with conventional RPLC separation. The
results obtained with an RPLC × RPLC setup show similar selectivity for the compounds
in comparison with a one-dimensional setup. This can be concluded from the 2D plot
which shows that the analytes are located in a straight slope. In the HILIC × RPLC
setup, the high organic modifier gradient (99–80% of ACN) of the first-dimension
HILIC mobile phase hampered the separation on the second-dimension RPLC column.
Most test compounds did not exhibit any retention at all. An intermediate dilution step
with water resulted in unfavorable sample dilution.
Further experiments showed that a combination of RPLC and a mixed-mode CEX/RPLC
column results in the highest stationary phase orthogonality for the compounds in our
test-set.
The mixed-mode column was preferred over conventional CEX or AEX columns because
it can be operated without the need for high salt concentrations in the mobile phase,
and is therefore more readily compatible with MS detection. With respect to the
separation of highly polar and especially early-eluting ionogenic compounds, a setup
consisting of RPLC in the first dimension and the mixed-mode CEX/RPLC column in the
second dimension is the best solution: compounds poorly retained in the first-
dimension RPLC are then separated in the second dimension. When using a reversed
column order with a mixed-mode CEX/RPLC column in the first dimension and RPLC in
the second dimension, the separation of most ionogenic compounds is strongly
hampered. Since the second-dimension mixed-mode CEX/RPLC column exists of two
different separation mechanisms, two different mobile-phase optimizations are
required. Our solution was to use step gradients in the second dimension. Additionally,
the chosen setup can readily be converted into a one-dimensional RPLC system,
whenever necessary for development purposes.
The LC × LC system consisted of a 50 mm × 1 mm I.D. RPLC first-dimension column,
operated at a flow-rate of 50 µL/min and a 30 mm × 4 mm I.D. Primesep A second-
dimension column, operated at a flow-rate of 1.5 mL/min. Transfer of 60-s fraction
from the first-dimension separation to the second-dimension column is performed by
means of an electronically controlled two-position switching valve, containing two 25-
µL sample loops, which are alternatingly used for fraction collection from the first-
dimension column and injection of the previous fraction into the second-dimension
Chapter 5
120
column. This means that the modulation time is 60 s. The first-dimension separation is
achieved using a linear mobile-phase gradient, while the second-dimension separation
is performed isocratically. Isocratic elution is easier to handle in the fast runs (typically
60 s) in the second-dimension separation and avoids the need for a column-
conditioning step and the negative effects of the rapid changes of the modifier
concentration on the background signals of UV or MS detectors. Nevertheless, gradient
elution in the second dimension would be more favorable, because it would reduce the
number of wrap-rounds. A wrap-around occurs, when the retention time of a
component in the second-dimension separation exceeds the modulation time. In that
case, the component elutes with compounds from the next modulation cycle, at a
retention time different than its real retention time, thus possibly resulting in co-elution
problems, and with a significantly broader peak, adversely affecting the detection limit
of that component.
Injection volume in first-dimension column
In choosing the injection volume in the first-dimension column, it must be considered
that a relatively large injection volume is preferred, given the expected low
concentrations of some of the relevant endogenous compounds in urine. On the other
hand, the inner diameter of the first-dimension column, and thereby the allowable
injection volume, should be limited in order to limit the flow-rate. Complete transfer of
the column eluate of the first-dimension column to the second-dimension column,
which is operated with a typical analysis time of 60 s, is most favorable.
The influence of the injection volume, varied between 1 and 25 µL, and the analyte
retention time and peak width in the first-dimension separation was evaluated for four
compounds. The results are summarized in Table 5.2. The injection volume had little
effect on the retention time in the gradient-elution system, but significant peak
broadening (20–30%) was observed for the earlier eluting compounds at higher
injection volumes. The peak areas were not affected. An injection volume of 25 µL is
used in further experiments.
LC × LC-UV/MS for the separation of complex samples
121
Table 2:
Retention time shift and peak widths related to injection volume of 4 compounds
m-Tyrosine Tryptophan Caffeine Metamitron
injection
volume
(µl)
tR
(minutes)
Width at
½ height
(minutes)
tR
(minutes)
Width at
½ height
(minutes)
tR
(minutes)
Width at
½ height
(minutes)
tR
(minutes)
Width at
½ height
(minutes)
1 2.89 0.63 6.03 0.81 9.73 0.27 11.48 0.30
5 2.86 0.68 6.00 0.86 9.70 0.27 11.43 0.31
10 2.86 0.62 6.01 0.86 9.73 0.27 11.45 0.30
25 2.74 0.80 5.80 1.06 9.68 0.27 11.39 0.31
Table 3:
Effects of eluent and the injection volume on the retention times of the re-injected fraction in the secondary column
Tryptophan(n=3) Caffeine (n=3) Metamitron(n=3) Carbamazepine(n=3) Hydrocortisone (n=3)
tR (minutes) tR (minutes) tR (minutes) tR (minutes) tR (minutes)
injection
eluent
A
84%A/
16%B
eluent
A
79%A/
21%B
eluent
A
72%A/
28%B
eluent
A
58%A/
42%B
eluent
A
62%A/
38%B
volume
(µL) 1stD eluent 1
stD eluent 1
stD eluent 1
stD eluent 1
stD eluent
25 3.85 3.53 1.09 1.04 3.21 3.05 5.61 5.37 3.94 3.64
10 3.88 3.55 1.09 1.05 3.22 3.08 5.56 5.50 3.90 3.79
5 3.87 3.52 1.10 1.06 3.21 3.09 5.12 5.08 3.92 3.82
1 3.63 3.52 1.09 1.07 3.25 3.12 5.28 5.10 3.92 3.84
Chapter 5
122
Effect of injection volume, organic and ionic modifier in the
second-dimension separation
The first-dimension separation is performed in gradient-elution mode. Fractions
transferred from the first-to the second-dimension column therefore differ in the
organic-modifier content. This may in turn affect the applicable transfer volume, that
is, the injection volume into the second-dimension column. This effect was studied for
five compounds, by evaluating the second-dimension retention times after direct
injection of volumes in the range between 1 and 25 µL in two solvent compositions,
that is a fixed composition of 1% acetonitrile in water with 0.1% formic acid and the
composition at which the particular compounds elutes in the acetonitrile gradient from
the first-dimension column. The results are summarized in Table 5.3. A complete
interpretation of the minor effects on retention time and peak width is complicated by
the differences in interaction of the components tested with the different retention
mechanisms active in the mixed-mode column. For instance, tryptophan shows little
volume-dependent behavior which is consistent with ion-exchange interaction, while
metamitron and hydrocortisone show more hydrophobic interaction, and is thus more
influenced by the injection volume. The most important conclusion from these
experiments is that the injection volumes, within the range of 1–25 µL, and the
mobile-phase composition have generally little influence on the retention behavior in
the second-dimension separation. This can to a large extent be attributed to the mixed
retention mechanisms in the second-dimension column.
Peak capacity of the two-dimensional system
The theoretical peak capacity of an LC × LC system can be calculated to show the
advantage of the two-dimensional LC × LC system over a one-dimension separation
system. The peak capacity nc,1 of a one-dimension separation in gradient-elution mode
can be calculated from Eq. (1) [29]:
W
ttn
RnR
c
1,,
1,
(1)
where tR,n and tR,1 are the retention times of the last and the first eluting peaks,
respectively, and W is the average 4σ peak width (at the baseline). Using this
equation, the peak capacity of the used C18 column under the conditions applied is
theoretically calculated to be 92 peaks in 46 min. The peak capacity nc,2 of the isocratic
second-dimension separation can be calculated from Eq. (2) [30]:
LC × LC-UV/MS for the separation of complex samples
123
1
2,ln
41
t
tNn
n
c
(2)
where N is the plate number, and t1 and tn are the retention times of the first and the
last peak, respectively. Using this equation, the peak capacity of the mixed-mode
CEX/RPLC column is 10 peaks in 1 min.
The total peak capacity nc, tot can be calculated by multiplying the calculated peak
capacities of the two dimensions according to Eq. (3):
(3)
Thus, the total peak capacity of our system is 920 peaks. However, during loop
trapping of the 60-s fractions in the transfer step, remixing occurs of components
initially separated in the first-dimension column but contained in the same collected
fraction. Therefore, the actual peak capacity of the first-dimension column is
approximately 46 rather than 92 peaks, and consequently, the total theoretical peak
capacity of the LC × LC system under the current operating conditions can be
calculated to be 460 peaks. Although in real practice the theoretical peak capacity
cannot be obtained, the two-dimension separation obviously provides a higher peak
capacity than a conventional one-dimension separation in the same analysis time (460
vs. 92).
Optimization of the second-dimension separation
Whereas the first-dimension separation is based on hydrophobicity, the second-
dimension separation is based on both cation-exchange and hydrophobic interactions.
This implies that despite the fact that the retention mechanisms show significant
differences, the phase systems are not completely orthogonal. Whereas ACN is used as
an organic modifier in the first-dimension separation, in the second-dimension
separation both an organic modifier (ACN) and an (acidic) ionic modifier are needed for
compound elution. Formic acid, acetic acid and TFA were tested as ionic modifiers,
because these volatile acids are compatible with MS detection. However, both formic
acid and acetic acid turned out to be too weak ionic modifiers for the primesep A
column. Therefore, TFA was applied as an acidic ionic modifier in the second-dimension
separation. With this system, featuring gradient elution (1–80% ACN) in the first-
dimension RPLC column and isocratic elution (40% ACN and 0.05% TFA) in the second-
dimension mixed-mode CEX/RPLC column, reasonable good results were obtained,
2,1,, cctotcnnn
Chapter 5
124
although not completely satisfactorily because of co-elution of components in some
parts of the chromatogram. Therefore, possibilities to improve the second-dimension
separation were considered.
Gradient elution is generally applied when complex samples need to be separated
which contain compounds covering a wide range in retention factors. However, in fast
LC × LC systems, the use of gradient elution in the second dimension is not readily
accomplished due to the short modulation time (30–60 s). With only a few exceptions,
for example, when using high temperature liquid chromatography (HTLC) in the second
dimension, as presented by Stoll et al. [31], the equilibration time for reconditioning of
the second-dimension column is found to be the time-limiting factor [32]. To overcome
these limitations and still apply more optimum mobile-phase compositions in different
parts of the chromatogram, the use of step gradients can be considered. This would
require a much shorter equilibration time. However, a complicating factor related to
the use of a mixed-mode column in the second dimension is that two step gradients of
two different modifiers (ACN and TFA) might be necessary to achieve an efficient
separation.
In order to evaluate the viability and potential of step gradients in the second
dimension, a series of experiments was performed where the test-set of 32 model
compounds was analyzed with the LC × LC system with the first-dimension gradient
elution, while three different percentages of organic modifier and four different ionic
modifier conditions were applied. In these twelve experiments, ACN percentage of 20,
40, and 60% were combined with TFA concentrations of 0.05, 0.10, 0.15 and 0.20%,
respectively. Preliminary experiments showed that modifier concentrations beyond the
ranges of 20–60% ACN and 0.05–0.20% TFA resulted in excessive broad peaks or
(almost) complete loss of separation. As expected, the results showed that under
different conditions different parts of the 2D chromatogram provided better peak
resolution. From an evaluation of the number of separated peaks and the peak
resolution in the second dimension, optimum separation conditions were selected. The
peak-to-peak resolution at half heights was calculated from Eq. (4) [33]:
21 2/12/1
1218.1
ww
ttR
RR
s
(4)
where tR2 and tR1 are the retention times of adjacent peaks 2 and 1, respectively, and
w1/21 and w1/22 are the peak width at ½ height of these peaks, respectively. The
results are summarized in Table 5.4: for each of the twelve experiments the number of
LC × LC-UV/MS for the separation of complex samples
125
peaks detected as well as the number of peak pairs showing a peak-to-peak resolution
better than 1.5 are given at three different time points during the LC × LC run. The
first fraction was 3–5 min, the second 27–31 min, and the third 33–37 min. The goal of
these experiments was to find second-dimension mobile-phase conditions that provide
optimum resolution in each part of the selected time windows. In some cases, that is
when the first-dimension peak was divided over two fractions, two peaks of the same
compound might be observed in two second-dimension chromatograms. These were
counted as two peaks in a time fraction. In addition, the occurrence of wraparound
peaks adversely affects the resolution in the time window. Typical baseline peak widths
in the second dimension are 6–8 s, while peaks with widths exceeding 15 s are
probably due to wrap-arounds.
Table 5.4: Number of detected peaks and resolutions in the 2nd dimension slices
Experimental conditions Fraction(1) 3-5
minutes
Fraction(2) 27-31
minutes
Fraction(3) 33-37
minutes
Experiment %TFA/%
ACN
# of
peaks
Rs>1.5 # of
peaks
Rs>1.5 # of
peaks
Rs>1.5
1 0.05/20 5 3 (A) 6 6 2 2
2 0.05/40 7 5 (B) 11 6 5 5
3 0.05/60 4 3 (C) 7 4 8 5
4 0.10/20 4 2 9 8 (D) 3 2
5 0.10/40 4 2 10 7 7 4
6 0.10/60 4 2 9 4 8 7 (G)
7 0.15/20 5 2 10 9 (E) 2 2
8 0.15/40 4 2 8 5 10 9 (H)
9 0.15/60 3 0 8 6 7 6
10 0.20/20 6 2 9 8 (F) 4 4
11 0.20/40 3 2 8 4 9 7 (I)
12 0.20/60 2 0 9 6 6 4
For each of the three fractions, the 2D chromatogram obtained under the three best
conditions in terms of the number of peaks detected and number of peaks with a
resolution better than 1.5 were studied in some more detail. The chromatograms are
shown in Figure 5.3, where Figure 5.3A, B and C correspond to the three best results
for fraction 1 (3–5 min), Figure 5.3D, E, and F with fraction 2 (27–31 min), and Figure
5.3G, H and I with fraction 3 (33–37 min) (see also Table 5.4).
Chapter 5
126
For the separation of fraction 1, representing compounds that hardly show any
retention in the first-dimension RPLC column, 0.05% TFA seems sufficient, but in fact
20% of ACN (Figure 5.3A) is not sufficient to elute these compounds without wrap-
arounds. With 60% of ACN (Figure 5.3C), on the other hand, the compounds co-elute
and show poor resolution. Therefore, a mobile phase with 40% of ACN and 0.05% of
TFA (Figure 5.3B) provides optimum results with seven peaks in one 60-s fraction and
with five peaks with a resolution above 1.5. Furthermore, the strength of the ionic
properties of this mixed-mode column is proven; even with 40% ACN the compounds
are retained.
The chromatograms of fraction 2 (Figure 5.3D, E, and F) show high similarity. The fact
that the mobile phase with 20% ACN and 0.15% TFA gave ten peaks and nine peaks
with a resolution above 1.5 indicates that probably only one peak is split over two
fractions. A somewhat closer look at the data indicates that in this fraction a low
percentage of both ACN and TFA provide the best separation.
In fraction 3 (Figure 5.3G, H, and I), both TFA and ACN play an important role in the
separation. A mobile phase with 40% ACN and 0.15% TFA (Figure 5.3H) resulted in the
highest number of separated peaks and highest number of peaks with a resolution
above 1.5.
The results of this series of experiments indicate, that optimum performance of the
second-dimension separation can be achieved by an independent and stepwise change
in the %TFA and %ACN in the mobile phase applied, while maintaining isocratic
conditions in each individual chromatogram. This will result in a higher peak capacity of
the entire LC × LC system. In addition, the number of changes in mobile-phase
composition during the LC × LC run must be limited in order to avoid excessive
background shifts which may hinder the construction of the 2D chromatograms (see
below). For instance, given the limited effect of %TFA in fraction 2 and the optimum
%TFA in fraction 3, a %TFA of 0.15% is kept throughout final part of the
chromatogram. Limiting the number of changes in the mobile-phase composition in the
second-dimension separation will also improve the reliability of the complete system.
From these experiments, it may be concluded that the conditions applied for Figure
5.3B, E and H can be considered optimum for the test-set of 32 components. The 2D
chromatograms from complete LC × LC runs acquired using either of these conditions
which are depicted in Figure 5.4. These chromatograms indicate, that optimum
separation can be achieved by the use of three different mobile-phase compositions for
the second-dimension separation, that is 0.05% TFA and 40% ACN during the first 12
min of the first-dimension separation, 0.15% TFA and 20% ACN between 12 and 31
LC × LC-UV/MS for the separation of complex samples
127
min, and 0.15% TFA and 40% ACN between 31 and 50 min of the first-dimension
separation.
Figure 5.3: Expanded 2D chromatograms of three different time windows obtained under different isocratic solvent conditions in the second-dimension separation. Gradient elution in the first-dimension separation (see Figure 5.2 for conditions). Fraction 1 (timer window 3–5 min): (A) 0.05% TFA, 20% ACN, (B) 0.05% TFA, 40% ACN, (C) 0.05% TFA, 60% ACN. Fraction 2 (time window 27–31 min): (D) 0.10% TFA, 20% ACN, (E) 0.15% TFA, 20% ACN, (F) 0.20% TFA, 20% ACN. Fraction 3 (time window 33–37 min): (G) 0.10% TFA, 60% ACN, (H) 0.15% TFA, 40% ACN, (I) 0.20% TFA, 40% ACN.
Figure 5.4: 2D chromatograms of the 32-component test-set obtained with gradient elution in the first-dimension separation (see Figure 5.2 for conditions) and with isocratic elution under the three different optimal conditions in the second-dimension separation: (A) 0.05% TFA, 40% ACN; (B) 0.15% TFA, 20% ACN; (C) 0.15% TFA, 40% ACN.
Chapter 5
128
These time windows are shown as boxes in the chromatograms in Figure 5.4. Note that
serious co-elution and wrap-arounds occur outside these boxes in either of the 2D
chromatograms in Figure 5.4.
Subsequently, these optimized conditions were applied within one LC × LC run. A
typical chromatogram obtained is shown in Figure 5.5. The data show that good
separation is obtained for most of the 32 components in our test mixture. The data
also clearly show the effect of the step gradient, especially of the change in TFA
concentration on the construction of the 2D chromatogram: the higher TFA
concentration in the second and third time zone results in a more significant UV cut-off,
represented by a color change of the drawing of the background.
Figure 5.5: 2D chromatograms of the 32-component test-set obtained with gradient elution in the first-dimension separation (see Figure 5.2 for conditions) and with the three step-gradient isocratic elution in the second-dimension separation: time window 1–12 min: 0.05% TFA, 40% ACN; time window 12–31 min: 0.15% TFA, 20% ACN; and time window 31–50 min 0.15% TFA, 40% ACN
Under these optimized conditions, repeatability and reproducibility of the retention
times in the chromatograms were tested. The intraday (n = 3) and interday (n = 4)
variability of the retention times (expressed as RSD) was 0.5% and 0.7%, respectively.
Good linearity (R2 higher than 0.996) was also achieved when acquiring calibration
LC × LC-UV/MS for the separation of complex samples
129
curves in the concentration range between 5 and 100 µM for three different compounds
varying in polarity, that is caffeine, metamitron and hydrocortisone. Detection limits of
caffeine, metamitron and hydrocortisone were between 1 and 2 µM.
Combining LC × LC with APCI-MS detection
When the LC × LC system developed will be applied to the separation of endogenous
compounds in urine, for instance, for metabolomics or biomarker studies, the use of
MS detection rather than UV detection will be an important aspect. Therefore, in
developing the mobile-phase conditions for the LC × LC system, mobile-phase
compatibility issues were already taken into account: no non-volatile mobile-phase
additives were considered. However, the use of TFA was found to be important in the
efficient use of the second-dimension mixed-
Figure 14: 2D chromatograms of the 32-component test-set with (A) UV-detection at 220 nm, and (B) total-ion current of positive-ion APCI-MS detection. Compounds: 1, 5–7, 9–18, 20–21, 24–32 were identified. Compound numbers are corresponding with the compounds in Table 5.1. Condition: see Figure 5.5 and Section 2.
mode CEX/RPLC column, as both formic and acetic acid turned out to be too weak ionic
modifiers. The presence of TFA in the mobile phase as well as the high flow-rate of the
second-dimension column, that is typically 1.5 mL/min, indicates that the use of APCI
in positive-ion mode for analyte ionization in MS is most likely more appropriate than
the use of electrospray ionization (ESI). TFA is known to provide ionization suppression
effects in ESI and high flow-rates are not optimum for ESI either. Figure 5.6 shows the
2D chromatogram of LC × LC separation with the optimized conditions using UV and
MS detection simultaneously. A combination of the two detectors is favorable because
some compounds do not ionize under these conditions, while some compounds do not
show UV activity. Compound 30 and 32 (naproxen and ibuprofen), for instance, are
readily detected by UV detection (Figure 5.6A), but are not detected by positive-ion
APCI-MS (Figure 5.6B), whereas compounds 29 and 31 (desmethyldiazepam and
Chapter 5
130
daizepam) are only weakly detected in the UV chromatogram (Figure 5.6A), while they
clearly show up in the APCI-MS chromatogram (Figure 5.6B). By the use of both UV
and MS detection, at least 26 compounds of 32 components can be detected and
quantified. The peaks of compounds 22 and 31 are relative broad, which indicates that
these compounds are wrapped around. This has to be taken in account if these
compounds have related to a peak in the first-dimension analysis.
Figure 5.7: 2D chromatograms of (A) a four-time diluted human urine sample, and (B) the same four-time diluted urine sample spiked with 25 µM of the 32-component test-set. Compounds identified by APCI-MS, (7) sulfaguanidine; (1) creatinine; (14) caffeine; (24) hydrocortisone; (25) carbamazepine; (28) oxazepam; (31) diazepam. Condition: see Figure 5.5 and Section 2.
As a final step in the developmental project for comprehensive two-dimensional LC ×
LC, four-time diluted urine samples, with and without spiking with the test-set of 32
components, were analyzed with the complete system, featuring both UV and positive-
ion APCI-MS detection. The resulting chromatograms are shown in Figure 5.7. No
attempts were made to identify the various endogenous components that showed up in
the urine sample, although both creatinine and caffeine were readily recognized. A
number of the components from the test-set detected in the chromatogram of the
spiked urine sample are also indicated in Figure 5.7B. These data show the potential of
the developed technology in urine analysis.
LC × LC-UV/MS for the separation of complex samples
131
Conclusion
In this paper, the development and optimization of a comprehensive two-dimensional
LC × LC system was described. The system features a 60-min gradient elution on a
first-dimension RPLC column, subsequent transfer of the column eluate in 60-s
fractions, and isocratic elution based on a step gradient of both organic and ionic
modifier (ACN and TFA) on a second-dimension mixed-mode cation-IEC/RPLC column.
The multi-step isocratic elution in the second dimension resulted in a higher peak
capacity and improved peak resolution of the complete system, without compromising
stability and background of the detectors used. Simultaneous UV and APCI-MS detec-
tion was demonstrated, which expands the applicability range of this approach in the
analysis of complex mixtures. Development and testing of the system was performed
using a 32-component test mixture of compounds widely differing in physico-chemical
properties (pKa, polarity measured as log D, ranging from −3.4 to +3.7). Some
preliminary results with (spiked) urine samples were also obtained.
The final system is highly flexible and has a broad applicability. The changes of the
organic and ionic modifier concentrations in the multi-step program are readily adapted
to optimize separation and peak capacity for other or even more complex mixtures.
The experimental strategy outlined can be applied in the optimization of two-
dimensional separation of other such mixtures. The compatibility with both UV and MS
detection provides chromatographic information with a high information content
concerning the sample constituents.
References
[1] J. Dalluge, J. Beens, U.A.Th. Brinkman, J. Chromatogr. A 1000 (2003) 69. [2] M.A. dahchour, J. Beens, R.J.J. Vreuls, U.A.Th. Brinkman, Trends Anal.
Chem. 25 (2006) 726. [3] H.-G. Janssen, S. de Koning, U.A.Th. Brinkman, Anal. Bioanal. Chem. 378 (2004) 1944. [4] S. de Koning, H.-G. Janssen, U.A.Th. Brinkman, J. Chromatogr. A 1058 (2004) 217. [5] C.R. Evans, J.W. Jorgenson, Anal. Bioanal. Chem. 378 (2004) 1952. [6] T. Stroink, M.C. Ortiz, A. Bult, H. Lingeman, G.J. de Jong, W.J.M. Under- berg, J. Chromatogr. B 817 (2005) 49. [7] P. Jandera, J. Fischer, H. Lahovska, K. Novotna, P. Cesla, L. Kolarova, J. Chromatogr. A 1119 (2006) 3. [8] P. Dugo, F. Cacciola, T. Kumm, G. Dugo, L. Mondello, J. Chromatogr. A 1184 (2008) 353. [9] R.A. Shellie, P.R. Haddad, Anal. Bioanal. Chem. 386 (2006) 405. [10] J.K. Killgore, S.R. Villasenor, J. Chromatogr. A 739 (1996) 43. [11] X. Chen, L. Kong, X. Su, H. Fu, J. Ni, R. Zhao, H. Zou, J. Chromatogr. A 1040 (2004) 169.
Chapter 5
132
[12] M.J. Gray, G.R. Dennis, P.J. Slonecker, R.A. Shalliker, J. Chromatogr. A 1041 (2004) 101. [13] L. Hu, X. Chen, L. Kong, X. Su, M. Ye, H. Zou, J. Chromatogr. A 1092 (2005) 191. [14] N. Tanaka, H. Kimura, D. Tokuda, K. Hosoya, T. Ikegami, N. Ishizuka, H. Minakuchi, K. Nakanishi, Y. Shintani, M. Furuno, K. Cabrera, Anal. Chem. 76 (2004) 1273. [15] T. Ikegami, T. Hara, H. Kimura, H. Kobayashi, K. Hosoya, K. Cabrera, N. Tanaka, J. Chromatogr. A 1106 (2006) 112. [16] A. van der Horst, P.J. Schoenmakers, J. Chromatogr. A 1000 (2003)693. [17] R.E. Murphy, M.R. Schure, J.P. Foley, Anal. Chem. 70 (1998) 1585. [18] S. Ma, L.X. Chen, G.A. Luo, K.N. Ren, J.F. Wu, Y.M. Wang, J. Chromatogr. A 1127 (2006) 207. [19] P. Dugo, T. Kumm, M.L. Crupi, A. Cotroneo, L. Mondello, J. Chromatogr. A 1112 (2006) 269. [20] G.J. Opiteck, K.C. Lewis, J.W. Jorgenson, R.J. Anderegg, Anal. Chem. 69 (1997) 1518. [21] G.J. Opiteck, S.M. Ramirez, J.W. Jorgenson, I.I.I. Moseley, Anal. Biochem. 258 (1998) 349. [22] M.M. Bushey, J.W. Jorgenson, Anal. Chem. 62 (1990) 161. [23] H.A. Holland, J.W. Jorgenson, Anal. Chem. 67 (1995) 3275. [24] K. Wagner, T. Miliotis, G. Marko-Varga, R. Bischoff, K.K. Unger, Anal. Chem. 74 (2002) 809.
[25] L.A. Holland, J.W. Jorgenson, J. Microcol. Sep. 12 (2000) 371. [26] D.A. Wolters, M.P. Washburn, J.R. Yates, Anal. Chem. 73 (2001) 5683. [27] E. Nagele, M. Vollmer, P. Horth, J. Chromatogr. A 1009 (2003) 197. [28] G. Mitulovic, C. Stingl, M. Smoluch, R. Swart, J.P. Chervet, I. Steinmacher, C. Gerner, K. Mechtler, Proteomics 4 (2004) 2545. [29] X. Wang, D.R. Stoll, P.W. Carr, P.J. Schoenmakers, J. Chromatogr. A 1125 (2006) 177. [30] J.C. Gidding, Anal. Chem. 39 (1967) 1027. [31] D.R. Stoll, J.D. Cohen, P.W. Carr, J. Chromatogr. A 1122 (2006) 123. [32] D.R. Stoll, P.W. Carr, J. Am. Chem. Soc. 127 (2005) 5034. [33] P.J. Schoenmakers, J.K. Strasters, A. Bartha, J. Chromatogr. 458 (1988) 355.