Anal Bioanal Chem (2004) 378 : 1952–1961
DOI 10.1007/s00216-004-2516-2
PA P E R I N F O R E F R O N T
Charles R. Evans · James W. Jorgenson
Multidimensional LC-LC and LC-CE
for high-resolution separations of biological molecules
Received: 23 December 2003 / Revised: 22 January 2004 / Accepted: 27 January 2004 / Published online: 12 February 2004
© Springer-Verlag 2004
Abstract In multidimensional separations, two or more
independent separation methods are coupled in an effort
to resolve complex mixtures. The displacement mecha-
nisms of each method should be orthogonal, such that lit-
tle correlation exists between the retention of compounds
in each dimension. When multiple orthogonal separation
methods are coupled such that all sample components are
subjected to complete analysis on all dimensions, the
method is considered “comprehensive”. The primary ad-
vantage of comprehensive multidimensional separations
over their one-dimensional counterparts is the potential for
dramatically enhanced resolution. High resolving power
can be achieved because the peak capacity of a compre-
hensive multidimensional separation is roughly equal to
the product of the individual peak capacities of each di-
mension. In this review, the theory and instrumentation of
two-dimensional liquid chromatography (LC-LC) and liq-
uid chromatography-capillary electrophoresis (LC-CE) sep-
arations are discussed. Some applications of these tech-
niques to the separation of biological molecules are high-
lighted. Future directions for the development of multidi-
mensional separations are also considered.
Keywords Multidimensional separations · Liquid
chromatography · Capillary electrophoresis · LC-LC ·
LC-CE · Biomolecules
Introduction
A multidimensional separation involves the coupling of
two or more separation mechanisms in order to carry out
a single analysis. Although simple in concept, multidimen-
sional separations can be exceedingly powerful in terms
C. R. Evans · J. W. Jorgenson (✉)
Venable and Kenan Laboratories, Department of Chemistry,
University of North Carolina at Chapel Hill,
Chapel Hill, NC 27599, USA
e-mail: jj@unc.edu
of resolution. The need for multidimensional separations
arises from the inability of one-dimensional (1D) separa-
tion methods to adequately resolve highly complex sam-
ples. One field of research that highlights the need for the
enhanced separation power of multidimensional tech-
niques is proteomics, which involves the characterization
of all proteins expressed by an organism. For example, the
simple single-celled organism E. coli produces a total
complement of over 4,000 proteins [1]. The most power-
ful 1D column chromatography techniques typically have
a peak capacity of only a few hundred [2], where peak ca-
pacity is defined as the maximum number of peaks that
can fit into the accessible separation space side by side
with a resolution of 1.0 from each neighboring peak [3].
Moreover, due to the usually random distribution of peaks
and the statistical probability of peak overlap, the actual
number of detectable components that can be expected to
be fully resolved is substantially smaller than the peak ca-
pacity [4, 5, 6]. The disparity between the capabilities of
traditional 1D separations and the complexity of real-
world samples has driven researchers to push toward mul-
tidimensional separations, due to the greatly enhanced re-
solving power they offer.
Multidimensional separation theory
In a multidimensional separation, a sample is first sub-
jected to separation via one method, and then the sepa-
rated components are further separated by at least one ad-
ditional independent method. Although there is no inher-
ent limitation to the number of independent separation
methods that can be coupled, practical constraints have
limited the vast majority of the multidimensional separa-
tions reported to date to two dimensions. Two-dimen-
sional (2D) separations provide enhanced resolution and
peak capacity as compared to 1D separation systems. In
an ideal case, the total peak capacity of a 2D separation
system is approximately equal to the product of the peak
capacities of both dimensions [4]. Hypothetically, if two
techniques with a peak capacity of 100 each were com-

bined, the resulting 2D system would have a maximum
theoretical peak capacity of 10,000. Achieving peak ca-
pacities this high using 2D separations is not a trivial task,
however, due to the inherent difficulties in coupling two
dissimilar techniques.
Due to the large number of different types of 2D sepa-
rations reported in the literature, it is important to make
clear distinctions between the different classes of 2D sep-
arations. One important concept is the difference between
separations that are 2D-in-space versus 2D-in-time. A sep-
aration which is 2D-in-space results in a separation of an-
alytes over a two-dimensional planar surface, such as a
gel. Techniques that are 2D-in-time are carried out by first
performing a 1D separation, often on a LC column, and
then sequentially subjecting individual fractions from this
first-dimensional separation to a second 1D separation.
Among techniques that are 2D-in-time, another distinc-
tion must be made between off-line and on-line multidi-
mensional separations. Off-line methods involve collect-
ing fractions from the first dimension, then later subject-
ing them to separation on the second dimension. On-line
techniques employ switching valves or other instrumenta-
tion that allow the fractions from the first dimension to be
transferred directly to the second dimension, usually while
the first-dimensional separation continues simultaneously.
Both off-line and on-line methods have notable advan-
tages and disadvantages. Off-line fraction collection al-
lows for easier optimization of the separation in both di-
mensions and permits sample manipulations, such as con-
centration or reconstitution, between dimensions. The abil-
ity to optimize each step of the separation independently
causes the entire analysis to take a substantial period of
time, which can become prohibitive when collecting large
numbers of fractions unless automated sample-handling
methods are used. On-line techniques tend to be substan-
tially faster because all the second-dimension separations
are run within the time that the first-dimension separation
takes to complete. However, run times and mobile phase
compositions tend to be more limited than they are with
off-line 2D separations. Instead, compromises in both di-
mensions must usually be made to allow successful on-
line coupling. Although both off-line and on-line techniques
have merit, the majority of recent efforts to develop 2D
separations have been focused on on-line methods, which
are the primary subject of this review.
Certain theoretical considerations apply to all multidi-
mensional separations, regardless of whether a technique
is 2D in space or time, off-line or on-line. Giddings estab-
lished two such fundamental requirements for ideal multi-
dimensional separations [4, 7]. First, the separation mech-
anisms must be orthogonal. This means that the chemical
or physical property by which analytes are separated must
be different for each dimension. For example, reversed-
phase liquid chromatography (RPLC), which separates based
on hydrophobicity, is orthogonal to ion-exchange chro-
matography (IEC), which separates according to coulom-
bic interactions. If a 2D system is truly orthogonal, the dis-
tribution of components (peaks) in one dimension should
not correlate with the distribution of components in the
1953
Fig. 1a,b Hypothetical orthogonal (a) and nonorthogonal (b) 2D
separations (Adapted from Ref. [35])
other dimension [8]. Hypothetical examples of orthogonal
and nonorthogonal 2D separations are shown in Fig. 1.
Giddings’ second criterion states that no resolution
gained in the first dimension of separation may be lost in
any subsequent dimension. For on-line, 2D-in-time sepa-
rations, this requirement is satisfied by designing instru-
mentation that transfers fractions from the first dimension
to the second dimension at a rate sufficient to ensure the
resolution of the first dimension is not substantially di-
minished. Therefore, the second dimension must be capa-
ble of completing separations very rapidly compared to
the first dimension. In order to achieve optimal resolution,
each peak in the first dimension should be sampled at
least three times by the second dimension [9]. The result-
ing need for high-speed analyses in the second dimension
represents an experimental challenge in carrying out 2D
separations. In spite of this and other difficulties, 2D sep-
arations that satisfy Giddings’ criteria and produce good
separations of complex mixtures have been demonstrated.
2D gel electrophoresis versus LC-LC and LC-CE
Before discussing 2D LC-LC and LC-CE, which are both
2D-in-time techniques, it is worthwhile to consider a 2D-
in-space technique known as two-dimensional polyacry-
lamide gel electrophoresis (2D-PAGE). At present, most
separations of complex mixtures of proteins are carried
out using 2D-PAGE, which was first reported in 1975 by
O’Farrell [10]. In 2D-PAGE, proteins are resolved into in-
dividual spots on a flat gel using two different separation
mechanisms. Proteins are separated first according to
charge via isoelectric focusing (IEF), and then according to
molecular size via sodium dodecyl sulfate polyacrylamide
gel electrophoresis (SDS-PAGE). Due to the high resolu-
tion and orthogonal nature of both separation mechanisms,
at least 2,000 proteins can be resolved in a single sample
[11]. 2D-PAGE also satisfies Giddings’ criteria for multi-
dimensional separations. In spite of its power, 2D-PAGE
has certain fundamental limitations. It is useful only for
proteins and is a labor-intensive analysis that usually re-
quires at least two days to complete. It cannot be coupled
in an on-line fashion to mass spectrometry (MS), thereby
requiring further manual intervention or complex robotics
1954
systems in order to implement MS analyses. Most signifi-
cantly, 2D-PAGE is only semiquantitative and may not
separate all the components present in a given sample;
proteins that are extremely large or hydrophobic may not
enter the gel, very acidic or basic proteins may be poorly
resolved, and low abundance proteins may be below the
detection limit of the silver-staining techniques used in
2D-PAGE [2].
Many of the limitations of 2D-PAGE are not shared by
2D separation methods based on liquid chromatography
(LC) and/or capillary electrophoresis (CE). When appro-
priate separation modes are selected, LC and CE can sep-
arate a broader range of biomolecules than 2D-PAGE, in-
cluding those that are large, hydrophobic, acidic or basic.
Depending on the detection technique used, LC and CE
can have greater sensitivity to analytes present in small
amounts than 2D-PAGE. Perhaps most importantly, sepa-
ration methods such as LC and CE are instrumental tech-
niques which are capable of full automation of sample in-
jection and interfacing between dimensions, and can be
coupled with a variety of detection methods such as UV
absorbance, laser-induced fluorescence or MS [2]. Given
these advantages, it is not surprising that researchers have
looked to 2D LC and related techniques to provide an al-
ternative to 2D-PAGE. The ongoing challenge for separa-
tion scientists has been to devise 2D separation methods
and instrumentation that can equal or surpass 2D-PAGE in
resolving power, speed, and reliability.
On-line multidimensional liquid chromatography
(LC-LC)
Heart-cutting multidimensional LC
One multidimensional method that has been used com-
monly is an on-line, 2D-in-time technique called heart-
cutting two-dimensional liquid chromatography [12]. In
this technique, the first-dimension separation is carried
out like a standard 1D analysis on a chromatography col-
umn. A single desired segment of the first-column efflu-
ent is then transferred, typically via a switching valve, to
a second column for further separation via a different
method. Heart-cutting is useful when extra resolution is
needed to examine a small segment of peaks in a complex
chromatogram, but it requires foreknowledge of the sam-
ple’s composition. Since heart-cutting does not subject the
entire sample to 2D separation, it is of little use when
every component of a sample is an analyte of interest, as
in proteomics. Heart-cutting multidimensional chromato-
graphy has been reviewed elsewhere in the literature [12];
this method of 2D separations is not considered further in
this report. The alternative to heart-cutting is known as
“comprehensive” multidimensional separations, because
all sample components are subjected to displacement on
both dimensions [13]. Comprehensive multidimensional
separations are more powerful than heart-cutting separa-
tions because they provide greater peak capacity.
Fig. 2 Schematic diagram of a directly coupled-column 2D LC in-
strument. Columns 1 and 2 represent two distinct, orthogonal sep-
aration modes
Directly coupled-column multidimensional LC
The simplest means of performing comprehensive 2D sep-
arations is to directly couple two or more columns with
orthogonal separation mechanisms in series, as shown in
Fig. 2. If such a system were run using isocratic elution,
this arrangement would be in violation of Giddings’ sec-
ond criterion for multidimensional separations, since ma-
terials separated on the first column could recombine on
the second dimension [4]. Furthermore, it is unlikely that
the same mobile phase would provide effective separa-
tions on both columns. However, useful 2D separations
can be carried out on certain arrangements of directly cou-
pled columns when both dimensions are operated in a
carefully controlled gradient elution mode [2]. Specifi-
cally, fractions from the first column are eluted onto the
second column by using a stepwise gradient of increasing
eluent strength with respect to the first column. After a
fraction is eluted from the first column, a continuous gra-
dient appropriate to the second dimension is run to sepa-
rate and elute the components of the fraction. Then an-
other first-dimension gradient step is performed and the
process is repeated.
The most prominent recent method that follows the
scheme of directly coupled-column multidimensional LC is
a technique known as multidimensional protein identifica-
tion technology (MudPIT), developed by Yates and colleagues
[14, 15, 16, 17, 18, 19, 20]. The method was developed
for proteomics research as an alternative to 2D-PAGE.
This technique follows the general scheme of 2D directly
coupled-column separations just described, although in-
stead of two separate columns in tandem, a single micro-
capillary column packed with two different stationary phases
is used. The microcapillary column is a packed fused-sil-
ica capillary with a pulled tip at the outlet end for direct
coupling to MS instrumentation via electrospray ionization.
The end of the capillary nearest the pulled tip is packed
with several centimeters of reversed-phase material, fol-
lowed by several centimeters of strong cation-exchange
(SCX) material. The column is connected to a PEEK mi-
crocross which splits the flow from a conventional quater-
nary LC pump, reducing it to 0.15–0.25 µL/min, and also
provides a connection where the electrospray voltage is
applied, as shown in Fig. 3. The 2D separation is carried
out using a multistep gradient elution profile, an example
of which is shown in Fig. 4. To elute fractions of the sam-
ple from the ion-exchange portion of the column onto the

Fig. 3 Diagram of MudPIT column and electrospray interface
(Adapted from Ref. [16])
Fig. 4 An example five-step MudPIT gradient profile. The bal-
ance of the mobile phase at any time is made up of buffer A.
Buffer A is 5% acetonitrile (ACN), 95% H2O, and 0.02% hepta-
fluorobutyric acid (HFBA). Buffer B is 80% ACN, 20% H2O, and
0.02% HFBA. Buffer C is 250 mM ammonium acetate/5% ACN,
95% H2O, and 0.02% HFBA. Buffer D is 400 mM ammonium ac-
etate/5% ACN, 95% H2O and 0.02% HFBA (Adapted from Ref.
[15])
reversed-phase material, a step-gradient of buffer with in-
creasing ionic strength is used. In between each of these
incremental steps, a linear gradient of increasing acetoni-
trile content is used to separate the eluted fraction on the
reversed-phase segment of the column. The outlet of the
column is interfaced with an electrospray mass spectrom-
eter. Typically the mass spectrometers used with MudPIT
have tandem mass spectrometry capabilities to enable the
identification of peptides via shotgun sequence analysis.
A database searching algorithm is then used to determine
the proteins present in the sample based on sequence
matches with the peptides detected by MS.
The capabilities of MudPIT as a separation method for
the purpose of proteomics have been assessed by several
studies. The peak capacity of the 2D separation carried out
in one 15-fraction MudPIT analysis was estimated as ap-
proximately 3,200 [16]. When the inherent peak capacity of
the mass spectrometer was included, the estimated overall
peak capacity increased to 23,000, which compares quite
favorably with 2D-PAGE techniques. In practice, MudPIT
is capable of identifying well over 1,000 proteins in a sin-
1955
gle sample. Washburn and colleagues applied MudPIT to
analysis of the proteome of the yeast S. cerevisiae and
identified 5,440 peptides using MS, which were assigned via
database searching to 1,484 unique proteins [15]. Among
these were a substantial number of low-abundance and
trans-membrane proteins, which suggests that the tech-
nique is largely unbiased and gives a representative sam-
pling of the yeast proteome. Lack of bias represents a sig-
nificant advantage over 2D-PAGE, which has limited dy-
namic range and typically performs poorly with trans-
membrane proteins since they are hydrophobic and do not
easily enter into the gel.
Another report indicates that the number of protein
identifications can be further improved by using a three-
phase MudPIT column in which an additional section of
reversed-phase material is added at the inlet of the column
to provide on-line removal of salts from the sample [18].
MudPIT has also been used for quantitative studies of
proteomic samples via the use of 15N isotope labeling
[17]. An assessment of quantitative MudPIT suggests that
the technique can provide useful data about relative levels
of protein expression, but that experimental reproducibil-
ity could be improved by enhancing resolution in order to
increase the number of peptide hits per protein [20].
One of the greatest advantages of MudPIT over other
2D techniques is the relative simplicity of the instrumen-
tation needed to carry out 2D separations. Only a single
quaternary gradient LC pump, an appropriately packed
microcapillary column, and minimal interfacing compo-
nents are required. However, the simplicity of this arrange-
ment also limits its flexibility. The two separation modes
coupled in tandem must both utilize gradient elution. Fur-
ther, the gradient in the first dimension must always be
run in a stepwise fashion in order to elute concise frac-
tions of sample onto the second column. The resolution
contributed by the first dimension is limited by the num-
ber of fractions transferred, which is equal to the number
of steps used in the first-dimension gradient. Each step re-
quires substantial time, typically 100 min, which includes
periods for reequilibration and washing of salts off the col-
umn. A maximum of 15 steps in the first-dimension salt
gradient have been reported [15], which may suggest that
the first dimension is undersampled in spite of long run
times. Notably, the only combination of separation modes
that has been used thus far in MudPIT is strong cation ex-
change–reversed phase, perhaps because this is the only
arrangement where suitable gradients can be applied alter-
nately to two different stationary phases in a single col-
umn in order to bring about an effective, orthogonal 2D
separation.
Column-switching multidimensional LC
Unlike directly coupled-column LC-LC, column-switch-
ing multidimensional LC uses valves to interface the first
column with the second and any subsequent columns.
This allows greater flexibility than directly coupled 2D
systems since, in principle, any two separation modes can
1956
Fig. 5 Schematic diagram of a
2D IEC-SEC instrument
(Adapted from Ref. [13])

be combined in any order. In general, the trade-off for this mental setup of this 2D LC system is provided in Fig. 5.
flexibility is greater instrument complexity [2]. Most The effluent of the cation-exchange column is directed to
comprehensive 2D techniques require the use of multiple a storage loop. When the valve is switched, the isocratic
LC pumps, automated switching valves, and computer- LC pump forces the contents of the loop onto the size-ex-
controlled instrumentation. In spite of the extra complex- clusion column where they are further separated. Mean-
ity, the majority of the true comprehensive multidimen- while the effluent of the first dimension is directed to the
sional separations reported to date have been based on second storage loop. The flow rates on the two columns
column-switching multidimensional LC. are selected such that the analysis time of the second di-
A wide variety of combinations of liquid chromato- mension exactly matches the amount of time it takes for
graphic modes have been used in column-switching 2D the first dimension to fill the storage loop. This allows all
LC separations, including ion exchange-reversed phase the effluent of the first column to be transferred to the sec-
(IEC-RPLC) [21, 22, 23, 24, 25, 26, 27], size exclusion- ond dimension. A computer is used to control the switch-
reversed phase (SEC-RPLC) [28, 29, 30, 31, 32], reversed ing of the valve and the acquisition of data from a UV ab-
phase-size exclusion (RPLC-SEC) [9], ion exchange-size sorbance detector at the end of the second column. In this
exclusion (IEC-SEC) [13], and normal phase-reversed manner the SEC column can analyze a large number of
phase (NPC-RPLC) [33]. Although many different instru- fractions from the ion-exchange column during the course
mental setups for 2D LC have been designed, most of the of a run lasting a total of 2.5 to 6 h. The chromatographic
components used are similar, and include an injector, two
isocratic or gradient LC pumps, a single column for the
first dimension, one or more columns for the second di-
mension, one or more computer-controlled switching valves,
and an appropriate detection system.
The first report of a 2D separation carried out using a
method resembling the comprehensive column-switching
approach was published in 1978 by Erni and Frei [28].
Their system, which was used to separate plant extracts,
consists of a size-exclusion column in the first dimension
coupled to a reversed-phase column in the second dimen-
sion. An eight-port switching valve is used to interface the
two dimensions. This configuration is very similar to
those presently used for comprehensive 2D separations,
but in this report only a total of seven fractions from the
first dimension were transferred to the second. This se-
verely reduces the amount of resolution that can be con-
tributed by the first dimension, thereby limiting the total
peak capacity of the system.
The first truly comprehensive system for 2D LC sepa-
ration was reported in 1990 by Bushey and Jorgenson
[13]. This system was used to analyze a mixture of protein
standards and a sample of human serum. A cation-ex-
change column, operated in gradient elution mode for the
first dimension, is coupled to a size-exclusion column Fig. 6 2D IEC-SEC chromatogram of protein sample. Peak iden-
used for the second dimension. As in the report by Erni tities: A glucose oxidase, B ovalbumin, C β-lactoglobulin, D tryp-
sinogen, E α-lactoglobulin, F conalbumin, G ribonuclease A,
and Frei, an eight-port switching valve is used to interface H hemoglobin; M exclusion volume “pressure” ridge, N inclusion
the two columns and allows on-line fraction transfer from volume “salt” ridge. Valve actuated every 6 min. UV detection at
the first column to the second. A diagram of the instru- 215 nm (Reproduced with permission from Ref. [13])
 1957

data are presented as a three-dimensional view of a 2D

chromatogram as shown in Fig. 6. The system was esti-
mated to have a total peak capacity of approximately 130,
which is the product of the peak capacities visually esti-
mated for each dimension. Other examples of column-
switching 2D LC separations using dual storage loops
have since been reported in the literature, some with
greater peak capacity [8, 9, 21, 22]. A 2D anion-exchange
reversed-phase system that was used to analyze a tryptic
digest of reduced porcine thyroglobulin gave a peak ca-
pacity over 2,000 [25].
A different approach, in which the effluent from the
first column is not captured by storage loops but instead is
transferred directly to the head of one of two parallel sec-
ond-dimension columns, was reported in 1997 by Opiteck,
Jorgenson, and Anderegg [29]. This system was used to
analyze the fragments produced by tryptic digests of the
proteins ovalbumin and bovine serum albumin. The first
dimension of their system consisted of six size-exclusion
columns connected in series. These were coupled via a
pair of four-port valves to two parallel reversed-phase LC
columns, which serve as the second dimension. This study
was also one of the first reports of on-line MS detection
used to analyze a 2D separation. A diagram of the entire
setup is shown in Fig. 7. As sample elutes from the series
of SEC columns, it is routed to RPLC column alpha. Fig. 7 Schematic diagram of a 2D SEC-RPLC instrument with
on-line MS detection. The two four-port valves are switched si-
Since the aqueous buffers used for SEC are weak eluents multaneously (Adapted from Ref. [29])
for RPLC, the sample material is concentrated in a narrow
zone at the head of the column. When the two four-port
valves are switched simultaneously, a second LC pump ceding the first dimension for on-line sample cleanup and
starts a gradient to elute the sample from RPLC column size-selective fractionation. The theoretical peak capacity
alpha, while the effluent from the first dimension is of the entire system is estimated to be as high as 3,000.
loaded onto RPLC column beta. The effluent from the Although this method is a powerful example of a fully au-
RPLC column being eluted is directed to a UV absor- tomated technique for proteomics separations, it involves
bance detector followed by a 10:1 flow splitter and an complex instrumentation with many switching valves and
electrospray mass spectrometer. The total peak capacity of LC pumps. An additional disadvantage is that on-line
this system was estimated to be about 500. Other 2D LC- coupling of the second-dimension RPLC columns to a
LC techniques using parallel columns in the second di- single electrospray mass spectrometer is not possible since
mension have been reported, many of which replace the the effluent from two columns must be monitored simul-
two four-port switching valves with a single ten-port valve taneously.
that accomplishes the same purpose [23, 24, 27, 31, 34].
Wagner and colleagues have described a novel com-
prehensive 2D anion-exchange reversed-phase system tar- Multidimensional liquid chromatography-capillary
geted at analysis of peptides and proteins of molecular electrophoresis (LC-CE)
weight below 20 kDa [26]. The most significant new fea-
ture of their system is the use of four parallel reversed- As has already been emphasized, the best multidimen-
phase columns in the second dimension. In order to inter- sional separations are achieved when techniques with very
face all four second-dimension columns to the single first- different separation mechanisms are coupled. Liquid chro-
dimension column, a set of three ten-port valves is used. matography-capillary electrophoresis (LC-CE) is there-
At any time during an analysis, effluent from the first di- fore a logical candidate for a 2D separation method, since
mension is being deposited onto one of the RPLC columns, separations in CE are based on electrophoretic mobility, a
two of the RPLC columns are being eluted, and one is be- mechanism nearly completely orthogonal to LC tech-
ing regenerated. This parallel elution allows the gradient niques. CE is a relatively fast separation method com-
time in the second dimension to be slowed down com- pared to LC, which makes it well suited as the second di-
pared to a system using only two reversed-phase columns mension in a 2D system because of its ability to sample
[24], which substantially enhances the peak capacity of the first dimension at a relatively high frequency. Al-
the second dimension. Two parallel UV detectors are used though LC and CE are complementary separation techniques,
to sample the system. This system also uses a novel re- their dissimilarity also makes coupling them somewhat
stricted access material (RAM) column positioned pre- complicated. Elution volumes of conventional LC columns
1958
Fig. 8 Schematic diagram of a LC-CE instrument based on the
transverse flow-gated interface (Adapted from Ref. [44])
are much greater than sample volumes of CE capillaries.
Therefore only a small fraction of the LC effluent can be
transferred to the CE dimension. Alternatively, the flow
from the LC dimension can be reduced by using capillary
LC columns [35]. Even when flow rates are relatively
well matched, the transfer of fractions from an LC column
to a CE capillary is usually more complex than in LC-LC
methods. As in LC-LC, off-line coupling is a simple op-
tion that has been used frequently [36, 37], but it is slower
than on-line methods and usually sacrifices resolution
from the first dimension. In order to take full advantage of
both dimensions, on-line interfacing techniques must be
employed. In spite of the difficulties, on-line LC-CE sep-
arations have been successfully performed and the tech-
nique has been shown to be a powerful multidimensional
separation method.
Since LC-CE is not yet commonly used, the number of
variations of the technique and the range of its applica-
tions are somewhat limited compared to LC-LC. To the
present, most work with comprehensive LC-CE systems
has focused on the separation of complex biological mix-
tures, particularly peptides and proteins, although some
work with smaller neutral molecules has been reported
[38]. The two modes of LC that have been coupled most
often with CE are reversed-phase chromatography [39,
40, 41, 42, 43, 44] and size-exclusion chromatography
[41, 45, 46]. Some other variants of CE have also been
used in 2D LC-CE separations, such as pressurized capil-
lary electrochromatography (pCEC) [47]. Due to the speed
and high resolution of CE, LC-CE offers substantial promise
in terms of separation power. The peak capacity of one
RPLC-CE system was estimated as 20,000 for a 5-h sepa-
ration [35].
LC-CE instrumentation
The first automated comprehensive LC-CE system was
developed by Bushey and Jorgenson in 1990 [39]. It was
based on the coupling of a commercial microbore RPLC
column with a fused-silica capillary using an electrically
actuated six-port valve with an external sample loop. This
system was used to separate mixtures of peptide standards
and tryptic digests of ovalbumin with fluorescence detec-
tion. The peak capacity of the original system was esti-
mated as 420, which was an order of magnitude improve-
ment over either of the two dimensions alone. An improved
version of this system was also devised which, among
other modifications, used a CE capillary with a smaller in-
ner diameter to allow higher electric field strengths to be
applied, thereby resulting in faster separations [40].
Many improvements in LC-CE instrumentation were
made during the 1990s. Enhanced separation efficiency was
achieved by using packed fused-silica capillaries for the
LC dimension as opposed to conventional LC columns
[46]. This also facilitated coupling the two dimensions be-
cause flow rates in capillary LC are more similar to those
typically used in CE. Other improvements in instrumenta-
tion focused on the interface used to couple LC columns
and CE capillaries. The transverse flow-gated interface,
first reported in 1993 [45] and improved in 1997 [44],
eliminates the need for sample loops that contribute to ex-

Fig. 9 2D chromatoelectropherogram from a micro-RPLC-CZE
separation of FITC-tagged human urine (Reproduced with permis-
sion from Ref. [44])
tra-column broadening. In a LC-CE system using this in-
terface, the LC column outlet and CE capillary inlet are
positioned directly opposite to one another, separated by a
narrow channel, as shown in Fig. 8. During most of a run,
CE buffer is continuously flushed through the channel,
which carries the LC column effluent to waste. To inject a
portion of the LC column effluent onto the CE capillary,
the flow of flush buffer is stopped and some of the LC ef-
fluent is drawn onto the CE capillary by electromigration.
Fast, reproducible injections can be performed by using
an air-actuated switching valve to control the flow of
flush buffer. A 2D chromatoelectropherogram of a sample
of fluorescently labeled human urine that was separated
using a transverse flow-gated LC-CE instrument is shown
in Fig. 9. In this chromatoelectropherogram over 400 peaks
are resolved and appear scattered randomly over the sepa-
ration space. This suggests that the two dimensions are or-
thogonal and that the system has a high total peak capac-
ity.
An even faster method of performing injections from a
LC column onto a CE capillary is provided by optical gat-
ing. In this “inverse” injection technique, a LC capillary
column and the CE capillary are coupled using a simple
interface tee, as shown in Fig. 10, so that some of the ef-
fluent of the first dimension is always being pulled onto
the CE capillary by electromigration [42]. An intense
1959
Fig. 10 Schematic diagram of an optically gated 2D RPLC-fast
CZE system (Adapted from Ref. [42])
laser beam is focused on the CE capillary at some point
beyond the interface tee; this beam photodegrades the fluo-
rescently labeled sample passing through the capillary.
When an injection is desired, a shutter is used to momen-
tarily block the laser beam, allowing a small amount of un-
degraded sample to be introduced onto the capillary. This
undegraded sample is then separated and detected. Al-
though the method is limited to fluorescently labeled sam-
ples, it offers the fastest injection method of any LC-CE
interface.
On-line coupling of LC-CE with MS represents a fur-
ther enhancement in LC-CE instrumentation [43, 47]. Nu-
merous challenges exist in terms of coupling CE with on-
line MS, however. One challenge is the selection of a CE
buffer that gives suitable separation performance and also
performs well when used with electrospray ionization.
A more fundamental difficulty in interfacing CE and MS
is the fact that most mass spectrometers are not capable of
sampling a CE capillary as rapidly as would be desired to
generate true 2D separations. Peak widths in CE separa-
tions are usually of the order of a second or less, which is
an inadequate period of time for most mass spectrometers
to acquire spectra over a reasonable m/z range [43]. As
more rapid mass spectrometers are produced, this problem
will be ameliorated. In spite of the difficulties, the use of
on-line MS with LC-CE still provides much additional in-
formation as compared to other detection methods.
Future directions for multidimensional separations
As for any area of research as diverse and wide-reaching
as multidimensional separations, it is difficult to predict
1960
the future directions of the field in the long term. Never-
theless, certain trends are evident within recent publica-
tions that may indicate where major research effort will be
focused in the immediate future. One possibility is the de-
velopment of separation methods that employ more than
two dimensions. Although research in this area has been
very limited due to the complexity of the instrumentation
required, three-dimensional separations using a SEC-
RPLC-CZE system have been reported [48]. As 2D sepa-
rations become faster and the instrumentation more robust,
incorporation of additional orthogonal separation methods
to perform n-dimensional separations will result in further
enhancement of peak capacity.
The coupling of multidimensional separations with on-
line MS will likely become more commonplace in future
years. MS provides more information than any other de-
tection method available for LC-LC or LC-CE separa-
tions. Indeed, the ability to obtain mass spectra of separated
components is essentially a prerequisite for any 2D sys-
tem intended for proteomics research. When tandem mass
spectrometry (MS-MS) is used, peptides can be fragmented
and sequence information can be obtained. Electrospray
ionization (ESI) allows the direct coupling of a LC col-
umn to a mass spectrometer. This gives LC-LC a dramatic
speed advantage over techniques such as 2D-PAGE, where
obtaining mass spectra involves extensive off-line sample
preparation. Given the versatility of MS, it is logical to
expect many future multidimensional separations to in-
corporate on-line MS as a detection method.
Another trend to be anticipated is the miniaturization
of multidimensional separation methods. In recent years
much attention has been given to the concept of perform-
ing chemical analyses on a very small scale using mi-
crofluidics devices – the so-called lab-on-a-chip [49]. Var-
ious separation methods using microchips have been re-
ported, including free-zone electrophoresis [50, 51, 52],
open-channel electrochromatography (OCEC) [53, 54],
gel electrophoresis [55, 56, 57], and micellar electroki-
netic chromatography (MEKC) [58, 59, 60]. Two-dimen-
sional separations using microchips have also been dem-
onstrated. In one system, a reversed-phase capillary LC
column was coupled to on-chip CE to produce a hybrid
off-chip/on-chip 2D separation method [61]. Ramsey and
colleagues reported another system in which MEKC was
coupled with open-channel electrophoresis on a single mi-
crofluidic device for a true on-chip 2D separation [62].
Due to their advantages over conventional 2D methods in
size and speed, on-chip multidimensional separations will
likely gain importance for certain separation tasks.
Regardless of the specific developments that take place,
the field of multidimensional separations seems poised for
substantial growth. There is no fundamental limit to the
number or type of separation methods that can be cou-
pled, so substantial increases in resolving power are pos-
sible. Although many 2D separation techniques being de-
veloped at present are targeted for use in the field of pro-
teomics, many other types of samples can be analyzed us-
ing similar methods. In practical terms, there is also much
room for the improvement of instrumentation such that
multidimensional separations can become a routine, reli-
able technique for analysis of real-world samples.
References
1. Regnier FE, Amini A, Chakraborty A, Geng M, Ji J, Riggs L,
Sioma C, Wang S, Zhang X (2001) LC GC N Am 19:200–213
2. Wehr T (2002) LC GC N Am 20:954–957
3. Giddings JC (1967) Anal Chem 39:1027–1028
4. Giddings JC (1987) J High Res Chromatogr 10:319-323
5. Davis JM, Giddings JC (1985) Anal Chem 57:2168–2177
6. Davis JM, Giddings JC (1985) Anal Chem 57:2178–2182
7. Giddings JC (1984) Anal Chem 56:1258A-1270A
8. Liu Z, Lee ML (2000) J Microcolumn Sep 12:241–254
9. Murphy RE, Schure MR, Foley JP (1998) Anal Chem 70:
1585–1594
10. O’Farrell PH (1975) J Biol Chem 250:4007–4021
11. Wehr T (2001) LC GC N Am 19:702–711
12. Cortes HJ (1990) Multidimensional chromatography: tech-
niques and applications. Marcel Dekker, New York
13. Bushey MM, Jorgenson JW (1990) Anal Chem 62:161–167
14. Link AJ, Eng J, Schieltz DM, Carmack E, Mize GJ, Morris
DR, Garvik BM, Yates JR III (1999) Nat Biotechnol 17:676–
682
15. Washburn MP, Wolters DA, Yates JR III (2001) Nat Biotech-
nol 19:242–247
16. Wolters DA, Washburn MP, Yates JR III (2001) Anal Chem
73:5683–5690
17. Washburn MP, Ulaszek R, Deciu C, Schieltz DM, Yates JR III
(2002) Anal Chem 74:1650–1657
18. McDonald WH, Ohi R, Miyamoto DT, Mitchison TJ, Yates JR
III (2002) Int J Mass Spectrom 219:245–251
19. Wu CC, MacCoss MJ, Howell KE, Yates JR III (2003) Nat
Biotechnol 21:532–538
20. Washburn MP, Ulaszek RR, Yates JR III (2003) Anal Chem
75:5054–5061
21. Holland LA, Jorgenson JW (1995) Anal Chem 67:3275–3283
22. Opiteck GJ, Lewis KC, Jorgenson JW, Anderegg RJ (1997)
Anal Chem 69:1518–1524
23. Wagner K, Racaityte K, Unger KK, Miliotis T, Edholm LE,
Bischoff R, Marko-Varga G (2000) J Chromatogr A 893:
293–305
24. Unger KK, Racaityte K, Wagner K, Miliotis T, Edholm LE,
Bischoff R, Marko-Varga G (2000) J High Res Chromatogr
23:259–265
25. Holland LA, Jorgenson JW (2000) J Microcolumn Sep 12:
371–377
26. Wagner K, Miliotis T, Marko-Varga G, Bischoff R, Unger KK
(2002) Anal Chem 74:809–820
27. Haefliger OP (2003) Anal Chem 75:371–378
28. Erni F, Frei RW (1978) J Chromatogr 149:561–569
29. Opiteck GJ, Jorgenson JW, Anderegg RJ (1997) Anal Chem
69:2283–2291
30. Opiteck GJ, Jorgenson JW, Moseley MA III, Anderegg RJ
(1998) J Microcolumn Sep 10:365–375
31. Opiteck GJ, Ramirez SM, Jorgenson JW, Moseley MA III
(1998) Anal Biochem 258:349–361
32. Stroink T, Wiese G, Lingeman H, Bult A, Underberg WJM
(2001) Anal Chim Acta 444:193–203
33. Murphy RE, Schure MR, Foley JP (1998) Anal Chem 70:
4353–4360
34. Murahashi T (2003) Analyst 128:611–615
35. Jeffery DJ, Hooker TF, Jorgenson JW (1997) Two-dimensional
liquid chromatography-capillary electrophoresis. In: Handbook
of capillary electrophoresis, 2nd edn. CRC, Boca Raton, pp 765–
790
36. Issaq HJ, Chan KC, Liu C, Li Q (2001) Electrophoresis 22:
1133–1135
37. Issaq HJ, Chan KC, Janini GM, Muschik GM (1999) Electro-
phoresis 20:1533–1537

38. Zhang X, Hu H-l, Xu S, Yang X, Zhang J (2001) J Sep Sci 24:
385–391
39. Bushey MM, Jorgenson JW (1990) Anal Chem 62:978–984
40. Bushey MM, Jorgenson JW (1990) J Microcolumn Sep 2:293–
299
41. Larmann JP Jr, Lemmo AV, Moore AW Jr, Jorgenson JW
(1993) Electrophoresis 14:439–447
42. Moore AW, Jorgenson JW (1995) Anal Chem 67:3448–3455
43. Lewis KC, Opiteck GJ, Jorgenson JW, Sheeley DM (1997)
J Am Soc Mass Spectr 8:495–500
44. Hooker TF, Jorgenson JW (1997) Anal Chem 69:4124–4142
45. Lemmo AV, Jorgenson JW (1993) Anal Chem 65:1576–1581
46. Lemmo AV, Jorgenson JW (1993) J Chromatogr 633:213–220
47. Stahl M, Jakob A, von Brocke A, Nicholson G, Bayer E (2002)
Electrophoresis 23:2949–2962
48. Moore AW, Jorgenson JW (1995) Anal Chem 67:3456–3463
49. de Mello AJ (2002) Anal Bioanal Chem 372:12–13
50. Effenhauser CS, Manz A, Widmer HM (1993) Anal Chem 65:
2637–2642
51. Harrison DJ, Fluri K, Seiler K, Fan Z, Effenhauser CS, Manz A
(1993) Science 261:895–897
1961
52. Jacobson SC, Hergenroder R, Koutny LB, Warmack RJ, Ram-
sey JM (1994) Anal Chem 66:1107–1113
53. Jacobson SC, Hergenroder R, Koutny LB, Ramsey JM (1994)
Anal Chem 66:2369–2373
54. Kutter JP, Jacobson SC, Matsubara N, Ramsey JM (1998) Anal
Chem 70:3291–3297
55. Effenhauser CS, Paulus A, Manz A, Widmer HM (1994) Anal
Chem 66:2949–2953
56. Woolley AT, Mathies RA (1994) Proc Natl Acad Sci USA
91:11348–11352
57. Jacobson SC, Ramsey JM (1996) Anal Chem 68:720–723
58. Moore AW, Jacobson SC, Ramsey JM (1995) Anal Chem 67:
4184-4189
59. Kutter JP, Jacobson SC, Ramsey JM (1997) Anal Chem 69:
5165–5171
60. von Heeren F, Verpoorte E, Manz A, Thormann W (1998)
Anal Chem 68:2044–2053
61. Yang Z, Zhang X, Li A, Zhu S, Huang Y (2003) Electrophore-
sis 24:1451–1457
62. Rocklin RD, Ramsey RS, Ramsey JM (2000) Anal Chem 72:
5244–5249