496
zyxwvuts
zyxwvutsrqpon
Anal. Chem. 1991, 63, 496-502
(6) Moscowitz, A. Adv. Chem. Fhys. 1962, 4 , 67-112.
(7) Djerassi, C. Optical Rotatory Dispersion; McGraw-Hill: New York,
1960.
(8) Stephens, P. J. Annu. Rev. fhys. Chem. 1974, 25, 201-232.
(9) Djerassi, C.; Bunnenberg, E.; Elder, D. L. Pure Appl. Chem. 1971, 25,
57-90.
(10) Shashova, V. E. J. Am. Chem. SOC. 1960,8 2 , 5505-5506.
(11) Pidgeon. C. R.; Smith, S. D. Infrared fhys. 1964, 4 , 13-28.
(12) Bloembergen, N.: Shen, Y. R. J . Opt. SOC.Am. 1964, 54, 551-552.
(13) Church, D. A,: Hadeishi, T. Appl. fhys. Le??. 1973, 2 4 , 185-187.
(14) Ito.. M.; Murayama. K.; Kayama, K.: Yamamoto. M. Spectrochim.
Acta, Part B 1977, 328, 347-355.
(15) Kitagawa, K.: Shigeyasu, T.; Takeuchi. T. Ana/yst 1978, 103,
1021-1030.
(16) Kersey, A. D.; Dawson, J. B. Anal. R o c . (London) 1981, 18,
187- 189.
(17) Debus, H.; Hanle, W.; Scharmann, A.; Witz, P. Spectrochim. Acta,
far? B 1981, 368, 1015-1021.
(18) Yamamoto, M.; Murayama, S.; Ito, M.; Yasuda, M. Spectrochim.
Acta, Part B 1980, 358, 43-50.
(19) Kawazumi, H.; Nishimura, H.: Ogawa, T. Anal. Sci. 1990, 6 ,
135-136.
(20) Partington, J. R. Advanced Treatise on Fhysical Chemistry;Longmans,
Green & Co.: London, 1954: p 602.
(21) Hougen, J. T. J . Chem. fhys . 1960,32, 1122-1 125.
(22) Warlng, C. E.: Hyman, H.; Steingiser, S. J. Am. Chem. Soc. 1940,
62, 2028-2031.
(23) Broersma, S.;Waterman, H. I.; Westerdijk, J. B.; Wiersma, E. C.
fhysica 1943, IO, 97-129.
(24) Washburn, E. W. International Critical Tables; McGraw-Hill: New
York. 1926; Vol. V I , p. 425.
(25) Schronrock. V. 0.Z . fhys. 1932, 78, 707-721.
(26) Chuckerbutti, B. N. Indlan J. fhys. 1932, 8 , 387-395.
(27) Yeung, E. S., Steenhoek, L. E.; Woodruff, S. D.; Kuo, J. C. Anal.
Chem. 1980,52, 1399-1402.
(28) Xi, X.: Yeung, E. S. Appl. Spectrosc. 1989, 43, 1337-1341.
(29) Montgomery, D. B. I n SolenoidMagnet Design; Krieger, R. E., Ed.: R.
E. K. Publishing Co.: Huntington, NY, 1980; p 6.
(30) Bobbitt, D. R.; Yeung. E. S. Appl. Spectrosc. 1986, 40, 407-409.
(31) Monnig, C. A.; Madison, R. T.; Helftje, G. M. Appl. Spectrosc. 1990,
4 4 , 216-219.
(32) Monnig, C. A.; Heiftje, G. M. J. Anal. At. Spectrosc. 1988, 3 ,
679-682.
(33) Oldenbourg, R.; Phillips, W. C. Rev. Sci. Instrum. 1986, 57,
2362-2365.
(34) Yeung, E. S. Talanta 1985, 32, 1097-1100.
(35) Stephen, M. J. Mol. fhys. 1958, 1 , 301-304.
(36) Partington, J. R. Advanced Treatise on Physical Chemistry; Longmans,
Green & Co.: London, 1954: p 607.
(37) Foehr. E. G.; Fenske, M. R. Ind. Eng. Chem. 1949, 4 1 , 1956-1966.
(38) Dovichi, N. J.; Harris, J. M. Anal. Chem. 1979, 51, 728-731.
(39) Woodruff, S. D.; Yeung, E. S. Anal. Chem. 1982, 5 4 , 2124-2125.
(40) Forsythe, J. G.; Kieselbach, R.; Shashoua, V. E. Appi. fhys. 1967, 6 ,
699-702.
zyxwvutsrqp
zyxwvutsrqponm
zyxwvutsrqp
RECEIVED
for review September 4,1990. Accepted November
30, 1990. The Ames Laboratory is operated by Iowa State
University for the U.S. Department of Energy under Contract
W-7405-Eng-82. This work was supported by the Director
of Energy Research, Office of Basic Energy Sciences, Division
of Chemical Sciences.
zyxwvut
Fluorescence Detection in Capillary Zone Electrophoresis Using
a Charge-Coupled Device with Time-Delayed Integration
J. V. Sweedler, J. B. Shear, H. A. Fishman, and R. N. Zare*
Department of Chemistry, Stanford University, Stanford, California 94305
R. H. Scheller
Howard Hughes Medical Institute, Department of Molecular and Cellular Physiology, Stanford University, Stanford,
California 94305
A fluorescence detection system for caplllary zone electrophoresis Is descrlbed In which a charged-coupled device
(CCD) vlews a 2-cm sectlon of an axially Illuminated capillary
column. The CCD Is operated In two readout modes: a
snapshot mode that acquires a series of Images in wavelength
and capillary posltion, and a time-delayed lntegratlon mode
that allows long exposure times of the movlng analyte zones.
By use of the latter mode, the aMllty to differentiate a species
based on both Its fluorescence emlsslon and mlgratlon rate
Is demonstrated for fluorescein and sulforhodamlne 101. The
detectlon limit for fluoresceln isothlocyanate (FITC) Is 1.2 X
mol; detection limits for FIX-amino acids are in the
(2-8) X
mol range.
INTRODUCTION
Numerous research areas in biochemistry depend on the
ability to analyze minute quantities of nucleic acids, amino
acids, and peptides, yet many present detection schemes have
inadequate sensitivity. Examples are the analyses of single
cells and subcellular compartments. One of the few techniques
to permit successful assays of the contents of a single cell (1-4)
is capillary zone electrophoresis (CZE), a powerful separation
technique for the analysis of small sample volumes. Sepa-
ration efficiencies routinely exceed several hundred thousand
theoretical plates, and typical injection volumes are 10 nL or
less. Several recent reviews describe the capabilities and
performance of CZE in detail (5-9).
For CZE, channel diameters usually range from 25 to 100
gm; therefore, designing methods to detect low concentrations
in these small-diameter capillaries is a challenge. Laser-induced fluorescence (LIF) is currently the most sensitive detection method for CZE; detection limits are in the low attomole range (10-12). In these systems, the capillary is used
as the “flow cell”, the laser illumination is perpendicular to
the capillary, and a photomultiplier tube (PMT) monitors the
fluorescence. Dovichi and co-workers (13-15) developed a
more sensitive method that has the same excitation geometry
but uses a sheath flow cuvette as the sample cell and thereby
eliminates much of the scattered light and luminescence from
the fused silica capillary. With this technique, detection limits
in the low zeptomole range for fluorescently tagged amino
acids have been reported. [Zepto (z) and yocto (y) have been
proposed as the new SI prefixes for
and
respectively, by the Comit6 Consultatif des Unit&; final approval
is pending the next meeting of the Comit6 GBn6ral des Poids
et Mesures.]
The LIF/CZE system described here uses a two-dimensional charge-coupled device (CCD) containing an array of 516
0003-2700/91/0363-0496$02.50/00 1991 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 63, NO. 5, MARCH 1, 1991
497
zyxwvutsrqponmlkjihgfed
+J
L -
Image Diream
zyxwvutsrqp
zyxwvutsrqpo
Figure 1. Diagram of the CCDILIF system illustrating the snapshot
mode. (a) The shutter is opened to expose the CCD to the fluorescence of two analyte bands. (b) After exposure, the shutter is closed
and the photogenerated charge information is read. (c) When the
shutter next opens, the analyte bands have traveled along the capillary,
and one band remains in the observation zone.
by 516 detector elements. Several reviews describe the advantages and operating characteristics of CCDs for fluorescence work (26-19). Although CCDs can have a lower dark
current and higher quantum efficiency than the best PMTs,
they suffer from a small but significant readout noise. An
important advantage of CCDs is their two-dimensional format;
CCDs are available in arrays ranging from 64 by 64 to more
than 2048 by 2048 elements. T o date, several reports have
described optical array detectors employed with CZE, including the use of a photodiode array as a multichannel UVvis absorbance detector (20,221 and a CCD as a multichannel
fluorescence detector (22).
We present a unique axial illumination arrangement for
CZE that has several advantages when used with multichannel
detectors. The output of a laser is focused into the end of
the capillary, the fluorescence emission from the analyte is
collected over a 2-cm section of the channel, and the entire
fluorescence spectrum is measured by using the CCD array.
In this way, the fluorescence cell is on-column, and the complete fluorescence spectrum is acquired simultaneously.
Residence times for analytes in the 2-cm observation zone
range from 2 to 45 s. The axial illumination method allows
the CCD to be operated in two readout modes and provides
significant advantages over conventional illumination and
CCD detection. These readout modes, the snapshot mode
and the time-delayed integration (TDI) mode, are described
below.
Snapshot Mode. Although the large format of many CCDs
offers considerable flexibility to the researcher, their slow
readout rate is often a drawback. Most scientific CCDs have
a readout rate of 50 kHz/element, requiring several seconds
to read out a large array. Consequently, CCDs are used
primarily with shutters to take sequences of snapshots (19).
After exposure of the detector to the fluorescence signal, the
shutter is closed and the photogenerated charge information
is digitized. The long delays between subsequent exposures
limit the usefulness of this readout mode in acquiring information from a changing scene, such as occurs in CZE.
In the one published report on CZE with a CCD, more than
5 s was required to read the CCD and transfer the data to the
host computer after each 0.2-s exposure (22). In CZE, peak
widths typically range from 0.5 to 5 s; thus, an entire analyte
band can be missed because the shutter is closed during the
time the band is in the observation zone. Even if some of the
sample is observed, quantitation becomes problematic because
the exposure time of the analyte to the CCD is not known.
Axial illumination eliminates the problems of missing bands
and incomplete analyte exposures. Figure 1 illustrates the
snapshot mode for an axially illuminated capillary by using
a simplified 3 by 6 element CCD. Although the analyte bands
migrate during the period that the shutter is closed, the second
band is still in the observation zone during the next exposure.
The choice of a 2-cm observation zone means that a t least 1
-
rime
zyxw
c
Figure 2. Dgram of the CCD/LIF system illustrating the time-dehyed
integration (TDI) mode. The CCD is oriented so that the parallel shift
direction is the same as the analyte band motion. As the emission of
the analyte bands moves across the CCD, the CCD shifts the resultant
photogenerated charge information at the same rate. When the
photogenerated charge from each band reaches the serial readout
register, the spectrum is read and digitized.
zyxwvu
but usually from 10 to 30 complete exposures of every analyte
are obtained. In addition, the snapshot mode is useful for
optical system diagnostics, as it allows tracking of individual
bands and real-time feedback for focusing the optics.
The low duty cycle of the detector (i.e., most of the time
the shutter is closed) is far from ideal. Moreover, the use of
a CCD to take a series of snapshots creates the need to extract
information from multiple exposures. With CCDs, the
problems with data accumulation are severe. For a 516 by
516 element CCD, such as the one employed in our work, each
complete CCD read generates 0.5 Mbytes of data. If, for
example, the CCD is read every 3 s, nearly 300 Mbytes of data
per CZE run is generated; extracting the fluorescence intensity
from this information is not trivial.
Time-Delayed Integration Mode. The TDI readout mode
was developed for satellite surveillance work when a CCD has
a limited time to integrate the signal from a fixed point on
the ground as a satellite passes overhead (23). It is also ideally
suited to LIF/CZE with axial illumination, in which the
analyte band passes through an extended detection zone. By
use of this technique, the CCD becomes a highly sensitive and
flexible multichannel fluorescence detection system. TDI does
not use a shutter, can acquire an entire spectrum in 50 ms,
and can spatially resolve multiple bands that are in the observation zone. Furthermore, far less data are generated than
in the snapshot mode.
To understand the TDI technique, a brief description of
CCD operation is necessary. In a CCD, all the photogenerated
charge in the photoactive elements is transferred toward the
serial register one row at a time, and the charge information
in the serial row is read by using the single on-chip amplifier
(16). For a 516 by 516 element CCD, each time a single
imaging area is transferred to the serial register, 516 readouts
are performed; each readout corresponds to a different spectral
element. This process continues until all 516 rows have been
read 516 times.
Normally these transfers are done with the shutter closed
to prevent exposure of the CCD to the illumination source;
if such exposure occurs, the image becomes blurred. In our
work, we eliminate the shutter and synchronize the shifting
of rows to the migration rate of the analyte band in our capillary. Figure 2 illustrates the TDI mode by using a simplified
3 by 4 element CCD and two analyte bands. As an analyte
band enters the laser excitation zone, the fluorescence is
collected and illuminates the first row of the CCD. From here,
the band takes a period of time to migrate to the point on the
zyxwvut
498
ANALYTICAL CHEMISTRY, VOL. 63, NO. 5, MARCH 1, 1991
capillary from which its image is projected onto the CCD one
row closer to the serial register. After this time, the charge
in the CCD is shifted toward the serial register by one row
so that the analyte fluorescence signal still contributes to the
same charge information. I t is important to distinguish between the physical rows of the CCD and the continually
moving row of accumulating photogenerated charge. The
effective integration time for a given analyte band is the entire
time the band is in the observation zone; however, multiple
bands can be within this zone simultaneously and still be
resolved because of their different spatial positions.
The TDI mode has several important advantages. Only one
row of the CCD is read at a time, reducing the data produced
and the readout time up to 516-fold (depending on the spatial
binning factor). Also, a row contains the fluorescence from
a single analyte and not from a point on the capillary; the
fluorescence is integrated over the entire time the band is in
the observation zone. A third advantage is that the fluorescence information from the analyte band can be obtained from
a single CCD readout instead of the approximately 4-20
readouts required with the snapshot mode. The single readout
produces a 2- to 5-fold reduction in read noise.
In CZE, the TDI mode is complicated by the fact that the
analyte bands move at different rates because of their different
electrophoretic mobilities. Hence, the shift rate must be
continually decreased during the run to follow different bands.
The synchronization of the shift rate to the analyte velocity
needs to be accurate to avoid a loss in separation efficiency.
Fortunately, such synchronization is not difficult to obtain.
Because both the start time of the CZE run and the distance
from the injector to the center of the observation zone are
known, the shift time can be determined by
A simple method exists to preclude the need for a changing
shift rate and thus to avoid the potential blurring caused by
the TDI method. If the capillary is grounded just prior to
the observation zone (25, 26), the analyte bands would not
undergo electrophoresis while being detected and instead
would move a t the solution flow velocity. Because the separation efficiency obtained with the present method is sufficient
for a wide range of applications, these experiments have not
yet been undertaken.
Using the TDI mode, we can obtain CZE electropherograms
with over 700 000 theoretical plates. In addition, we obtain
sensitivities for fluorescent tags in the low zeptomole range
and can differentiate between multiple fluorophores based
on different migration times and spectral characteristics.
zyxwvutsrqp
Tshift
= ( TelapsedXobsNbin)
EXPERIMENTAL SECTION
Axial Illumination. To achieve high sensitivity with the CCD
system using either of these CCD readout modes, longer exposure
of the CCD to the analyte fluorescence is advantageous. In most
LIF/CZE systems, the laser beam diameter is focused to less than
50 pm and strikes the capillary perpendicular to the separation
channel; thus, the analyte is illuminated by the laser for only
several milliseconds. In the system described here, the capillary
is illuminated end-on, and the resultant fluorescence from a 2-cm
section is imaged onto the CCD. Fluorescence is collected during
the entire residence time of the analyte band in this section.
Problems could occur when axially illuminating a capillary that
contains a leading band at high concentration closely followed
by other bands. The leading band could absorb a significant
fraction of the channeled excitation light, thus reducing the
fluorescent signal from later bands also resident in the observation
zone. However, we do not observe such a shadowing effect at the
fluorophore concentrations used in this work.
A buffer-filled capillary does not fulfill the requirements of a
light pipe; the index of refraction of the aqueous center is lower
than that of the fused silica “cladding”. Thus, laser illumination
tends to travel in the fused silica wall rather than propagate
through the center of the capillary, unless care is taken in the
choice of laser-focusing lens and the alignment of the laser beam
with respect to the capillary. The focusing lens is chosen to prove
a beam waist just smaller than the capillary inside diameter. This
lens provides a large depth of field and hence an extended illumination zone. Even with these measures, illumination travels
only several centimeters in the capillary channel before significant
attenuation occurs.
An axially illuminated open tubular liquid chromatography
(OTLC) system has been described using capillaries of similar
dimensions, but in that case, the solvent was chosen to provide
a higher index than the quartz capillary material (27). For axial
illumination of the OTLC system, the laser beam must be
propagated through the length of the capillary. In our case,
however, the loss of light after several centimeters is desirable.
Many fluorophores, including fluorescein, photodegrade rapidly
under moderate-intensity illumination; thus, illumination of the
analyte band prior to the detection zone is deleterious. For a
water-filled fused silica capillary, light radiates from the aqueous
channel into the walls as the capillary bends (27,28). Thus, the
channel is bent sharply (90’ bend over 1 cm) before the 2-cm
observation zone. This bending prevents laser light from propagating much past the observation zone and thus greatly reduces
photodegradation of the fluorophore before detection. Additionally, the bend in the capillary prevents ambient room light
from propagating through the capillary to the detection region.
In conventional systems, when the laser power is high, photodegradation occurs but the fluorescence of the fluorophores is
still monitored. In our system, a 1-cm distance separates the bend
in the capillary and the beginning of the observation zone. Thus,
for 1 cm, or up to 20 s, the fluorescently tagged analyte is subject
to laser illumination without collection of the resultant fluorescence. T o limit premature photodegradation and a consequent
reduction in sensitivity, laser power must be extremely low.
Currently, we are developing an improved capillary holder that
will sharply reduce this dead zone distance. The low-intensity
illumination and the long observation time are expected to provide
the highest signal-to-noise ratios obtainable from a fluorophore
zyxwvutsrq
/ (NccdLcap)
(1)
where Tshittis the time until the next CCD shift,
is the
time since the start of the CZE run, X o b s is the length of the
observation zone, Nbinis the binning factor in the observation
(imaging) dimension, Nccdis the number of CCD elements the
capillary image illuminates, and Leap is the length of the capillary to the center of the observation zone. The integration
times range from 2 to 45 s for a 80-cm capillary, 2-cm observation zone, and analyte elution times from 1 to 30 min.
A CCD must shift charge from all rows in concert (Le., the
shift rate is the same for all rows). Because each row receives
a signal that originates from a different point in the capillary,
the shift rate must be incorrect for all but one specified row.
To visualize this concept, consider two analyte bands that are
resolved and are in the detection region simultaneously. They
must be moving a t slightly different velocities because they
are resolved; the CCD, which shifts at a single velocity, cannot
independently track both. In these experiments, in which we
have matched the shift rate to the center of the 2-cm observation zone a t the end of an 80-cm capillary, a f l %difference
between the zone velocity a t the observation boundaries and
TDI shift rate is expected. Thus, in the configuration we
employ, almost no increase in analyte zone width is expected.
Immediately after injection, the required time between
shifts is less than 50 ms. Because the system cannot shift this
fast, the tracking of this extremely fast moving (hypothetical)
band is not exact. Currently, the system shifts a t the maximum rate until the calculated shift rate is slower than 20 Hz
(for approximately 2 min), a t which point it shifts a t the
appropriate rate. Because the shift rate is always decreasing,
data are obtained with a nonuniform time resolution. although
the spatial resolution remains constant (the system images
every band with the same 80-pm resolution). Thus, the early
fast bands are sampled faster than later bands. This type of
sampling scheme fits well with constant resolution and data
acquisition per band ( 2 4 ) .
zyx
zyxwvu
zyxwvutsrqp
zyxw
zyxw
zyxwvutsrqpon
zyxwv
ANALYTICAL CHEMISTRY, VOL. 63, NO. 5, MARCH 1, 1991 499
Ar ton Laser
I
lnterlodted
Figure 3. Schematic diagram showing the Ar ion laser, capillary
arrangement, optics, and CCD detector.
PM512
CCD
Flgure 4. Detailed schematic of the optical system from Figure 3,
showing the axially illuminated capillary. The optics form a 250-pm
by 6-mm image of a 63-pm by 2-cm section of capillary on the entrance slit of the spectrograph.
susceptible to photodegradation (29).
An argon-ion laser (Model 164, Spectra-Physics, Mountain
View, CA) provides the laser illumination, and a 488-nm or 514-nm
laser-line interference filter (Oriel Corporation, Stratford, CT)
blocks all but the desired wavelength. For most studies, the laser
is operated on the lowest power light-stabilized mode (100 mW),
and several neutral density filters are used to reduce the power
illuminating the capillary to between 35 and 175 pW. These
illumination levels may be the lowest reported for LIF detection.
For all studies except the multiple fluorophore experiment, 488
nm is employed. A 100-mm focal length (fl) fused silica lens
(LQP007, Melles Griot, Irvine, CA) is used to focus the laser into
the 63-pm-i.d. capillary. The capillary alignment is controlled
by two linear translation stages and a two-axis tilt stage (Newport
Research Corp., Irvine, CA) to allow precise alignment with the
laser beam.
Detection System. The overall detection system is shown in
Figure 3, and a more detailed view of the optical system appears
in Figure 4. As shown in Figure 3, the optics form an image of
the capillary on the CCD, and a spectrograph disperses the image.
Hence, the two dimensions of the CCD array contain different
information; one contains the image of the capillaryand the other
contains wavelength information. For clarity, we define rows as
the dispersive dimension and columns as the capillary image
dimension. Thus, the fluorescence from one spot on the capillary
(or one analyte band at a particular time) is dispersed along a
row of the array. Because the requirements for the optics are
different in the two dimensions, cylindrical optics are used.
The optical system consists of five elements. It was designed
with the aid of Beam 3 Optical Ray Tracer (Stellar Software) to
maximize the light collection while minimizing aberrations. The
first element is a 20-cm fl cylindrical mirror (LCP 178, Melles
Griot), used instead of a lens to reduce chromatic and spherical
aberrations. Next are two achromats (used as the imaging elements) and two cylindrical lenses (LAO 288, LA0126, LCP155,
and LCN129, Melles Griot, respectively). To simplify the
alignment of the system, the two cylindrical lenses are held in
a single mount, the two achromats are in another mount, and the
slit of the spectrograph is fixed. The overall system is designed
to form an image of a 2.5-cm by 50-pm capillary on the 6- by
0.25-mm slit of the spectrograph. This system is F/1 in the
wavelength dimension and F / 12 in the imaging dimension; both
are matched to the F/3 of the spectrograph.
An aberration-corrected imaging spectrograph (CP200, Instruments SA, Edison, NJ) is used because it maintains a
point-bpoint relation between the image of the slit and the CCD
focal plane (30,32). By use of the CP200 equipped with a 133
grooves/mm grating, the field of view of the CCD is 370 nm; this
wavelength window is adjustable from 200 to 1000 nm. The
spectral bandwidth is approximately 0.72 nm/CCD element; the
resolution is 3.6 nm using a 100-pm slit and is 9 nm using a 250-pm
slit.
The detection system consists of a 516 by 516 element CCD
(PM512, Photometrics Ltd., Tucson, AZ), supplied with the
Metachrome I1 overcoating to extend the wavelength range; the
quantum efficiency is more than 10% from 200 to 900 nm, with
a maximum of more than 50% at 700 nm. The CCD is controlled
with CCD electronics and signal-processing modules (CE200 and
CC200, Photometrics Ltd.).
Characteristics that make PM512 CCD well suited to CZE
include its large format, high quantum efficiency, low dark current,
and low read noise. In addition, this CCD does not suffer from
the charge trapping and poor charge transfer efficiency of some
other scientific-grade CCDs (32,143). With the particular CCD
used, the overall system readout noise can be fewer than 6
electrons; at 600 nm, this read noise corresponds to fewer than
12 photons of noise per detector element. The CCD is contained
in a liquid-nitrogen-cooled cryostat to reduce the dark current.
When the system is operated at -125 "C, the measured dark
current is less than 2 electrons/min per detector element, an
insignificant value considering the background luminescence levels
found in typical "clean" solvents.
System Control. The system uses a unique set of CCD readout
methods. The Photometrics CC200 camera controller is a
multibus 68OOO computer system with a variety of CCD operations
and commands preprogrammed. Several additional commands
have been added to the command library by using FORTH. The
CC200 computer performs all CCD control and executes all
high-level commands to the CCD detector. The CC200 computer
is controlled with a high-speed IBM AT clone (System 220, Dell
Computer Systems) using ASYST version 3.1 (Asyst Software
Technologies, Rochester, NY) through a high-speed GPIB interface (PCBA, National Instruments, Austin, TX). considerable
effort has gone into designing computer code to optimize data
acquisition and system control. The data reduction from twodimensional spectral and spatial information to electropherogram
is performed in real time by data-reduction algorithms in both
computers. In addition, the Dell computer controls voltage and
monitors voltage, current, and other aspects of system performance; all spectra and electropherograms are stored and displayed
by using routines written in ASYST.
The CCD detector utilizes the read mode commonly referred
to as binning (26,18). Binning is an electronic method of changing
the effective detector element size, using the actual 20- by 20-pm
detector elements as building blocks to make binned elements
of any combination desired. Binning is accomplished by combining the photogenerated charge on-chip and then reading out
the information in this combined charge packet with a single read.
Because the binned information is read by using a single readout,
the overall read noise is lower than in systems in which the
individual detectors are read and their readouts summed in
computer memory. In our system, we routinely read out the CCD
in the wavelength dimension with 4-fold binning, thereby producing 128-point spectra. Because of the broad nature of molecular fluorescence spectra in solution, little information is lost
when using 4-fold binning; moreover, both the amount of data
generated and the effective read noise are reduced. Binning in
the image dimension also reduces the amount of data generated,
but in this case, spatial resolution is decreased.
As each row of the CCD is read, an entire fluorescence spectrum
is acquired. For several applications, observing complete spectra
is useful. For quantitative analysis, we extract the analyte intensity information by using the following method. First, a
background array is subtracted from each spectrum to remove
the electronic offset for each element. Next, the spectrum is
multiplied by an array that contains weighting factors for each
zyxwv
zy
zyxwvut
500
zyxwvutsrq
zyxwvutsrq
zyxwvutsrqpo
ANALYTICAL CHEMISTRY, VOL. 63,NO. 5, MARCH 1, 1991
wavelength. These weighting factors constitute a digital filter
that selects the optimum wavelengths for determining the
fluorescence intensity. Although sophisticated methods can be
used to optimize the extraction of information, we have not yet
taken such measures. For the present studies, the weighting
factors are set to 1for the wavelengths of maximum fluorescence
intensity, all other weighting factors are zeroed, and the resulting
intensity information is summed. For several studies, the resulting
electropherograms are digitally smoothed by using Blackman's
window for the convolution weights and a noise cutoff frequency
of 0.33 cycles/point (34). This procedure reduces the higher
frequency noise but has little effect on the peak width and height.
CZE Apparatus. The CZE portion of the system is similar
to that described previously (35-38). A 4-cm section of the polyimide from the 63-pm-i.d., 363-pm-0.d. capillary (Polymicro,
Phoenix, AZ) is burned off by using gentle heating in a flame.
Precautions must be taken to avoid damaging the fused silica
during this process. After the polyimide is removed, the end of
the capillary is carefully scored and broken to leave a surface as
flat as possible. Because the laser illumination is directly through
the end of the capillary, any imperfections contribute greatly to
scattered light. This end of the capillary is held at ground potential in a modified fluorescence quartz cuvette. The incident
laser illumination is focused through the cuvette containing the
buffer solution and into the capillary. The inlet end of the
capillary, held a t high positive voltage, is placed in a 4-mL vial
containing approximately 2-3 mL of electrolyte buffer solution.
The electrical circuit is completed by strips of Pt foil submersed
in each buffer reservoir. All separations employ a 20-kV potential
across the capillary. The current through the system is measured
by monitoring the voltage drop across a 10-kQ resistor at the
ground side of the capillary. The capillary inlet is contained in
an interlocked Plexiglas box, and the outlet, where detection
occurs, is held in a light-tight optical enclosure.
All samples are introduced by using gravity injection in which
the inlet of the capillary is removed from the buffer vial and placed
in an elevated sample vial. This injection method is chosen to
avoid the sampling bias associated with electrokinetic injection
(39). For these experiments, the height displacement is 5 cm and
the injection time is from 10 to 30 s, which introduces from 4 to
13 nL of sample. To prevent sample carryover from an injection,
we dip the inlet into a vial of water after each injection. In
addition, we have removed the polyimide from the first 5 mm of
the capillary.
Reagents. The water used to prepare solutions is freshly
purified (Ion-X Adsorber and Research Deionizer, Cole Parmer,
Chicago, IL; LD2AjDemineralizer and Mega-Pure Still, Corning
Glassworks, Corning, NY). The supporting electrolyte for all the
experiments is 50 mM borate (pH 9.0) prepared from reagentgrade sodium borate decahydrate and boric acid (Mallinckrodt).
Stock solutions of fluorescein (Sigma), sulforhodamine 101 (Exciton), and fluorescein isothiocyanate (FITC) isomer I (Aldrich)
are prepared a t approximately
M and used without further
purification. All low-concentration solutions are made daily by
serial dilutions from these stock solutions. All amino acids are
purchased from Sigma and used without further purification.
Reagent-grade sodium bicarbonate (Sigma) and acetone (Baker)
are used in the derivatization reactions described below.
Derivatization. Amino acids are derivatized with FITC isomer
I in a modified version of the procedure described by Kawauchi
et al. (40). Briefly, individual solutions of amino acids are prepared
at approximately 100 mM in 200 mM NaHCO, (pH 9.0). Approximately 10 mM FITC is dissolved in a 95% acetone, 5% water
solution that contains 0.001% pyridine (by volume) to catalyze
the reaction. Derivatization is initiated by the addition of 100
p L of FITC solution to 500 r L of amino acid solution. The
reaction is carried out in the dark a t 20 "C for approximately 8
h. FITC-amino acid solutions are maintained separately in
opaque vessels at 4 O C for up to 2 days and are mixed immediately
prior to separation.
Wavelength
Dimension
Figure 5. Snapshots (0.5sexposures) of the focal plane output for
a 4-nL injection of 8 X lo-'' M FITC: (a) taken after 13 min, (b) 4
s later, and (c) 6 s after (b). The Rayleigh and the main Raman bands
of water are visible along with the moving FITC band.
plane output of the CCD for a single 4-nL injection of 8 X
M FITC. T h e CCD is used in the conventional snapshot
mode, in which a shutter is placed between the capillary and
the spectrograph slit to time the exposure of the CCD. After
the shutter is closed, the CCD is read. Figure 5a is a 0.5-s
exposure taken after approximately 13 min, 5b is taken 4 s
later, and 5c, 6 s after 5b. The data in Figure 5 show the CCD
output corresponding to approximately 46MOO nm, less than
50% of the spectral information stored in the CCD. T o reduce
the data further, the CCD is read with 4 by 4 binning, the
resulting data are smoothed in the wavelength dimension, and
every second wavelength point and every second spatial line
are plotted. This procedure reduces the focal plane output
from 250 000 detector elements to the approximately 2000
points plotted. With 4 by 4 binning, each element corresponds
to an 80 by 80 pm area on the detector focal plane and a
320-pm length of capillary. Both the Rayleigh line and the
strongest Raman line are clearly visible along with the
fluorescence from FITC. Comparison of these images shows
the motion of the FITC band.
Figure 6 illustrates the ability of the TDI mode to track
an analyte. These electropherograms show a series of 4-nL
injections of 8 X
M FITC for a variety of shift rates. In
Figure 6a, the shift rate is that calculated by using eq 1;
electropherograms in Figure 6b and 6c are obtained with 10%
faster and 20% slower shift rates, respectively. For the optimum shift rate, the efficiency (N)is 400 000 theoretical
plates, but as the shift rate is changed, N decreases; however,
with a 10% error in shift rate, efficiencies of 250000 theoretical
plates are obtained. Thus, extremely accurate estimations
of the shift rate are not required for most purposes.
Multiple Fluorophores. To determine the flexibility of
the system to detect multiple fluorophores, we analyze a
mixture of two commonly used derivitizing agents. Two
changes in the detection system are made for this experiment:
the laser illumination is changed to 514 nm, and a 525-nm
zyxwvutsrqp
RESULTS AND DISCUSSION
System Performance. Figure 5 illustrates the ability of
the optical system to image the capillary accurately and of
the CCD/LIF system to measure the fluorescence from an
analyte undergoing electrophoresis. This figure shows the focal
zyx
zyxw
ANALYTICAL CHEMISTRY, VOL. 63, NO. 5, MARCH 1, 1991
2so*
1
.ooo+
501
Rayletgh
;
;
;
;
;
;
;
;
;
1
-.
250.
500
740
Wavelength (nm)
Figure 7. TDI electropherogram of sulforhodamine and fluorescein.
The increase in the Raman and Rayleigh lines as the run progresses
is caused by the accumulation of signal for a longer integration time
for each succeeding spectrum.
1
loo.
zyxwvutsrqponmlk
zyxwvu
zyxwvu
.ooo+
loo*
:
;
;
;
;
;
;
;
;
4
ci
zyxwvut
I
I -
.ooo
;
i2.
w.0
1
I (c)
*
U
.r(
R
U
6
w
C
1
;
;
;
;
13.
;
;
;
;
24
20.0
k4
i3.
l2.
1
mO.o
]
(d)
(c)
,000 13.
14.
1s
Figure 8. TDI electropherograms of FITC for the following concentrations: (a) 7 X lo-'' M, (b) 6 X lo-'' M, (c) 6 X 10-l2 M, and (d)
6X
M. The injection volume for (a), (b), and (c). is 4 nL, and
for (d), 13 nL. Fluorescence intensity is obtained by using a spectral
window between the Rayleigh and major Raman band. The sloping
base line is caused by a continually increasing integration time corresponding to a decreasing shift rate.
zyxwvutsrqp
tlmo (mlnutar)
Flgure 6. Electropherogramsillustrating the effect of dmnging the TDI
shift rate on electrophoretic efficiency for a series of 3amol FITC
injections. I n (a), the shift rate is optimized and the efficiency (N)is
400 000. I n (b), the shift rate is increased 10% and Nis 250 000. I n
(c), the shift rate is reduced 20 % from (a), and Nis 150000. The small
peak to the right of the FITC peak is a contaminant.
cutoff filter is inserted before the spectrograph slit to reduce
the Rayleigh scattering. Figure 7 shows the TDI electropherogram of 3 X lo3 M sulforhodamine 101 and 9 X lo4
M fluorescein. The CCD is binned 4-fold in the wavelength
dimension, and the resulting 128-point spectra are stored. To
aid visualization of the data, the CCD is read by using 16-fold
binning in the image dimension, thus reducing the number
of spectra from more than 25000 to less than 2000. Every
fourth spectrum collected between 9 and 14 min is plotted
in Figure 7. The spectra shown in this figure are plotted with
equal spacing; because of the nonlinear data-acquisition rate,
the time axis is nonlinear.
Figure 7 shows the spectra of the fluorophores, as well as
the 514-nm scattered light and two Raman bands from water.
The fluorescence emission of the sulforhodamine is to the red
of emission from fluorescein. This figure also illustrates another effect of the changing shift rate; because the data acquisition rate is continually slowing down, the background
(i.e., Rayleigh and Raman bands) is continually increasing in
intensity. This effect does not alter the sensitivity or utility
of this technique because both the analyte fluorescence and
the background emission increase a t the same rate with increasing integration time. If desired, however, each spectrum
can be divided by its integration time. The ability to detect
and differentiate multiple fluorophores is important to several
applications, notably in DNA sequencing when identifying
each of the four bases (10,41).
Sensitivity. Figure 8 illustrates the sensitivity of the
system with a series of electropherograms of FITC. Parts a-c
of Figure 8 represent 4-nL injections of 7 X 10-lo,6 X lo-",
and 6 X 10-l2M solutions, whereas part d represents a 13-nL
injection of a 6 X 10-l2 M solution. The limit of detection
(LOD) for a 4-nL injection is 3 X
M, or 12 zmol, calculated by using two times the peak-to-peak base-line noise. The
peak in Figure 8c represents approximately 14000 FITC
molecules. Increasing the injection to 13 nL does not improve
the mass detection limit but improves the LOD concentration
to 1 X 10-l2 M while retaining good efficiency.
502
zyxwvutsrq
150.
>
.
U
.
I
n
zyxwvutsr
zyxwvutsrqponm
ANALYTICAL CHEMISTRY, VOL. 63, NO. 5, MARCH 1, 1991
I
(IO) Nickerson, B.; Jorgenson, J. W. HRC&CC, J . Huh Resdut. Chromatogr.. Chromatogr. Commun. 1988. 1 1 , 533-534.
EI
U
U
C
5 0 . 0 4 ;
IO.
i
i
i
(1 1) Drossman, H.; Luckey, J. A,; Kastichka, A. J.; D'Cunhan. J.; Smith, L.
M. Anal. Chem. 1990,62, 900-903.
(12) Kuhr, W. G.; Yeung, E. S.Anal. Chem. 1988, 60, 2642-2646.
(13) Cheng, Y.-F.; Dovichi. N. J. Science 1988,242, 562-564.
(14) Wu, S.; Dovichi, N. J. J . Chromatogr. 1989,480, 141-155.
(15) Dovichi. N. J.; Cheng, Y.-F. Am. Biotechnol. Lab. 1989, 7 , 10-14.
(16) Sweedler, J. V.; Bilhorn, R. B.; Epperson. P. M.; Sims, G. R.; Denton.
M. 8. Anal. Chem. 1988960, 282A-291A.
(17) Epperson, P. M.; Sweedler, J. V.; Bilhorn, R. B.; Sims, G. R.; Denton,
G. R. Anal. Chem. 1988. 6 0 , 327A-335A.
(18) Epperson, P. M.; Denton, M. B. Anal. Chem. 1889, 6 1 , 1513-1519.
(19) Epperson. P. M.; Jalkian, R. D.; Denton, M. B. Anal. Chem. 1989,6 1 ,
282-285.
(20) Koyashi, S.; Veda. T.; Kikumoto, M. J . Chromatogr. 1989, 480,
179-184.
(21) Brownlee. R. G.; Compton, S. W. Am. Lab. (Fairfield, Conn.) 1988,
20, 10-17.
(22) Cheng, Y.-F.; Piccard, R. D.; VoDinh, T. Appl. Spectrosc. 1990, 44,
755-765.
(23) Appendix E, Photometric Ltd. CC200 Operating Manual, Photometrics
Ltd., 1989, Tucson, AZ.
(24) Huang, X.; Coleman, W. F.; Zare, R. N. J . Chromatogr. 1989,480,
a K - i i .".
n
(25) Huang. X.; Zare. R . N. Anal. Chem. 1990,6 2 , 443-446.
(26) Wallingford. R . A.; Ewing, A. G. Anal. Chem. 1987,59, 1762-1766.
(27) Xi, X.; Yeung, E. S.Anal. Chem. 1990, 62, 1580-1585.
(28) Lei, W.; Fujiwara, K.; Fuwa, K. Anal. Chem. 1983,55, 951-955.
(29) Mathies, R. A.; Peck, K.; Stryer, L. Anal. Chem. 1990, 6 2 ,
1786-1791.
(30) Jalkian. R. D.; Pomeroy, R. S.; Kolczynski, J. D.; Denton, M. B.; Lerner, J. M.; Grayzel, R. Am. Lab. (Fairfield, Conn.) 1989, 2 1 , 80-88.
(31) Kolczynski, J. D.; Pomeroy, R. S.; Jalkian, R. D.; Denton, M. B. Appl.
Spectrosc. 1989,43, 887-891.
(32) Bredthauer, R. A.; Chandler, C. E.; Janesick, J. R.; McCurnin, I.W.;
Sims, G. R. I n Instrumentation for Ground-Based Optical Astronomy;
Robinson, L. B., Ed.; Springer-Veriag: New York, 1987; p 468.
(33) Pemberton, J. E.; Sobocinski. R. L.; Sims, G. R. Appl. Spectrosc.
1990,44, 328-330.
(34) Blackman, R. B.; Tukey, J. W. In The Measurement of Power Spectra; Dover: New York, 1958.
(35) Gassmann, E.; KUO,J. E.; Zare, R. N. Science 1985,230, 813-814.
(36) Pentoney, S. L., Jr.; Huang, X.; Burgi, D. S.; Zare, R. N. Anal. Chem.
1988,60,2625-2629.
(37) Huang. X.; Pang, T.-K.; Gordon, M. J.; Zare, R. N. Anal. Chem. 1987,
59, 2747-2749.
(38) Huang, X.; Gordon, M. J . ; Zare, R. N. Anal. Chem. 1988, 6 0 ,
1837- 1838.
(391
. . Huana. X.; Gordon. M. J . : Zare. R. N. Anal. Chem. 1988. 6 0 .
375-577.
(40) Kawauchi, H.; Tuzimura, K.; Maeda, H.; Ishida, N. J . Biochem. 1969,
66, 783-789.
(41) Prober, J. M.; Trainor, G. L.; Dam, R. J.; Hobbs, F. W.; Robertson, C.
R.; Zagursky, R. J.; Couzza, A. J.; Jensen, M. A,; Baumeister, K. Science 1987, 238, 336.
zyxwvu
zyxwvutsrqponmlkjihgfed
zyxwvut
i
15.
i
i
i
I
20
Time (minutes)
Figure 9. TDI electropherogram of four FITC-amino acids, with the
concentration of injected FITC-arginine, FITC-valine, and FITCglycine at 7 X lo-" M and FITC-glutamate at 8 X lo-" M. The
fluorescence intensity is obtained by using a spectral window between
the Rayleigh and major Raman band. The sloping base line is caused
by a continually increasing integration time corresponding to a decreasing shift rate.
Figure 9 shows an electropherogram of four FITC-amino
acids: arginine, valine, and glycine a t 7 x lo-" M, and glutamate a t 8 X lo-" M. The injection volume is 13 nL. The
LODs for the FITC-amino acids are in the 20-80 zmol range,
which compares favorably to the best previous work. Moreover, further improvements in the optical system, faster imaging spectrographs, optimized fluorophores, and optimized
separation conditions are expected to push the sensitivity into
the high yoctomole range in the near future.
ACKNOWLEDGMENT
We gratefully acknowledge the helpful advice and assistance
of Stephen L. Pentoney, Jr., Xiaohua Huang, and Manny J.
Gordon.
LITERATURE CITED
Olefirowicz, T. M.; Ewing, A. G. Anal. Chem. 1990,62, 1872-1876.
Wallingford, R, A,;
A , G, A
~ Chem,
~ 1988,
~
, 60, 1972-1975,
Chien, J. B.; Wallingford, R. A.; Ewing, A. G. J . Neurochem. 1990,5 4 ,
633-638.
... ~ . .
Kennedy, R. T.; Oates, M. D.; Cooper, B. R.; Nickerson. B.; Jorgenson,
J. W. Science 1989. 246, 57-63.
Jorgenson, J. W.; Lukacs. K. D. Science 1983, 222, 266-272.
Gordon, M. J.; Huang, X.; Pentoney, S. L., Jr.; Zare. R. N. Science
1988,242, 224-228.
Ewing, A. G.; Wallingford, R. A,; Oleferowicz. T. M. Anal. Chem.
1989, 6 1 , 292A-303A.
Karger, B. L.; Cohen, A. S.; Guttman, A. J. Cbromatogr. 1989,480,
35-67.
Roach, M,:
p,; zare, R, N, J , Chromatogf, 1988, 426,
129- 140.
.
RECEIVED
for review September 24,1990. Accepted November
30, 1990. J.V.S. thanks the National Science Foundation for
a postdoctoral fellowship administered under NSF Grant
CHE-8907446. J.B.S. is a Howard Hughes Medical Institute
Predoctoral Fellow. The support of Beckman Instruments,
Inc., and of the National Institute of Mental Health under
Grant No. MH45423 is gratefully acknowledged.