Anal. Chem. 1995, 67, 3448-3455
Rapid Comprehensive Two=Dimensional Separations of Peptides via RPLC-Optically Gated Capillary Zone Electrophoresis Alvin W. Moore, Jr., and James W. Jorgenson* CB 3290, Department of Chemistry, University of North CarolineChapel Hill, Chapel Hill, North Carolina 27599
Coupled-columntwo dimensional (2D)separation systems offer potentially high peak capacity and are amenable to automation. Reversed phase liquid chromatography (RPLC) and capillary zone electrophoresis (CZE) are complementary techniques well suited for use in a 2D system. Optically gated CZE is a means of performing rapid CZE analyses. In a 2D system, these rapid CZE analyses enable more frequent sampling of the RPLC separation and thus more freedom in control of the RPLC analysis conditions. Complete 2D separations can be done in the time usually required to do the RPLC analysis alone. With the present system, a complete 2D analysis can be done in under 10 min. Because the peak capacity of a 2D method is inherently high, some of the available peak capacity can be exchanged for speed of analysis. Acceleration of the RPLC elution gradient will decrease analysis time, but RPLC resolution will suffer. However, because it is a 2D system, some of the resolution lost in the RPLC can be regained in the second dimension CZE analysis. Here, the RPLC gradient was done over only 2 min, but CZE analyses done every 2.5 s restore some of the resolution lost in the rapid RPLC gradient Two-dimensional (2D) separation methods are of interest because of the potentially high peak capacity possible with such methods. In particular, a coupled-column approach is appealing because of the possibilities of automation and on-line detection. Giddings has shown that the peak capacity of a 2D method is the product of the peak capacities of its component one-dimensional (1D) methods.’ This is true if the component separation methods are orthogonal. Two methods are considered orthogonal if their selectivities are based on different and uncorrelated chemical or physical characteristics of the molecules of interest. Reversed phase HPLC and capillary zone electrophoresis (CZE) are well suited for use in a 2D system. Their operating buffers are compatible, while their separating mechanisms are very different. In the case of reversed phase LC (RpLC) and CZE, separations depend on component hydrophobicity in the former and electrophoretic mobility (mass-tocharge ratio) in the latter. The orthogonality of these mechanisms enables many of the potential gains from 2D operation to be realized in practice. A common example of multidimensional coupled-column separation is “heartcutting”? in which particular regions of interest (1) Giddings,J. C. HRC CC, J. High Resolut. Chromatogr. Chromatogr. Commun. 1987,10,319-323. ( 2 ) Majors, R. E. J. Chromatogr. Sci. 1980,18, 571-579.
3448 Analytical Chemistry, Vol. 67, No. 19, October 1, 1995
from a separation done in one column are automatically reinjected onto a second column of different chemistry. Another method of 2D separation involves manually collecting fractions of effluent from one separation method and reinjecting them into a second separation system. Heart-cutting methods are of limited use if we desire to separate all of the components of the sample in both dimensions (a comprehensive 2D separation). Also, for a comprehensive method of this type, the manual fraction collection method is too labor-intensive and slow. We would like a continuous on-line system in which the entire effluent of the first dimension method is sampled into the second. In such a system, detection is usually done only after the second method. Detecting sample zones between the two 1D methods is of little practical utility and might contribute additional band broadening. If detection is done only after the second separation method, then the sampling of the separation in the first dimension is limited by the analysis time of the second dimension separation. Each entire analysis in the second dimension represents only a single “point” in the first. To make full use of the resolution provided in the first dimension separation, the second dimension must have a signiticantly faster analysis time. This places a limitation on most 2D systems and generally results in very long 2D analysis times. In CZE, unlike LC, separation efficiency is not a direct function of capillary length.3 In LC,with all other factors held constant, a column 20 cm long will give twice as many theoretical plates as a column 10 cm long. In contrast, in CZE, if the applied voltage remains constant, the separation efficiency will remain constant, independent of capillary length. The CZE efficiency is related to capillary length only indirectly, in that sufficient capillary surface area must be maintained to eliminate any Joule heat generated by the passage of current through the buffer-filled capillary. However, the migration time of solutes in CZE is a function of the capillary length squared. Assuming voltage is held constant, if capillary length is reduced by half, then the migration times of the analytes will be reduced by a factor of 4. It has previously been shown that the combination of short capillaries with high applied voltages will give very fast and efficient CZE analy~es.47~ Because of the speed of these analyses, a unique optical-gatinginjection system is used to obtain injections that are rapid relative to the CZE analysis time. The “fast-CZE” system makes possible complete CZE analyses in as little as 3 s while high separation efficiency is maintained. (3) Jorgenson, J. W.; Lukacs, K D.Anal. Chem. 1981,53, 1298. (4) Monnig, C. A; Jorgenson, J. W. Anal. Chem. 1993,63, 802-807. (5) Moore, A W., Jr.; Jorgenson, J. W. Anal. Chem. 1993,65, 3550-3560. 0003-2700/95/0367-3448$9.00/0 0 1995 American Chemical Society
The high speed of the fast-CZE system makes it particularly well-suited for use as a second dimension in a 2D system. Because the second dimension separation is very fast, there is more freedom in control of the first dimension separation. The LC analysis need not be made unusually slow to allow time for the CZE analysis. In fact, the opposite is true. Because 2D methods have inherently high peak capacity, if not all of the peak capacity of a 2D separation is needed, the unneeded peak capacity can be traded for greater overall speed of analysis. Because the second dimension does part of the “work of the separation, the first dimension can be run faster than normal. Thus, a separation that might be achieved in a 1D method in 1 h can be done with a 2D system in a few minutes, because peak capacity equivalent to that of a 1D system can be generated in a 2D system in a much shorter time. Fast-CZE. In brief, fast-CZE is a means of performing CZE analyses in seconds rather than minutes. Characteristic CZE highvoltage levels (5-25 kv) are applied to short lengths of capillary, which results in very short CZE analysis times. The capillary is only l@pm internal diameter (Ld.) and low-concentration CZE buffers (10 mM) are used. Together these limit the current through the capillary at the high applied field, so that Joule heating is not a signifcant problem. Because the fast-CZE analyses are done in seconds, a unique injection system is needed. Sample injections must be short relative to total analysis time, and mechanical fluidic valves are too slow for this purpose. A high-speed optical gating injection system was developed by Monnig and Jorgen~on.~ This system is based on an argon ion laser operated at 350 mW. Samples to be analyzed are tagged with fluorescein isothiocyanate (FITC) to be sensitive to the 488nm line of the laser. The beam from the laser is split into a gating beam, focused nearer the injection end of the capillary and containing 95%of the laser power, and a probe beam, focused nearer the exit end of the capillary and containing the remaining 5%of the laser power. The optical-gating injection may be thought of as an “inverse” injection method. A constant high voltage is applied to the capillary, and sample is migrated continuously through the capillary by a combination of electroosmosisand electrophoresis. As long as the gating beam is focused on the capillary, the majority of sample passing through the beam is photodegraded by the intense light of the gating beam. Only a residual background fluorescence is seen at the probe beam further along the capillary. To make an “injection”,the gating beam is momentarily blocked (usually for 5-50 ms) with a computer-controlled shutter. This allows a small slug of unbleached material to pass through into the region of the capillary between the beams, where its components are separated and detected by their fluorescence as they pass through the probe beam. Figure 1 is a timing diagram for the optical-gating injection. The upper trace represents laser power at the gating beam. The sharp spike downward is the point of sample injection, when the beam is momentarily blocked by the shutter (here, for 20 ms). The lower trace is the fluorescence signal seen at the probe beam for an actual fast-CZE run of tryptic digest of cytochrome c. A dip in the background fluorescence signal is seen at the point of sample injection, because the mechanical shutter used to block the gating beam actually blocks the laser beam before it is split and thus blocks both beams. This dip in background fluorescence serves as a convenient injection marker in each run. Seconds
0.0
I
0.5
I
I
1.o
1.5
CZE (
I
2.0
I
2.5
8 ~ )
Figure 1. Timing diagram and fast-CZE analysis of tryptic digest of horse heart cytochrome c injection time, 20 ms; 20 kV applied over total capillaty length of 8 cm, with 2 cm between gating and probe beams. Data were smoothed with a five point moving average filter to reduce baseline noise.
after the injection, peaks are seen in the fluorescence trace as sample components separated between the beams pass through the probe beam. EXPERIMENTAL SECTION Samples and Reagents. Deionized water for all sample solutions and buffers was generated from distilled water with a Barnstead Nanopure system. All other reagents were used as obtained. All aqueous buffers were prepared with 0.005%(w/v) sodium azide as an antimicrobial agent. Buffers and samples to be injected directly into the fast-CZE system were first filtered with 0.2-pm Acrodisc syringe filters (Gelman, Inc.). Methoxyarginine, horse heart cytochrome c, and trypsin were purchased from Sigma. Water-soluble sodium fluorescein was purchased from Aldrich. Fluorescein isothiocyanate was purchased from Molecular Probes (Eugene, OR) and Sigma. Acetone, pyridine, sodium phosphate, boric acid, and NaOH were purchased from Fisher Scientific. A solution of 0.01 M sodium phosphate, adjusted to pH 6.85 with NaOH, was used to optimize the operation of the CZE system. This buffer is hereafter referred to as CZE buffer. Digest and FTIX-TaggingConditions. A 10 mg/mL solution of horse heart cytochrome c was prepared in 0.1 M boric acid buffer, pH 8.4. To 3 mL of this solution was added 1mL of O.lmg/ mL trypsin solution in the same buffer, and digestion was allowed to proceed for 24 h at 37 “C. The digest was then filtered with a 0.2-pm syringe filter (Gelman, Inc.). A 5GpL aliquot of FITC in a 9O:lO (v/v) acetone/pyridine solution was added to 200 pL of filtered digest solution and allowed to react in darkness for 24 h at room temperature. The concentration of the FITC solution was adjusted to give a ?-fold molar excess of tryptic peptides over FITC in the sample. The final concentration of FlTGtagged components in this sample was 4.4 mM. Samples were diluted lwfold into CZE buffer before analysis by RPLC/fast-CZE. Methoxyarginine was tagged with FITC in a manner similar to that described above, with the methoxyarginine in sfold molar excess to the FITC tag. This procedure has been described in detail el~ewhere.~ The resultant fluorescein thiocarbamyl (FTC) derivative, FTC-methoxyarginine, was used as a neutral marker to monitor electroosmotic flow in CZE. Instrumentation. The fast-CZE instrument arrangement has been described in detail e l s e ~ h e r e .Figure ~ 2 is a diagram of the 2D RPLC/fast-CZE instrument. The fast-CZE part of the system Analytical Chemistry, Vol. 67, No. 19, October 1, 1995
3449
P
Close up of interface lee Figure 2. Two-dimensional RPLCIfast-CZE instrumental diagram. is shown in the left side of the figure. The CZE capillary is mounted vertically, but there is no significantgravitationally driven flow because the capillary is only 10pm id.. A stainless steel tee (750/rm i.d., Valco Instrument Co., Inc., Houston,?x) is connected to the top of the CZE capillary. The internal volume of the tee replaces the buffer resewoir normally seen in CZE, and the tee serves as a simple interface to the LC system . The tee is connected to ground, and negative high voltage is applied to the lower buffer reservoir so that electroosmotic flow is from top to bottom in the diagram. The positions of the gating and probe laser beams focused onto the CZE capillary are also shown. The laser is a 1-W Coherent Innova 70 argon ion laser (Coherent Inc., Palo Alto, CA) operated at 400 mW at a wavelength of 488 mn. The right side of Figure 2 shows the LC system, made up of a gradient-capable HPLC pump (Hewlett-Packard Model 1050), a sample injection valve (Rheodyne) with a 20pL sample loop, and an LC column. Also shown is a post-LC flnsh valve (Valco Instrument Co., Inc.) between the LC column and the fastCZE system and a shut-off valve (Alltech Associates Inc., Deerdeld, IL) on the waste line from the fast-CZE tee. Connecting tubmg between the outlet of the LC column and the flush valve is 0.005 in. (127-pm) i.d. PEEK tubing (Upchnrch Scientific, Oak Harbor, WA). The waste line on the fast-CZE tee is 0.040in. (1-mm) i.d. PEEK tubing (Upchurch). The waste line shut-off valve is used in the pretreatment of a new piece of fused-silica capillary before use in CZE. During 2D or fast-CZE analysis, this valve is open so that there is no significant back pressure on the waste line. When the waste valve is closed, the pump can be used to force liquids through the CZE capillary under pressure. The post-LL flush valve has a large (0.7 mL) sample loop which is filled using a 1-mL Glenco flushing syringe (Alltech Associates Inc.). It is used to send a slug of sample solution or base solution used in capillary pretreatment through the fast-CZE tee without going through the LC system. Separation Conditions. The RPLC column was a Vydac 15 cm x 2.1 mm i.d. protein and peptide C18 column, 5pm particle size, 3Wf pores Nest Group, Inc., Southborough, MA). Solvent A for RPLC was CZE buffer. Solvent B for RPLC was 60% acetonitrile (Fisher Optima)/40% CZE buffer (v/v). The RPLC flow rate was 250 pL/min, with a linear elution gradient from 10%B to 50% B over 5 min, hold 1min at 50%B, and then return to initial conditions. This gave a linear gradient from 6 to 30%acetonitrile over 5 min. The fast-CZE capillary was l0pm i.d., 350pm o.d., and 8 an long. The high voltage is applied across the entire length of the
me
3450 Analyiical Chemistry. Vol. 67,No. 19, October 1, 1995
capillary, hut the actual length available for CZE separation is less. Since the gating beam is the actual point of injection, and the probe beam is the point of detection, only the voltage dropped over that length of the capillary between the beams is actually available for the separation. Due to the design of the mount used to hold the capillary,for an 8cm length of capillary, the maximum beam separation is about 2 cm. Thus, for 20 kV applied with 2 cm between the beams, only 25%or 5 kV is actually used in the CZE separation. Similarlywith only 0.45 cm between the beams, only 5.6% or 1.1kV is actually used in the CZE separation. Data Acquisition. Two-dimensionalsystems inherently generate large amounts of data. For the LC/fast-CZE, because of the speed of the CZE analyses, that data must also be acquired rapidly. The data acquisition hardware PMT,current ampliier, etc.) is as described previously for the fast-CZE system.',5 For 2D analysis, data acquisition is done at 400 Hz. Aprogram written in IabVIEW2 (National Instruments, Austin, TX) running on a Macintosh I1 computer acquired data in doublebuffered mode through a multipurpose A/D interface board (Model N R M l 0 16X, National Instruments). In doublebuffered mode, data are written to disk in the background as they are acquired in the foreground. The 2D data file is displayed using SpyGIass Transform and Format color raster imaging software (Spyglass, Inc.). Transform can be used to view the data as a twcdimensional color or gray-scale map. Only gray-scale plots are shown here. In these plots, RPLC time in minutes is on the y-axis, fast-CZE time in seconds is on the X-axis.and fluorescence intensity of the sample peaks is shown by gray-scale intensity. Transform can also be used to generate three-dimensional surface plots of the 2D data set, similar to those shown in Figure 4. Plots of 1D data for individual CZE and LC analyses were done with IGOR plotting software (Wavemetrics Inc., lake Oswego, OR). Procedures. A new CZE capillary is prepared as follows. Eight centimeters of capillary is installed in the fast-CZE mount, and the polyimide coating is removed over about 4 cm using a heated nichrome wire device constructed in-house. The capillary mount is then installed into the system, and the fluid lines are connected. To fill the capillary, the LC column is replaced with a '/,sin. union, and the LC pump is used to flow buffer through the tee and out the waste line. Then the waste valve is closed, and the pump back pressure is allowed to rise to 1000 psi. The maximnm pressure safely applied to the tee is about loo0 psi, due to the constraints of the 1-mm i.d. PEEK (Upchurch Scientific) waste line. At loo0 psi, the pump is stopped but the pressure on the tee is not released. The hack pressure slowly bleeds off as buffer flows through the small bore of the CZE capillary. To pretreat the capillary, the flnsh valve loop is filled with 1 M NaOH solution and the valve is turned to the inject position. The waste valve is opened, and the base is pumped from the flush valve sample loop into the tee. Once the base is in the tee, the waste valve is closed and the pump is used to pressurize the tee as described above. In this way, 1 M NaOH solution is passed under pressure through the CZE capillary for 15 min. The waste valve is opened, the tee is flushed with CZE buffer, and the waste valve is closed again to pressurize the tee and flush the CZE capillary with buffer. This is done for 30 min, and the capillary is ready for use. Before beginning a 2D analysis, the RPLC column is reinstalled and the fast-CZE optics are aligned using a solution of 1 x 10-6 M sodium fluorescein in CZE buffer. The post-LC flush valve can
be used to fill the CZE capillary without pumping a large volume of fluorescein solution through the LC column. The sample loop of the flush valve is filled with fluorescein solution, the pump is set to 50 pL/min (no gradient), and the flush valve is turned to the inject position to send a constant flow of fresh fluorescein solution to the fast-CZE tee. The large sample loop and low flow rate allow fluorescein to be pumped through the tee for about 15 min. When the high voltage is on, a fraction of the fluorescein solution is continuously electromigrated into the CZE capillary. The optics that focus the laser beams can then be adjusted to give the maximum fluorescence signal. Fast-CZE analyses of this solution can also be done to check CZE system performance prior to a 2D analysis. Fast-CZE is well-suited for use in a 2D LC/CZE instrument. In a 2D analysis, sample is injected with the sample injection valve and separated on the LC column, usually by gradient elution. Effluent from the LC column flows into the tee of the fast-CZE system. As shown in the inset in Figure 2, this effluent flows across the end of the fast-CZE capillary and out to waste. Because the waste line is 1-mm id., there is no signilicant pressureinduced flow through the 10ym i.d. CZE capillary. As long as high voltage is applied to the capillary, a small fraction of the LC effluent is continuously electromigrated into the capillary. No automated switching valve is needed between the LC and fast-CZE systems, because the actual fast-CZE “injections”from the LC effluent are done with the gating beam. Fast-CZE analyses can be done so rapidly that second dimension analysis time plays a much smaller role in determining the first dimension LC analysis conditions. LC Background Correction. The residual fluorescence background seen in fast-CZE is a result of the optical-gating injection method. When only CZE is being done, the background is constant for a given sample. However, in 2D analysis, the background level changes as peaks elute from the LC column. This makes interpretation of the 2D data dif6cult. To solve this problem, the original 2D data is digitally filtered with a moving median filter.‘j The moving median filter is similar to a moving average filter except that, instead of calculating the average of a group of data points, it calculates the statistical median. The moving median discriminates against sharp momentary changes in signal level but retains slow changes and edgelike characteristics in a signal. In the 2D LC/fast-CZE data, the CZE peaks are less than 100 ms wide, while the LC peaks are 5-20 s wide. Thus the CZE peaks are very sharp relative to the LC peaks. In fact, the CZE peaks appear as spikes “riding” on the LC peaks. Figure 3A shows 2D data the way it is actually collected, as one long array of data points, one CZE run after another. In these data, no LC column was present, but flow injection of a slug of fluorescein was done. The sharp spikes upward are the CZE peaks, while the spikes downward are the injection markers for each run. The broad “hill” is the change in background as the slug of fluorescein passes through the CZE tee. In Figure 3B, the data from (A) have been digitally filtered with a moving median filter of rank 32. The number of points in the filtering window is twice the rank plus 1,or 65 in this case. This removes the spikes due to the fast-CZE, leaving only the “LCpart” of the data. In Figure 3C, the filtered data in (B) have been subtracted from the raw data in (A). This removes the changing background due to the LC peak and leaves only the CZE peaks. (6)Moore, A W., Jr.; Jorgenson, J. W. Anal. Chen. 1993,65,188-191.
r 1 .o
1.5
2.0
2.5
2.0
2.5
Tlme (min)
1 .o
1.5
Time (mln)
I
I
I
-r
1.0
1.5
2.0
2.5
Time (mln)
Figure 3. RPLC/fast-CZE background correction. (A) shows the raw data as it is actually acquired, as a series of 1D CZE analyses. (B) is the data from (A) after median filtering with filter of rank 32 (window width, 65 points). (C) is the difference data, the result of subtracting the filtered data in (B) from the raw data in (A).
The same data in a three-dimensional surface plot are shown in Figure 4. Figure 4A is the raw data. The broad “hill” is the peak in the LC dimension, while the sharp knifelike feature is the CZE peak. In the rear of the figure a “trench” is noticeable, which is the injection marker for each CZE analysis. Figure 4B is the data from Figure 3C, the result of subtracting the filtered data from the raw data. Notice that the knifelike CZE peak is now visible on a flat background, with the injection marker trench still seen in the rear of the figure. RESULTS AND DISCUSSION
Because the LC analysis involves a solvent gradient, and because the LC effluent is the buffer used in fast-CZE, CZE times for analyses late in the LC run are not strictly comparable to those of CZE analyses early in the LC run. The increasing concentration of organic solvent (ii this case acetonitrile) causes a decrease in electroosmoticflow. Thus, CZE analyses later in the LC gradient show later migration times than those earlier in the LC gradient. The data in Figure 5 were obtained by replacing the LC column with a tee (25bpm id., Valco) through which a small volume of FTC-methoxyarginiie solution was continuously added to the effluent of the pump. In this way, FTC-methoxyarginine was present throughout the gradient. The same LC gradient was run, Analytical Chemistty, Vol. 67,No. 19,October 1, 1995
3451
=-
7.0
-
n
75
-
6.0
-
8.5
-
-s E
U
1.4
Figure 4. 20 surface plot 01 data from Figure 3.(A) is the raw data from Figure 3A, now plotted as a two-dimensional surlace. (E) is the difference data from Figure 3C plotted as a two-dimensional surface.
6.1 6.5 6.9
-
E
7.3
=
7.7
sn
8.1
8.5 8.9
0.0
0.4
0.8
1.2
i.6
2.0
2.4
2.6
CZE(Snc)
Figure5. Effect of LC elution gradient on CZE electroosmotic flow. Sample is FTC-methoxyarginine continuously added to the RPLC mobile phase as a neutral marker of electroosmotic flow in CZE. RPLC conditions: 5-min linear gradient from 10% B to 50% B, hold 1 min at 50% E, and return to initial conditions over 0.1 min. CZE conditions: 20 kV applied over 8-cm total capillary length, with 2 cm between gating and probe beams. CZE injection time. 10 ms: total CZE analysis time, 2.5 s. The region of the 20 data set shown is the same as in Figure 6.
and the change in migration time of the FTGmethoxyarginiie over the course of the LC gradient is obvious as the sloping dark gray line in the right side of the figure. In practice, this change in CZE migration time does not change the use of the method. If standards were being analyzed to compare to migration and retention times in an unknown sample, the standards would be run in the same 2D system as the unknown. If the LC pump reproducibly forms the gradient each time, the shift in CZE migration times will be reproducible as well. Two-dimensional migration/retention times behveen standards and unknowns would still be directly comparable. Figure 6 is a gray-scale plot of the 2D analysis of a hyptic digest of horse heart cytochrome c. The LC gradient is only 5 min long, but the actual times on the LC axis are longer hecause there is about 4.5 min of analysis "dead time". At a flow rate of 250 p W 3452 Analytical Chemisiy, Vol. 67, No. 19, October 1, 1995
1.6
2.2
2.4
Figure 6. 2 0 RPLCffast-CZEanalysis with 5min LC graolent. 2.5-5 CZE analyses. Sample is FTC-tagged trypt c d gest of horse heatl cytochrome c. RPLC conditions: 5-min linear grad ent from 10". B io 50'. B. and nold lmin at 50% B, retun lo initial condlions over 0.1 min. CZE condtions: 20 kV apple0 over 8-cm total capillary length. with 2 cm between gating and probe beams. CZE iniection t me, 10 ms: total CZE analysis time 2.5 s Labeled peaG are lor referencewithin me text but are not ioentilieo as specific compounos.
min, it takes 4.5 min for changes in solvent composition formed in the proportioning valve of the pump to pass through the LC column and connecting tubing to reach the head of the CZE capillary. In fact, sample injection is simultaneous with the start of the elution gradient, but data collection is not begun until four minutes after sample injection. RPLC and CZE are truly complementary techniques. As shown in Figure 6, components unresolved in one dimension are often resolved in the other. To see this, consider first the LC dimension. Any peaks found on the same horizontal line through the data in Figure 6 are peaks with the same LC retention time. By summing horizontally, we get the reconstructed LC. Components represented by those peaks would not be resolved by U: alone. Similarly,peaks found on the same vertical h e represent sample components with similar CZE migration times. Sample components responsible for those peaks would not be resolved by fast-CZEalone. The data in the timing diagram in Figure 1is a fast-CZE analysis of the same digest sample showing a number of unresolved peaks. Clearly RPLC or CZE alone is unable to resolve the multiple sample components under these conditions. Through the combination of WLC and CZE, most components in this sample are separated. The peak capacity of a separation with unit resolution is defined as
n, = L/4o where L is the total distance (or time) over which the sample zones (peaks) are distributed and u is the average standard deviation of the peaks separated.' Using the data shown in Figure 6, we can estimate the peak capacity of the 2D method. In Figure 6, the RPLC peaks elute over a range of 2.8 min, from 5.8 to 8.6 min. The average peak width in the RPLC dimension is 0.19 min 0 Giddings, J. C.Unilrpd Separation Seience;John W W and Sons: New York, 1991.
A
c
0.30
.
,
,
.
,
,
,
0.35
0.40
0.45
0.50
0.55
0.60
0.65
CZE(ses)
I 0.5
1
1.0
I
1
1.5
2.0
CZE 1-1
Figure 7. 2D RPLCffast-CZEanalysis with 5-min LC gradient, 0.7-s
CZE analyses. Sample is FTC-tagged tryptic digest of horse hearl cytochrome c. RPLC conditions: 5-min linear gradient from 10% B to 50% B, hold lmin at 50% B. return to initial conditions over 0.1 min. CZE conditions: 20 kV applied over 8-cm total capillary length, with 0.45 cm between gating and probe beams. CZE injection time, 10 ms; total CZE analysis time, 0.7s. Labeled peaks are for reference within the text but are not identified as specific compounds. (4 standard deviations). Using eq 1and assuming unit resolution, the number of peaks 0.19 min wide that fit into the 2.smin space gives the peak capacity of the RPLC separation, or 15 in this case. Similarly for CZE, the peaks span 1.0 s, from 1.4 to 2.4 s. An average peak width is 0.023 s (4 standard deviations), and this gives a peak capacity with unit resolution of 43. The peak capacity of the 2D RPLC/CZE system is then 15 x 43 650. One possible application of this method is the rapid fingerplinting of proteins. Ttyptic digests of known and unknown proteins could be fluorescentiy tagged and analyzed by 2D RPLC/fastCZE. The 2D data obtained act as a "fmgetprint" of the protein structure based on its component peptides. Comparison of known proteins to unknowns could he done to identify unknowns as well as to assess protein purity. A contaminant in a sample protein or peptide should be visible as spurious peaks not found in the 2D data for that protein or peptide standard. Sampling Considerations. In a continuous 2D system, the analysis time of the second dimension separation limits the rate at which the first dimension separation is sampled. As mentioned earlier, each entire analysis in the second dimension gives only a single "poinv in the first. Sampling of the first dimension determines how well the fust dimension separation is characterized. ' h a t is, if the first dimension is not sampled often enough, we do not make full use of the resolution obtained in that dimension. However, if the second dimension is made too rapid, we generally lose resolution in the second dimension. In fast-CZE, with a given applied voltage and total capillary length, the actual CZE analysis time and separation efficiency are determined by the separation of the gating and probe beams. The distance between the beams can be adjusted to give the desired compromise between time of analysis and CZE separation efficiency. Greater distance between the beams gives a more effective CZE separation, because a greater fraction of the total applied voltage is used, but the analysis time is also greater. Thus the LC separation is sampled less often, and we may not take full advantage of the resolution obtained in the LC column. Alter-
0.0
I
1
I
0.1
0.2
0.3
I 0.4
I 0.5
I 0.6
CZE (us)
Figure 8. Individual fast-CZE analyses taken from 2D data sets. (A) is a CZE analysis taken from the data in Figure 6 at 6.3-min LC time, with 2 cm between the gating and probe beams and a total CZE analysis time of 2.5 s. (B) is a CZE analysis taken from the data in Figure 7 at the same LC time, with 0.45 cm between the beams
and a total CZE analysis time of 0.7 s. nately, less distance between the beams results in faster CZE analyses (so that the LC separation is sampled more rapidly) but less CZE separation efficiency. In Figure 6, the fastCZE analysis time is 2.5 s. This maximizes efficiency in the CZE analysis but somewhat undersamples the LC analysis. Figure 7 is an analysis of the same sample with a beam separation of 0.45 cm. The CZE analysis time is much shorter, only 0.7 s, so that the LC analysis is better sampled. This increase in sampling of the LC has come at the expense of lost efficiency in the CZE. Panels A and B of Figure 8 are individual CZE analyses taken from the 2D data sets in Figures 6 and 7 at 6.5min LC time. The peaks are much sharper and much farther apart in the 2.5s analysis than in the 0.7-s analysis, though they are still baseline resolved in the faster analysis. Notice in Figure 7, the pair of peaks labeled B are fully baseline resolved in the LC. In Figure 6, these peaks are beginning to run together. The LC gradient is the Same in both cases, but with the slower LC sampling rate (due to the longer CZE analysis time), the LC separation in Figure 6 is not as well characterized. In Figure 6, the LC resolution available is not completely utilized due to the slower CZE sampling rate. Compare also the cluster of peaks labeled A in Figures 6 and 7. In Figure 6, these peaks Analytical Chemistry, Vol. 67,No. 19, October 1, 1995 3453
0.W 0.0
11.1
7.0
Wu: n m
-
7.6
8.0
l.W
0.50
1.50
2.04
2.50
8.8
om",
C E (.as)
-
Reconstructed LC chromatoorams for the 2D data shown in Figures 6 and 7.These plots are created by integrating (summing) the signal across each CZE analysis (each horizontal row) in the 2D data. Fiaure 9.
are well resolved by CZE but show some overlap in the LC dimension. In Figure 7,they are not as well separated by CZE but are better resolved in the LC. Again, the faster CZE analysis time makes more complete use of the LC resolution. The particular compromise chosen between the necessary sampling rate in the U:and the desired efficiency in the CZE will depend somewhat on the particular sample being analyzed. Figure 9 shows the reconstructed LC chromatogram for the data shown in Figures 6 and 7. These plots are created by integrating (summing) the signal across each CZE analysis (each horizontal row) in the 2D data. Each data point in these plots then represents an entire fastCZE analysis. The difference in CZE sampling of the LC chromatogram is more obvious in the data shown in this way. In Figure 9, there is a group of four peaks near 6.0 min in the LC chromatogram in which a small peak is visible between the last two large peaks in the analysis sampled every 0.7 s. At the same position in the analysis sampled every 2.5 s, the small peak is not seen. Here, because of the slower sampling of the LC by the CZE, resolution gained in the LC has been lost in the 2D analysis. Similarly, in the last peak near 8 min, a small shoulder on the larger peak is more distinctlyvisible as another peak in the lower trace than it is in the upper. One means of dealing with the compromise between CZE analysis time and thorough sampling of the LC dimension is through overlapped injections. This technique has been described earlier! Figure 1OA shows the entire 2D data set for the analysis shown in Figure 6 without overlapped injections. Notice in the far left of the figure a white vertical line is visible. This is the CZE injection marker described earlier. It is purposefully offset from the very edge of the 2D plot, but it is at zero time on the CZE axis. The unused part of the 2D separation space is obvious, because most of the sample components have CZE migration times in the last half of the CZE analysis. To make better use of the total 2D separation space, the injections in the second dimension may be overlapped. To do so, before the peaks from one injection have reached the detector, a second injection is done. This doubles the rate at which the LC analysis is sampled, with no loss in CZE efficiency. Figure 1OB shows the same analysis (8) larmann. 1. P..Jr.: Lemmo. A V.: Moore. A W..
lr.: lorsenson, I. W.
Electrophonris 1993.14. 439-447.
3454 Analytical Chemistry, Vol. 67, No. 19, October 1, 1995
ko 8.5
1 1.2
11
1.6
1.8
2.0
2.2
2.4
CZE (ss) Figure I O . Overlapping of CZE injections to fill the 20 separation space. (A, top) is the same data as in Figure 6 but shows the entire 20 data set. The vertical white line at the left of the figure is the fastcze injection marker. (B. bottom) is the same tryptic digest sample, with the same LC and CZE conditions but with overlapped CZE injections. Notice that the vertical white injection marker is no longer at zero on the CZE axis.
with overlapped injections. The injection marker is also visible here, but no longer at zero because of the overlapped injections. The effect is to obtain the efficiency of 2.5s CZE analyses, while sampling the LC run every 1.25 s. Figure 11 compares the sampling of the LC run every 2.5 s with that of every 1.25 s obtained through the use of overlapped injections. Overlapped injections may not be useful for all samples. A drawback of overlapped injections is the possibility of unusually late peaks from one CZE analysis "aliasing" into the following analysis, resulting in an incorrect CZE migration time for that peak. Faster 2D An&is. Because 2D methods have inherently higher peak capacity, they generate that peak capacity more rapidly. If not all of the peak capacity is needed, it can be exchanged for a decrease in analysis time. Figure 12 shows a 2D analysis in which the LC gradient is completed in only over 2 min. This is faster than optimum for the LC separation, but because it is pad of a 2D system, the second dimension can be used to regain some of the resolution lost in the accelerated LC gradient In Figure 1% the fast-CZE beam separation is only 4.5 nun. This gives a rapid 0.7-s CZE analysis time, with 0.35s sampling of the LC separation using the overlapped injection
8.0
7.0
7.5 RPLC n m m n ,
8.0
1.1
O X
0.45
0.40
Figure 11. Reconstructed LC chromatograms from Figure 10A.B. The reconstructed LC separations allow comparison of sampling for nonoverlapped and overlapped injection 2D analyses.
technique. Unfortunately, with the gating and probe beams so close together, the CZE efficiencyis low and does not adequately make up for the resolution lost in the LC. In Figure 12B, the beams are 2 cm apart The greater beam separation gives greater CZE efficiency,which does begin to restore the resolution lost in the LC. The large peaks Oabeled A) in the early part of the LC gradient had merged in Figure 12A to give an unresolved multiplet. We h o w from Figures 6 and 7 that this multiplet is actually three peaks. In Figure 12B, the high-efficiency CZE analyses have separated the multiplet into two components, so that at least some fraction of the lost LC resolution is compensated for in the second dimension. An interesting artifact of the fast LC gradient is seen in F i r e 12B. All of the peaks appear to be swept toward the lower right comer of the figure. This is because of the change in CZE migration time with the changing solvent gradient. Because the LC solvent composition is changing so rapidly, the migration times of peaks in the CZE are changing even over the course of a single peak in the LC. Thus, the front of an LC peak has a shorter CZE migration time than the rear of the same peak, and seen in two dimensions, the appearance is that of a peak slewed to the right. The same effect is present in Figure 12.4 but is not as obvious because of the lower CZE separation efficiency. In practice this is an indicator that the gradient is changing too rapidly for the given LC column. ACKNOWLEDGMENT
This research was supported by grants from the National Science Foundation (CHE 9215320) and the National Institute of Health (GM 39515) and by a gift from Hewlett-Packard. The
0.50
0.55
0.60
0.
CZE(*nc)
5.0
=
5.5
6.5 6'o
1 1.2
1.4
1.6
1.8
2.0
2.2
m(snc)
Figure 12. 2D RPLCKZE analysis with 2-min LC gradient and overlapped injections. Sample is FTC-tagged tryptic digest of horse heartcytochromec. RPLCconditions: 2-min linear gradient from IO% B to 50% B, hold 1min at 50% B, and then return to initial conditions over 0.1 min. (A, top) 0.7-s CZE analyses, with CZE conditions as in Figure 7, but with overlapped injections. (B) 2.5-s CZE analyses, with CZE conditions as in Figure 6 but with overlapped injections. Labeled peaks are tor reference within the text but are not identified as specific compounds.
HP1050 pump and UV absorption detector were donated by Hewlett-Packard. Received for review October 25, 1993. Resubmitted July 7, 1995. Accepted July 14, 1995." AC9506793 *Abstract published in Advnnce ACS Abrhoctr, September 1. 1995.
Analyticai Chemistry, Voi. 67,No. 19,October I, 1995
3455