Automated instrumentation for comprehensive two-dimensional high

Pier Giorgio Righetti , Roberto Sebastiano , Attilio Citterio ..... Amanda E. Sinha , Bryan J. Prazen , Carlos G. Fraga , Robert E. Synovec. Journal o...
0 downloads 0 Views 3MB Size
978

Anal. Chem. 1990, 62,978-984

Automated Instrumentation for Comprehensive Two-Dimensional High-Performance Liquid ChromatographyKapillary Zone Electrophoresis Michelle M. Bushey and James W. Jorgenson* Department of Chemistry, University of North Carolina, Chapel Hill, North Carolina 27599-3290

A cornprehenslve two-dimenslonai (2-D) separation system utiiizlng reversed-phase (RP) chromatography as the first dimension Separation and capillary zone electrophoresis (CZE) as the second dimension separation is presented. Effluent from the RP column, operated under gradient conditions, fills a loop on a computer-controlled six-port valve. A second pump flushes this loop material over the grounded (anode) end of a CZE capillary at specific intervals. Eiectromigration injections are performed on the CZE portion of the system from a flowing stream. Fluorescence detection is used on the CZE capillary. The system is used for the analysis of peptide standards and fiuorescentiy labeled peptide products from a tryptic digest of ovalbumin. The 2-D system has a much greater resolving power and peak capacity than either of the two systems used independently of each other. The entire system is automated and operated under computer control. Three-dimensional data representation provides a means of viewing peak profiles in either separation dimension, and contour mapping of the 3-D data provides a "fingerprint" of protein digests.

INTRO DUCT10N A variety of two-dimensional (2-D) and coupled separation mechanisms has been developed for the analysis of complex samples. The coupling of liquid chromatography and gas chromatography to mass spectrometry, tandem mass spectrometry, LC with electrochemical detection, coupled column techniques, and other such systems can produce a wealth of information on complex samples. Recently we presented a method termed "comprehensive two-dimensional HPLC" ( I ) , which coupled ion exchange chromatography and size exclusion chromatography in a manner such that all of the effluent from the first column was reanalyzed on the second column without the use of stopped flow methods. This is an important aspect to consider if fraction collection of the purified peaks is desired in a reasonable time frame. The entire 2-D system was automated and used to separate protein standards and serum samples. Substantially greater peak capacity and resolution can be obtained by two-dimensional separation systems. This is because, assuming no resolution is lost during 2-D operation, the two coupled systems work at the same level of performance as when they are uncoupled, and the methods are orthogonal, the peak capacity becomes the product of the peak capacities for each dimension. The resolution becomes the square root of the sum of the squares of the resolution in the two independent systems (2). Giddings has pointed out that many samples need to undergo more than one separation mechanism to reduce peak overlaps ( 3 , 4 ) . For two techniques to be successfully and satisfactorily coupled, however, several criteria need to be addressed. First, and most importantly, the two techniques should be as orthogonal to each other as possible. This means the two techniques should base their respective separations

on as different sample properties as possible. This will reduce the amount of cross-information and make the most out of 2-D operation. The orthogonality aspect creates an interesting problem in the design of 2-D separation systems; the more orthogonal two separation mechanisms are, the more dissimilar they will be in operation, and the more dissimilar the two systems are in operation, the more difficult it will probably be to couple the two systems. Another aspect to consider in designing 2-D separation systems is that it is necessary for the second dimension separation to sample the first dimension separation as frequently as possible; this is because it is desirable for the second dimension separation to sample a peak eluting from the first dimension separation several times across each peak's width. In this way, peak profiles can be obtained in both separation dimensions, and the first column's resolution is preserved. Three-dimensional (3-D) data representations should be employed to take advantage of obtaining peak profiles in two dimensions. Short analysis times are not only convenient but can be necessary if biological or other sensitive samples of interest degrade or change with time. Automation of the entire system is also desirable and again can be necessary if a large number of first column fractions is sampled by the second separation system. Fraction collection is another feature that may be desired. Ideally, the option for fraction collection should be after the second dimension separation, so full advantage can be taken of the high separation power of 2-D operation. Reversed-phase chromatography and capillary zone electrophoresis (CZE) are presumably highly orthogonal separation methods. Reversed-phase high-performance liquid chromatography (RP HPLC) separates analytes on the basis of hydrophobicity and CZE separates analytes mainly on the basis of charge, and to a lesser degree on size. On this basis, they are good candidates for pairing in a 2-D system. Several groups have already recognized this and have used CZE to analyze collected RP HPLC fractions of enzymatic digests of proteins and to compare the tryptic digest fingerprints of R P HPLC and CZE (5-7). The unsurprising result in cases like these is that CZE is found to be able to resolve some peptides that coelute on R P HPLC. Although no one has actually reexamined collected CZE fractions by R P HPLC, Nielson and co-workers (6) have shown that R P HPLC is capable of resolving peptides that comigrate in CZE. This was done by careful examination of RP HPLC and CZE tryptic digest maps of human growth hormone. This is not a surprising result, especially considering those species with zero net charge that migrate at the velocity of the electroosmotic flow and are thus unresolved by CZE. Pairing electrophoretic separations with chromatography has also been done by using isotachophoresis as a purity check for R P HPLC analysis of peptides (8). While several other groups have analyzed the same samples by LC and CZE, and a handful of groups have reanalyzed LC fractions by CZE, only one other group that we are aware of has attempted to automate the coupling of HPLC with CZE. That attempt coupled a Sephadex G-50 size exclusion column as the first dimension separation to isotachophoresis as the

0003-2700/90/0362-0976$02.50/06 1990 American Chemical Socletv

ANALYTICAL CHEMISTRY. VOL. 62. NO. IO, MAY 15, 1990

second dimension separation. Second dimension analysis times, however, were very long a t 18 min each, stopped flow methods were used on the first column during the second dimension separation, and 3-D “chromatoelectropherograms” were not presented. The system was used to analyze a sample containing bovine serum albumin, myoglobin, and tyrosine

RUN

979

INJECT

n 01 il“

(9).

The work presented here is the automated coupling of reversed-phase microbore chromatography and capillary zone electrophoresis. Many of the techniques used here parallel those in our earlier work ( I ) . The dissimilarity of R P HPLC and CZE both increases the difficulty of the coupling while a t the same time enhances the power and usefulness of this particular 2-D system. Some of the features of the system are as follows: the system is entirely automated; the entire sample is subject to separation in both dimensions (although the entire sample volume does not go through the second dimension separation (CZE)); a large number of first column fractions is reanalyzed on the CZE system: several forms of data presentation including 3-D chromatoelectropherograms are available; the analysis time is not unreasonably long considering the complexity of the sample. In addition, we present an alternative method of sample introduction on the CZE system. Electromigration is used as the injection method, hut no manual manipulation of the fused silica capillary is necessary since sample material or fresh buffer is constantly being transported past the anode of the capillary. EXPERIMENTAL S E C T I O N Samples and Reagents. Each analysis was performed with a 0.012 M potassium phosphate buffer at pH 6.9, which was prepared daily from a single 0.2 M stock solution to minimize day-to-day changes in the reversed-phase chromatography and electrophoresis. This solution is referred to as buffer A. Buffer A was used on the RP column as well as in the CZE system. Deionized water was further purified with a Barnstead Nanopure System (Boston, MA). All solutions were filtered with 0.22-pm pore size MAGNA Nylon 66 membrane filters from Micron Separations, purchased from Fisher Scientific (Raleigh, NC). Reagent grade acetonitrile was purchased from Fisher Scientific. The following were purchased from Sigma Chemical Co. (St. Louis, MO): chicken egg albumin (ovalbumin),fluorescamine (fluram), riboflavin, bovine pancreatic trypsin, angiotensin I, Met-Leu-Phe, methionine enkephalinamide, and leucine-enkephalin. Digest Conditions. One-tenth gram of ovalbumin was dissolved in 10 mL of buffer A. One drop of an indicator solution (brilliant yellow) was then added to help monitor the pH change when the pH was adjusted to 8.0 with “,OH. The solution was heated in a boiling water bath for 6 min. After the solution cooled, 2.0 mg of trypsin was added to the solution and the digest solution was allowed to react for 4 h a t 37 “C. The digest was stored at 4 “C until needed. This procedure was based on previous work in both our lab (10) and other labs (11). Tagging Conditions for Digest. Five-tenth milligram of fluorescamine was dissolved in SO pL of acetonitrile. The digested protein sample was adjusted to pH 9.0 with KOH. Peptide solution (100 ALL) was added to the SO-pL fluorescaminesolution. Buffer A (500 pL) was added to this solution. The resulting solution was used as the injection sample. Samples were used within 5 min of tagging. Tagging Conditions for Peptide Standards. Two-milligram portions of Met-Leu-Phe and angiotensin I were each added separately tu I-mL aliquots of buffer A, which had been adjusted to pH 9.0. One milligram of methionine enkephalinamide and leucine-enkephalinwas each added separately to 1 mL of buffer A at pH 9.00. One milligram of fluorescamine was dissolved in 100 pL of acetonitrile. Twenty microliters of this solution was added to 40 pL of each of the four peptide standard solutions. Twenty-microliter portions of each of the resulting four solutions were mixed together, and 300 pL of buffer A at pH 6.9 was added to the sample mixture. In addition, 20-pL portions of each of the four tagged peptide solutions were also added to 300 pL of buffer A separately. These four separate samples were used on the CZE

Figure 1. Two configurations of six-port. computer-controlled valve. C1 is RP HFiC wlumn. P2 is pump 2. L is loop. CZE is capillary zone electrophoresis fused silica capillary. PW is paper wick. W is waste.

Figure 2. Schematic of 2-D LC/CZE insbumentation: A and 0, buffer A and acetonitrile, respectively; PI. Brownleee microgradient syringe pump; M. 52-pL mixer; V I . Valco six-port manual injection valve; S. injection syringe: L1. 50-pL loop: C1, reversed-phase column: P2,

Waters Associates Model 6000A piston pump; V2, grounded six-port electrically actuated Valco valve; L2, IO-pL loop: CZE. CZE capillary; D, fluorescence detector; 10. interlock box: $A, microammeter;GB, grounding box: HV. Spellman highvonage power supply. system for identification of the peptides by migration time. Riboflavin Solutions. Four-tenths milligram of riboflavin was dissolved in 20 mL of buffer A. Twenty-microliter portions of this solution were added to the following nine solutions: 80 pL of buffer A; 70 pL of buffer A, 10 pL of acetonitrile; 60 r L of buffer A, 20 pL of acetonitrile; SO pL of buffer A, 30 pL of acetonitrile;40 pL of buffer A, 40 pL of acetonitrile;30 pL of buffer A, SO pL of acetonitrile; 20 pL of buffer A, 60 pL of acetonitrile: 10 pL of buffer A, 70 pL of acetonitrile; 80 pL of acetonitrile. The resulting solutions were each M in riboflavin and used over a 180-minrun time to simulate injections from solutions that were 0%, lo%, 20%, 30%.40%. 50%,60%, 70%, and 80% acetonitrile, respectively. Instrumentation. The coupling of a liquid chromatography column with capillary electrophoresiswas accomplished through the use of a six-portvalve. The two valve configurations are shown in Figure 1. In the “run” position, effluent from the LC column (C1) filled the loop (L) while the second pump (P2) continuously forced fresh buffer coaxially past the grounded (anode) end of the CZE capillary. A paper wick (PW) carried excess buffer away from the valve. In the inject position, effluent from C1 went directly to waste while P2 forward flushed the contents of the loop past the grounded end of the CZE capillary for electromigration injection. At the end of the injection time the valve was returned to the “runXposition. Timing and control of the valve movements are described in the subsection Instrument Control. Figure 2 is a schematic diagram of the instrumental setup. P1 was a Brownlee microgradient syringe pump (Santa Clara, CA). The pump was fitted with a Brownlee 52-rL mixer column labeled M in Figure 2. V1 was a manually operated six-port valve with ‘/,6-in. fittings and a standard port diameter of 0.016 in. Valve material was Nitronic 60. The valve and its fittings were purchased from Valco Instruments Co., Inc. (Houston, TX). The loop (Ll) was a Valco 50-pL loop. S was a 100-pL Hamilton syringe fitted with a 2-in., 22-gauge needle purchased from Alltech Associates (Waukegon, IL). C1 was a Brownlee Aquapore RP-300 reversed-phase column, 250 mm X 1.0 mm, and was purchased through Rainin Instrument Co., Inc. (Woburn, MA). P2 was a Waters Associates (Milford,MA) Model MH)OA piston pump. V2 was a six-port electrically actuated valve similar to V1 except the valve material for V2 was Hastelloy C. The loop (L2) was a Valco IO-pL loop (19 cm length, 0.26 mm i.d.). All connecting tubing, unless otherwise noted, was 0.007 in. i.d., and in. fittings were used

980

ANALYTICAL CHEMISTRY, VOL. 62, NO. 10, MAY 15, 1990

The entire valve was held at electrical ground and functioned as the anode for the CZE system. An alligator clip placed on the stainless steel valve body ensures that the connection to ground is maintained. Fused silica capillary was purchased from Polymicro Technologies (Phoenix, AZ). No capillary surface pretreatments or coatings were used. The fused silica capillaries used were 41 pm i d . and 150 pm o.d., and also 50 pm i.d. and 150 pm 0.d. Each was cut to a total length of 38 cm. A short section of the polyimide coating was burned off of the fused silica to create a detection window. The distance from the center of this window to the grounded (injection) end of the capillary measured 6.5 cm. A 1.7-cm portion of the fused silica capillary extended down into a 2 cm length piece of 1/18-in.316 stainless steel tubing with an inner diameter of 0.040 in. (shown in Figure 1). This tubing was held in place on the six-port valve by a nut and ferrule. No rigid fittings were used to hold the capillary in this piece of tubing. The short fused silica length and rigid mounting of the valve prevented the CZE capillary from moving out of position. A short section of the polyimide was also removed at the very end of the capillary to prevent sample from becoming trapped between the polyimide and the fused silica during injections. A slit was cut near one end of a paper towel, which was folded lengthwise and wrapped in aluminum foil. This slit was placed around the tubing and nut, and this assembly acted as a wick to carry away excess solvent to a waste container and prevent corrosion of the valve and detector. The aluminum foil allowed the wick to rest against the valve without wetting the valve. The detector (D) used in this work has been described elsewhere (12). It was a variable-wavelength fluorescence detector. An excitation wavelength of 365 nm was isolated with a double monochromator. The slit width was 10 nm. A 470-nm cut on emission filter was used. A f30 kV dc power supply (HV) with the remote voltage programming option was obtained from Spellman High Voltage Electronic Corp. (Plainview, NY) and was used in the negative voltage mode. The direction of migration for all analytes was from positive to negative electrode. The negative end of the CZE capillary terminated in a vial containing the same buffer as that being pumped by P2, which was the same as buffer A on C1 (A). The buffer in this vial was leveled to the same height as the top of the 2 cm length piece of tubing in V2 to prevent hydrodynamic flow in the capillary. An in-house made grounding box (GB) and interlock box (IB) were used for operator safety at the high voltage end of the CZE system. A microammeter (FA)was placed between the high-voltage electrode and CZE buffer vial inside the interlock box to monitor current. The first dimension separation could be run independently of the CZE system by moving C1 to the P2 port in Figure 1 and replacing the CZE system with a short piece of fused silica, held in the valve by a Valco fused silica adapter. The valve (V2) was moved to the “run” position for this type of operation. This procedure allowed for coupling of the microbore RP column to the capillary fluorescence detector. Effluent from C1 flowed through the valve directly to the piece of fused silica tubing mounted in the detector and emptied to waste on the opposite side of the detector. The second dimension (CZE) separation could be run independently of the LC system by replacing C1 in Figure 1 with a Valco syringe port. A 50-pL Hamilton syringe was then used to fill the 10-pL loop for a single injection on the CZE portion of the system. The valve could then be actuated by manual means or through computer control. Instrument Control. An IBM/PC/XT, purchased from IBM (Boca Raton, FL), was fitted with a Labmaster multifunction input/output interface board which was purchased from Scientific Solutions (Solon, OH). The computer was used to record data, control the valve switching of V2 shown in Figures 1 and 2, and control the high-voltage power supply. The interface board was used as follows: a 16-bit analog-to-digital converter (ADC) was used to acquire data from the fluorescence detector photometer output, a 100-pF capacitor was placed across the ADC input; a 12-bit digital-to-analog converter was used to control the level of high voltage applied during the analysis through the remote voltage programming option of the power supply; a programmable parallel port (Intel 8255) was used for control of the valve position and the on/off control of the power supply; a programmable timer (Advanced Microsystem 9513) was used for time data acquisition.

VOLTAGE

RUN IKlECT

VALVE

IKlECT RUN

L 4

DATA ACQUISITION

OFF ON

~

RUN TIME (60 Si

ACQUISITION TIME (55Si

-

-

HV SLEW TIME (2 Si INJECTION TIME 13 SI

Figure 3. Computer program timing diagram drawn to scale. Assumes a 1-min “run” time, 3-s injection time, and 2-s HV slew time; actual data acquisition time is equal to run time minus HV slew time and injection time.

timed valve movements, and timed voltage changes. Data acquisition rates depended upon the particular application. Computer Programs. Hardware and LC-CZE instrumentation were controlled by software written in Microsoft QuickJ3ASIC. Programs were written in-house. The program allowed for user entered parameters such as CZE injection and run voltages, CZE injection time lengths and run times, number of CZE runs, and data acquisition rate. The injection time and run time variables determined the frequency of V2 movements. Bookkeeping data with regards to other run conditions could also be entered. At the completion of the analysis, another portion of the program processed the data and converted it to a compatible form for use with the plotting software, Surfer version 4.0 (Golden Graphics Software, Golden, CO). Several options were available for data processing. By use of Golden Graphics Software, the following results could be obtained: (a) The data could be presented as a 3-D chromatoelectropherogram viewed from any angle or height, where the x axis represented the migration time on the CZE system, the y axis represented the chromatogram retention time, and the z axis was detector response. Any combination of X,Y , or 2 lines may be plotted. (b) Gold Graphics Software also allowed for contour maps of the 3-D data; entire chromatoelectropherograms or portions thereof could be plotted in either of these two forms. In-house-written software allowed for the following types of data presentations: (a) individual electropherograms could be displayed (b) any number of electropherograms could be summed together and displayed, the resulting display of the summation of all electropherograms was a simulation of the electropherogram of the total original sample; (c) a chromatogram “slice”of the 3-D plot could be displayed; (d) the width of this “slice” was varied by summing individual chromatogram “slices”,summing all slices produced a simulation of the chromatogram of the total original sample, smaller widths were the simulated chromatograms for particular electrophoretic mobility range. There were several considerations in the timing aspects of this system’s operation. A timing diagram is illustrated in Figure 3. In this diagram, a user-entered value of 1 min was assumed for the length of the CZE run time. A user-entered value of 3 was assumed for the length of the CZE injection time. Because the shortest injection time possible was limited by the valve switching time, the power supply voltage must be lowered for these injections. Otherwise, band broadening due to excessively large injection slugs was the result. Another consideration was that it takes approximately 2 s for the power supply voltage to drop from -30 to 0 kV. A faster drop can be facilitated by operation of the grounding box, but this had a tendency to interfere with the computer and over the course of a several hour run would cause unnecessary wear on the grounding box solenoid. Therefore, prior to each valve movement to the inject position, there was a 2-s interval during which no data were collected and the power supply voltage was allowed to drop from the run voltage to the injection voltage. This 2-s high-voltage slew time interval was constant regardless of any other run parameters. A t the end of the 2-s interval the valve was turned to the inject position and the electromigration injection was begun and continued for the length of time entered by the user. No data were collected during the injection period. A t the end of the injection time the valve was returned to the “run” position, voltage was returned to the run voltage value, and data acquisition was resumed, in that order.

ANALYTICAL CHEMISTRY, VOL. 62, NO. 10, MAY 15, 1990 981 This entire sequence of 2-s interval, injection time, and data collection time took place during the user entered "CZE run time". Therefore, CZE data acquisition times were actually shorter than the user-specified time, and the time allowed for the RP column to fill the loop on V2 was equal to the user-specified run time minus the injection time. This also means C1 always underfilled L2 by about 0.5 pL each injection cycle. However, since P2 overflushed the loop during a CZE injection, this small amount of underfilling was unimportant. System Preparation. Each day fresh buffer A was aspirated through the CZE capillary; -15 kV was then applied for 15 min. Prior to the start of the 2-D run, the sample was manually injected on the CZE capillary to verify that electrophoretic analysis would be completed during the specified run time and to recheck the injection parameters such as voltage, time, and P2 flow rate. The CZE capillary was aspirated again with buffer A before reconnection of the 2-D system and the start of the first column gradient. Due to limited computer memory, the operation of the CZE portion of the system was not begun until a short time before the elution of the first peak from the RP column. This time was determined by prior operation of the R P column independent of the CZE system. Only relatively minor elution time variations from day-to-day have been observed on the RP column. Separation Conditions. Peptide Standards. A 10 pL/min flow rate was used on C1. The mobile phase was held isocratic in buffer A for 10 min. The gradient was begun a t 10 min and the mobile phase was changed from 0% to 25% acetonitrile from 10 to 100 min; from 100 to 200 min the mobile phase was changed from 25% to 50% acetonitrile; from 200 to 230 min the mobile phase was changed from 50% to 75% acetonitrile; from 230 to 290 min the mobile phase was changed from 75% to 90% acetonitrile. Due to limited computer memory, the CZE injections were not begun until 204 min and continued until 264 min. Sixty injections on the CZE system were made. One CZE injection was made every minute. The flow rate on P2 was 0.5 mL/min, the injection voltage was -2 kV, and V2 was held in the inject position for 5 s to ensure complete flushing of the 10-pL loop. The CZE run voltage was -19 kV. A 50 pm i.d. CZE capillary was used. The data collection rate was 5 points/s. Tryptic Digest. A 10 bL/min flow rate was used on C1. The mobile phase was held isocratic in buffer A for the first 10 min. A gradient from 0% acetonitrile to 30% acetonitrile ran from 10 to 175 min. From 175 to 300 rnin the mobile phase was changed from 30% to 90% acetonitrile. CZE injections were begun at 95 min and continued until 275 min; 180 injections were made. One injection every minute was performed with a flow rate of 0.4 mL/min on P2. The valve was held in the inject position for 3 s for each injection and the injection voltage was -3 kV. The CZE run voltage was -22 kV. A 41 pm i.d. CZE capillary was used. The data acquisition rate was 5 points/s. To investigate any changes in the electroosmotic flow in the CZE system, manual injections of riboflavin solutions were made approximately every 20 min on a 50 pm i.d. capillary. One hundred eighty valve rotations were performed; every 20th rotation the loop was filled with the appropriate riboflavin sample to simulate a gradient from 0% to 80% acetonitrile from 0 to 160 min. The flow rate on P2 was 0.4 mL/min. The injection voltage was -3 kV; the run voltage was -15 kV. The valve was held in the inject position for 5 s. One CZE injection was made every minute. The data acquisition rate was 3 points/s.

RESULTS Figure 4 is a contour map of nine injections of riboflavin on the CZE portion of the 2-D system. One injection was made approximately every 20 min from solutions made to simulate a gradient of 0% t o 80% acetonitrile over a 160-min time frame as described in the Experimental Section. The R P column was removed and replaced with a syringe port for these injections. Between injections the valve was operated, but no injections were made. The entire data acquisition time for each CZE run is shown to illustrate the magnitude of electroosmotic flow change with respect t o the full run time. Tic marks on the CZE migration time axis represent 1 s each. Tic marks on the injection number axis represent five injections each. Table I lists the migration time (first statistical

A0 m 0

CZE MIGRATION TIME (SEC) 15 30 45 530 I

11l-rTl~-

r7

-9

A

9-

w

Q

CZE MIGRATION TIME (SEC) Figure 4. Contour plot of riboflavin injections: P2 flow rate, 0.4 mL/min; injection voltage, -3 kV; injection time, 5 s; run voltage, -15 kV; run time, 1 min; 1 riboflavin injection every 20 min, valve.operated every minute; injections 1 through 180 plotted; samples simulate a

gradient from 0% to 80% acetonitrile as described in the text; buffer A, 0.012 M potassium phosphate, pH 6.9; capillary, 50 km i.d. 150 krn o.d., 6.5 cm to detector, 38 cm total length; data acquisition rate is 3 pointsls; every point is plotted. Table I. CZE Migration Time as a Function of Sample Solvent Composition %

acetonitrile

0

10

20

30

40

50

60

70

80

first moment, s 32.0 32.1 31.8 31.6 31.3 31.2 31.2 31.2 31.2 standard deviation, 0.372

relative standard deviation, 1.18%

moment) for all nine injections and their compositions as well as the standard deviation and relative standard deviation. The relative standard deviation of the first statistical moment for these peaks is 1.18%. This variability, as can be seen in Table I and examination of Figure 4, is due more to a slight gradual increase in the electroosmotic flow rather than a more scattered type of variation from injection to injection. This means a single peak eluting from the LC column over the course of several CZE injections would be largely unaffected by this variability (if this type of variability is typical of all 2-D runs). Although the electroosmotic flow can change significantly from one day t o another (which necessitates daily optimization of the CZE parameters prior t o 2-D operation as stated in the Experimental Section), no evidence has been seen indicating significant variation within the time period of one 2-D run. Figure 5 contains Surfer generated plots of a chromatoelectropherogram of leucine-enkephalin, Met-Leu-Phe, angiotensin I, and methionine enkephalinamide, all of which were labeled with fluorescamine. The symmetrical peak shapes are evidence that electroosmotic flow variation between neighboring runs on the CZE system is insignificant. A factor that may have contributed to the extremely steady electroosmotic flow is the continuous supply of fresh buffer at the CZE capillary anode. These plots nicely demonstrate the increased separating power of 2-D operation. Neither method, used alone, could separate all four analytes under these conditions, but in combination with each other, these four peptides are cleanly resolved. Figure 6a is a single dimension chromatogram of a R P separation of the tryptic digest of ovalbumin, again labeled with fluorescamine. Although a slower gradient might effect a more complete separation, computer memory limits the total length of time during which the CZE portion of the system can operate at a sufficient data acquisition rate. There are two very small peaks that elute a t approximately 23 min, which is not in the time frame sampled by the CZE system.

982

* ANALYTICAL CHEMISTRY. VOL. 62. NO. 10. MAY 15. 1990 i

R

RP ELUTION TIME IMIN)

CZE MIGRATION TIME (SEC) Figure 5. (a). Surfer generated 3-D chromatoelectropherogramof fluorescamine labeled peptide standards. Peaks are identified as follows: A. Met-Leu-Phe: 6, ieucineenkephalin: C, angiotensin I: D. methionine enkephalinamide: P1 flow rate 10 pL1min: gradient conditions described in Experimental Section: P2 flow rate, 0.5 mllmin: injection vonage. -2 k V injection time, 5 s: run voltage, -19 kV: run time, 1 min: injections 1 through 30 plotted; data acquisition rate. 5 oointsls: everv other ooinl dotted: CZE caoilhw. um ~,50 um i.d... 150 ~ o 0 , 6 5 c m ~ tdetector. i 3 8 c m iota engtn. Ib) Conio,r plot of Same ~~

.

chromatoelectropherogram.

In any case, this is a complex sample which is clearly not completely resolved by use of the RP column alone. Tic marks on the x axis are a t 1-min intervals, indicating how this separation would have been "fractionated" for CZE runs had a full 2-D analysis been performed. Assuming an elution time range from 120 to 260 min and peaks 4 min wide a t the base, this dimension has a peak capacity of approximately 35 under these gradient conditions (probably a conservative estimate). Figure 6b is a single dimension electropherogram of the same fluorescamine-labeled digest sample. This electropherogram was obtained as described in the Experimental Section by replacing C1 in Figure 2 with a syringe port and a user-enetered value of one for the number of CZE injections. Assuming a migration time range of 18-55 s and 3 s wide peaks, this dimension has a peak capacity of approximately 12. On the basis of these estimates for peak capacity in both dimensions, the 2-D operation of this system should have a peak capacity of a t least 420. One aspect that should be noted in Figure 6b is the mobility discrimination observed with electromigration injection. Late eluting peaks which have low mobilities are discriminated against in the injection. Smaller peaks later in the electropherogram are the result. A hydrodynamic injection method for use with 2-D operation, which should correct this problem, will be investigated in this lab. Both of these figures demonstrate the complexity of the sample and the inadequacy of either technique to fully resolve the sample when operated independently of each other. Figure 7a is a chromatoelectropherogram of the 2-D separation of the digest sample. One injection on the CZE system was made every minute. Each tic mark on the RP elution time axis represents five CZE injections. Injection numbers 15 through 150 are displayed. The gradient used on the LC column was identical with that used for obtaining the data for Figure 6a. Well over 30 peaks can be counted in Figure

25

35

45

55

CZE MIGRATIONTIME ISECI

Figure 6. (a). Single dimension RP chromatogram of fluorescamine labeled tryptic digest of ovalbumin: Ct flow rate. to pLlmin: gradient, 0 to 10 min, isocratic in buffer A 10-175 min. 0% to 30% acetonilrile: 175 to 300 min. 30% to 90% acetonitrile: data acquisition rate is 0.5 pointsls; Xaxis tic ma&s are at 1-min intervals. (b) Single dimension

electropherogram of fluorescarnine labeled tryptic digest ovalbumin: P2 flow rate. 0.4 mLImin, bufferA: injection voltage, -3 k t injection time, 3 s: run voltage, -22 kV; run time, 1 min: data acquisition rate, 6 pointsls: capillary, 41 pm i.d.. 150 pm 0.d.. 6.5 cm to detector. 38 cm total length 7a, although not all are fully resolved. No attempt was made to identify any of these peaks. Because this sample is so complex it is difficult to visually distinguish all peaks with this type of data display, and other viewing angles simply obscure different peaks. A more useful type of data display for complex samples is shown in Figure 7b. This is a contour map of Figure 7a. Here the number of peaks is more easily established. and distineuishine . .one peak from another becomes easier. Other disolav ootions are shown in Fieure 8. Each d o t in this figure ibobtained from the s a m e i a t a set displayed in Figure 7. Summing together all points in each electropherogram produces a simulation of the RP chromatogram of the total original sample shown in Figure 8a. The final result of this process appears undersampled since the number of points is equal to the number of CZE injections, and some peaks may only appear in one or two CZE injections. The peak-containing region of Figure 8a appears to he more compressed than that in Figure 6a, the actual first column separation of this sample. Because the first column is undersam-

ANALYTICAL CHEMISTRY, VOL 62. NO. 10. MAY 15. 1990

-

983

be noted that this type of plot is only approximate. Very large peaks lying on the outer edge of the CZE summation range will prejudice the final plot in their favor. Small peaks lying totally within the desired mobility range may be poorly represented in the final plot. Figure 8d is a plot of a single CZE injection from the chromatoelectropherogram in Figure I . Injection number 103, which contains the fastest migrating peak in the sample, is plotted.

DISCUSSION

b -18

CZE MIGRATIONTIME (SEC) 25

35

45

55

-z

5. w 2 c

‘3

k-

3 _I

LL

a a CZE MIGRATION TIME (SEC) Figure 7. (a). Surfer generated chromatoelectropherogram of fluorescaminelabeled tryptic digest of ovalbumin. The condiiions are the same as in Figure 6 with the following exceptions: data acquis#on rate is 5 pointsls; all points from 18 to 55 s are plotted: injections 15 through 150 are pbned. (b) Contour pbt of same data set. Tic marks on injection number axis represent five injections and 5 min each. pled by the CZE system, this undersampling may severely distort the simulated chromatogram. This may account for the different profiles evident in Figures 8a and 6a. In Figure 7a, the base plane can be seen to drift slightly downward. This manifests itself in Figure 8a as a large sloping base line evident most noticeably at the beginning of the plot. Although the simulated method appears to be a poor approximation of the actual first column chromatogram, it may provide a method for roughly estimating system performance and reproducibility. Figure Bb is a simulated electropherogram of the total original sample obtained by summing together corresponding points on each electropherogram or, put another way, summing all points together in each “chromatogram”. The result shows good agreement with Figure 6b, the actual electropherogram of the entire original sample. It should be noted that these two data sets were obtained on different days, as were the data sets compared in Figures Ba and 6a, and that may explain the slight differences in migration times for corresponding peaks. Figure 8c is a chromatogram “slice” of the 3-D chromatoelectropherogram. The width of this slice contains all points in the range of 3C-35 s on the electropherogram. The width of this “slice” may also he enlarged to cover a larger specific mobility range by summing together other “slices”. It should

Clearly RP HPLC and CZE are orthogonal separation methods. As such, their 2-Doperation is effective, yet not trivial to achieve. The fact that they are orthogonal is demonstrated in Figure 7b. The lack of a specific diagonal pattern of peaks shows these methods to be highly orthogonal. In addition, both methods are shown to be able to separate species the other cannot separate. As stated previously, CZE has already been shown to be able to separate that RP HPLC cannot (5-7), hut the method presented here shows the opposite to be true also. While it may be expected that R P HPLC can separate comigrating species a t the electroosmotic flow migration time, examination of Figure 7b shows R P HPLC to separate species that comigrate at a variety of CZE migration times. As stated earlier, the more orthogonal two separation mechanisms are, the more effective and the more difficult will be their coupling. One dissimilarity of these two techniques that must be considered is the use of increasing organic mobile phase in the R P HPLC portion of the system. Although anticipated BS a CZE injection problem, we discovered no such problems when trying to inject samples on the CZE portion of the system from solutions with varying amounts of organic component. Another aspect to consider is that of analysis time. Most of the data presented here demanded over 4 h of data acquisition time alone. While a large amount of information is obtained, shorter analysis times are always desirable. Greater resolution could be obtained on the RP column by use of a different gradient at the expense of a longer chromatographic elution time. However, as stated earlier, computer limits determine the longest amount of time during which the CZE system can operate with a sufficiently high data acquisition rate. CZE resolution could also he improved by increasing the useful capillary length. However, it is vitally important for the CZE injection system to sample the first column as frequently as possible. The shorter the CZE analysis time for each injected sample, the more frequent the CZE sampling rate. Even a t 1-min sampling times, we are undoubtedly undersampling the R P column. Another aspect of incompatibility is that of system volumes. In general it would be desirable for effluent volumes of the first dimension to match the sample volumes of the second dimension as was the case in our earlier work in LC/LC ( I ) . Such a situation provides for a more efficient use of sample. The microbore column used in this work produces far too much volume for the CZE system to sample entirely. Coupling an open tubular LC column or a packed capillary column with CZE would more closely match system volumes but would he more technically difficult. We believe this to be the first example of electromigration injection from a flowing stream. This technique warrants more investigation and characterization, and it has several interesting advantages. First, injections are performed at the grounded electrode and no capillary manipulations are necessary. This feature has obvious safety advantages but also has advantages for applications where capillary manipulation is either difficult or undesired. Another feature of this injection method is that it provides for a convenient method of producing gradients during the electrophoresis. Although not used in the application described here, gradients may he

984

ANALYTICAL CHEMISTRY, VOL. 62, NO. 10, MAY 15, 1990

-.

260

100

RP ELUTION TIME (MIN)

CZE MIGRATION TIME (SEC)

I

d

2 , $

1

$ 1

s

3

U

11

- 1 d 0

1 1

/ I

I

1

'

I

L? 100

-

260

1

1

1

I

'8

25

35

45

55

CZE MIGRATION TIME (SEC)

RP ELUTION TIUE (MIN)

Flgure 8. All plots from data set in Figure 7. (a) Simulated chromatogram of total original sample produced by summation of electropherogram data points from injections 5 through 165, data points from 18 to 55 s included. (b) Simulated electropherogramof total original sample produced by summation of chromatogram data points from injections 15 through 150, data points from 18 to 55 s included. (c) Plot of simulated chromatogram for Points between 30 and 35 s on the CZE miaration time axis. iniections 5 through 165 included. (d) Plot of single electropherogram, number 103, occurring at 198 min on the RP elution time axis.

desired for CZE, or particularly for MECC applications. If P2 is replaced with a pump capable of creating a gradient, the electroosmotic flow will continuously pull this gradient into the CZE capillary during the course of the analysis. It should be noted that while specific migration times and elution times can change for a particular sample from dayto-day, the general appearance and relative intensities, at least to a first approximation, of the 3-D chromatoelectropherogrmn of different tagged samples of the same ovalbumin digest are similar. Tryptic digests of other proteins will be attempted in the near future. The contour type data display may provide a means of obtaining digest fingerprints of different proteins which would be more reliable than single dimension fingerprinting since three variables are obtained for each peak, migration time, elution time, and relative intensity. Although specific migration/elution times can change, it should be possible based upon the general pattern, to determine if one sample is a different protein or if one sample has more or less peaks than another sample. Single amino acid changes in proteins would have a greater possibility of being detected with this 2-D method than reliance on single dimension fingerprints. We have shown that it is possible to couple HPLC with CZE in a two-dimensional, automated manner. Frequent sampling of the R P column by the CZE portion of the system coupled with commercial software provides a means of viewing peak profiles in both separation dimensions. Because these two separation techniques are orthogonal methods, their coupled

operation is effective and may provide for definitive peptide digest maps. Greater resolution and peak capacity are achieved with 2-D systems. Other applications of this particular instrumentation and pairings of other column separations are currently under investigation in our laboratory.

LITERATURE CITED (1) Bushey. M. M.; Jorgenson, J. W . Anal. Chem. 1990, 62, 161-167. (2) Giddings, J. C. HRC CC, J . High Resolut. Chromatogr. Chromatogr. Commun. 1987, 10, 319-323. (3) Davis, J. M.; Gddings, J. C. Anal. Chem. 1985. 5 7 , 2168-2177. (4) Davis, J. M.; Giddings, J. C. Anal. Chem. 1985, 5 7 , 2178-2182. ( 5 ) Puma, P.; Young, P.; Fuchs, M. Poster M-P-126 presented at HPCE 89. Boston. MA. Aoril 10-12. 1989. (6) Nielsen, R . G.; Riggin, R . M.yRickard, E. J . Chromatogr. 1989, 480,

393-401. (7) Grossman, P. D.; Colburn, J. C.; Lauer, H. K.; Nieisen, R . G.; Riggin, R.

M. Sittampalam. G. S.; Richard, E. C. Anal. Chem. 1989, 67. 1 186- 1 194. (8) Janssen, P. S. L.; van Nispen, J. W.; van Zeeland, M. J. M.; Melgers, P. A. T. A. J . Chromatogr. 1989, 470, 171-183. (9) Yamamoto, H.; Manabe, T.: Okuyarna, T . J . Chromatogr. 1989, 480,

277-283.

(IO) Jorgenson, J. W.; Lukacs. K. D. HRC CC, J . High Resolut. Chromarogr. Chromatogr. Commun. 1981, 4 , 230-231. ( 1 1 ) Canfield, R. E.; Anfinsen, C. B . J . Blol. Chem. 1983, 238,

2684-2690. (12) Green, J. S.; Jorgenson, J. W . J . Chromatog. 1986, 352, 337-343.

RECEIVED for review December 27,1989. Accepted February 23,1990. Support for this work was provided by American Cyanamid and the National Science Foundation under Grant No. CHE-8912926.