ANALYTICALBIOCHEMISTRY

185,51-56

(1990)

Characterization of Protein Behavior in HighPerformance Capillary Electrophoresis Using a Novel Capillary System Sally A. Swedberg Hewlett-Packard Laboratories, ChemicalSystemsDepartment, Building 28C, P.O. Box 10490,Palo Alto, California 94303

Received

June

1,1989

Original calculations of over a million theoretical plate efficiency for macromolecular solutes in the open tubular high-performance capillary electrophoresis experiment considered axial diffusion to be the efficiency limiting factor. In practice, interactions of biopolymers, such as proteins, with the capillary wall has had a significant impact on readily achieving high efficiencies for a wide variety of proteins. This paper reports a capillary system in which protein-surface interactions have been minimized, resulting in high efficiencies (>300,000 theoretical plates). This system allows the analysis of a set of protein standards over a wide pl range at neutral pH and moderate ionic strength. The characterization of the behavior of those protein standards in this capillary system is described. 0 1990 Academic Press, Inc.

The potential for high-performance capillary electrophoresis (HPCE)l to provide high efficiency, rapid, quantitative, fully automated analysis for biosciences and biotechnology has attracted significant attention to this technique as witnessed by the explosion of literature on this topic over the past few years. The problem of protein-capillary interactions which degrade the efficiency and reproducibility of open tubular HPCE is now commonly known. Few examples are available which demonstrate the potential of HPCE as a viable approach to protein analysis (l-7). Of these studies, not all give detailed information on the limitations of the approach described (1,6,7) or approaches which were unsuccessful (25). ’ Abbreviations used: HPCE, high-performance capillary electrophoresis; EOF, electroosmotic flow; APF, arylpentafluoro; pZ, isoelectric point; IEF, isoelectric focusing; DMSO, dimethyl sulfoxide; N, designation for theoretical plates, a chromatographic measure of efficiency. 0003-2697/90 $3.00 Copyright 0 1990 by Academic Press, All rights of reproduction in any form

Inc.

reserved.

Using unmodified fused silica capillaries, Lauer and McManigill demonstrated with a selected set of proteins that coulombic repulsion between the negatively charged capillary wall and negatively charged proteins resulted in high efficiencies (4). This study emphasized the potential problem of coulombic attraction as an efficiency-degrading process and verified that very high efficiencies for proteins were obtainable if protein-capillary interactions could be significantly minimized. Additionally, the investigation included an attempt at dynamic deactivation of the fused silica by the addition of cationic species to the electrophoresis buffer. The authors concluded that chemically bonded wall deactivation would be important in providing a separation system which could analyze proteins of any charge. Chemical modifications of the silica capillary wall may result in drastic reduction or elimination of electroosmotic flow (EOF), as discussed by Hjerten (3). Systems which lack an EOF of sufficient magnitude to transport proteins of opposite electrophoretic migration past the point of detection may be limited in their ability to analyze proteins of diverse pl values. To avoid this limitation, the use of low pH (pH 1.5 to 2.0) to promote homogeneity of charge-type and therefore the same direction of migration has been reported (6). There are two potential problems with the use of pH extremes: (i) Loss of separation of similarly charged species may result (6). (ii) At pH extremes, many proteins are significantly denatured. Conditions promoting denaturation may enhance irreversible sticking to capillary walls by exposing additional binding sites (8). In other investigations using chemically bonded wall deactivation for protein analysis, either the efficiencies were significantly lower than expected (1,7) or the elution order of the proteins was unexplained (6,7). 51

52

SALLY

A. SWEDBERG

This investigation utilized a capillary deactivated with a terminal aryl pentafluoro (APF) group. An idealized depiction of this type of modification is shown below: N

-0-S,i

-(CH2)3.N 0 L

A

were purchased as lyophilized powders from Sigma Chemical Co. (St. Louis, MO). Protein mixtures were made in the range of 3-6 mg/ml per protein so that the mixture could range between 21 and 42 mg/ml in total protein. Protein samples were prepared in the electrophoresis buffer used during a run. DMSO was prepared as a 2.5% solution in the running buffer. Running buffers were prepared from reagent grade mono- and dibasic phosphate salts. When additional electrolyte was added, Fluka Chemical Corp. (Ronkonkoma, NY) purissis grade potassium chloride was used. Various concentrations and compositions of buffers used in this study are noted on the electropherograms.

Preparation Capillaries were prepared such that a significant EOF (ca 0.05 cm/s) resulted at neutral pH and moderate ionic strength. A set of proteins commonly used as isoelectric point (~1) standards for isoelectric focusing (IEF) was used. The hydrodynamic radii of this protein set varies in a way that elution order by pl is still anticipated in the HPCE experiment. EXPERIMENTAL

Apparatus The apparatus used has been previously described and consists of a direct current power supply (Hipotronics, Inc., Brewster, NY) and a modified ultraviolet detector (ISCO, Lincoln, NE) (4). All injections were made hydrodynamically by creating a low pressure at the elution end of the column. Pressure difference required and duration of injection were determined using the HagenPoiseulle equation (9). Typical injection time for a 20pm capillary 1 m length with a pressure difference of 20 cm of mercury was 3 s resulting in sample volumes of 0.3 nl. The runs were done in the range of 200-300 V/cm. Data acquisition was accomplished using an HP 3455A digital voltmeter (Hewlett-Packard, Palo Alto, CA) in conjunction with an HP Vectra computing system. Data analysis was accomplished using Asyst graphics and analysis software specially programmed for the analysis of HPCE data. Efficiencies were determined using statistical moments analysis (10). Peak width is corrected by taking into account the different mobilities of the individual protein solutes.

Materials Silica capillary tubing was purchased from Polymicro Technologies (Phoenix, AZ). All other chemicals mentioned in the column modification section were purchased from Aldrich Chemical Co., Inc. (Milwaukee, WI) in the highest grade available. Dimethyl sulfoxide (DMSO), used as a neutral marker, was purchased from Alltech Associates, Inc. (Deerfield, IL). All proteins used

of Silica Capillaries

One-meter lengths of 20-pm-i.d. and 375-pm-o.d. capillary tubing were silylated using 0.1% y-aminopropyltrimethoxysilane solution using standard procedures (11). Solution was pumped through the capillaries using syringe pumps at a rate of l-2 column vol per minutes for 30 min. The capillaries were then flushed with He overnight. Columns were then rinsed with excess dry toluene before a solution of 0.2 M pentafluorobenzoyl chloride in toluene was pumped through the capillaries. The columns were then reequilibrated back to aqueous conditions through toluene, methanol, and finally water washes. After being washed, the columns were equilibrated with the appropriate running buffer and were ready for use. RESULTS AND DISCUSSION Performance Characteristics of APF Columns Figure 1 shows the comparison of seven protein markers on an APF column vs unmodified silica. It can be seen in Table 1 that the deactivation is successful in providing respectable efficiencies. Table 2 shows representative within-day and day-to-day reproducibility. Since the columns are not thermostated in these studies, the EOF can be influenced by the thermal effects on viscosity. For water, viscosity varies inversely with temperature at 2% per degree (12). This value may be used to estimate viscosity effects of aqueous buffers which impact solute mobility (1). The differences between withinday and day-to-day relative standard deviation (RSD or CV%) could be accounted for by the thermal fluctuations in the laboratory. This type of small, random fluctuation is significantly different from the large and nonrandom effects on EOF as a result of irreversible protein accumulation on the capillary wall. It has been observed in this lab and others (1,3,6) that contamination of the capillary wall has a significant impact on EOF. This is probably due to effects on the zeta potential (3,4).

Influence of Ionic Strength

on Protein Mobility

In an electrophoresis experiment, it is anticipated that as ionic strength increases, the mobility of the proteins

HIGH-PERFORMANCE

CAPILLARY

53

ELECTROPHORESIS

TABLE 2

Statistical Evaluation of Column Peformance Evaluated over a 5-Day Period Within

day

Day 2 data

24.0 0.4 0.35

xcte, (min)

SD

RSD

X(L) (min)

SD

RSD

20.8 24.4 25.3 28.8 31.3

20.3 kO.7 f0.7 L1.2 k1.3

+1.4% f2.9% +3.9% _+4.2% k4.2%

19.1

kO.4

23.1 24.1 27.5 29.6

rto.1 fO.l

kO.5% ?0.4% f0.4% 20.7% ?0.7%

28.0

,

I t

B

I

0.3-

Lyso Ribo %P

HuCA BovCA

0.25! 0.2L A 0.15B a O.l-

Between

0.05-

-0.05 !

27.0

day

31.0

35.0

Xi

43.0

47.0

FIG. 1. Elution

profile of seven protein markers and DMSO on an column (A) vs untreated fused silica (B). Conditions of were 200/100 mM phosphate/KC1 solution, pH 7, in 20-pm tubing using applied voltage of 250 V/cm. Detection: on-colat 219 nm. Length of column to detection, 100 cm. L, hen lysozyme; D, DMSO; R, bovine ribonuclease A, T, bovine trypsinogen; WM, whale myoglobin; HM, horse myoglobin; human carbonic anhydrase B; BCA-B, bovine carbonic anhy-

should decrease as a result of the counterion layer shielding the intrinsic protein charge (13). The mobilities of the protein markers were studied over a range of 125 + 500 mM ammonium phosphate buffer, pH 7.0. Mobilities of the proteins were calculated based on an uncharged, unretained EOF marker, DMSO. In Fig. 2, the plot of mobility vs buffer concentration reveals the anticipated decreasing mobility with increasing ionic strength for the seven proteins studied. Figure 3 shows the electropherograms which accompany the data plotted. The behavior of the charged protein solutes as a function of ionic strength was significantly different

Efficiencies

for a Protein

Protein Lysozyme Ribonuclease Trypsinogen Whale myoglobin Horse myoglobin Human carbonic Bovine carbonic

B B

RSD

Ribo

25.1

f1.9

Trw

26.2

r7.6% +7.6%

HuCA BovCA

29.7

22.0 f2.1 k2.2

32.0

f7.1% f6.9%

7.0 I~ 6.0 5.0 4.0

1

w

3.0 I

2.0 1

> 1.04 XE 0.0 -2.0 -

1

-3.0 -

Series PI

anhydrase anhydrase

SD

from the behavior of the uncharged EOF indicator, DMSO. In Fig. 3A no DMSO is indicated since at this ionic strength DMSO, ribonuclease, trypsinogen, and whale myoglobin were not clearly resolved. This lack of resolution obscured the behavior of the proteins. In Fig. 3B, ribonuclease, trypsinogen, and whale myoglobin elute after the EOF market, even though they are positively charged at pH 7.0. This difference in behavior between uncharged species with no counterion layer orga-

;-Lo

TABLE

(min)

I

MINUTES

APF-treated analysis capillary umn uv egg white pancreatic HCA-B, drase B.

f0.2 to.2

X(te)

o-

Typical

Day 3 data

11.0 9.6 9.3 8.1 7.4 6.6 5.9

on

APF

-4.0 -

Capillaries Efficiency 325,000 620,000 525,000 610,000 669,000 524,000 540,000

(N)

-5.0 -6.0

-

-7.0

-

>a

-Lo* 0

, , , , , , , , , , , , , , , , , , , ( 100

200

Buffer

FIG. 2.

Concentration;

MO

400

500

mM

Mobility vs ionic strength of the seven protein markers in a pH 7 ammonium phosphate buffer. Conditions of analyses: 17-pm APF run at voltages of 250 V/cm using on-column uv detection at 219 nm. Length of column to detection, 65 cm.

54

SALLY

A. SWEDBERG

ysis in this study will be expanded upon in a subsequent section. The Influence

0.14

i7.4

21.4

25.4

I 33.3

29.3

0.6

-

of Ionic Strength

on Efiiency

Table 3 shows the relationship between ionic strength and efficiency (N) as they were measured under the conditions of the study. In Fig. 2 an apparent difference in selectivity is demonstrated in the graph for all species studied. Yet, in comparison of Fig. 3A vs Figs. 3B and 3C, it is obvious that not all components are resolved at lower ionic strengths. The data were evaluated by inserting the respective peak widths and elution times into the equation (15)) t2 -

t1

Rs = (w2 + WI)/2 ’

-O.:1.4-

’

2i.2

---xc-+

I

31.9

C

1.2-

ly o.s9 ij

;:I-

The InfEuence of Sample Storage on Analysis

0.2 0 17.0

where Rs = resolution between two Gaussian bands; tl , t2 = elution times for the first and second Gaussian bands being evaluated; wl, w2 = baseline bandwidths for Gaussian bands with elution times tl and t2. On the basis of this analysis, it became evident that the drop in efficiency with lower ionic strength results in lack of resolution of ribonuclease, trypsinogen, and whale myoglobin.

21.0

25.0

llIWTES

29.0

32.9

3

FIG. 3. Comparison of protein elution profiles at 125 mM ammonium phosphate (A), 250 mM ammonium phosphate (B), and 500 mM ammonium phosphate (C), all pH 7. Other conditions of analyses were 17-pm APF capillary using 250 V/cm and on-column uv detection at 219 nm. Length of column to detection, 65 cm. See legend to Fig. 1 for abbreviations.

nization vs charged species with a counterion layer organization is anticipated by electrophoresis theory. The net migration of the protein solutes is the average mobility of both the macromolecule and counterion layer. These mobilities are in opposite directions and the impact of the counterion layer on apparent protein mobility may be considerable (13). It should also be noted that the lysozyme peak appears in Figs. 3B and 3C, but is absent in Fig. 3A. This result was consistently achieved under the conditions of sample preparation and analysis described. Lysozyme self aggregation has been extensively studied and is known to be influenced by multiple factors including protein concentration (14). Aggregation of lysozyme was apparent under the conditions of sample preparation described for this study. It was particularly problematic for the 125 mM phosphate buffer condition, resulting in the apparent absence of this species from Fig. 3A. The topic of the impact of sample preparation and storage on anal-

Samples prepared as previously described for this study displayed obvious signs of change upon either long-term standing (i.e., 2-4 h) at room temperature or after thawing poststorage at -15°C. What was consistently observed in samples aged through either of these processes included: (i) Formation of a precipitate at the bottom of the sample container. (ii) Degradation of peak profile for lysozyme, accompanied by eventual loss of the peak entirely with repeated injection. (iii) The coelution of trypsinogen and whale myoglobin. As previously mentioned, lysozyme self aggregation has been well documented. In addition to protein concentra-

TABLE

3

Efficiency vs Ionic Strength for Three Closely Eluting Proteins Efficiency 125 Ribonuclease Trypsinogen Whale myoglobin

mM

pi

250

(N X 10w3) mM

Pi

500

mM

117

611

773

273

485 477

824 795

195

Pi

HIGH-PERFORMANCE

CAPILLARY

Subsequent experiments where whale myoglobin and trypsinogen were spiked into separate thawed samples confirmed the comigration of these two proteins in such samples. That these observations are sample-related and not column-related is borne out by two factors:

0.45 0.4

A

I

0.35 0.3 j

0.05

7l

-

.I

o-0.05

'V

I

15.0 0.7 ,

Y 18.6

16.6

20.5

**

22.3

55

ELECTROPHORESIS

24.1

(i) Freshly prepared samples run subsequent to aged samples always displayed the previously well-established elution profile (Fig. 4 vs Fig. 3). (ii) The EOF is routinely determined between protein injections. The EOF component monitored in this fashion for the capillary used in this study remained constant except for thermal fluctuations as previously noted. CONCLUSIONS

I

I

-0.1

! 15.0 0.6 ,

16.6

20.5

16.6

I 24.1

22.3 I

I

C I-I

0.1 0 1 12.0

13.6

15.7

17.5

19.3

•/

2 1.1

MINUTES

FIG. 4. Time course of sample aging. Freshly thawed protein standard solution spiked with lysozyme and injected immediately (A); subsequent injection after sample was spiked with whale myoglobin (B); freshly prepared solution (C). Conditions: 250 mM ammonium phosphate pH, 7. For other conditions see legend to Fig. 3.

tion, other factors cluding (16):

may affect lysozyme

aggregation,

in-

(i) Lysozyme self aggregation is particularly sensitive to temperature. In fact, no lysozyme peak was initially evident for thawed samples. (ii) Aggregates present in a sample may act as nuclei to promote further aggregation. Figures 4A and 4B show subsequent injections of a thawed sample freshly spiked with lysozyme. Figure 4C shows a freshly prepared sample containing lysozyme. For the thawed sample, as aggregation proceeded and heterogeneity resulted, the broadened peak shape seen in Fig. 4A degraded into the pattern shown in Fig. 4B. That the broadening occurred immediately upon spiking in the thawed sample is likely due to the presence of aggregates in the thawed sample acting to promote further aggregation. -I I

For HPCE of proteins, the capillary system should not only be defined by the type of surface modification of the silica capillary but also by the electrophoresis medium (i.e., buffer type, concentration, pH, etc.). The capillary system in this study differs from previously described systems in that it demonstrates fairly high efficiencies (>300,000 theoretical plates) at neutral pH and moderate ionic strength for a selected set of proteins. In this system, a considerable EOF is retained and allows for the transport of both positively and negatively charged proteins past the point of detection. Hjerten (3) suggested that EOF may cause peak shape aberrations in systems where nonuniform adsorption of protein solute with the wall has occurred. According to this model, local changes in zeta potential and hence EOF would result. The local inhomogeneity of EOF would cause peak asymmetry to occur with possible lowering of efficiency. In light of this model, Hjert& proposed that systems with no EOF would be necessary. In contrast, Jorgenson and Lukas (20) proposed a model for HPCE separation systems which included EOF. For proteins, the impact of EOF could be to increase efficiency. This was based on an expression of efficiency specifically for HPCE, NC

(P+Pr,m)v 20

’

where N = efficiency, theoretical plates; ~1= electrophoretie mobility of a solute; kosm = electroosmotic mobility; V = electric field gradient; D = diffusion coefficient of a solute. This model assumes that no significant proteincapillary wall interaction will occur during the separation process and can be ignored. In these systems, the magnitude of the EOF may decrease resolution, but not efficiency (20). Previous work in this laboratory proposed an efficiency model for HPCE which allowed for high efficiencies with considerable EOF (17). Using the Giddings

56

SALLY

A. SWEDBERG

equation for capillary chromatography (18) and modifying the coefficients to reflect a plug profile anticipated for HPCE (19) vs a Poiseuille flow profile known for chromatography, the efficiency model for HPCE proposed is

(iv) In this capillary system, there appears to be a relationship between ionic strength and efficiency. Studies are currently underway to characterize k’ as a function of ionic strength in this system. Additionally, some precautions in the analyses of proteins in HPCE are noted. Particular emphasis should be placed on conditions promoting protein stability during sample preparation, storage and analysis. ACKNOWLEDGMENTS

where 1 = injected plug length; L = length of column to detection; DS,r = diffusion coefficient of solute; lz’ = capacity factor (15); u = EOF; r = internal radius of the capillary; kd = first order dissociation constant. According to this model, efficiency is unaffected by EOF if the protein-capillary wall interaction (k’) is very small. Subsequent work inspected a fit to that model using the capillary system described in this study and supported its validity (21). Three criteria for high efficiency separations of proteins in open tubular HPCE remain: (i) Reversible protein-capillary wall interactions must occur during the electrophoresis experiment. (ii) These reversible interactions must be fairly small (k’ 6 0.07) for efficiencies to be fairly high (>300,000 theoretical plates). (iii) To ensure optimum resolution in a system with EOF, enough time must be allowed in the experiment for separation of solute species by electrophoretic migration. Factors affecting this separation time include the applied voltage and the length of the capillary to detection. The present report gives more detailed information about the observed behavior of a set of protein standards in a system providing reversible protein-capillary wall interactions with low k’. The conclusions of the current study are: (i) High efficiency separations of protein solutes over a wide pl range are possible in a capillary system with considerable EOF. (ii) With reversible protein-capillary wall interactions obtainable, reproducibility in these HPCE systems is achievable. (iii) In such a system, protein behavior as anticipated by electrophoresis theory is observed. Primarily, predictable elution order will result, as well as expected behavior of observed mobility as a function of ionic strength.

The author gratefully acknowledges Dr. Mark Bateman for his contribution in development of the data acquisition and analysis system used. The author also thanks Drs. Anthony Fink, Joseph Pesek, and Douglass McManigill for helpful comments on the manuscript.

REFERENCES 1. Jorgenson, J. W., and Lukacs, K. D. (1983) Science 222,266. 2. Lukacs, K. D. (1983) Ph.D. thesis, University of North Carolina, Chapel Hill, NC. 3. Hjerten, S. (1985) J. Chromatogr. 342,191. 4. Lauer, H. H., and McManigill, D. (1986) And. Chem. 5. Rose, D. J. (1988) Ph.D. thesis, University of North Chapel Hill, NC. 6. McCormick, R. M. (1988) And. Chem. 60,2322. 7. Bruin, G. J., Chang, J. P., Kuhlman, R. H., Zegers, J. C., and Poppe, H. (1989) J. Chromutogr. 471,429. 8. Hjer&, S., Rosengren, J., and Pohlman, S. (1974) 101,281. 9. White, F. M. (1979) Fluid Mechanics, McGraw-Hill,

J.

58,166. Carolina,

K., Kraak,

Chromatogr. New York.

10. K&era, E. (1965) J. Chromatogr. 19,237. 11. Weetall, H. H. (1976) in Methods in Enzymology (K. Mosbach, Ed.), p. 139, Academic Press, New York. 12. Hamer, W. J., and Wood, R. E. (1958) in Handbook of Physics (Condon, E. U., and Odishaw, H., Eds.), pp. 4-153, McGraw-Hill, New York. 13. Cantor, C. R., and Schimmel, P. R. (1980) Biophysical Chemistry, p. 677, Freeman, New York. 14. Banerjee, S. K., Pogolotti, A., and Rupley, J. A. (1975) J. Bid. Chem. 250,826O. 15. Kirkland, J. J., and Snyder, L. R. (1979) Introduction to Modem Liquid Chromatography, 2nd ed., Wiley, New York. 16. Fink, A. L., personal communication, University of California, Santa Cruz. 17. McManigill, D., and Lauer, H. H. (1986) 6th International Symposium on HPLC of Proteins, Peptides and Polynucleotides, Baden-Baden, FRG, October 20-22,1986. 18. Giddings, J. C. (196-5) Dynamics of Chromatography; Principles & Theories, Dekker, New York. 19. Whitehead, R., and Rice, C. L. (1965) J. Phys. Chem. 69,4017. 20. Jorgenson, J. W., and Lukacs, K. D. (1981) Anal. Chem. 53,1298. 21. Swedberg, S. A., and McManigill, D. (1989) in Techniques in Protein Chemistry (Hugli, T., Ed.), Academic Press, San Diego.