Chromatographic properties of silica-immobilized antibodies

High-performance immunoaffinity chromatography and chemiluminescent detection in the automation of a parathyroid hormone sandwich immunoassay. David S...
2 downloads 5 Views 720KB Size
Anal. Chem. 1880, 52, 2013-2018

ACKNOWLEDGMENT

(10) Karger, Barry L.; Gant, Russel J.; Hartkoph, Arleigh; Weiner, Paul H. J . Chromatoor. 1976. 728. 65-78. (1 1) Colin, Hen&; Guiochon, deorges J . Chromafogr. 1978, 758, 183-205. (12) Hemetsberger, H.; Maasfeld, W.; Ricken, H. Chromatcgraphia 1976, 9 , 303-3 10. (13) Hemetsberger, H.; Behrensmeyer, P.; Henning, J.; Ricken, H. Chromatographia 1979, 72, 71-76. (14) Locke. David C. J . Chromatogr. Sci. 1974, 72, 433-437. (15) Scott. R. P. W.; Kucerna, P. J . Chromatogr. 1977, 742, 213-232. (16) Knox. John H.; Vasvari, Gabor J . Chromatogr. 1973, 8 3 , 181-194. (17) Hwvath, Csaba; Mebndec, Wayne; Molnar, Imre Anal. Chem. 1977, 49, 142-1 54. (18) Sander, Lane C.; Field, Larry R., unpublished work. (19) Parris, Norman E. 1. duPont de Nemours 8 Co., Inc., Wllmlngton, DE, personal comunicatlon, Dec 18, 1978. (20) Sokolowskl, A.; Wahlund, K. G. J . Chromatogr. 1980, 789, 299-316. (21) Knox, J.; Jurand, J. J . Chromatogr. 1977, 742, 651-670. (22) Twkchett, P.; Moffat, A. J . Chromatogr. 1975, 7 7 7 , 149-157. (23) Lumry, Rufus, Rajender, Shyamala Biopolymers 1970, 9 , 1125-1227. (24) Melander, Wayne; Campbell, David E.; Horvath, Csaba J. Chromatogr. 1978, 158, 215-225.

The authors wish to thank Spectra Physics Inc. for the generous loan of the SP8000 liquid chromatograph, and Dupont, Inc., for a gift of the Zorbax CN column. Special thanks are also due to Dave Herman for his helpful discussions and ideas on the subject.

LITERATURE CITED (1) Cox, C. 0. J . Chromatogr. Sci. 1977, 385-392. (2) . . Colin. Henri: Ward. Norman: Guiochon. Georaes J . Chromatwr. 1978. 749, 169-197. (3) Horvath, Csaba; Melander. Wayne J . Chromatcgr. Sci. 1977, 75, 393-404. (4) Unger, K. K.; Becker, N.; Roumeliotls, P. J . Chromatogr. 1976, 725, 115- 127. (5) Roumeliotis, P.; Unger, K. K. J. Chromatogr. 1978, 749, 211-224. (6) Karch, Karl; Sebestian, Imrlch: Halasz, Istvan J . Chromatogr. 1978, 722, 3-16. (7) Hennion, M. C.; Picard, C.; Cam,M. J. Chromatogr. 1978, 766,21-35. (8) Simpson, C. F. “Practical High Performance Liquid Chromatography”, 1st ed.;Heyden and Son Ltd.: London, 1976; Chapter 7. (9) Horvath, Csaba; Melander, Wayne; Molnar, Imre J. Chromatogr. 1976, 725, 129-156.

-

2013

-

RECEIVED for review February 29,1980. Accepted August 4, 1980.

Chromatographic Properties of Silica-Immobilized Antibodies J. Richard Sportsman and George S. Wilson’ Department of Chemistry, University of Arizona, Tucson, Arizona 85779

Antibodies against protein antigens have been immobilized on silane-treated porous glass and used in a technique designated “high-performancelmmunoaffinity chromatography” (HPIC). A theory has been developed for HPIC which adequately describes the blnding of antigen by immobillred antibody in terms of an apparent binding constant, K’, whose value is flow-rate dependent. The concentration and heterogeneity of blnding sites on the chromatographic column also are successfully predicted by the theory; ail parameters are shown to be in agreement with those determined by batch methods. The suitability of porous silica as a support for the immobiliratlon of proteins and the possibility of antibody fractionation are discussed.

T h e use of antibodies immobilized on insoluble supports for immunoassays and protein purification is now common practice. In the case of the radioimmunoassay (RIA), an immobilized antibody, or immunosorbent, enables a more convenient separation of antibody-bound antigen from unbound antigen; more important is the fact that the precision and sensitivity of the assay is noticeably improved over homogeneous RIA methods ( I , 2). The proliferation of methods for affinity chromatographic purification of innumerable substances has also resulted in the routine use of immunosorbents for these purposes. Comparatively little work has been done in an attempt to better understand the kinetics and thermodynamics of the reactions of antigens with specific immunosorbents ( 3 , 4 ) . Of particular interest is the question of whether immobilization affects the characteristics of immunochemical reactions. Antibodies against protein antigens have been shown to retain considerable binding capacity when attached to gly0003-2700/80/0352-2013$01 .OO/O

cerylpropylsilane (GOPS) derivatives of silica particles and used in a high-pressure liquid chromatographic (HPLC) configuration (5). This appeared to offer a convenient method for studying the reaction of antigen with immobilized antibody, since when so configured, the immunosorbent is reusable and need not be handled or transferred. Furthermore, the HPLC configuration enables one to quantitate the distribution of antigen between solution and binding site by spectrophotometric or fluorometric detection in a flow cell. Finally, it becomes quite simple to vary the conditions of the antigenantibody interaction by adjustment of the mobile-phase composition. Porous silica was chosen as a support because of its low cost and high dimensional stability. It was necessary to use 10 pm diameter particles because the method described below requires maximum surface area and minimum volume. The use of silane-treated porous glass for immobilization of proteins has not generally been as widespread as that of agarose, dextran, and other organic polymers owing largely to some unfavorable reports regarding the supposed lack of stability of protein ligands when attached to silica surfaces (6). Additionally, the tendency of such surfaces to irreversibly absorb proteins has earned porous glass in general a bad reputation (9, in spite of recent work which shows t h a t diol bonded phases on silica provide a suitably deactivated surface for protein chromatography (8, 9). For this study we investigated the binding of two antigens, human immunoglobulin G (IgG) and fluorescamine-labeled beef insulin, by their respective antibodies immobilized on silica. We report the results of experiments which demonstrate that under the appropriate conditions, the binding of an antigen by a “high-performance i m m u n o a f f i n i t y chromatographic” (HPIC) column can be described in terms of an apparent binding constant, which can be related to values 0 1980 American Chemical Society

2014

ANALYTICAL CHEMISTRY, VOL. 52, NO. 13, NOVEMBER 1980

obtained from conventional batch methods. Furthermore, the problems of nonspecific binding of antigen to and leakage of antibody from the surface of the support can be made insignificant.

EXPERIMENTAL SECTION Materials. Human immunoglobulin G (IgG) was obtained from Armour Pharmaceutical Co., Phoenix, AZ, and was dialyzed against several changes of 0.05 M phosphate buffered saline (1%1, pH 7.4 (PBS), prior to use. Goat antihuman IgG was obtained as the immunoglobulin G fraction of the antiserum (lot 284L013) from Kallestad Laboratories, Inc., Chaska, MN. Porous glass spheres (LiChrosphere Si lo00 (10 pm diameter, loo0 A pore size)) were obtained from Altex Scientific, Inc., Berkeley, CA. Glycidoxypropyltrimethoxysilane(GOPS) was obtained from Petrarch Systems, Inc., Levittown, PA. Na’=I, 25 mCi/mL, was obtained from ICN Pharmaceuticals, Irvine, CA. Crystalline bovine zinc insulin was obtained from Lilly Research Laboratories, Indianapolis, IN. Fluorescamine (4-phenylspiro[furan-2(3H),1’(3’H)-isobenzofuran]-3,3’-dione) was obtained from Sigma Chemical Co., St. Louis, MO. Monoclonal antibodies against beef insulin were the gift of Joyce Schroer of the National Institutes of Health, Bethesda, MD. The monoclonal preparation was obtained by cell fusion of beef insulin immune BALB/c lymph node cells and the myeloma cell line NSI/l-Ag4-1 that makes only cytoplasmic MOPC 21 light chain. These so-called hybridoma antibodies bore the designation BE CB6 and were raised in ascites (10). The hybridomas had been purified from serum by ammonium sulfate precipitation of globulin fraction three times followed by extensive dialysis against phosphate-buffered saline. Apparatus. Spectral measurements were made by using a Cary 219 UV-VIS spectrophotometer. The spectrofluorometer was a Perkin-Elmer Model 204A equipped with an ultramicroflow cell. Excitation and emission wavelengths were 283 and 335 nm, respectively, for monitoring IgG natural fluorescence. Insulin, whose native fluorescence is weak, was labeled with fluorescamine as described below and monitored at excitation and emission wavelengths of 390 and 470 nm, respectively. Counting of all radiolabeled materials was performed on a Packard Model 9012 multichannel analyzer system. Procedures. [‘251]-IgGwas prepared by using the iodine monochloride method of Helmkamp et al. (11)and was shown to be reactive toward antibody by double diffusion in 170agar. The specific activity was 1600 Ci/mol, indicating that only 1 IgG molecule in 1000 is labeled. This low specific activity provided sufficient sensitivity for our purposes and minimized radiation damage. Insulin* was prepared by a method similar to that given by Sung et al. (12). The number of fluorophors attached per molecule of insulin was determined to be 0.7 by using a molar absorptivity of 1.7 X lo4 M-’ cm-’ for fluorescamine at 390 nm ( 1 3 ) . For fluorescamine-conjugated insulin it was necessary to determine protein concentrations by Folin-Lowry assay (14) rather than by simple absorbance at 280 nm because of apparent interference by the fluorophor. The porous glass spheres (1g) were silanated after first being treated with hot chromic acid cleaning solution followed by rinsing in hot 1M HNOB(100 mL) and water (100 mL). The silanation was performed by suspending the cleaned glass spheres in a 10% aqueous solution of GOPS, degassing with ultrasonic vibration 10 min, and then allowing the reaction to proceed at 90 “C for 2 h, during which time the pH was maintained at 3.0with 1 M HC1 (9). The glass spheres were then collected on a mediumporosity glass frit, rinsed with 10 mL of HzO, and dried in vacuo overnight at 105 “C. The GOPS derivatized glass spheres were then subjected to quantitative periodate oxidation (15)from which a value of 340 pequiv of diol/g of glass was obtained. The resulting aldehydic glass particles (1 g) were collected on a medium-porosity glass frit and washed with water (200mL) and 0.1 M borate-buffered saline pH 8.5 (BBS), 200 mL. The spheres were then suspended in 3 mL of BBS containing 5-10 mg of antibody protein/mL. The mixture was degassed with ultrasonic vibration for 15 min and then allowed to mix by gentle inversion for 20 h at 4 “C. At this time, NaBH4 (5 mg) was added at three 15-min intervals to effect reduction of the Schiff base formed between glass-bound aldehyde

reservoir E

switching

valve

-Q

injection p o r t

c o / umn

Figure 1. Block diagram of HPIC apparatus, detector fluorometric (variable excitation/emission).

and amino groups on the protein (16). The amount of protein coupled to the glass was determined by measuring the absorbance at 280 nm of the solution before and after coupling. These results were confirmed by repeating the analysis using a method which we developed for this purpose (17,in which the protein is hydrolyzed from the glass in 6 N HCl and the resulting amino acids are determined spectrophotometrically, For the goat antihuman IgG, both methods gave a value of 6.8mg of protein/g of glass (dry weight basis); in the case of the BE CB6 antiinsulin antibodies, a value of 3.2 mg/g was determined by protein balance method whereas the hydrolysis/spectrophotometric method gave a value of 2.9 mg/g glass. Chromatographic Method for Studying Immunosorbent-Antigen Interaction. Stainless steel columns, 4 cm X 2.1 mm i.d., were packed by slurrying the immunosorbent in 50% saturated sucrose. For most studies a dual-reservoir (“diisocratic”) chromatograph was used, a diagram of which is shown in Figure 1. The column void volume was determined by measuring the elution volume of 10 pL of bovine serum albumin (BSA) or pure water, both of which are nonretained. Method for IgG HPIC. Initially, the pH 7.4 buffer supplied from reservoir A flows through the column at a constant flow rate. When a 10-pL sample of human IgG is injected onto the column, a peak which elutes with the void volume of the column is seen. This is represented by the event marked “A’ on the chromatogram shown in Figure 2. When the mobile phase is changed to pH 2.2 buffer by switching to reservoir “B”, the antigen-immunosorbent complex formed during event “A’ dissociates (5,18),giving rise to the peak which follows event “B” on the chromatogram. The eluting antigen is detected by monitoring the native fluorescence of the protein. Since this fluorescence is quenched at pH 2.2, a postcolumn pH-adjusting buffer (0.3M phosphate, pH 8.0)is introduced as shown in Figure 1. It was determined, by Folin-Lowry protein assay of collected fractions, that the peak areas on the chromatograms may be correlated to the mass of eluted antigen. Method for Insulin*. The chromatographic behavior of the insulin*-antiinsulin system is shown in Figure 3. In this case, elution of the antibody-bound insulin* (Le., event “B”) is accomplished by applying 25% acetonitrile in PBS from reservoir B to the column, since this elution could not be obtained by using an acid buffer as in the case of the IgG system. Naturally, the postcolumn pH adjustment was not needed.

ANALYTICAL CHEMISTRY, VOL. 52, NO. 13, NOVEMBER 1980

2015

88-

11

60-

48

I),

E

1 1 1 1

1 1 1 1

1

1

1

~ 1 1 1 1

1 1 1 1

1

1

1

~

TIHE, MUMITES

Figure 2. HPIC chromatogram for IgG system: immunosorbent, antihuman-IgG attached to LiChrosphere Si 1000; column 2 mm i.d. X 4 cm; flow rate 0.5 mL/min; mobile phase “ A ” PBS, pH 7.4: mobile phase “B” 0.01 M phosphate buffer, pH 2.2; antigen (IgG) concentration 1.4 mg/mL; volume injected 10.0 pL; fluorometric detection, excitation 283 nm, emission 335 nm.

8

3

4

5

TIME, WNJTES

Flgure 3. HPIC chromatogram for insulin‘: immunosorbent, BE CB6 monoclonal antiinsulin attached to LiChrosphere Si 1000; column 2 mm i.d. X 4 cm; flow rate 0.30 mL/min; mobile phase “A” PBS pH 7.4; mobile phase “B” 28% acetonitrile in PBS; antigen (insulin’) concentration 4.0 pg/mL; volume injected 50 FL; fluorometric detection, excitation 390 nm. emission 470 nm.

Initially it was found necessary to collect fractions and determine by spectrofluorometry the amount of insulin* in each, since under the conditions of the experiment, the fluorescence yield of insulin* differs for the two eluents. It was determined that the peak area of that peak which appears just after injection (Le., event “ A ’ in Figure 3) must be multiplied by a response factor of 0.51 in order to be correlated to the area of the subsequent peak (peak “B”). This response factor was found to be constant for the range of concentrations used. Determination of Apparent Binding Constants. Studies were made for both systems by using varying concentrations of antigen at a given flow rate. It was observed that the ratio of the amount of antigen which elutes immediately upon injection to that which subsequently elutes upon changing the mobile phase is a hyperbolic function of concentration, as shown in Figure 4 for the IgG system. The similarity of this curve to those obtained in ordinary immunoassay procedures led us to describe the data in terms of an apparent binding constant, K’, given by

B is the moles of antigen bound by the immobilized antibody upon injection, F is the moles of antigen which pass unretarded through the column, (-Ab) is the moles of immobilized antibody binding sites left unoccupied during the interaction, and brackets signify

for

CBI X 1 0 * MOL/L Flgure 5. Scatchard plot (eq 5) for data of Figure 5, antigen-lgG flow rate 0.30 mL/min. Dashed line shows linear portion of curve extrapolated to abscissa.

that the parameters are expressed as molar concentrations. Since the unretarded antigen elutes a t uo, the void volume of the column, the interaction with immobilized antibody is assumed to occur in an effective volume equal to this void volume. Taking (Ag), to represent the total moles of antigen injected, the concentration terms of eq 1may then be rendered into the following relations:

(3)

[-Ab] = (-Ab)/uo

(4)

Since (Ab), defines the total moles of antibody binding sites on the column, then by invoking the appropriate mass balance relationships, we derive an equation similar to that of Scatchard (19)from eq 1-4. B / F = [Abt]K’- [BIK’

(5)

A plot of the area ratio of the peaks labeled “A” and “B” in Figures 2 and 3 against [B] gives a slope of -K‘and an intercept on the abscissa from which [Ab,] may be calculated. For example, when the data of Figure 4 are plotted in this fashion, the nonlinear curve of Figure 5 results. The hyperbolic shape of this plot indicates that a heterogeneous population of binding sites is present (I9).

2016

ANALYTICAL CHEMISTRY, VOL. 52, NO. 13, NOVEMBER 1980

: i Table I. Apparent Binding Constants for Antibody- Antigen Reaction

immobilized

D n t i body

0 0 0

CBI X ! 0 * M O L / L Scatchard plot (eq 5) for binding of [’“I]-IgG immunosorbent (antihuman IgG attached to LChrosphere Si 1000) in a batch experiment. Figure 6.

anti-IgGglycophase ant i-IgGglycophase anti-IgGglycophase anti-IgGgly cop h ase an ti-IgG glycophase CBGdglycophase

antigen

IgG

flow rate,a mL/ min K , M-’ 0.1

ub

(Ab),,‘ mg/g

0.52

0.10

0.77

0.072

0.75

0.039

0.50

0.027

0.73

0.098

0.94

0.19

5.6 x

107

IgG

0.3

IgG

0.8

1gG

2.0

[”’I]-IgG

0

1.7 x 107 7.3 X 1O6 1.6 x 105

4.4 X 107

insulin*

0.3

8.8 X 105

a A zero flow rate indicates a static or batch measurement. Heterogeneity parameter determined from Sips plot. Expressed as weight of specific active antibody per gram dry weight of glass; value obtained from eq 5. Monoclonal antiinsulin obtained from NIH.

4

\

WASH NUMBER

Leakage of immobilized [”’I]-IgG. Rate is expressed as t h e ratio of supernatant/solid-phase activity. One gram of immunosorbent was suspended in 4 mL of PBS whh 24 h of agitation between washes. Figure 7.

The data can be plotted according to an equation similar to Sips (13) r log -= a log K’ + a log [F] n-r where r = [B]/[Ab,], n is the number of antigen binding sites per molecule of antibody, and a is the so-called “heterogeneity index”, whose value is unity for a homogeneous population of binding sites and less than unity according to the extent of population heterogeneity (19). Extrapolation to the abscissa of the linear portion of the curve of Figure 5 yields [Ab,]; from this, r and thus log ( r / ( n- r ) ) may be calculated. When log (r/(n- r ) ) is plotted vs. log [F] using data as in Figure 5 , the value of log K‘ is then obtained as the ordinate intercept while a is the slope. Values of the mean equilibrium constant were determined from eq 5 and 6 in a “static” or batch mode. The immunosorbent is uniformly suspended in a known volume of PBS. Aliquots (100 wL) of suspension are pipetted into each of several plastic tubes which have been previously treated with BSA to minimize nonspecific adsorption. Various concentrations of [ 1251]-IgGare added to the tubes which are capped and equilibrated overnight at 4 “C. The tubes are then centrifuged, and the supernatant is collected in separate tubes as the “free” fraction. The bound fraction is washed once with water and immediately centrifuged. The amount of antigen in each fraction is determined by counting the 35-keV y-ray activity. The data are then evaluated by using eq 5 and 6. Figure 6 is a Scatchard plot of these data.

RESULTS A N D D I S C U S S I O N Evaluation of Immunosorbent Stability. Figure 7 shows the results of a study designed to evaluate the stability of

0 8

0.5

I 0 FLOU RATE,, ML/MIN

I .5

2.0

Figure 8. Flow rate dependence of apparent binding constant K’: antigen, human IgG; immunosorbent,antihuman IgG bound to LiChrosphere Si 1000. Conditions are as in Figure 2.

immunosorbents prepared as described above. An immunosorbent was prepared from [ 1251]-IgGand subjected to a series of successive washes, each of which consisted of end-over-end mixing for 24 h in 4 mL of PBS. The amount of activity found in the wash solutions relative to the immunosorbent was measured after each wash. The initial losses on washes 1-4 undoubtedly represent primarily desorption of noncovalently bound protein. Thereafter the leakage declines to an insignificant level. It can be calculated that through the seventh wash, the total loss is less than 12% of the amount of protein initially present, which was 5 mg/g of glass. The end-over-end mixing exerts considerable mechanical stress upon the individual particles. It seems unlikely that the chromatographic configuration in which the immunosorbent is used would result in more rigorous treatment, and i t has indeed been observed by us and others (5)that the activity of such antibody columns is stable for several months provided that the column is stored a t 4 “C when not in use. Nonspecific Adsorption. Regnier and Noel (9) and Karger e t al. (8) have reported that GOPS bonded phases exhibit little tendency to irreversibly adsorb immunoglobulins. Our own recovery studies showed that at those antigen concentrations which were used in these experiments, more than 97% of the protein injected could be subsequently eluted. Furthermore, nonspecific binding of bovine serum albumin

ANALYTICAL CHEMISTRY, VOL. 52. NO. 13, NOVEMBER 1980 4

S-

lA

\

m

2

e

\

,

,

1

1

,

1

,

1

,

,,,,

,

2

,

3

,

I

1

I

4

I

,

I

6

Flgure 9. Scatchard plot (eq 5) for binding of insulin’ to immunosorbent (BE CB6 monoclonal antiinsulin attached to LiChrosphere Si 1000). Conditions are as in Figure 3.

(BSA) was not observed on either column under normal operating conditions, confirming the results of Mosbach and co-workers ( 5 ) . Verification of Theory. T h e values of the apparent binding constant K’and heterogeneity index a for the anti-IgG column were determined a t several flow rates. The results are shown in Table I along with the batch results. Figure 8 shows the flow-rate dependence of K’. It is to be noted that the least-squares straight line extrapolates to a value for log K’of 7.8 a t the ordinate intercept. This compares favorably to the batch (static) value of 7.6 h 0.2 (one standard deviation) for the same immunosorbent. A comparison of Figures 5 and 6 further serves to indicate the validity of the chromatographic method for determining K’. It can be seen that both curves display a similar hyperbolic shape. Inspection of Table I shows that the values of the heterogeneity index for these two sets of data are indeed similar suggesting that the heterogeneity is not an artifact of the chromatographic process. In all cases the IgG-anti-IgG system shows considerable heterogeneity as seen in Table I. The values for [Ab,] as shown in Table I for the IgG system seem to increase with decreasing flow rate. This indicates that the reaction of antigen with immobilized antibody binding sites is sufficiently slow that the residence time of the antigen in the column dictates the availability and hence the apparent concentration of the binding sites. The limiting value for [Ab,] appears to be 0.10 mg/g as shown in Table I. A value of 0.82 mg of specific antibody/g of immunosorbent was determined by analysis of the immunosorbent. The specific antibody content of the antiserum as supplied was assumed to be 12.0% of the total protein on the basis of data provided by the manufacturer. It can be concluded that the coupling procedure results in loss of activity (20), such that 11% of the original antibody activity remains. This is c o n f i i e d by the results of the batch experiment, from which a value for [Ab,] of 0.098 mg/g glass was calculated (Table I), again representing a recovery of 11% of the original antibody activity. It is interesting to compare the above results for the IgG system with those obtained for the insulin* system, whose Scatchard plot (eq 5) is shown in Figure 9. This plot demonstrates essentially linear behavior as opposed to the distinctly hyperbolic shape of the Scatchard plots (Figures 5 and 6) for the IgG system. As shown in Table I, the Sips plot of these data yields a value for K’ of 8.8 X lo6 M-’ with a heterogeneity index of 0.94. This is consistent with the fact that the antibody is monoclonal and possesses a single type of binding site (18). This hybridoma antibody has a binding

2017

affimity for beef insulin of 1.5 X lo6 M-’ as determined by cold competition radioassay (21). T h e monoclonality of the antibody is proven both by the linearity of a Scatchard plot and by the isoelectric focusing pattern which this antibody demonstrates (22). Thus the chromatographic method for determination of K’ adequately describes the behavior of the insulin* system. The difference between the value of the homogeneous equilibrium constant and the value of K‘ is undoubtedly due to the fact that K’ was here determined at a finite flow rate. By analogy with Figure 8, a reduction in the value of K’ by about a factor of 2 would be expected as the flow rate goes from 0 to 0.3 mL/min. This is indeed what is observed. The comparison of a binding constant for native insulin with that for fluorescamine-labeled insulin might appear questionable, but there is good reason to expect that this comparison is valid. Fluorescamine becomes attached solely to primary amino groups, of which there are only three in the beef insulin molecule. T h e antibody appears to be directed against a hydrophobic region on the insulin molecule, since dissociation of this antigen-antibody complex is achieved by increasing the organic solvent content of the solution. Such a hydrophobic region would not be expected to bear a primary amine. Given the sparsity of suitable positions on the insulin molecule for the fluorophor and the fact that the fluorophor/insulin molecular ratio is 0.7, it is most probable that the fluorophor does not affect the antigenic site of insulin. The value for [Ab,] shown in Table I for the insulin system is 0.19 mg antibody/g of glass; using the assay value of 2.9 mg total antibody/g, the activity remaining is calculated to be 7 YO,although this figure is probably biased toward the low side since the value of [Ab,] was determined a t 0.3 mL/min. Work is currently being done in our laboratory to further clarify the nature of the flow-rate dependence of parameters calculated by this chromatographic method. In conclusion it can be said that the binding of an antigen by an HPIC column may be described in terms of an apparent binding constant, whose value is dependent upon the flow rate. Further, a t diminishing flow rates, this binding constant appears to approach the value of an equilibrium constant for the immunosorbent in the batch mode and for the antibody in a homogeneous phase reaction. The values of other parameters determined by this chromatographic method, that is, the heterogeneity index a and total column binding site concentration [Ab,], also agree with their corresponding batch values. It is expected that this method will prove useful for investigating the kinetics and thermodynamics of antigen-antibody reactions in heterogeneous reaction phases. Such fundamental knowledge will be indispensible for optimizing the performance of immunoaffinity techniques. T h e model described herein should provide a mathematical basis for probing the conditions, of both mobile phase and immunosorbent, which can minimize unwanted adsorption effects while maximizing the recovery of immunosorbent-bound ligands, an area in which sound theoretical understanding is noticeably lacking (23). It is also possible, by careful adjustment of mobile-phase conditions, to fractionate heterogeneous antibody populations rapidly and with good resolution. Because recovery rates are both high and reproducible, this technique also provides the possibility of selective and sensitive bioassays. ACKNOWLEDGMENT We thank Perry Cole and Art Barry for valuable advice. The gift of bovine insulin from Mary Root of Lilly Research Laboratories is gratefully acknowledged. LITERATURE CITED (1) Wae, L. In “Radioimmunoassay Methods”;Klrkham, K. E., Hunter, W. M.. Eds.; Churchill Llvlngston: Edinburgh, 1971; p 405.

2018

Anal. Chern. 1980, 5 2 , 2018-2022

(2) Daughaday, W. H.; Jacobs, L. S. In “Principles of Competiive ProteinBinding Assays”; Odeli, W. D.; Daughaday, W. H.. Eds.; J. B. Lippincott: Philadelphia, 1971; p 405. (3) Hertl, W.; Odstrchel, G. Mol. Immunol. 1979, 16, 173-178. (4) Eilat, D.; Chaiken, I. M. Biochemistry 1979, 78. 790-795. (5) Ohlson, S.; Hansson, L.; Larsson, P. F.; Mosbach, K. FEBS Lett. 1978, 93, 5-9. (6) Cuatrecasas, P. Proc. Abtatl. Acad. Sci. U.S.A. 1972, 69, 1277. (7) Messing, R. A. J. Am. Chem. SOC.1969, 91, 2370-2371. (8) Schmidt, E. E., Jr.; Giese, R. W.; Conron, D.; Karger, B. L. Anal. Chem. 3980, 52, 177-182. (9) Regnier, F. E.; Noel, R. J. Chromatogr. Sci. 1976, 14, 316-320. (10) Schroer, J. A.; Kim. K. J.; Rosenthai, A. S. fed. Roc., Fed. Am. Soc. Exp. Biol. 1979,39(3), 1421. (11) Helmkamp, R . W.; Contreras, M. A.; Bale. W. F. Int. J. Appl. Radiat. m. 1967, 18, 737-746. (12) Sung, M. T.; Bozzola. J. J.; Richards, J. D. Anal. Biochem. 1978, 8 4 , 225. (13) Richardson, J. H. Anal. Biochem. 1977,83. 754-762. (14) Peterson, G. L. Anal. Biochem. 1977, 83, 346-356. (15) Siggia, S. ”Quantitative Organic Analysis via Functional Groups”; Why: New York, 1949; pp 8-9.

(16) Royer, G. P.; Liberatore, R . A.; Green, G. M. Biochem. 6iophys. Res. Commun. 1975, 64. 478-484. (17) Smtsman. J. R.; Wilson, G. S., unpublished work, Universtty of Arizona. ’ Tucson, AZ. 1980. (18) Weliky. N.; Weetali, H. H. Immunochemistry 1972. 9 , 967-978. (19) Richards, R . F.; Rosenstein, R. W.; Varga, J. M.; Konigsberg, W. H. I n “Immunogiobulins”; Litman, G. W.. Good, R . A., Eds.; Plenum: New YOrk. 1978: OD 138-141. (20) Line,’W. F.;Sbgel, S. J.; Kwong, A,; Frank, C.; Ernst, R. Clin. Chem. ( Winston-Salem, N.C.) 1973, 19. 1361-1365. (21) Farr, R . S. J. Infect. Dis. 1958, 703,239. (22) Schroer, J.. personal communication, 1980, National Institute for Allergies and Infectious Diseases, NIH, Bethesda, MD. (23) Kristiansen, 1.In “Immunoadsorbents in Protein Purification”; Rmslahti. E., Ed.; University Park Press: Baltimore, 1976; p 22.

RECEIVED for review May 12, 1980. Accepted July 29, 1980. This investigation was supported in part by National Science Foundation Grant PCM 78-13226.

Borate Complex Ion Exchange Chromatography with Fluorimetric Detection for Determination of Saccharides Kenneth Mopper



Department of Analytical and Marine Chemistry, University of Goteborg, S-4 72 96 mteborg, Sweden

Rodger Dawson, Gerd Liebezelt, and Hans-Peter Hansen Sonderforschungsbereich9 5 and Institut fur Meereskunde, University of Kiel, 2300 Kiel, Federal Republic of Germany

An analytical technlque is described for the detection of fluorescent products formed by the reaction of reducing and nonreducing saccharldes with ethylenediamlne. The saccharides are separated by anion exchange chromatography In borate media with the amine reagent already present in the mobile phase. Fluorescent products are formed on-line In a postcolumn reactor whose temperature may be varied to optimlre the response toward reducing and/or nonreduclng saccharides. The detection limit Is under 1 nmoi for most saccharides. Appllcatkns to environmental and natural product samples involving little or no sample pretreatment are given.

Borate complex anion exchange chromatography has been shown to be an extremely powerful tool in the rapid separation of complex mixtures of carbohydrates ( 1 , 2 ) . Almost any combination of sugars can be resolved by adjustment of chromatographic conditions, e.g., column temperature, buffer molarity, and pH. Thus, mixtures of 15-25 sugar components have been successfully separated ( 2 , 3 ) . Furthermore, a wide variety of sample materials, e.g., urine, blood plasma, acid hydrolysates, beverages, etc., can be analyzed either directly or after only a simple filtration step. This is advantageous since tedious and sometimes contaminating clean-up, desalting, and derivatization steps (as required by gas or partition liquid chromatography) are often superfluous. College of Marine

19958.

Studies, University of Delaware, Lewes, DE 0003-2700/80/0352-2018$01 .OO/O

Despite these positive attributes, automatic sugar analyzers have not received the widespread acceptance as,for example, automatic amino acid analyzers. The main hindrance appears to have been the lack of a safe and sensitive detection system applicable to all classes of sugar compounds. Reagents for postcolumn reaction systems may be divided into two broad categories: corrosive and noncorrosive. An example of the former is concentrated sulfuric acid-orcinol ( 4 , 5 ) . This type of reagent reacts with most classes of sugars and is fairly sensitive (nanomole level); however, its corrosive nature generally detracts from its use in routine clinical and industrial analyses. Furthermore, precipitation problems leading to blocked reaction coils and leakage have been reported (6, 7). Noncorrosive colorimetric reagents, such as copper(I1)-aspartic acid-bicinchoninate (7,8), copper(I1)-neocuprion (9),and p-hydroxybenzoic acid hydrazide ( l o ) ,are also sensitive but react only with reducing sugars;nonreducing oligosaccharides, e.g., sucrose, raffinose, etc., are not detected. Honda et al. (11,12) reported that reducing sugars form fluorescent products when reacted with ethylenediamine (EDA) in weakly basic phosphate buffer solutions. The present investigators have adapted this reagent to the on-line fluorimetric detection of sugars in borate complex anion exchange chromatography. The diluted reagent in buffer is safe to handle and, by proper control of the reaction conditions, can be made to react with nonreducing oligosaccharides as well as reducing sugars. Furthermore, EDA can be added directly to the mobile phase, the borate buffer, thereby eliminating the need for a separate reagent stream after the chromatographic column. 0 1980 American Chemical Society