Proc. Natl. Acad. Sci. USA Vol. 77, No. 3, pp. 1632-1636, March 1980 Medical Sciences Prediction of peptide retention times in high-pressure liquid chromatography on the basis of amino acid composition (lipophilicity/separation techniques) JAMES L. MEEK Laboratory of Preclinical Pharmacology, National Institute of Mental Health, Saint Elizabeths Hospital, Washington, D.C. 20032 Communicated by Bruce Merrifield, December 17,1979

ABSTRACT Analysis of peptides by reverse-phase high- on octadecylsilyl silica gel is a quite different process from oc- pressure liquid chromatography would be simplified if retention tanol/water partition and from the unavailability of hydro- times could be predicted by summing the contribution to re- groups of the peptides. It is the tention of each of the peptide's amino acid side chains. This phobicity data for terminal paper describes the derivation of values ("retention coeffi- purpose of this paper to show that "retention coefficients" can cients") that represent the contribution to retention of each of be derived directly from HPLC data for all amino acids and the common amino acids and end groups. Peptide retention end groups such that the retention time of a peptide can be times were determined on a Bio-Rad "ODS" column at room predicted from the sum of the retention coefficients for each temperature with a linear gradient from 0.1 M NaCIO4, pH 7.4 amino acid and end group. or 2.1, at 0 min to 60% acetonitrile/0.1 M NaCIO4 at 80 min. The NaClO4, a chaotropic agent, was added to improve peak shape and to minimize conformational effects. Retention coefficients MATERIALS AND METHODS for the amino acids were computed by using a Hewlett-Packard Peptides were obtained from Sigma and Peninsula Laboratories 9815A calculator programmed to change the retention coeffi- cients for all amino acids sequentially to obtain a maximum (San Carlos, CA). HPLC grade acetonitrile was obtained from correlation between actual and predicted retention times. Cor- Fisher. The HPLC system was assembled from modular com- relations of 0.999 at pH 7.4 and 0.997 at pH 2.1 were obtained ponents: a one-chamber glass gradient maker, a pump (Milton for 25 peptides including glucagon, oxytocin, [Metlenkephalin, Roy, Riviera Beach, FL), and a sample valve (Rheodyne, neurotensin, and somatostatin. This high degree of correlation Berkeley, CA). The column effluent was passed in series suggests that, for peptides containing up to 20 residues, retention through a variable wavelength photometer (Altex, Berkeley, is primarily due to partition processes that involve all the resi- dues. Although steric or conformational factors do have some CA) and a filter fluorometer (Farrand, Valhalla, NY). The effect on retention, the data suggest that under the above mobile phase gradient normally used was from 0.1 M chromatographic conditions the retention of peptides con- NaCIO4/0% acetonitrile at 0 min after injection to 0.1 M taining up to 20 residues can be predicted solely on the basis of NaCl04/60% acetonitrile (vol/vol) at 80 min. For separations their amino acid composition. This possibility was tested by at pH 7.4, the starting buffer contained 5 mM phosphate buffer, using data taken from the literature. pH 7.4. For operation at pH 2.1, both starting and final buffers contained 0.1% phosphoric acid (3). The possibilities for separating and isolating small peptides have Linear gradients from 100% A to 100% B are generated if a been markedly improved by the introduction of reverse-phase solution A is placed in a mixing chamber and then B is added high-pressure liquid chromatography (HPLC) (1-5). This at 1/2 the rate at which the mixture is withdrawn (6). The technique depends upon the hydrophobic interactions between of a 3 X 13 cm with two a hydrocarbonaceous column and the peptides to be separated: gradient maker consisted glass cylinder the more hydrophobic (lipophilic) the compound, the stronger Teflon three-way stopcocks and a magnetic stirrer. To generate its retention by the column. To elute strongly retained com- the gradient, the chamber was filled with 40 ml of starting pounds, aqueous solutions (the "mobile phase") containing a buffer. Final buffer was then added to the chamber at 0.5 large amount of organic solvent must be pumped through the ml/min, while the mixture was pumped into the column at 1.0 column. Choice of the optimum mobile phase and chromato- ml/min. trial and Peptides were detected by absorbance'at 200 or 220 nm or graphic conditions for given peptides can be found by by fluorescence [after the primary amino groups of the peptides error after qualitatively examining the balance of hydrophobic had reacted with fluorescamine (7)]. With a mobile phase of and hydrophilic amino acids present in the peptide. However, pH 7.4, the column effluent could react directly with fluores- as noted by Molna'r and Horvath (1), it should be possible to camine, because the pH optimum for many peptides is near obtain quantitative estimates of the hydrophobicity of the ml of amino acids contained in a peptide, which will reflect their neutral (8). The fluorescamine (10 mg/100 acetonitrile) of was added to the effluent at 0.1 ml/min. Continuous neutrali- retention on the reverse-phase column. Estimates hydro- zation of the pH 2.1 mobile phase was achieved by adding the phobicity based on octanol/water partition coefficients exist organic base imidazole (1.0 M final concentration) to the for many but not all amino acids. By using such values, O'Hare fluorescamine/acetonitrile mixture. and Nice (2) noted that the retention order for small peptides The retention coefficients were computed by repetitive re- was generally correlated with the sum of the values for the most amino acid were hydrophobic residues of the peptides. The many deviations they gression analysis: values for each successively the changed by 0.2 min until maximum correlation between actual found between the observed order of elution and lipophil- and predicted retention times was obtained. Starting values for icity estimates presumably derive from the fact that retention the retention coefficients of the neutral and hydrophilic amino acids were initially assumed to be zero. Starting values for the The publication costs of this article were defrayed in part by page amino acids were obtained the retention charge payment. This article must therefore be hereby marked "ad- lipophilic by plotting vertisement" in accordance with 18 U. S. C. §1734 solely to indicate this fact. Abbreviation: HPLC, high-pressure liquid chromatography. 1632 Downloaded by guest on September 28, 2021 Medical Sciences: Meek Proc. Natl. Acad. Sci. USA 77 (1980) 1633 times of oligomers (e.g., diphenylalanine, triphenylalanine, were only slightly retained. Small peptides with lipophilic side tetraphenylalanine, etc.) vs. the number of residues; the slope chains (e.g., triphenylalanine) and most of the larger biologi- of the plot equals the retention per residue. To compute the cally active peptides required much higher concentration of retention coefficients, a Hewlett-Packard 9815A calculator was acetonitrile for elution. Acidification of the mobile phase in- programmed to store these starting retention coefficients for creased the retention of peptides with free terminal carboxyl the 26 amino acids and end groups, to store the actual retention groups ([Metlenkephalin and angiotensin II) or with acidic times for the 25 peptides studied, and then to calculate pre- residues [gastrin-(12-15)]. Most peptides with masked carboxyl dicted retention times for these peptides by summing the re- or carboxyl and amino groups (thyrotropin-releasing factor, tention coefficients for each amino acid contained. After cal- oxytocin) had similar retentions at both pHs. Peptides con- culating the correlation coefficient between predicted and taining the basic residues lysine or arginine (luteinizing hor- actual retention times, 0.2 min was added to the retention mone-releasing factor, eledoisin-related peptide) exhibited a coefficient for an amino acid; the predicted retention times and decreased retention at lower pH due either to increased ioni- correlation coefficient were then again calculated. If the cor- zation of the amino group or to formation of an ion pair with relation had been improved by adding 0.2 min to the retention the perchlorate in the mobile phase. Fig. 1 shows the separation coefficient, the change was kept; otherwise the value was re- at pH 2.1 of various peptides. The sharpness of the peaks with turned to that previously used, and 0.2 min was added to the minimal tailing demonstrates the high resolution possible with next amino acid and so on. After checking all amino acids to see HPLC. Fig. 1 also demonstrates the difference in detector se- whether increasing the retention coefficient could increase the lectivity between measurement of absorbance at 220 nm (top correlation, 0.2 min was sequentially subtracted from each trace) and measurement of fluorescamine-induced fluorescence amino acid in turn and correlations were again calculated after (bottom trace). Compounds such as thyrotropin-releasing factor each subtraction. At the end of these two cycles, the slope of the (peak 5) without a free amino-terminal group give little or no plot of predicted vs. actual retention times was calculated be- fluorescence. Fig. 2 shows that there is an approximately linear cause the predicted and actual times should be equal, not increase in retention time for phenylalanine oligomers as merely be correlated (proportional). Therefore, if the slope was phenylalanine residues are added to diphenylalanine. This greater than 1.0, all retention coefficients were multiplied by finding indicates that, with the linear gradient used in these 0.99; if the slope was less than 1.0, the values were multiplied experiments, the addition of each phenylalanine residue adds by 1.01. These cycles were repeated until a near maximum approximately the same retention time to the peptide. The slope correlation had been obtained. thus equals the retention added per side chain and peptide bond, but does not include the contribution of the terminal RESULTS amino or carboxyl groups. Extrapolation of the line to 0 residues Table 1 lists the 25 peptides used for calculation of retention gives a positive value, which represents the contribution to re- coefficients and their retention times at pH 7.4 and 2.1. As ex- tention of the end groups. The retention for phenylalanine itself pected, small neutral peptides (triglycine and pentaalanine) 0.05' Table 1. Retention times of peptides separated by HPLC F Retention time, min Compound pH 7.4 pH 2.1 CD 1. Triglycine 2.0 3.0 c o 2. Pentaalanine 4.6 8.1 .0 3. Divaline 6.9 14.5 0 4. Dimethionine 10.5 21.0 5. TRF 11.5 11.2 6. Tuftsin 11.7 12.0 7. Trityrosine 19.5 29.7 8. [Met]Enkephalin 27.5 38.0 9. Trileucine 28.0 36.8 -0 10. [Leu]Enkephalin 29.3 42.0 11. Ditryptophan 31.6 44.3 I1 I 12. Angiotensin II 32.2 47.5 10 13. a-Endorphin 32.3 47.7 14. Caerulein 34.2 42.2 CQ) 24 C: 7 15. Oxytocin 36.4 37.9 2 16. Gastrin-(12-15) 36.5 42.4 C.1)IL) 17. Neurotensin 39.0 48.0 0) .18. Physalaemin 41.0 43.0 LL 19. Triphenylalanine 41.6 49.5 b~~~~ 20. LHRF 42.8 49.5 25 21. a-Melanotropin 46.2 46.2 22. Bradykinin 48.0 45.0 . ~~~ _V 23. Eledoisin-related peptide 53.0 44.0 0 10 20 30 40 50 60 24. Glucagon 53.6 60.0 Retention time, min 25. Somatostatin 57.5 55.0 FIG. 1. Separation of peptides by reverse-phase HPLC on a Li- chrosorb RP18 column. (Upper) Absorbance at 220 nm (0.05 ab- Peptides were chromatographed at room temperature on a Bio-Rad sorbance units full scale). (Lower) Fluorescamine-induced fluores- ODS column with a linear acetonitrile gradient (0.75%0/min). TRF, cence. A 40-gl sample containing 300-600 ng of each peptide was thyrotropin-releasing factor; LHRF, luteinizing hormone-releasing chromatographed at pH 2.1 at room temperature with a flow rate of factor; Eledosin-related peptide, Lys-Phe-Ile-Gly-Leu-Met-NH2. LO ml/min and an acetonitrile gradient of 0.75%/min. Downloaded by guest on September 28, 2021 1634 Medical Sciences: Meek Proc. Natl. Acad. Sci. USA 77 (1980)

Table 2. Retention coefficients of amino acid residues Retention coefficients Amino acid (N) pH7.4 pH 2.1 Tryptophan (7) 14.9 18.1 Phenylalanine (13) 13.2 13.9 Isoleucine (4) 13.9 11.8 c *Leucine (9) 8.8 10.0 E Tyrosine (11) 6.1 8.2 Methionine (9) 4.8 7.1 Valine 2.7 3.3 C (5) 0 Proline (10) 6.1 8.0 w a) Threonine (5) 2.7 1.5 a1) Arginine (7) 0.8 -4.5 Alanine (4) 0.5 -0.1 Glycine (13) 0.0 -0.5 Histidine (5) -3.5 0.8 Cystine (2) -6.8 -2.2 Lysine (8) 0.1 -3.2 Serine (6) L2 -3.7 Asparagine (5) 0.8 -1.6 1 2 3 4 5 Glutamine (4) -4.8 -2.5 Number of phenylalanine residues Aspartic acid (5) -8.2 -2.8 FIG. 2. Linearity of retention time with the number of phenyl- Glutamic acid (3) -16.9 -7.5 alanine residues. Phenylalanine oligomers were chromatographed by Amino- (19) 2.4 -0.4 using a linear acetonitrile gradient. The slope of this plot indicates -COOH (17) -3.0 6.9 the HPLC retention per phenylalanine residue. -Amide (8) 7.8 5.0 Pyroglutamyl- (5) -1.1 -2.8 Acetyl- (1) 5.6 3.9 was not plotted because the pKs for amino acids differ consid- Tyrosine (1) 10.9 6.5 erably from those of peptides and the extent of ionization sulfate markedly affects retention. N, number of peptides used for calculation of retention coefficients Table 2 lists the retention coefficients calculated by repeated that contained this amino acid. Retention coefficients (in min) were determined by reverse-phase HPLC on a Bio-Rad ODS column. The regression analysis. Amino acids with aromatic or aliphatic side predicted retention time for a peptide equals the sum of the retention chains have a marked positive contribution to retention which coefficients for the amino acids and end groups plus to (the time for changes relatively little with pH. Residues with acidic side elution of unretained compounds). In these experiments, to = 2.0 min. chains have a marked negative contribution to retention which increases in magnitude as ionization increases. Basic and neutral residues have little effect on retention. Fig. 3 shows the re- with predicted times. The best available compilation of re- markably high correlation obtained when plotting the actual tention times for peptides of biological interest was made by retention times from Table 1 vs. the times predicted by O'Hare and Nice (2). These authors used an acetonitrile gra- summing the retention coefficients listed in Table 2 for each dient at pH 2.1 (as in the present study), although the inorganic peptide. ion added (0.1 M phosphate), Hypersil ODS column, and tri- It should also be possible to predict retention times when phasic linear gradient differed from the conditions used here. using various other chromatographic conditions. The retention They chose a gradient consisting of a rapid rate of change for times for several di- and tripeptides (numbers 1, 2, 3, 9, and 19 about the first 6 min, then a 40-min slower rate of change, and of Table 1) were determined with several gradient rates and a final 5-min rapid rate. The retention characteristics of com- with columns of several manufacturers. With a gradient rate pounds eluting during the second phase of their chromato- of 1.5% acetonitrile per min (twice the usual rate), all retention graphic system should be comparable to those used in the times were 70% of normal. With a gradient rate of 0.5%/min, present study, although retention of compounds eluting be- retention times were increased to 120% of normal. Retention tween 10 and 15 min might be overestimated due to the in- times for these compounds obtained with the Bio-Rad column fluence of the initial rapid gradient. To examine the compar- were similar to those seen with columns from Waters Associates ability of the two chromatographic systems, the actual retention (10-,m particle size), Lichrosorb (Rheodyne) (5 ,mn), and times of phenylalanine oligomers was plotted (Fig. 4). Least Dupont (5 ,um). That there are minor differences in retention squares analysis gave a line with a slope of 0.836 (representing with different columns can be noted by comparing Table 1 the difference in gradient rates) and a y intercept of -12.3 min (Bio-Rad column) with Fig. 1 (Lichrosorb column). Thyrotro- [due to the rapid initial gradient used by O'Hare and Nice (2)]. pin-releasing factor (peak 5) was shifted about 2 min relative The high correlation for these three points (0.9983) indicates to pentaalanine and divaline; luteinizing hormone-releasing that, for these compounds at least, the chromatographic con- factor was shifted about 2 min relative to [Met]- and [Leu]- ditions can be compared. Table 3 lists the actual retention times enkephalin. of compounds tested by O'Hare and Nice of 31 amino acid residues or less, excluding insulin chains A and B. These latter TEST OF PREDICTIVE ABILITY OF compounds contain cysteine for which no retention coefficient RETENTION COEFFICIENTS had been calculated because cysteine was not contained in any To examine how well these retention coefficients are able to of the peptides tested in Table 1. predict times for other peptides and other chromatographic Predicted retention times were calculated for the compounds conditions, data were taken from the literature for comparison in Table 3 by summing the retention coefficients, multiplying Downloaded by guest on September 28, 2021 Medical Sciences: Meek Proc. Natl. Acad. Sci. USA 77 (1980) 1635

610

540 2 25 .E_ 21

C ) 1 7 c 1 6 14 w 7+ 0) 0) 30 12 4- 1

" 2C .2

0)

10 4.56 3~~ 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Actual retention time, min FIG. 3. Correlation of actual retention times vs. times predicted by summing retention coefficients for the amino acids and end groups. Numbers adjacent to the data points indicate the peptides listed in Table 1. (Left) pH 7.4; correlation = 0.9996. (Right) pH 2.1; correlation = 0.9970.

by the slope from Fig. 4 (0.836), and adding the y intercept (30,000 for 100-A pore size particles) makes it unlikely that (-12.3 min). As seen in Table 3, the predicted and actual re- steric exclusion has a significant role in separation of these small tention times agree reasonably well, especially considering the peptides. Quite large peptides and even small enzymes can be differences in columns, mobile phases, and gradients. For all chromatographed with high resolution by HPLC (2), although peptides up to 20 residues, the average error was 4 min. it is likely that these compounds are in a partly folded config- uration. In theory, even higher resolution might be obtained DISCUSSION if conditions could be found in which all the residues in the molecule The basic premise of this was it were available for interaction with the surface of the study that should be possible stationary phase. to derive values that reflect the hydrophobicity (positive or negative) of the amino acids in peptides and that it is the sum Table 3. Comparison of data from the literature with predicted of these values that might primarily determine the extent of retention times retention on reverse-phase HPLC columns. Amino acid com- No. of Retention time, min Error, position cannot be the only factor determining the extent of Compound residues Predicted Actual* min retention because it is possible to separate position isomers with the same composition (e.g., Gly-Trp and Trp-Gly) (1) and ste- [Met]Enkephalin 5 18.4 19.0 -0.6 reoisomers. However, it seemed likely that for small linear [Leu]Enkephalin 5 20.9 22.0 -1.1 peptides, sequence and conformation should have relatively ACTH-(5-10) 6 11.9 17.0 -5.1 little effect on retention. As shown by these data, this premise ACTH-(34-39) 6 27 31.0 -4.0 appears justified for peptides up to about 20 residues long. ACTH-(4-10) 7 12.4 20.5 -8.1 8 33.1 30.0 There is no evidence that the size of these peptides has any ef- ACTH-(4-11) 3.1 Angiotensin II 8 27.4 23 fect on retention. The high exclusion limit of these columns 4.4 Substance P-(4-11) 8 33.1 30.0 3.1 Oxytocin 9 27.4 23.0 4.4 [Arg]Vasopressin 9 9.0 14.0 -5.0 [Lys]Vasopressin 9 10.2 13.0 -2.8 [Arg]Vasotocin 9 7.2 12.0 -4.8 Substance P 11 33.3 29.0 -4.3 c a-MSH 13 27.3 26.0 1.3 I E Neurotensin 13 28.9 24.5 4.4 z _. Somatostatin 14 25.5 32.0 -6.5 0 Phe 2, Bombesin 14 22.6 0) 26.0 -3.4 Gastrin-1 17 26.8 28.5 -1.7 az ACTH-(18-39) 22 20.2 30.5 -10.3 C ACTH-(1-24) 24 34.9 21.5 13.4 0) 20 30 40 50 60 Melittin 25 61.3 46.0 15.3 Retention time in NaCIO4, min Glucagon 29 39.7 36.0 3.7 ,B-Endorphin 31 41.5 34.0 7.5 * Data were from O'Hare and Nice (2). Predictions of retention times FIG. 4. Comparison of retention times for phenylalanine oli- were made by summing retention coefficients for each peptide and gomers in two chromatographic systems. Data of O'Hare and Nice (2) adding to (2 min). To correct for the different gradient rates used were obtained with a triphasic acetonitrile gradient containing in these studies, the sum of the coefficients was multiplied by 0.836 NaH2PO4 on a Hypersil ODS column. Data in this paper were ob- (the slope in Fig. 4). To correct for the biphasic gradient used by tained with a linear acetonitrile gradient containing NaClO4 on a O'Hare and Nice, they intercept of Fig. 4 (-12 min) was then added. Bio-Rad ODS column. ACTH, corticotropin; MSH, melanocyte-stimulating hormone. Downloaded by guest on September 28, 2021 1636 Medical Sciences: Meek Proc. Natl. Acad. Sci. USA 77 (1980) CHOICE OF CHROMATOGRAPHIC CONDITIONS separation a plot of log k' (the adjusted retention time/to) vs. The ideal mobile phase should allow sharp peaks with minimal the number of alanine residues in alanine oligomers yielded a tailing, have low absorbance at 200 nm to simplify high-sensi- straight line.,Snyder and Kirkland (10) reported that with linear tivity detection, be easy to neutralize to facilitate reaction with gradients on chemically bonded phases log k' was linear with fluorescamine, and have only volatile components to allow vol% of B over the range 20-90% B. It therefore follows that a concentration of the effluent for subsequent analysis. Early linear gradient should give a linear increase in retention time reports in which peptides were chromatographed with as residues are added to a homologous series. That this is ap- methanolic mobile phases containing no inorganic salts showed proximately true is shown in Fig. 3. broad peaks of little use considering the complexity of real Complete gradients might not be required for routine sep- samples. Molnar and Horvath (1) showed that extremely high arations of a few compounds; a near optimal gradient program resolution could be obtained with acetonitrile gradients con- could be readily designed by predicting the retention times, taining 0.6 M HC104 or 0.1 M phosphate, pH 2.1. However, calculating the acetonitrile concentration at that point both these solutions are difficult to neutralize continuously. Use (0.75%/min), and choosing a slightly lower concentration of of the strong phosphate buffer at neutral pH also gave excellent acetonitrile to bring the retention time into the desired range results, but the low solubility in acetonitrile of sodium phosphate for k' of 2-10 (10). limits its usefulness. Rubinstein et al. (4) obtained excellent Although HPLC has so far not proven useful in the direct results with an isopropanol gradient and volatile pyridine chemical estimation of peptides in crude tissue samples, HPLC buffers, but these buffers preclude the use of ultraviolet de- appears to be the most powerful tool currently available for the tectors. The present experiments were performed with NaClO4 preparative separation and isolation of small peptides and seems because it permitted both reaction with fluorescamine and to be the ideal method to check the specificity of measurements detection at 220 nm. Unfortunately, NaClO4 is not volatile. made by site-specific tests such as radioimmunoassays or Preliminary experiments showed that many small peptides gave bioassays. Hopefully, the ability to readily predict retention sharp peaks with only the dilute phosphoric acid recommended times and to simplify choice of chromatographic conditions will by Hancock et al. (3) or very dilute sodium phosphate buffer, aid such studies. pH 7.4. However, one compound (substance P) gave very broad peaks unless NaCl04 was added. It may be that the perchlorate 1. Molnar, I. & Horvath, C. (1977) J. Chromnatogr. 142, 623- 640. blocks adsorption to some exposed sites on the silica beads of the 2. O'Hare, M. J. & Nice, E. C. (1979) J. Chromatogr. 171, 209- column. However, a theoretical reason for using NaClO4 is that 226. it is a strong "chaotropic" agent-i.e., an inorganic ion that 3. Hancock,-W. S., Bishop, C. A., Prestidge, R. L., Harding, D. R. favors the transfer of nonpolar groups to water by altering water K. & Hearn, M. T. W. (1978) Science 200,1168-1170. structure (9). Such agents increase the solubility of lipophilic 4. Rubinstein, M., Stein, S. & Udenfriend, S. (1977) Proc. Nati. Acad. polymers and are likely to break down secondary and tertiary Sci. USA 74,4969-4972. structures of peptides. It may be that high concentrations of 5. Rivier, J. (1978) J. Liquid Chromatogr. 1, 343366. NaClO4 will aid in chromatography of very large peptides. 6. Lakshmanan, T. K. & Lieberman, S. (1954) Arch. Biochem The choice of a linear gradient, started at the time of injec- Biophys. 53, 258-281. 7. Udenfriend, S. (1972) Science 178, 871-872. tion, was dictated by several considerations: (i) it provides op- 8. Frei, R. W., Michel, L. & Santi, W. (1976) J. Chromatogr. 126, timal peak shape in all regions of the chromatogram, (ii) it is 665-677. reproducible in all laboratories regardless of the equipment, 9. Hatefi, Y. & Hanstein, W. G. (1969) Proc. Nati. Acad. Sci. USA and (iii) it was hoped that linear gradients would permit esti- 62, 1129-1136. mation of retention times from amino acid composition. Molbrfr 10. Snyder, L. R. & Kirkland, J. J. (1974) in Introduction to Modern and Horvaith (1) reported that in an isocratic (nongradient) Liquid Chronatography (Wiley, New York), p. 466. Downloaded by guest on September 28, 2021