Anal. Chem. 1990, 62, 332-338
332
Evaluation of the Effect of Organic Modifier and pH on Retention and Selectivity in Reversed-Phase Liquid Chromatographic Separation of Alkaloids on a Cyclodextrin Bonded Phase Daniel W. Armstrong,* G a r y L. B e r t r a n d , K a r e n D. Ward, a n d Timothy J. Ward Department of Chemistry, University of Missouri-Rolla,
Rolla, Missouri 65401
H e n r y V. Secor a n d J e f f r e y I. Seeman* Philip Morris Research Center, P.O. Box 26583, Richmond, Virginia 23261
A mathematical model is developed Using a basic equlllbrium drlven approach to explain the effect of mobile-phase composltlon on the relatlve retentlon of Ionizable solutes In a reversed-phase llquid chromatography (LC) separation. Three speclflc cases (Le., dlfferent organic modifiers, dtfferent pHs, and different solute pK,s) and one general case are consldered. Twelve structurally related tobacco alkalolds and metabdnes are separated under varlous conditkns desabed by the model. The test solutes include anabaslne, anatabine, cotlnlne, P,S'-Mpyrhlyl, N'methylanabasine, myosmine, nicotine, the two nicotlne N'sxldes, nlcotyrlne, norcotlnlne, and nornkotlne. The bask model appears to explain the effects of pH and organic modlfler on the relatlve retention of these solutes. The same general model should be appllcable to most groups of ionizable solutes separated by reversebphase LC.
"Optimizing the separation" is an important part of any methods development project involving liquid chromatography (LC). In doing this, one must choose the best stationary phase and mobile phase combination. There are several types of reversed-phase bonded stationary phases in liquid chromatography, including hydrocarbon-based materials (which include the well-known and c8 phases as well as diphenyl and related bonded phases), cyclodextrin-based materials, internal surface reversed-phase materials, and others ( 1 , Z ) . In classic reversed-phase LC, hydroorganic mobile phases are used, nonpolar solutes tend to be retained more than polar solutes, and retention of nonpolar solutes tends to decrease with increasing concentration of organic modifier in the mobile phase (provided solubility is not a problem). Once a stationary phase is chosen, only the mobile-phase composition and temperature significantly affect the separation. Most researchers have focused attention on mobilephase optimization, and with good reason, since this is the easiest way to control retention and selectivity in LC ( 1 , Z ) . Obviously, the understanding and prediction of retention behavior in reversed-phase LC are of fundamental importance to chromatographers. Both empirical (3-6) and theoretical (7-9) approaches have been used to explain chromatographic retention. The first goal of this work is to use a basic equilibrium driven approach to explain the relative retention behavior of ionizable solutes on a cyclodextrin bonded phase in three ideal situations. These include (1)the relative retention behavior of an ionizable solute when chromatographed with two different mobile phases in which the organic modifier is different
but the pH of the aqueous component is identical, (2) the relative retention behavior of different ionizable solutes in the same hydroorganic solvent at thesame pH, and (3) the relative retention behavior of an ionizable solute in the same hydroorganic solvents at different pHs. A general theoretical treatment is given in the Appendix. The retention behavior expected from the theoretical treatment is compared to actual separation data for 12 naturally occurring tobacco alkaloids and metabolites (1-12, Chart I). Qualitative and quantitative analysis are important for programs involving alkaloid isolation and phytochemistry (10-13) as well as for investigations involving metabolic, enzymatic, and chemical transformations (14-1 7). While an extensive literature is available on the quantitative determination of nicotine (7) (18-30), very few studies have been reported in which more than two or three of the other related tobacco alkaloids have been analyzed simultaneously (26,29). The second goal of this study was to develop an LC method for 1-12 uving a 8-cyclodextrin (p-CD) bonded phase, which seemed to have the greatest selectivity for these compounds. EXPERIMENTAL SECTION Methods. All separations were done at room temperature (21 "C) with Shimadzu LC-4A and 6A liquid chromatographs. The compounds were detected at 254 nm with a variable wavelength detector containing 13- and 8-rL flow cells, respectively. All samples were dissolved in acetonitrile or methanol (depending on the mobile-phase composition) prior to injection. Typically, 5-10 WLof solution (0.1-0.5% nicotinoid) was injected. Columns (25 x 0.46 cm) containing P-cyclodextrin bonded to 5-hm silica were obtained from Advanced Separation Technologies, Whippany, NJ. The void volume of the column was determined by injecting neat methanol. The peak-trough combination caused by the change in refractive index was used as a marker. Flow rates, solvent compositions,and pHs are given in the respective tables and figures. Materials. High-performanceliquid chromatography (HPLC) grade methanol, acetonitrile, triethylamine, and water were obtained from Fisher Scientific Co. Buffers were prepared by making a 1% solution of triethylamine in water and adding glacial acetic acid until the desired pH was obtained. Nicotine (7), cotinine (3), and 2,3'-dipyridyl (4) are available commercially (Aldrich). (l'R,Z'S)-Nicotine "-oxide (8) was prepared by sodium tungstate (Alfa Products)/hydrogen peroxide oxidation of nicotine following literature procedures (31)and isolated by chromatography on a 4-mm Chromatotron with sequential elution with chloroform followed by chloroform:ethanol:a"onium hydroxide (8511411) followed by chloroform:methanol:a"onium hydroxide (50/20/ 1). (l'S,Z'S)-Nicotine "-oxide (9) was prepared by m-chloroperbenzoic acid oxidation of nicotine (32) and purified by column chromatography. The other seven tobacco alkaloids were prepared by literature procedures: 1 (33-37), 2 (38,39),5 (37),6 (33,34), 10 (401, 11 (411, and 12 (33, 42).
0003-2700/90/0362-0332$02.50/00 1990 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 62, NO. 4, FEBRUARY 15, 1990
!
i
9
Table I. Retention Data for Tobacco Alkaloids and Nicotine Metabolites
4,lO
k’ with acetonitrile modifier at pH of 4.Ia 5.5O 7.1a
compound
5
0
RETENTION TIME, 5,11
.
15
10
333
20
MiN
1 (&(-)-anabasine 2 (R,S)-anatabine 3 @)-(-)-continine 4 2,3’-dipyridyl 5 @)-(-)-”-methyl-
anabasine 6 myosmine 7 @)-(-)-nicotine 8 (lR,2S)-unti-nicotine ”-oxide 9 (lS,BS)-syn-nicotine ”-oxide 10 nicotyrine 11 (R,S)-norcotinine 12 (R,S)-nornicotine
8.9
k’with M~OH modifier at pH 5.5O
1.87
0.36 0.22 1.43 3.97 0.79
1.48 0.55 0.98
1.60 1.53 0.64
2.03 0.65 1.08
0.60
0.98
0.64
0.98
1.41 0.37 0.15
2.87 0.66
2.94 0.56 0.80
3.97 0.83 0.22
0.20 0.15 0.78 1.87 0.37
0.36 0.30
0.60 0.28 0.71
1.21
2.59 0.66
0.30
0.80 0.80 1.15 2.46
~~
5
0
15
13
20
RETENTION TIME, M I N Flgure 1. Separation of tobacco alkaloids and nicotine metabolites. A. Chromatographic conditions: 5 % methanol, 95 % aqueous triethylammonium acetate (1%) pH = 5.5, flow rate = 1.0 mL/min, one 25-cm P-cyclodextrin column. B. Chromatographic conditions were the same as in A except 5 % acetonitrile was used. 1
a The mobile-phase composition was (5/95 (v/v)) acetonitrile:l% triethylammonium acetate (aq). The flow rate was 1.0 mL/min.
t
/
A
5
0
10
RETENTION TIME,
MiN
11
B
15
0
1,2,12
5
10
RETENTION TIME,
15
MiN
Flgure 2. Separation of tobacco alkaloids and nicotine metabolites using a 5 % acetonitrile, 95 % triethylammonium acetate mobile phase: flow rate = 1.0 mL/min; one 25-cm PCO column: (A) pH = 4.1 and (B) pH = 7.1.
RESULTS AND DISCUSSION The 12 naturally occurring tobacco alkaloids and metabolites (1-12) listed in Chart I cover a broad spectrum of structural types: imine, lactam, secondary and tertiary amine, amine oxide, and pyrrole. All are 3-substituted pyridine derivatives with either a pyrrolidine ring (nicotine family) or piperidine ring (anabasine family) substructure. The very close similarity of some of these compounds (e.g., 1, 2, and 12; 5 and 7; 8 and 9), the similarity of the basicity of these compounds, and the similar molecular size throughout the entire set makes the analytical challenge significant. We will first disclose our recent investigation using HPLC with improved 0-cyclodextrin bonded phases to effect the optimum separations. We will then provide a discussion of an equi-
I k 1.0
2.0
3,O
4.0
k’ (pH = 5.5, methanol) Figure 3. Relationship between k ’ s with acetonitrile and methanol as modifiers at pH = 5.5 at the same flow rate, temperature, and column. See eq 1.
librium based model to explain relative retention behavior of ionizable solutes in reversed-phase HPLC. Figures 1 and 2 show the separations obtained in this work by using either methanol or acetonitrile as the organic modifier with the aqueous phase buffered at pH = 4.1,5.5 or 7.1. The data is summarized in Table I, where the k’s can be directly compared, at different pHs, since differences in flow rate are normalized by using k’s. At least one set of mobile-phase conditions could be found to separate all components except for anatabine (2) and nornicotine (12), the two secondary amines. As may be expected, significant changes in retention and selectivity could be obtained by altering the organic modifier and pH of the mobile phase. These figures demonstrate the capability of this reversed-phase LC method for the analysis of related alkaloids. A comparison of Figures 1 and 2 indicates little change in the selectivity of the compounds at pH = 5.5 buffer. This is
334
ANALYTICAL CHEMISTRY, VOL. 62, NO. 4, FEBRUARY 15, 1990
Chart I. Structures of Tobacco Alkaloids and Nicotine Metabolites
compound 1 @)-(-)-anabasine
structure
I/
I, Y
Ir
2
(R,S)-anatabine 2 3
1.3
3 3
I
k' (PH= 5.5,acetonitrile)
Figure 4. Relationship between k ' s at pH = 4.1 and 5.5 with acetonitrile as modifier in both cases at the same flow rate, temperature, and column. See eq 2.
3 (SI-(-bcotinine
4 2,3'-dipyridyl
,r-
*'O
r
0
/
N 5
(5')-(-)-N'-methylanabasine
I 6
myosmine
7 @)-(-)-nicotine
I
1
I
1.C
I
2.0
I
3.1
k' (pH = 7.1, acetonitrile) Figure 5. Relationship between k's at pH = 4.1 and 7.1 with acetonitrile as modifier in both cases at the same flow rate, temperature, and column. The correlation line A (eq 3) describes the behavior for the seven less basic nicotinoids (types I1 and 111, open circles). The correlation line B (eq 4) describes the behavior for the five more basic nicotinoids (type I, solid circles). The data was taken from Table I.
shown by an excellent correlation (Figure 3 and eq 1)between the k's with acetonitrile vs methanol as modifier (note that pH, flow rate, temperature, and column remain the same). k'(acetonitri1e) = 0.187
+ 0.647k '(methanol)
(1)
p H = 5.5
r = 0.994, std dev of residuals = 0,099, n = 12, p c 0.0001 Figure 3 10 nicotyrine
11
(R,S)-norcotinine
12
( R$1-nornicotine
As the pH increases, three types of behavior are observed: type I (for 1 , 2 , 5 , 7 , and 12), where solute retention increases with pH; type I1 (for 8 and 9), where solute retention goes up from p H = 4.1 to pH = 5.5 and then decreases to less than the original retention a t pH = 7.1; and type 111 (for 3, 4, 6, 10, and l l ) , where solute retention goes up and more or less stabilizes at pH > 5.5. Examination of these three groups shows a structural clustering. All 12 substrates are 3-substituted pyridines. Type I compounds are moderately strong organic bases, all secondary or tertiary amines. At pH = 5.5, nicotine is >99% monoprotonated (43). Type I11 compounds are all far less basic, the second ring being part of a lactam, a pyrrole, an imine, or a pyridine. At pH = 5.5, these compounds exist primarily as their unprotonated free bases (44). The two compounds in type I1 are both "-oxides of the methyl-
ANALYTICAL CHEMISTRY, VOL. 62, NO. 4, FEBRUARY 15, 1990
R
Stationary Phase Surface
will be considered subsequently. Given the basic model shown in Figure 6, two equilibrium expressions can be written for the solute, R. The first will be referred to as the observed equilibrium expression (where (KR)obs is the equilibrium constant). actually is a distribution coefficient that takes into account all possible forms of R (i.e., ionized, un-ionized, associated, unassociated, etc).
KRf
R’
Figure 6. Basic model describing the equilibria involved in the HPLC retentions. R and R* refer to the solute being chromatographed,the former the amine free base and the latter the protonated species. OM is the organic modifier.
pyrrolidine ring. These structural features and basicity properties suggest that the protonated species does not complex strongly with the /3-CD support. Figure 4 illustrates that there is a linear correlation between the k’s observed for 1-12 in acetonitrile at pH = 4.1 vs pH = 5.5 (eq 2). However, completely different behavior is found
k’(pH = 4.1) = -0.0099
+
0.588k’(pH = 5.5) acetonitrile r = 0.951, std dev of residuals = 0.172, n = 12, p < 0.001
(2)
Figure 4
+
where nR(s)is the number of adsorption sites occupied by solute R, n: is the total number of adsorption sites, and [R],, is the total concentration of R regardless of its form. A true equilibrium expression for each form of R can be written as well:
where nOM(s) is the number of stationary-phase adsorption sites occupied by the organic modifier and [R,] is the concentration of R in bulk mobile-phase solution. The ratio of the observed to the true equilibrium constant (provided n: >> nR(B), which is the normal chromatographic condition) is
The equilibrium expression for the organic modifier (OM) is
when the k’s are compared at pH = 4.1 vs pH = 7.1. As shown in Figure 5, a spread of points is observed. If the data is divided into two groups, one consisting of type I substrates and the second consisting of type I1 and type I11 substrates, two correlations are found, as indicated by eq 3 and 4 and the correlation lines A and B in Figure 5, respectively.
k’(pH = 4.1) = 0.225 + 0.477k’(pH = 7.1) type I1 and type I11 substrates, acetonitrile r = 0.851, std dev of residuals = 0.307, n = 7, p < 0.015 Figure 5, line A k’(pH = 4.1) = 0.022 0.180k’(pH = 7.1)
335
nOM(s)
KOM
=
[OM]
(nso - nOM(s))
where [OM] is the concentration of organic modifier in the mobile phase. Rearranging eq 8 gives
KOM[OMl
1 - -nOM(s) =1nsO
(3)
+ KOMIOMl
(9)
or
(4)
Equation 10 can be substituted into eq 7 to give
type I substrates, acetonitrile
r = 0.971, std dev of residuals = 0.026, n = 5, p < 0.005 Figure 5, line B Expressions can be derived that explain the observed elution profiles (vide supra) and can be used to predict relative rekntion behavior of other classes of acidic o; basic solutes. The derivations are based on the simple model shown in Figure 6. There are a finite number of adsorption sites on the stationary phase on which the solute being chromatographed (R or R*) and the mobile-phase components (i.e., the organic modifier (OM) and water) can bind. It is assumed that the organic modifier has a greater affinity for the binding site than water has, and that the solute (R)has the greatest affinity for the adsorption site. If given sufficient time, there would be an equilibrium distribution of R,R* and OM between the mobile phase and the stationary-phase adsorption sites. It is further assumed that the equilibrium constants involving R and OM are independent of their concentration. Having made these assumptions, we point out that there can be secondary effects of concentration on both the thermodynamic equilibrium constants and activity coefficients. These could cause deviation from the ideal cases being outlined here and
(KR)obs --
(KR)true
-
[Raql [RIbtal(1 + KOM[oMl)
If R is a weak base (i.e., R
(Kb)R =
+ HzO
-
RH+
(11)
+ OH-), then
[OH-] [RH+] [Raql
(12)
or
(Kb)R =
[OH-li[Rltntal- [RaqlJ [Raql
(13)
Equation 13 can be rearranged to
-[Raql iR1tntaI
1 1 + (K~)R/[OH-]
(14)
If R is a weak acid, the analogous expression is obtained:
-[Raql -
-
1
[Rt~tdI 1 + (K~)R/[H+]
(15)
In general, [Raql/[Rtota~l= ~(PH,PKR). Taking the case where R is a weak base and substituting eq 14 into eq 11 yields
336
ANALYTICAL CHEMISTRY, VOL. 62, NO. 4, FEBRUARY 15, 1990
-(KR)obs -
-
(KR)true
1 (1 + (Kb)~/[oH-])(l + KoM[OM])
(16)
For two different solutes (Rl, R,), (Kb)RI/(Kb)R2 and (KRl)true/(KRP)tmeare fairly independent of small changes in pH and [OM] or the nature of the organic modifier. KOM is independent of pK, [R], and the nature of R. Special Case No. 1. To evaluate the reversed-phase retention behavior of a solute, R, in two different mobile phases (i.e., in which the organic modifier is different) and with the pH of the buffer portion of the mobile phase identical, one must use eq 16. For the mobile phase containing the first organic modifier
( K R1 ~
' = (1
(KRl)true
(Kb)R1/
[OH-l)(l
K o ~[OMiI) i
(17)
and
This time when the ratio of eq 22 and 23 is taken, both ( K R ~and ) ~(1~4-~K~o M ~ [ O M ~cancel ]) to give
Consider two limiting cases for eq 24. The first case is a t a very high pH (relative to Kb) where [OH-], and [OH-], >> (Kb)R1.In this instance, eq 24 becomes
For the mobile phase containing the second modifier The second case is for a relatively strong base chromatographed with a mobile phase a t lower pHs were (Kb)R1 >> [OH-], and [OH-],. In this case one obtains
Dividing eq 17 by eq 18 gives
where both (KRl)true and (1+ ((Kb)Rl/[OH-])) have cancelled. In liquid chromatography, the capacity factor is directly proportional to the equilbrium constant of a solute between stationary and mobile phases (i.e., k' 0: KeJ. Therefore eq 19 also could be written as the ratio of the capacity factors. Thus, eq 19 can relate the relative retentions of a solute (R,) chromatographed with two different mobile phases. In order to compare the retention behavior (or k') of two or more different solutes chromatographed with two different mobile phases, the derivation that led to eq 19 must be repeated for the second solute, giving
Since the right sides of eq 19 and 20 are identical for any two different (but related) solutes, this quantity may be taken as a constant, CY, which is independent of solute.
In plots of the limiting cases for a hypothetical series of strong and weak bases, the weak bases would fall along a line with a slope of 1, while the relatively strong bases would tend to fall along a line parallel to the x axis if this axis represents the high pH. Solutes of intermediate PKbs would fall somewhere between these lines. Indeed this type of behavior was observed for the alkaloids used in this study. Figure 5 is a plot of k'values for a series of the 12 alkaloids chromatographed at pH 4.1 and 7.1 (all other conditions were identical). Clearly two types of retention behavior are observed. One group of alkaloids (type I: anabasine, anatabine, N'-methylanabasine, nornicotine, and nicotine) approaches the retention behavior described by eq 26 while the other group (types I1 and 111) approaches the behavior described by eq 25. The segregation of these alkaloids into two groups is a result of their relative PKbs and the pH of the mobile phase as described by eq 24.
CONCLUSIONS According to eq 21, plots of k'or (KRJObs for different solutes in two different mobile phases (i.e., two different organic modifiers, all other conditions identical) should be a straight line with an intercept of zero. The slope of this line would vary with the type and concentrations of the organic modifiers. The intercept also can deviate from zero somewhat if other factors are operative, such as adsorption to residual silanol groups or deviations from the assumptions made for this derivation. An analogous plot of In (KR):zl vs In (KR);z2 should produce a straight line of slope 1 and an intercept of In a. An example of the type of behavior predicted by eq 21 is shown in Figure 3 in which the capacity factors of several alkaloids separated with two different mobile phases (at pH = 5.5) are plotted. Special Case No. 2. To evaluate the reversed-phase retention behavior of a compound in the same mobile phase (i.e., the same organic modifier) but with the aqueous portion of the solution buffered a t two different pHs, eq 16 again can be used. In this case eq 16 must be written for each of the two pH situations where
-
The treatment derived in this work adequately explains the reversed-phase retention behavior of these alkaloids on a p-CD column. In fact, an analogous treatment should be able to explain the retention behavior of most ionizable compounds on any reversed-phase packing, given standard reversed-phase separation conditions (e.g., hydroorganic mobile phases; simple buffers; avoidance of temperature, flow, pressure, or concentration extremes; and so on). A general treatment describing the retention behavior of ionizable solutes in reversed-phase separations is given in the Appendix. The important procedure of optimizing a liquid chromatographic separation has been, at best, an empirical process. By providing a theoretical basis for this, one may be able to obtain a better separation and a more complete understanding of retention behavior.
APPENDIX Given the model in Figure 6, the following general equilibrium and associated constant can be written: R(aq)
If
+ nH+(aq) F= R*(aq)
n = +1, R is a base n = -1, R is a n acid
KD = K,/K, KD = K ,
The retention of R is determined by an effective equilibrium constant that does not distinguish between the various forms of R, R", etc. Translation of retention volume to an equilibrium constant also assumes that all sites (n,O)on the column are available to the analyte: InR(s) (KR)obs
=
+ nR+-(s))
{n: - nR(s) - nRt(s)lI[Rl+ [Rill' nS0 >> n ~ + ( ~~ R) L ( ~(28) )
However, the actual equilibria depend on the amounts of each form and the possible occupancy of sites by the organic modifier, OM:
For a basic analyte, if pH1 < pH2, K D [ H + ~>> ] KD[H+S]
strong base (KR)!g'
