Anal. Chem. 1994,66, 147-154
CydodextrbModified MSceliar Electrokinetic Capillary Chromatography Separations of Benzopyrene Isomers: Correlation with Computatlonally Derived Host-Guest Energies chrktlrte L. Copper and Mkhael J. Sepanlak' DspaHmnt of Ch&stry, Untversny of Tennessee, Knoxvk, Tennessee 37996 1600
Ceaeral adjwtment of system reteation is often inadequate to resohertnxturllysidarcempcnmdshmiccllnrekctrelriDetic crpilhry chromrtograpky (MECC). Tbe use of cyclodextTim (a) aa laobilephwe rdditivea is described for sepnmtions of structarrlisonen. CDS msbomto providedramatic pnd se&ctin effects om tbe retention of beytopyrcse isomers. Etficient sepratiolls of six methyl-substitnted .ad t h e l - p i t h - beazopyrene ~ isomers are presented. DerkrLlzcdy-CDdiscrhninrtesi~4~eusrdrptitutioanlisomers Lcssthannative yCD. A comparisonof sodiumdadecyld a t e (SDS)and sodium cholate (NaC) surfactant systems indicates that S D s C D mowt pbasea are more favorable for separation of benmpyreaeiroseft. Possible separation mechinism are dlscdandevdPttedbasedonresultsof tbmesfudiea. The c o m p r t . W proecdues of a commercial moleculard e l i n g system are modificdad " Ito ' interactioaenergies fanriollshost-gwet ( i . e . , ~ - C D - b e n z o p y n a e ) c o ~ ~ By aae of thc average of tbe five best energy values from inter8ction ene#gy matrsceg correct elution order is predicted fatkl-poaitandmoetoftbemetbyl-sukrtituted aio-". Camiderationof differ@ poesiBkCDbslrzopyre!~oriamtntiom mmt be mule to correctly predict elutiom order. bpectba of the interaction eaergy matrices reo~roab~eaetgyburiers~twouldiahibitincllrrion compkx fowrtioa A very useful characteristic of capillary electrophoresis (CE) and micellar electrokinetic capillary chromatography (MECC) is the ability to change quickly and easily the mobilephase composition of these techniques to allow for increased separation selectivity. As representative examples, CEMECC running buffers have incorporated soluble polymers for size discrimination in oligonucleotide separations,'J complexing agents or crown ethers to separate metal cati0nsP1~ and organic solvents and/ar bile salt micelles to separate hydrophobic c o m p o ~ n d s . ~Since * ~ the micellar (secondary) phase in MECC is a fluid component of the system, it can also be easily changed to alter ~electivity.2~ (1) C h k , B.K.; S e p W M. J. 1. Microcolumn Sep., in press. (2) SepU M. J.; Powell, A. C.; Swaile, D. P.;Cole, R. 0. In Cupillory E l s r r a p M q Omumn. P., Colkun, J. C., Edr.; Academic Pm8, Inc.: New Yotk, 1992; Chapter 6. (3) Smile. D. P.;Scpni.t M. J. A w l . Chem. 1991,63,179-84. (4) swpDt J.; Jcliask, I.; Smkm-Keul-& E. 1. Chromufogr. lWLB, 452,571-90. (5) B.lchunu, A. T.; scpnirk,M. J. AM/. Ch" 1987.59,1466-70. (6) Cole, R.0.;Oorre, J.; Ohgcd, K.; Sepaniak, M.J. 1.Chromurogr. 1991,557, 113-23.
Recently, cyclodextrins (CDs) have been shown to improve CE-MECC separations by imparting unique selectivity based on solute size and geometry or through chiral recognition.8*9 Examples of their successful use include separations of polycyclic aromatic hydrocarbons (PAHS)~JOand watersoluble and fat-soluble vitamindl and chiral separations of binaphthyl compounds,12 dansyl amino acids,13 and several drug mole~ules.~ While these reports demonstrate the utility of CD mobile-phase additives, they provide little fundamental information regarding the electrophoretic-chromatographic characteristics of CD separation systems. In particular, the interaction energies between the CD host and the guest solute have not e n correlated with separation performance. Such correlations could provide useful guidelines for designing capillary electrokinetic separationsinvolving various CDs (see below) or other organized media. The work reported herein concerns the chromatographic behavior of benzopyrene isomers in CD-modified MECC. Benzopyrene compounds are members of an important class of compounds, the PAHs. PAHs are a significant type of environmental pollutant. One particular benzopyrene isomer, benzo[a]pyrene (BaP), is a proven ~arcin0gen.l~ This creates a need to develop analytical methodologies for isolating and characterizing benzopyrenes in real samples. In our previous work, we quantitatively determined BaP in a shale oil sample by using CD-modified MECC, noting a very strong interaction with y-CD.l0 The magnitude of this interaction provided impetus for the studies presented in this report. CDs are a product of the enzymatic digestion of starch. They are cylinderically shaped with a hydrophobic cavity and a hydrophilicexteri~r.~~ Themast commonCDs arecomprised of six (a-CD), seven (8-CD), or eight (7-CD) glucopyranose units. There is considerablediscrepancyin the literature with regard to cavity dimensions. However, cavity diameters of 5.0, -6.3, and -8.0 A for a-,8-,and y-CDs, respectively, are commonly quoted.l5 Crystallographic data used in this
-
(7) Holland, R. D.; Scpaniak, M. J. AM/. Chem. 1993,65, 1140-6. (8) Terabe, S.;et al. 1. Chromurogr. 1.990,516, 23-31. (9) Nishi, H.;Fukuyama. T.;Tcrabc. S.1. Chromurogr. 1991,553, 503-16. (10) Copper, C. L.; Staller, T.D.;Sepniak, M.J. Polycyclic Aromuf. Compd. 1993,3 (2), 121-35. (1 1) Oag, C. P.; Ng,C. L.; Lee, H. K.;Li. S.F. Y.1. Chromurogr. 1991, 547. 419-28. (12)Sepaniak. M.J.; Cole, R. 0.;Clark, E. K. 1. Uq.Chromurogr. 1992, 15, 102340. (13) Miyasbita, Y.;Tersbe, S.Chromurogrum 1990.11, 6 7 . ( 14) Ejoneth, A. H d h k of Polycyclic AromuficHydmurbow, Marcel Dekkcr, Inc.: New Yarlr. 1983; Chapter 4. (lS)Strattcn, C. BioPhurm 1991.4.44-51.
Analyilcel ChemEStty, Vd. 66,No. 1, Jencrery 1, 1994 j 4 1
work (see below) place the inner hydrogen distance across the opening of the -y-CD cavity at 9.2 A. Inclusion complex formation between guest solutes and the CD's cavity is dependent upon the geometry, size, and physiochemical properties of the solutes. Hydrophobic interactions predominate in the cavity. However, these interactions can act in concert with polar or hydrogen-bondinginteractionsthat occur with hydroxyl groups located on the outer lip of the CD cavity. Derivatized CDs have also been used as mobile-phase additives. The hydroxypropyl cyclodextrin derivative HP-yCD used in this work has a larger, less rigid cavity than its native counterpart, which allows for the inclusion of an even wider range of solute molecules.16 Another characteristic of HPy-CD is that it is more water soluble than native y-CD. This permits the use of higher mobile-phase concentrations, which increases its chromatographic effects. Moreover, the chiral nature of native and derivatized CDs provides discrimination in separations of many optical isomers.9J2J3J7J* Computer-aided molecular modeling and conformational analyses of solute inclusion by CDs have been performed and some of these studies, which are significant to this work, have been summarized.19-21 Lu et al. performed molecular mechanical calculations on 8-CD complexes and found a correlation between the van der Waals stabilization energies of the complexesand dissociationconstants measured by cyclic voltammetry.20 Kostense et al., through systematic rotation and translation of the guest molecule about the central axis of the host's cavity, were able to derive the position of the guest that results in the most favorable interaction energy with the host.z1 These authors demonstrated that their data were in good correlation with crystallographic data. Data obtained through molecular modeling studies have been used to describe the chromatographic behavior of solutes in HPLC.22,23Arnold et al. compared the retention times of several solutesusing a 8-CD bonded column to computationally derived interaction energies of the corresponding &CD inclusion complexes.23 They were able to correlate interaction energies and HPLC retention times within a given series of similarlysubstitutedbenzenes. Correlationsbetween different classes of substituted benzene compounds were not observed. Similar studies could be useful in explaining the effects of CD additives in MECC. In this paper, we demonstrate effective separations of various benzopyrene isomers and illustrate the effects of CDs and other mobile-phase components on retention. We also describe efforts to correlate the retention behavior of benzopyrene isomers in CD-modified MECC with computationally derived host-guest interaction energies. Such correlations (16) PersonalcommunicationwithDr.StweJ.Herschleb,Pharmatec, Inc.Alachua, FL. (17) Szcjtli, J. Cyclodextrins and Their Inclurion Complexes; Akademiai Kiado: Budapest, Hungary, 1982. (1 8) Bender, M. L.; Komiyama, M. Cyclodextrin Chemistry; Spcinger-Verlag: Berlin, 1978. (19) van Helden, S.P.;van Drooge, M. J.; Claesacns, A. J.; Jansen, A. C.; Janssen, L. H. Carbohydr. Res. 1991, 215, 25140. (20) Lu, T. X.; Zhang, D. B.; Dong, S.J. J. Chem. Soc., Faraday Trans. 2 1989, 85, 1439-45. (21) Kostensc, A. S.;Van Helden, S.P.; Janssen, L. H. M. J . Compur.-Aided Mol. Des. 1991, 5, 525-43. ( 2 2 ) Camilleri, P.; Murphy, J. A.; Saunders. M. R.; Thorp, C. J. J . Cornput.Aided Mol. Dcs. 1991,5, 277-84. (23) Arnold, E. N.; Lillie, T. S.;Bccsley, T. E. J. Llq. Chromafogr.1989,12 (3). 337-43.
140 Ana!~ticaIChem&tty,Vol. 88, No. 1, January 1, lQg4
are complicated by the complex and dynamic nature of CDmodified MECC. In this separation system, CD and mobile phases are transported at the electroosmotic flow velocity, The which is very dependent on experimx$.il micellar phase is electrophoretically retarded and involves a dynamic equilibrium between free surfactant and micelle. Moreover, solute-micelle association can w u r by a variety of mechanisms? and solute-CD interactionsdo not necessarily require inclusion complex formation.23 The general effm of CD-solute interactionis to reduce retentiontimes by inhibiting solute-micelle association. Despite the aforementioned complications, elution orders for most of the benzopyrene isomers studied are correctly predicted on the basis of calculations of host-guest interaction energies.
EXPERIMENTAL SECTION Apparatus. The MECC apparatus used in this work is described in detail el~ewhere.2~ Briefly, it consisted of the following: a Hipotronics Model 840A high-voltage power supply (Brewster, NY) connected with platinum wire electrodes toinlet and outlet reservoirs (made from microcentri€uge vials). Separations were performed using fused-silica cap illaries purchased from Polymicro Technologies,Inc. (Phoenix, AZ) with inner diameters (i.d.'s) of 25 or 50 pm,the spaaller diameter finding utility with laser fluorometricdetection (see below). Two detection schemes were used in this work. Excitation for laser fluorometric detection was provided by an Omnichrome (Chino, CA) Model 307420M He-Cd laser (20 mW, 325 nm). Specific details concerning the optical arrangement and alignment p r d u r e s for this mode of detection can be found in refs 10and 24. Alternately, a linear (Reno, NV)Model 204detector,with a CZE flow-cell adaptor, was used (at 254 nm) to perform UV abmrbance detection. The laser fluorometric approach was desirable in that the higher sensitivity permits the use of smaler inner diameter capillaries and lower sample concentrations, both measures being important in achieving high separation efficiency.'O Host-guest interaction energies were calculated using the SYBYL 6.0 molecular modeling software developedby T r i p Associates, Inc. (St. Louis, MO) and run on a Evans & Sutherland (Salt Lakecity, UT) workstation. TheseSYBYL energy calculationsconsider ionic, hydrogen-bonding,and van der Waals interactions between host and guest. For some of this work, the SYBYL DOCKING routine was employed. This routine, which translates the molecules in order to attain the optimum interaction complex, suffered from a general problem of stopping at localized energy minima. In order to circumventthis problem, weutilized the SYBYL programming language (SPL) to generate a routine that permitted operator control over the host-guest positioning prior to the energy calculation (see Energy Calculations below). Reagents. HPLC grade solvents were obtained from Baxter Scientific (McGaw, IL). Buffer and surfactant reagents were obtained fromSigma Chemical Co. (St. Louis, MO)and were reagent grade or better. PAH samples were obtained from Aldrich Chemical Co. (Milwaukee, WI) or donated by the National Cancer Institute Chemical Carcinogen Repository (24) Scpaniak, M. J.; Swaile, D. F.; Powell, A. C. J. Ckromafogr. 19g9.480.18596.
(Kansas City, MO). Native and derivatized CDs were obtained from Pharmatec, Inc. (Alachua, FL). Structures for molecular modeling were generated using the sketch option of the SYBYL system, for the benzopyrene isomers, or obtained from the CambridgeCrystallographicData Files (Cambridge, England), for y-CD. The running buffer consisted of HPLC grade water and was 10 mM in NaH2P04 and 6 mM in NazB407. All mobile phases were comprised of runningbuffer with either sodium dodecyl sulfate (SDS) or sodium cholate (NaC) surfactants added at the concentrations stated in the text. Cyclodextrins were used at either 10 or 15 mM in concentration. When appropriate, 2-propanol was employed in the described proportions (% v/v). Procedures. The fused-silicacapillaries were cut to a length of 60 cm. A region of the polyimide coating was removed to generate a transparent region for detection 10 cm from the outlet end of the capillary. The capillary was then placed in a cell holder compatiblewith the mode of detection to be used. A syringe needle was affixed to the inlet end to facilitate rinsing. Before use, the capillaries were rinsed with 0.10 M NaOH for 5 min to remove any impurities from the surface. The NaOH solution was thoroughly rinsed from the column with HPLC grade water before filling the column with mobile phase. The column was then suspended between two mobilephase reservoirs making sure that the inlet and outlet ends were at equal heights so as to prevent hydrostatic flow. Operating voltage was applied across the capillary for 10 min after each mobile-phase change to allow the system to equilibrate. Hydrostatic sample injections were manually performed by placing the inlet end of the capillary in the sample solution and raising it 10 cm above the outlet end for 10-15 s. The inlet end of the capillary was then rinsed with HPLC grade water and replaced into the inlet reservoir. Separations were performed using an applied voltage of 20 (50-pm4.d. capillaries) or 25 kV (25-pm-i.d. capillaries). Approximate capacity factor (k') values were calculated as described by Terabe et al. using eq 1.25 The variable to is k' = (t, - to)/(to(l-
tJt,))
(')
the column void time, appearing as a solvent disturbance, while t m is the effective micelle migration time. Typically, a hydrophobic molecule that associates completely with the micelle can be used to measure tm. However, when using mobile phases that contain y-CD, we have yet to find a compound for which retention time ( t r ) is not altered by the presence of the CD; e.g., the PAH corenene (FW 300) exhibited almost as strong as an interaction with y-CD as BaP, despite being much larger than the y-CD cavity dimension. For this reason, t m was measured by injecting BaP into a mobile-phase system containing no CD (i.e., t m = tr(BaP without CD)). Subsequently, duplicate injections of the separation mixture (containing various combinations of benzopyrene isomers) were performed using a mobile phase containing CD, followed by another measurement of tm. The values of tm, to, and t r used to calculate k'are averages from these injections. (25) Terabe, S.;Otsuka, K.;Ando, T.Anal. Chem. 1985,57, 834-41.
1
1
9
10
I1
FblJre1- Depictions of B e (on left) and B@ guest molecules and y-cyclodextrin host molecule.
Energy Calculations. Interaction energies of various CD (host)-benzopyrene (guest) inclusion complexes were computed using the T r i p Force Field of the SYBYL 6.0 molecular modeling program. All structures used in these calculations were first minimized using the energy minimization function of SYBYL. Specifically, Gasteiger-Huckel charges were assigned to the molecule and minimization was then performed with 10 000 designated as the number of possible iterations to ensure complete minimization. The SPL routine that was generated in-house (see above) allowed the guest molecule's position to be systemmatically altered relative to that of the CD cavity. The initial and final positions of the guest molecule and translational increments could be defined. Rotational increments at the various translational positions could also be specified. For this work, the starting position of the benzopyrene isomer was centered directly above the cavity with 10 A between the centers of mass (see Figure 1). From this position, the isomer was translated toward the CD cavity at 0.25-A intervals to a final position 5 A beyond the point at which the centers of mass were aligned. At each translational position, the benzopyrene isomer was rotated 360' at 15' increments. In our studies, it was determined that a symmetrical PAH such as anthracene has equivalent interaction energies upon 180' rotation while the unsymmetrical BaP isomers do not. Consequently, a full rotation about the CD cavity is necessary to ensure the collection of the full range of energy data. Increments less than 15' were desired but precluded by computing time constraints. The total energy of the complex was computed and recorded at each translational-rotational position. The interaction energy of the complex at each of these positions was calculated by subtracting the energies of the uncomplexed guest and host molecules from the total energy values obtained. Thus, a three-dimensional energy matrix (translation, 60 positions/ rotation, 24 positions/energy value) was generated. In selected cases, the SYBYL DOCKING routine was used to locate interaction energy minima starting at given translationalAnalyticallhemistry, Vol. 66, No. 1, January 1, 1994
149
rotational positions (see below). However, this was very time consuming and generally not employed. Different initial orientations between benzopyrene and CD were used to generate energy matrices. Referring to Figure 1, showing structures for BaP and benzo[e]pyrene (BeP), orientation 1 corresponds to an initial entry of positions 2 and 3 of BaP or 4 and 5 of BeP (“fat end”) into the cavity, while orientation 2 corresponds to an initial entry of positions 8 and 9 of BaP or 10 and 11 of BeP (“skinny end”). Orientation 3 (used only for unsubstituted BaP and BeP) corresponded to a sidewise approach (note that for BeP there are two equivalent approaches). In all thesecases, the long axis of the benzopyrene is either parallel (orientations 1 and 2) or perpendicular (orientation 3) to the cavity axis during translation. Safety. Benzopyrene compounds are known carcinogens, so caution should be exercised when one is working with them.I4 The benzopyrene solutions were prepared under a ventilated hood. Disposable latex gloves were worn while working with these compounds. Care was taken to disposeof waste solutions properly.
RESULTS AND DISCUSSION
CenerrJ!3eparationConsiderations. As with reversed-phase HPLC, solutes eaa be separated using MECC on the basis of differencesin their hydrophobic character. However, MECC‘s utility is greatly diminished when one is dealing with compounds that are highly hydrophobic. These compounds associate totally with the micellar phase and coelute at t,.5 Resolution in MECC is optimal when solutes have relatively small k’ values (typically less than 5).25 If conventional reversed-phaseorganic modifiers (e.g., acetonitrileand certain alcoholQare added to the MECC mobile phase, the resolution of hydrophobic molecules is improved due to a reduction in k’ and an extention of the elution range.2 A small concentration of 2-propanol (5% v/v) is added to the mobile phases used in this work to moderately extend the elution range. Unfortunately, large organic modifier concentrations disrupt micellization. Micelles formed from bile salts (e&, NaC) have been found to bevery useful in separationsof hydrophobic compounds.6 Bile salt micelles are more polar than those formed from SDS and also tolerate higher concentrations of organic modifier. Another approach to the separation of hydrophobic compounds involves adding CD to the MECC mobile phase. In addition to specific solute-CD interactions, the CD can function in a general sense to impart organic character to the mobile phase and thereby reduce k’. The distribution of BaP between aqueous (as), CD, and micellar (M) phases can be expressed in terms of the micellar-aqueous ( K M )CD-aqueous , (KcD),and CD-micellar (KcD-M)distribution coefficients as depicted below. BaP(*q)
-
I60 42 38 23 30 25 >60 >60 9.3
a Mobile phase contains SDS (0.02 M) and CD (0.01 M). Mobile phrrse.mnt”SDS(0.02M) CD(O.O16M),and6%(v/v)2-prppanoL M o b l phaw contam SDd (0.02 M), 22.5%(v/v) acetonitrtle, and 6% (V/V) 2-RrOWOL
of -5.8 A for this isomer. The larger, more flexible cavity of the H P y C D results in relatively uniform interactions with the isomers. Thenarrow range in k’values (3.54.4) indicates that H P y C D is functioning largely as an organic modifier (although k’s cannot be reduced to those low values by use of conventional organic solvents as modifiers). Conversely, the more rigid 7-CD cavity provides excellent discrimination between the isomers, with k’values ranging from 1.5 for BaP to 29 for 6-Me BaP. Duplication of the procedure described in the Experimental Section for determining k’values resulted in the run 1 and run 2 data appearing in the table. While
these k’s are only approximate capacity factor^,^^^^ good repeatabilityin the measurements is evidenced by the closeness of the two runs. When coupled with efficiencies of greater than 1Os plates/m (see below), successful separations of benzopyrene isomeric mixtures are possible (as demonstrated in Figures 3 and 4). The elution order of the 1-position isomers in Figure 3 is consistent with structurally related polarity considerations. The polar, hydroxy-substituted isomer can hydrogen bond with the hydroxyl groups found on the lip of the CD when the skinny end of the BaP isomer is inserted into the cavity. Hence, association with the CD is expected to be stronger than with the other 1-position isomers, and this is consistent with the observed elution order. It is much more difficult to predict the elution order of the methyl-substituted isomers observed in Figure 4 on the basis of a general inspection of structure. The high efficiency of these separations is somewhat surprising and may provide insights into the separation mechanism. When BaP is injected into a mobile phase without y-CD, it elutes at tmwith an efficiency of - 8 X 104 plates/m. This modest efficiency for MECC may be limited by the adverse effects of micelle polydispersity. The contribution of micelle polydispersity to band dispersion in MECC increases in magnitude with increasing k’.2 The addition of CD greatly reduces k’and improves efficiency (e.g., N k p is -2 X lo5 plates/m in Figures 3 and 4). While this would be expected Anel).ticsl Chefnktry, Vol. 88, No. 1, Janwty 1, 1994
161
Tablo 2. NotmaUzed and Actual Hort-&~ed Inlrrotkn Ernrgka (korUmol) for Dmr.nt orkntrtkn and Mothod8 ol Tram Emgy Yatrbt V a h a
compds (in elution order) Figure 2 BQP BeP Figure 3, BaP 1-hyd BQP 1-eth BQP 1-Me BQP Figure 4 BaP 3-Me BQP 5-Me BQP 7-Me BaP 1-Me BQP 10-Me BaPe 6-Me BQP'
guest-host orientationa (average best five values -min) orientation 1 -minb +mid orientation 2 orientation 3 (-9.9) (-5.4)
lood (-9.9) 63 67 54
(-5.8) X 2 (-7.6) X 2
(-14) (-6.5)
rood (-19.4)
lood (-14.0)
89 81 84
92 83 87
composite'
composite using adjusted contour vole
lood (-35.5) 76 (-27.1) 100" (-23.9) 78 75 71
100" (-3204) 65 27 69
lood (-9.9)
lood (-14.0)
rood (-23.9)
lood (-3204)
81 81 105 54 102' 1W
94
88 83 71 71 102e 87e
111 78 46 69 111e 7 le
84
37 87 103e 73'
0 Orientations and procedures for com uting composite values are described in the text. SYBYL DOCKING energy minimiition waa performed using the coordinatesof the 10 Best (nonminimized)values as starting points. The average of the five best of these minimizedvalues is reported. e Total contour volume is computed aa the sum of all localized, nonminimized interaction energies lower than a cutoff value determined by the average value with guest-host se aration of 10.0 A. d A normalized value of 100 is assigned to the average (or sum for the last column) negative interaction energy (in kcal/mo$ of BQP.e 6-Me BQPand 10-Me BaP have independent minimized energies that are -3 kcal/mol higher than the other isomers (-13 versus -10 kcal/mol).
to reduce the polydispersity problem, it might be expected to create problems with slow mass transfer. Slow mass transfer should be particularly problematic if one of the phases involved in the mass transfer is the 5% (v/v) 2-propanol mobile phase, in which BaP and its isomers are essentially insoluble. Consequently, the observed high efficiency provides strong evidence that the mass-transfer process actually involves a direct exchange of BaP between micelle and CD, rather than the aforementioned stepwise mechanism. Molecular Modeling Considerations. Data Presentation. Data from the interaction energy matrices generated via implementation of the SYBYL molecular modeling procedures, described in the Experimental Section, are presented in Table 2. Interaction energy (in kcal/mol) is reported for BaP and BeP; the former is assigned a value of 100. The interaction energies of the other isomers are based on this normalized 100 scale. Three benzopyreneCD orientations (see Experimental Section) and a composite sum of the interaction energies for those orientations are found in the table. Two approaches to representing the interaction energy data appear in the table. Most of the reported interaction energies are the averages of the five best (most negative) energy values in a given matrix. However, the last column is effectively the totalvolumeof the contour formed by the energy values in the matrix (see Table 2 footnotes). The excessive time required to perform the interaction energy calculations with the computing system employed in this work necessitated certain omissions in the table. In particular, the sidewise approach of the benzopyrene isomer (orientation 3; see Experimental Section for details) was performed only for the unsubstituted benzopyrene molecules. In the case of BeP there are two equivalent sidewise orientations (see Figure I), each yielding an average energy of -7.6 kcal/mol. While BaP does not possess the same symmetry with regard to this orientation, the SYBYL calculation yielded the same value of -5.8 kcal/mol for the two sidewise approaches of that isomer. 152
Analytical Chemistry, Vol. 66, No. 1, January 1, 1994
Scope of Approach. The main focus of the molecular modeling work described herein was to (1) assign micropositional energy states (Le,, create the interaction energy matrices) and (2) properly interpret the significance of the states to establish correlations (if they exist) with benzopyrene isomer retention in CD-modified MECC. However, it is instructive at this point to enumerate what is not considered in these computational studies of benzopyrene isomer elution behavior. In relation to eq 2, elution order should correlate with trends in KCD.M. The modeling that was performed provides energy (stability) information without consideration of solvation effects (i.e., isomers, y C D , and complexes are all considered to be in sterile environments for energy calculations). Moreover, entropy contributions to KCD.Mare not considered. The similar hydrophobicity, molar volume, and lack of symmetry of the solutes used in these studies provides some justification for these simplifications. It is expected that solvation effects (e.g., the expulsion of solvent or SDS monomer from CD cavity by the isomers) will be similar among the isomers and not contribute significantly to trends in KCD-M. Entropy changes can be determined by measuring k' as a function of temperaturea2S However, our apparatus is not equipped for such studies; furthermore, the entropy changes are also expected to be similar for these nonsymmetric BaP isomers. A near-linear relationship was observed between retention and [CD]/ [BaP] over a fairly wide concentration ratio range. A distinctive change in slope might indicate a change in the stoichiometry of complexation. Based on this study, CD and benzopyrene isomers were assumed to form 1:1 complexes in this work. The benzopyrene isomers all coelute at rm when 7-CD is not included in the mobile phase, indicating that KM is very large. However, the last column of k'values in Table 1 (no C D maximum tolerable percent (v/v) organic modifier2) illustrates that there are significant and detectable differences in k'(hence K M )between some of the isomers. Nevertheless,
we did not consider KM when predicting elution order. As discussed above, the observed efficiency indicates that a stepwise process involving the KM distribution (see eq 2) is probably not the principal separation mechanism. This is further supported by a lack of correlation between the high organic modifier and r-CD k’data in Table 1. Except for the early elution of BaP using both systems, the ordering of k’ values is very different. The efficiency was very poor for the high organic modifier case. In fact, for certain isomers, it was obvious that concentrations required for detection exceeded or nearly exceeded solubility in the 22.5% (v/v) acetonitrile mobile phase. Predictions of Elution Order. The correct elution order for the separation of BaP and BeP (see Figure 2) is predicted without considering orientation 3. However, when orientation 3 is not included, the disparity in composite interaction energies is very large, leading to the prediction that the isomers should be easily separated with the BeP exhibiting a very large k’. When the sidewise orientation% considered, the BaP and BeP composite interaction energies are considerably closer in magnitude and more in line with the k’data in Table 1 and separation shown in Figure 2. The potential significance of the sidewise insertion of BeP into the cavity (Le., partial inclusion complex formation) is fairly obvious from its structure. Such is not the case for the substituted BaPisomers, and orientation 3 was not considered for those isomers. The importance of considering both logical orientations (fat and skinny end approaches) for inclusion complex formation with the substituted isomers also can be seen from the data in Table 2. Correct elution order is not predicted for the separations in Figures 3 and 4 based individually on orientation 1 or orientation 2 data. However, the composite (five-value average) is nearly correct for these separations. The two exceptions are the late-eluting isomers, 10-Me BaP and 6-Me BaP, in Figure 4. The SYBYL interaction energies predict a much stronger interaction with y C D than is indicated by their chromatographic behavior. Although we are uncertain as to the reason for the poor predictions for these isomers, we did note that they exhibited unusually large independent energies following use of the minimization function of SYBYL (see Table 2 footnotes). Since these independent values are subtracted from the energy of the complex to determine the interaction energy, this would indicate a stronger interaction with 7-CD than would otherwise be predicted. In principle, the SYBYL DOCKING routine could be employed to minimize the energy of the interaction complex using each of the translational-rotational positions in the energy matrices as starting points. Because this would require a prohibitive amount of computer time, a modified version of this approach was applied to the 1-positional isomers for orientation 1 (see Table 2 footnotes). For this selected case, the DOCKING routine rendered the computed interaction energies more similar and changed the ordering of the isomers. Clearly, the utilization of the minimization routine represents a refinement of the molecular modeling procedures used in this work. However, this would require greater computing capability than exists with our system, but could yield better correlations between interactions energies and chromatographic behavior.
The presence of energy barriers that inhibit the formation of stable inclusion complexes could result in greater retention (less CD interaction) than would be predicted on the basis of the five best energy values used in Table 2. These considerations were also made by Arnold et al. concerning HPLC with c y ~ l o d e x t r i n snoting ~~ that other researchers have observed alterations in elution order based on changes in flow rate and temperature.28 If energy barriers are significant, the total contour volume (see last column in the table) might be expected to correlate better with chromatographic behavior, since that method of representation would include consideration of nearly all the energy values in the matrices generated in this work. As can be seen from the table, the correlation between elution order and composite interaction energy based on contour volume is weak at best. Future efforts will include evaluating statistical mechanical approaches toward treating the interaction energy data. The significance of energy barriers in influencing retention should also depend on the relative velocities of the chromatographic phases involved. The systems used to generate Figures 3 and 4 exhibited velocities for the mobile, y C D , and SDSmicellar phases of approximately 17, 17, and 3.8 cm/min, respectively. Thus, if solute mass transfer between phases involves separate CD-mobile phase exchange and mobilephase-SDS micelle exchange steps, energy barriers would not be important since there is no relative velocity between CD and mobile phase. Conversely, if the solute is exchanging directly between micelle and CD (as hypothesized above), energy barriers could exert an influence on retention. The high separation efficiency achieved in this work indicates that significant energy barriers do not exist. Using the interaction energy matrices, we generated plots for both orientations of each isomer of number of rotational positions (out of 24 possible) versus translational position that have interaction energies above a reasonable cutoff value. A low number indicates stability of the complex throughout a complete rotation of the benzopyrene isomer. An inspection of these plots did not reveal any significant energy barriers that would inhibit insertion into the cavity to reach a position of maximum stability (Le., the observed efficiency and the computational energy values are consistent in this regard). In summary, high-efficiency separations of hydrophobic benzopyrene isomers are shown to be possible using MECC with CD mobile-phase additives. The ability to utilize molecular modeling techniques to make predictionsconcerning the effects of CDs on retention can greatly simplify methods development and provide insights into pertinent retention and band-broadening mechanisms. Despite the complexity of the MECC system and the similarity between benzopyrene isomers, it is encouraging that elution order is predicted with fair accuracy. Only three orientations and two-dimensional positioning (translation and a perpendicular rotation) were considered in this work. Improvements in molecular modeling systems (hardware and software) should reduce the time for computations and providegreater freedom with regard to hostguest positioning. It is expected that this will enhance the predictive capabilities of this methodology. With other (28) Isaaq, H. J.; Glennon, M. L.; Weiss, D. E.; Fox, S. D. In Ordered Media in
ChemicalSeparationr;Hinze,W. L., Armstrong, D. W., Eds.;ACSSympium Series 342; American Chemical Society: Washington,DC, 1987; Chapter 15.
A~&ticaiChemistry,Vol. 66, No. 1, January 1, 1994
159
separation systems and samples, considerations of entropy and solvationchanges may be important to predicting retention behavior.
ACKNOWLEDGMENT This work was sponsored by the Division of Chemical Sciences,Office of Basic Sciences, United States Department of Energy, under Grant DE-FG05-86ER13613, with The University of Tennessee, Knoxville. Support was also contributed by The Procter & Gamble Co. and by Merck & Co.,
154 Anelytical Chemktry, Vol. 66, No. 1, &nary 1, 1994
Inc. The authors thank the NCl Chemical Carcinogen Repository at Midwest Research Institute for their generous gift of methyl-substituted BuP samples. They also thank Mike Lawless, T r i p Associates, Inc., and Edward Wolpert, University of Tennessee, for their help in developing the computer programs used in this work. Rece~hredfor review June 9, 1993. Accapted October 6, 1093.. Aktract published in Adoancc ACS Absrracrs, November 15, 1993.
