DAVID R. BUSS THEODORE VERMEULEN
OPTICAL ISOMER SEPARATION Quest for a New Biochemical Technology Unwanted optical isomers are effectively removed from synthetic biochemical materials by stereoselective sorption and extraction. These methods owe their success to the nonidentical manner in which each of a pair of optical isomers interacts with another asymm etri ca 1 m ol e c ul e AUTHORS David R. Buss is a Chemical Engineer in Chemical Process Research and Development at T h e Upjohn Go., Kalamazoo, Mich. Theodore Vermeulen is Professor of Cfiemical Engineering at the University of California, Berkeley. This review i s part of a study of biochemical separations supported by the Sational Institute of General Medical Sciences, U. S. Public Health Service, under grant GM-08042. T h e authors thank Profescors C. W . Tobias, D . S. Noyce, J . Cason, and D . G. Howery and Messrs. C. Edson, J . Woodrow, J . M . Krochta, G. H. Robertson, and F. Y . Pan for their assistance in this study. 12
INDUSTRIAL A N D ENGINEERING CHEMISTRY
ptical isomers are encountered widely in nature, and are often needed for pharmaceutical purposes, for experimental biochemical studies, or as intermediates in organic syntheses of products destined for such uses. When an asymmetric-carbon structure is synthesized by living media, almost always only one of the two possible mirror-image isomers results. Thus optical activity is an innate property of life and of most chemical materials of biological origin. Because an ever-increasing number of syntheses of biochemical materials is conducted in vitro, a steadily growing need arises for supplies of pure enantiomers. Figure 1 shows the L- and D-isomers of aspartic acid, (one of the amino-acid components of protein) with their corresponding structural formulas. The particular configuration used here to portray these isomers brings all three polar groups into a triangular arrangement in a single plane. Here the asymmetric carbon atom, to which four different groups are bonded, is indicated by a bold-face C ; in the schematic structure underneath, the two polar groups attached directly to C are shown by heavy circles. An asymmetric pair, such as L- and maspartic acid, will rotate the plane of polarized light by equal amounts in opposite directions ; hence the alternate designation, “optically active.” The initials d- and 1- refer to righthand and left-hand rotations; while D - and L- indicate configurations of families of compounds related to some one dl- pair. Asymmetric isomers or “enantiomers” (opposites) will interact identically with any symmetric molecule but nonidentically (stereospecifically) with any other asymmetric molecule. The market price of various optical isomers is a measure of present-day supply and demand. Approximate current prices, normalized to a 1-kg basis, and
0
+
0
-c/
-
HH ,\3, N,\/'/ C'\,\ 0 C '0'
/\ 0
0
H
00 0
H
\
H
RC\ H0
0
U
Figure 1. L- and D-isomers of aspartic acid with structural formulas
TABLE I. M A R K E T PRICE OF SOME OPTICAL ISOMERS ($/KG) L
Compound
Alanine Valine Leucine Isoleucine Serine Threonine Lysine Arginine Aspartic acid Asparagine Glutamic acid Glutamine Cysteine Cystine Methionine Phenylalanine Tyrosine Proline Tryptophan Histidine Ornithine Norvaline Nor leu cine Homoserine Ethionine Adrenaline a-Aminobutyric acid 2-Amino-I -butanol Amphetamine sulfate 2-Benzylamino-I propanol Camphene Camphoric acid Camphorsulfonic acid a-Chlorosuccinic acid Citrulline Coniin Ephedrine a-Hydroxy isocaproic acid Limonene Mandelic acid a-Methyl benzyl amine Menthol Malic acid Pantothenic acid a-Phenylglycine p-Pinene Thyroxine (Na-salt) Tocopheryl acetate Usnic acid
14
DL
300 600 60 600 800 450 25 50 100 35 12 150 100 35 135 400 20 280 400 90 1,650 6,000 6,000 18,000 14,000 2 )000 10,000 75 700
25 35 60 150 90 100 65 2,000 15 30 75 200 1,300 1,500 15 80 65 1,000 140 180 680 80 80 1,800 1,900 ,800 80 10 70
1,000
2,000
,000 3 75 40 1,400 5,000 350 1,800
12,000 55 75
14,000 35 25
75 60 65
25 40 5 140 50
55 000 14,000 ~
140 30,000 800 700
D
1,200 1,500 1,200 2,500 3,000 1,500 15,000 10,000 200 95 210 4,500 2,000 20,000 900 900 9,000
500 600 9,200 9,200 75,000 14,000 10,000 75 400
25 30 40
12,000
40 75 75 65 140 150 4 60,000 3,500 4,500
INDUSTRIAL AND ENGINEERING C H E M I S T R Y
quoted usually for laboratory quantities, are shown in Table I. ‘4 comparatively low price for a D L - mixture usually indicates a synthetic source, whereas a lower price for the L- (or D-) isomer suggests a natural origin. When a DL-mixture is produced by large-scale synthesis, usually the L-isomer is wanted, and the D-iSOmer serves as inert filler or even as a damaging impurity-. Removal of the unwanted isomer, possibly with subsequent racemization (conversion to DL-mixture) and recycle, would have evident economic advantages. Thus a strong incentive exists to study the resolution of optical isomers, together with the stereospecific molecular interactions that make resolution possible. From a unir-operations viewpoint the processes of solvent extraction and sorption will often be more efficient and more generally applicable than crystallization. This survey therefore omits the relatively wellknown crystallization techniques, reviewed authoritatively by Secor (97) and Eliel (28),and focuses on optically active solvents and sorbents as resolving agents. This approach to separating optical isomers has received limited attention thus far, probably because of the small separation factors observed. However, as more is learned about stereospecific interactions, it is likely that specially prepared sorbents and solvents of higher selectivity will become available. This paper surveys existing experimental knowledge on the resolution of optical isomers by means of their stereospecific interactions with asymmetric sorbents and solvcnts, as published u p to about January of 1968. Also, as an aid to the further use of synthetic asymmetric sorbents, the known methods of preparation of such sorbents are reported (excluding optically active polymers formed from optically inactive monomers), The number of reported instances of enantiomer “resolution” (separation) by stereoselective sorption on asymmetric sorbents is quite small. Partial reviews of sorption separations exist (73, 28, 38, 66, 69, 97, 175). However, even today, some confusion persists between the separation of diastereoisomers (nonmirror-image compounds containing two asymmetric centers) on an inactive column packing, and the resolution of enantiomers by stereospecific sorption on an asymmetric column packing. The first case is not so unusual, since diastereoisomers (unlike enantiomers) have different gross physical properties. Although the thermodynamics for the resolution of optical isomers have been considered (3, 97))’precise data on such systems are almost nonexistent. This fact is believed due to the general lack of availability of opdcally active solvents for experimentation, rather than failure of such solvents generally to exhibit stereospecificity. Many familiar synthetic organic solvents are racemic mixtures of optical isomers. A list of such compounds
TABLE II.
INEXPENSIVE UNRESOLVED MATERIALS ($/KG) . .
Comoound
1-Amino-2-propanol sec-Butyl alcohol sec-Butylamine Butyl lactate 2-Chlorobutane 3-Chloro-l,2-propanedio 2-Chloropropionic acid 1,2-Diaminopropane
1,2-Dichlorobutane 1,2-Dichloropropane 1-Dimethylamino-1 -propano 2-Ethylhexanoic acid 2-Ethyl-1 -hexanol 2-Ethylhexylamine Ethyl lactate 1,2,G-Hexanetriol Mercaptosuccinic acid 3-Methoxy-1-butanol a-Methylbenzyl (or a-phenylethyl) alcohol 2-Methyl-2,4-pentanediol 4-Methyl-2-pentanol 2-Methylvaleric acid 2-Phenoxypropionic acid 2-Phenylbutyric acid 1-Phenyl-2-amino-l-propanol 1,2-Propanediol
2,2,4-Trimethyl-1,3-pentanedio
Price
3 2 20 4 40 7 20 7 40 20 10 2 1
5 4 5 18
3 6 2 2
5 20 20 50 2 3
is given in Table 11. If indeed strong stereoselectivities can be identified between particular enantiomers and such optically active solvents, these solvents might be resolved, and in turn be used repetitively to resolve optical isomers. Three-Point-Interaction Model
Diastereoisomer formation is a classical method used to resolve enantiomers (28, 38, 97). I n this method a racemate (equimolal mixture of enantiomers) is allowed to form a chemical derivative with a different, pure optically active material. For example, a racemic acid, DL-A, reacts with a pure optically active base, L-B, forming two salts, D-A.L-B and L-A-L-B which are diastereoisomers. Since these two derivatives are not mirror images, a separation may be devised for them that is based on differences of their physical properties. Usually the most efficient method of separation is crystallization, as crystal structure is particularly sensitive to small variations in molecular architecture. T h e success of such a separation depends on interactions of A . B with several other AeB’s in the crystal matrix. Chromatographic resolution of racemates by sorption
onto an optically active sorbent from optically inactive solutions is explained by a different type of diastereoisomer formation (28). I n this case the asymmetric sites are fixed at isolated spots within an insoluble porous matrix. When a racemate, DL-X,sorbs onto the asymmetric site, L-S, two diastereoisomers, D-X. L-S and L-X.L-S are formed, which do not have identical stability. The enantiomer forming the less stable diastereoisomer passes through a chromatographic column faster. Here the dominant factor is the interaction between X and S; the effect of the surrounding matrix or of other sorbed species seems entirely negligible. Moreover, the “diastereoisomer,” to be effective, cannot be formed by a single bond. Both theory and experience indicate that contact between sorbate and sorbent must occur at several points for resolution to occur. Since enantiomers differ only in their three-dimensional structure, the positions of a minimum of three groups must be identified to distinguish between members of an enantiomorphic pair. The importance of “three points of contact” was first recognized for stereospecific enzyme reactions (83). Later it was used to explain the resolution of racemates by asymmetric sorbents (24) and by optically active solvents ( 3 ) . I n practice, it is both accurate and informative to represent the interactions of S and X by a three-point interaction model. Aspartic acid molecules can be used to typify the interaction of either a sorbent-surface site or a solvent molecule with a pair of enantiomers. Specifically, in Figure 2, we consider the possible configurations for contact between maspartic acid (on the right, in the figure) with either D- or L-aspartic acid (on the left), the projections being made by folding the left-side “solute” over the right-side “sorbent” so that the relatively planar triangular faces of the two molecules achieve a three-point vertex-to-vertex contact. Because both molecules are zwitterionic, a strong dimer will be formed by the “sorbent” with either a Dor an L-solute, having a two-point interaction represented by the quadrupole
OCO
The two-point interaction is almost entirely nonselective; the precise arrangement of atoms away from the two points of contact, in either solute molecule entering into the complex, cannot have much effect on the sorbent molecule or upon the bonds between them. I t is when the molecules already bound in a two-point contact swing around to develop a third contact, and find an additional energetically favored point of attachment-and only then-that the asymmetries affect the relative strength of complex formation. A dimer strongly bonded at three contact points will greatly increase the equilibrium constant for forming the total (three-point plus two-point) of that type (for instance, D-D or L-L) of dimer. Figure 2A shows D-solute and n-sorbent in a configuraVOL. 6 0
NO. 0
AUGUST 1 9 6 8
15
tion which will superimpose to form the strong twopoint interaction and also- an acid dimerization by hydrogen bonding at the third point. The probable structure of this third contact is
hO--H-T -C
\
0-H--0
pc-
An L-solute with an L-sorbent would form the mirrorimage complex, relative to Figure 2A, and its formation would involve the same AH and equilibrium constant. (In the molecular models of Figure 2, each individual functional group in the plane of contact is rotated into the location that provides the clearest view of its structure. These rotations will not correspond exactly with the atomic positions for the strongest interaction,) Figures 2B and 2C show two of the strongest threepoint contacts that can be formed between an L-solute and a D-sorbent (or between an L-sorbent and a D solute), but probably neither structure is quite as strong as the two-point complex; hence these structures will only slightly influence the overall L-~IUS-Dequilibrium. I n Figure 2B, the L structure is rotated 120" clockwise, relative to its position in Figure 1. I t can then form a n ionic bond a t one contact point, and these two hydrogen bonds a t the remaining two points:
Figure Za. Sfmng 2-poinl inleracfion between 0-solure and D-sorbcnf
. In Figure 2C, the negative charge of the L-zwitterion (solute) has shifted from the carboxyl group nearer the amine group to the carboxyl farther away; this form is only about one eighteenth as stable (27a, 37a). After this unfavorable shift has been made, the complex that forms involves the same bonds as in Figure 2A, but still carries the handicap of being about one eighteenth as probable. The relative likelihood of forming these particular structures becomes more evident if one surveys all the potentially possible forms of three-point complex, as in Figures 3 and 4. For the normal D-D combination, in Figure 3, the three orientations of solute molecule relative to solvent give rise to three possible complexes; one is the strongly bonded structure of Figure 2A, while the other two bring like charges together and are therefore antibonding. With a shift in negative-charge location on the solute (or alternatively on the solvent), the results are two weakly bound complexes plus a strongly antibonded one. For the normal L-D combination, in Figure 4, two structures of equal strength are formed (one of them shown by Figure 2B), along with a strongly antibonded one. With the negative-charge shift, we regain one moderately strong structure (Figure 2C, discussed above) together with two antibonded forms. An equilateral-triangle representation of the groups 16
INDUSTRIAL AND ENGINEERING CHEMISTRY
.
Figure 2b. Strong 3-poinf interacfion between ~ - m l u t cand D-sorbent (or belween L-swbenf and D-solufe)
participating in selective three-point contact seems to be a good approximation, a t least for aspartic acid. A minor extent of selective solubility for L-aspartic acid in an aqueous solution of L-glutamic acid, HOCOCHzCHzCH(NHa')COO-, has been observed by the present authors (77). For other asymmetric structures in contact, as observed from molecular models, the asymmetric selectivity will tend to disappear as the arrangement of the contact-point functional groups becomes more linear. What about the "back side" of the molecule? If we accept the equilateral-triangle model for the contact face, the molecule as a whole must be viewed as a tetrahedron with four contact faces and six edges. Between a three-point adsorbent surface and a tetrahedron, the possible structures for complexing are 12 one-point contacts, 36 two-point contacts, and 12 three-point contacts. Between two tetrahedra, there are 16 onepoint contacts, 72 two-point contacts, and 48 threepoint contacts. For pliable molecules the alternatives are still more exhaustive. In the present discussion of aspartic acid, we have considered only the contact-edge and the contact-face that are most likely to be included in strongly associative structures. Used in the manner just outlined, the three-point interaction model is even today a valuable qualitative tool. It can be used for interpreting chromatographic resolutions of optical isomers on asymmetric sorbents. In some cases it can be used to predict the type of asymmetric site needed for resolving a given pair of enantiomers; the model predicts that an amino acid attached to an insoluble matrix in a way that preserves its zwitterionic character will be stereospecificfor other zwitterionic amino acids, and such a result has been observed (93). Further, the theory indicates that an alkaloid or other optically active base which successfully resolves racemic acids by diastereoisomer formation and crystallization will not necessarily effect the samc - 4 u t i o n s when it is
incorporated into an insoluble matrix as an asymmetric adsorbent or anion-exchange center; the mechanism of resolution is quite different in the two cases. Sorption Techniques
The use of asymmetric sorbents for resolution of optical isomers has .been retarded by a lack of availability of such sorbents and by a limited understanding of how they function. The time is approaching when synthetic resins will bc prepared with stereoselectivity comparable to that of enzymes. For this purpose, additional general methods are needed of attaching asymmetric sites to sorbents, along with more understanding of the best structures. Generally only partial resolutions have been achieved on asymmetric sorbents A problem then is how to detect the fact that resolution has occurred. Usually optical rotation is the analytical method of choice. However, unless large quantities of materials are used in high concentrations and the resolutions are quite good, rotation measurements often are not very sensitive. This fact accounts for the large number of experiments with mandelic acid (a-hydroxyphenylacetic acid), and its derivatives and complex ions; they have high specific rotations and dissolve well in numerous solvents. When separations are complete and pure isomers are available, the problem is much simpler as in the case of some paperchromatography resolutions. Batch experiments showing stereospecific sorption. Early attempts to show stereospecific sorption centered around the use of naturally occurring protein materials such as wool, silk, and casein. Since chromatographic techniques were not then developed, these experiments were necessarily of batch type. Willstatter (111) is credited with having first considered the problem in 1904; he used aqueous solutions of racemic alkaloids in contact with wool, but found no selective sorption. In 1922 Ingersoll and Adams (45) prepared an azo-dye
VERY STRONG (Fig. Za)
WEAK
DOLIBLY REPELLENT (Fig. 1)
REPELLENT
REPELLENT
WEAK
MODERATE
MODERATE (Fig.2r:)
REPELLENr
DOUBLY REPELLENT
MODERATE [Fig..zbl
REPELLENT
00: 00 00 00 0 0 0 0 Fiprc 3. Possible
D-D
complexes: 7 p r i m p (
cturc, 2 lesser
Fipre 4. Possible L-D complexes: 3 major structures
s ~ t w e (L-L s would be minor imagcr of this) VOL. 6 0
NO. 8
AUGUST 1 9 6 8
17
TABLE 1 1 1 .
BATCH EXPERIMENTS SHOWING SPECIFIC SORPTION
Sorbate
Sorbent
Dye produced by reaction of diazotized dl-phenyl($-aminobenzoy1amino)acetic acid with dimethylaniline dl-rn-Azo-0-naphthol-mandelic acid dl-2,3-Di- [p-( p '-sulfophenylazo)anilino]butane [Co(en)3]Br3.3Hzo [Co(en)n(NH3)Cl]Brz [Co(en)zCl~lCl.H~O Kg[C0(0~)31.3.5H 2 0
Solvent
Wool Wool Wool d-Quartz
[Co(dg)z(NHs)CIl [CO( C O ( N H ~ ) ~ ( OalC1e' H ) ~ ]2 H z 0 d-Quartz 1-Quartz Wool Casein Wool Wool Wool
[Co(dg)z("s)C11 dl-Mandelic acid dl-Mandelic acid dl-a-Naphthyl-glycolic acid dl-Mandelic acid &Mandelic acid dl-p-Methoxy mandelic acid dl-p-Ethoxymandelic acid dl-p-Pentoxymandelic acid dl-p-Heptoxymandelic acid dl-p-Decoxymandelic acid
dl-Diphenyleneoxide-glycolicacid dl-$-Phenylmandelic acid dl-0-Naphthylglycollic acid dl-3-Anthrylglycollic acid dl-a-Methoxyphenylacetic acid dl-a-Methoxymandelic acid dl-or-Methoxymandelic acid
-
b
pair by diazotizing d- and Z-a-p-aminobenzamido-aphenylacetic acid and coupling the products with dimethylaniline. Preliminary results indicated that the rate of sorption of the isomers onto wool was different, but further studies by Brode and Adams in 1926 (75) showed them to be the same. Other reports (79,88) of the specific sorption of asymmetric dyes by wool about the same time are also quite questionable. The first unequivocal indication of specific sorption came not with the use of proteins, but with quartz powder. Tsuchida and coworkers in 1935 (707)found that aqueous solutions of cobalt complexes became optically active when placed in contact with finely ground d-quartz (Tablc 111). I n 1936 Tsuchida (706) also found that if d-quartz sorbed a d-cobalt complex more strongly, then 1-quartz would sorb the 1-isomer more strongly. This expected reciprocity of selectivity cannot be checked, many times, since naturally occurring specific sorbents such as wool usually exist in one enantiomorphic form only. I n 1951 Bradley and coworkers (8, 77) began a series of experiments on selective sorption using wool, casein, and silk as sorbents. Aqueous solutions of racemic 18
Hz0
(?1 d-Isomer d-Isomer 1-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer I-Isomer d-Isomer d-Isomer d-Isomer d-Isomer &Isomer d-Isomer &Isomer d-Isomer &Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer
Refs.
en = ethylenediamine, ox = oxalate, dg = dimethylglyoxirne. in each table are listed i n chronological order.
a Others (75 43) were unable to rcproduce this work.
,Vote-Referdnces
'Iliool Wool Wool
E tOH-H 2 0 HzO EtOH-HBO(1: 1 ) (3:l) (1:l) (1:l) (1:l) (3:l) (1:l) EtOH-HzO(1: 1 ) (4:l) (3:l) EtOH-H*O(l:1 ) (1:l)
Optical isomer more strongly sorbed
INDUSTRIAL A N D E N G I N E E R I N G CHEMISTRY
mandelic acid and related materials (Table 111) became optically active in contact with wool and casein; silk effected no change. They took care to ensure that matter from the sorbent did not dissolve and thus influence optical-rotation measurements. I'Vool was found to ha\e a capacity of 0.45 mol/kg for mandelic acid, about the same as for hydrochloric acid and chloroacetic acid, indicating that sorption resulted mainly from salt formation with lysine or arginine residues within the protein. This would explain the absence of specific sorption by silk, since it contains no lysine or arginine. Despite this indication of selective sorption, wool has never been used for a chromatographic-type resolution. T h e reasons are probably that the packing density of wool is low and that the kinetics of sorption are very poor, due to wool's heterogeneous nature. Wool is believed to have amorphous and crystalline regions ( 4 ) , the former being accessible for sorption. Bradley (72) found that 757, of wool's ultimate capacity was used within 4 minutes, but that full equilibrium was "reached" only after 5 to 6 days. With mandelic acid and wool, after 4 min the excess of d- over I-isomer sorbed was 2.1'%, but at equilibrium it was 3.5y0,.
Diffusion of a racemate through an optically active The first chromatographic resolution of a compound membrane was considered by Klingmuller and Gedenk optically active because of an asymmetric carbon atom was performed by Henderson and Rule in 1938 (44). in 1957 (53). They reported the recovery of an optically They eluted dl-p-phenylene-bis-iminocamphor on a lacactive solution when racemic tartaric acid diffused tose column, using a light petroleum fraction as solvent. through a synthetic collodion membrane which contained 2-(~-gluco-~-gluco-hepto-hexahydrohexyl)benz-After the color of the camphor derivative reached the bottom of the glass column, extrusion and analysis of the imidazole. lactose packing showed the adsorbate to be dextrorotaI n 1941 Martin and Kuhn (76) resolved partially tory in the upstream part, and levorotatory downstream. racemic mandelic acid by a rotating wool belt, the ends They found too that dl-m-p-naphtholazomandelic acid of which were maintained a t different temperatures. is also resolved partially on lactose (43). I t was felt that if optical isomers had different sorption Lactose columns have since been used with nonaqueenergies, the temperature coefficients of sorption should ous solvents to resolve partially ephedrine (65),“Troger’s be different also. The experiment was not very rebase” (89),and hexahydrojoulolidine (68). Also Garproducible, and today’s chromatographic techniques maise and Colucci (31) have patented the process of would be much superior to a rotating belt. using lactose for the chromatographic resolution of Resolution by adsorption chromatography. Quartzamino acids and their derivatives, specifically including powder studies by Tsuchida, as mentioned above, DL-lysine methyl ester dihydrochloride and DL-proline. opened the way to further study of metal ion complexes. Cellulose has not shown much success as a column I n 1938 Karagounis and Coumoulos (48, 49) partially packing, despite the many resolutions performed by resolved a chromium complex chromatographically on paper chromatography discussed in the next section. d- and I-quartz (Table IV). For quartz powder, the Only DL-kynurenine (47), a single cobalt complex (59), asymmetric surface exchange sites are uniformly accessiand a single nickel complex ( 7 0 2 ~have ) been resolved ble, and the rates of exchange should be very high even partially. The last case, reported by Taylor and Busch though the sorptive capacity is low. Moreover, the in 1967 on microcrystalline nonionic cellulose, gave [MID high molecular rotation of the complexes-e.g. = 3000 for the chromium complex studied, makes it essentially complete resolution. I t appears that sorption onto a normal cellulose matrix is too slow or that possible to identify any partial resolution obtained. I n the solvent-to-sorbent ratio in a column is too high. 1950 Schweitzer and Talbott (96) extended quartzMicrocrystalline cellulose may have a more available, powder column studies to 12 complexes of cobalt and compact, and homogeneous structure thus making chromium, and showed that specific adsorption of resolution possible. transition-metal complexes is a general phenomenon. Similar to cellulose are the cross-linked dextran maEven so, the precise nature of the site-sorbent molecular terials (polyglucoses) sold under the trade name of interactions is not yet ucderstood. Sephadex (87). A partial resolution of mandelic acid Starch columns were used in 1954 by Krebs and was achieved by Leitch, Rothbart, and Rieman in 1967 Rasche (60) for resolving cobalt complexes. Starch by using a column packed with Sephadex G-25 and has a high absorption capacity, since adsorption takes eluting with an aqueous sodium chloride solution (67). place throughout the particles. Consequently, howI t was calculated that a column of 6.7 X lo6 cm would ever, the rate of adsorption is lower, and the columns be needed for complete separation. must be run at a slow rate; this factor was found to Even though quartz powder has a very high stereoaffect some separations because many complexes racespecificity, its capacity is limited by its surface area. mize with half-lives varying from a few minutes to T o overcome this type of disadvantage, Karagounis and several days at room temperature. Krebs and Diewald coworkers in 1959 (47) prepared chromatographic sorin 1956 (59) extended starch-column studies to nine bents of large surface area by coating porous alumina complexes of cobalt, two of chromium, and one of iron with optically active compounds in a thickness of only a (Table IV). This work confirmed that starch has widely few molecules. With these sorbents many racemic applicable selectivity as an asymmetric adsorbent, the mixtures were resolved (Table IV). I n a similar apbest selectivity being observed with aqueous solutions. proach, Klemm and Reed in 1960 (52) coated silicic I n 1956 Krebs and coworkers (62) also reported using acid with an optically active compound, 1-a-(2,4,5,7starch columns to resolve partially a large number of tetranitro - 9 fluoroeny1ideneaminooxy)-propionic acid, mandelic acid derivatives, and some amino acid derivawhich is known for its complexing ability with other tives (Table IV). Consistent with Dalgliesh’s concept aromatic compounds. Several aromatic racemates were of “three-point interaction” (24) they attributed the resolved partially when eluted on a column of this observed stereospecificity to hydrogen bonding between material with petroleum ether or cyclohexane as the solthe starch hydroxyls and the carboxyl, hydroxyl, and vent (57,52). amine groups of the asymmetric solutes. Other NSilica gel, which would ordinarily not be expected to derivatives of amino acids have since been resolved on exhibit specificity, may be specially prepared to sorb starch columns (64). Most recently Krebs and Schuoptical isomers selectively. Pauling (86) first suggested macher (67) have resolved various amino acid derivathe “molding” of an adsorbent to a particular molecule; tives and camphor derivatives, and isolated 100% pure and Dickey, a student of Pauling’s, first made a specific fractions in some cases.
-
VOL.
6 0 NO. 8
AUGUST 1968
19
TABLE IV.
PARTIAL RESOLUTION OF OPTICAL
-
Optical isomers
Adsorbent
[Cr(en)a]Cla. 3.5 HzO 1-Quartz powder d-Quartz powder [Cr (en)31 CIS. 3.5 Hz0 1-Quartz (300 mesh) Ka[Cr(ox)a]. 3 H20 KB[CO(OX)I] . 3 Hz0 K [Cr ( o ~ ) ~ e n ] [Cr(ox)(en)zlCl [Co(ox)(en)z]Cl~HeO [Cr(en)a]Cls.5 H20 [cr(en)a!~(SO4)r [Co (en 1 3 1Cis [Coien)zCis]Cl [CoNHa(en)nCl]Cl2 [Co(dg)z(”aPI [Co(dg)z(XHa)2IC1 Starch [Co (en ) 3 1 Cl3 [Co(NH2CHzCOO)a] Cobalt dithiocarbamate of 4methylaminophenol Starch [Co enzCOslC1 [Co en2 H4NCH2COOlC12 [Co en2 (NOz)z]N02 [Co en2 NOzClIC1 [Co enz(l\;Oa)p]NOa [Cr ( e n ) ~ l C L K~[C~(OX)~I [Co(NHzCHzCOO)a] Trimethylpyridylketoxime Co(II1) (“~)a[Fe(ox)al Cellulose Cellulose and starch Sephadex G-60 dl-fi-Phenylene-bzs-iminocamphor Lactose
K3 [co(ox)al Ni(TRI)(NOI)LHzO dl-Mandelic acid
dI-m-Naphtholazomandelic acid dl-Ephedrine “Troger’s base” Hexahydrojoulolidine DL-Lysine methyl ester.2HCl DL-Proline dl-hlalic acid (2NH4-salt) dl-Mandelic acid p-Bromo-dl-mandelic acid m-Nitro-dl-mandelic acid o-Nitro-dl-mandelic acid dl-Phenylglycine a-Chloro-dI-phenylacetic acid a-Methoxy-dl-phenylacetic acid &Mandelic acid methyl ester a-Acetyl-dl-mandelic acid a-Acetyl-&mandelic acid (”4salt) o-Nitro-dl-mandelic acid (“4salt) m-Nitro-dl-mandelic acid (“4salt) a-Acetyl-dl-mandelic acid methyl ester DL-Benzoylalanine (NH4-salt) DL-Benzoylisoleucine (NH4-salt) DL-Benzoylvaline (NH4-salt) DL-Acetylphenylalanine (“4salt)
20
Lactose Lactose Lactose Lactose Lactose Lactose Starch
Starch
INDUSTRIAL A N D ENGINEERING CHEMISTRY
Solvent
EtOH-H20(85: 15)
He0
MeOH-Hz0 (1 : 1 at 3 5 O C.) Hz0 Ha0 3.0‘14 NaCl (H20) Kaphtha-benzene (8:l) Benzene CHC13, CHCla-benzene Petroleum ether Petroleum ether Chloroform Chloroform Hz0 HzO HzO-MeOH(7 : 3 ) H20-MeOH(l : 1) HZO-MeOH(1: 1 ) l N HC1 HzO-MeOH(7 : 3 ) H20-MeOH (7 : 3 ) HzO-MeOH(7 : 3 ) HZO-MeOH(1: 1 )
Optical isomer more strongly sorbed
&Isomer 1-Isomer &Isomer d-Isomer d-Isomer &Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer 1-Isomer d-Isomer d-Isomer 1-Isomer I-Isomer 1-Isomer 1-Isomer 1-Isomer 1-Isomer d-Isomer d-Isomer I-Isomer d-Isomer (?1 1-Isomer 1-Isomer &Isomer
Hz0
&Isomer 1-Isomer I-Isomer 1-Isomer D-Isomer D-Isomer 1-Isomer d-Isomer d-Isomer d-Isomer d-Isomer d-Isomer [-Isomer d-Isomer &Isomer 1-Isomer 1-Isomer
HzO
&Isomer
I320
1-Isomer
H*O-MeOH(Z: 8 )
1-Isomer
HzO with NH3 H a 0 with NH, H20 with NHI H2O with NHB
!-Isomer I-Isomer I-Isomer 1-Isomer
Refs. 1
I SOM ERS BY ADS0 RPTl ON CH RO MATOG RAPH Y
Adsorbent
Optical isomers
DL-Formylphenylalanine 4"-( salt) DL-N- Acet ylphenylalanine DL-N-Benzoylalanine DL-N-Benzoylnorleucine DL-N-Trichloroacetylalanine nL-N-Trichloroleucine dZ-Mandelic Acid dl-Mandelic Acid DL-Kynurenine DL-Alanine DL-N-Acetylalanine DL-N-Chloroacetylalanine ~~-N-Dichloroacetylalanine
Optical isomer more strong1 sorbed
Solvent
H 2 0 with NHa
I-Isomer
Starch MeOH-HzO(2 : 3 )
HzO MeOH-HgO(2: 3 ) HzO Starch Cellulose
HzO MeOH-BuOH-PhHHzO ( 2 : l : l : l ) (and 1% HOAc)
Starch
&Mandelic acid dl-Mandelic acid dl-Mandelic acid dl-Mandelic acid dl-Phenylglycine methyl ester dl-Menthol dZ-2,2 '-Dinitro-b,b'-biphenyldicarboxylic acid dl-Phenylacetic acid dl-3,4,5,6-Dibenzo-9,1 O-dihydrophenanthrene dl-1 -Naphthyl-2-butyl ether d-9-sec-Butylphenanthrene d-Camphorsulfonic acid d-2-Butanol d-2-Bromobutane d-a-Bromopropionic acid ethyl ester d-a-Bromobutyric acid ethyl ester N-Trifluoroacetyl (TFA)-a-amino acid esters N-Trifluoroacetyl derivatives of primary amines
N-TFA-DL-alanine-t-butylester
I-Isomer I-Isomer I-Isomer d-Isomer &Isomer &Isomer &Isomer D-Isomer 1-Isomer d-Isomer &Isomer d-Isomer d-Isomer &Isomer (?)
HzO
~~-N-Trichloroacetylalanine oL-N-Trimethylacetylalanine nL-Camphorsulfonic acid ("4salt) DL-3-Bromocamphorsulfonic acid ) ("4-salt dl-6-Nitrophenic acid (Na-salt) dl-4,6,4 ',6 '-Tetranitrophenic acid (Na-sal t ) dI-Mandelic acid
Refa.
(?)
d-Isomer (?1 d-Tartaric acid on alumina (0.45 g per g AlzOa) &Tartaric acid on alumina (0.45 g per g Alz08) d-Tartaric acid on alumina (0.40 g per g A1z03) D-Glucose on alumina (0.42 g per g A1203)
Acetone
I-Isomer I-Isomer
&Tartaric acid on alumina (0.43 g per g Alz0.3) &Tartaric acid on alumina (0.43 g per g AlzOa) L-Sodium glutamate on alumina (0.35 g per g AlnOs) Lactose on alumina (0.39 g per g A1203)
Chloroform-acetone (4:l) Petroleum ether-acetone (2:l) Petroleum ether-acetone Benzene Benzene-ligroin, benzene, ether Petroleum ether
&Isomer
Petroleum ether-benzene (1 : 9 )
1-Isomer
d-Fenchane on alumina
HzO
I-a-(2,4,5,7-Tetranitro-9-fluorenylideneaminooxy)propionic acid (8,9y0 by wt. on silicic acid) I-a-(2,4,5,7-Tetranitro-9-fluorenylideneaminooxy)propionic acid (8.9% by wt. on silicic acid) Specific silica gel (made in presence of disomer) I-Ethyl tartrate
Petroleum ether
1-Isomer d-Isomer
Petroleum ether Cyclohexane
d-Isomer 1-Isomer
I-Isomer I-Isomer 1-Isomer
I-Isomer
d-Isomer Hz Carrier gas for sorbates
(? )
(?) (?)
N-TFA- isoleucine lauryl ester (capil-
Nz Carrier gas for sor-
lary column) Condensation product of phosgene with Lvaline isopropyl ester (capillary column) N-TFA-L-valyl-L-valine cyclohexyl ester on Chromosorb W (5%) (packed column)
bents NZ Carrier gas for sorbents Nz Carrier gas for sorbent at 100' C.
(?) (?) (?1 L.
somer ~~
en = ethylenediamine. ox = oxalate. dg = dimethylglyoxine. TRI = tribenzo[b f '][1 5 9]triazacycloduodecine; TFA = trifluoroacetate. for dl-2-bromobutane are now believed to r e h t from dehydrogenation during injection (20). ' ;esults were reported unreproducible (36). 5
VOL. 6 0
NO. 8
C
Two peaks
AUGUST 1968
21
silica gel (27) which potentially could be stereospecific. I n 1952 Curti and Colombo (22) prepared a stereospecific silica gel and partially resolved two racemic acids chromatographically. However, specific silica gels are unstable with respect to moderately high pH, high temperature, certain chemical agents, and even long storage (27). This instability, along with their difficulty of preparation, may account for the small amount of attention given to silica gel. The resolution of optical isomers by gas-liquid partition chromatography has attracted much interest in recent years. Karagounis and Lippold reported in 1959 resolving four compounds including 2-butanol and 2bromobutane using l-ethyl tartrate, a high boiling liquid, as the selective sorbent on an inert packing (50). I t is now believed that the two peaks for 2-bromobutane resulted from dehydrohalogenation during injection (20); the other resolutions could not be reproduced, either (36). More recently Gil-Av and coworkers have reported a series of experiments which show conclusively that optical isomers can and do have different solubilities in optically active solvents (29,33-35). They used 100-m (140,000-plate) capillary columns with N-trifluoracetylL-isoleucine lauryl ester as the optically active stationary phase to resolve a series of N-TFA-alanine esters and N-TFA-valine and N-TFA-leucine esters of 2-butanol. Derivatives with one asymmetric center gave two peaks, and derivatives with two asymmetric centers gave four peaks. T h e steric influence of the alcohol chain appeared to be important; with the IV-TFAalanine esters, the methanol ester gave only a single peak, whereas those esters with two to four carbon alcohols showed progressively better resolution. With the cyclopentyl ester, resolution was complete (34). I n another set of experiments using the condensation product of phosgene with L-valine isopropyl ester as the optically active stationary phase, a series of :V-TFA-2amino-n-alkanes and other amines were resolved (29). The data indicate that the essential feature for resolution on this stationary phase is an amino group attached directly to the asymmetric carbon. Capillary columns were used by Gil-Av and coworkers originally because of the low ratio of retention times of the enantiomers, 1.01 to 1.05. They have now dernonstrated a resolution of ;L’-TFA-dl-alanine t-butyl ester on a 2-m packed column using iV-TFA-~-valyl-~-valine cyclohexyl ester as the optically active stationary phase (35). Other lower-molecular-weight sLationary phases were not satisfactory because they bled rapidly out of the column under operating conditions. Resolution was practical because of a higher retention-time ratio, 1.197. The experiment shows that it is now possible to use preparative gas-liquid chromatography (GLC) to resolve enantiomers in certain cases. The dipeptide derivative exhibits remarkable efficiency for resolving amino acid derivatives, and further studies may identify the stereospecific interactions (analogous to an enzyme?) which in turn would point IO better stationary phases. Such studies coupled with the advances being made in 22
INDUSTRIAL A N D ENGINEERING CHEMISTRY
preparative GLC could lead to a new resolution technology. Some of the optically active stationary phases developed by Gil-Av and coworkers are commercially available in 1- to 5-g quantities for laboratory studies (701). Resolution by paper chromatography. I n 1951, after several years of analytical use of paper chromatography, it became dramatically apparent that this method was not purely a partition technique but involved active participation of the asymmetric cellulose as well. At that time Fujisawa (30) and others (58, 80, 95), in Osaka, Japan, resolved racemic tyrosine-3sulfonic acid by paper chromatography without using an optically active solvent. Soon afterward, Dalgliesh (24) published the first systematic study on resolution of optical isomers by using paper chromatography, working with derivatives of phenylalanine (Table IV). Later these results were fully substantiated by Lambooy (63). Resolution was shown by the appearance of two spots on the paper chromatogram using a chemically pure racemate, these being of equal size, shade, and intensity. Pure optical isomers gave only a single spot which corresponded to one of the spots given by the racemate. Dalgliesh pointed out that “three-point” attachment of a sorbate is the minimum necessary for stereospecificity, and that hydrogen bonding of the a-amino and carboxyl groups to the cellulose is probably the dominant interaction. Dalgliesh also supposed that a flat aromatic portion was necessary within the sorbate molecule. This qualification was loosened when a,€-diaminopimelic acid (9 7) and cystathionine ( 7 ) were resolved in 1955-56. Also the resolution of a series of alkaloids (Table V) in 1954 by Alessandro (2) by paper chromatography shows conclusively that cellulose selectivity is not limited to sorbates containing the a-aminocarboxyl group. I n 1963 DeLigny and coworkers (26) studied the effects of pH, solvent, ionic strength, temperature, and various papers on the resolution of racemic histidine, tryptophan, and kynurenine. T h e R , ratios showed that the best separations occur with neutral or slightly acidic 90y0 methanol of low ionic strength. While the type of paper did not seem to be important, the separations were enhanced by low temperature. For these compounds it was concluded that the simultaneous presence of --NH3+ and -COOgroups is required to effect separation, and that, in methanol concentrations above 90%, the equilibrium between the zwitterionic sorbate and the uncharged sorbate shifts too far toward the latter. Among the best resolutions achieved by paper chromatography are those of Zaltzman-Nirenberg, Daly, Guroff, and Udenfriend in 1966 with certain enantiomers of aminonaphthylalanines ( 7 74). I n the case of 3 - ( 3 ’amino-I ’-naphthyl)-DL-alanine, the D-isomer advanced 50Oj, faster than the L-isomer ; however, no separation occurred when 3-(4 ’-amino-2 ’-naphthyl)-DL-alanine was used.
TABLE V.
RESOLUTION OF OPTICAL ISOMERS BY PAPER CHROMATOGRAPHY Rf values
Optical isomers
DL-Phenylacetic acid DL-Kynurenine DL-Kynurenine sulfate n~-Tyrosine-3-sulfonicacid ~~-Tyrosine-3-sulfonic acid ~~-Tyrosine-3-sulfonic acid ~~-Tyrosine-3-sulfonic acid ~~-Tyrosine-3-sulfonic acid
DL-Kynurenine
0.378
0.233
(6)
0.61 0.59 0.50 0.31 (? 0.63
0.58 0.55 0.59 0.36 (?) 0.79
(30)
BuOH-HOAC (4 : 1 ) BuOH [dl-Methyl-( &phenylisopropyl)amine]-HOAc-H2O-BuOH (1 :1:1:1) 500 g BuOH, 167 g HzO, 33 g 1methyl-2-phenylisopropylamine, 33 g HOAc (paper chromatopile experiment) BuOH-HOAC-H~O (4: 1 :5 ) BuOH-HOAc-HzO (4: 1 :5 )
0.88 0.60 0.84 0.34 0.40 0.74 0.39
0.306 0.61f 0.31d
0.46 0.31
0.350 0.17d
0.303 0.293 0.274 0.258
0.317 0.307 0.293 0.272
MeOH-HzO-pyridine (80 :20 :4) MeOH-H20 (3 : 1) MeOH-H20-1 ON HC1-Pyridine (80:7.5:2.5:10) MeOH-H20 (9 :1) (1 70Glucose) MeOH-H20 (9 : 1) (1 % Raffinose) MeOH-H20 (9:1), 25°C
0.26 (LL)
0.17 (DD)
BuOH-pyridine-HBO (1 : 1 :1) (TLC using cellulose)
0.36 0.52 0.47 0.43 0.36
DL-Leucine
BuOH (170D-tartaric acid)* satd. with HzO" BuOH-H~O-HOAC
MeOH-H20
n-Butanol-H20-pyridine
(1:1 : 1)
