Effect of the mobile phase on antibody-based enantiomer separations of amino acids in high-performance liquid chromatography.

Abstract

The effect of the mobile phase parameters flow rate, temperature, pH and ionic strength, as well as the addition of various organic modifiers on the enantiomer separation of various aromatic alpha-amino acids was investigated using two antibody-based chiral stationary phases that have opposing stereoselectivity. On both columns, a decrease in flow rate or temperature resulted in increased interaction with the retained enantiomer. It was found that the retention factor k2 depends on the affinity between the analyte and the immobilized antibody and is not independent of the flow rate. Optimum separations of all amino acids investigated were obtained at pH 7.4 on both columns. While increased k2 values were obtained at low ionic strength on the anti-D-amino acid antibody column, no such effect was observed on the anti-L-amino acid antibody column. The addition of organic modifiers did not improve separations. In all studies, the unretained enantiomer eluted with the void volume.

Full text

Page 1

Journal of Chromatography A, 1049 (2004) 85–95
Effect of the mobile phase on antibody-based enantiomer separations of
amino acids in high-performance liquid chromatography
Oliver Hofstetter∗, Heather Lindstrom, Heike Hofstetter
Department of Chemistry and Biochemistry, Northern Illinois University, DeKalb, IL 60115, USA
Received 12 May 2004; received in revised form 29 July 2004; accepted 2 August 2004
Available online 28 August 2004
Abstract
The effect of the mobile phase parameters flow rate, temperature, pH and ionic strength, as well as the addition of various organic modifiers
on the enantiomer separation of various aromatic ?-amino acids was investigated using two antibody-based chiral stationary phases that
have opposing stereoselectivity. On both columns, a decrease in flow rate or temperature resulted in increased interaction with the retained
enantiomer. It was found that the retention factor k2depends on the affinity between the analyte and the immobilized antibody and is not
independent of the flow rate. Optimum separations of all amino acids investigated were obtained at pH 7.4 on both columns. While increased
k2values were obtained at low ionic strength on the anti-d-amino acid antibody column, no such effect was observed on the anti-l-amino
acid antibody column. The addition of organic modifiers did not improve separations. In all studies, the unretained enantiomer eluted with
the void volume.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Enantiomer separation; Antibodies; Chiral stationary phases, LC; Mobile phase composition; Amino acids; Proteins
1. Introduction
One of the most popular techniques for the analysis and
direct separation of enantiomers is HPLC utilizing a chi-
ral stationary phase (CSP). Chiral selectors commonly used
for the preparation of CSPs include oligo- and polysaccha-
rides, macrocyclic antibiotics, alkaloids, synthetic polymers,
?-donor/?-acceptor systems, crown ethers, ligand exchange
selectors, and various proteins [1,2]. Despite their success,
these chiral selectors are generally not tailor-made for spe-
cific analytes, which can make the identification of a suitable
CSP a time-consuming exercise. We have recently demon-
strated that tailor-made CSPs for the direct separation of
enantiomers in HPLC can be prepared by immobilizing suit-
ably raised stereoselective antibodies onto a solid support
material [3]. Using the stereoselective interaction between
monoclonalanti-?-aminoacidantibodiesandd-andl-amino
∗Corresponding author. Tel.: +1 815 753 6898; fax: +1 815 753 4802.
E-mail address: ohofst@niu.edu (O. Hofstetter).
acids as a model system, we showed that such immunoaffin-
ity stationary phases possess predicted selectivity and, if op-
erated under mild isocratic conditions, are surprisingly sta-
ble. By employing phosphate buffered saline (PBS), pH 7.4,
as the sole mobile phase, we were able to reuse the same
antibody-columns for more than 2000 separations over a pe-
riodof3years.Thisisincontrasttovariousotherreportsthat
demonstrated separation of stereoisomers in immunoaffinity
systems[4–9].Inthosestudies,mixturesofenantiomers,dis-
solvedinaneutralbuffer,werepassedthroughacolumncon-
taining antibodies covalently linked to agarose beads [4–9]
or silica [9]; after the unbound enantiomer was washed clear,
elution of the bound enantiomer was achieved by altering to
a mobile phase that disrupted its interaction with the anti-
body. As is typical for classical affinity and immunoaffinity
chromatography systems [10–12], elution of bound material
required harsh conditions, i.e., a drastic change in pH [9], or
addition of organic solvents [4,6–8] or chaotropic salts [5].
Such severe elution conditions invariably cause protein de-
naturation and considerably shorten column lifetime, which
rarely exceeded 100 separation cycles [9].
0021-9673/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2004.08.002

Page 2

86
O. Hofstetter et al. / J. Chromatogr. A 1049 (2004) 85–95
Antibodies are glycoproteins, which are produced by
the immune system of vertebrates in response to invading
pathogenic microorganisms and “non-self” biological mate-
rial, called antigens. Through their specific interaction with
discretestructuresontheantigentheymarkitforelimination
by other components of the immune system such as phago-
cytes and the complement system [13]. It is well known that
antibodies can be raised against low-molecular weight com-
pounds, haptens, if these are conjugated to suitable carriers,
e.g.,proteins;suchantibodiesmaybestereoselectiveandmay
distinguish between the enantiomers of chiral haptens [14].
Despitethefactthatantibodiesaremadefromchiralbuilding
blocks, namely l-amino acids and sugars, and therefore pos-
sess an inherent chirality, their stereoselectivity appears to
be primarily based on the specific interaction between amino
acid residues in the binding site and complementary func-
tional groups and moieties in the hapten structure [3]. The
most abundant antibodies in blood are immunoglobulins of
the IgG class, which possess two identical binding sites and
have a molecular weight of approximately 150,000Da [13].
As is the case for other types of protein–ligand interactions,
such as enzyme-substrate, receptor-hormone, or lectin-sugar
systems, the binding forces between an antibody and an anti-
gen are of purely non-covalent nature and involve electro-
static interactions, electron acceptor–electron donor forces
(hydrogen bonds), and non-polar interactions (Lifshitz-van
der Waals forces) [15,16]. The strength of interaction, there-
fore,isaffectedbyenvironmentalparameterssuchasthetem-
perature,andthepH,ionicstrengthandpolarityofthesolvent
in which binding occurs. The effect of the mobile phase on
protein–ligand interactions has successfully been utilized in
a number of protein-based chiral separations to modulate the
interaction between the immobilized chiral selector and the
analyte, and to optimize separations [17,18].
Here, we describe how the mobile phase parameters flow
rate, temperature, pH and ionic strength, as well as the addi-
tion of various organic modifiers influence enantiomer sep-
aration of several aromatic ?-amino acids on two antibody-
based CSPs that possess “opposing stereoselectivities” and
bind to either d- or l-?-amino acids.
2. Experimental
2.1. Chemicals
POROS-OH was obtained from PerSeptive Biosystems
(Cambridge,MA).N,N?-disuccinimidylcarbonate(DSC)and
dimethylaminopyridine (DMAP) were from NovaBiochem
(La Jolla, CA), HPLC-grade acetonitrile, ethanol, methanol,
1-propanol, and 2-propanol were purchased from Sigma (St.
Louis, MO). Inorganic salts were from ACROS/Fisher (Fair
Lawn, NJ); all other chemicals were from Sigma (St. Louis,
MO).
Water was purified using a MilliQ water system (Milli-
pore, Bedford, MA). Phosphate buffered saline (PBS) was
prepared according to reference [19] and adjusted to pH 7.4
with 0.1N HCl. All amino acids were of the highest purity
available.d-Tryptophan,d-andl-phosphotyrosine,d-andl-
p-aminophenylalanine,andd-andl-phenylalaninewerepur-
chased from Sigma (Deisenhofen, Germany). l-Tryptophan
andd-tyrosinewerefromAldrich(Munich,Germany),d-and
l-DOPA were from Fluka (Neu-Ulm, Germany). l-Tyrosine
was kindly provided by Degussa (Frankfurt, Germany).
2.2. Monoclonal antibodies
Monoclonal antibodies were produced as previously de-
scribed [20]. In brief, 8-week-old BALB/c mice were immu-
nized with conjugates of keyhole limpet hemocyanin and ei-
ther p-amino-d-phenylalanine or p-amino-l-phenylalanine,
preparedbydiazotization,incompleteFreund’sadjuvantfol-
lowing a standard immunization protocol. For the produc-
tion of monoclonal antibodies splenocytes were fused with
NS0 myeloma cells using polyethylene glycol and hybrido-
mas were selected in hypoxanthine/aminopterin/thymidine
medium [21]. Large quantities of the anti-d-amino acid anti-
body secreted by clone 67.36 and the anti-l-amino acid anti-
body produced by clone 29.2 were obtained by the prepara-
tion of ascites fluid. The antibodies were purified by ammo-
nium sulfate precipitation followed by ion-exchange chro-
matography on DEAE-Sephacel (Amersham Biosciences,
Piscataway, NJ). The antibodies did not contain any impuri-
tiesasdeterminedbysodiumdodecylsulfate-polyacrylamide
gel electrophoresis. The binding characteristics of the anti-
amino acid antibodies have been described in detail else-
where [3,20,22]. Their affinities are highest for structures
that resemble the hapten, i.e., bear aromatic side chains;
however, also amino acids with different side chains, in-
cluding aliphatic, charged and non-charged residues, are
stereoselectively recognized by the respective antibodies.
The exquisite stereoselectivity of the antibodies has previ-
ously been utilized in enzyme-linked immunosorbent as-
says [20,22,23], flow-injection immunoassays [24,25], im-
munosensors [26–28], and chromatography [3,20] for enan-
tiomer detection and separation.
2.3. Chiral stationary phase
Following activation with DSC and DMAP, the POROS-
OHstationaryphasematerialwasreactedwitheithertheanti-
d-amino acid antibody or the anti-l-amino acid antibody as
described elsewhere [3]. In brief, 3.5g POROS-OH (20?m
particles)werereactedwith350mgDSCand287mgDMAP
in 17.5mL dry acetone for 1.5h at 4◦C. One fraction of the
support material (1g) was reacted overnight with 30mg of
the anti-d-amino acid antibody in PBS at 4◦C, while the
other fraction (1.5g) was reacted with 50mg of the anti-
l-amino acid antibody overnight at 4◦C under salting-out
conditions. Remaining active groups on the support were
quenched by treatment with 0.2M TRIS for 1h, followed
by extensive washing. The concentration of immobilized an-

Page 3

O. Hofstetter et al. / J. Chromatogr. A 1049 (2004) 85–95
87
tibody was determined using the dye binding method pub-
lished by Bonde et al. [29], and was found to be as fol-
lows: anti-d-amino acid antibody column, 12mg per gram
of support material (wet); anti-l-amino acid antibody col-
umn, 15.5mg/g. As previously reported, differences in col-
umn performance of the two antibody-columns are based on
thefactthattheanti-d-aminoacidantibody67.36hasahigher
affinity towards d-amino acids than the anti-l-amino acid
antibody 29.2 has towards the corresponding l-enantiomer
[3]. In addition, it was found that the column capacities of
the POROS-based antibody-columns are too low to enable
enantiomer separation of aliphatic amino acids [3]. Genetic
analysis of the antibody-binding sites has recently revealed
that both antibodies possess considerable differences in their
amino acid sequences and that they belong to different fam-
ilies [30]. Thus, it can be expected that the two antibodies
interact with their corresponding binding partners via differ-
ent amino acids, and that these interactions may show dif-
ferent susceptibilities to changes of environmental parame-
ters.
2.4. Chromatography and instrumentation
The HPLC system consisted of a Hitachi L-7100 pump
with a degasser, an L-7400 UV-detector equipped with an
analytical flow cell, and a D-7000 interface with System
Manager V 4.0 software. Injections were performed using
a Rheodyne 7725i injection valve with a 20?L loop (Hi-
tachi,Naperville,IL).ColumnswerepackedusinganAlltech
Slurry Packer Model 1666 (Alltech, Deerfield, IL).
Stainless steel columns (anti-d-amino acid antibody col-
umn: 2.3mm × 200mm; anti-l-amino acid antibody col-
umn: 4.6mm × 250mm) were slurry packed at a pressure of
160bar in PBS. No leakage of antibody was detected during
column equilibration. Unless stated otherwise, all chromato-
graphic separations were performed at room temperature un-
der isocratic conditions using PBS, pH 7.4, as mobile phase.
Columns were stored under azide-containing PBS at 4◦C
only when not used for an extended period of time. Flow
rates used in this study varied from 0.5 to 4.5ml/min. Mix-
tures of the pure enantiomers in PBS (10?L) were injected.
Tryptophan was detected at 280nm, p-aminophenylalanine
at240nm,phosphotyrosineat254nm,tyrosineandDOPAat
220nm, and phenylalanine at 205nm. The elution order was
determined by injection of the pure enantiomers as well as
by spiking. The void volume for the determination of chro-
matographic data was measured using water or buffer [31].
For the short retention times obtained in this study, errors in
the determination of the void volume affect the calculation
of chromatographic parameters. Slight variations may be ex-
plained by this fact. In addition, the results described in this
paper were obtained over a period of approximately 2 years
duringwhichtheantibodycolumnsweresubjectedtovarious
conditions, e.g., organic modifiers, which may have affected
the structural integrity of the immobilized antibodies. To en-
surethereproducibilityoftheresults,allmeasurementswere
carried out at least in triplicate. Standard deviation of chro-
matographic parameters was typically less than 10%.
2.4.1. Influence of the flow rate
The effect of the flow rate on antibody-based chiral sep-
arations was investigated at flow rates between 0.5ml/min
and 4.5ml/min. Due to the higher affinity of the anti-d-
amino acid antibody towards d-amino acids, compared to
the anti-l-amino acid antibody towards l-amino acids, more
amino acids, namely DOPA, p-aminophenylalanine, phenyl-
alanine, phosphotyrosine, and tryptophan could be baseline
separated on the anti-d-amino acid column. In contrast, only
p-aminophenylalanine, phenylalanine, and tyrosine yielded
calculable results on the anti-l-amino acid column.
2.4.2. Effect of temperature
Thetemperaturedependenceoftheenantiomerseparation
of d,l-phenylalanine was studied between 5◦C and 40◦C.
Antibody columns and buffer reservoir were immersed in a
water bath (Haake, Berlin, Germany), and the temperature
was increased in increments from 5◦C to 40◦C. The anti-
body columns were equilibrated for at least 1.5h before the
corresponding series of injections were performed at the re-
spective temperatures.
2.4.3. Effect of pH
The effect of the pH on enantiomer separations was stud-
ied using 10mM phosphate buffer ranging from pH 4.5 to
pH 10, to which sodium chloride and potassium chloride
were added to yield final concentrations of 2.7mM KCl and
137mM NaCl, respectively, which corresponds to their mo-
larity in PBS. Columns were equilibrated with each buffer
for at least 1.5h. Separations using the anti-d-amino acid an-
tibody column were carried out at flow rates of 1ml/min for
p-aminophenylalanine, tryptophan, and DOPA and 2ml/min
forphenylalanine.Separationsoftyrosine,phenylalanineand
p-aminophenylalanineontheanti-l-aminoacidantibodycol-
umn were carried out at 1ml/min.
2.4.4. Effect of ionic strength
To study the influence of the ionic strength on the sep-
aration of d,l-phenylalanine, the molarities of NaCl and
KCl contained in the phosphate buffer mobile phase were
changed by adding the chlorides to a solution of 8.5mM
Na2HPO4and 1.5mM KH2PO4adjusted to pH 7.4. The an-
tibody columns were equilibrated for at least 1.5h before the
corresponding series of injections were performed at the dif-
ferent ionic strengths. Separations were carried out at flow
rates of 2ml/min using the anti-d-amino acid antibody col-
umnand1ml/minontheanti-l-aminoacidantibodycolumn,
respectively.
2.4.5. Effect of organic modifiers
To study the effect of organic modifiers on the separa-
tion of d,l-phenylalanine, the organic modifiers methanol,
ethanol, 1-propanol, 2-propanol, and acetonitrile were added

Page 4

88
O. Hofstetter et al. / J. Chromatogr. A 1049 (2004) 85–95
to the PBS buffer. Using the anti-d-amino acid antibody col-
umn, the following maximum percentages of organic mod-
ifiers were employed: methanol and acetonitrile, 5% (v/v);
ethanol, 9%; 1-propanol and 2-propanol, 8%. Using the anti-
l-aminoacidantibodycolumnallorganicmodifierswereem-
ployed up to a maximum of 5%, except acetonitrile, which
was added up to 3%. Flow rates of 1ml/min and 2ml/min
were used with the anti-l-amino acid antibody column and
the anti-d-amino acid antibody column, respectively.
3. Results and discussion
3.1. Influence of the flow rate
Both chiral stationary phases used in this study were pre-
paredbyimmobilizingstereoselectivemonoclonalantibodies
to either d- or l-?-amino acids onto the synthetic high flow-
through type perfusion material POROS [32,33]. This sup-
port contains a network of large throughpores, which allow
the liquid phase to flow through the sorbent and enable con-
vective transport within the particle. Thus, diffusional mass
transfer limitations are overcome, and, based on the reduced
flow-resistance of the packing, analyses can be performed at
higher flow rates compared to conventional silica supports
[32,34]. In order to evaluate the effect of the flow rate on
antibody-based chiral separations, several aromatic amino
acidswereinvestigatedunderisocraticconditionsusingPBS,
pH 7.4, as mobile phase at flow rates of up to 4.5ml/min.
DOPA, p-aminophenylalanine, phenylalanine, phosphotyro-
sine,andtryptophanwerestudiedusingtheanti-d-aminoacid
column,whilep-aminophenylalanine,phenylalanine,andty-
rosine were separated on the anti-l-amino acid column. As
exemplified in Fig. 1, which shows the separation of d,l-
phenylalanine on the anti-l-amino acid column at 1ml/min
and 4ml/min, increasing the flow rate leads to an improve-
ment in peak shape and enables rapid separations within a
fewminutes.PlotsofRsversusflowratefortheanti-l-amino
acid column (Fig. 2) show a considerable decrease in resolu-
Fig. 1. Enantiomer separation of d,l-phenylalanine on the anti-l-amino
acid column at 1ml/min and 4ml/min. The first peak corresponds to the
d-enantiomer, the second to the l-enantiomer. Separations were carried out
at RT using PBS as mobile phase. Other conditions are given in Section 2.
Fig. 2. Resolution values of the enantiomer separations of phenylalanine
(?), tyrosine (?), and p-aminophenylalanine (?) on the anti-l-amino acid
column in PBS at different flow rates. Separations were carried out at RT.
Other conditions are given in Section 2. Values represent means of triplicate
determinations. Missing error bars are obscured by the symbols.
tion as the flow rate is increased. The same general trend was
observed using the anti-d-amino acid column for all amino
acids investigated. With the perfusion material, however, the
loss in resolution is somewhat smaller than predicted by the
theory for conventional supports. According to the extended
plate height equation for large-pore supports developed by
Rodrigues [35,36], the plate height becomes independent of
the mobile phase velocity at high velocities, and the resolu-
tion Rsis expected to be independent of the flow rate. This,
however, may not hold true for affinity-based separation sys-
tems, where the slow kinetics of non-covalent molecular in-
teractions can contribute to an extra term in the plate height
equation, which is proportional to the velocity. A decrease
in column efficiency was also observed with both antibody
columns studied here. The plate number N for the separation
of phenylalanine on the anti-l-amino acid column, for exam-
ple, dropped from 1080/m at 1ml/min to 580/m at 4ml/min.
A similar loss in efficiency was observed using the anti-d-
amino acid column, where plate numbers were lower by a
factor of eight. The overall low efficiencies can be attributed
to the slow dissociation kinetics, the large particle size of
the POROS material as well as the high flow rates. In a pre-
viously published study on the separation of various amino
acidderivativesanddrugsusingPOROS-immobilizedbovine
serumalbuminasachiralstationaryphase[37],wefoundthat
although the loss in resolution at increased flow rates is less
significant than predicted by the theory for conventional sup-
ports, it is not independent of the flow rate as suggested by
the Rodrigues equation. Based on chromatographic theory,
the product of the resolution Rsand the square root of the
flow rate√F is expected to be constant. Therefore, the prod-
uct of
two different flow rates, and the ratio R?sof the resolutions
at these flow rates should have a value of one [37]. Table 1
summarizes the relationship between the change in flow rate
andthechangeinactualresolutionR?sfortheenantiomersep-
√F?, which represents the square root of the ratio of

Page 5

O. Hofstetter et al. / J. Chromatogr. A 1049 (2004) 85–95
89
Table 1
Relationship between the change in flow rate and the change in actual resolution R?s
Chiral selectorAnalyte
F2/F1=F?
4/1 = 4
3/1 = 3
4/1 = 4
3/1 = 3
4/1 = 4
3/1 = 3
4/1 = 4
3/1 = 3
4/0.5 = 8
3/0.5 = 6
4/1 = 4
3/1 = 3
4/2 = 2
4/3 = 1.33
4.5/2 = 2.25
4/2 = 2
√F?
2
1.73
2
1.73
2
1.73
2
1.73
2.83
2.45
2
1.73
1.41
1.15
1.5
1.41
Rs,2/Rs,1= R?s
0.91/1.53=0.59
1.02/1.53 = 0.67
1.71/2.75 = 0.62
1.83/2.75 = 0.67
1.81/3.11 = 0.58
2.03/3.11 = 0.65
0.69/1.17=0.59
0.8/1.17 = 0.68
0.78/1.73 = 0.45
0.98/1.73 = 0.56
1.01/1.62 = 0.62
1.07/1.62 = 0.66
0.92/1.19 = 0.77
0.92/1.03 = 0.89
1.19/1.84 = 0.65
1.20/1.84 = 0.66
R?s
1.19
1.16
1.24
1.16
1.16
1.12
1.18
1.18
1.27
1.37
1.24
1.14
1.09
1.02
0.98
0.93
√F?
Anti-l-amino acid antibody
p-Aminophenylalanine
Tyrosine
Phenylalanine
Anti-d-amino acid antibody Phosphotyrosine
DOPA
Tryptophan
Phenylalanine
p-Aminophenylalanine
Conditions: PBS, pH 7.4; room temperature.
aration of several amino acids on both antibody columns. In
most cases, the values of R?s
indicates that the perfusion-type support material has some
advantage over conventional supports. However, a slightly
higher loss in resolution at increased flow rates was observed
for the enantiomer separations of p-aminophenylalanine on
the anti-d-amino acid column, which might be based on the
higheraffinityofthed-enantiomerofthisaminoacidtowards
the immobilized antibody. The major advantage of the per-
fusive support appears to be the greater range of flow rates
that can be applied; this is especially valuable in cases where
the retained enantiomer possesses a higher affinity towards
theimmobilizedantibody,whichwouldleadtolongretention
times and extreme peak-broadening at lower flow rates [3].
The use of higher flow rates in such cases allows relatively
rapid elution of the retained enantiomer under mild isocratic
buffer conditions as a well-defined peak.
√F?are greater than one, which
Fig. 3. Relationship between the retention factor k and the flow rate. Separation of (a) p-aminophenylalanine (triangles) and phosphotyrosine (diamonds) on
the anti-d-amino acid column, and of (b) tyrosine (squares) and p-aminophenylalanine (triangles) on the anti-l-amino acid column. Filled symbols represent
the d-enantiomer, open symbols the l-enantiomer. Separations were carried out at RT using PBS as mobile phase. Other conditions are given in Section 2.
Values represent means of triplicate determinations. Missing error bars are obscured by the symbols.
Aplotoftheretentionfactorkasafunctionoftheflowrate
(Fig. 3) suggests that, for the immunoaffinity-based enan-
tiomer separations investigated here, k is not independent
of the flow rate as would be expected according to chro-
matographic theory for isocratic separations. As exemplified
in Fig. 3, the decrease in k with increasing flow rates ap-
pears to be more prominent for analytes that have a relatively
higher affinity towards the immobilized antibody. Thus, the
k2values for d-p-aminophenylalanine decrease from 7.33 at
2ml/min to 5.60 at 4.5ml/min on the anti-d-amino acid anti-
bodycolumn,whilek2ford-phosphotyrosineremainsalmost
constant between 1ml/min and 4ml/min (Fig. 3a). Similarly,
the decrease in k2is more prominent for l-tyrosine than for
l-p-aminophenylalanine on the anti-l-amino acid antibody
column (Fig. 3b). This disparity may be based on different
dissociation rate constants. While association rate constants
of antibody–hapten interactions are typically high and may