`fatty acids to human serum albumin
`
`Marie A. Kenyon and James A. Hamilton'
`Department of Biophysics, Housman Research Center, Boston University Medical School, 80 E. Concord
`Street, Rill, Boston, MA 02118-2394
`
`Abstract Binding of the medium-chain fatty acids (MCFA),
`octanoic (OCT) and decanoic (DEC) acid, to human serum al-
`bumin (HSA) has been studied by I3C NMR spectroscopy.
`NMR spectra at 35OC showed an apparently homogeneous
`binding environment (a single, narrow resonance for the 13C-
`enriched carboxyl carbon) at different mole ratios and pH
`values. Changes in the chemical shift of this peak with mole ratio
`and protein concentration demonstrated rapid equilibration
`( I msec) of bound and unbound MCFA and permitted a direct
`quantitation of bound/unbound MCFA. Spectra of OCT/HSA
`mixtures at 6°C revealed at least three distinct binding sites that
`fill sequentially. The observed heterogeneity of binding at low
`temperature, compared to 35"C, is attributed to a slower ex-
`change rate of OCT between binding sites. The highest affinity
`sites for both OCT and DEC have properties similar to those of
`binding sites for longer-chain fatty acids, such as the close prox-
`imity of the fatty acid carboxylate to basic amino acid residue(s).
`Interestingly, chemical shift data showed that the first mole of
`OCT and DEC either bind differently to the same site or bind
`to different sites on HSA. The rapid desorption of MCFA from
`HSA binding sites has implications for dietary regimens with
`medium chain triglycero1s.-Kenyon, M. A., and J. A. Hamil-
`ton. I3C NMR studies of the binding of medium-chain fatty
`acids to human serum albumin. J Lipid Res. 1994. 35: 458-467.
`
`binding
`
`affinities - exchange rates
`
`Supplementary key words octanoic acid
`decanoir acid
`parenteral feeding
`
`Human serum albumin (HSA) is a key transport pro-
`tein in plasma that plays an important role in lipid
`metabolism by virtue of its high-affinity and high-capacity
`binding of free (unesterified) fatty acids (FA). The physio-
`logical importance of plasma free FA and the pathological
`effects resulting from abnormally high levels of free FA (1)
`have encouraged ongoing investigations of the binding
`and transport of free FA by albumin. The high-resolution
`X-ray structure of FA-free HSA (2) shows elegantly and
`in great detail the loops and domains that were predicted
`by numerous investigators (3). Nevertheless, molecular
`details of the binding of FA, particularly those of medium-
`chain length (8 or 10 carbons), are incomplete. Although
`medium-chain fatty acids (MCFA) normally constitute a
`very minor fraction of the FA in plasma, interest in the
`
`458
`
`Journal of Lipid Research Volume 35, 1994
`
`binding of MCFA exists as the levels of these FA can be
`greatly elevated in certain disease states (4), in patients
`fed intravenous medium chain triacylglycerols (4-6) and
`in infants treated for low birth weight (7). Interest in the
`binding of MCFA to HSA also stems from the observation
`that these acids may compete with certain drugs and with
`tryptophan for binding sites on HSA (8-11).
`Interactions of MCFA with albumin have previously
`been probed by analysis of equilibrium binding data in
`the presence and absence of competing ligands (9, 12) and
`as a function of p H (12). However, equilibrium binding
`studies do not differentiate multiple binding sites with
`similar affinities and do not provide direct information
`about structural features of the binding sites. Additional
`strategies have attempted to locate MCFA binding sites
`on serum albumin by covalent modification of HSA (13)
`and by fragments of bovine serum albumin (BSA) (14, 15).
`A more recent approach for examining interactions of
`FA with albumin has been l3C N M R spectroscopy. This
`spectroscopic approach, which uses native FA with a non-
`perturbing modification (13C enrichment), can provide in-
`formation about molecular interactions in individual bind-
`ing sites, in contrast to methods that report on average
`behavior. New information has been provided about ionic
`interactions between the FA carboxyl group and the amino
`acids in the binding sites on BSA, about the location of
`high affinity sites on BSA for long-chain fatty acids, and
`about the dependence of interactions on FA chain length
`(16-20). Like classical approaches, 13C N M R spectroscopy
`has provided less information about the binding of MCFA
`to albumin (21) than about long-chain FA and, to date, most
`studies have focused on BSA rather than HSA. In this study,
`interactions of MCFA ( O C T (8 carbon) and DEC (10 car-
`
`Abbreviations: MCFA, medium-chain fatty acids; HSA, human
`serum albumin; FA, fatty acid; OCT, octanoic acid; DEC, decanoic acid;
`BSA, bovine serum albumin; CD, circular dichroism; UV, ultraviolet;
`TMS, tetramethylsilane.
`'To whom correspondence should be addressed.
`
`MPI EXHIBIT 1032 PAGE 1
`
`MPI EXHIBIT 1032 PAGE 1
`
`
`
`bon)] with HSA were investigated by high-resolution I3C
`NMR spectroscopy. NMR measurements (primarily chem-
`ical shift and lineshape) were made as a function of pH,
`temperature, and the mole ratio of FA to albumin. Our
`results illuminate molecular aspects of MCFA binding to
`HSA.
`
`MATERIALS AND METHODS
`
`Materials
`Lyophilized, crystallized HSA (A3782, primarily Lot
`127F-9310) was obtained from Sigma Chemical Go.,
`St. Louis, MO. The essentially FA-free HSA contained
`less than 0.01 mole of FA per mole of protein as deter-
`mined by gas-liquid chromatography. Ninety percent 13C
`carboxyl-enriched OCT and DEC used in this study were
`purchased from CIL, Cambridge, MA.
`
`Sample preparation
`A measured amount of HSA was dissolved in 0.56%
`KCl and the protein concentration was determined from
`the absorbance at 279 nm (22) of filtered 1:lOO dilutions
`in 7.5 mM KCl; the extinction coefficient used was
`0.55 ml mg-1cm-I (23). The protein concentration was
`
`- 93 mg/ml for all NMR samples unless otherwise stated.
`
`OCT/HSA and DEC/HSA complexes were prepared with
`aqueous potassium OCT and with aqueous potassium
`DEC, respectively. The concentration of the FA, dissolved
`in chloroform-methanol 2:1, was determined by measur-
`ing dry weights on an electrobalance (Cahn model 25,
`Cerritos, CA). The aqueous solutions of potassium FA
`were made by combining a known amount of 13C-
`enriched FA with 1.2 eq of base (1 N KOH, 0.1 N KOH).
`Sodium could be substituted for potassium, yielding iden-
`tical results. FAIHSA samples were made by the addition
`of an appropriate amount of potassium FA (depending on
`the desired mole ratio of FA to HSA) directly to a measured
`volume of aqueous protein (1.4-1.6 ml). To increase the mole
`ratio, FA was added directly to the FA/HSA complex.
`OCT/HSA and DEC/HSA mixtures were gently vortexed
`and these samples were adjusted initially to a pH of 7.4.
`During the course of some experiments, the pH of the
`samples was changed by adding small amounts of KOH
`(0.1 N, 1 N) or HCl (0.1 N, 1 N) directly to the NMR
`tube, using a microliter syringe. The pH values were
`measured by a Beckman model 3560 pH meter equipped
`with a 5-mm diameter microelectrode. pH values for the
`FA/HSA samples before and after NMR analysis differed
`by 5 0 . 2 pH units. Final pH values are reported.
`
`I3C NMR spectroscopy
`A Bruker WP-200 spectrometer operating at 4.7 T
`(50.3 MHz) was used, unless otherwise noted, to obtain
`I3C NMR spectra (16). Selected spectra were obtained
`
`with a Bruker AMX 300 spectrometer operating at 7.05 T
`(75 MHz for 13C). Spectra were obtained with 16K time
`domain points and a 10 KHz (4.7 T) or 15 KHz (7.05 T)
`spectral width. Aqueous FA-albumin complexes included
`100 pl of D 2 0 for a lock and a shim signal. The NMR
`sample tubes supported an insert containing tetramethyl-
`silane (TMS) in CDCl,. All chemical shifts were mea-
`sured with respect to this external reference. The uncer-
`tainty of all chemical shift values reported is k0.05 ppm,
`based on multiple measurements of samples with similar
`compositions prepared and analyzed at different times.
`Each NMR experiment reported in this study was re-
`peated one to three times. Data shown are representative
`data. The temperature dependence of the reference signal
`was estimated to be <0.1 ppm in the temperature range
`investigated. Temperature was controlled ( k1-Z0C) with
`a Bruker B-VT-1000 variable temperature unit. The inter-
`nal temperature of selected samples was measured as fol-
`lows. After the sample had equilibrated in the magnet for
`- 10 min at a fixed temperature that was regulated by the
`variable control unit, the sample was removed (time
`zero). A thermocouple was immediately inserted in the
`sample and several temperature values were recorded at
`15-sec intervals. The temperature at time zero (cor-
`responding to the true sample temperature) was obtained
`by extrapolation of the linear relationship of temperature
`versus time. A temperature calibration curve was con-
`structed from measurements at several temperatures.
`Spin lattice relaxation time was measured for selected
`samples by the fast inversion-recovery method (24). Pulse
`intervals for obtaining standard spectra were generally
`
`chosen for optimal signal to noise ratios ( - 1 x TI) rather
`
`than for equilibrium intensities (5 x Ti). Deconvolution
`of spectra was performed with NMRl (New Methods Re-
`search, Inc., East Syracuse, NY).
`
`CD spectroscopy
`Near-UV C D spectra were recorded using a Cary 61
`CD spectropolarimeter (Varian, Palo Alto, CA). Three
`continuous spectra per sample were recorded in the wave-
`length range of 320-260 nm. Data points were analyzed
`at 1-nm intervals; trough depth measurements were made
`with respect to the baseline value. Molar ellipticity values
`[e], in units of deg.cm2/dmol, were calculated by the
`standard equation for a path length of 0.02 cm (25).
`For temperature-dependent C D spectra, the sample
`temperature was maintained by circulating ethylene
`glycol-water through the lamp compartment by means of
`a thermostated refrigerator-heater bath (NESLAB, Ports-
`mouth, NH). Temperature was measured to within 0.1OC
`by means of a copper-constantan thermocouple posi-
`tioned in contact with the CD cell. With each tempera-
`ture change C D spectra were recorded after a short time
`interval (- 15 min) to allow the sample to equilibrate at
`the desired temperature setting.
`
`Kenyon and Hamilton
`
`Interactions of octanoic and decanoic acids with albumin
`
`459
`
`MPI EXHIBIT 1032 PAGE 2
`
`MPI EXHIBIT 1032 PAGE 2
`
`
`
`RESULTS
`NMR studies at 35°C
`13C NMR spectra for mixtures of I3C carboxyl-
`enriched O C T or DEC with HSA were obtained as a
`function of mole ratio of FA to protein at T = 34.5OC and
`pH = 7.40 + 0.15. The MCFA carboxyl carbon gave a
`single peak at all mole ratios; the intensity of this peak in-
`creased relative to that of protein peaks (e.g., the broad
`carbonyl at -170-180 ppm) with increasing FA to HSA
`mole ratio. Under similar experimental conditions, FA
`with a chain length of 212 carbons in the presence of
`HSA or BSA give rise to multiple narrow signals whose
`individual intensities increase with increasing mole ratio
`(18, 26). For O C T and DEC the chemical shift of the sin-
`gle carboxyl peak showed a dependence on mole ratio
`(Fig. 1A). For OCT/HSA mixtures, the chemical shift in-
`creased steadily with increasing mole ratio of FAIHSA,
`from 181.80 ppm (1:l OCT/HSA) to 182.83 ppm (15:l
`OCT/HSA). The value of the chemical shift approached
`but did not reach that of unbound OCT (184.32 ppm; see
`below). In contrast, the chemical shift of DEC/HSA mix-
`tures showed only a small dependence on mole ratio ex-
`cept at very high ratios (Fig. 1A). Between a 3:l and 11:l
`mole ratio of DEC/HSA, there was a small (0.28 ppm)
`linear increase in chemical shift. Above 11:l DECIHSA,
`a progressive shift to higher ppm was seen.
`
`The dependence of the FA carboxyl chemical shift on
`protein concentration was examined at 34.5OC and pH =
`7.4 by diluting samples with a fixed mole ratio of OCT or
`DEC to albumin (3:l and 1O:l) from 90-100 mg protein/
`ml to the lowest concentration feasible for NMR studies
`(10-20 mg/ml). In all cases a single, narrow resonance
`from the FA carboxyl carbon was observed. The FA chem-
`ical shift for the OCT/HSA system was more dependent
`on protein concentration than that of the DEC/HSA sys-
`tem (Fig. 1B). The OCT carboxyl peak shifted downfield
`(to higher ppm) with decreasing HSA concentration, from
`182.05 ppm (92 mg/ml) to 182.50 ppm (23 mg/ml) for a
`3:l OCT/HSA complex and from 182.30 ppm (98 mg/ml)
`to 183.15 ppm (12 mg/ml) for a 10:l OCT/HSA complex.
`The chemical shift for the 3:l DEC/HSA complex re-
`mained constant at 182.20 ppm for all HSA concentra-
`tions, whereas the 1O:l DEC/HSA mixture showed a
`slight increase in chemical shift (from 182.30 ppm to
`182.45 ppm) at the lowest concentration (12 mg/ml).
`NMR spectra obtained as a function of pH can provide
`important information about binding interactions from
`the ionization behavior of FA in the presence of protein
`(16, 19). Therefore, 13C NMR spectra were obtained at
`34.5OC as a function of pH for several mole ratios of
`OCT/albumin (1.5:1, 3:1, and 5:l). These spectra showed
`a single resonance under all conditions. Fig. 2 compares
`the titration behavior of O C T in the presence of HSA (3:l
`
`c, 3 3 183.0
`
`; t
`
`6
`
`d
`
`A
`
`1's 2b h '
`
`116
`
`0 10 20 30 40 50 60 70 80 90 100
`[HSAl (mr/ml)
`
`114
`1;
`110
`Mole Ratio
`Fig. 1. A: Plot of FA carboxyl '3C chemical shift versus mole ratio at T = 34.5'C and pH = 7.4 + 0.2. The mole ratio represents the stoichiometric
`amount of FA and albumin in the mixture. Filled circles represent DEC/HSA mixtures; unfilled triangles correspond to OCT/HSA mixtures. All sam-
`ples had a protein concentration of 93 mg/ml. The inset shows a Scatchard plot of the mole ratio data for a three-term fit. B: Plot of FA carboxyl
`I3C chemical shift versus HSA concentration at T = 34.5% and pH = 7.4 + 0.2. Open and closed circles represent 3:l and 1O:l OCT/HSA mixtures,
`respectively. Closed and open squares correspond to 3:l and 1O:l DEC/HSA mixtures, respectively. Conditions for obtaining NMR spectra were as
`follows: spectral accumulations ranged from 250 to 1000 for DEC/HSA and 6800 to 10,000 for OCT/HSA over 16,384 time domain points with a
`pulse interval of 2.0 sec. Chemical shift values were measured with respect to an external reference (tetramethylsilane) in units of ppm (parts per
`million).
`
`460
`
`Journal of Lipid Research Volume 35, 1994
`
`MPI EXHIBIT 1032 PAGE 3
`
`MPI EXHIBIT 1032 PAGE 3
`
`
`
`184
`
`..........
`
`182
`
`h 183
`a a
`v s
`-z 181
`al 8 179
`
`4 d
`
`180
`
`178
`
`177
`
`1
`3
`
`I
`4
`
`I
`5
`
`I
`6
`
`I
`7
`
`I
`8
`
`PH
`Fig. 2. Plot of OCT carboxyl 13C chemical shift versus pH at
`T = 34.5OC and mole ratios of 3:l (filled square) and 5:l (filled triangle)
`OCT/HSA. The protein concentration of all samples was 92-93 mg/ml.
`The solid line is the theoretically calculated Henderson-Hasselbach
`curve for a 3:l OCT/HSA mixture. The dashed curve is the '3C NMR
`titration of aqueous octanoic acid at 1.6 mM and 40% (20).
`
`and 5:l mole ratios) with that of O C T in the absence of
`protein. The plot of chemical shift versus pH closely fol-
`lows the sigmoidal Henderson-Hasselbach behavior that
`is characteristic of an acid-base titration (27). The appar-
`ent pKa value for OCTIHSA complexes was the same as
`the pKa of aqueous OCT (pKa = 4.6). However, the titra-
`tion curve for OCT with HSA is shifted to lower ppm
`compared to the curve for unbound OCT, a shift that
`reflects partitioning of MCFA into a less hydrophilic en-
`vironment than water (Le., interactions of OCT with the
`protein). pH-dependent spectra (not shown) were also ob-
`tained at 34.5OC for DEC/HSA complexes at 1.5:l and 3:l
`mole ratios. For the 1.5:l mole ratio, the chemical shift
`was monitored between pH 8.1 (182.16 ppm) and pH 6.1
`(182.01 ppm), below which the carboxyl signal became too
`broad to measure the chemical shift. For the 3:l mole
`ratio, it was possible to measure the chemical shift at lower
`pH values. The chemical shift decreased from 182.26 ppm
`at pH 8.4 to 181.21 ppm at pH 4.7. These data for the 3:l
`DEC/HSA complex suggest a titration of DEC similar to
`that for OCT in the presence of HSA (Fig. 2).
`
`NMR studies at low temperature
`In the above NMR studies at 35OC, the FA carboxyl
`group appeared to experience a single "binding" environ-
`ment under various experimental conditions: different
`FA/HSA mole ratios, pH values, and protein concentra-
`tions. These results could mean that binding sites for
`MCFA are not structurally heterogeneous (unlike the case
`for long-chain FA) or that the NMR experiments did not
`detect heterogeneous sites. The latter case could occur if
`
`FA in different binding sites exchange rapidly to produce
`an apparent "single" site, which has been shown to occur
`with binding of MCFA to BSA (21). As it was possible to
`detect multiple binding sites for MCFA on BSA at lower
`temperatures (21), a similar strategy was applied to HSA
`complexes with OCT and with DEC at fixed mole ratios
`
`of MCFA/HSA (1.5:1, 3:1, 5:l) and at pH 7.5 * 0.1. Fig. 3
`
`illustrates 13C NMR spectra (FA carboxyl region) for a 3:l
`mole ratio complex of OCT/HSA at five temperatures. At
`the higher temperatures (41.5OC and 34.5OC), a single
`narrow peak (181.97 ppm and 182.01 ppm, respectively)
`was observed from the W-enriched MCFA. As the tem-
`perature was decreased, the carboxyl resonance first
`broadened (Fig. 3C) and then separated into two major
`signals (Fig. 3D). At 6.5OC (Fig. 3E) the spectrum ex-
`hibited one signal at higher ppm and the other at lower
`ppm relative to the single signal at 35OC, characteristic of
`exchange (see below). A spectrum at -6' obtained at
`75 MHz (not shown) did not show improved resolution
`relative to the spectrum at 50 MHz (Fig. 3E).
`Temperature-dependent spectra for a 3:l DEC/HSA
`
`4 4
`11a2-01
`13 HzII
`
`41.5OC
`
`34.5%
`
`B
`
`182.17
`
`28
`
`C
`
`D
`
`16.5"C
`
`1 15°C
`
`6.5"C
`
`160
`
`170
`
`PPM
`Fig. 3. The carboxyl and carbonyl regions of 13C NMR spectra for a
`3:l mole ratio of 90% [l-13C)OCT/HSA complex at various temperature
`values: A) T = 41.5OC after 1000 accumulations; B) T = 34.5OC after
`1000 accumulations; C) T = 16.5OC after 4000 accumulations; D)
`T = 11.5OC after 8000 accumulations; and E) T = 6.5OC after 1186 ac-
`cumulations. All spectra were recorded with a pulse interval of 2.0 sec
`and processed with a line broadening 4.0 Hz. The sample pH was
`7.4 + 0.1 and the protein concentration was 93 mg/ml.
`
`Kenyon and Hamilton
`
`Interactions of octanoic and decanoic acids with albumin
`
`461
`
`MPI EXHIBIT 1032 PAGE 4
`
`MPI EXHIBIT 1032 PAGE 4
`
`
`
`are shown in Fig. 4A. For this mole ratio, as well as the
`two other mole ratios investigated (1.5 and 5.0), a single
`carboxyl resonance was observed at T>2l0C (Fig. 4A,
`top spectrum). As the temperature was decreased (e.g.,
`20.5OC), the chemical shift of the predominant peak
`shifted downfield slightly and a second small signal was
`seen upfield from this signal (Fig. 4A, middle spectrum).
`At 6.5OC the major signal was at 182.30 ppm and a
`broader, less defined signal was seen between 181.64 ppm
`and 181.15 ppm (Fig. 4A, bottom spectrum). A low tem-
`perature spectrum at higher field (75 MHz) showed evi-
`dence of a third signal at 181.96 ppm between the two
`separated peaks (Fig. 4A, right side). The carboxyl spec-
`tral region for 1 mole DEC/mole HSA at T = 6.5OC is
`shown in Fig. 4B (top spectrum). The spectrum of FA-free
`HSA (Fig. 4B, bottom spectrum) contains signals around
`
`181.09 ppm from amino acid carboxyl groups that are
`shifted away from the intense, broad envelope of carbonyl/
`carboxyl resonances, similar to the case for BSA (16). Sub-
`traction of the protein component from the spectrum for
`1:l DEC/HSA revealed a narrow peak centered at
`182.28 ppm (Fig. 4B).
`Temperature-dependent spectra in the same tempera-
`ture range as for OCT and for DEC were obtained for the
`12-carbon fatty acids (dodecanoic or lauric). These spec-
`tra gave three resonances (181.8, 182.0, and 182.3) for a 3:l
`mole ratio of FA/HSA at high temperature, as shown pre-
`viously (26). There was little effect of temperature in the
`carboxyl spectrum. Therefore, the temperature-dependent
`changes described above are specific for MCFA and are not
`due to protein aggregation or other general temperature-
`dependent structural changes.
`
`182.18
`
`I
`
`181.16
`
`182.30,
`
`81.1 5
`
`B
`
`182.28,
`
`3:l DECIHSA
`T=34.5'C
`
`3:l DEC/HSA
`
`3:l DEC/HSA
`T=6.5'C
`
`1:l DECiHSA
`T=6.5'C
`
`0:1 DEC/HSA
`
`w
`
`1
`
`I kI3
`
`'\
`
`i?,
`
`182.28 il
`
`1:1-0:1
`
`180
`PPM
`
`170
`
`.
`
`.
`,
`170
`
`'
`
`.
`
`.
`
`'
`
`
`
`I
`
`'
`
`.
`
`.
`
`,
`
`,
`180
`
`~
`
`l
`PPM
`Fig. 4. The carboxyl and carbonyl regions of 13C NMR spectra for A) a 3:l mole ratio of 90% [l-"C]DEC/HSA
`mixture at various temperatures: T = 34.5OC, 1200 scans (top spectrum); T = 20.5OC, 2000 scans (middle spec-
`trum); and T = 6.5OC, 3003 scans (bottom spectrum) and for B) a 1:1 mole ratio of90% [l-'jCC]DEC/HSA sample
`at 6.5OC, recorded after 4000 scans (top spectrum); a spectrum of FA-free HSA at 6.5OC (bottom spectrum); and
`the difference of the 1:l and 0:l mole ratio of DEC/HSA (right). A spectrum (4000 scans) of a 3:l DEC/HSA sample
`at 6.5OC recorded at a higher field (7.05 T) and the corresponding deconvolution spectrum are shown in panel A
`(right column). The deconvolution shows, in addition to the narrow signals at 182.27 ppm and 181.13 ppm, a broader
`signal at 181.96 ppm, representing a third binding environment for DEC, and a weaker broad signal from protein
`carboxylates underlying the FA carboxyl signal at 181.13 ppm (see panel B, bottom spectrum). All other NMR and
`sample conditions were the same as those stated in Figs. 1 and 3.
`
`462
`
`Journal of Lipid Research Volume 35, 1994
`
`MPI EXHIBIT 1032 PAGE 5
`
`MPI EXHIBIT 1032 PAGE 5
`
`
`
`Additional NMR studies of OCT binding to HSA were
`conducted at low temperature ( T = 6.5OC) where multiple
`binding environments were detected. Because the spectrum
`for 3:l OCT/HSA showed (at least) two binding environ-
`ments, it was of interest to examine different mole ratios
`of O C T to HSA in order to determine whether the sites
`fill sequentially or simultaneously. Fig. 5 shows spectra
`(carboxyl region) for 0, 1, 2, and 3 moles of OCT/mole of
`HSA. At the lowest mole ratio of O C T studied (1:l OCT/
`HSA), a strong carboxyl signal was seen at 181.83 ppm,
`
`along with weaker signal(s) at - 181.16 ppm (Fig. 5B) that
`
`are coincident, or nearly so, with the signal from the pro-
`tein amino acid carboxylates (Fig. 4B, bottom spectrum;
`Fig. 5A). The difference spectrum of B-A (Fig. 5E) re-
`vealed a single peak at 181.83 ppm which represents the
`primary binding site for O C T (see below). This chemical
`shift is the same as that for a 1:l mole ratio of O C T to
`HSA at 35°C (Fig. 1). (The temperature insensitivity of
`this chemical shift was confirmed in a separate experi-
`ment in which the same sample (1:l OCT/HSA) was ana-
`lyzed at 35OC and 5OC.) With 2 moles of OCT the car-
`boxyl spectrum at 6.5OC showed a peak at 181.83 ppm
`with a shoulder at 182.14 ppm, representing a new en-
`vironment (Fig. 5C). The difference spectrum of C-B,
`which reflects the signal intensity correlated with the
`addition of a second mole of OCT, showed a single peak
`at 182.14 ppm (Fig. 5F). Addition of a third mole of O C T
`to the OCT/HSA mixture yielded a spectrum (Fig. 5D;
`see also Fig. 3E) with two resonances (181.84 ppm and
`182.30 ppm). The difference spectrum, representing the
`
`addition of a third mole of OCT (Fig. 5G), showed a
`single peak at 182.30 ppm. No additional environments
`were detected at higher mole ratios of OCT to HSA (spec-
`tra not shown). The above results show three major en-
`vironments represented by signals at 181.8 ppm, 182.1
`ppm, and 182.3 ppm and suggest sequential filling of
`different binding sites whose affinity for OCT differs by
`2 5-fold (see discussion).
`Using low temperature as a probe of FA-protein inter-
`actions, pH-dependent data were obtained to assess struc-
`tural features of individual binding sites for MCFA on
`HSA. Spectra of a 3:l and 5:l mole ratio of OCT/HSA
`showed multiple binding sites at various pH values, in
`contrast to the titration data at 35OC (Fig. 2). Low tem-
`perature spectra for a 3:l mole ratio DEC/HSA sample
`revealed two resonances (-182.3 ppm and -181.2 ppm,
`as in Fig. 4A) that were both independent of pH (between
`5.0 and 8.5).
`Fig. 6 shows the NMR titration plot for 3:l OCT/HSA
`at low temperature. As there was a signal at 181.8 ppm for
`all pH values, it was assumed that this represented a peak
`with an invariant chemical shift. The peak representing
`the third binding site for OCT (182.3 ppm at pH 7.4)
`showed a change in chemical shift with pH. The peak for
`the second mole (182.1 ppm at pH 7.4) was submerged be-
`tween the other two peaks and apparently did not shift
`with pH. These titration data suggest stronger ionic inter-
`actions between amino acids and O C T in the first (181.8
`ppm) and second binding sites (182.1 ppm) than in the
`third site.
`
`D l
`
`l
`
`182.30 l8IB4
`
`
`
`3: 1
`
`A
`
`I
`
`.
`
`'
`
`.
`
`.
`
`.
`
`,
`180
`
`.
`
`.
`
`.
`PPM
`
`.
`,
`170
`
` ,
`
`.
`
`.
`
`
`
`PPM
`
`Fig. 5. The carboxyl and carbonyl regions of 13C NMR spectra for 90% [1-13C]OCT/HSA complexes at low temperature ( T = 6.5OC), at a fixed
`sample pH of 7.4 + 0.1 and at various mole ratios (A-D) with corresponding difference spectra (E-G). The spectra were recorded at a mole ratio
`of OCT/HSA of A) 0:l (FA-free HSA), 3000 scans; B) 1:1, 3000 scans; C) 2:1, 2000 scans; and D) 3:1, 1186 scans. The difference spectra are displayed
`as follows: E) 1:l-O:l, F) 2:1-l:l, and G ) 3:l-2:l OCT/HSA. All other NMR and sample conditions were equivalent to those stated in Figs. 1 and 3.
`
`Kenyon and Hamilton Interactions of octanoic and decanoic acids with albumin
`
`463
`
`MPI EXHIBIT 1032 PAGE 6
`
`MPI EXHIBIT 1032 PAGE 6
`
`
`
`1
`
`182.5
`
`182.0
`
`181.5
`
`181.0
`
`180.5
`
`5.5
`
`7.5
`
`8.5
`
`6.5
`PH
`Fig. 6. Plot of the carboxyl chemical shift versus pH of the resonances
`represent the nontitrating peak at - 181.8 ppm and the filled squares cor-
`for O C T in a 3:l OCT/HSA complex at T = 6.5OC. The filled circles
`
`respond to the pH-dependent resonance. The unfilled symbol signifies
`overlapping chemical shift data points.
`
`NMR studies of aqueous MCFA
`The CMCs of K' octanoate and K' decanoate (400 mM
`and 100 mM, respectively; ref. 28) are high compared to
`the concentrations used in this study. Decanoic acid can
`form acid soap complexes at pH 7.4 at concentrations
`lower than 100 mM, but '3C NMR signals from this
`phase are broad (29). Therefore, the relevant chemical
`shifts for comparison with data for these FA in the pres-
`ence of HSA are those of the aqueous monomers. The
`chemical shift for OCT at a concentration equivalent to
`that in the 3:l OCT/HSA mixture (6.5 mM OCT) was
`184.32 ppm at 34.5"C (pH 7.4); at 6.5OC (pH 7.2) the sig-
`nal shifted to 184.51 ppm. The pH dependence of the
`chemical shift of the carboxyl peak of aqueous OCT has
`been reported (19). Aqueous DEC at a concentration
`(6.5 mM) equivalent to a 3:l DEC/HSA complex at
`pH 7.4 and 34.5"C gave a value of 184.29 ppm, nearly
`identical to the shift for aqueous OCT under the same ex-
`perimental conditions.
`
`Near-UV circular dichroism (CD) studies
`A previous study using fluorescence-energy transfer
`measurements at 25OC found a very small change
`(<0.1 nm) in the distance between Trp-214 of HSA and
`bound bilirubin with the addition of up to 4 moles of
`OCT or DEC and a much larger change in this distance
`(0.2 nm) with long chain FA (30). Thus, to examine the
`influences of temperature and binding of MCFA on the
`tertiary structure of HSA, near-UV CD spectra were
`recorded for FA-free HSA (94 mg/ml; pH 7.4) as a func-
`tion of temperature (42OC, 36OC, 25"C, 15OC, 1l0C, and
`7OC). No significant change in molar ellipticity as a func-
`tion of temperature was seen for the HSA sample without
`added FA (data not shown). Further, near-UV CD spec-
`tra were recorded for a 3:l OCT/HSA complex (pH 7.4)
`at protein concentrations of 94 mg/ml, 47 mg/ml, 24 mg/
`
`464
`
`Journal of Lipid Research Volume 35, 1994
`
`ml, and 12 mg/ml and at temperatures of 42OC, 36"C,
`25"C, 15"C, 11"C, and 7OC. The molar ellipticity of HSA
`was not changed by the presence of OCT (3:l OCT/HSA)
`and this FA/HSA complex was also unaffected by temper-
`ature. A very small decrease in molar ellipticity at some
`wavelengths was observed upon decreasing the HSA con-
`centration from 94 to 24 mg/ml for the 3:l OCT/HSA
`mixture. Both results suggest no large conformational
`changes in the protein or aggregation within the tempera-
`ture range examined.
`
`DISCUSSION
`
`Exchange of MCFA between bound and unbound
`pools
`All NMR spectra of MCFA/HSA mixtures exhibited
`one or more peaks from the '%-enriched carboxyl carbon
`of the MCFA. These peaks occurred at chemical shifts
`upfield (lower ppm) from that for the aqueous MCFA in
`the absence of HSA and therefore reflect MCFA binding
`interactions with the protein. OCT/HSA complexes at
`35% (Fig. 1A) showed a single carboxyl signal whose
`chemical shift increased continuously as a function of
`mole ratio of OCT/HSA. This result is best explained by
`assuming rapid equilibration of bound and unbound
`OCT. NMR theory predicts a single, narrow peak at a
`chemical shift between the values for the two pools that
`are in fast exchange (31). The chemical shift measured for
`the OCT carboxyl signal in spectra of OCT/HSA mix-
`tures would then reflect the weighted average of bound
`and unbound OCT. Taking the chemical shift of 184.3 ppm
`to represent unbound OCT and 181.8 ppm to represent
`bound OCT (see below), the chemical shift of 182.85 ppm
`at 15 mol OCT/mol HSA indicates 40% bound (or 60%
`unbound) OCT. The data in Fig. 1A show that the frac-
`tion of unbound OCT increases continuously with in-
`creasing mole ratio; in contrast, the fraction of unbound
`DEC does not become detectable until mole ratios > 12:l.
`A quantitative analysis of OCT binding to HSA was
`made by calculating the ratio of bound/unbound OCT
`from the chemical shift data in Fig. 1A and plotting these
`results in a Scatchard form (Fig. lA, inset). The curve
`that best fits the data gave three distinct classes of binding
`sites with nl = 2.0 and K1 = 2.15 x io4 M P ; n2 = 5.1
`and K2 = 4.3 x lo2 M-1; and n3 = 6.6 and K3 = 78 M I
`.
`Previous studies have yielded variable predictions about
`the highest affinity sites on HSA for OCT. For example,
`Scatchard analysis has suggested nl = 4.2 and K1 =
`6.5 x lo3 M-' (32) and nl = 0.9 and K1 = 2.5 x lo5 M-'
`with the next class of sites having n2 = 10 and K2 =
`2.7 x lo3 M-' (10). A recent study of binding of OCT to
`HSA by equilibrium dialysis gave a single high affinity
`binding site with K1 = 1.6 x lo6 M-' (9). An alternative
`treatment of binding data, the stepwise equilibrium
`
`
`
`MPI EXHIBIT 1032 PAGE 7
`
`MPI EXHIBIT 1032 PAGE 7
`
`
`
`model, suggests that binding affinities decrease continu-
`ously as a fimction of added FA, Le., the binding of the
`first mole of OCT occurs with K1 = 3.4 x lo4 M-' and
`the binding of each consecutive mole occurs with approxi-
`mately a 3-fold lower affinity (33). Our results fall in the
`range of previous estimates, suggesting that our interpre-
`tation of the NMR chemical shift changes is valid. How-
`ever, our analysis is limited to high ratios of OCT/HSA
`and cannot be used to validate or invalidate previous
`predictions of absolute affinities. NMR data that reflect
`the filling of the binding sites for O C T give unique infor-
`mation about relative affinities of the highest affinity bind-
`ing sites (see below).
`Dilution experiments provided further evidence for fast
`exchange between bound and unbound pools. Upon dilu-
`tion of samples having fixed mole ratios of OCT/HSA, the
`O C T chemical shift moved to higher ppm, reflecting a
`higher ratio of unbound/bound OCT. Fig. 1B shows that
`for 10 OCT/HSA the fraction of unbound OCT was 30%
`for 49 mg/ml HSA and 54% for 12 mg/ml HSA. At a
`3 mole ratio of OCT/HSA the fraction of unbound O C T
`was -16% for 46 mg/ml HSA and -28% for 23 mg/ml
`HSA. At lower OCT/HSA mole ratios, the fraction of un-
`bound O C T will be significantly lower because the affini-
`ties of the first and second moles of O C T are much higher
`than that of subsequent moles (see be