Initial Photochemistry of Bilirubin Probed by Femtosecond

Oct 11, 2007 - 120 fs in the multiexponential fluorescence decay, being only visible in the bilirubin molecule, is interpreted as exciton localization...
1 downloads 0 Views 197KB Size
J. Phys. Chem. B 2007, 111, 11997-12003

11997

Initial Photochemistry of Bilirubin Probed by Femtosecond Spectroscopy Burkhard Zietz*,† and Tomas Gillbro Department of Chemistry, Biophysical Chemistry, Umeå UniVersity, SE-90187 Umeå, Sweden ReceiVed: May 4, 2007; In Final Form: August 8, 2007

Bilirubin is a breakdown product from heme catabolism, and reduced excretion of bilirubin can lead to jaundice. Phototherapy is the most common treatment for neonatal jaundice, a condition frequently encountered in newborn infants. Knowledge of the photochemistry of bilirubin, which is dominated by (ultra)fast components, is necessary for the profound understanding of the processes in phototherapy. Here, we report results from femtosecond fluorescence upconversion measurements on bilirubin and half-bilirubin model compounds, as well as pump-probe absorption measurements on bilirubin. A fast component of ca. 120 fs in the multiexponential fluorescence decay, being only visible in the bilirubin molecule, is interpreted as exciton localization within the molecular halves. The slower components of several hundreds of femtoseconds and a few picoseconds, occurring in bilirubin and the half-bilirubin model, are interpreted as relaxation to a (twisted) intermediate, which decays further with ca. 15 ps to the ground state.

Introduction Bilirubin ((Z,Z)-bilirubin IXR, BR), a yellow-orange waterinsoluble tetrapyrrole, is constantly formed in humans (and other mammals). Catabolism of heme starts with ring opening by heme oxygenase and removal of the central iron atom, and the biliverdin thus formed is then reduced to bilirubin.1,2 Previously seen as just a waste product, evidence has emerged that it also acts as a potent antioxidant and is now seen as a major physiologic antioxidant cytoprotectant.3-5 Due to its water insolubility, direct excretion is impossible. Instead, glucuronidation occurs in the liver, and the product (bilirubin diglucuronidate) is excreted into bile. In healthy persons, formation and excretion are balanced and happen unnoticed. Damage to the hepatic system can, however, lead to a situation where the formation of soluble BR diglucuronidate (often called conjugated bilirubin) is decreased, and a buildup of bilirubin in blood serum can occur. A similar situation is seen after birth, where the higher rate of red cell breakdown is not compensated by the initially low liver activity. This condition, neonatal jaundice, is very common, but in most cases abates within a few days. Higher concentrations of BR have to be avoided though, as this can lead to bilirubin encephalopathy. The resulting brain damage can be permanent, even lethal. Several methods are available to minimize BR levels in jaundiced patients; by far the most widespread is phototherapy. Having been applied routinely for over 40 years, it is now a standard treatment for a great number of infants in most hospitals. The structure of bilirubin consists of two nearly identical dipyrrinone subunits, bridged by a methylene groups that allows flexible rotation. The crystal structure, published in the late 1970s6,7 and a refined study with higher resolution shortly after8 have given an explanation for the nonpolar properties of BR, unsuspected from the linear representation of the structure. A ridge-tile formation is taken, which is stabilized by six strong intramolecular hydrogen bonds to the opposite dipyrrinone’s * Corresponding author. E-mail: [email protected]. Tel.: +33 (0)3 88 10 71 93. Fax: +33 (0)3 88 10 72 45. † Present address: IPCMS-GONLO, 23 rue du Loess, PB 43, F-67034 Strasbourg Cedex 2, France.

propionic acid group, see Figure 1. This renders the molecule’s polar groups locked, and explains the insolubility in water and methanol, and that it easily dissolves in, e.g., CHCl3. On the physiological side, these properties are important, as the produced BR needs to be transported to the liver, which is accomplished by binding to albumin. Strong, noncovalent binding to human serum albumin (HSA, binding constant ca. 1 × 107 M-1)9 is also of importance in buffering higher temporary concentrations of BR, delaying the onset of encephalopathy. The two dominant processes involved in the mechanism of phototherapy are formation of the configurational (4Z,15E)isomer10 and formation of a structural isomer, called lumirubin, involving internal cyclization.11 The former occurs with medium efficiency (φZE ≈ 0.1),12 whereas the latter has a low quantum yield (φLR ≈ 0.001).13 Both processes are believed to contribute to a reduction of BR in blood plasma as the products have a considerably improved water solubility, enabling direct excretion. No consensus has been reached on the exact contributions of these processes, with some authors regarding the formation of lumirubin as the most efficient excretion channel,14 while others stress the importance of the (E)-isomer in reducing BR levels.10,15 Quantum yields of processes besides the ones of isomerization are very low; the fluorescence is extremely weak in solvents at room temperature (φfl < 0.0002 in CHCl3) and only slightly stronger for the BR-HSA complex (φfl ∼ 0.003).13 Phosphorescence is also extremely weak (φisc < 0.01),16 indicating that a very efficient, nonradiative de-excitation channel must exist. Fluorescence can be considerably increased, however, by a rigid environment (φfl ) 0.71 in PMMA at room temperature),13 low temperature (φfl ) 0.92 for BR/HSA at 77K),17 and chemically limiting the twisting motion (see below). A special and noteworthy property of bilirubin is its wavelength-dependent photochemistry.12,18-21 A reduction of the quantum yield of (E)-isomer formation by a factor of 2 is seen when going from 458 to 514 nm in the case of HSA-bound BR, and a smaller, but similar, trend is seen in alkaline methanol solution. This has been ascribed to excitonic effects,12 i.e., the dipole moments of the two conjugated subunits are close enough

10.1021/jp073421c CCC: $37.00 © 2007 American Chemical Society Published on Web 10/11/2007

11998 J. Phys. Chem. B, Vol. 111, No. 41, 2007

Figure 1. Structure of (4Z,15Z)-bilirubin IXR (BR); broken lines indicate hydrogen bonds.

Zietz and Gillbro described recently38 could well explain the fast deactivation observed for the BR system. The geometry seen not only shows a twisted double, but also single bond, not unlike the hulatwist mechanism.39,40 The (E-Z) (or cis-trans) isomerization process is of importance in a wide range of biological processes. The best known example is vision in higher animals, where rhodopsin undergoes a cis-trans isomerization. Bacteriorhodopsin is closely related to rhodopsin and functions as a simple photosynthetic system in halobacteria. Phytochrome, a light receptor in plants,41-43 provides another biologically important example and is very closely associated with bilirubin. Biliverdin, phytochrome’s chromophore, is a fully conjugated tetrapyrrole that differs from bilirubin by just one double bond. As in phototherapy, the isomerization around an exocyclic double bond is triggered by light. Bilirubin’s photochemistry, which forms the basis for phototherapy, poses a range of unanswered questions, not least of medical importance. Combined with the biological importance of the configurational isomerization process in biology, this gives incentive to study bilirubin’s complex photochemistry. Here, we report results from an ultrafast time-resolved study of bilirubin in solution and a comparison with the half-bilirubin model compounds. Experimental Section

Figure 2. Fluorescence upconversion kinetics of BR in different solvents.

to interact, with two different halves having slightly different excitation energies. Photoisomerization is supposed to occur on the half with lower energy, a property that can be affected by binding to protein. This can also explain the regiospecific isomerization to the (4Z,15E)-isomer, which is greatly preferred over the (4E,15Z)-isomer when bound to HSA. As a consequence of the above, the optimal choice of wavelength used for phototherapy is not straightforward. Absorption is maximized at ca. 450 nm, but efficiency of lumirubin formation is higher at 500 nm and above. Also, absorption of blue light is higher in human skin, allowing deeper penetration of green light.22 An optimal wavelength has yet to be determined unequivocally.23 In order to elucidate bilirubin’s chemistry, complicated by exciton coupling and hydrogen bonding, model compounds as xanthobilirub(in)ic acid, XBR, and its methyl ester, MeXBR, have been synthesized24 and used in several studies.25-27 Due to the structural similarities with BR, even these dipyrrinone compounds tend to form hydrogen bonds by dimerization in nonpolar solvents.28-30 To mimic BR’s intramolecular hydrogen bonding in single dipyrrinone units, on the other hand, compounds with flexible alkane chains have been synthesized and proven to engage in internal hydrogen bonding.31-33 The twisting motion of bilirubin has been identified as a deactivation channel in several studies15,17,34,35 and confirmed by synthesis of N,N′-bridged dipyrrinones.36 These show strong fluorescence in the case of a short methano bridge (φfl ) 0.81), rapidly decreasing with increasing bridge length (φfl ) 0.26 for ethano and φfl ) 0.0012 for propano bridge), i.e., decreasing fluorescence with increasing rotational flexibility around the exocyclic bonds.37 The existence of a conical intersection between the ground and first excited state for a half-bilirubin model that we have

Bilirubin (for biochemistry, Merck) was used as received, and solutions were prepared fresh under dim light conditions. Xanthobilirubic acid (XBR) and its methyl ester (MeXBR) were a kind gift from D. A. Lightner. Steady-state absorption spectra were recorded on a Beckman DU-70 spectrometer. Solvents used (chloroform, dichloromethane, tetrahydrofuran (THF), dimethylformamide (DMF), dimethyl sulfoxide (DMSO), triethylamine (Et3N)) were of pro analysi or spectroscopic grade. The experiments were complicated by the very limited solubility of BR in most common solvents, as well as the sensitivity of BR to photooxidation. A flowing or stirred sample (alternatively a rotating cell) had to be used in all experiments. Absorption spectra were recorded before and after all experiments to check the status of the sample. Fluorescence Upconversion Measurements. The fluorescence upconversion spectrometer used in this work has been described in detail earlier;44 a short description is given here. Near-infrared femtosecond pulses of ca. 60 fs from a modelocked, 82 MHz Ti:Sapphire laser (Spectra Physics Tsunami), pumped by an argon ion laser, were frequency doubled to produce light with wavelengths between 420 and 480 nm. After passing through a dichroic filter, the fundamental wavelength beam was sent to a delay line, and the frequency-doubled pulses were used to excite the sample in a rotating cell. Detuning of the doubling crystal decreased frequency-doubled pulse energies to typically 0.1 nJ, reducing bleaching of the sample. The fluorescence was collected and focused on a rotating BBO crystal, together with the delayed near-infrared gate beam. The upconverted light was dispersed by a prism, sent through an iris and a UV filter, and focused onto the slits of a monochromator (ISA H10 UV, f/3.5). A Peltier-element-cooled, low-noise photon-counting photomultiplier (Hamamatsu R4220P) detected and a gated photon counter (Stanford Research Systems SR400) counted the signal. For isotropic decay measurements, the polarization of the pump beam was set to the magic angle with a variable waveplate (Berek’s polarization compensator, New Focus). The sample (200-1500 µL) was held in a rotating sample cell of 1-2 mm path length. In several solvents,

Photochemistry of Bilirubin

J. Phys. Chem. B, Vol. 111, No. 41, 2007 11999

TABLE 1: Fluorescence Decay Lifetimes for BR in a Range of Solvents Obtained by Fitting Exponential Decays to Single-Wavelength Upconversion Kinetics; λex ) 454 nm, λem ) 526 nm CHCl3 CH2Cl2 THF DMF DMSO 2% Et3N/CHCl3 5% Et3N/CH2Cl2 1% NH4OH/MeOH

τ1/ps (amp1)

τ2/ps (amp2)

τ3/ps (amp3)

τ4/ps (amp4)

0.134 (51%) 0.110 (56%) 0.138 (58%)

0.58 (37%) 0.386 (42%) 0.444 (34%) 0.599 (39%) 0.780 (39%) 0.411 (22%) 0.383 (25%) 0.297 (19%)

2.2 (9%) 5.16 (2%) 2.67 (6%) 2.48 (54%) 2.74 (52%) 3.69 (15%) 1.66 (60%) 0.953 (57%)

9.4 (3%)

0.033 (52%)

bleaching occurred, and the solution was changed regularly. In a typical measurement, 5-15 scans were accumulated and averaged. Comparison of the first and last scans showed only slight deviations. Transient Absorption Measurements. Time-resolved absorption measurements, using the pump-probe absorption technique, were made with a laser setup comprising a regenerative amplifier (Spitfire; Positive Light) and an optical parametric amplifier. A Ti:Sapphire laser (Spectra Physics Tsunami), pumped by a continuous wave frequency-doubled diode-pumped Nd:YAG laser (Millennia V, Spectra Physics) served as a seed for the regenerative amplifier, which was pumped by a Nd: YLF laser (Merlin; Positive Light). The 800 nm output, ca. 80 fs pulses with an energy of ca. 0.2 mJ per pulse at a repetition rate of 5 kHz, was split into two parts, 5% being used for white light continuum (probe) generation in a sapphire plate and 95% being used as input for the optical parametric amplifier (OPA800, Spectra Physics). The fourth harmonic of the idler at around 450 nm was used as excitation beam and, in order to compensate for pulse stretching in the optical components, in some experiments compressed through an SF14 prism pair. After being chopped, the beam was sent into an optical delay line and focused onto the sample. The white light continuum was split into signal and reference beam, focused by spherical mirrors onto the sample (with signal beam overlapping the pump beam), and passed onto the slits of a computer-controlled monochromator ((TRIAX 190; Jobin Yvon Horiba). The intensities of the signals of pump, signal, and reference were recorded by integrating silicon photodiodes (FD-4; EKSPLA, Lithuania). The sample of ca. 10 mL was pumped continuously through a 1 mm flow cell (except for THF as solvent). Pumping at lower rates led to a reduction in signal, and a high rate was chosen that assured the signal to be unaffected by small variations in pump speed. All absorption measurements were carried out with a magic angle polarization between the pump and probe beam. The data analysis for upconversion and pump-probe measurements, carried out with the Spectra Solve program, allowed fitting of multiexponential decay curves with variable background, convoluted with a Gauss response function with adjustable time shift. Decay-associated spectra (DAS) were calculating by a global analysis of a series of kinetics measured at different wavelengths. The lifetimes were variable, but locked for all wavelengths, and the amplitudes, backgrounds, and time shifts were allowed to float freely. The preexponential factors thus determined were normalized to the steady-state fluorescence intensities taken from an independent experiment.

12.2 (2%) 12.9 (7%) 10.9 (9%) 19.2 (11%) 10.0 (15%) 2.74 (23%)

τ5/ps (amp5)

19.9 (0.5%)

ponents. An even faster decay is seen for CH2Cl2 and somewhat slower in DMF. However, a dominant subpicosecond component is observed in all solvents investigated. Lifetimes obtained in a range of different solvents are summarized in Table 1. All decays required multiexponential components to obtain good fits to the data. In the less polar solvents examined, like CHCl3, CH2Cl2, and THF, a fast component of 110-140 fs is obtained, with amplitudes between 40% and 60%. Two to three further components, with lifetimes from several hundreds of femtoseconds to a few picoseconds, were also obtained by fitting the data. In more polar solvents like DMF, the fast (ca. 120 fs) component disappears. An overall slower decay is seen, with major components of 600 fs (40% amplitude) and 2.5 ps (50%) for the DMF case. It is interesting to compare fluorescence upconversion results of bilirubin in dichloromethane and dichloromethane with 5% (v/v) Et3N added. The addition of Et3N is not expected to change much the general solvent properties (like dielectric constant and polarizability). Bilirubin, possessing two propionic acid groups, can however be assumed to be deprotonatedsthe pKa values for mesobilirubin XIIIR have been measured as 4.2 and 4.9.46 As can be seen in Figure 3, a striking difference in the kinetics is obtained, which is also reflected in the lifetimes (see Table 1). The fast component in CH2Cl2 (110 fs) is lost, and the major component with 60% amplitude in Et3N/CH2Cl2 is ca. 1.7 ps. Also in CHCl3 a major difference is observed in the kinetics upon addition of 2% Et3N (data not shown). The decay becomes slower, but at the same time a very fast component of ca. 30 fs with major amplitude is obtained. Although being on the edge of what the upconversion system is able to resolve, the high data quality allows very accurate fitting, and changes to the response function during the fitting procedure cannot substitute this very fast component. Decay-Associated Spectra. The dependence of the fluorescence on the emission wavelength was investigated by measur-

Results Fluorescence Upconversion. Single-WaVelength Kinetics and SolVent Effects. Fluorescence upconversion kinetics of bilirubin in three different solvents are shown in Figure 2. As we have reported earlier for BR in CHCl3,45 a very rapid decay of the initial excitation is visible, dominated by subpicosecond com-

Figure 3. Fluorescence upconversion kinetics of BR in CH2Cl2 with and without Et3N.

12000 J. Phys. Chem. B, Vol. 111, No. 41, 2007

Figure 4. Decay-associated spectrum for bilirubin in CHCl3; the steady-state fluorescence is also shown for comparison.

Figure 5. Decay-associated spectrum for bilirubin in DMSO.

ing a series of decays at different wavelengths. The results were fitted in a global fitting procedure and are plotted as DAS for chloroform in Figure 4, and for DMSO in Figure 5. For the former, the fastest component of ca. 120 fs shows a blue-shifted spectrum with a maximum at ca. 500 nm, compared to the second fastest component (ca. 600 fs), which is centered at around 525 nm. The longer lifetimes (3 and 30 ps) have small amplitudes; slightly higher amplitudes are discernible at 700 nm for the slowest one. The results are of similar nature for other less polar solvents like CH2Cl2 and THF. A different picture is obtained in the case of DMSO. The fastest component (ca. 80 fs) shows high amplitudes at 525 nm and negative amplitudes at longer wavelengths (580 and ca. 610 nm). The zero-crossing occurs at around 550 nm. For the 0.67 ps component, high amplitudes are seen at the blue side of the spectrum and low amplitudes at longer wavelengths. The 2 and 5.5 ps components both show a maximum at around 550 nm, and the slowest component has small amplitudes, slightly higher at the red side of the spectrum. Excitation WaVelength Dependence. In the exciton picture, the dipole moments of monomers at medium close range (i.e., no van der Waals contact) interact, and a splitting in the absorption spectrum is visible. Instead of two independent absorptions at the same energy, a splitting occurs, with some higher energy components and one lower energy component. The intensities of these are governed by the orientation of the dipole vectors, and this can lead to a bathochromic or hypso-

Zietz and Gillbro

Figure 6. Fluorescence upconversion decay of BR in CH2Cl2 with excitation at different wavelengths.

Figure 7. Fluorescence upconversion decay of BR and XBR in chloroform. Symbols represent the measured data; the lines are the best fits to the data. The inset shows the initial decay at an enlarged time scale.

chromic shift of the absorption or result in a splitting or broadening. Irradiation into the blue and red side of the absorption spectrum is therefore mainly leading to excitation of the higher and lower excitonic state, respectively. We have probed the time-resolved fluorescence with excitation into the blue and red wings of BR’s absorption, in addition to the absorption maximum. Kinetic traces are shown in Figure 6. Even with the excellent signal-to-noise ratio of the data, the curves are virtually identical. Half-Bilirubin Model XBR. Results from fluorescence decay measurements of the half-bilirubin model XBR in chloroform in comparison with results for bilirubin are shown in Figure 7. The decay of XBR, although considerably slower, is still fast, and most of the intensity has decayed by 10 ps. The results for MeXBR, the methyl ester of XBR, are practically identical to XBR in this solvent (data not shown). Lifetimes obtained from fits to the data are given in Table 2. Even here, several lifetime components are needed to achieve good agreement with the data. In contrast to bilirubin, no fast lifetimes (100-150 fs) were obtained in less polar solvents. The dominant component, ca. 3 ps, has an amplitude of about 75% in CH2Cl2 and CHCl3. It is interesting to note that although a great difference between the kinetics of BR and XBR is observed in CHCl3, the decays become very similar in alkaline methanol, which also is reflected in similar lifetimes. This is mainly due to the lack of the ca.

Photochemistry of Bilirubin

J. Phys. Chem. B, Vol. 111, No. 41, 2007 12001

TABLE 2: Fluorescence Decay Lifetimes for XBR in a Range of Solvents Obtained by Fitting Exponential Decays to Single-Wavelength Upconversion Kinetics; λex ) 421 nm, λem ) 524 nm CHCl3 DMSO 1% NH4OH/MeOH MeOH

τ1 (amp1)

τ2 (amp2)

τ3 (amp3)

0.712 (27%) 0.327 (19%) 0.254 (42%) 0.200 (28%)

2.58 (71%) 6.67 (56%) 2.04 (40%) 1.13 (32%)

20.3 (2%) 15.9 (25%) 7.63 (18%) 2.45 (40%)

120 fs decay component of large amplitude observed for BR in, e.g., CHCl3. Decay-associated spectra of MeXBR in CHCl3 show very little dependence of amplitudes on wavelengths in the range of 485-580 nm; a slightly higher amplitude of the longest lifetime component is seen at long wavelength (data not shown). Transient Absorption. Whereas the fluorescence upconversion method is able to provide very accurate data on the decay of initial excitation, little information is gained about the later processes of the system, such as nonemitting, “dark” intermediate states. Therefore, transient absorption spectroscopy was used to study the changes induced by light after the initial relaxation. Changes of absorbance at different time delays for BR in DMF are shown in Figure 8. At short wavelengths (