Supporting Information

Transcription

1 Supporting Information Wiley-VCH Weinheim, Germany

2 Metal-Organic-Frameworks as Efficient Materials for Drug Delivery Patricia Horcajada, Christian Serre*, María Vallet-Regí, Muriel Sebban, Francis Taulelle and Gérard Férey* Materials and methods Synthesis. Ordered chromium carboxylates, denoted as MIL-100 and MIL-101, were hydrothermally synthesized as previously reported [1,2]. MIL-100 is hydrothermally synthesised [1] from a suspension of metallic chromium (Aldrich, 99%), hydrofluoric acid (HF, Prolabo 40%), 1,3,5 benzene tricarboxylic acid H 3 BTC (Aldrich, 99%) and H 2 O with a molar ratio of 1:1:0.67:265 using a Teflon Liner and a metallic Paar Bomb; the whole is then 96 hours at 220 C with a heating ramp of 12 hours. MIL-101 was also synthesized hydrothermally [2] from a suspension containing chromium (III) nitrate Cr(NO 3 ) 3 9H 2 O (Aldrich, 97%), terephthalic acid HO 2 C-(C 6 H 4 )-CO 2 H (Alfa, 97%), hydrofluoric acid (HF, Prolabo 40%) and H 2 O in the molar ratio 1:1.5:1:280 and further heated 10 hours at 220 C with a heating ramp of 1 hours, still using a Teflon Liner and a metallic Paar Bomb. After a cooling ramp down to room temperature of 3 hours, a significant amount of recrystallized terephthalic acid is present. To eliminate most of the carboxylic acid, the mixture is filtered first using a large pore fritted glass filter (n 2); the water and the MIL-101 powder passes through the filter while the free acid stays inside the glass filter. Then, the free terephthalic acid is eliminated and the MIL-101 powder is separated from the solution using a small pores (n 5) paper filter and büchner. The solid is then redispersed in ethanol (1g of dried MIL101 for 50 ml of EtOH) and using a Teflon liner and a Paar Bomb, heated overnight at 100 C to extract BDC molecules trapped in the pores of MIL-101. After cooling, the product is again filtered and dried overnight at 100 C under air in an oven. Incorporation of Ibuprofen. The Ibuprofen loading was performed introducing under stirring during 24 h 100 mg of the dehydrated powder materials (dried overnight at 100 C in an oven) in a 10 ml solution of hexane containing variable amounts of Ibuprofen. The best ratio porous solid to Ibuprofen (wt) were 1:2 and 1:4 for MIL-100 and MIL-101, respectively. The adsorbed amount of Ibuprofen into the porous solids was estimated by UV-Vis spectroscopy, thermogravimetric analysis (TGA), X-Ray Fluorescence (XRF) and elemental analysis.

3 Preparation of the pellets. 200 mg of the Ibuprofen-containing porous solids were compacted by uniaxial (2.75 MPa) and isostatic pressure (3 MPa) to obtain disks (13 mm in diameter) used for the delivery assays. The delivery assays were carried out soaking the samples in a simulated body fluid (SBF) [3] at 37ºC, maintaining the ratio ml SBF/mg adsorbed Ibu equal to 1 and stirring continuously to avoid a delivery by external diffusion constraints [4]. The Ibuprofen concentration in the solution was measured by High Performance Liquid Chromatography (HPLC). The samples were characterized by XRD, 13 C and 1 H-NMR, TGA, elemental analysis, XRF and N 2 adsorption. Determination of Ibuprofen content Optimisation of Ibuprofen adsorption Several parameters have been tested for this optimization of the drug loading in MIL-100 and MIL-101. Here are summarized some of the resulting estimated drug capacities. The dehydrated materials adsorb more Ibuprofen than the original ones which contain 30-50% weight of water. In the same way, Ibuprofen insertion increases with the Ibuprofen/material ratio. In fact, the optimum Ibu/material ratio (wt) was 2:1 and 4:1 for MIL-100 and MIL-101, respectively. Although these values are limited by the Ibuprofen solubility in hexane, the very fast and high adsorption of Ibuprofen for MIL-101, allowed the use of a higher Ibuprofen amounts in the starting solution for MIL-101. Finally, the influence of the contact time and repeated impregnations on the final drug loading was not significant. Table S1. Estimated Ibuprofen content according to several impregnation parameters. Immersion time Ibu/material ratio Dehydratation (days) 1:1 1:2 1:3 no yes 1 3 Consecutive impregnations MIL MIL Thermogravimetric analyses The thermogravimetric analyses (TGA) were carried out between ºC under air atmosphere (100 ml min -1 ), using a Perkin Elmer Diamond TGA/DTA. The heating rate was 5ºC min -1 for the starting materials and was 1ºC min -1 for the Ibuprofen-containing samples; in the latter case, an isotherm step at 250ºC for 6 h was applied. This plateau was used to give a better estimation of the Ibuprofen content knowing that the adsorbed Ibuprofen will

4 leave the structure before the degradation of these solids, due to the departure of the carboxylic groups from the framework, which occurs within the C range. In Figure S1, three weight losses are observed within the , and C. The first weight loss is due to the departure of water, the second weight loss departure to the Ibuprofen and the last one to destruction of the porous solid. 100 MIL-100 Material Material-Ibu % Weight loss MIL Temperature (ºC) Figure S1. TGA under air of MIL-100 and MIL-101 materials, before and after the Ibuprofen adsorption. Estimating the Ibuprofen content based on TGA is difficult for several reasons. First, the desorption temperature of drug depends on the strength of interaction between the Ibuprofen moieties and the framework. MIL-100 starts loosing Ibuprofen at 200ºC while in MIL-101 the Ibuprofen departure begins at 150ºC. Secondly, it is hard to distinguish the loss of free Ibuprofen from the incorporated drug in weak interaction within the pores. Finally, estimating the loss of an organic (drug) in a hybrid solid is hard since it cannot be excluded that some of the Ibuprofen will leave the structure at temperatures very close to the destruction of the framework. Despite these considerations, Ibuprofen content can be estimated by TGA (Table S2), by considering for MIL-101 the loss between 150 C and 275 C (even if some residual Ibuprofen could leave the pores at higher temperatures). This gives an approximate 1.01 g/g drug content. The lower temperature for the desorption of Ibuprofen from MIL- 101, compared with MIL-100, is due to the larger windows (~12-16 Å) of MIL-101 which favors a quick release of the Ibuprofen molecules upon heating.

5 In MIL-100 the aperture of the windows is much smaller (~5-8.6 Å) than in MIL-101, which can explain a lower Ibuprofen content and a higher release temperature. Thus, the content is based on losses between 200 and 300 C; the initial loss observed at lower temperatures is probably due to some water and free Ibuprofen moieties. A content of 0.31 g/g is estimated in MIL-100. In both cases, the drug content estimated by TGA is underestimated since it is likely that some Ibuprofen molecules in strong interaction with the framework are leaving the solid when the degradation of the framework occurs. UV-Vis Spectroscopy The Ibuprofen content was indirectly estimated using UV-Vis spectroscopy by subtraction of the Ibuprofen amount in the hexane solution and after drug adsorption at 273 nm (see Table S2 below). To avoid errors, due to the hexane volatility, samples were weight before and after the impregnation assays, replacing the volume of lost solvent. The Ibuprofen content is overestimated by UV-Vis spectroscopy, due to a precipitation of the drug in the concentrated solutions. In fact, the calculated Ibuprofen content increases when the ratio Ibuprofen/material increases, as consequence of a lower stability of the solution. Elemental analysis and X-Ray Fluorescence Elemental analysis of C, N and H was carried out in a Leco CNS-200 analyzer. The Cr and F contents were determined by X-Ray Fluorescence using a Philips PANalytical AXIOS spectrometer (RhK α λ= Å). To perform the XRF analysis, samples were compacted as 0.2 g disks by uniaxial (2.75 MPa) and isostatic pressure (3 MPa) and dehydrated at 100ºC for 24 h. The Ibuprofen amount was determined by elemental analysis and XRF is shown in Table S3. Taking into account the Ibuprofen content obtained by the different techniques, elemental analysis/xrf data are the most reliable. Table S2. Estimated Ibuprofen contents in MIL-100-IBU and MIL-101-IBU. g IBU/g dried material TGA UV-Vis Elemental analysis/xrf MIL100-Ibu MIL101-Ibu Taking into account the effective volume of an Ibuprofen molecule (~200 Å 3 ) and considering the volume of each cage for MIL-101 (12700 Å 3 and Å 3 ) present in a 1:2 ratio, this gives about 45% Ibu is located in the bigger

6 cages while 55% is hosted into the smaller cages. Besides, one should expect a much higher Ibuprofen content for MIL-100 considering the pore volume ratio MIL-100/MIL-101=1.2 / 2.0=0.6. The observed Ibu ratio in both materials is equal to 0.35/1.4 = 0.25!! This is due to the fact that Ibuprofen molecules are probably present only in the larger cages of MIL-100 since the smaller cages possess windows with an aperture of 4.8 x 5.8 Å, ruling out the introduction of Ibuprofen; thus, as the ratio between the volume of the large cage in MIL-100 (Volume of the cages = 8200 Å 3 and Å 3 ) are 44% of the total volume, this should give a 0.6*0.44=0.26 ratio of Ibu content in MIL- 100 compared with MIL-101. This is in agreement with the observed results. X-Ray Powder Diffraction The XRD patterns were collected in a conventional high resolution (θ-2θ) Philips X Pert MDP diffractometer (λ Cu Kα??K α 2 )?from 1.5 to 10º (2θ) using a step size of 0.03º and 4 s per step in continuous mode (Figure S5). MIL-100 and MIL-101 present a cubic structure (Fd3m) [1,5] which maintain this crystalline structure after the Ibuprofen adsorption despite a slight decrease in crystallinity. This reversible phenomenon is also observed when MIL-101 and MIL-100 are impregnated with water and the initial crystallinity is fully recovered when the samples are dried. This is probably due to the presence of a high degree of disorder for the guest molecules within the giant cages. Thus, the loss of crystallinity is probably due to the filling of the pores by Ibuprofen molecules. Moreover, the application of the compacting pressure to form the pellets does not affect the integrity of the crystalline structure, as proved by XRD before and after compression. However, compacted MIL-101-Ibu showed a contraction of the porous structure, due to a relative flexibility of the framework under the pressure conditions. After the delivery assays, XRD confirms that the integrity of the framework is kept for both phases.

7 MIL-100 powder MIL100-Ibu powder MIL100-Ibu compacted MIL100-ID compacted MIL-101 powder MIL101-Ibu powder MIL101-Ibu compacted MIL101-ID compacted Figure S2. XRD patterns of the MIL-100 and MIL-101 materials. º2θ Nitrogen adsorption Surface area of the materials was determined by N 2 adsorption using BJH method in a Micromeritics ASAP 2010 porosimeter. MIL-100 and MIL-101 samples were previously dehydrated under vacuum overnight at 200ºC and the Ibuprofen-containing samples at 65ºC to avoid the degradation of drug.

8 Figure S3 shows the nitrogen isotherms for MIL-100 and MIL-101 materials and the Ibuprofen-containing samples before and after the delivery, whereas the Langmuir surface areas and pore volumes are collected in Table S MIL MIL V (cm 3 g -1 ) MIL100-ID MIL100-Ibu V (cm 3 g -1 ) MIL101-ID MIL101-Ibu P/P P/P 0 Figure S3. Nitrogen isotherms of MIL-100 and MIL-101, and the Ibuprofen-containing samples before and after delivery. A slight secondary uptake indicates the presence of micro and mesopores in the starting materials. After Ibuprofen adsorption, there is almost no residual porosity which indicates that the drug fills completely and/or blocks the pores of the material leaving approximately no accessible pore volume for nitrogen. Table S3. Surface area and pore volume of MIL-100 and MIL-101 samples. MIL-100 MIL-101 S LANGMUIR (m 2 g -1 ) V P (cm 3 g -1 ) Ibu ID Ibu ID Finally, after the drug delivery assays, MIL100-ID and MIL101-ID showed a partial recovery of pore volume and surface (Table S3). This fact could be explained by the presence of ions (coming from the SBF) into the cages, which could block the nitrogen access to the cavities or increase the framework density of the porous solids. MIL101-ID recovers only half of its surface area and volume, while MIL100-ID only 20 % of it. This is in agreement with a additional steric hindrance at the smaller windows of MIL-100. XRF data (Table S4) show a higher retention of ions PO 3-4, Mg 2+, Ca 2+ and Na + in MIL100-ID than in MIL101-ID, being in total agreement with this explanation. An other possibility is the grafting on the Lewis acid sites (i.e. chromium atoms) of the organic base, aminoethane, used to fix the ph of the SBF; this latter would also blocks some pores and decrease the accessibility of nitrogen. The presence of % N is in agreement with a possible grafting of the amino groups.

9 Table S4. XRF data of the MIL-100 and MIL-101 after the delivery assays. % Elements Na Mg P S Cl K Ca N MIL100-ID MIL101-ID Infrared Spectroscopy The infrared spectra were collected with a Nicolet Nexus spectrometer. The spectra of the starting porous materials show bands ν(arc-h) corresponding to aromatic groups at 3050 cm -1. and vibrational bands characteristics of the O-C-O- group around 1550 and 1430 cm -1 (Figure S4). The Ibuprofen spectrum shows ν(c-h) and ν(c=o) bands at 2900 and 1708 cm -1, respectively, corresponding to the carboxylic group and aliphatic groups of the drug. IBU n (C-H) MIL-101 n Ar(C-H) n (C=O) n Ar(C-H) n (C=O) Ar 1,4 MIL101-Ibu n (C-H) Ar (C=C) n (C=O) Transmitance n (C=O) MIL-100 n Ar(C-H) n (C=O) MIL100-Ibu Ar Ar 1,3,5 (C=C) n (C=O) n (C=O) n (C-H)

10 Figure S4. Infrared spectroscopy of MIL-100 and MIL-101 before and after Ibuprofen adsorption, together with the Ibuprofen spectrum. In the Ibuprofen-containing materials spectra can be observed the presence of ν(c-h) bands around 2900 cm -1, corresponding to the C-H groups from the methyl moieties, and an intensity increase of the ν(arc-h) bands, due to the presence of Ibuprofen aromatic groups. Moreover, the shift of the ν(c=o) band from 1704 cm -1 to 1714 and 1707 cm -1 in MIL100-Ibu and MIL101-Ibu, respectively, is characteristic of the Ibuprofen carboxylic groups within the impregnate solids. ndicates the presence of the Ibuprofen carboxylic group. Delivery assays: HPLC determination The delivered Ibuprofen concentration was determined using a RP-HPLC system (Reversed phase liquid chromatography) equipped with a Waters Alliance 2695 separations module (Waters, Milford, MA, USA), a variable-wavelength diode array detector Waters 2996 and controlled by Millenium 32 software. Zorbax Eclipse XDB-C18 reverse-phase column (5 µm, 4.6x150 mm), supplied for Waters, were employed. The mobile phase consisted of 90% solution (v/v) of acetonitrile in water. The flow rate was 1.5 ml min -1 and the column temperature was 50ºC. The effluent was monitored at 273 nm and the injection volume was 10 µl. Several Ibuprofen solutions at concentrations of 0, 0.25, 0.5 and 1.0 mg ml -1 in SBF were used as standards. The calibrated plot showed a good correlation coefficient > 0.99 (Figure S5). Besides, the σ 2 obtained from 10 consecutive injections of each concentration, was < Integrated area (mv) IBU (mg ml -1 ) Figure S5. Calibration plot of standard Ibuprofen by HPLC method.

11 Chromatogram of standard solutions presents two retention times, at 0.8 and 1.1 min. The peaks at 1.1 min were identified by UV spectra, with absorption maximum at 264 nm, in standard solutions as well in Ibuprofen release as Ibuprofen. However, MIL101-Ibu chromatogram (Figure S6) shows another retention peak at 1.4 min corresponding to the terephthalic acid, as it is verified by UV spectrum. It is interesting to note that the intensity of this peak increases until 8 h of delivery and is maintained at longer times. The released terephthalic acid amount was determined, achieving a maximum value < 0.6 mg g -1 of dried material. Besides, the concentration of delivered organic acid does not depend on the Ibuprofen presence, as it was verified by soaking of the starting MIL-101 material at the same conditions. Crystalline structure maintenance (XRD, Figure S2) confirms that the released organic acid did not come from the material degradation, but probably from a organic acid fraction trapped initially into the MIL-101 material AU UA MIL101-Ibu (48h) AU l (nm) l (nm) AU l (nm) SBF Ibuprofen Terephtalic acid Retention time (min) Figure S6. Chromatogram and UV-Vis spectra of MIL101-Ibu alter 48 h of delivery assay. On the other hand, MIL100-Ibu did not show any release of the trimesic acid (1,3,5-BTC). For MIL100 material, the first stage of Ibuprofen delivery (2 h) can be empirically adjusted to a zero order kinetic ([Ibu] = Kt) (Figure S7), drug concentration independent, with regression factor >0.99 and a value of the slope (K) of 22.9 ± 1.9. However, Ibuprofen delivery in the first 8 h from MIL101 is empirically adjusted to a Higuchi model ([Ibu] = Kt 1/2 ) (Figure S7), with a regression factor >0.99, pointing to an Ibuprofen delivery controlled by an internal diffusion process of drug through the channels. The slope value (K) is 13.0 ± 0.6.

12 MIL100 % IBU released MIL Time (h) Time (h 1/2 ) Figure S7. Fit-curves of MIL-100 ([Ibu] = Kt) and MIL-101 ([Ibu] = Kt 1/2 ) in the first hours of Ibuprofen delivery. Nuclear Magnetic Resonance Spectroscopy Experimental Part Solid state NMR spectra were performed on a spectrometer Bruker Avance 500 at 125 MHz for 13 C and MHz for 1 H using MAS 3.2mm and MAS 4mm probes. Samples of starting porous materials MIL101, pure Ibuprofen and Ibuprofen-containing materials MIL101-Ibu were put in 3.2mm diameter ZrO 2 rotors. MIL100 and MIL100-Ibu were put in 4mm diameter ZrO 2 rotors. Data were acquired using the Bruker software XWINNMR, and processed with WINNMR 1D. 1 H and 13 C spectra with magic angle spinning and proton decoupling (xix, 95kHz) were recorded for pure Ibuprofen ( 1 H- 13 C cross polarization experiment), the starting porous materials, and Ibuprofen-containing materials (one pulse experiment). 1 H and 13 C acquisition and processing parameters are reported in Tables 1 and 2. Adamantane was used as external secondary reference. 13 C signals assignment of ibuprofen was realized following the literature.

13 Table S5. Acquisition and processing parameters of 1 H NMR experiments 1 H Acquisition Parameters Ibu / MIL101 / MIL101-Ibu MIL100, MIL100-Ibu Bruker avance 500MHz / MAS Bruker avance 500MHz / MAS Spectrometer / Probe 3.2mm 4mm Temperature 25 C 25 C MAS 20kHz 12.5kHz Pulse program (PULPROG) one pulse (zg) one pulse (zg) 90 Pulse 1H (P1 / PL1) 4.49u / 4.66dB (Rf=55.7kHz) 2.5u / -3.07dB (Rf=100kHz) Scans (NS) 8 8 TD 4k / 32k 4k / 32k SWH 24 khz / 1 MHz 24 khz / 1 MHz Relaxation Delay (D1) 1 sec 1 s 01 (Hz) Referencing Processing Parameters No Zero Filling LB=0 Chemical Shift referencing SR= Hz (left signal of adamantane at 1.74 ppm). FT / Phase/ Baseline correction Chemical Shift referencing SR= Hz (left signal of adamantane at 1.74 ppm).

14 Table S6. Acquisition and processing parameters of 13 C NMR experiments 13 C Acquisition Parameters MIL101 / MIL101-Ibu MIL100, MIL100-Ibu Bruker avance 500MHz / MAS Bruker avance 500MHz / MAS Spectrometer / Probe 3.2mm 4mm Temperature 25 C 25 C MAS 20kHz 12.5kHz One pulse with proton decoupling One pulse with proton decoupling Pulse program (PULPROG) (HPDEC) (HPDEC) 90 Pulse 1H (P3 / PL2) 2.5u / -3.49dB (Rf=100kHz) 2.5u / -3.07dB (Rf=100kHz) 90 Pulse 13C (P1 / PL1) 6.17u / 4.25dB (Rf=40.5kHz) 2.5u / -5.1dB (Rf=100kHz) Scans (NS) TD 4k 4k SWH 35 khz 35 khz Relaxation Delay (D1) 5 sec 30 s 01P (ppm) 100 ppm 100ppm Xix 1H decoupling Pl12=1dB (RF=95kHz) / Pl12=1dB (RF=95kHz) / p31=145u p31=148u Referencing Chemical Shift referencing SR= Hz (left signal of adamantane at ppm). Chemical referencing SR= Hz (left signal of adamantane at ppm). Contact time (CP) (p15) Hartmann-Hahn Pure Crystalline Ibuprofen 4 msec pl22 (1H)=-2.08dB / pl1(13c)=5db Processing Parameters Zero Filling : Si=16k LB=0 FT / Phase/ Baseline correction Results Starting materials The pure crystalline Ibuprofen was first characterized. Its 1 H spectrum was obtained with a good signal to noise ratio. It displays three resonances typical of aliphatic CH 3, CH 2 and CH carbons in the range 0-4 ppm, signals of sp2 carbons in the aromatic ring in the range 6-10 ppm, and a signal in the range ppm characteristic of the proton of the carboxylic acid. Its 13 C CPMAS spectrum was obtained with a good signal to noise ratio and high resolution. It displays 13 C resonances typical of aliphatic CH 3, CH 2 and CH carbons in the range ppm, signals of sp2 carbons in the aromatic ring in the range ppm, and a signal at 183 ppm characteristic of the carbon of the carboxylic acid.

15 One pulse and 13 C CPMAS experiments were used to acquire 1 H and 13 C NMR spectra of starting porous materials. No signal except probe signal appears due to the paramagnetic effect of chromium atoms. Ibuprofen-containing materials MIL101-Ibu and MIL100-Ibu 1 H and 13 C NMR spectra (one pulse experiment with proton decoupling) recorded for Ibuprofen-containing materials MIL101-Ibu and MIL100-Ibu display broad signals corresponding to Ibuprofen in the material. The comparison of NMR spectra of pure crystalline Ibuprofen and Ibuprofen-containing materials MIL101-Ibu and MIL100-Ibu confirms the insertion of Ibuprofen in the porous materials MIL-101 and MIL H NMR The 1 H NMR spectra show the disappearance of the broad signal close to 14 ppm characteristic of the proton of the carboxylic acid function. In order to establish if this loss of signal is due to a proton exchange (with water for example) or to an Ibuprofen-containing materials MIL101-Ibu was analyzed before and after heating. Identical spectra were obtained, revealing the existence of interaction between this part of the Ibuprofen molecule and the porous materials. 1 H NMR spectra were also acquired with a spectral window of 1MHz for MIL101-Ibu and MIL100-Ibu (Figure S8). The spectrum of MIL100-Ibu displays more spinning side bands than the spectrum of MIL101-Ibu. So, chemical shift anisotropy due to dipolar coupling with the chromium electrons is more important in MIL100-Ibu, than in MIL101-Ibu. Interactions between Ibuprofen and MIL 101 are weaker than interactions between Ibuprofen and MIL 100. MIL100-ibu MIL101-ibu (ppm)

16 Figure S8. 1 H NMR spectra of MIL101-Ibu and MIL100-Ibu (spectral width: 1MHz) 13 C NMR 13 C NMR spectra of MIL100-Ibu and MIL101-Ibu (before and after heating) are obtained with a weak signal to noise ratio because of the paramagnetic effect of chromium atoms. The broadening of the signals is due not only to the paramagnetic effect of chromium atoms, but also to a distribution of chemical shifts. This distribution of chemical shifts can be related to a conformational distribution of Ibuprofen in the cavities of MIL 101 and 100, or/and to an increase of the Ibuprofen mobility. This effect is more important in the case of MIL100 and concerns essentially the carboxylic acid part of Ibuprofen.

17 References [1] G. Férey, C. Serre, C. Mellot-Graznieks, F. Millange, S. Surblé, J. Dutour, I. Margiolaki, Angew. Chem. Int. Ed., 2004, 43, [2] C. Serre, F. Millange, C. Thouvenot, M. Noguès, G. Marsolier, D. Louër, G. Férey, J. Am. Chem. Soc., 2002, 124, [3] T. Kokubo, H. Kushitani, S. Sakka, T. Kitsugi, T. Yamamuro, J. Biomed. Mater. Res., 1990, 24, [4] B. Muñoz., A. Rámila., J. Pérez-Pariente., I. Díaz., M. Vallet-Regí, Chem. Mater., 2003, 15, [5] G. Férey, C. Mellot-Draznieks, C. Serre, F. Millange, J. Dutour, S. Surblé, I. Margiolaki, Science, 2005, 309,