Development of a self-emulsifying formulation that reduces the food effect for torcetrapib.

Torcetrapib is a highly lipophilic (Clog P=7.45) and water insoluble cholesteryl ester transfer protein (CETP) inhibitor developed for the treatment of atherosclerosis. Self-emulsifying drug delivery system (SEDDS) formulations have been developed to reduce the food effect observed in early clinical trials using medium chain triglyceride (MCT) softgels and to increase the dose per capsule. MCT/Triacetin/Polysorbate 80/Capmul MCM (20/30/20/30) (MTPC) increased fasted exposure and thus reduced the food effect from 5- to 3-fold in dogs at a dose of 90 mg. Self-emulsifying formulations also accelerated absorption and generally decreased variability. Use of the lipophilic, GRAS cosolvent triacetin allowed a 2-fold increase in the dose per capsule. For the three formulations evaluated in dogs that showed significant differences in absorption, emulsion droplet size did not appear to play a significant role. Absorption did increase with Cremophor RH40 content, and at 50% Cremophor RH40 there was no food effect (at 30 mg). High polar surfactant content also resulted in poor dose proportionality, while there was good dose proportionality for MTPC. Studies of in vitro lipolysis are being conducted to aid in understanding the in vitro/in vivo relationships for torcetrapib SEDDS absorption.

Solution-phase, parallel synthesis and pharmacological evaluation of acylguanidine derivatives as potential sodium channel blockers.

Solution-phase synthesis of various acylguanidine derivatives and the evaluation of a small library of compounds as potential sodium channel blockers are described.

Identification and characterization of a potential ischemia-selective N-methyl-D-aspartate (NMDA) receptor ion-channel blocker, CNS 5788.

The identification and characterization of a potentially ischemia-selective and orally-active sulfoxide based NMDA ion-channel blocker showing good neuroprotective activity, (R)-(+)-N-(2-chloro-5-methylthiophenyl)-N'-(3-methylsulfinylphenyl)-N'-methylguanidine (CNS 5788), is described.

Synthesis and binding characteristics of N-(1-naphthyl)-N'-(3-[(125)I]-iodophenyl)-N'-methylguanidine ([(125)I]-CNS 1261): a potential SPECT agent for imaging NMDA receptor activation.

N-(1-Naphthyl)-N'-(3-[(125)I]-iodophenyl)-N'-methylguanidine ([(125)I]-CNS 1261) was synthesized as a potential radioligand to image N-methyl-D-aspartate (NMDA) receptor activation. [(125)I]-CNS 1261 was prepared by radioiodination of N-(1-naphthyl)-N'-(3-tributylstannylphenyl)-N'-methylguanidine using Na(125)I and peracetic acid. [(125)I]-CNS 1261 uptake in vivo reflected NMDA receptor distribution in normal rat brain, whereas in ischemic rat brain, uptake was markedly increased in areas of NMDA receptor activation. Radiolabeled CNS 1261 appears to be a good candidate for further development as a single photon emission computed tomography tracer in the investigation of NMDA receptor activation in cerebral ischemia.

Synthesis and pharmacological evaluation of N,N'-diarylguanidines as potent sodium channel blockers and anticonvulsant agents.

Synthesis and structure-activity relationships (SAR) are described for a series of N,N'-diarylguanidines related to N-acenaphth-5-yl-N'-(4-methoxynaphth-1-yl)guanidine (3) as anticonvulsants through blockade of sodium channels. SAR studies on compound 3 led to several simpler diphenylguanidines with improved in vitro and in vivo activity. Compounds were screened for blockade of sodium channels in a veratridine-induced [14C]guanidinium influx assay (type IIA sodium channels) and for anticonvulsant activity in the audiogenic DBA/2 mouse model. Results indicated that N, N'-diphenylguanidines substituted with flexible and moderate size lipophilic groups were preferred over aryl and/or hydrophilic groups for biological activity. Among the compounds studied, n-butyl- and/or n-butoxy-containing guanidines showed superior biological activity. A possible relationship between in vitro and in vivo activity of this compound series and their measured/calculated lipophilicities was investigated. Compounds of this series showed only weak NMDA ion channel-blocking activity indicating that the anticonvulsant activity of these compounds is unlikely to be mediated by NMDA ion channels but, more likely, by acting at voltage-gated sodium channels.

Design, synthesis, and pharmacological evaluation of conformationally constrained analogues of N,N'-diaryl- and N-aryl-N-aralkylguanidines as potent inhibitors of neuronal Na+ channels.

In the present investigation, the rationale for the design, synthesis, and biological evaluation of potent inhibitors of neuronal Na+ channels is described. N,N'-diaryl- and N-aryl-N-aralkylguanidine templates were locked in conformations mimicking the permissible conformations of the flexible diarylguanidinium ion (AS+, AA+, SS+). The resulting set of constrained guanidines termed "lockamers" (cyclophane, quinazoline, aminopyrimidazolines, aminoimidazolines, azocino- and tetrahydroquinolinocarboximidamides) was examined for neuronal Na+ channel blockade properties. Inhibition of [14C]guanidinium ion influx in CHO cells expressing type IIA Na+ channels showed that the aminopyrimidazoline 9b and aminoimidazoline 9d, compounds proposed to lock the N,N'-diarylguanidinium in an SS+ conformation, were the most potent Na+ channel blockers with IC50's of 0.06 microM, a value 17 times lower than that of the parent flexible compound 18d. The rest of the restricted analogues with 4-p-alkyl substituents retained potency with IC50 values ranging between 0.46 and 2.9 microM. Evaluation in a synaptosomal 45Ca2+ influx assay showed that 9b did not exhibit high selectivity for neuronal Na+ vs Ca2+ channels. The retention of significant neuronal Na+ blockade in all types of semirigid conformers gives evidence for a multiple mode of binding in this class of compounds and can possibly be attributed to a poor structural specificity of the site(s) of action. Compound 9b was also found to be the most active compound in vivo based on the high level of inhibition of seizures exhibited in the DBA/2 mouse model. The pKa value of 9b indicates that 9b binds to the channel in its protonated form, and log D vs pH measurements suggest that ion-pair partitioning contributes to membrane transport. This compound stands out as an interesting lead for further development of neurotherapeutic agents.

Direct measurements of the dissociation-rate constant for inhibitor-enzyme complexes via the T1 rho and T2 (CPMG) methods.

The unimolecular dissociation rate constant, k-1, for the inhibitor-enzyme complex tubercidin-Escherichia coli purine nucleoside phosphorylase (PNPase) has been determined directly via two related 1H NMR methods for studying exchange-mediated transverse relaxation. One method involves measurements of the decay rate, 1/T1 rho, of spin-locked magnetization in the rotating frame as a function of the strength of the spin-locking field, omega SL. The second method involves measurements of the Carr-Purcell-Meiboom-Gill (CPMG) spin-echo decay rate, 1/TCPMG2, as a function of the repetition rate, 1/tcp, of the refocusing pulses. Expressions describing the dependence of TCPMG2 as a function of 1/tcp and k-1 have been previously derived with sufficient generality to include the two-site inhibitor-enzyme exchange case. Existing expressions for T1 rho as a function of kex and omega SL, however, had to be reformulated to take into account differences between Tb2 and Tb1 for the bound form of the inhibitor as well as offset corrections important at low values of omega SL. A new expression for exchange-mediated T1 rho has been derived to take these factors into account and is shown to provide a more accurate description of observed T1 rho data than previous models. Numerical analysis of relaxation rates, measured independently by either the rotating-frame or the spin-echo method for the H1' and H2 protons of tubercidin at different inhibitor:enzyme ratios, yields comparable values for k-1 of 2400 (+/- 350) and 900 (+/- 80) s-1 at 20 and 10 degrees C, respectively.(ABSTRACT TRUNCATED AT 250 WORDS)

Studies of inhibitor binding to Escherichia coli purine nucleoside phosphorylase using the transferred nuclear Overhauser effect and rotating-frame nuclear Overhauser enhancement.

NMR studies of the adenosine analog tubercidin have been carried out in the presence of Escherichia coli purine nucleoside phosphorylase (PNP) in order to characterize the conformation of the enzyme-complexed nucleoside. Although analysis of transferred NOE data at various enzyme/inhibitor ratios indicated a predominantly syn nucleoside conformation in the enzyme-complexed state, the results, particularly the 8(1') and 8(3') NOE interactions, were not quantitatively consistent with any single bound conformation. Dissociation rate constants for the tubercidin-PNP complex were determined based on analysis of chemical shift and line width data as a function of enzyme/inhibitor ratio, Carr-Purcell-Meiboom-Gill measurements of the transverse relaxation rate as a function of pulse rate, and T1 rho experiments as a function of the spin-lock field strength. Dissociation rate constants of 2100 s-1 at 20 degrees C and 1400 s-1 at 10 degrees C were determined using the latter two methods. These rates are sufficiently high to justify the validity of the transferred NOE method for an enzyme as large as PNP. The possible significance of spin diffusion was investigated by the use of the deuterated analog [2'-2H]tubercidin, for which many of the intraligand spin diffusion pathways are eliminated, and by performing a series of transferred ROE experiments. A comparison of data obtained using transferred NOE and ROE measurements provides a basis for separating direct and indirect relaxation pathways. Both approaches indicated that the relatively significant 8(3') NOE interaction was not dominated by spin diffusion. Furthermore, analysis of chemical shift and transverse relaxation data for the tubercidin H-2 resonance gave inconsistent results for the chemical shift of the bound species and was inconsistent with the assumption of a single, bound conformation. These results were interpreted in terms of a 2:1 ratio of a syn, 3'-exo:anti, 3'-endo geometry for bound tubercidin. Ligand competition experiments using 9-deazainosine show that all of the tubercidin TRNOE effects are reversed by addition of the second nucleoside, suggesting that the TRNOE data for tubercidin arise due to interactions at the active sites of PNP rather than as a consequence of nonspecific binding to the enzyme.

Mechanism of preconditioning. Ionic alterations.

The mechanism by which preconditioning (brief intermittent periods of ischemia and reflow) improves recovery of function and reduces enzyme release after a subsequent 30-minute period of ischemia was investigated in perfused rat hearts. Specifically, it was hypothesized that ischemia after preconditioning would result in a decreased production of H+ and therefore a smaller rise in [Na+]i and [Ca2+]i via Na(+)-H+ and Na(+)-Ca2+ exchange. To test this hypothesis we measured pHi, [Na+]i, [Ca2+]i, and cell high-energy phosphates during ischemia and reflow, and we correlated this with recovery of contractile function and release of creatine kinase during reflow. 31P nuclear magnetic resonance (NMR) was used to measure pHi and cell phosphates. [Na+]i was measured by 23Na NMR using the shift reagent thulium 1,4,7,10-tetraazacyclododecane-N,N,'N",N"'-tetramethylenephosph onate to distinguish intracellular from extracellular sodium. [Ca2+]i was measured by 19F NMR using hearts loaded with 1,2-bis(2-amino-5-fluorophenoxy)ethane-N,N,N',N'-tetraacetic acid, termed 5F-BAPTA. Basal time-averaged levels of pHi, [Na+]i, and [Ca2+]i were 7.07 +/- 0.08, 9.4 +/- 0.8 mM, and 715 +/- 31 nM, respectively. After 30 minutes of ischemia, in preconditioned hearts, pHi was 6.5 +/- 0.06, [Na+]i was 2.09 +/- 4.4 mM, [Ca2+]i was 2.1 +/- 0.4 microM, and ATP was negligible. In untreated hearts, after 30 minutes of ischemia, pHi was 6.3 +/- 0.08, [Na+]i was 26.7 +/- 3.8 mM, [Ca2+]i was 3.2 +/- 0.6 microM, and ATP was undetectable. During reperfusion after 30 minutes of ischemia, preconditioned hearts had significantly better recovery of contractile function than untreated hearts (71 +/- 9% versus 36 +/- 8% initial left ventricular developed pressure), and after 60 minutes of ischemia, preconditioned hearts had significantly less release of the intracellular enzyme creatine kinase (102 +/- 12 versus 164 +/- 17 IU/g dry wt). We also found that unpreconditioned hearts arrested with 16 mM MgCl2 (to inhibit calcium entry via calcium channels and Na(+)-Ca2+ exchange) before 30 minutes of ischemia recover function on reflow to the same extent as preconditioned hearts with or without magnesium arrest. Thus, preconditioning has no additional benefit in addition to magnesium arrest. In addition, in hearts that received 16 mM MgCl2 just before the 30-minute period of ischemia, preconditioning had no effect on the rise in [Ca2+]i during the 30-minute period of ischemia. These data support the hypothesis that preconditioning attenuates the increase in [Ca2+]i, [Na+]i, and [H+]i during ischemia, most likely because of reduced stimulation of Na(+)-H+ and Na(+)-Ca2+ exchange.(ABSTRACT TRUNCATED AT 400 WORDS)

Synthesis of potential dual-acting radiation sensitizer antineoplastic agents: 2,2-dimethylphosphoraziridines containing 2-nitroimidazoles or other electron-affinic moieties.

In view of the in vivo demonstrated radiation-potentiating activities of several previously studied 2,2-dimethylphosphoraziridines, six new compounds incorporating the bis(2,2-dimethyl-1-aziridinyl)phosphinyl moiety, together with an electron-affinic group such as 2-nitroimidazole or nitrobenzyl, have been synthesized and tested (1) in vitro for ability to increase the effect of X-irradiation under hypoxic conditions on V-79 Chinese hamster lung fibroblast cells, (2) in vivo for antitumor activity in the absence of radiation against P388 leukemia in mice, and (3) in a preliminary experiment with compound 10 only, in combination with whole-body gamma-radiation, using the P388 leukemia mouse model for in vivo radiation-potentiating activity. The chemical-alkylating activities and hydrolytic behavior of these compounds, as well as their antitumor activities without radiation, were found to be comparable to those of other 2,2-dimethylphosphoraziridines, while their in vitro radiosensitizing activities were at low concentrations generally comparable to that of misonidazole, with compound 8 showing superior activity. At higher concentrations, only compound 10 was sufficiently soluble and nontoxic to the cells for evaluation in this assay. Thus, the bis(2,2-dimethyl-1-aziridinyl) phosphinyl moiety does not seem to have contributed to the hypoxic radiosensitizing activities (only to the cytotoxicities) of the electron-affinic moieties in this in vitro assay. In comparison, the prototype 2,2-dimethylphosphoraziridine, ethyl [bis(2,2-dimethyl-1-aziridinyl) phosphinyl]carbamate (AB-132), showed at nontoxic doses no radiosensitizing activity in this assay, and at cytotoxic doses increased the cell-killing effect of each given dose of X-radiation additively under both hypoxic and oxic conditions. Conversely, only the 2,2-dimethylphosphoraziridine moiety appeared to participate in the moderate "therapeutic radiation-potentiating" activity indicated by compound 10 in the in vivo experiment using the P388 leukemia model (on day 1), as the misonidazole standard was inactive in this nonhypoxic system. Clearly, the mechanism of the in vivo observed radiation-potentiating effect of AB-132 and other 2,2-dimethylphosphoraziridines is different from that of the hypoxic radiosensitizers, but the possible synergism between the two biologically active moieties of the new compounds could not be demonstrated with the experimental models so far employed.

Uridine diphospho sugars and related hexose phosphates in the liver of hexosamine-treated rats: identification using 31P-[1H] two-dimensional NMR with HOHAHA relay.

The effects of administration of galactosamine (GalN) and glucosamine (GlcN) on the levels of UDP-sugars and hexose monophosphates in rat livers were studied by a variety of 31P NMR methods. The flux of metabolites in the liver was monitored by in vivo NMR and showed elevated levels of UDP-sugars, and even greater increases in resonances at 4.6 ppm for GlcN treatment and at 2.0 ppm for GalN treatment. The individual compounds corresponding to these changes were identified in PCA liver extracts by 31P-[1H] two-dimensional relay spectroscopy with a HOHAHA-type 1H spin-lock. This method of transferring proton magnetization allows for nearly all of the proton chemical shifts to be observed for the hexose moiety of a UDP-sugar present in a complex mixture. The UDP-sugars in the extracts from treated rats were predominantly UDP-hexosamines. Relay spectra were also used to determine that GalN-1-P was the major component (16.0 mumol/g of liver) of the GalN-treated liver, while both alpha and beta anomers of GlcNAc-6-P were readily identified as the major hexose monophosphates in the GlcN experiment. Spectra from the 1H dimension of relay experiments conducted on extracts were nearly superimposable on relay spectra obtained under the same conditions for mixtures of standard compounds of known structure. UDP-GlcN and UDP-GalN were not commercially available, but their presence was established in the extracts after GalN treatment by obtaining relay spectra for a mixture of the compounds produced in situ enzymatically, without purification.(ABSTRACT TRUNCATED AT 250 WORDS)

13C and 15N nuclear magnetic resonance evidence of the ionization state of substrates bound to bovine dihydrofolate reductase.

The state of protonation of substrates bound to mammalian dihydrofolate reductase (DHFR) has significance for the mechanism of catalysis. To investigate this, dihydrofolate and dihydropteroyl-pentaglutamate have been synthesized with 15N enrichment at N-2. 15N NMR studies have been performed on the binary complexes formed by bovine DHFR with these compounds and with [5-15N]dihydrobiopterin. The results indicate that there is no protonation at N-5 in the binary complexes, and this was confirmed by 13C NMR studies with folate and dihydrofolate synthesized with 13C enrichment at C-6. The chemical shift displacements produced by complex formation are in the same direction as those which result from deprotonation of the N-3/C-4-O "amide" group and are consistent with at least partial loss of the proton from N-3. This would be possible if, as crystallographic data indicate, there is interaction of N-3 and the 2-amino group of the bound ligands with the carboxylate of the active site glutamate residue (Glu30).

In vivo nuclear magnetic resonance studies of hepatic methoxyflurane metabolism. I. Verification and quantitation of methoxydifluoroacetate.

The elimination and metabolism of the fluorinated inhalation anesthetic methoxyflurane (2,2-dichloro-1,1-difluoroethyl methyl ether) in rats has been monitored using in vivo 19F nuclear magnetic resonance at 8.45 T. The elimination of methoxyflurane from rat liver as measured using a surface coil is a first order process when measured beginning 2-3 hr after the end of methoxyflurane anesthesia over a period of 12 hr. The rate constant for hepatic methoxyflurane elimination is dependent upon the duration of anesthesia, varying from 0.24 hr-1 for 15 min of anesthesia to 0.07 hr-1 for 1 hr of anesthesia. Methoxyflurane was shown to be metabolized in the liver to methoxydifluoroacetate using the surface coil method. No resonance for hepatic fluoride ion could be observed in vivo. Pure sodium methoxydifluoroacetate was synthesized in order to confirm the identity of the resonances in liver and urine. 19F NMR spectra of urine collected from anesthetized rats contain resonances for two methoxyflurane metabolites, methoxydifluoroacetate and inorganic fluoride. Studies with liver homogenates imply that fluoride is quickly cleared from the liver and eliminated from the body through the urine, explaining the inability to observe hepatic fluoride using a surface coil. The 19F NMR resonance for inorganic fluoride in urine was found to be broadened by interaction with metal ions, since the broadening could be eliminated by treatment with chelating resin.

In vivo nuclear magnetic resonance studies of hepatic methoxyflurane metabolism. II. A reevaluation of hepatic metabolic pathways.

Methoxyflurane (2,2-dichloro-1,1-difluoro-ethyl methyl ether) is believed to be metabolized via two convergent metabolic pathways. The relative flux through these two metabolic pathways has been investigated using a combination of in vivo surface coil NMR techniques and in vitro analyses of urinary metabolites. Analysis of the measured concentrations of inorganic fluoride, oxalate, and methoxydifluoroacetate in the urine of methoxyflurane-treated rats for 4 days after anesthesia indicates that the anesthetic is metabolized primarily via dechlorination to yield methoxydifluoroacetate. The methoxydifluoroacetate is largely excreted without further metabolism, although a small percentage of this metabolite is broken down to yield fluoride and oxalate, as determined by urine analysis of rats dosed with synthetic methoxydifluoroacetate. At early times after methoxyflurane exposure, the relative concentrations of methoxyflurane metabolites indicate that a significant fraction of the metabolic flux occurs via a different pathway, presumably demethylation, to yield dichloroacetate as an intermediate. Direct analysis of dichloroacetate in the urine using water-suppressed proton NMR indicates that the level of this metabolite is below the detection threshold of the method. Measurements made on the urine of rats dosed directly with dichloroacetate indicate that this compound is quickly metabolized, and dichloroacetate levels in urine are again found to be below the detection threshold. These results demonstrate the quantitative importance of the dechlorination pathway in the metabolism of methoxyflurane in rats.

Synthesis and properties of bis(2,2-dimethylaziridinyl)phosphinic amides: a series of new antineoplastic agents.

In continuation of efforts to improve the antitumor selectivity of the 2,2-dimethylaziridine class of alkylating agents, a series of N-substituted bis(2,2-dimethyl-1-aziridinyl)phosphinic amides has been synthesized and evaluated. All of these compounds (3-15) were tested in vivo against leukemia P-388 in mice, where most of them caused significant increase of survival time at nontoxic dose levels. Some of the most active compounds were also tested against leukemia L1210, B16 melanoma, and colon 26 carcinoma; in the latter tests, the parent unsubstituted amide 3 appeared to show the highest antitumor activity. Since the dose-limiting toxicity of the clinically tested prototypes of this class of anticancer agents AB-132 (1) and AB-163 (2) had been found to be CNS toxicity attributable mainly to the inhibition of cholinesterase, the compounds were tested in vitro against the cholinesterases from horse serum, electric eel, and bovine erythrocytes, as well as in vivo for the inhibition of the cholinesterase present in the whole blood of mice. In all of these assays, the various members of the present series showed a wide range of anticholinesterase activities, ranging from almost zero (for 3) to even higher potency than that of the prototype 2. A similarly wide range of stability was observed toward hydrolytic ring opening of the 2,2-dimethylaziridine moieties. Several of the compounds, particularly 3, deserve further study.

Nucleosides. 133. Synthesis of 5-alkenyl-1-(2-deoxy-2-fluoro-beta-D-arabinofuranosyl)cytosines and related pyrimidine nucleosides as potential antiviral agents.

The synthesis of 1-(2-deoxy-2-fluoro-beta-D-arabinofuranosyl)cytosines with a halovinyl or vinyl substituent at C-5 was accomplished from the corresponding 5-iodo (FIAC, 1) and/or 5-chloromercuri nucleoside analogues with use of Li2PdCl4- and Pd(OAc)2-mediated coupling reactions. Thiation of the benzoylated derivative of the 5-ethyluracil nucleoside 3 followed by S-methylation and then ammonolysis provided 5-ethyl-2'-fluoro-ara-C. 5-Ethynyl-2'-fluoro-ara-C (19a) and 5-ethynyl-2'-fluoro-ara-U (19b) were also obtained from the persilylated 5-iodo nucleosides 1 and 16, respectively, by PdII/CuI catalyzed coupling with (trimethylsilyl)acetylene. With use of selective sugar deprotection of the initial coupling products with H2O/Me2SO, the corresponding 5-[2-(trimethylsilyl)ethynyl] derivatives 18a and 18b could be isolated. Most of the new compounds showed activity in vitro against both HSV-1 and HSV-2, as did the known corresponding 5-alkenyluracil nucleosides synthesized earlier. The 5-vinylcytosine and -uracil nucleosides 10 and 24, respectively, were highly effective against HSV-1 (ED90 = 0.40 and 0.043 microM, respectively) and HSV-2 (ED90 = 0.59 and 0.56 microM, respectively). Unlike BVDU, the 2'-fluoroarabinosyl derivatives of 5-(halovinyl)cytosine and -uracil showed activity against both types of herpes simplex virus. The therapeutic indices of these compounds are in some cases superior to those of 2'-fluoro-5-methyl-ara-U (FMAU, 2). Moderate antileukemic activity was observed in vitro for the 5-alkynyl and 5-vinyl compounds. The competition of these compounds with thymidine for viral-induced thymidine kinases was also studied.

Synthesis and purification of the anti-viral agent 1-(2-deoxy-2-fluoro-beta-D-arabinofuranosyl)-5-iodocytosine (FIAC) labeled with iodine-125.

The 125I labeled analog of 1-(2-deoxy-2-fluoro-beta-D-arabinofuranosyl)-5-iodocytosine (FIAC) has been prepared for studies on the imaging of herpes simplex virus (HSV) encephalitis infections in animals. Iodination of 1-(2-deoxy-2-fluoro-beta-D-arabinofuranosyl)cytosine X HCl (FAC X HCl) with Na 125I and NaI/HIO3 in aqueous acetic acid, and subsequent removal of residual iodine by CCl4 extraction and ion exchange chromatography, yielded [125I]FIAC in aqueous solution with a specific activity of 45.5 mCi/mmol in 44% radiochemical yield, and with a radiopurity of 96-97%. Methods for further purification are described.