ANALYTICALBIOCHEMISTRY

Analytical Biochemistry 341 (2005) 214–219

www.elsevier.com/locate/yabio

Fluorimetric HPLC detection of endogenous riboXavin 4�,5�-cyclic phosphate in rat liver at nanomolar concentrations

José Canales, Alicia Cabezas, Rosa María Pinto, José Carlos Cameselle ¤

Unidad de Bioquímica y Biología Molecular, Facultad de Medicina, Universidad de Extremadura, 06080 Badajoz, Spain

Received 31 August 2004Availabel online 1 April 2005

Abstract

We have investigated the biological occurrence of riboXavin 4�,5�-cyclic phosphate (cyclic FMN or cFMN), the Xavin productknown to be formed in vitro from FAD by the rat liver enzyme FAD–AMP lyase (cyclizing) or FMN cyclase (EC 4.6.1.15). Theexpected diYculties were the low concentration of the compound, the tendency of the more abundant FAD to decompose chemicallyto cFMN, and the acid lability of cFMN itself. A protocol was devised to prepare deproteinized rat liver extracts, avoiding condi-tions prone to the chemical formation of cFMN and making exposure to conditions of cFMN degradation as short as possible. Inthese extracts, cFMN was assayed by HPLC with Xuorimetric detection. The identity of liver cFMN was conWrmed by its HPLC sep-aration from other known Xavins, its coelution with authentic cFMN, and its susceptibility to acid degradation, yielding a mixture of5�-FMN and 4�-FMN. The amount of total cFMN recovered in the liver extracts was 22 § 11 pmol/g fresh tissue. Careful controlexperiments were performed to rule out the possibility that this could be a chemical product of FAD degradation during extractpreparation. These controls indicated that, on average, 97% of the measured extract concentration of cFMN, i.e., about21 § 10 pmol/g, was already present in the liver at the beginning of the process and was extracted from the tissue. This Wgure is likelyto be an underestimation of the hepatic content, as indicated by control experiments. 2005 Elsevier Inc. All rights reserved.

Keywords: Cyclic FMN; Flavin nucleotide; Cyclic nucleotide; FMN cyclase; Rat liver

RiboXavin 4�,5�-cyclic phosphate (cyclic FMN orcFMN1; Fig. 1) was identiWed very early as a product ofthe chemical decomposition of FAD in alkaline medium[1,2]. Later it was found also as a contaminant ofunknown origin in commercial preparations of 5�-FMN[3], and more recently, its presence in extracts obtainedfrom baker’s yeast and food materials has been reported,supposedly as a consequence of FAD degradation dur-

* Corresponding author. Fax: +34 924289468.E-mail address: camselle@unex.es (J.C. Cameselle).

1 Abbreviations used: cFMN, riboXavin 4�,5�-cyclic phosphate; FAD,Xavin–adenine dinucleotide; 4�-FMN, riboXavin 4�-monophosphate;5�-FMN, riboXavin 5�-monophosphate; HPLC, high-performanceliquid chromatography; ribo-NDP-X, ribonucleoside 5�-diphosphateesteriWed by an X-OH group.

0003-2697/$ - see front matter 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.ab.2005.01.060

ing the extraction and processing of the samples [4,5].The Wrst hint that cFMN could be biologically relevantwas the discovery of an enzyme in rat liver that catalyzesFAD decomposition by an internal attack of the 4�-OHgroup of the ribityl chain over the proximal P atom, withthe reaction stoichiometry FAD ! cFMN + AMP [6]. Inaddition to FAD, this enzyme acts on a selection ofother ribo-NDP-X substrates. However, it has beennamed FAD–AMP lyase (cyclic FMN-forming), FAD–AMP lyase (cyclizing), or just FMN cyclase (EC 4.6.1.15)because it displays the highest catalytic eYciency (kcat/Km) toward FAD. ADP-glucose is also a good substrate,with a kcat/Km only 2-fold lower than FAD, but it is not aphysiological compound in mammals. The other ribo-NDP-X substrates of the enzyme display 50- to 500-foldlower kcat/Km values [7]. Therefore, the possibility should

Fluorimetric HPLC detection of riboXavin 4�,5�-cyclic phosphate in liver / J. Canales et al. / Anal. Biochem. 341 (2005) 214–219 215

be considered that FAD could be the physiological sub-strate, and that cFMN is the speciWc biological productof FMN cyclase.

The physiological occurrence of cFMN in biologicalsources and any possible physiological role of thisintriguing compound remain to be demonstrated. There-fore, we decided to look for this compound in rat liver,aiming to show its physiological occurrence in the samebiological source from which FMN cyclase was isolated.The expected diYculties were the presumably low con-centration of cFMN against the complex background ofa hepatic extract, including a high FAD content (about20–40 nmol/g [8,9]), and the possibility that the presenceof cFMN could be accounted for by the chemical degra-dation of FAD [1,2]. The Wrst drawback was overcomethrough the use of HPLC and the high sensitivity andrelative speciWcity of Xavin-containing compoundstoward Xuorimetric detection. The second drawbackforced us to carry out a series of carefully controlledexperiments to monitor how each step of the procedureaVected the Wnal cFMN content of the extract, to ruleout the possibility that its presence was exclusively dueto the chemical lability of FAD.

Materials and methods

Materials

Female Wistar rats fed ad libitum, weighing 200–250 g, were used throughout the experiments. FAD and5�-FMN were high-purity (97%) preparations fromSigma. Authentic cFMN, used as a standard for theassay and characterization of the compound in liverextracts, was also a high-purity (99%) preparation chem-ically obtained from FAD as described [10].

Fig. 1. RiboXavin 4�,5�-cyclic phosphate, cyclic FMN, or cFMN. Thisproduct is the Xavin product formed by the decomposition of FADcatalyzed by FAD–AMP lyase (cyclizing) or FMN cyclase, an enzymeidentiWed and puriWed from rat liver [6,7].

Standard preparation of rat liver extracts containing cFMN

To recover cFMN from rat liver, the following protocolwas used, where labels a–f correspond to the time pointsindicated in Table 1. Livers were excised (a) under etheranesthesia and dropped (b) into ice-cooled 10mM EDTAkept in a 50-ml metallic chamber of a Sorvall Omni-Mixerblender. After rapid weighing, more cold EDTA solutionwas added up to a total of 1ml per gram of fresh tissue.Homogenization was performed at once for 60 s at setpoint 7 of the Omni-Mixer blender. Immediately, enoughice-cooled 0.4M perchloric acid was added (c) to thehomogenate to reach a Wnal concentration of 0.2M, andanother round of homogenization was immediatelyapplied for 60s. The acid homogenate was immediatelycentrifuged (d) for 5min at 27,000g and 4 °C in a SS-34Sorvall rotor. The deproteinized supernatant was decantedand neutralized to near pH 7 (e) by the gradual addition ofice-cold KOH solutions under rapid magnetic stirring (anysample that reached pH 7.4 or higher during the neutral-ization was discarded). After 10min at 4°C, the neutral-ized supernatant was centrifuged for 5min at 27,000g, toremove the potassium perchlorate formed, and was kept at4°C in the dark until the time of HPLC analysis (f).

Estimation of the amount of cFMN formed or degraded during each step of the standard procedure for the preparation of liver extracts

The standard preparative protocol was divided intoWve strictly timed steps as indicated by the six timepoints marked by letters a–f (see above and Table 1). Tocheck how the conditions of each step aVected the WnalcFMN content of the extract, a series of controls (Exper-iments 1–5, following; Fig. 2) were run in which thelength of one of the steps was intentionally increased,

Table 1Standard preparative protocol

Note. The standard preparative protocol was divided into Wve strictlytimed steps as indicated by the six time points shown. To determinehow each step of the standard procedure aVected the Wnal cFMN con-tent of the extract, control experiments were run as described in Mate-rials and methods (Experiments 1–5).

¤ The indicated time values correspond to total time elapsed sincepoint a during preparative runs following the standard procedure. Ineach particular experiment, the length of each step was exactly timed.

Time point¤ Action

a 0 min Liver excision from anesthetized ratb 1–2 min Start Wrst homogenization in EDTAc 2–3 min Start second homogenization in

perchloric acidd 5–6 min Start centrifugation of acid homogenatee 20–25 min End of neutralization to pH 7.0–7.3f t24 h HPLC analysis

216 Fluorimetric HPLC detection of riboXavin 4�,5�-cyclic phosphate in liver / J. Canales et al. / Anal. Biochem. 341 (2005) 214–219

while the other steps were carefully adjusted to the con-ditions of the standard procedure according to Table 1.

Experiment 1 (to check step a-b). After excision, theliver was divided into two portions. One was processed

Fig. 2. Control experiments performed to analyze the stability ofcFMN during each step of the extraction protocol. (A) Primary dataobtained in Experiments 1–5 described under Materials and methods.Each panel shows the results of several independent experiments inwhich one of the steps was lengthened with respect to the standardprocedure. The data are expressed as picomoles of cFMN found in therat liver extract per gram of fresh tissue. In the experiments with morethan two points, the lines were drawn by linear regression. (B) Estima-tions of net changes of cFMN during each step. The upper panelshows the average rates of cFMN change, with error (SD) bars, calcu-lated from the experiments of the corresponding panel in (A). Thelower panel shows the average change of total cFMN per gram of liverfor each step, as calculated from the average rate changes and the timelength (Table 1) of each step in the standard protocol.

immediately and the other was left on ice before resum-ing the procedure, so that the hepatic material wasexposed to these conditions longer (23–43 min) than inthe standard procedure.

Experiment 2 (to check step b-c). After the Wrst homog-enization in EDTA was Wnished, the homogenate wasdivided into two or three parts. One part was processed atonce and for the others the addition of perchloric acid andthe start of the second homogenization were delayed for8–40 min before the procedure was resumed.

Experiment 3 (to check step c-d). After the secondhomogenization in perchloric acid was Wnished, thehomogenate was divided into two or three parts. Onewas processed at once and for the others the start of cen-trifugation was delayed for 10–110 min before the proce-dure was resumed.

Experiment 4 (to check step d-e). After the centrifuga-tion and removal of the protein precipitate were Wnished,the acid extract was divided into two or three parts. Onewas processed at once and for the others the start of theneutralization was delayed for 8–100 min.

Experiment 5 (to check step e-f). After the neutraliza-tion and removal of the potassium perchlorate wereWnished, the extract was stored at 4 °C, and HPLC analy-ses were performed immediately and repeated daily forup to 10 days.

Control experiments carried out with aqueous solutions of Xavin nucleotides

Several experiments simulating the preparation of thedeproteinized liver extracts were carried out with aque-ous solutions of FAD, 5�-FMN, or cFMN substitutingfor liver tissue. Ten-milliliter samples of one Xavin nucle-otide were treated strictly as if they were 10 g of rat liver(see extract preparation and Table 1). The cFMN con-tent of all the samples was assayed by HPLC, both in theuntreated solutions and at the end of the mock-extrac-tion procedure.

HPLC analysis of cFMN in liver extracts

The concentration of cFMN in the deproteinizedextracts was assayed by HPLC on a Hewlett–Packard1100 chromatograph with Xuorescence detection (excita-tion at 450 nm, emission at 520nm). For data handlingand peak integration, the HP ChemStation software wasused. The HPLC method was that used previously [6] withminor modiWcations. The chromatographic column wasof Hypersil ODS 5�m (150£4 mm) from Teknokroma,protected with a precolumn (10 £4 mm) of the samematerial. Samples were applied to the column with a 50-�linjection loop. The chromatography was developed at aXow rate of 0.5ml/min with a 20-min linear gradient ofmethanol 15–30% (vol/vol) in 10 mM sodium phosphate(pH 7.0). Under these conditions, the cFMN peak dis-

e in liver / J. Canales et al. / Anal. Biochem. 341 (2005) 214–219 217

played a retention time of about 15min, well resolvedfrom other Xavins present in the rat liver extract (Fig. 3A).The area of this peak was used to measure the concentra-tion of total cFMN in the extract by comparison to thearea increase of the same peak after addition of a knownamount of standard cFMN as a calibrator (Fig. 3B). Rou-tinely, 5–20�l of 200 nM cFMN was added over 195–180�l of liver extract (after neutralization of the per-chloric acid with KOH and centrifugation) to make 200-�lspiked samples that were analyzed by HPLC in parallelwith nonspiked extract samples (as in Fig. 3). The dilutionof the extract in the spiked samples was taken intoaccount to calculate the size of the signal corresponding tothe calibrator. Each sample was analyzed at least in tripli-cate using undiluted extract and a 1:1 dilution in water.

Preparative isolation of cFMN from liver extracts

To allow the characterization of rat liver cFMN, thepeak that showed the same HPLC retention time as theauthentic compound was puriWed. Twenty samples of250 �l each were chromatographed under the same con-ditions employed for analysis, except that a 500-�l injec-tion loop was used. The cFMN peak was collected at theend of the column. The material thus obtained waspooled, lyophilized, and dissolved in 1 ml of water.

Results and discussion

The chemical decomposition of FAD to cFMN isfavored by the presence of divalent metals, alkaline pH,or high temperatures [1,4,6–8,10]. On the other hand,acidic conditions are known to favor the degradation ofcFMN to a mixture of 4�-FMN and 5�-FMN [10,11].

Fig. 3. HPLC identiWcation of cFMN in deproteinized rat liver extracts.(A) A 50-�l sample of rat liver extract analyzed as described underMaterials and methods. The cFMN peak in this chromatogram corre-sponds to 8 nM cFMN in the processed extract (25 pmol/g fresh liver).According to the control experiments (Fig. 2B) and to the length of thepreparative steps in this particular case, only 3% of this concentrationcan be accounted for by the net formation of cFMN during liver pro-cessing and extract preparation. (B) Same as (A) except that the samplewas spiked with standard cFMN to a Wnal concentration of 10 nM.

Under near neutral conditions, without deproteiniza-tion, cFMN can be formed in rat liver extracts by FMNcyclase [6,7] and can be degraded by a so-far uncharac-terized cyclic phosphodiesterase [6]. FMN cyclase isinhibited by EDTA [6,7], but the cyclic phosphodiester-ase is not (results not shown). Hence we devised a proce-dure to prepare deproteinized liver extracts, payingparticular attention to avoiding conditions prone to thechemical and enzymatic formation of cFMN and tomaking as short as possible the exposure of the extractsto conditions of chemical and enzymatic degradation ofcFMN. The protocol included rapid liver homogeniza-tion in EDTA, deproteinization by perchloric acid, andrapid neutralization to pH 7 under rapid stirring. Everystep of the procedure was performed at 0–4 °C, with pre-cooled solutions, under strictly timed conditions (seeMaterials and methods and Table 1).

The quantitation of cFMN in deproteinized extractswas performed by HPLC analysis with Xuorimetric detec-tion. Fig. 3A shows the Xuorescence tracing of a liverextract sample in which cFMN appears as a well-resolvedpeak with a retention time of 15.3 min. Fig. 3B shows theanalysis of the same sample after supplementation withan authentic standard of 10 nM cFMN. The Xuorescencesignal (area) of the cFMN peak was linear with respect toits concentration both when a pure standard was ana-lyzed and when standard cFMN was added as a supple-ment to a liver extract. The results of Fig. 4 show thelinear character of the cFMN assay in both the absenceand the presence of liver extract. The parallelism betweenthe two lines indicates that the presence of the extract didnot interfere with the Xuorimetric quantitation of cFMN.This was also supported by the routine practice of assay-ing cFMN in undiluted and in a 1:1 dilution of theextract, which gave the expected values (not shown).

The authenticity of cFMN in the liver extract wasstrongly indicated by two criteria. Standard cFMN coelutedstrictly with the extract peak at 15.3min, and it was easilydistinguishable from other Xavin derivatives expected to bepresent in the liver extract (FAD, 5�-FMN, and riboXavin;

Fig. 4. Calibration of the HPLC assay of cFMN with Xuorimetricdetection. Standard cFMN dissolved in water (�) or added to liverextract after deproteinization and neutralization (�) was analyzed byHPLC without further treatment. The area of the cFMN peak wasmeasured in every case. FLD, Xuorescence detector.

Fluorimetric HPLC detection of riboXavin 4�,5�-cyclic phosphat

218 Fluorimetric HPLC detection of riboXavin 4�,5�-cyclic phosphate in liver / J. Canales et al. / Anal. Biochem. 341 (2005) 214–219

Fig. 3). As is known for the standard compound [10,11],cFMN in the liver extract decomposed in 1M HCl at 45°Cto a mixture of 4�-FMN and 5�-FMN (Fig. 5).

To correct the concentrations of cFMN found in theliver extract, the time-dependent changes of cFMN dur-ing each step of the preparation of the liver extracts weremeasured in a series of control experiments (see Experi-ments 1–5 in Materials and methods).

Step a-b. Four independent experiments of type 1were performed and in all cases it was found thatthe concentration of cFMN was decreased (Fig. 2A, stepa-b). However, since each independent experiment con-tained only two points, no evidence that the decreasewas linear with time was obtained. Therefore, the estima-tion of the rate of decrease of cFMN (Fig. 2B, upperpanel) must be considered a minimum estimation. Fromthis, the total negative contribution of step a-b to theWnal measured cFMN concentration was calculated tobe about ¡0.7 pmol/g (Fig. 2B, lower panel). The actualnegative value can be higher, but this does not interfere

Fig. 5. Susceptibility of rat liver cFMN to acid degradation to a mix-ture of 5�-FMN and 4�-FMN. cFMN, preparatively isolated from ratliver extract (see Materials and methods), was incubated in 1 M HCl at45 °C. The Wgure shows the results of analyses performed on untreatedcontrols, and on cFMN samples treated for 1 or 40 min. Similar condi-tions were applied to liver extract cFMN, standard cFMN and anequimolar mixture of both.

with the conclusion of the work in relation to the biolog-ical character of cFMN in rat liver. The endogenoushepatic cFMN content may actually be higher than thatcalculated in the present work.

Step b-c. Seven independent experiments of type 2were performed, and in all cases it was found that theconcentration of cFMN was decreased (Fig. 2A, stepb-c). Three of the experiments that contained threepoints gave results indicative of a linear decrease. Fromthe estimated change rate (Fig. 2B, upper panel), thetotal negative contribution of this step to the Wnal mea-sured cFMN concentration was calculated to be about¡0.8 pmol/g (Fig. 2B, lower panel).

Step c-d. Seven independent experiments of type 3were performed, and in all cases it was found that theconcentration of cFMN was increased (Fig. 2A,step c-d). The two experiments that contained threepoints yielded results indicative of a linear increase.From the estimated change rate (Fig. 2B, upper panel),the total positive contribution of this step to the Wnalmeasured cFMN concentration was calculated to beabout +0.2 pmol/g (Fig. 2B, lower panel).

Step d-e. Four independent experiments of type 4 wereperformed, and in all cases it was found that the concen-tration of cFMN was increased (Fig. 2A, step d-e). Two ofthe experiments contained three points and yielded resultsindicative of a linear increase. From the estimated changerate (Fig. 2B, upper panel), the total positive contributionof this step to the Wnal measured cFMN concentrationwas calculated to be +2.0 pmol/g (Fig. 2B, lower panel).

Step e-f. Three independent experiments of type 5 wereperformed (Fig. 2A, step e-f), and in all of them cFMNremained stable well after the 24-h period within which theHPLC assays were performed under standard conditions.

The decrease of cFMN during step a-b and b-c can bedue to the action of a cyclic nucleotide phosphodiester-ase present in rat liver [6]. The source for the increases ofcFMN seen during steps c-d and d-e is uncertain. Thecontrol experiments carried out with aqueous solutionsof 70 �M FAD or 12 �M 5�-FMN, which were subjectedto the full extraction protocol in the absence of liver tis-sue (see under Materials and methods), failed to showthe formation of cFMN. The recovery of cFMN in thesecontrols (cFMN is a known contaminant of commercialFAD and 5�-FMN, according to [3] and unpublishedresults from our laboratory) was 57 § 11% (n D 3) in thecase of FAD samples and 86 § 11% (n D 4) in the case of5�-FMN samples. However, the failure to detect net for-mation of cFMN from commercial FAD or 5�-FMNcould be the result of a turnover in which cFMN wasgenerated at a rate lower than that of its degradation. Onthe other hand, the increase of cFMN in liver extractsduring steps c-d and d-e could result from cFMNgeneration at a rate higher than that of its degradation.In summary, the controls carried out with aqueoussolutions of FAD and 5�-FMN do not make it possible

Fluorimetric HPLC detection of riboXavin 4�,5�-cyclic phosphate in liver / J. Canales et al. / Anal. Biochem. 341 (2005) 214–219 219

to rule out the possibility than any of these Xavins couldbe the source of cFMN during steps c-d and d-e.

Nonetheless, the measured rates of cFMN change inliver extracts during steps a-b to e-f can be used to calcu-late the amount of cFMN formed or degraded in thehepatic material between the sacriWce of the animals andthe HPLC assay (Fig. 2), and then one can estimate howmuch of the total cFMN measured represents theendogenous pool of cFMN in liver.

In a series of 21 independent experiments, we assayedcFMN in the livers obtained from 21 rats. Themean § SD of the total cFMN content in the extractswas 22 § 11 pmol/g fresh liver. Summing up the contri-butions of steps a-b to e-f (Fig. 2B, lower panel), the netcontribution of the whole processing of rat liver extractsto the Wnal cFMN content was on average +0.7 pmol/g(considering the variability of the results shown in theupper panel of Fig. 2B, this value could be between ¡0.7and +2.1 pmol/g). Since the net amount of cFMNformed during extract processing was about 3% of themeasured content, one can conclude that 97% of thecFMN concentration measured in the extract, i.e., about21 § 10 pmol/g, was present in liver and was extractedfrom the tissue. This Wgure is possibly an underestima-tion of the actual hepatic cFMN content, because whenaqueous cFMN solutions were submitted to a mockextraction protocol, including all the standard preparativesteps, except that liver tissue was not included, the recov-ery of cFMN was only about 50% (44% of an 8.5�McFMN sample, 43% of a 100 nM cFMN sample, and60§ 12% of three 20 nM cFMN samples). Therefore weestimate the endogenous cFMN concentration in rat liverto be about 40 nM, assuming an even tissue distribution.

The low concentration of cFMN in liver and itsknown potential formation by degradation of FADexplain why so far the biological character of cFMN hasremained unrecognized. For instance, when a few yearsago the compound was detected in baker’s yeast extracts,it was presumed to be formed by nonenzymatic degrada-tion of FAD [4]. Our results show that cFMN is a naturalcomponent of rat liver, where it may arise from the actionof FMN cyclase, known to be present in this tissue [6,7].

Although the biological role of cFMN remains to beelucidated, we suggest that it may be a signal molecule, aminor redox Xavocoenzyme, or an intermediate of a deg-radative pathway for FAD. While a nanomolar concen-tration is a low level for many intracellular metabolites,it is noteworthy that other biologically relevant cyclicnucleotides occur at similar or even lower concentra-tions, for instance, 3�,5�-cyclic GMP in rat liver [12].

Conclusion

cFMN has been identiWed in deproteinized extractsfrom rat liver as an endogenous hepatic compound.

Control experiments indicated that most part of cFMNin liver extracts is not formed from other Xavin nucleo-tides during tissue processing and extract preparation.The measured concentration of cFMN corresponds to aliver content of about 20 pmol/g, but the actual tissuecontent may be higher, as indicated by a 50% recovery ofcFMN in control experiments and by the possibility thatthe rate of cFMN degradation measured in one of thesteps of the extraction protocol was an underestimation.We propose that cFMN is a low-abundance naturalcompound in rat liver, where cFMN-forming FAD–AMP lyase is also present [6,7].

Acknowledgment

This work was supported by Grant BMC2001-0719from the Dirección General de Investigación (MCYT)and FEDER.

References

[1] H.S. Forrest, A.R. Todd, Nucleotides. Part V. RiboXavin-5� phos-phate, J. Chem. Soc. (1950) 3295–3299.

[2] F.M. Huennekens, D.R. Sanadi, E. Dimant, A.I. Schepartz, Theoccurrence of a new Xavin dinucleotide (FAD-X), J. Am. Chem.Soc. 75 (1953) 3611–3612.

[3] P. Nielsen, P. Rauschenbach, A. Bacher, Phosphates of riboXavinand riboXavin analogs: a reinvestigation by high-performance liq-uid chromatography, Anal. Biochem. 130 (1983) 359–368.

[4] A. Gliszczynska, A. Koziolowa, Chromatographic determination ofXavin derivatives in baker’s yeast, J. Chromatogr. A 822 (1998) 59–66.

[5] A. Gliszczynska-Swiglo, A. Koziolowa, Chromatographic deter-mination of riboXavin and its derivatives in food, J. Chromatogr.A 881 (2000) 285–297.

[6] F.J. Fraiz, R.M. Pinto, M.J. Costas, M. Ávalos, J. Canales, A.Cabezas, J.C. Cameselle, Enzymic formation of riboXavin 4�,5�-cyclic phosphate from FAD: evidence for a speciWc low-Km FMNcyclase in rat liver, Biochem. J. 330 (1998) 881–888.

[7] A. Cabezas, R.M. Pinto, F.J. Fraiz, J. Canales, S. González-Santi-ago, J.C. Cameselle, PuriWcation, characterization, and substrateand inhibitor structure–activity studies of rat liver FAD–AMPlyase (cyclizing): preference for FAD and speciWcity for splittingribonucleoside diphosphate-X into ribonucleotide and a Wve-atomcyclic phosphodiester of X, either a monocyclic compound or acis-bicyclic phosphodiester-pyranose fusion, Biochemistry 40(2001) 13710–13722.

[8] K. Yagi, Simultaneous microdetermination of riboXavin, FMNand FAD in animal tissues, Methods Enzymol. 18 (1971) 290–296.

[9] A. Hinkkanen, K. Decker, Luminometric determination of FADin subpicomole quantities, Anal. Biochem. 132 (1983) 202–208.

[10] R.M. Pinto, F.J. Fraiz, A. Cabezas, M. Ávalos, J. Canales, M.J. Costas,J.C. Cameselle, Preparation of riboXavin 4�,5�-cyclic phosphate byincubation of Xavin–adenine dinucleotide with Mn2+ in the absence ofriboXavin 5�-phosphate cyclase, Anal. Biochem. 268 (1999) 409–411.

[11] G. Scola-Nagelschneider, P. Hemmerich, Synthesis, separation, identi-Wcation and interconversion of riboXavin phosphates and their acetylderivatives: a reinvestigation, Eur. J. Biochem. 66 (1976) 567–577.

[12] E.K. Frandsen, G. Krishna, A simple ultrasensitive method for theassay of cyclic AMP and cyclic GMP in tissues, Life Sci. 18 (1976)529–542.