Thermospray and electrospray mass spectrometry of flavocoenzymes. Analysis of riboflavin sulphates from sugar beet

ANALYTICA CHIMICA

ACT4 ELSEVIER

Chimica Acta 302 (1995) 215-223

Analytica

Thermospray and electrospray mass spectrometry of flavocoenzymes. Analysis of riboflavin sulphates from sugar beet J. Abihn a,*, S. Susin b, J. Abadia b, E. Gelpi a a Department of Medical Bioanalysis, CID, CSIC, Barcelona, Spain b Department of Plant Nutrition, Estacibn Experimental de Aula Dei, CSIC, Zaragoza, Spain Received 14 June 1994; accepted 2 October 1994

Abstract Thermospray (TSP) liquid chromatography mass spectrometry and electrospray (ESP) mass spectrometry were applied to the analysis of the flavocoenzymes riboflavin (Rfv), riboflavin 5’-monophosphate (FMN) and flavin adenine dinucleotide (FAD). Positive and negative ion spectra are presented. TSP spectra of Rfv and FMN were characterized by the presence of intense and characteristic fragment ions containing the flavin ring whereas in those from FAD, ions containing the adenine moiety predominate. ESP affords mainly molecular weight information as well as some indication of the number of acidic places in these molecules. Application of these techniques to the analysis of root extracts from iron-depleted sugar beet (Beta uulgaris) was instrumental in the characterization of two novel riboflavin compounds, riboflavin 3’-monosulphate (3-FMS) and riboflavin S-monosulphate (S-FMS), as the major flavins in roots. Keywords:

Mass spectrometry;

Flavocoenzymes;

Riboflavin

sulphates;

Sugar beet

1. Introduction Flavins

are redox

cofactors

in many

biological

redox processes such as mitochondrial electron transfer, oxidation by oxidases and oxygenases and reduction by dehydrogenases. The more common flavocoenzymes are riboflavin (Rfv, vitamin B,), riboflavin S-monophosphate (FMN) and flavin adenine dinucleotide (FAD). Flavins can participate in either one- or two-electron redox reactions and when reduced, also react with molecular oxygen to form stable radicals. The ubiquity of some of these compounds and their capacity to serve as a link between

* Corresponding

author.

0003-2670/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved SSDf 0003-2670(94)00490-O

one- and two-electron transfer processes places these coenzymes among the most versatile redox cofactors known. The chemical structures of Rfv, FMN and FAD are characterized by a isoalloxazine group substituted in the nitrogen at position 10 by a ribityl moiety (Fig. 1). Analytical methods for these compounds generally use liquid chromatography (LC) with spectrophotometric detection [l-3] taking advantage either of the fluorescence or UV-visible absorbance properties of the isoalloxazine chromophore. Although sensitive, the spectrophotometric detection is not specific and compound identification is based on retention time. Due to their lability, high polarity and ionized phosphate moieties these compounds are not suitable for conventional mass spec-

J. Abicin et al. /Analytica

216

Chimica Acta 302 (1995) 215-223

!

2.2. Preparation

of flauin extracts

B HZ HW

-H

HO-

6

HOI'

-H -H H,

! e n&

I ‘,”

“u ’ ,” 0

Fig. 1. Chemical structures of riboflavin (R = H), FMN (R = PO,H,) and FAD (R = 5’-diphosphatidyl adenosine).

trometric (MS) techniques and can not be analyzed by gas chromatography-mass spectrometry. MS data from these compounds are, thus, scarce and mostly directed to the study of fragmentation processes [4,5]. A good approach for the analysis of flavins could be the use of soft ionization techniques capable to be coupled to liquid chromatography. Continuous-flow fast atom bombardment coupled to LC through a frit interface has been recently used for the analysis of Rfv [6]. A full scan spectrum required, however, around 300 ng of standard riboflavin. As far as we know, no other attempts have been reported on the analysis and characterization of these compounds by other common soft ionization techniques such as thermospray (TSP) or electrospray (ESP). In this work, we compare the results obtained with these two techniques using Rfv, FMN and FAD as model compounds. We recently reported the presence of two new riboflavin sulphates in roots of sugar beet (Beta vulgaris) and spinach (Spinacea oleracea) grown in iron deficient media [3,7,8]. To our knowledge, this was the first time that these compounds were reported to be present in biological media. The application of TSP and ESP MS to their analysis is also described in some detail.

2. Experimental 2.1. Chemicals Rfv, FMN and FAD were obtained from Sigma (St. Louis, MO). Other reagents were obtained from Fluka (Madrid) or Merck (Darmstadt) and were chromatographic or p.a. grade.

Root extracts from sugar beet were prepared as described earlier [3]. Briefly, extracts were obtained from plants grown without iron for lo-12 days. Root tips exhibiting a yellow color were excised and ground in dim light at 4°C in a PTFE pestle glass homogenizer with ice cold ammonium acetate 0.1 M, pH 6.1 (7 ml extractant per gram of fresh root). The extracts were centrifuged for 10 min at 12,000 g. Supernatants were lyophilized and stored at -40°C until analysis. 2.3. Mass spectrometry TSP and ESP mass spectra were obtained using a Finnigan (San Jose, CA) TSQ 700 triple quadrupole mass spectrometer equipped with a Finnigan interface for TSP and an Analytica of Branford (Branford, CT) interface for ESP. 2.4. TSP LC-MS Chromatography was carried out by using a Pump System 32X from Kontron (Barcelona), a Model 7125 Rheodyne (Cotati, CA) injector with a 20-~1 sample loop. A 5-pm (15 X 0.46 cm i.d.) Spherisorb ODS-2 column (Phase Separations, Norwalk, CT) and a 3-pm (5 X 0.46 cm i.d.) Spherisorb ODS-2 column were used. The 15-cm column was used for gradient elution of standards and flavin extracts and the 5-cm column for isocratic chromatography of the standards. Gradient elution was carried out using two different solvents: (A) a 0.05 M ammonium acetate solution in water (1% formic acid), and (B) a 0.05 M ammonium acetate solution in water-methanol (l:l, v/v, 1% formic acid). Ammonium acetate buffers were prepared from ammonia and acetic acid as previously described [9]. Chromatography was carried out at a flow rate of 1 ml/min using a solvent program starting at 5% solvent B. After injection, a linear gradient up to 40% B in 4 min and then to 65% B in 8 min was initiated. For chromatography of the standards the first gradient step was eliminated. The interface temperature was set around 105°C. The source and manifold temperatures were kept at

217

J. Abicin et al. /Analytica Chitnica Acta 302 (1995) 215-223

[M-H]

80

255

[M-BH+FO]-

40’

/

179

\ ‘7

20-

TM’/’ 241

[M+FO]

[Al‘

421

193 27’ zp1 200

250

1:.

, :. 100

I

FMN (+)

%I

loo1

,? :. 350

:‘,:. (00

.I?,, . w (50 500

455

FMN (-)

I [M-H]

37

[M+H]

Wd+

FAD (+) F4H,]+ /

100

1

I

V-41’

.

I

I

FAD i-1 I

[F3H+FO]-

269 I

tF2t-41’ 348

K-W 439

KW’ [M-H]

‘AH,]+ \.

7v

m/z

m/z

Fig. 2. Positive and negative ion TSP mass spectra of Rfv, FMN and FAD. Spectra were generated from an injection of ca. 500 ng on -oIumn. Isocratic elution at 35% methanol in the eluent. See Fig. 3 for fragment ion nomenclature. FO, formate; AcO, acetate.

218

J. Abirin et al. /Analytica

Chimica Acta 302 (1995) 215-223

NH

An

RIBOFLA’JINE

Fl

Mr 242

Fig. 4. Tentative mechanism

Fig. 3. Major fragments

observed in the TSP spectra of flavins.

250°C and 70°C respectively. Acquisition was made at a rate of 0.5 scans/s. The electron multiplier was set at 1400 V. 2.5. ESP MS For ESP analysis, standards were dissolved in methanol at a concentration of 10 pmol/pl. The flavin concentration in the plant extracts was approximative and based on solution absorptivity at 445 nm. Flavin solutions were introduced into the ESP source at a rate of 1 $/min using a Harvard Apparatus (Southnatick, MA) perfusion pump. Operating conditions were as follows: ESP voltage, 3500 V (positive ions) and 2800 V (negative ions); drying nitrogen flow, 5 ml/min; scan range, 200-1000 amu; scan rate, 2 s. Spectra presented here are the averaged signal from 5 min acquisition.

3. Results and discussion

Mr134

for fragment AH formation

in TSP.

particularly intense for Rfv and FMN. In the negative ion (NH mode the corresponding signals are observed at m/z 241 and 255. The origin of these and other ions in the spectra could be explained as suggested in Fig. 3. Fragments AH and BH that originate from the common ions [AH,]+ and [BH,]+ (m/z 243 and 257) and [A]- and [B]- (m/z 241 and 255) contain the flavin ring system and can be used as tracers for the presence of this moiety. These ions were also the major fragments detected during the FAB analysis of Rfv [6]. The structures of AH and BH could correspond to those of the Rfv photodealkylation products lumichrome and lumiflavin, respectively. A tentative mechanism for the formation of AH from Rfv in TSP is depicted in Fig. 4. Formation’ of AH through this process should give rise to the concomitant formation of fragment Fl. This is supported by the presence in the TSP spectra of an intense signal at m/z 152 ([Fl + NH,]+). Formation of fragments AH and BH is strongly dependent on probe temperature and eluent composition (Table 1). The relative abundance of the fragment AH and BH derived signals diminished with

3.1. TSP LC-MS of riboflavin compounds Positive and negative TSP spectra of these compounds are presented in Fig. 2. Molecular weight information can be obtained from all the spectra except for FAD in the positive ion (PI) mode. Besides ions derived from the intact molecule such as the [M+H]+, [M+Na]+ and [M-H]ions, several diagnostic fragment ions are present in the spectra. In the PI mode the signals at m/z 243 and 257 are common to these structures and are

Table 1 Effect of methanol concentration ion formation by TSP LC-MS [Ion] +

%Methanol

in 0.1 M ammonium

Rfv

MNa MH BH AH

in the TSP buffer on fragment acetate

FMN

35%

50%

35%

50%

25 75 100 85

70 100 12 14

20 40 50 100

60 100 45 25

J. Abicin et al./Analytica

219

Chimica Acta 302 (1995) 215-223

the proportion of water in the eluents. Thus, a high concentration of methanol in the eluents is recommended in order to obtain more intense molecular adducts. Rfv and FMN show very similar spectral patterns. FMN show additional losses of HPO, to give signals at m/z 377 ([CH,]+, where C refers to fragment C as in Fig. 3) and 399 ([CH + Na]+) in the PI mode and at m/z 375 (C-) in the NI mode (See Fig. 2). Elimination of H,PO, accounts for signals at m/z 359 (PI mode) and 357 (NI mode). In addition to the isoalloxazine-containing fragments, the TSP spectra of FAD are characterized by other fragment ions derived from the presence of the adenylic moiety. Fragmentation at the carbonnitrogen bond between the adenine and the ribose moieties (F4) generates the signals at m/z 136 ([F4H,]+) in the PI mode and at m/z 134 (F4-), 180 ([F4H + formate]-) and 194 ([F4H + acetate]-) in the NI mode. The adenosine moiety (F3) is observed at m/z 268 ([F3H,]+) in the PI mode and at m/z 266 (F3-), 312 ([F3H + formate]-) and 326 ([F3H + acetate]-) in the NI mode. The adenylic acid is observed at m/z 348 ([F2H,]+) and at m/z 346 (F2-). Signals at m/z 439 (PI) and 437 (NI) could be assigned to a dehydrated FMN and thus could be formed after elimination of water and breakdown of the diphosphate bond. However, water elimination does not seem to be an important process for these compounds because FMN does not show water losses in its spectrum. On the other hand, hydrolysis of the diphosphate bond to give an FMN molecule does not seem to take place either because the FMN ions are not observed in the PI and NI spectra of FAD, Thus, ions at m/z 439 and 437 ([IX-II+ and [D - HI- in Fig. 2, respectively) could be better explained as the result of the elimination of adenosine monophosphate by an intramolecular substitution reaction giving rise to a cyclic phosphate diester. Cyclic diester structures are well known in the FMN chemistry [lo] and are intermediate compounds in the FMN isomerization that takes place in the liquid phase [ll]. Rfv, FMN and FAD were readily separated by LC using ammonium acetate buffer-methanol mixtures as indicated in Methods (Fig. 5). Full scan spectra (m/z 130-800) could be obtained with less than 100 ng of each compound. Taking into consideration the

1.1 Ed4

a) TSP (+)

Rfv FAD

I

I

FMN

0

2

4

6

6

10

12

14

8

10

12

14

mln

31E1P3 1 b) TSP (-)

0

2

4

6

min

Fig. 5. TSP LC-MS of Rfv, FMN and FAD (300 ng each) in the (a) positive and (b) negative ion modes. LC conditions as indicated in Experimental.

[M + H]+ and [M - HI- signals for Rfv and FMN and the [F3H,]’ and [F3]- signals for FAD, response follows the order: FAD (PI) > Rfv (PI) > FAD (NI) > FMN (PI) > Rfv (NI) > FMN (NI) (35% methanol in the eluents). 3.2. ESP MS of ribojlauin

compounds

TSP makes use of a heated stainless steel capillary tube to produce the evaporation of that part of the eluent necessary to generate the spray. Thus, during TSP ionization, the sample molecules can be submitted to temperatures as high as 200-3OO”C, depending on the temperature of the different heated zones, the characteristics of the liquid droplet in the spray from which the sample will be evaporated or desorbed as well as its location along the system [12]. Also, the analyte can interact with active points on the hot interface and source surfaces, so that fragmentation of labile substances is often observed. In contrast, the dissolved analyte is only slightly heated by a N, countercurrent gas flow during ESP

220

J. Abicin et al. /Anafytica

Chimica Acta 302 (1995) 215-223

,

Rfv (+)

100-

[hl+Na]+

[M-H]-

PI

3;

80-

[M+H]+ GO-

[2M+Na]+

I

455

FMN (+)

I

[M+2N&HJ+

1

[M-H] ,-

[B+2Na-H]+

80

FMN(_1

413 1 60’

I

“-

[BH+Na]+ IM+Na]+

\ 20-

I 100

t

28

740

[2M+Na-ZH]

I

I

301

I

1

/ ail-

{BH+2Na-HI+

3

I

392

FAD (-)

FAD (+) [M-2Hj-*

413

[M+?Na-H] [M-SH-HIGo-

806

526

&l-

J

233

20-

146

Fig. 6. ESP MS spectra of Rfv, FMN and FAD (10 pmol/Fl

I

in methanol-water,

955).

ESP conditions

as indicated

in Experimental.

J. Abicin et al. /Analytica

221

Chimica Acta 302 (1995) 215-223

analysis. This gas flow is necessary to desolvate the charged liquid droplets that constitute the spray. Under these conditions, liquid phase fragmentation processes and thermal degradation on instrument surfaces or in the gas phase are less favored. These differences between TSP and ESP are clearly observed by comparing the spectra in Figs. 2 and 6. The ESP spectra of flavins showed mainly molecular weight information with signals due to proton abstraction in the NI mode and proton and sodium adducts in the PI mode (Fig. 6). Except for FAD, only low abundance ions derived from the fragment BH were present in the PI and NI spectra of these compounds. The relative abundance of these fragments is only important when the complete molecule itself shows low response to the ionization mode. The best responses were obtained for FAD and FMN using the NI mode. Only for FAD in the PI mode a molecular ion could not be observed. In the NI mode, FMN and FAD showed doubly charged ions at m/z 227 (FMN) and 391.7 (FAD). These multicharged ions reflect the minimum number of acidic places in the FMN and FAD structures and thus are useful for characterization. NI ESP spectral patterns and sensitivity were very dependent on solvent parameters such as pH or solvent composition. This is a particular characteristic of ESP ionization where the observed spectra are determined by the ions pre-existent in solution and thus by the relative liquid phase proton affinity (or basicity) of analyte and other solvent components. On the contrary, in TSP, the observed spectral pattern for a given molecule is directed by the gas phase chemical ionization processes that take place between the analyte and solvent molecules and ions in the high pressure source volume. The spectra shown in Fig. 6 were obtained from methanolic solutions of the pure standards containing ca. 3-5% water. Under these conditions, PI spectra showed to be reproducible from day to day, but this was not the case for NI spectra where FMN and FAD doubly charged and molecular ion relative abundances were found to change significantly. This effect is clearly observed comparing the NI spectrum from 10 pmol of FMN in Fig. 6 and the one presented in Fig. 7. In the latter spectrum, trace amounts of triethylamine (ca. 0.05%) were added to the solvent (10% water in methanol). Addition of triethylamine produced two major ef-

P 30

Fig. 7. ESP MS spectrum of FMN (10 pmol/pl) in the negative ion mode in methanol-water (9O:lO) with 0.05% of triethylamine.

fects: a decrease of the total ion current and an increase of the relative abundance of the doubly charged ion relative to the [M - HI- ion. The loss in total intensity is probably due to the increase of the spray electrical conductivity reducing the actual ESP potential. Several methods to overcome this problem and to prevent corona discharges when increasing the ESP voltage have been discussed before [13] and could be also applied in the analysis of flavins. The increase of the [M - 2Hlm2 relative abundance is the result of the higher liquid phase deprotonation due to the more basic media as discussed above. In practice, trace amounts of contaminating acids or bases from the samples, extraction buffers or solvents, could be responsible for these differences. ESP spectrum reproducibility is thus strongly related to sample preparation. A good solution to solve this problem is the prepurification of the flavin by absorption on a C,, solid phase extraction cartridge and posterior elution with methanol (not shown). 3.3. Analysis of root extracts In a previous report we showed that LC-UV analysis of root extracts from plants grown in iron deficient media affords clean profiles showing the presence of only two absorbance peaks [3]. These peaks were isolated by LC and characterized by ESP, nuclear magnetic resonance (NMR) and inductively coupled plasma spectroscopy (ICP) as Rfv 3’-sulphate (3-FMS) and Rfv S-sulphate (5-FMS), two compounds not previously described in biological media. The chemical synthesis of the FMN analogue 5-FMS

222

J. Abicin et al. /Analytica

Chimica Acta 302 (1995) 215-223

from Rfv was reported by Egami and Yagi 30 years ago [14]. In these studies, it was shown that this analogue inhibits D-amino oxidase by competing with FAD [14] and shows bacteriostatic activity by acting as an antivitamin B, [15]. After the characterization of Rfv sulphates as endogenous substances in plants, their physiological role and metabolic pathways stand to be elucidated. In order to develop methods to check for the presence of FMN, Rfv sulphates and other related compounds in plant extracts we tested the utility of direct TSP LC-MS or ESP analysis of the crude extracts. TSP LC-MS profiles obtained from a root extract are shown in Fig. 8. Important differences were observed between the PI and NI acquisition modes. In the PI mode (Fig. 8a) the profile showed several signals due to compounds that were not detected in the UV analysis. In contrast, the NI acquisition mode afforded a clean profile with two major, partially

I I.30

I

2‘42

1

a) TSP (+) 359

2.6 E47

1 5

a) TSP (+) In/z Fig. 9. TSP LC-MS spectra of 5-FMS in the (a) positive and (b) negative ion acquisition modes.

mln

10

15

b) TSP (-) 3-FMS

5-FMS

Fig. 8. TSP LC-MS profiles obtained from root extracts in (a) the positive and (b) negative ion acquisition modes. Conditions as indicated in Experimental.

resolved signals corresponding to the Rfv monosulphates. The spectra of these compounds (Fig. 9) show a pattern related to that of FMN (Fig. 2) although some differences in the relative ion abundances allow their differentiation. Elimination of H,SO, from Rfv sulphates takes place to a higher extent than the elimination of H,PO, from FMN. Thus, in the PI mode, the [M + H]+ ion observed in the spectrum of FMN was not observed for Rfv sulphates and, in the NI mode, the base peak ([M HI- for FMN) corresponded to the ion at m/z 357 derived from the elimination of H,SO, from the parent [M - HI-. Also, 3-FMS can be differentiated from 5-FMS by comparing their different relative abundances of the ions at m/z 357 and 359 in their TSP spectra (not shown). A more detailed study on the fragmentation of Rfv sulphates and phosphates will be reported elsewhere. Direct analysis of crude extracts by negative ion

.I. Abirin et al. /Analytica

Chimirn Acta 302 (1995) 215-223

ESP allowed the rapid characterization of the presence of Rfv sulphates. The ESP spectra obtained from these samples showed a unique and strong signal at m/z 455 ([M - HI-) (not shown) characteristic of Rfv sulphates [3,7]. Due to the acidic sulphate moiety and given that the samples are conveniently diluted, the ionization of Rfv sulphate does not seem to be interfered by other components in the sample. This characteristic could be used to monitor plant extracts or LC fractions for the presence of FMN and Rfv sulphates. It should be noted that detection using this method takes advantage of both the presence of the sulphate (or phosphate) moiety and the atomic composition of the analyte (molecular weight). This contrasts with more conventional spectrophotometric methods where detection relies only on the flavin chromophore and no information can be obtained on the ring substituents.

4. Conclusions TSP LC-MS and ESP MS can be used complementarily for the analysis of flavocoenzymes. ESP MS allows molecular weight determination for Rfv and nucleotide compounds by selecting the positive and negative ion mode detection, respectively. TSP ionization produced some fragmentation probably due to thermal degradation in the heated TSP probe. Fragments detected in the PI mode can be used with advantage for the characterization of the isoalloxazinc moiety. Riboflavin monosulphates can be easily detected and characterized by TSP LC-MS. Initial complete characterization required, however, the use of other techniques such as ESP, NMR and ICP spectroscopy. Work is in progress on the study of the application of tandem MS to afford structural information.

223

Acknowledgements The authors gratefully acknowledge Montse Carrascal and Albert Llonch for their skillful assistance in obtaining mass spectra.

References 111R.P. Hausinger,

J.F. Honek and C. Walsh, Methods Enzymol., 122 (1986) 199. 121P. Nielsen, J. Harksen and A. Bather, Eur. J. Biochem., 152 (1985) 465. M.L. Peleato, A. [31 S. Susin, J. Abiin, F. Sgnchez-Baeza, Abadia, E. Gelpi and J. Abadia, J. Biol. Chem, 268 (1993) 20958. [41 P. Brown, CL. Hornbeck and J.R. Cronin, Org. Mass Spectrom., 6 (1972) 1383. M. Bock, M. Elsner, II. Kurreck and F. 151 G. Holzmann, Miiller, Org. Mass Spectrom., 23 (1988) 789. [61N. Asakawa, H. Ohe, M. Tsuno, Y. Nezu, Y. Yoshida and T. Sato, J. Chromatogr., 541 (1991) 231. 171 J. Abiln, F. Sanchez-Baeza, E. Gelpi, S. Susin and J. Abadia, Proceedings of the 41st ASMS Conference on Mass Spectrometry and Allied Topics, San Francisco. May 31-June 5, 1993, p. 33a. A. Bl S. Susin, J. Abiln, M.L. Peleato, F. Sinchez-Baeza, Abadia, E. Gelpi and J. Abadia, Planta, (1994) in press. [91 J. AbiBn, F. SBnchez-Baeza, E. Gelpi and D. Barcelb, J. Am. Sot. Mass Spectrom., 5 (1994) 186. [IO1 P. Nielsen, P. Rauschenbach and A. Bather, Anal. Biochem., 130 (1983) 359. P. Nielsen, J. Harksen and A. Bather, Eur. J. Biochem.. 152 (1985) 465. A.L. Ycrgey, Ch.G. Edmonds, LAS. Lewis and M.L. Vestal, [a in D. Hercules (Ed.), Liquid Chromatography/Mass Spectrometry. Techniques and Applications, Plenum Press, New York, 1990. [131 R.F. Straub and R.D. Voyksner, J. Am. Sot. Mass Spectrom., 4 (1993) 578. [I41 F. Egarni and K. Yagi, J. Biochem., 43 (1956) 1.53. 1151 F. Egarni, M. Naoi, M. Tada and J. Yagi, J. Biochem., 43 (1956) 669. [Ill