8448

Natural Toxins Nat. Toxins 7: 365-370 (1999)

RESEARCH ARTICLE

HPLC/MS Analysis of Fusarium Mycotoxins, Fumonisins and Deoxynivalenol t

Ronald D. Plattner*
Mycotoxin Research, United States Department of Agriculture, Agricultural Research Service, National Center for Agricultural Utilization Research, 1815 N. University Street, Peoria, IL 61604, USA

Fusarium fungi are widely found in agricultural products, worldwide and can produce a great variety of mycotoxins. Fumonisins, produced by F. moniliforme, and deoxynivalenol, produced by F. graminearum, are two such mycotoxins that have received considerable attention as food safety concerns by regulatory agencies. High Performance Liquid Chromatography/Mass Spectrometry (HPLClMS) was found to be a convenient analytical method to detect and quantify the naturally occurring fumonisin homologs and deoxynivalenol in extracts from grains and food products. The fumonisins are detected primarily as protonated molecules in the positive ion electrospray ionization (ESI) mode as they elute from a C-18 reverse phase column during a methanol water gradient containing acetic acid to facilitate chromatography. Deoxynivalenol can be detected as positive or negative ions in the atmospheric pressure chemical ionization (APCI) mode or in the negative ion ESI mode. One nanogram amounts of fumonisins or deoxynivalenol injected into the HPLC system are easily detected with signal to noise allowing detection limits of 1 ~lg g-1 or better to easily be achieved with minimal clean-up of grain extracts. Published in 1999 by John Wiley & Sons, Ltd.
Key words: Fusarium; fumonisin; trichothecene; deoxynivalenol; HPLC; electrospray

ABSTRACT

INTRODualON
Fungi of the genus Fusarium are widely found in plant debris and crop plants worldwide (Marasas et al., 1984). Several species from this genus are economically important because, in addition to their ability to infect important crops such as com, wheat and other small grains in the field, they produce mycotoxins in the crops in the field and in storage. The presence of high levels of mycotoxins in grains poses a significant economic threat since animals that consume contaminated grains may perform poorly or become sick and die. The threat of mycotoxins entering the human food supply through contaminated grain products and exposed animals is also a concern for regulatory officials. Two important families of Fusarium mycotoxins that have received considerable attention as food safety concerns by regulatory agencies are fumonisins and trichothecenes. Fumonisins are produced by the fungus F. monilijorme and some closely related species in section Liseola. Fumonisin B I is generally the most abundant member of the family of mycotoxins and is known to cause animal diseases including equine leucoencephalomalacia and porcine pulmonary edema. Additionally fumonisins are potent liver toxins in most
Published in 1999 by John Wiley & Sons, Ltd. -

animal species (Nelson et aI., 1993) and are suspected human carcinogens (Gelderblom et al., 1991). Among the trichothecenes, deoxynivalenol (DON, also known as vomitoxin) has probably had the greatest attention by regulatory agencies. DON is produced by various species of Fusariulil, most notably F. graminearum which causes fusarium head blight (FHB) or scab on wheat and other small grains. FHB occurs sporadically in wheat growing regions worldwide and has severely impacted wheat and barley in the upper Midwest over the last 5 or 6 years, causing severe economic losses to producers. The most widely used laboratory methods to analyze for fumonisins and DON are based on High Performance

*Correspondence to: Ronald D. Plattner, USDA-ARS, NCAUR, 1815 N. University Street, Peoria, IL 61604, USA. E-mail: plattnrd@mail.ncaur.usda.gov 7This article is a US Government work and is in the public domain in the USA. USDA disclaimer. Names are necessary to report factually on available data; however, the USDA neither guarantees nor warrants the standard of the products, and the use of the name by USDA implies no approval of the product to the exclusion of others that may also be suitable. Received 3 November 1999; Accepted 20 July 2000.

366

PLATINER

Liquid Chromatography (HPLC) with detection by ultraviolet (UV) absorbance or fluorescence (Sydenham et aI., 1996; Trucksess et aI., 1996). Fumonisins are typically extracted (water/methanol or water/acetonitrile) and concentrated by solid phase extraction before HPLC. Fumonisins have no UV chromophore so they are derivatized for fluorescence detection on the free amine group (Sydenham et al., 1996; Bennett and Richard, 1994). This procedure has very good sensitivity (detection limit typically 10-50 ~lg kg -1) but cannot detect the N-acetylated fumonisin analogs and other fumonisins that lack the free amine group that are sometimes present, albeit at much lower levels. DON is extracted from grain samples with acetonitrile/water (84116) and the sample extract is cleaned up by passing the extract through a clean-up column, concentrated under nitrogen, and an aliquot is analyzed by HPLC. Quantitation is based on UV absorbance at 220 nm. In recent years significant improvements in the coupling of HPLC and mass spectrometry (MS) has resulted in the emerging availability of relatively low cost HPLCIMS systems which are well suited to the analysis of small molecules. HPLCIMS utilizing electrospray ionization (ESI) offers similar detection limits to published methods for the analysis of fumonisins and DON as well as significant advantages.
MATERIALS AND METHODS Sample Preparation and Extraction

Because mycotoxins are likely to be unevenly distributed in naturally contaminated grains, a proper sample selection and preparation protocol are critical to generating useful data (Whitaker et aI., 1998). For fumonisin analysis com samples were ground and extracted with 5 ml g-I of acetonitrile/water (Ill) for 3h with shaking. The extract was filtered and stored at 4 DC until analysis. Alternatively, the extraction procedure reported by Shephard et al. (1990) using methanol/water also works well. For DON analysis ground samples were extracted with 4 ml g-I of acetonitrile/water (84/16) in Erlenmeyer flasks with shaking for 2 h. The extract was then filtered through filter paper and samples were stored in capped vials at 4 DC until analysis. The selection of a clean-up method for sample extracts prior to HPLC depends on what compounds need to be measured and what the needed detection limits are. For analysis of fumonisin B J, B 2 , B 3 and B4 in com and com products the best clean-up protocol is based on use of a strong ion exchange (SAX) column as reported by Sydenham (Sydenham et al., 1996). However, in this clean-up protocol any hydrolyzed fumonisins elute in the void volume of the SAX column because they do not contain the acidic side-chains and are lost. Additionally any partially hydrolyzed fumonisins with only one sidePublished in 1999 by John Wiley & Sons. Ltd.

chain have a very different retention on the SAX column and also can be lost dming clean-up. Thus, to measure the hydrolyzed and partially hydrolyzed series fumonisins as well. one must either analyze the extract without cleanup, modify the SAX clean-up, or use the C I8 column clean-up protocol (Rice et al., 1995) which does not lose the partially hydrolyzed and hydrolyzed B series fumonisins. Lime treated samples, such as masa or tortillas and products which produce extracts with pH values >6 should be acidified to pH 4 with drop-wise addition of phosphoric acid before clean-up on C 18 columns or, preferably, 0.1 M NaH 2 P0 4 at pH 3.0 should replace the water in the extraction solvent (Stack, 1998) to insure efficient extraction and retention of all fumonisins to be measured. Imuno-affinity columns have also been used for clean-up for HPLC analysis (Newkirk et aI., 1998), but the user must take care to assure that all analogs that need to be measured cross-react with the antibody and are retained on the imuno-affinity column so that they are properly recovered in any protocol used. Also the capacity of the imuno-affinity column must be sufficient to retain all desired components that may be present in the extract so that some are not lost due to column overload and the level is therefore underestimated. For DON analysis extracts of samples such as severely scabby wheat, in which the DON level is expected to be high and measurement of samples with DON levels below 2 ppm is not required, an aliquot of the extract (1 0 ~tl) can be analyzed without sample clean-up. For lower detection limits a slight modification of the cleanup reported for HPLC with UV detection (Trucksess et aI., 1996) is satisfactory. Briefly, a 6 ml portion of the extract is placed into a 15 x 85 mm test tube. A MycoSep 225 column (Romer Labs, Union, MO, USA) is inserted into the top of the test tube and slowly pushed to the bottom of the tube. Then 4 ml are taken from the top of the column and evaporated to dryness under nitrogen and redissolved in 0.4 ml of extraction solvent. A 10 ~tl aliquot is used for the HPLC analysis.
HPLC Analyses

HPLC was done on a Thermal Separations HPLC consisting of a model P4000 solvent delivery system and an AS3000 autosampler. Separations were done using an Intersil 5U ODS-3 column (ISO x 3.0 mm, MetaChem Technologies, Inc, Torrance, CA, USA). The flow was 0.3 ml min -I and the column effluent was directly coupled to a Finnigan LCQ mass spectrometer. The MS was operated in the ESI mode with an inlet capillary temperature of 220 DC and the sheath gas was nitrogen at 80 ml min -I. For DON analysis the instrument was operated in the negative ion mode. Fumonisins were detected in the positive ion mode.
Nat. Toxins 7: 365-370 (1999)

HPLC/MS ANALYSIS OF FUMONISINS AND DON

722.5

367

744.5

704.6

30

500

mfz

700

eo o

90

Figure 1. ESI mass spedrum of FB i

For DON analysis 10 ~tl aliquots of standard solutions or sample extracts were injected after the column was initialized with water/methanol (99/1) with 0.0 I % acetic acid. After injection, the solvent was held for 1 min and then programmed to water/methanol (67/33) with 0.3% acetic acid in 6 min and held until 17 min. The solvent was then returned to the starting solvent over I min and held until 25 min to re-equilibrate the column prior to injection of the next sample. Quantitation of DON in the samples was by the external standard method. The standard curve was made from 10 ~ll injections of DON solutions containing 0.1, 0.5, 1,5 and 10 ng ~tl-l of DON. A linear response was observed across this range (1100 ng injected). The detection limit for DON was consistently better than 0.5 ng injected on the column. This corresponded to a detection limit of approximately 2 ppm without any sample extract pretreatment or concentration and 0.2 ppm for the extracts that were cleaned up and concentrated as described above. For fumonisins 10 ~tl aliquots of sample extracts or standard solutions in acetonitrile water 1/1 were injected on to a column conditioned with water/methanol/acetic acid (65/35/0.35) at a flow rate of 0.3 ml min -I. The solvent was programmed in 10 min to water/methanol! acetic acid (5/95/0.35) and held for 20 min. The solvent was returned to the Oliginal composition in I min and equilibrated for 9 min before the next sample was injected. Quantitation of fumonisins was made by the external standard method. The standard curve was made from 10 III injections of solutions of fumonisins at
Published in 1999 by John Wiley & Sons. Ltd.

concentrations of 0.1, 0.5, I, 5, and lOng Ill-I. The detection limits for FB I, FB 2 and FB 3 were consistently below 0.1 ng injected on column. Response was linear across the sample concentration range and to at least 500 ng injected. Injection of larger amounts of fumonisins gave reasonable results, but created problems of sample CaITy-over between injections (discussed below) so samples were diluted to stay within the concentration range of the calibration standards.

RESULTS AND DISCUSSION Fumonisins

Electrospray MS is an ideal technique to detect and measure fumonisins (Plattner, 1995). Fumonisins tend to be ionic and produce abundant signals in both the positive and negative ion modes. In the positive ion mode, the protonated molecule (m/z 722 for fumonisin B I) is the base peak in the mass spectrum (Figure I). Little fragmentation is observed. Protonated sodium or potassium salts may be seen in samples when the salts aI'e present and the pH of the elution solvent system is not sufficiently acidic to exchange them away. In the negative mode the base peak in the mass spectrum is the M-H anion (m/z 720 for fumonisin B 1). Injection of 0.1 ng of FB I to a reverse phase column produces an easily detected peak in the ESI positive ion mode. Because fumonisins are ionic in solution, separations on reverse phase columns tend to be based on a mixture of reverse phase and ion exchange mechanisms. Columns
Nat. Toxins 7: 365-370 (1999)

368
10

PLATTNER

9
B

., 7

~ 6 c:

~ .,
.~

.!!! 4

a:

.,

Figure 2. HPLC chromatogram: total MS signal from injection of mixture of FBi (mlz722 at 12.5 min), FB 3 (mlz706 at 14.17 min), and FB 4 (m/z 690 at 14.88 min)

from various vendors have dramatically different selectivities. Fumonisins do not elute well from most reverse phase columns when they are injected and eluted with neutral unbuffered solvent systems. For best results the mobile phase should be acidic. This can be accomplished either by the addition of acetic or formic acid to the mobile phase, or by the use of a volatile buffer of 0.1-1 N ammonium acetate or ammonium formate. Under the conditions reported above separation was achieved for FBj, FB 2 , FB 3 and FB 4 (Figure 2) as well as their A series analogs and P series analogs (data not

shown). The C series fumonisins are incompletely separated from the B series fumonisins under these conditions except for FC 4 which is completely separated from FB 4 The hydrolyzed (HFB I), the partially hydrolyzed isomers (HHFB I) and FBI are also resolved although separation of the two HHFB I isomers is not complete (Figure 3). Qualitative identification of known fumonisins is relatively straightforward with retention time and the observation of a protonated molecule of the appropriate mlz value. Fumonisins can be quantified based on the external

Figure 3. HPLC chromatogram of total MS response for mixture of HFB i (9.89 min), HHFB 1 isomers (11.40 and 11.68 min) and FBi (12.61)

Published in 1999 by John Wiley & Sons. Ltd.

Nat. Toxins 7: 365-370 (1999)

HPLClMS ANALYSIS OF FUMONISINS AND DON

369

295

Figure 4. Positive and negative ion APCI spectra of DON

standard method using a response curve generated from injection of known quantities of reference fumonisins. The response of FBj, FB 2 and FB 3 under the conditions reported is virtually identical. The free amino group is the more easily protonated than the carboxyl groups on the side-chain. Thus the A series fumonisins which are acetylated on the amino group respond significantly less well. The spectra of the A series fumonisins are considerably different from fumonisins with a free amino group. Both the M + H ion (m1z 764 for FBI) and the M + Na ion (m1z 786 for FBI) are abundant. Purified FA I (as well as FA 2 and FA3 ) can undergo a rearrangement

which transfers the N-acetyl group at C-2 to the oxygen on C-3. This rearranged isomer elutes earlier from the HPLC column (just after FA l ) and has a spectrum that only has the M + H ion. A complicating effect on quantitation of fumonisins by LCfMS is their tendency to be reversibly adsorbed by the HPLC system. This is particularly a problem with solvent gradient conditions. After injections of relatively large amounts (> 100 ng) of fumonisin B I to the HPLC column, some FB I elutes from the column at the proper retention time when subsequent blank (solvent only) or sample injections are made. Similar reversible absor-

295.3

355.3

(a)
2 0
2 0 3 0 m/z 4 0

(b)
1 0

2 0

2 0

Figure 5. Negative ESI spectra of DON eluted with (a) water/methanol and (b) water/methanol plus 0.1 % acetic acid

Published in 1999 by John Wiley & Sons; Ltd.

Nat. Toxins 7: 365-370 (1999)

370

PLAnNER

bance was observed with the other fumonisin homologs but was not observed for any other compounds investigated. Thus it was concluded that when larger amounts of fumonisins were injected on the column they were being bound to the column packing material itself and then being desorbed in subsequent injections. This effect is generally small; the amount of FB I that elutes is generally < 1 % of the previous injection but after injections of more that 100 ng sample carry-over can persist for many injection cycles, making it nearly impossible to have a chromatogram that is completely negative for fumonisins. Thus considerable care must be exercised in interpretation of the significance of small FBI peaks in chromatograms. Injection of blank samples, considering the fumonisin levels in the last injected sample and care to avoid injection of high fumonisin concentrations can provide references to interpret the validity of small positive peaks. In spite of the above difficulties, FE b FB z and FB 3 can be reliably quantitated at levels that meet or exceed the levels that can be measured using OPA derivatives, with the advantages of not needing derivatiation, sample stability, more rigorous detection based MS and the ability to detect homologs without a free NH z group. Excellent agreement was achieved between analysis of corn extracts by the OPA method and by HPLCIMS without any sample clean-up with detection limits of about 0.5 J..1.g g-I. Clean-up using either published SAX or C18 solid phase extraction columns allowed detection at 10 J..1.g kg- I levels.
Deoxynivalenol

candy less cleaning and maintenance than the APCI interface to maintain good performance. For samples of scabby wheat from field and greenhouse studies, excellent agreement was achieved between the fluoroquant method (Malone et al., 1998) and LCIMS analysis of the sample extract across the range 5200 ppm. The detection limit for the LCIMS method without clean-up was >2 ppm. Using the clean-up procedure DON could be detected at lO-fold lower levels. The sensitivity of the MS detector in either the ESI or APCI modes was comparable to that achieved by measuring UV absorbance while providing the higher degree of specificity of mass-to-charge detection.

REFERENCES
Bennett GA. Richard. JL. 1994. Liquid chromatographic method for analysis of the napthalene dicarboxaldehyde derivative of fumonisins. J AOAC Inti 77:501-506. Gelderblom WCA. Kriek NPJ. Marasas WFO. Thiel PG. 1991. Toxicity and carcinogenicity of Fusarium monili/orllle metabolite fumonisin B j in rats. Carcinogenesis 12:1247-1251. Malone BR. Humphrey CW, Romer TR. Richard JL. 1998. One-step solid-phase extraction clean-up and fluorometric analysis of deoxynivalenol in grains. J AOAC Inti 81:448-452. Marasas WFO, Nelson PE. Toussoun TA. 1984. A Toxigenic Fusariulll Species: Identity and Mycotoxicology. Pennsylvania State University Press: University Park, Pennsylvania; 216. Nelson PE, Desjardins AE, Plattner RD. 1993. Fumonisins, mycotoxins produced by Fusarium species: biology, chemistry and significance. Ann Rev Plzytopatlzol 31:233-252. Newkirk DK, Benson RW, Howard Pc. Churchwell MI. Doerge DR, Roberts DW. 1998. On-line immunoaffinity capture, coupled with HPLC and electrospray ionization mass spectrometry, for automated detection offumonisins. J Agric Food Clzem46:1677-1688. Plattner RD. 1995. Detection of fumonisins produced in Fusarium monili/onne cultures by HPLC with electrospray MS and evaporative light scattering detectors. Nat Toxins 3:294-298. Rice LG. Ross PF. Dejong J, Plattner RD, Coats JR. 1995. Evaluation of a liquid chromatographic method for the determination of fumonisins in com, poultry feed and Fusariulll culture material. J AOAC Inti 78:1002-1009. Shephard GS, Sydenham EW, Thiel PG, Gelderblom WCA. 1990. Quantitative determination offumonisins B j and B 2 by high pressure liquid chromatography with fluorescence detection. J Liquid Clzromatogr 13:2077-2087. Sydenham EW, Shephard GS, Thiel PG, Stockenstrom S, Snijman PW, VanSchalkwyk EJ. 1996. Liquid chromatographic determination of fumonisins B io B 2 and B3 in com: AOAC-IUPAC collaborative study. J AOAC Inti 79:668-696. Stack ME. 1998. Analysis of fumonisin B j and its hydrolysis product in tortillas. J AOAC Inti 81:737-740. Trucksess MW, Read DR, Pender MK, Ligmond CA, Wood GE, Page SW. 1996. Determination and survey of deoxynivalenol in white flour. whole wheat flour. and bran. J AOAC Inti 79:883-887. Whitaker TB. Trucksess MW, Johansson AS. Giesbrecht FS, Hagler WM. Bowman DT. 1998. Variability associated with testing shelled com for fumonisin. J AOAC Inti 81:1162-1168. ' .;'i-; .' ..
~.

Analysis of extracts of wheat for DON is typically done by either electron capture GC of a suitable derivative or by HPLC with detection by UV at 220 nm (Trucksess et al., 1996). Coupling of the HPLC effluent to the ESI or APCI interface to the MS provides a complementary detection protocol with advantages over UV detection. When the APCI interface is used DON is protonated to yield a signal at mlz 297 and fragment ions. In the negative ion mode the DON gave an ion at mlz 295 (Figure 4). In this mode, it was noticed that the addition of small amounts of acetic acid to the solvent system resulted in the prominent formation of an adduct ion at mlz 355 instead of the ion at mlz 295. Similarly in ESI negative ion mode the adduct at mlz 355 (DON + acetic acid -H-) is detected (Figure 5). While APCI and negative ESI give similar detection limits for DON, ESI data are reported and used for routine DON assays because it was found that the ESI interface was much more rugged and forgiving when dirty samples were analyzed. The ESI interface maintained sensitivity after many injections of dirty samples and required signifiPublished in 1999 by John Wiley & Sons, Ltd.

SuP.'" u.s. DeDL" AlllcultulJat. Toxins 7: 365-370 (1999)

National Center for AgrIcultural
Utihzation Research, Peoria, Illinois