Simultaneous Quantification of Fumonisins and Their Hydrolyzed Metabolites in Donkey Matrices: A Tool for Exposure Assessment and Toxicokinetic Studies

Abstract

A novel, sensitive, and robust LC-MS/MS method was developed and fully validated for the simultaneous determination of fumonisins (FB₁, FB₂, FB₃) and their hydrolyzed metabolites (HFB₁, HFB₂, HFB₃) in donkey plasma, urine, and feces—three critical matrices for toxicokinetic studies. Sample preparation was optimized for each matrix: salting-out assisted liquid–liquid extraction (SALLE) with perchloric acidification for urine and feces, and a dilute–evaporate–shoot (DES) approach for plasma. Chromatographic separation was achieved on a BEH C₁₈ column with water-ACN containing 0.5% formic acid. The method demonstrated excellent linearity (R² ≥ 0.99), acceptable accuracy (mean recoveries: 73.3–111.5%), and good precision (intra- and inter-day RSDs < 20%). The limits of quantification (LOQ) for FBs and HFBs were 0.1–0.15 μg/L in plasma, 1.0 μg/L in urine, and 60 μg/kg in feces. To our knowledge, this is the first reported method capable of quantifying this comprehensive panel of analytes across multiple biological matrices in donkeys, providing an essential tool for future exposure assessments and pharmacokinetic research in this species.

1. Introduction

Fumonisins are a group of mycotoxins produced primarily by fungi such as Fusarium verticillioides and Fusarium proliferatum [1]. Among the numerous fumonisin analogues identified, Fumonisin B₁ (FB₁), B₂ (FB₂), and B₃ (FB₃) are the most prevalent and toxicologically significant, with FB₁ being the most abundant form found in nature [2]. Chemically, these compounds share a common 20-carbon aminopolyol backbone that is esterified with two tricarballylic acid (TCA) groups at the C₁₄ and C₁₅ positions (for FB₁), which is crucial for their biological activity [3]. These mycotoxins are common contaminants of maize and maize-based products worldwide. Contamination can occur in the field before harvest or during storage under conditions of high humidity and temperature, making preemptive control challenging [2]. Mycotoxin contamination is predominantly localized in the peripheral tissues of grains. Consequently, grain processing by-products commonly used in animal feed, such as wheat bran and distillers dried grains with solubles (DDGS), may contain particularly elevated levels of these toxins [4,5].

The widespread occurrence of fumonisins (FBs) in animal feed poses a significant risk to livestock health and production economics. FB₁ and FB₂ are hepatotoxic and nephrotoxic in a variety of animal species and have been classified by the International Agency for Research on Cancer (IARC) as a Group 2B carcinogen (possibly carcinogenic to humans) [6]. Their primary toxicity is mediated through ceramide synthase inhibition—triggered by structural similarity to sphingoid bases—resulting in disrupted sphingolipid metabolism, sphingoid base accumulation, and altered S1P-dependent signaling [7]. This disruption can manifest in animals as reduced growth rates, immunosuppression, and specific toxicoses such as equine leukoencephalomalacia (ELEM) and porcine pulmonary edema (PPE), leading to substantial economic losses [8]. Hydrolyzed fumonisins (HFBs) are formed through the cleavage of the TCA side chains (Figure 1). This hydrolysis can occur during certain food processing techniques (e.g., alkaline cooking or nixtamalization of corn) or, critically, in vivo through the metabolic activity of gastrointestinal microbiota and host carboxylesterases [9,10]. Both the parent FBs and their hydrolyzed metabolites can be detected in biological fluids and excreta, making them valuable biomarkers for assessing animal exposure and for conducting toxicokinetic and biotransformation studies [11,12,13].

The donkey (Equus asinus) is a large herbivorous animal valued for its resilience and ability to thrive on high-fiber, roughage-based diets [14]. In China, the donkey has transitioned from a traditional draft animal to an important economic species [15]. Donkey hides are the primary raw material for ejiao, a precious traditional Chinese medicine, while donkey meat is also consumed [16]. Notably, equids are known to be highly sensitive to fumonisin toxicosis, with horses (Equus caballus), the closest relative to donkeys, developing the characteristic and fatal equine ELEM upon FB₁ exposure [2]. Donkey diets often incorporate significant quantities of roughages and cereal by-products, which are known to be potential sources of mycotoxin contamination, including FBs [17]. Despite the known sensitivity of the equine family and the economic importance of donkeys, scientific research on the absorption, distribution, metabolism, excretion, and toxicological effects of FBs in donkeys is extremely limited. Analyzing the concentrations of FBs and their metabolites in donkey plasma, urine, and feces is an essential prerequisite for conducting any in vivo study aimed at understanding FB₁ exposure, toxicokinetics, and overall health impact in this species. To date, such analytical data and validated methods specific to donkey matrices are lacking in the scientific literature.

A considerable number of analytical methods have been established for the quantification of FBs, predominantly in grains and feed matrices. Analytical techniques include high-performance liquid chromatography with fluorescence detection (HPLC-FLD) after pre-column derivatization [9,12], as well as the increasingly employed liquid chromatography-tandem mass spectrometry (LC-MS/MS), which offers superior sensitivity, specificity, and the ability to detect multiple analytes simultaneously without the need for derivatization [18,19,20]. Sample preparation approaches vary widely and include simple “dilute and shoot” protocols, solid-phase extraction (SPE) using various sorbents (e.g., strong anion exchange (SAX), C₁₈, mixed-mode phases), immunoaffinity columns (IAC), and modified QuEChERS (Quick, Easy, Cheap, Effective, Rugged, Safe) methods [3]. In contrast, the development of analytical methods for the determination of FBs and HFBs in complex biological matrices (e.g., plasma, urine, feces) is less advanced and shows considerable variability in approach and performance. For example, De Baere et al. (2018) developed a UPLC-MS/MS method for FB₁ and HFB₁ in broiler chicken plasma using a novel protein precipitation and phospholipid removal plate (Ostro™), achieving limits of quantification (LOQ) between 0.72 and 2.5 ng/mL and reporting mean recoveries ranging from 61.6% to 70.8% [10]. For the analysis of FB₁ in human urine, Shetty and Bhat (1998) employed a method involving Amberlite XAD-2 resin cleanup followed by strong anion-exchange (SAX) SPE and HPLC-FLD, reporting a limit of detection (LOD) of 8 ng/mL and recoveries between 93.6% and 94.4% [21]. Additionally, Zhang et al. (2022) developed a QuEChERS-UPLC-MS/MS method for the simultaneous determination of FB₁, FB₂, FB₃, HFB₁, HFB₂, and HFB₃ in broiler chicken feed and excreta [13]. Their method achieved an LOQ of 2.0 μg/kg for all analytes and reported mean recoveries between 82.6% and 115.8% with a precision (RSD) of 3.9–18.9% [13]. However, existing methodologies are limited by their narrow matrix scope, incomplete coverage of FBs and their hydrolyzed metabolites, and the limited availability of a universal sample preparation protocol. This clear research gap underscores the necessity for a novel analytical approach that is comprehensively validated for the simultaneous analysis of multiple biological matrices in donkeys—specifically, plasma, urine, and feces. These matrices are critical for assessing exposure, toxicokinetics, and excretion, and their analysis has been highlighted in prior reviews on mycotoxin biomarkers [6].

To bridge this gap, we developed a robust, sensitive, and specific LC-MS/MS method for the simultaneous quantification of FBs and HFBs in the challenging matrices of donkey plasma, urine, and feces. This method incorporates distinct yet optimized sample preparation procedures tailored to the specific characteristics of each matrix. Although tailored for donkeys, the integrated analytical strategy holds broader applicability, offering a valuable reference for method development in other species. This validated tool will facilitate critical in vivo research into fumonisin toxicology in donkeys, ultimately aiding in improved risk assessment and species protection.

2. Results and Discussion

2.1. Optimization of MS Parameters

To maximize sensitivity and specificity, mass spectrometric parameters were optimized for each analyte. Standard solutions (1 μg/mL) were directly infused to scout both ESI modes.

The results indicated that the FBs could form stable ions in both modes, corresponding to [M + H]⁺ in ESI⁺ and [M − H]⁻ in ESI⁻. In contrast, the HFBs were detectable only in ESI⁺ as [M + H]⁺ ions. Notably, the signal intensities of FBs in ESI⁺ were more than one order of magnitude higher than those observed in ESI⁻. This predominant response in positive mode, attributable to more efficient protonation pathways, led to the selection of ESI⁺ for all subsequent analyses, aligning with established literature [3,10].

For each analyte, the two most abundant and specific precursor ion → product ion transitions were selected for multiple reaction monitoring (MRM), in accordance with the identification criteria outlined in the SANTE/11312/2021 guidelines [22]. It is noteworthy that the two pairs of positional isomers, FB₂/FB₃ and HFB₂/HFB₃, share virtually identical primary fragment ions due to their highly similar chemical structures. Therefore, unambiguous identification and quantification of these isomer pairs relied exclusively on their chromatographic separation, which was successfully achieved (as detailed in Section 2.2). The optimized MRM parameters are summarized in

Table 1. Optimized parameters of the mass spectrometer for the mycotoxin analysis.

Analyte	Precursor	Product Ions (m/z) a	RT b (min)	CV c (V)	CE d (eV)
FB1	722.4	352.4/334.4	4.27	62	34/40
FB2	706.4	336.4/354.4	4.79	62	36/32
FB3	706.4	336.4/354.4	4.58	2	36/32
HFB1	406.3	370.2/352.2	4.18	16	18/22
HFB2	390.4	336.2/354.2	4.89	2	20/20
HFB3	390.4	336.2/354.2	4.58	2	20/20
[13C34]-FB1	756.5	374.5/356.5	4.27	34	16/10
[13C34]-FB2	740.5	358.5/376.5	4.79	20	16/10
[13C34]-FB3	740.5	358.5/376.5	4.58	26	14/10

^a m/z of the quantifier and qualifier ions, respectively; ^b RT = Retention time; ^c CV = Cone voltage; ^d CE = Collision energy.

.

2.2. Optimization of Chromatographic Conditions

The optimization of chromatographic conditions is pivotal for achieving satisfactory ionization efficiency, peak shape, and resolution of isomeric analytes in LC-MS/MS analysis [23]. Initially, the separation performance of two columns widely employed in mycotoxin analysis, the Waters ACQUITY BEH C₁₈ (100 mm × 2.1 mm, 1.7 µm) and the HSS T3 (100 mm × 2.1 mm, 1.8 µm) [10,17,24], was evaluated under identical mobile phase conditions. As illustrated in Figure S1, the BEH C₁₈ column demonstrated superior resolution for both FBs and HFBs compared to the T3 column. Notably, the C₁₈ column successfully achieved baseline separation for the positional isomers FB₂/FB₃ and HFB₂/HFB₃. Subsequently, the influence of the organic modifier was investigated by comparing acidified water-MeOH and water-ACN systems. While both systems yielded comparable peak areas, the water-ACN system produced sharper peaks with narrower peak widths (Figure 2). Consequently, the BEH C₁₈ column with a water-ACN mobile phase was selected as the optimal combination for this method.

The composition of mobile phase additives significantly influences chromatographic behavior and mass spectrometric response. We systematically evaluated the effects of FA, AA, NH₄OH, and NH₄OAc in the water/ACN system. The presence and type of additive were found to be critical for the detection of the target analytes. As shown in Figure 3, no detectable peaks were observed for FBs under no-additive, 0.1% NH₄OH, or 0.1% AA conditions. Similarly, HFBs exhibited severe peak tailing without an additive and were undetectable with NH₄OH. This phenomenon can be attributed to the molecular structures of FBs and HFBs, which contain multiple carboxylic acid groups and/or hydroxyl groups [24]. In insufficiently acidic conditions, these groups remain deprotonated, leading to poor retention on reversed-phase columns and inefficient ionization in the positive ESI mode due to the lack of readily protonatable sites. Further comparison of different concentrations of FA and AA revealed that FA provided a more significant signal enhancement for FBs. Consistent with prior literature reporting enhanced FB response with increased FA concentration [10,13], the present study found that increasing the FA concentration from 0.1% to 0.5% resulted in an average 40% increase in the peak areas of FBs, with no further improvement observed at 1.0%. The superior performance of FA over AA may be attributed to its stronger acidity (lower pKa), which could more effectively suppress the dissociation of the analytes’ carboxylic acid groups. This mechanism would promote better chromatographic retention and enhance the formation of [M + H]⁺ ions in the ESI source. Conversely, the addition of NH₄OAc to the FA-containing mobile phase strongly suppressed the signals of both FBs and HFBs, most likely due to ion suppression and the establishment of an unfavorable ionic equilibrium that compromises the protonation efficiency of the analytes in the ESI source. Therefore, water-ACN with 0.5% FA was selected as the final mobile phase. The extracted ion chromatograms of all analytes under the optimized conditions are presented in Figure 2B.

A significant carry-over effect in the chromatographic system was observed for FBs, estimated to be up to 5% of the previous injection’s concentration. Given that such carry-over has been a concern in prior method development for FBs [10,25], the effect is attributed to the strong adsorption of these polar, multi-carboxylic acid compounds to the metallic surfaces of the autosampler needle and injection port. To mitigate this, the needle wash protocol was optimized by adding 0.5% formic acid to the wash solvent (90% methanol) and extending the post-injection wash time from 10 s to 40 s. This modified wash procedure successfully reduced the carry-over value to below 0.5%, a negligible level, thereby ensuring the reliability of subsequent injections. To confirm the absence of long-term carry-over, solvent blanks injected after the highest calibration standard showed no detectable analyte peaks (S/N < 3), validating the effectiveness of the optimized wash protocol.

2.3. Optimization of Sample Pre-Treatment

The primary objective of this study was to develop a rapid and sensitive LC-MS/MS method for the simultaneous quantitation of FBs and HFBs in donkey plasma, urine, and feces. This method is intended to support toxicokinetic and exposure assessment studies, which typically involve a large number of samples. Consequently, a simple and practical sample preparation procedure is advantageous [10]. Conventional extraction and clean-up techniques, such as solid-phase extraction (SPE) [21,26] and immunoaffinity columns (IAC) [27,28], are often more complex, time-consuming, and costly. Given the inherent high sensitivity and specificity of LC-MS/MS detection, stringent purification of crude extracts is often not required [3]. The most common simplified preparation protocols for mycotoxin analysis in biological matrices using LC-MS/MS include the “dilute-and-shoot” (DAS) approach, the “dilute-evaporate-shoot” (DES) modification, and salting-out assisted liquid–liquid extraction (SALLE) [29]. Our investigations demonstrated that for complex matrices like donkey urine and fecal slurry, the SALLE method effectively removed interfering impurities while maintaining high extraction efficiency. Conversely, for plasma samples, a simpler DES procedure was found to be sufficient, providing adequate extraction efficiency and sensitivity.

2.3.1. Urine Sample

While the DAS approach has been reported as suitable for mycotoxin analysis in human urine, attributed to its relatively simple matrix [30,31], this method was not applicable to donkey urine in the present study. Donkey urine exhibited intense coloration, a strong odor, and a complex matrix, which led to significant signal suppression for both FBs and HFBs when using DAS, rendering the method unsuitable under these conditions. Therefore, the SALLE method was adopted to mitigate matrix interferences [29]. Preliminary experiments revealed that the presence of two carboxylic acid groups in FB molecules caused them to remain predominantly in the aqueous phase during the salting-out process. Efficient transfer of FBs to the organic phase (ACN) required sufficient acidification of the urine sample. The acidification efficiency of AA (organic) and perchloric acid (inorganic, used as a 35% v/v aqueous solution, PCA solution) was evaluated. As shown in Figure 4, satisfactory extraction recoveries (80–120%) for HFBs were achieved under all tested acidification conditions. However, for the parent FBs, recoveries exceeded 80% only when the concentration of AA reached 30% (v/v of sample) or PCA solution reached ≥5% (v/v of sample). This requirement for a higher acid loading compared to the 1% FA used for pig urine SALLE [29] is likely due to the alkaline nature and high buffering capacity of donkey urine, resulting from its high salt content.

The matrix effects (ME) for the target analytes were further compared between urine acidified with 30% AA and 10% PCA solution (Figure 5). The ME profiles were largely similar between the two conditions, with FB₂ and FB₃ showing strong signal enhancement, while FB₁ and the HFBs exhibited moderate signal suppression. Notably, when PCA solution was used, the organic layer turned black if not separated from the aqueous layer promptly after SALLE, possibly due to oxidation of urinary components by the strong acid. Therefore, rapid processing is essential when using PCA solution. Considering the large volume and high volatility of AA required, acidification with 10% (v/v of sample) PCA solution was selected as the optimal condition. Unlike a reported DAS method for human urine involving a 9-fold final dilution [31], the sample concentration factor in our final extract is 1, leading to improved analytical sensitivity.

2.3.2. Feces Sample

Donkey feces contain undigested forage, feed components, and endogenous metabolites. Importantly, true blank fecal samples (free of FB contamination) were unavailable, as preliminary analysis detected FBs in all samples collected. For method development and validation, a fecal sample with the lowest measured FBs concentration was selected as the representative matrix for optimization and standard addition experiments. FBs originate from both unabsorbed dietary fractions and biliary excretion, while HFBs derive from hepatic metabolism and microbial degradation in the gut [9,10]. To obtain a homogeneous matrix, fecal samples were first lyophilized and ground. A slurry was prepared by adding 5-fold (w/v) water to the lyophilized powder, followed by SALLE extraction. Similar to urine, effective extraction of FBs from the fecal slurry required adequate acidification. The optimal acid requirement and its impact on recovery were consistent with findings for urine; addition of 10% (v/v of sample) PCA solution provided satisfactory extraction efficiency (Figure S2). Considering that the concentrations of FBs and HFBs in feces are generally much higher than those in urine and plasma, we adjusted the workflow to balance matrix suppression with analytical sensitivity. Specifically, after SALLE treatment and complete evaporation of the extract, the residue was reconstituted to a final volume that resulted in a 15-fold dilution of the original fecal concentration. This approach effectively mitigated matrix effects to an acceptable level (Figure S3) while maintaining sufficient analytical sensitivity for quantification. Compared to a reported DAS method for chicken feces involving an 80-fold dilution [13], our method employs a lower dilution factor and yields a cleaner extract, making it more suitable for analyzing the extremely complex donkey fecal matrix.

2.3.3. Plasma Sample

Toxicokinetic studies indicate that FB₁ is poorly absorbed and rapidly metabolized, resulting in very low residual concentrations in plasma [32], necessitating a method with high sensitivity. When the SALLE procedure optimized for urine was applied to plasma, the recovery for most analytes was below 80% (Figure S4). This suppression is likely attributable to interactions (e.g., hydrogen bonding) between the carboxyl, hydroxyl, and primary amine groups on the analytes and denatured plasma proteins, hindering their partitioning. Interestingly, a simpler DES approach, omitting the salting-out step, provided superior recovery. While acidification remained crucial for efficient FBs extraction, the required acid amount was substantially lower than for urine. As shown in Figure 6, a 1% AA content in the extraction solvent (ACN, equivalent to ~3% v/v of the plasma sample) was sufficient to achieve satisfactory recovery (>80%) for FBs, especially FB₁. Overall, reported extraction recoveries for FBs from biological matrices tend to be lower compared to other mycotoxins. For instance, a DES method for multiple mycotoxins in dried blood spots reported an average recovery of only 61% for FB₁ without acidification, while recoveries for other mycotoxins exceeded 80% [33]. Another study reported recoveries around 60% for FB₁ and FB₂ from rat plasma using 1% AA in ACN or MeOH [30], suggesting species-dependent variations. Although one study using SALLE for pig plasma reported an internal standard-corrected recovery of 96.1% for FB₁ without acidification, the absolute recovery was not provided [25]. Another method for chicken plasma involved protein precipitation with 1% AA in ACN followed by a phospholipid-removal plate, with a final 4-fold dilution [10]. In contrast, our DES method incorporates a 4-fold concentration of target analytes from plasma, and no significant interference from phospholipids was observed during LC-MS/MS analysis (Figure S5).

2.3.4. Purification

Solid-phase sorbents such as C₁₈, PSA, and NANO are commonly employed in sample pretreatment to remove interfering matrix components, mitigate matrix effects, and enhance sensitivity [3,20,34]. However, due to the presence of both carboxyl and hydroxyl groups in their molecular structures, FBs exhibit significant affinity for PSA, rendering it unsuitable as a purification material for these analytes [20]. This study further evaluated the purification efficiency of C₁₈ and NANO for the organic phase obtained after salting-out from donkey urine and fecal samples. The results demonstrated that C₁₈ treatment did not substantially improve either the visual appearance of the samples or the associated matrix effects. Although NANO effectively adsorbed pigments (Figure S6), it also strongly retained hydrolyzed metabolites (HFBs), leading to a marked reduction in HFBs recovery (Figure S7), while offering minimal alleviation of matrix effects (Figure S8). These observations may be attributed to the fact that, during subsequent chromatographic separation, the target analytes are already effectively resolved from those interfering matrix components that would otherwise be retained by C₁₈ or NANO. Moreover, introducing an additional purification step would significantly increase method complexity and consume more reagents, labor, and time. Consequently, the present method ultimately omitted any solid-phase purification procedure. Instead, an internal standard calibration was applied to correct the quantification of FBs, and matrix-matched standard curves were employed for the accurate determination of HFBs.

2.3.5. Results of Method Validation

The established method was subjected to comprehensive in-house validation, with the key performance parameters summarized in

Table 2. Overview of the spiked concentration (Conc.), mean recovery (Re), repeatability (RSD_r), and intermediate precision over 3 days (RSD_R) in the three matrices used for method validation.

Analyte	Cocn.	Plasma			Cocn.	Urine			Cocn.	Feces
	(μg/L)	Re (%)	RSDr (%)	RSDR (%)	(μg/L)	Re (%)	RSDr (%)	RSDR (%)	(μg/kg)	Re (%)	RSDr (%)	RSDR (%)
FB1	0.5	96.8	14.7	14.7	2	97.2	13.3	4.3	300	94.3	9.3	2.8
	5	96.3	6.3	4.4	10	97.5	14.0	14.5	750	105.6	14.4	8.3
	20	93.8	7.7	2.8	50	103.8	7.1	15.0	1500	109.0	12.3	2.9
FB2	0.5	96.9	1.5	3.1	2	96.8	5.8	5.5	300	110.8	5.4	9.2
	5	96.1	4.2	2.4	10	95.6	12.3	3.3	750	94.1	12.4	3.5
	20	93.9	9.7	5.8	50	99.0	9.0	13.3	1500	109.5	15.8	2.9
FB3	0.5	100.5	4.3	13.5	2	105.0	15.1	5.8	300	95.6	14.3	9.1
	5	86.7	16.6	12.1	10	104.3	4.4	14.5	750	111.5	13.8	6.0
	20	95.4	3.3	12.6	50	103.3	7.0	9.5	1500	106.7	8.7	7.7
HFB1	0.5	80.7	14.5	9.5	2	84.5	6.2	3.6	300	87.7	7.0	8.1
	5	85.6	1.9	6.5	10	84.8	2.4	6.7	750	80.8	5.8	7.5
	20	74.8	14.5	9.7	50	81.4	10.9	9.1	1500	89.9	4.9	3.8
HFB2	0.5	81.4	7.7	9.2	2	82.3	7.2	5.8	300	74.4	11.9	4.2
	5	79.1	13.1	14.6	10	88.1	11.2	11.0	750	78.3	7.7	14.4
	20	92.1	15.8	4.3	50	77.8	5.3	3.3	1500	74.9	16.0	12.7
HFB3	0.5	81.3	3.8	4.6	2	81.6	1.9	11.2	300	76.9	2.8	10.9
	5	76.0	11.3	6.2	10	73.3	11.8	8.1	750	75.9	7.1	14.1
	20	81.4	3.8	7.3	50	89.8	8.2	2.7	1500	79.8	5.6	11.3

and Table 2 and Table S1. Preliminary tests indicated peak saturation when the concentrations of FBs and HFBs in the injection solution exceeded 200 μg/L and 100 μg/L, respectively. Therefore, a series of standard working solutions with FB concentrations of 0.1, 0.5, 2, 10, 20, 50, 100, and 200 μg/L, HFB concentrations of 0.1, 0.5, 2, 10, 20, 50, and 100 μg/L, and internal standards (ISs) each at a concentration of 8 μg/L. LC-MS/MS analysis demonstrated excellent linearity (R² ≥ 0.99) for all six analytes across their respective calibration ranges. Since animal feeds are commonly contaminated with FBs [17,18,19] but exhibit low absorption and rapid metabolism in animals, leading to minimal residues in plasma and urine with predominant excretion via feces [20], method sensitivity requirements differ by matrix. Consequently, the required limits of detection (LOD) for plasma and urine were pushed as low as possible, whereas for feces, they were set to align with realistic exposure levels, considering the concentrating effect during digestion, which can lead to fecal concentrations exceeding those in feed. Specificity was confirmed in blank donkey plasma and urine, where no endogenous interference was observed at the retention times (RTs) of the target analytes. As a true blank fecal sample was unavailable (the lowest FB₁ concentration found in all tested samples was 375 μg/kg dry matter), specificity for feces was assessed using the standard addition method. Spiking feces with 500 μg/kg of each analyte yielded chromatographic characteristics (RT, ion ratio, peak shape, separation) identical to those in solvent standards, confirming the absence of co-eluting interferences. Method accuracy and precision were evaluated across all three matrices by analyzing samples spiked at three concentration levels, with six replicate samples (n = 6) per concentration per matrix (Table 2). The mean recoveries ranged from 73.3% to 111.5%, with both intra-day (RSD_r) and inter-day (RSD_R) precision below 20%. The occasional recoveries slightly exceeding 100%, particularly for the FBs in feces, are likely a combined result of inherent matrix complexity and background interference. Despite the application of isotope-labeled internal standard correction for FBs, residual matrix-induced signal enhancement, possibly co-eluting with the analytes, could contribute to this observation. The limits of detection (LOD) and quantification (LOQ) were defined as the analyte concentrations yielding signal-to-noise (S/N) ratios of 3 and 10, respectively, based on the least sensitive transition for each compound (see Table S1 for details). Specifically, the LOQs for FBs were 0.1 μg/L in plasma, 1.0 μg/L in urine, and 60 μg/kg in feces, respectively. Correspondingly, the LOQs for HFBs were 0.15 μg/L in plasma, 1.0 μg/L in urine, and 60 μg/kg in feces. The sensitivity achieved with the present method is comparable to that of other reported methods using similar cleanup approaches [1,7,9,12], though lower than those utilizing immunoaffinity or dedicated SPE columns [5]. The key advantage of our method is its speed and cost-effectiveness, as it avoids the need for specialized purification columns. However, for samples with highly complex matrices, this simplified preparation may need further optimization to manage stronger matrix interference. Nevertheless, to the best of our knowledge, this is the first method capable of simultaneously determining this comprehensive panel of FBs and HFBs across donkey plasma, urine, and feces, thereby meeting the analytical demands for pharmacokinetic studies and exposure assessments in this species.

3. Conclusions

In conclusion, a robust and sensitive LC-MS/MS method was successfully developed and validated for the simultaneous determination of three major fumonisins (FB₁, FB₂, FB₃) and their hydrolyzed metabolites (HFB₁, HFB₂, HFB₃) in the complex biological matrices of donkey plasma, urine, and feces. The method employed matrix-specific sample preparation strategies—SALLE for urine/feces and DES for plasma—to effectively overcome matrix interferences while maintaining high throughput. Key validation parameters, including linearity, accuracy, precision, and sensitivity, met or exceeded the criteria set by relevant international guidelines. The achieved limits of quantification are adequate to detect anticipated exposure levels, supporting its application in real-world studies. By enabling comprehensive multi-matrix analysis, this method fills a significant methodological void and provides a reliable analytical foundation for conducting crucial in vivo toxicokinetic and exposure assessment studies of fumonisins in donkeys. Importantly, the core analytical principles and matrix-tailored sample preparation strategies (e.g., SALLE for complex fluids, DES for plasma) could provide a valuable starting point for method adaptation in other species. Therefore, this work not only addresses an immediate need in donkey research but also offers a methodological framework and reference for investigating fumonisin metabolism and exposure in other livestock and animal species, contributing to broader toxicological risk assessment efforts.

4. Materials and Methods

4.1. Chemicals and Reagents

HPLC-grade methanol (MeOH), acetonitrile (ACN), ammonium hydroxide (NH₄OH), acetic acid (AA), formic acid (FA), and ammonium acetate (NH₄OAc) were supplied by Fisher Scientific (Waltham, MA, USA). Analytical-grade perchloric acid (HClO₄), sodium chloride (NaCl), and anhydrous magnesium sulfate (MgSO₄) were obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Deionized water (18.2 MΩ·cm) was produced using a Milli-Q water purification system (Millipore Corp., Bedford, MA, USA). Clean-up sorbents, including C₁₈, primary secondary amine (PSA), and NANO materials, were provided by Bonna-Agela Technologies (Tianjin, China).

Mixed standard solutions of FBs (FB₁, FB₂, FB₃) and their hydrolyzed metabolites (HFB₁, HFB₂, HFB₃), each with certified concentrations, were purchased from Alta Scientific Co., Ltd. (Tianjin, China). Isotopically labeled internal standards ([¹³C₃₄]-FB₁, [¹³C₃₄]-FB₂, and [¹³C₃₄]-FB₃) were acquired from Romer Labs (Tulln, Austria). Two mixed internal standard working solutions were prepared in 50% (v/v) ACN-water: a 1 mg/L solution (for each labeled analyte) used for spiking urine and fecal samples, and a 0.1 mg/L solution (for each labeled analyte) used for spiking plasma samples. All standard solutions were stored at 4 °C before use.

Blank plasma, urine, and fecal samples used in this study were collected from six Dezhou donkeys. These samples were kindly provided by the Equine Research Center, College of Animal Science and Technology, China Agricultural University.

4.2. Instruments

The analysis was conducted with an ultra-high-performance liquid chromatography system (ACQUITY UPLC I-Class, Waters, Milford, MA, USA) interfaced with a triple quadrupole tandem mass spectrometer (Xevo TQ-S, Waters) featuring an electrospray ionization source. MassLynx software (version 4.1, Waters) controlled the UPLC-MS/MS system for both data acquisition and processing. The equipment utilized for sample preparation included an ultrasonic homogenizer (KQ-500DE, Kunshan Ultrasonic Instruments Co., Kunshan, China), a vortex mixer (GENIUS 3, IKA, Staufen Vajra, Germany), an automated solvent evaporation workstation (Vajra 1000, Beijing Giam Instrument Co., Ltd., Beijing, China), a high-speed refrigerated centrifuge (3K15, Sigma, Osterode am Harz, Germany), and a microcentrifuge (CF-10, Wisd, Wonju, Republic of Korea).

4.3. Sample Preparation

The sample preparation procedures were optimized for plasma, urine, and feces, respectively, as detailed below.

For plasma samples, a 1 mL aliquot was spiked with 20 μL of the mixed internal standard working solution (0.1 mg/L for each analyte). This addition introduced a nominal concentration of 2 μg/L for each internal standard in the plasma sample. After vortex mixing, proteins were precipitated with 3 mL of 1% AA in ACN (1% AA-ACN). The mixture was vortexed for 1 min, subjected to ultrasonic extraction for 10 min, and then centrifuged at 8000 rpm for 5 min. The entire supernatant was collected and evaporated to dryness using an automated solvent evaporation workstation at 50 °C. The residue was reconstituted in 250 μL of a 0.1% FA-10% MeOH solution (0.1:10:89.1, v/v/v, FA/MeOH/water). After centrifugation at 10,000 rpm for 2 min, the supernatant was transferred to a conical-bottom autosampler vial for LC-MS/MS analysis. Thus, the sample preparation procedure yielded a 4-fold concentration of the analytes in the final extract.

For urine samples, a 1 mL aliquot was fortified with 8 μL of the mixed internal standard working solution (1 mg/L for each analyte). This resulted in a nominal concentration of 8 μg/L for each internal standard in the urine sample. Subsequently, 100 μL of 35% (v/v) aqueous perchloric acid solution (PCA solution) and 3 mL of ACN were added for protein precipitation and liquid–liquid extraction. The mixture was vortexed, ultrasonically extracted for 10 min, and then salted out with 300 mg of NaCl. After centrifugation, a 2 mL aliquot of the supernatant was dehydrated with 100 mg of MgSO₄. Then, 900 μL of the purified supernatant was taken, dried at 50 °C, and reconstituted in 300 μL of the 0.1% FA-10% MeOH solution. The final supernatant was centrifuged and analyzed. This process maintained the original concentration of analytes without dilution or concentration.

For fecal samples, the specimens were first freeze-dried, pulverized, and sieved through a 40-mesh screen. An aliquot of 0.2 g of the homogenized powder was weighed and spiked with 24 μL of the mixed internal standard working solution (1 mg/L for each analyte). This spike corresponded to a nominal concentration of 120 μg/kg for each internal standard in the solid fecal sample. The sample was then extracted with 1 mL of deionized water, 100 μL of 35% PCA solution, and 3 mL of ACN. The subsequent steps, including vortexing, ultrasonic extraction, salting-out with 300 mg NaCl, centrifugation, and dehydration of 2 mL supernatant with 100 mg MgSO₄, were performed similarly to the urine protocol. Finally, 500 μL of the treated supernatant was dried, reconstituted in 500 μL of the reconstitution solution, and centrifuged. The overall sample preparation resulted in a 15-fold dilution of the original fecal sample.

4.4. UPLC-MS/MS Conditions

Chromatographic separation was achieved using an ACQUITY BEH C₁₈ column (100 mm × 2.1 mm, 1.7 µm, Waters). The column temperature was maintained at 40 °C. The mobile phase consisted of (A) 0.5% FA in water and (B) 0.5% FA in ACN. A binary gradient elution program was employed at a flow rate of 0.2 mL/min as follows: 95% A (0–1 min), linearly decreased to 70% A (1–2 min), then to 5% A (2–7 min), held at 5% A (7–8 min), returned to 95% A (8–8.5 min), and finally re-equilibrated at 95% A until 10 min. The injection volume was 5 µL.

Mass spectrometric detection was performed in the positive electrospray ionization (ESI+) mode with multiple reaction monitoring (MRM). The ion source parameters were set as follows: capillary voltage, 0.6 kV; source temperature, 150 °C; desolvation temperature, 450 °C. The desolvation gas (nitrogen) flow was 800 L/h, and the cone and nebulizer gas flows were both set at 40 L/h. The optimized MRM transitions, corresponding cone voltages, and collision energies for all target analytes and internal standards are detailed in Table 1.

4.5. Method Validation

The method was comprehensively validated for donkey plasma, urine, and feces according to the GB 5009.295-2023 [35] and SANTE/11312/2021 guidelines [22], assessing the parameters of selectivity, linearity, sensitivity, matrix effect, accuracy, precision, and carry-over. Selectivity was evaluated by analyzing blank samples from six individual animals to confirm the absence of interfering signals at the retention times of the analytes and internal standards. Given the identical mass spectrometric transitions of the isomeric pairs FB₂/FB₃ and HFB₂/HFB₃, their selectivity was assessed by verifying baseline chromatographic separation. Linearity was established across a concentration range of 0.1 to 200 ng/mL. Sensitivity was determined by establishing the limit of quantification (LOQ) based on the lowest spiked level that provided a signal-to-noise ratio ≥ 10 and could simultaneously meet the predefined criteria for accuracy and precision. The matrix effect (ME) was quantified for each matrix by comparing the analyte peak areas in post-extracted blank samples to those in pure solvent in three replicates. The ME was calculated using the formula ME (%) = [(Peak area in matrix/Peak area in solvent) − 1] × 100, where values between −20% and +20% were considered acceptable according to SANTE guidelines [22]. The matrix-corrected recovery was also calculated to evaluate the efficiency of the sample preparation procedure. It was determined by comparing the peak areas of analytes spiked into blank matrix before extraction with those spiked into the post-extracted blank matrix after extraction (matrix-matched standards), according to the following formula: Matrix-corrected recovery (%) = (Peak area of pre-spiked sample/Peak area of post-spiked sample) × 100. Accuracy and precision were assessed through recovery experiments at three concentration levels (low, medium, high) with six replicates per level, analyzed both within a single day and over three separate days. Finally, carry-over was checked by injecting a blank solvent sample immediately after the highest concentration calibration standard.