In many matrices, ochratoxin A is present at low levels (µg/kg) making precise and accurate analysis difficult. To add to this challenge, OTA is known to be heterogeneously dispersed in many bulk commodities such as raw grains [26, 27]. Although steps can be taken to mitigate heterogeneity, such as in the preparation of the CRM MYCO-1, it cannot be completely eliminated [23]. Therefore, the results obtained for each quantitation method described herein were calculated using this same set of extracted samples to eliminate any homogeneity issues as a potential source for any observed differences allowing for the direct comparison of each calibration strategy.

External standard calibration

Official AOAC methods and European Standards exist for the quantitation of ochratoxin A in a variety of different matrices including coffee, barley, wine, and beer [28,29,30,31,32,33,34]. These methods use liquid chromatography with fluorescence detection (LC-FLD) and require preparation of a calibration curve with ochratoxin A standard solutions [25, 35, 36]. These external standard calibration methods require complete extraction of the analyte or the results must be recovery-corrected. The percent recovery must be determined from an independent experiment typically using a blank matrix, which as previously mentioned is not always available. Additionally, a clean-up step is usually required such as an immunoaffinity separation. Use of LC-MS methods compared to LC-FLD increases sensitivity and selectivity, proving valuable for OTA analysis at trace levels. However, matrix interferences can be problematic, and for external calibration methods, sample clean-up may still be required [37, 38].

The mass fraction of OTA was calculated for all flour samples using external standard calibration. Preliminary work to determine the appropriate calibration range was performed by comparing the CWRS and CWAD samples to MYCO-1. The calibration curve was prepared by gravimetrically diluting the native standard solution containing the CRM OTAN-1 in the extraction solvent, 85% acetonitrile/water (by volume), to obtain a concentration range of 0.1–2.5 ng/g. Results are therefore traceable through OTAN-1. All sample extracts and calibration standard solutions (Cals) were analyzed as is, with no sample clean-up, by LC-HRMS to determine peak areas. The calibration curve was fitted using a linear function and the resulting equation solved for each sample to determine the mass fraction of OTA. As shown in Fig. 1a, the peak area for each sample fell toward the bottom of the calibration curve. The mass fraction of OTA in quality control samples of MYCO-1 was 18–38% lower than the certified value of 4.05 ± 0.88 µg/kg, k = 2 (MYCO-1 samples A–C in Table 1). Two out of the three samples even fell outside the certified uncertainty range with values of 2.91 and 2.50 µg/kg for the mass fraction of OTA by external calibration. As shown in Fig. 1b, an overlay of the chromatograms for calibration standard solution 2 (Cal-2, 1.2 ng/g) and MYCO-1 extract (1.3 ng/g) reveals significant ion suppression in the extracted sample. Evidence of ion suppression, combined with the poor results obtained for quality control samples of MYCO-1, indicates external calibration is not an ideal method for the quantitation of OTA in flour. It is therefore unlikely that the OTA values in CWRS and CWAD samples (Table 1) are an accurate representation of the true value.

Fig. 1

a External standard calibration curve with results of each sample (average of triplicate extractions). The error bars represent the standard deviation of the triplicate extractions and reflect the heterogeneity of the samples. All samples shown on the same calibration curve for simplicity. b LC-HRMS chromatograms of OTA in calibration standard solution 2 (Cal-2, 1.2 ng/g) and MYCO-1 extract (1.3 ng/g), revealing ion suppression

Table 1 The mass fraction of OTA in various flour samples obtained by external standard calibration, based on the average of triplicate extractions (N.D., not detected). The uncertainty associated with the mass fraction (u) is the average measurement uncertainty which includes contributions from the CRM OTAN-1 and triplicate analysis of the same extractIsotope dilution methods

Isotope dilution methods include the addition of an isotopically enriched compound to the sample to counteract matrix effects. Traceable quantitation can then be achieved through the natural reference standard with use of an internal standard as a control or it can be obtained directly through the internal standard, if the latter is a reference material or a CRM. Numerous labelled derivatives are available for mycotoxins; however, their concentration is not always known with confidence and they lack traceability. OTAN-1 and OTAL-1 are both CRMs produced in accordance with ISO 17034:2016 [39] with a certified mass fraction of OTA (11.03 ± 0.32 µg/g, k =2) or [13C6]-OTA (4.91 ± 0.16 µg/g, k = 2) respectively [40, 41]. They can be used for the quantitation of OTA in a wide range of matrices by employing isotope dilution strategies. An advantage of isotope dilution over external standard calibration is that complete analyte extraction may not be required, but rather complete equilibration between the analyte and internal standard must be achieved [42]. If the internal standard is added prior to extraction and complete isotopic equilibrium is reached, it is assumed that any loss of analyte will be compensated for even if the extraction is incomplete [16]. Additionally, as ratios of native analyte peak area to stable isotope-labelled internal standard peak area are used instead of absolute peak areas, solvent evaporation of samples is no longer an utmost concern. Calibration standard solutions (Cals) can then be analyzed multiple times, even days later, provided the analyte is stable.

Single isotope dilution mass spectrometry (ID1MS)

When quantitation is achieved directly through the internal standard, this is known as single isotope dilution mass spectrometry (ID1MS). It requires only an isotopically enriched standard to act as the primary calibrator, providing an advantage over higher order isotope dilution (IDnMS) methods due to its simplicity. Additionally, only a single sample is required, reducing sample preparation and analysis time as well as eliminating the need for any calibration standard solutions (Fig. 2a). To illustrate its effectiveness, the mass fraction of OTA in CWRS and CWAD wheat samples, as well as quality control samples of MYCO-1, was determined by ID1MS using OTAL-1 as the internal calibrant. As shown in Fig. 2b, both unlabelled OTA and [13C6]-OTA have similar elution and ionization profiles by LC-MS. Therefore, OTAL-1 is expected to reliably provide very accurate measurement values for OTA when used as a calibrant.

Fig. 2

a Experimental scheme for ID1MS (sample) and ID2MS (sample and calibration standard solution). OTA (A*) comes from the native working solution of diluted OTAN-1 and [13C6]-OTA (B) comes from the internal standard solution of diluted OTAL-1. b Unlabelled OTA and [13C6]-OTA exhibit similar elution and ionization profiles by LC-MS

Each sample was spiked with known amounts of the internal standard solution containing the CRM OTAL-1 and analyzed by LC-HRMS. Spectral windows centered at the exact monoisotopic masses of 404.08954 Da for OTA and 410.10967 Da for [13C6]-OTA were extracted and the ratio of peak areas of the two isotopic forms of ochratoxin A was measured in each sample (unlabelled:labelled). The mass fraction of OTA (wA) in each sample was calculated as follows:

$$_}=_}\bullet \frac_}-_}}_}-_}}\bullet \frac_\left(\mathrm\right)}}_\left(\mathrm\right)}}\bullet \frac_}_}$$

(1)

where wB is the mass fraction of [13C6]-OTA in the internal standard solution, rA is the isotope ratio of OTA in the sample, rB is the isotope ratio of [13C6]-OTA in the internal standard solution, rAB is the isotope ratio in the sample extract (AB), mB(AB) is the mass of [13C6]-OTA in the sample, mA(AB) is the mass of flour in the sample, MA is the molar mass of OTA, and MB is the molar mass of [13C6]-OTA. Results for quality control samples of MYCO-1 were in agreement with the certified value of OTA, 4.05 ± 0.88 µg/kg, k = 2 (Table 2).

Table 2 The mass fraction of OTA in various flour samples obtained by each quantitation method, based on the average of triplicate extractions (N.D., not detected). The uncertainty (u) is the average measurement uncertainty which includes contributions from the CRMs OTAN-1 or OTAL-1Double isotope dilution mass spectrometry (ID2MS)

As previously mentioned, if the exact concentration of the labelled compound is unknown or lacks the appropriate traceability, quantitation cannot be obtained by ID1MS. Alternatively, single-point calibration with internal standard known as double isotope dilution mass spectrometry (ID2MS) can be used. With this method, OTAN-1 and OTAL-1 serve as a primary calibrator and internal standard respectively, and wB (the mass fraction of OTAL-1) is no longer required. Samples were prepared, same as those for ID1MS, consisting of flour spiked with known amounts of the internal working solution containing the CRM OTAL-1 ([13C6]-OTA). Additionally, a calibration standard solution containing a known amount of the native standard solution (unlabelled OTA from OTAN-1) and internal standard solution ([13C6]-OTA from OTAL-1) was prepared (Fig. 2). The ratio of OTA to [13C6]-OTA in each sample extract and calibration standard solution (Cal) was measured using LC-HRMS and the mass fraction of ochratoxin A was then calculated according to the standard ID2MS equation [18] as follows:

$$_}=_*}\bullet \frac_*}-_}^\mathrm}}_}^\mathrm}-_}}\bullet \frac_}-_}}_}-_}}\bullet \frac_}^\left(}^\mathrm\right)}}_\left(}^\mathrm\right)}}\bullet \frac_\left(\mathrm\right)}}_\left(\mathrm\right)}}$$

(2)

where wA* is the mass fraction of OTA in the native standard solution, rA is the isotope ratio of OTA in the sample, rA* is the isotope ratio of OTA in the native standard solution (we assume rA = rA*), rB is the isotope ratio of [13C6]-OTA in the internal standard solution, rA*B is the isotope ratio in the calibration standard solution (A*B), rAB is the isotope ratio in the sample (AB), mA*(A*B) is the mass of OTA in the calibration standard solution, mB(A*B) is the mass of [13C6]-OTA in the calibration standard solution, mB(AB) is the mass of [13C6]-OTA in the sample, and mA(AB) is the mass of flour in the sample. The mass fraction of OTA for each sample was calculated using the calibration standard solution (Cal) which had an isotope ratio (rA*B) closest to that of the sample extract (rAB). Again, the results for quality control samples of MYCO-1 fell within the certified range (Table 2).

A similar method, known as exact-matching ID2MS, was used to certify the mass fraction of OTA in MYCO-1 [22, 23]. For exact-matching, the concentration of the unlabelled reference standard in the calibration standard solution (Cal) should match the estimated mass fraction of the analyte in the sample; then, both are spiked with equal amounts of internal standard (mB(A*B) = mB(AB)). Additionally, both sample and calibration standard solution are prepared such that the signal ratio of unlabelled to labelled is as close to 1:1 as possible (rA*B = rAB = 1). This method minimizes mass bias effects ultimately providing a more accurate result and reducing overall uncertainties [19]. Exact-matching ID2MS can prove challenging as initial knowledge about the concentration of analyte is necessary. Preliminary work is required to first estimate the mass fraction of analyte in each sample. Once known, the concentration of the unlabelled reference standard in the calibration standard solution can be determined and the ideal ratio (1:1) can be fully optimized. Additionally, as each sample will likely differ in the amount of analyte present, an independent calibration standard solution would need to be prepared for each sample. Although this approach can yield highly accurate results, it is very labor intensive and thus unsuitable for high-throughput analysis. Therefore, exact-matching ID2MS was not used to quantitate OTA in CWRS and CWAD samples.

Higher order isotope dilution mass spectrometry (IDnMS)

An alternate approach, ideal for measuring samples over a wider range of concentrations, employs multiple calibration standard solutions (Cals) of the unlabelled reference standard and internal standard across a calibration range (Fig. 3a). Depending on the number of calibration standard solutions prepared, this method can be referred to as triple (ID3MS, 2 calibration standard solutions), quadruple (ID4MS, 3 calibration standard solutions), and quintuple (ID5MS, 4 calibration standard solutions) isotope dilution mass spectrometry and so forth. The resulting mass fraction can be determined using a specific isotope dilution equation [18] or a calibration curve can be fitted similar to external standard calibration. However, rather than plotting peak areas, ratios of peak areas of native to labelled isotopes are used instead to mitigate matrix effects.

Fig. 3

a Experimental scheme for quintuple isotope dilution mass spectrometry (ID5MS) including four calibration standard solutions (Cal-1 to Cal-4) with different isotope ratios shown to bracket the isotope ratio in the sample (AB). b Isotope ratio calibration curve with results of each sample (average of triplicate extractions). The error bars represent the standard deviation of the triplicate extractions and reflect the heterogeneity of the samples. All samples shown on the same calibration curve for simplicity

For the quantitation of OTA in the CWAD and CWRS samples, the ID5MS method was used. Four calibration standard solutions (Cals) were gravimetrically prepared using varying amounts of the native standard solution, containing the CRM OTAN-1, and a similar amount of internal standard solution, containing the CRM OTAL-1, to obtain isotope ratios of OTA/[13C6]-OTA at ~0.08:1, 0.2:1, 1:1, and 2:1. The internal standard solution added to the calibration standard solutions was equal to the amount spiked in the sample extracts. All samples were analyzed by LC-HRMS and using the peak area ratio of OTA to [13C6]-OTA, a calibration curve was fitted using a linear function (Fig. 3b). While a linear model is commonly applied for routine use and adequate in many cases, we note that the theoretical shape of the calibration curve is inherently nonlinear [21]. The calibration curve equation was solved for each sample with the mass fraction of OTA in quality control samples of MYCO-1 falling within the certified range (Table 2). This approach can be beneficial for samples where the levels of OTA are unknown, as well as when samples contain a wide range of differing amounts of analyte. Although this method may require multiple calibration standard solutions, the calibration curve range can be tailored to include as many or as few points as desired. If the isotope ratio in a given sample extract falls outside of the calibration range, additional calibration standard solutions can be added and the samples re-analyzed. Provided the analyte is stable, a single calibration curve can be re-analyzed multiple times allowing the analysis of a large number of samples, maximizing throughput.

Method comparison

The final mass fractions of OTA obtained from each calibration method, external standard calibration, and single (ID1MS), double (ID2MS), and quintuple (ID5MS) isotope dilution mass spectrometry are shown in Fig. 4. For external calibration, the mass fraction of OTA obtained for each sample was much lower than those obtained by isotope dilution methods. It is evident from the results of quality control samples of MYCO-1 that isotope dilution strategies provide a more accurate quantitation than external calibration. Figure 4 also confirmed that ID2MS and ID5MS produce equivalent results. Moreover, the accuracy of these methods was confirmed, as they produced expected mass fractions for OTA in the quality control sample MYCO-1. However, the mass fractions of OTA obtained by ID1MS were on average 6% lower than those obtained by ID2MS and ID5MS for all samples (Table 2 and Fig. 4). This difference was attributed to isotopic enrichment bias [43,44,45] of the [13C6]-OTA relative to native OTA.

Fig. 4

Mass fraction of OTA obtained for each quantitation method (average of triplicate extractions with an average measurement uncertainty). The solid line represents the certified value for OTA in MYCO-1 and the dotted lines represent the expanded uncertainty

Since the bias appeared systemic and ID2MS and ID5MS results were deemed accurate, gravimetric or other experimental sources of error were discarded. Instead, the bias originates from the fundamental principle of calculating isotopic abundance in high-resolution mass spectrometry based on the distribution of the monoisotopic mass (M) alone. The percent abundances of the monoisotopic mass (M) are not constant, but rather vary among unlabelled and their labelled counterparts. As shown in Fig. 5, the monoisotopic signal for unlabelled and [13C6]-OTA compromises 59.9% and 63.9% of their respective isotopic patterns. The isotope ratio (unlabelled:labelled) measured in the sample extracts is from the monoisotopic ions, specifically m/z 410.10967 for [13C6]-OTA which accounts for 63.9% of the isotopologues and m/z 404.08954 for native OTA given by 59.9% of the isotopologues. Therefore, the ID1MS results are expected to be biased by 6%, i.e., by a factor of 59.9/63.9.

Fig. 5

Isotopic pattern of native OTA, [13C6]-OTA, and [13C20]-OTA. The percent contribution of each isotopologue toward the full isotopic envelope is highlighted

It should be noted that this bias would be even more significant in the case of commonly used fully labelled [13C20]-OTA, where an increased monoisotopic contribution of 74.2% (Fig. 5) would result in a measurement bias of approximately 19% if a simple ID1MS calibration approach was employed. Additionally, it should be pointed out that the isotopic enrichment does not impact IDnMS. This is because the ratio of labelled to unlabelled OTA was measured in both the sample and calibration standard solutions separately and these two values cancel out in Eq. 2. The overall effect is that the direct measurement of the monoisotopic ion of the labelled internal standard ([13C6]-OTA) cancelled out and had no bearing in the final calculations. Therefore, when ID1MS is chosen as the quantitation method, one must be aware of the limitations of using the internal standard as the calibrator.