Tips on the analysis of pesticides and mycotoxins in cannabis products: Matrix matters! (Part II)

In part I of this blog series, we learned about matrix effects, how to assess them, and how to address them. One of the main conclusions of part I was the importance of using matrix matched calibration as the means to account for possible ionization effects in LC-MS/MS, and to minimize bias in GC-MS/MS due to chromatographic response enhancement. In part II, I want to talk about important points to keep in mind when evaluating recoveries and performing a matrix matched calibration. The first aspect, which was discussed previously in part I, is to ensure that your surrogate matrix reflects the composition of your sample matrix. Additionally, this surrogate matrix should be a blank (or free of your target analytes, in our case pesticides and mycotoxins). Once you source your surrogate matrix and decide on the sample prep conditions, it is very important to investigate your analytes recoveries.

2. Spiking procedure and recoveries

Assessment of recoveries (% of analyte extracted/total amount of analyte spiked) is critical to ensure reliable quantitative data in typical pesticide analysis methodologies. This is important because matrix-matched calibrations are commonly prepared by post-spiking blank extracts with target analytes and internal standards at different concentration levels using the same dilution factor as used in real samples/extracts. A key assumption made when preparing the calibration curve is that the extraction efficiency is assumed to be 100%. Unfortunately, as we already showed in our technical article, there are some cases (e.g. daminozide in brownies) where getting close to exhaustive recoveries (close to 100% recoveries) is very difficult. Hence, the use of the right internal standard can be very critical to account for those variations. Alternatively, calibrators can be prepared by spiking matrix blanks at different concentration levels and running them through the entire extraction process to construct the calibration curve. I know that this can be tedious, especially when dealing with multiple matrices and many different pesticides, but if you are having difficulties with a particular cannabis matrix, this is in an option that you may want to consider. Undoubtedly, this approach in combination with the use of isotopically labeled standards will give you the best results in terms of accuracy and precision.

Typically, recoveries are tested by spiking the surrogate matrix with the analytes of interest and then performing your extraction procedure. Then you compare the amount of analyte recovered to analyte spiked in post-extraction blank assuming 100% extraction efficiency. How do we effectively use this approach? First, you need to pick the concentrations at which you want to test your method. Testing your method at a low (close to LOQ), medium, and a high concentration is a good idea. In our brownies workflow, we chose a concentration of 100 ng/g to test our methodology. This concentration was chosen because it is the lowest action level regulated for some of the California pesticides. Then, you need to spike your analytes in the surrogate matrix. Here I would like to bring a very interesting finding to your attention. A chemist from a cannabis testing lab shared with me that he first adds the extraction solvent to his sample, and subsequently spikes the samples with target analytes. In my case, I normally spike my analytes in the blank matrix, wait for the analytes to equilibrate with the sample, and then proceed to perform the extraction. In order to check whether the order of addition impacts the results, we evaluated the effect of spiking our target analytes (California list of pesticides and mycotoxins) before and after adding the extraction solvent. I know that this experiment may sound trivial, but please don’t forget that the devil is in the details. This experiment was performed in two matrices: brownies and dark chocolate. Although both matrices are delicious and have chocolate as one of their main ingredients, their compositions are very different. Figure 1 summarizes the results of the comparison for the three most impacted analytes.

 

Figure 1. Relative responses obtained in A) brownies and B) dark chocolate after comparing the effect of spiking our target analytes before and after adding the extraction solvent. Responses were normalized by the results obtained when spiking first our analytes and then adding solvent (n=3). Brownies samples were prepared as described in our technical article. Dark chocolate samples (0.5 g) were extracted by using isopropyl alcohol (0.5 mL) and acetonitrile acidified with acetic acid at 1% (2.5 mL); subsequently, 2 mL of the extract were passed through a Restek Resprep C18 cartridge (Restek Cat.#26030).

 

To better visualize our data, the responses (area counts) were normalized according to the response obtained for the samples where analytes were spiked first and solvent was added later. Interestingly, the effect depends on the compound and on the matrix type (and of course, it also depends on your extraction conditions). Daminozide (I guess that at this point you may think that it is our favorite analyte) displays 3-fold better recoveries when it is spiked in brownies after adding the extraction solvent vs. when it is spiked in the dry matrix. Conversely, in the case of dark chocolate, there is only a 10% difference between spiking daminozide before vs. after the extraction solvent. The main reason for this difference is that daminozide, a highly polar pesticide (logP=-1.5), displays a much higher affinity for a matrix like brownie compared to a more fatty sample such as dark chocolate. Such affinity can be modified when solvent is added to the matrix, meaning that the recovery of a surrogate matrix spiked by adding solvent and then the analytes won’t always be representative of a real sample. In the case of acephate, another polar pesticide (logP=-0.85), the spiking order results in a 17% difference in brownie whereas in dark chocolate the difference is negligible. Ochratoxin A, on the other hand, showed differences of 39% and 24% in brownies and chocolate, respectively. Based on these results, it is clear that the safer way to assess recoveries (also method accuracy and precision, which will be discussed in part III) is to spike your analytes and isotopically labeled analogues directly in the matrix before adding the extraction solvent. If you intend to use your isotopically labeled analogues to correct only for matrix effects, instrumental response drift, or injection variations, you can add them to your final extract. However, if you need to account for extraction variations, internal standards should be added directly to the sample matrix (before the extraction process).

After selecting the concentrations at which you want to test your recoveries and spiking analytes in your surrogate matrix, you need to prepare your post-spiked extract, which is the solution you will compare against your pre-spiked extracts. As we learned in part I, it is very important to have the same matrix components present in the solutions used to assess recoveries in order to account for matrix effects. To prepare a post-spiked extract, you simply perform your whole extraction procedure using blank matrix and then spike your post-extraction blanks with analytes assuming 100% recovery. Here it is worth  emphasizing that it is critical to ensure that your dilution factors are correct. For example, when using SPE cartridges to clean-up your extract, your final extract volume is going to be lower (~2.6 mL) than the original volume of extraction solvent used (3 mL). In that case, you may need to pool extracts collected from at least two replicates and then measure the volume you need to keep the same dilution factor. For instance, in our case, to prepare a post-spiked extract we added 50 µL of a 1 ppm analyte mix (same volume added to 0.5 g of surrogate matrix to attain 100 ng/g) in 3 mL of matrix blank extract (3 mL was the total volume of extraction solvent used for our brownies workflow). Recoveries are then estimated using the following equation:

% Recovery = (analyte response in pre-spiked extract/analyte response in post-spiked extract)*100

It is very important to emphasize that you can have amazing accuracy and precision results without having exhaustive recoveries. An example of this is the quantitation approach we used for daminozide in our technical article. In other words, recoveries and accuracy are two different things. We will talk more about this in Part III, so please stay tuned!