Matrix Effect on Butyltin Analysis of Sediments and Fish Tissues by GC-PFPD

A quantificação de compostos butílicos de estanho (BTs) em matrizes ambientais pode ser afetada pela ocorrência de interferentes encontrados em matrizes complexas tais como sedimentos e tecidos biológicos. O presente estudo investigou o efeito matriz em procedimentos analíticos para quantificação e especiação de BTs (TBT, DBT e MBT) em sedimentos e em dois tecidos (brânquias e fígado) de peixe, utilizando a técnica de cromatografia a gás com detecção fotométrica de chama pulsante (GC-PFPD). Diferentemente dos sedimentos avaliados, um efeito matriz significativo foi observado para os tecidos de peixe investigados, indicando que a quantificação de BTs deve ser realizada através da curva construída na matriz para evitar, dessa forma, erros de quantificação e redução da precisão analítica.


Introduction
Butyltin compounds (BTs) are a group of organic contaminants characterized by a tin atom (Sn) covalently bound to one or more butyl substituents. 1These compounds are utilized in many applications.However, their use has resulted in significant amounts of BT residues entering the ecosystem, mainly aquatic environments in which they can be found in the water, suspended particulate matter (SPM) and biota, 2 even after tributyltin (TBT) was banned as an antifouling paint agent in 2008. 3,4ttention has been focused on BTs in biological matrices due to their toxic effects on aquatic life even at lower concentrations (ng L -1 ). 5 Due to their hydrophobicity, 6 BTs are mainly found in sediments in aquatic systems and are particularly accessible for organisms in direct contact with the bottom. 7everal studies regarding TBT toxicity in fish have been reported, [8][9][10] and studies of demersal and detritivorous fishes, such as catfishes, have become important for the assessment of contaminant levels and potential risk to the aquatic trophic chain.In these organisms, butyltin residues are preferentially accumulated in liver tissues; thus, the monitoring of this organ for contamination is important. 11In addition to the liver, muscle tissue has also been used for BT analysis. 12Furthermore, the accumulation of contaminants in gills 13 is less frequently tested but it is also important to assess butyltin buildup from water and SPM.
The evaluation of the environmental impact of butyltins requires the development of accurate and precise analytical methods for several sample types, including speciation analysis of biological samples, which is a challenging task. 14ample extracts with high levels of organic matter or fats can contain substances that are co-extracted and deposited on the chromatographic column, reducing the efficiency of the separation.To minimize these interferences, a cleanup step, usually involving silica gel (Si−OH active sites), alumina (Al−OH active sites) and Florisil (for matrices with high lipid levels), is strongly suggested in addition to derivatization. 15][21] The selectivity, specificity, linearity and linear range as well as the quantification accuracy are evaluated using an internal standard (IS) and a certified reference material (CRM) to test accuracy and precision. 19A recovery assay is usually prepared using a surrogate, a compound that is chemically very similar to butyltins (TBT, DBT and MBT) but is not naturally present in samples. 20,22n important analytical parameter in BT analysis is the overestimation of analyte concentrations due to matrix effects in the chromatographic system, which is common for biological matrices. 22This effect can lead to an increase or decrease in the chromatographic response to butyltin analytes prepared in the sample matrix versus solvent. 18uring the chromatographic analysis, these effects can be observed as false-negative and false-positive peaks as well as a decrease in the detector signal for real samples compared with the analyte prepared in solvent. 23,24The matrix effect can manifest itself as an increase in the tested mass, 18 or as a blockage of active sites on the injector by matrix compounds and possible degradation or adsorption of analytes, 25,26 resulting in an increase or decrease in the detector response. 27As a consequence of these issues, the matrix effect is considered to be one of the most significant sources of error in analytical measurements and must be accounted for in studies involving organotin compounds, particularly in biota samples. 28Most studies involving this analytical parameter have cited the presence/absence of a matrix effect but do not clearly describe this effect.
In this study, our group presents quality control procedures for a current method for BT quantification in estuarine sediments and fish tissues.This study focuses on matrix effects present in these two types of environmental samples and their interference on the analytical system.Currently, there are few studies evaluating butyltin quantification, and this research is important to assess the real impact of these compounds on organisms and their associated environment.

Sample preparation and BT stock solutions
The estuarine catfish Cathorops spixii (Siluriformes, Ariidae) was collected at Guaratuba Bay (S25 o 49' W48 o 36') in Paraná State, South of Brazil.A total of 30 fish were sampled, frozen and transported to the laboratory.Gills and livers were removed and mixed by vortexing for the procedure validation tests.Sediments were sampled in the São Vicente Estuary (S23 o 58' W46 o 23') in São Paulo State, Southeast of Brazil.Sediment samples were homogenized and sieved (< 63 µm) in the laboratory for fine particle analysis.
The butyltin chloride stock solutions were prepared in hexane at concentrations of 5 µg mL -1 MBT, 6 µg mL -1 DBT and 4 µg mL -1 TBT.Surrogate (TPrT) and internal standard (IS) stock solutions were 6 and 10 µg mL -1 , respectively.These solutions were stored at −20 °C in the dark and used for a period of six months.

Analysis
The utilized methods were based on previous studies reported in the literature. 15,29,30The method for sediment samples was developed by Godoi et al. 15 , and the method for fish tissues was adapted from a method proposed by Cristale et al. 30 using 0.5 g of fresh tissue.The procedures for both methods were performed in triplicate (n = 3) for all tested analytical parameters.The tripropyltin (TPrT) surrogate compound was added at a concentration of 300 ng g -1 to samples prior to the extraction step (12 h before).The tetrabutyltin (TeBT) internal standard was added at a concentration of 1000 ng g -1 to the final extracts.The extraction of lipid content was based on the method proposed by Folch et al. 31 using a mixture of methanol and chloroform (2:1).

Quality control
Linearity was calculated according to the linear interval method proposed by Huber, 32 and described by Cristale et al. 30 Values below this confidence interval were rejected.Limits of detection (LOD) and limits of quantification (LOQ) for the analytical system (GC-PFPD) were based on the Huber test 32 in which the first note of linearity was considered to be LOD, and the first point in linearity was considered to be the LOQ.
The Thier and Zeumer 33 determination was used for LOD, considering Student's t-value (t n95% ) using three degrees of freedom and a confidence level of 0.05, the system sensitivity (S), the standard deviation of a low spiked level ( A ) and blank ( B ) and the number of spiked determinations (m) and blank measurements (n).LOQ was based on the lower spiked value, in which the recovery was above 70% (accuracy) and the relative standard deviation (RSD) was below 20% (precision).
Quality control was conducted using blanks, treated in the same manner as the samples, as well as the internal standard TeBT and surrogate TPrT.The accuracy and precision of the method were checked using mussel tissue ERM-CE477 for fish tissues and PACS-2 for sediment.Spiked tissues and sediments were also investigated at three different recovery levels on the calibration curve.

Matrix effects
Matrix effects were determined by adding data points to the calibration curve (n = 5) for liver and gill samples (n = 5) and sediment samples (n = 7) post-extraction.Then, the effect was measured based on the angular coefficient of the regression curve according to Thompson et al. 19 to verify a positive or negative influence on the results.This influence was calculated based on differences between the matrix and solvent angular coefficients multiplied by 100 or the point-by-point RSD of area/concentration values obtained from the matrix and solvent curves. 26The effect was considered to be significant if the obtained difference was above 10% for the angular coefficient method and above 20% for the RSD (%) method.A t-test was also applied with a significance level of 95% for significant differences between slopes of the curves. 34

Results and Discussion
For all of the tested BTs, the system linearity and the calibration curves were linear in the interval between 0.033 and 11.880 µg mL −1 .The correlation coefficients (R 2 ) of the calibration curves were above 0.95, and all of the curves were obtained using the internal standard TeBT.The LOD and LOQ values for the GC-PFPD method and for the tested methods are presented in Table 1.
The limits of detection observed in other studies involving BT analyses of sediment and fish are generally lower than the LOD value found in this study.Thus, the tested method, which uses a Grignard reagent and extraction with toluene/methanol, does not produce the highest analytical sensitivity for BTs in these matrices.In some cases, the use of a Grignard reagent as the  36 However, the quality control parameters can permit the safe application of our method because environmental and toxic levels of BTs are usually above the determined limit of quantification.Furthermore, the applied quality control procedures can assure the stringency of results using strict statistical parameters for determination of limit of detection instead of signal/noise parameters. 21,37Previous studies using the same analytical system 30,38 have found that the intermediate analytical response is reproducible for butyltin standards using GC-PFPD and the same analytical method. 39issue blanks were analyzed, and butyltin concentrations were not detectable for C. spixii liver or gills or for sediment samples used for quality control.For fish tissues, the evaluation of the accuracy and precision using the certified reference material resulted in recoveries of 98 ± 12, 61 ± 18 and 32 ± 6% for TBT, DBT and MBT, respectively.The recovery interval must fall between 50-150%, 21,40 so the obtained values satisfy the quality control parameters, with the exception of that for MBT.Lower analytical response is usually observed for MBT due to its higher polarity. 41Lower recoveries of DBT, as well as broader peaks, are generally due to the interaction of this BT with other co-extracted compounds present in biotic samples. 42The presence of foams and colloidal interfaces observed between the aqueous and organic phases can also reduce the BT recoveries and the reproducibility of the measurements. 34o obtain a realistic recovery in the absence of certified reference material for the fish tissues, spiked samples of blank fish tissues were also used for the recovery assay.The assay results ranged from 60 ± 7 to 95 ± 8% for gill samples and from 65 ± 2 to 88 ± 1% for liver samples spiked at 3 points on the calibration curve (33-196 ng g -1 ).These recoveries are acceptable for complex matrices, such as biological samples, 21 in which the lipid content can reduce the recovery efficiency.Lipid contents in the samples were approximately 7 and 2% in liver and gill tissues, respectively.Fats present in fish tissues can remain in the extract even after a clean-up step and consequently influence the chromatographic response, affecting the detection of butyltin compounds.Although lipid content and analyte adsorption in the matrix can decrease the recoveries, the use of a surrogate compound, i.e., TPrT, in each analysis can reduce potential errors.In these samples, surrogate (TPrT) recoveries were above 70% for both matrices and represent good method efficiency.
For sediment samples, the CRM analysis (PACS-2) using the method proposed by Godoi et al. 15 obtained recoveries between 98 ± 0.2 and 110 ± 0.13%, and the three spiked samples resulted in recoveries above 70% with RSD values below 20%.The RSD values for these samples are sometimes used to verify the matrix effect when values are considered to be high (above 20%). 43he matrix effect on the BT analyses has been discussed in the literature since 1990; however, these studies focused on the significant variations among the responses of the chromatographic system in different matrices when the same analytical method was applied.These studies do not address the different analytical response for a standard prepared in solvent and spiked matrices.The quantification is usually made based on calibration curves for standards prepared in solvents, resulting in significant quantification errors for samples prepared in interfering matrices.
As previously discussed, the matrix effect for a chromatographic system can result from the analytical method and the selected quantification system, as well as the presence of other matrix contaminants and components, such as organic matter, fats and hydrocarbons.The term "matrix effect" is also used to refer to the recovery of spiked samples compared with the recovery of an analyte from blanks, [42][43][44][45][46] or to the evaluation of interfering compounds at the same retention time as the analytes.
In the present study, chromatograms of BTs prepared in solvent and in the investigated matrices (sediment and fish tissues) do not exhibit differences in the retention times of analytes (Figure 1 and Table 2), and interfering peaks are only observed for sediment samples (Figure 1b) co-eluting with TPrT and TBT analyte peaks.
The matrix effect was observed for TBT in fish tissues when the calibration curves were overlapped (Figure 2), primarily in liver samples, and the signal intensity decreases.Tang and Wang 18 reported the same effect using the muscle of the milkfish (Chanos chanos); the effect was due to decrease ethylation (NaBEt 4 ) of organotins (butyltin and phenyltin compounds), which preferentially bond to cations and matrix sites.Thus, derivatized matrix components can occupy active sites during standard analysis, 24 and an internal standard and/or spiked sample are commonly used to reduce matrix effects caused by analyte losses during analytical procedures. 47atrix effects can lead to systematic errors that affect both the intercept and slope of the calibration curves. 46hen the evaluation method of Thompson et al. 19 was applied, a negative effect on TBT (−37.5%),DBT (−22.2%) and TPrT (−28.5%)angular coefficient curves was observed in liver tissues (Table 2).The slopes were also compared using the t-test, 34 and there was a significant difference (t-values higher than critical values) for the same analytes and tissue using the Thompson method.
These results for liver tissue suggest that the matrix effect is not only matrix-dependent but is also partially due to the chemical structure of each BT compound. 23he matrix effect was not observed using this method for TBT and DBT in gill samples based on the results presented in Figures 3a and 3b; an RSD value above 20% was found for some tested concentrations.The same RSD was found for MBT once matrix effect of liver tissues was observed in three points of curve as shown on Figure 3a.Differences above 20% RSD have been previously reported for fish tissues quantified by GC-FPD, resulting in different slopes for solvent and matrix calibration curves as well as MBT recovery below the confidence level and LOQ between 20 and 30 ng g -1 . 48he matrix effect for sediment samples was not observed when based on a difference in response of less than 10% for the angular coefficients (Table 2) and RSD values below 20% (Figure 3c).However, the analytical performance for MBT was worse than that for other BTs, i.e., lower recoveries.According to Ceulemans and Adams, 44 the higher sulfur content, as well as the presence of greater amounts of organic matter frequently found in estuarine environments, can strongly affect MBT extraction from sediments, suppressing the analyte signal. 22Thus,  MBT extraction depends on the sample composition, 49 and co-elution at the same retention time can be observed. 17In Figure 1, some sulfur interference can be observed in the chromatogram for the sediment, but co-elution occurs with TPrT and TBT, not MBT; this result is a potential focus for further studies using the tested method.The comparison of the results of this study with those reported in the literature indicates that the absence of the matrix effect in the tested sediment samples and the interference of biological matrices during analysis using PFPD are likely the result of variations in extraction/ derivatization efficiency or due to the use of different analytical systems.In complex matrices, the derivatization step with a Grignard reagent was reported as affecting the BT chromatographic response. 17Matrix interference was also previously observed in studies involving the BT determination using photometric detectors with different speciation methods, resulting in 40% higher response for fish tissues. 48,50In contrast, the matrix effect was not observed in biota samples using GC-PFPD for extraction based on solid phase micro extration (SPME), 5,17 demonstrating that the extraction method and the detection system can reduce the matrix effect. 24,27,51ess interference and higher selectivity were reported for pesticide analysis using PFPD analysis versus mass spectrometry (MS). 48According to Pinho et al. 24 , the MS detector can increase the matrix effect due to the metallic surface and combustion step that are not present on the  photometric detectors in PFPD.However, the matrix effect with spectrometric detectors can be reduced using extraction procedures that improve the selectivity. 51The reduced injection volume (2 µL) did not appear to have an influence on the matrix effects observed (fish tissues) and not observed (sediments) in the analytical system, and the mass of sample may also be investigated.
The observed differences in chromatographic response for fish tissues can lead to errors in the quantification of these compounds at reduced environmental concentrations, and their effect on the analytical system can reduce the method sensitivity. 52This issue is significant for BTs due to the banning of TBT after 2008; BT concentrations in the aquatic environment have experienced a considerable decrease, and the safe quantification of these compounds at lower levels is essential for determining toxic levels for aquatic organisms.

Conclusions
The method tested in this study can be applied to environmental samples of estuarine fish based on results obtained for quality control procedures for both types of tissues (liver and gills) used for BT analysis.Using GC-PFPD analysis, the matrix effect was observed in fish tissues but was absent in sediments.Although no matrix effect was observed for sediment samples, further studies will examine the interfering peaks observed in the sediment chromatogram by evaluating different estuarine sediment compositions of organic matter and sulfur.The establishment of a method based on quality control and the evaluation of matrix effects will improve the accuracy of analytical results, helping future environmental studies to determine the real impact of BT compounds in estuarine systems and nearby human communities.

Figure 2 .
Figure 2. Calibration curves for TBT in the matrix effect study: (a) fish tissue and (b) sediment.

Figure 3 .
Figure 3. Relative standard deviations (RSD) for the BT response in matrix and solvent calibration curves: (a) liver, (b) gills and (c) sediment.

Table 1 .
Limits of detection (LOD) and quantification (LOQ) for both the analytical system and the tested methods

Table 2 .
BT matrix effect study: retention times (t R ), linear range and angular coefficients of the calibration curves (n = 9)