LC-QTOF-MS as a Tool for Quantitative and Qualitative Analysis of Isoniazid and Its Metabolites in Dog Liver Samples

TL;DR

This study develops a method to measure isoniazid and its metabolites in dog liver samples, aiding in diagnosing poisoning cases.

Contribution

A validated LC-QTOF-MS method for quantifying isoniazid and analyzing its metabolite profile in canine liver is introduced.

Findings

Liver INH concentrations ranged from 11.822 to 30.484 μg/g in poisoned dogs.

Extensive necrotic lesions were observed in multiple tissues associated with high INH levels.

Isonicotinic acid was a prominent metabolite, while acetylated forms were negligible.

Abstract

Isoniazid (INH) is an antitubercular drug that exhibits high toxicity in dogs due to the absence of N-acetyltransferase activity in this species. Consequently, it has been implicated in both accidental and intentional poisonings in dogs. The aim of this study was to develop and validate an analytical method for the quantification of INH in canine liver samples and to apply it in the forensic investigation of seven suspected poisoning cases. The method, based on liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (LC-QTOF-MS), enabled both accurate INH measurement and analysis of the molecular pattern of its metabolite formation. In addition, histopathological examination of the stomach, pancreas, liver, and brain was performed. Liver INH concentrations ranged from 11.822 to 30.484 μg/g and were associated with extensive necrotic lesions across all examined tissues. A strong signal for isonicotinic acid was observed in all samples, whereas the acetylated metabolite was negligible. The developed method allows precise quantification of INH in canine liver and facilitates identification of the characteristic molecular profile of its metabolites.

Figures

Full text

1. Introduction

Isoniazid (INH) is one of the most effective anti-tuberculosis drugs, introduced into clinical use over 70 years ago (1952). Despite its widespread therapeutic success, especially when combined with other agents such as rifampicin, its precise mechanism of action remains only partially understood. INH is a prodrug that, once activated by the mycobacterial catalase-peroxidase enzyme (KatG), inhibits an enoyl-acyl carrier protein reductase, InhA. This effect disrupts the biosynthesis of mycolic acids, which are vital components of the mycobacterial cell wall [1].

In humans, the potential toxicity of INH is reduced through acetylation mediated by the N-acetyltransferase 2 (NAT 2). Dogs, however, lack this enzymatic system, which leads to increased toxicity in this species [2]. Due to the global prevalence of Mycobacterium tuberculosis infections, cases of INH poisoning in dogs, both accidental and intentional, have been reported worldwide. Notably, in Ukraine, intentional isoniazid poisonings have been associated with the control of the stray dog population [3]. This situation poses a significant public health and ethical challenge and is a frequent subject of scientific investigation by local researchers [3].

The clinical presentation of INH poisoning in dogs is well characterized, with neurological symptoms predominating due to the drug’s ability to cross the blood–brain barrier. The most commonly observed signs include seizures, often accompanied by hyperthermia, as well as ataxia and hypersalivation. The largest case series to date was reported by Schmid et al., who presented a retrospective analysis of 137 dogs with INH toxicosis between 2004 and 2014 [2]. However, this publication reports only cases in which INH exposure was confirmed based on history and provides no information on analytical methods that could be effectively used to confirm suspected INH poisoning, highlighting the need for reliable diagnostic techniques.

To date, only a few analytical methods have been reported for the toxicological assessment of INH in clinical samples. This likely reflects significant analytical issues arising from the complex biological matrix of the sampling material. In the study by Bayer et al., liquid chromatography-tandem mass spectrometry (LC-MS/MS) was used to analyze gastric content collected from poisoned dogs. While this material provides valuable diagnostic information, it is not optimal for the assessment of true internal exposure, as it does not allow for the determination of the concentration in tissues, making the clinical interpretation of the results less accurate [3]. In an earlier study, Wang et al. described INH analysis in canine plasma using LC-MS [4]. This type of sampling material enables a more accurate determination of systemic exposure. However, serum or plasma is rarely available in post-mortem toxicological analyses due to its limited biological stability [5].

This study aimed to develop a quantitative method for detecting INH in liver samples, the most readily available and commonly archived biological material in post-mortem examinations. The method applied liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (LC-QTOF-MS), which enables the quantitative determination of INH as well as the identification of its metabolites using untargeted screening, which may potentially serve as biomarkers of intoxication. Although potentially less robust and sensitive, LC-QTOF-MS allows metabolite identification without the need for analytical standards for each compound, which are typically required in targeted LC-MS/MS analyses [6,7]. The strength of this method is that the same acquisition permits both quantitation and non-targeted metabolite screening, which makes it a viable alternative to classical LC-MS/MS in a forensic setting. The method was applied to liver samples from seven dogs with suspected INH poisoning. Additionally, histological analysis of selected tissue samples was performed to relate measured INH concentrations to organ-specific damage.

2. Results

2.1. Analytical Method Validation

2.1.1. Selectivity, Linearity, and Carry-Over

Minor background signals were observed in 3 out of 18 blank samples; however, they did not interfere with quantification. Calibration curves (0.208–213 µg/g) showed excellent linearity with the regression equation y = 0.12976x and a correlation coefficient of R^2^ = 0.9998. No significant carry-over was detected when blank samples were injected after the highest analyte concentration.

Representative chromatograms are shown in Figure 1. The upper panel shows the chromatogram of a standard solution containing INH and the internal standard (IS, INH-d_4_), where both analytes eluted at 0.598 min. The middle panel shows a blank liver sample spiked with the IS, demonstrating the absence of interfering peaks at the retention times of INH and INH-d_4_. The lower panel illustrates the chromatogram of a clinical liver sample, where clear peaks corresponding to INH and INH-d_4_ were detected. These results confirm the selectivity of the method and its applicability for quantitative analysis of liver tissue samples.

2.1.2. Precision, Accuracy, Limit of Quantification and Recovery

Intra- and inter-day precision and accuracy are summarized in Table 1. Both precision and accuracy values complied with the acceptance criteria at concentration levels between 0.83 and 213 µg/g. According to [8], the accuracy is accepted in the range of 70–140% if precision is ≤20%. The limit of quantification (LOQ) was selected as the lowest calibration level (0.83 µg/g) at which the criteria for precision and accuracy were met. The mean recovery was 89.7% (range: 74–120%, which is within regulatory acceptance limits).

2.1.3. Relative Matrix Effect

The relative matrix effect (%SSE) was 87%, 111% and 95% at 0.83, 13.31, and 213 µg/g, respectively. These results indicate a minor influence of matrix components on the analytical response and lie within the acceptable limit of 80–120%.

2.2. Clinical Cases

The study was performed on tissues from seven privately owned domestic dogs from Lviv (Ukraine). In all of these cases, following routine walks in urban parks, forest belts, or within private household yards, the animals ingested discarded meat or meat products (baits) suspected to contain toxic substances. Credible estimation of the ingested dose was not possible. Shortly after ingestion, the dogs exhibited acute clinical signs of poisoning, characterized by hypersalivation, vomiting, and pronounced neurological disturbances, including ataxia. Owners of four affected dogs sought emergency veterinary care in Lviv; however, despite therapeutic intervention, all animals died. In three other cases, the owners did not seek veterinary assistance due to the rapid progression of clinical signs, which led to death within approximately three hours of symptom onset. The individual information on clinical cases is summarized in Table 2. The cases were considered as peracute if the time to death was less than 3 h and acute if it was between 3 and 6 h.

2.3. Pathology Findings

At necropsy, the predominant pathological findings included cyanosis, pancreatic necrosis, right ventricular dilatation, acute congestive hyperemia and pulmonary edema with focal alveolar emphysema. Additional lesions comprised hepatic and renal hyperemia, edema, and dystrophic–necrotic changes, as well as acute gastric catarrh and catarrhal enteritis. Hyperemia, punctate hemorrhages, and edema of the brain and meninges were also observed. Histopathological evaluation was performed using a semiquantitative scoring system ranging from 0 to 4, where 0—no lesion; 1—mild hemodynamic disturbance and destruction of cellular structures, edema of stromal-vascular structures, plasma extravasation into the interstitium; 2—moderately expressed circulatory disorders and degenerative–necrotic changes in the parenchymal and stromal-vascular structures, presence of isolated hemorrhages; 3—severe disturbances in the organ’s architecture and hemodynamics, massive hemorrhages, erythrocyte hemolysis, and pronounced necrosis of the organ’s parenchymal and stromal-vascular structures; 4—significant necrosis, lysis of parenchymal and stromal-vascular structures of the organ [9]. For each organ, five randomly selected, non-overlapping high-power fields (×400) were examined per animal and compared to age-matched controls from the same laboratory. Individual organ scores were expressed as the median per organ per dog (Table 3). Full graphical documentation of histopathological lesions in individual cases is provided in Supplementary File S1.

2.3.1. Stomach

Histological examination revealed primarily dystrophic and necrotic lesions. In peracute cases (Figure 2A), the changes were limited to degeneration, necrosis, desquamation of the epithelium of the apical folds of the stomach, and hypersecretion. There was strong PAS positivity, indicating the accumulation of glycoprotein-rich mucus. The submucosa showed edema, venular dilatation with hemolyzed blood, and endothelial edema. In acute cases (Figure 2B), necrosis was widespread, the mucosal surface was poorly contoured, and covered with fine-grained necrotic material. Principal, parietal and superficial cells were lysed, and the lamina propria was thickened. Edema, swelling, delamination, deformation of collagen and smooth muscle fibers, and karyolysis were observed in the submucosa and muscular layers.

2.3.2. Liver

In the liver, a marked disruption of the angioarchitectonics and parenchymal structure was observed. In cases of peracute poisoning (Figure 3A), pronounced congestive hyperemia, dilatation and deformation of sinusoidal walls and central veins, centrilobular hemorrhages, and disorganization of the hepatic laminae were evident. Hepatocytes exhibited indistinct cell borders, nuclear lysis, and frequent cytoplasmic accumulation of bilirubin, while hemosiderin deposits were detected in macrophages (Perls staining). In the acute course (Figure 3B), erythrocyte hemolysis and accumulation of blood plasma within dilated sinusoids and central veins were noted. The portal connective tissue appeared disorganized; vascular connective tissue fibers were loosened and deformed, and the nuclei of smooth muscle myocytes and endothelial cells showed predominant necrosis. These changes indicate a severe disturbance of blood rheology, reflected by the extensive necrotic alterations within the hepatic parenchyma.

2.3.3. Pancreas

INH exposure was associated with pancreatic necrosis. In peracute cases (Figure 4A), hemorrhages, edema, blood infiltration of the stroma, and degenerative–necrotic changes in vascular and stromal structures were noted. Acinar cells appeared edematous and lysed in hemorrhagic areas, consistent with hemorrhagic pancreatic necrosis. In the acute course (Figure 4B), extensive necrosis involved vascular–stromal structures, excretory ducts, and both exocrine and endocrine cells. Acinar cells were swollen, with cytoplasmic and nuclear fragmentation. Lytic processes within the stroma produced a uniform, homogeneous appearance containing clear vesicles (lipid droplets), characteristic of fatty pancreatic necrosis (pancreatosteatosis).

2.3.4. Brain

In the brains of dogs with INH poisoning, no inflammatory changes were observed, indicating dyscirculatory encephalopathy of toxic–hypoxic origin. In the peracute course, hemocirculatory disorders (angiopathy, perivascular edema, and hemorrhages), pericellular edema, and vacuolar dystrophy of stellate neurons were noted. In the acute course, pyknotic forms of pyramidal cells, “shadow cells” among neurons of the nuclear formations of the medulla oblongata, as well as disruption of the structural organization of nerve fibers and fragmentation of myelin fibers, were observed (Figure 5A,B).

2.4. Determination of INH Concentration in Clinical Liver Samples

The mean concentrations of INH in the liver samples ranged from 11.822 ± 1.378 μg/g (Liver 5) to 30.484 ± 1.726 μg/g (Liver 7). Very low signal intensity seen in the blank samples should be considered as noise responsible for a relatively high LOQ value. A summary of the measured concentrations is presented in Table 4.

2.5. Detection of INH Metabolites in Clinical Liver Samples

Apart from the detection of INH, the developed LC-QTOF method allowed for the detection of its major stable metabolites: N-acetylisoniazid and isonicotinic acid. The retention times were 0.6 min and 0.59 min, respectively. Because nicotinic acid (vitamin B3) and isonicotinic acid share an identical monoisotopic mass and differ only in the position of the nitrogen atom within the aromatic ring, their initial identification proved difficult. An additional analysis of a sample spiked with a pharmaceutical vitamin B3 formulation confirmed that the peak at 0.89 min corresponded to nicotinic acid, so the peak of an identical mass eluted at 0.59 min was identified as isonicotinic acid. This was further confirmed by fragmentation at 10 eV, which revealed that the peak at 0.59 min was characterized by a dominant ion at m/z 96.044, specific for isonicotinic acid. This signal was observed exclusively in INH-exposed liver samples and was absent in the blank controls (Figure 6).

Signal intensities for N-acetylisoniazid and isonicotinic acid were compared between blank samples and INH-exposed liver samples (Table 5). In all cases, signal values were higher in the INH-exposed samples compared with the blanks.

2.6. Exploration of Trends and Statistical Relations Between Pathological and Toxicological Data

Table 6 summarizes the individual information on the tissue damage score together with liver INH concentration and the signal intensity for isonicotinic acid and N-acetylisoniazid, and provides the Pearson correlation coefficient that describes the relation between these values. Due to a uniform value for the stomach score in all individuals, no assessment was possible. Although none of the correlations seem to be strong, it was found that the signal for isonicotinic acid and the N-acetyl derivative seems to correlate slightly stronger with the lesions (particularly in the pancreas) as compared to the parent compound. Surprisingly, the Pearson correlation value was higher for the pancreas than for the liver, in which the xenobiotics were actually measured. The weakest correlation was observed for brain lesions and INH concentration, as well as the intensity of its metabolites.

Additionally, the data were analyzed from the perspective of clinical severity. The individual measurements and scores were divided into two groups: history of acute and peracute course, and were compared by means of the Mann–Whitney U test (Table 7). The only statistically significant difference was the tissue lesion score for the pancreas.

3. Discussion

Although INH is primarily known as a first-line drug for the treatment of tuberculosis in humans, its high species-specific sensitivity in dogs makes it an emerging concern in veterinary toxicology. Therefore, the objectives of this study were twofold: first, to develop and validate an analytical method for quantifying the toxicant in the most commonly collected material for toxicological analysis in companion animals—the liver; and second, to apply this method to clinical samples obtained from real-life poisoning cases. In addition, an effort was made to qualitatively characterize the specific molecular pattern of canine INH metabolites and to evaluate its potential utility as a biomarker of exposure.

To address these aims, we successfully implemented an LC-QTOF-MS approach for the quantification of INH and detection of its metabolites in canine liver tissue. This represents the first reported application of LC-QTOF-MS in this context. Previous quantitative analyses of INH have focused primarily on human plasma [10,11,12,13], as well as urine [14], rat tissues [15,16], and plasma or gastric content in dogs [3,4]. Our results demonstrate the suitability of LC-QTOF-MS for analyzing INH in canine liver, thereby expanding the range of applicable biological matrices.

Bayer et al. [3] reported the use of the LC-MS/MS method to measure INH in canine gastric content. Unfortunately, the authors did not specify a formal LOQ, only a working measurement range of 10–1000 µg/mL. While this data provides a useful reference point, gastric content may not accurately reflect systemic drug exposure and is not routinely collected during postmortem examinations or may be lost due to extensive vomiting before death [17]. A different study by Wang et al. presents a method for INH determination in canine plasma using LC-MS, achieving a LOQ of 0.025 µg/mL [4]. In our study, the LOQ for canine liver tissue was 0.83 µg/g. Higher LOQ for the liver is not surprising due to the limitations associated with tissue processing, the necessity for homogenization, the presence of interfering compounds and the matrix effect. Despite these analytical challenges, the liver remains the preferred matrix in postmortem toxicology, as blood is highly unstable after death due to autolysis and drug redistribution [18]. In the present study, the matrix effect was corrected using an internal standard, resulting in a value below ± 20%, which is compliant with SANTE guidelines [8].

The study of Fang et al. [19] reported the quantitative determination of INH in both plasma and liver tissue of mice using LC-MS/MS, with LOQ values of 0.11 µg/g and 0.018 µg/mL for liver and plasma, respectively. These findings are lower than our results (LOQ 0.83 µg/g) and support the common conclusion that LC-QTOF-MS methods exhibit lower sensitivity compared to LC-MS/MS. They also highlight that liver tissue represents a more challenging matrix than plasma, requiring additional optimization to minimize the matrix effect. The intra- and inter-day precision and accuracy values in our study were comparable to those reported in the literature [19] and compliant with SANTE guidelines. The inter-day coefficient of variation (CV) of 20.62% at the lowest calibration point (0.208 µg/g) slightly exceeded the recommended 20% limit. Even though the deviation was very small (0.62%), we decided to adopt the next dilution level (0.83 µg/g) as our LOQ. Reverse phase HPLC-based methods with Photodiode Array Detection may provide even better sensitivity, as has been shown for rat liver (LOQ of 0.2 µg/g) [20]. However, they do not allow for the detection of metabolites if their standards are not used in the analysis. It should be underlined that the exploration of the metabolite profile was a key aspect of our study using LC-QTOF-MS. Although combined approaches employing LC–MS/MS for quantitative analysis and LC–QTOF–MS for metabolite identification have been reported [21], in the present study, the sensitivity achieved with LC–QTOF–MS was sufficient for both quantitative analysis and metabolite detection in a single run.

The developed method was subsequently applied to a series of suspected INH poisoning cases in dogs from the Lviv region. The affected animals exhibited neurological signs, including seizures and ataxia, accompanied by vomiting and hypersalivation, and died within 3–6 h after ingesting the suspected bait material. Similar clinical manifestations have been previously reported in canine INH intoxications [2,22,23]. Furthermore, a documented history of the drug’s use for malicious dog poisonings in this area provided additional justification for analyzing the samples for the presence of INH [24].

Measured INH concentrations in liver samples ranged from 11.822 ± 1.378 µg/g (Case 5) to 30.484 ± 1.726 µg/g (Case 7). All these values were much above the LOQ, indicating that the analytical method was sufficiently sensitive for determining INH in this biological matrix. Due to the lack of published data on INH concentrations in canine liver, only indirect comparisons are possible. Bayer et al. [3] reported INH concentrations in stomach content ranging from 0.0905 to 5.7998 µg/g. Although these values are much lower compared to the levels we found in the livers, the authors interpreted them as lethal. On the other hand, Wang et al. [4] reported that after oral administration of 150 mg INH to six dogs weighing 20 ± 5 kg, the maximum plasma concentrations measured one hour post-dose were 3.90 ± 1.71 µg/mL. As Wang et al. did not report the clinical response in dogs to the administered dose, it is difficult to interpret whether such concentrations may lead to toxicity. The literature also indicates that the lethal dose of INH for dogs is approximately 75 mg/kg body weight [2], approximately 10 times higher than the dose used by Wang et al. (~7.5 mg/kg). Surprisingly low concentrations measured by Bayer et al. may suggest that massive vomiting, which is often observed in canine cases of INH poisoning [3], may lead to autodecontamination, and gastric content samples may not represent the systemic exposure well. Moreover, any inference relating tissue concentrations to toxicity remains speculative as the dose information is not available in either the present study or the one by Bayer et al. In view of the limited data available for dogs, interspecies extrapolation might seem a valuable option. However, known interspecies differences in INH metabolism considerably limit the validity of such comparisons.

In species possessing an active NAT2 enzyme, INH is primarily converted into the relatively non-toxic N-acetylisoniazid. In contrast, dogs are classified as so-called non-acetylators (lack of NAT2 activity), resulting in an alternative biotransformation pathway leading to the formation of isonicotinic acid and hydrazine. Hydrazine subsequently undergoes further metabolism in the liver, producing ammonia and reactive intermediates that bind covalently to cellular macromolecules. The reactive intermediates may be generated from INH also by free radical processes, e.g., catalyzed by metal ions or myeloperoxidase [25]. These mechanisms are associated with marked hepatotoxic and neurotoxic effects [26]. The application of LC-QTOF-MS allowed us to explore potential biomarkers of INH poisoning in canine liver tissue. Among the metabolites analyzed, isonicotinic acid showed the strongest correlation with INH levels, suggesting its potential as a reliable indicator of exposure. In contrast, the signal of N-acetylisoniazid was very low, only slightly exceeding the levels of blank samples, likely due to the absence of NAT2 activity in dogs [2]. These findings highlight the value of including metabolite profiling in postmortem toxicological investigations, as it may provide more robust evidence of drug exposure than the measurement of the parent compound alone. This is particularly important in poisonings where metabolic activation of the parent compound plays the central role in the pathomechanism. In INH poisonings, the cytotoxic effect is caused primarily by hydrazine [26]. Yet hydrazine is highly reactive and unstable and is produced in relatively small amounts making it a difficult target for analysis [26]. It seems likely that other, more stable metabolites, like isonicotinic acid, may serve as a proxy not only of the INH exposure but also of the intensity of metabolism and hydrazine formation. Interestingly, the correlation analysis between the INH concentration and tissue lesion scores did not reveal any meaningful trend. However, for the metabolites, this correlation seems to be stronger. The highest value (0.4734) was seen for the isonicotinic acid signal and the severity of pancreatic damage, which may suggest that indeed this metabolite could serve as a potential biomarker of poisoning. However, there was no significant difference in INH concentration or metabolite signal intensity between the acute and peracute cases. Again, the pancreatic lesions seem to be the only difference, as all acute cases showed a higher lesion score in this organ compared to the peracute ones. The limited number of samples in this study underlines the need for further investigation to confirm the reliability of isonicotinic acid as a biomarker in canine INH toxicity.

In addition to the toxic action of hydrazine and ammonia, a significant role in the development of neurological symptoms is played by the direct effect of INH on pyridoxine metabolism. INH forms inactive complexes with vitamin B_6_, resulting in the depletion of its active form—pyridoxal-5′-phosphate—the essential cofactor of glutamic acid decarboxylase (GAD). Consequently, the synthesis of γ-aminobutyric acid (GABA), the main inhibitory neurotransmitter, is reduced, leading to a decreased seizure threshold and the occurrence of convulsions [27].

All these mechanisms may have contributed to the morphological changes observed in this study, including extensive necrotic lesions in the hepatic parenchyma as well as neuronal damage and destruction of myelin fibers within the medulla oblongata. Based on literature reports indicating the presence of INH in pancreatic juice [28], as well as the observed macroscopic appearance of pancreatic necrosis during necropsy, a histopathological examination of the pancreas was performed. The observed lesions included extensive necrosis, which suggests that INH exerts a multi-organ toxic effect. Furthermore, due to the oral route of exposure and the presence of gastrointestinal symptoms, a histopathological examination of the stomach was conducted. As expected, it revealed toxic lesions, confirming the likely local impact of INH on the epithelium and mucosal architecture. These findings suggest that INH not only induces systemic organ damage but also causes direct local tissue injury. To our knowledge, this study provides the first detailed histopathological description of pancreatic and gastric lesions in dogs following INH exposure.

While the results provide valuable insights, several limitations should be acknowledged, and some questions remain open for future investigation. The primary limitation is the lack of screening for other common toxicants that could have been involved. This prevents us from ruling out the possibility of co-intoxication among the investigated cases. Second, a larger number of individuals would allow for more robust statistical analysis and offer greater insight into the molecular pattern of metabolites and their relationship to the severity or clinical presentation of poisoning.

Although dogs lack functional NAT2 enzyme activity, N-acetylisoniazid was detected in clinical samples at slightly higher levels than in controls. Previous studies have described the biotransformation of certain drugs by endogenous bacteria [29,30], which may offer a potential explanation for this finding. Additionally, preanalytical sample stability should be assessed with respect to factors such as the time from death to sampling, storage temperature, and possible effects of freeze–thaw cycles, to ensure accurate interpretation of postmortem concentrations.

Quantitative metabolite analysis could further elucidate the relationship between the extent of toxic damage and the intensity of metabolism; however, it would substantially increase analytical costs, raising questions about its applicability in clinical diagnostics. Finally, extending the study to compare INH levels in the liver with those in other organs susceptible to damage (such as the stomach, pancreas, and brain, as confirmed by histopathological studies) would be valuable. Unfortunately, this was not feasible in the present work, as these tissues are not routinely archived for toxicological examination.

Despite these limitations, this study provides a valuable tool for confirming INH exposure in dogs, with applications in both veterinary forensic and experimental toxicology. Furthermore, the use of LC-QTOF-MS enabled deeper insight into species-specific molecular metabolic patterns, contributing to a better understanding of the underlying pathomechanisms of INH toxicity.

4. Materials and Methods

4.1. Chemicals and Reagents Used in the Toxicological Analysis

INH (analytical grade, >99% purity) was purchased from Sigma-Aldrich (St. Louis, MO, USA). The internal standard (INH-d_4_, 99.3% purity) was obtained from LGC Standards (Toronto, ON, Canada). Acetonitrile, formic acid, methanol and water (all LC-MS grade) were obtained from Merck (Darmstadt, Germany). All chemicals and solvents were of analytical or LC-MS grade and were used without further purification.

4.2. LC-QTOF-MS Conditions

Liquid chromatography was performed on an Agilent 1260 Infinity II system equipped with a binary pump, autosampler, and thermostatted column compartment (Agilent Technologies, Santa Clara, CA, USA). The system was coupled with an Agilent 6546 LC/Q-TOF mass spectrometer with a Dual Jet Stream electrospray ionization (ESI) source operating in positive ion mode. Data acquisition and processing were performed with Agilent MassHunter Workstation software (v. 10.1.48, Agilent Technologies, Santa Clara, CA, USA). Chromatographic separation was achieved on an InfinityLab Poroshell 120 EC-C8 column (2.1 × 100 mm, 2.7 µm; Agilent Technologies, Santa Clara, CA, USA). The mobile phases were (A) water with 0.1% formic acid (v/v) and (B) methanol with 0.1% formic acid (v/v). Analysis was performed on the injection volume of 1 µL with the flow rate of 0.5 mL/min for a total runtime of 16 min. The column temperature was set to 30 °C and the autosampler to 4 °C. The gradient program is presented in Table 8.

Mass spectrometric detection was carried out in centroid mode with a scan range of 100–1000 m/z for the MS experiment and 20–1000 m/z for the targeted MS/MS experiment. For All-Ions experiments (data-independent analysis), the acquisition range was 20–200 m/z. Source parameters were as follows: capillary voltage 4000 V, fragmentor voltage 150 V, nebulizer pressure 60 psi, drying gas 320 °C at 13 L/min, and sheath gas 400 °C at 12 L/min. Collision energies were automatically optimized by the MassHunter software (v. 10.1.48, Agilent Technologies, Santa Clara, CA, USA). Reference masses at m/z 121.0509 and 922.0098 were continuously monitored for mass correction. The targeted ions included INH (138.066 m/z) and INH-d4 (142.091 m/z) used as the internal standard.

4.3. Preparation of INH Standard and Quality Control Samples

Stock solutions of INH (2.13 mg/mL) and IS (1 mg/mL) were prepared in LC-MS grade methanol (−80 °C). Fresh working solutions were obtained daily by appropriate dilution with the same solvent. Calibration samples were prepared by spiking 1 g of drug-free liver homogenate with working standard solutions to yield final INH concentrations of 0.208, 0.83, 3.31, 13.31, 53.25, and 213 µg/g, with a constant IS concentration of 6 µg/g. Samples were incubated for 15 min with gentle shaking. Blank liver samples (n = 18) were obtained during routine necropsies from dogs that died due to defined, non-toxicological reasons (origin: Poland and Ukraine). Blank samples were processed using the same procedure as described above, with pure methanol replacing the INH solution. Solvent calibration samples were prepared using 10% methanol as a matrix.

4.4. Sample Preparation

Liver samples from the poisoned dogs were homogenized (1 g) and processed as blank samples. Following incubation with the IS, 6 mL of 10% methanol was added to each homogenate, and the samples were shaken. The mixtures were centrifuged at 4500 rpm for 15 min at 4 °C. A 150 µL aliquot of the supernatant was transferred to a new tube and mixed with 750 µL of methanol. The mixtures were centrifuged at 13,500 rpm for 2 min at room temperature. The clear supernatant was transferred into autosampler vials for mass spectrometry analysis, and three injections were performed for each sample.

4.5. Method Validation

The analytical method validation protocol was based on SANTE/11312/2021 Guidance Document [8]. The validation covered the key performance parameters, including selectivity, linearity, carry-over, precision, accuracy, matrix effects, LOQ, and recovery.

4.5.1. Selectivity, Linearity, and Carry-Over

Selectivity was evaluated by analyzing blank liver samples obtained from dogs that died due to causes other than poisoning. Linearity was assessed using matrix-matched calibration curves prepared at six concentration levels across the validated range. Calibration curves were constructed by plotting analyte-to-IS peak area ratios against nominal concentrations and evaluated by linear regression. Carry-over was tested by injecting blank samples immediately after the highest calibration standard to confirm the absence of analyte residues.

4.5.2. Precision, Accuracy, LOQ and Recovery

Intra-day precision and accuracy were determined at six concentration levels using four replicates per level within a single batch. Inter-day precision and accuracy were assessed with twelve replicates per level obtained over three consecutive days. Acceptance criteria were based on [8]: accuracy within the range of 70–140% with precision ≤20%. The LOQ was defined as the lowest calibration level that met the criteria for accuracy and precision. Recovery was assessed by comparing responses of blank samples spiked with isoniazid prior to extraction with those fortified after extraction, thereby evaluating extraction efficiency.

4.5.3. Relative Matrix Effect

The relative matrix effect was assessed as the percent signal suppression/enhancement (%SSE) by comparing the slopes of calibration curves prepared in pure solvent and in an analyte-free matrix, using the following formula:

$$[eqn]$$

The calibration curve was constructed using matrix-matched standards representative of the composition of the actual samples. To minimize the impact of matrix effects, IS was applied for signal normalization.

4.6. Identification of INH Metabolites

In addition to INH, two metabolites were investigated: N-acetylisoniazid and isonicotinic acid. They were identified using a data-independent All-Ions MS/MS acquisition mode supported with the fragmentation data from a commercial spectral library (Agilent PCDL). The precise m/z values for their parent ions were 180.07675 and 124.03931 for N-acetylisoniazid and isonicotinic acid, respectively. Although isonicotinic acid has exactly the same m/z value as endogenic niacin, the differences in the fragmentation pattern allowed for identification of the signal specific to the INH metabolite. The value of the signal for the untargeted compounds (INH metabolites) was compared between the clinical liver samples and the blanks using the Mann–Whitney U test. Additionally, INH concentrations as well as metabolite signal intensities between the peracute and acute cases were compared. Statistical significance of the difference between the two groups was defined at p < 0.05.

4.7. Data Analysis

Peak integration, compound identification, and quantification were performed using Agilent MassHunter Qualitative 10 and Profinder 10 (Agilent Technologies, Santa Clara, CA, USA). Compounds were identified based on the [M + H]+ ion and confirmed by at least two fragment ions from a spectral library. Identification criteria included a retention time tolerance of 0.1 min, a co-elution score ≥ 90, and a signal-to-noise ratio ≥ 5. Statistical analyses and visualization were performed in R v. 4.4.1 (RStudio v. 2024.09, R Foundation for Statistical Computing, Vienna, Austria).

4.8. Processing of Tissue Samples

During necropsy, samples of the liver, pancreas, stomach, and brain were collected and fixed in 10% neutral buffered formalin, Carnoy’s fixative, Bouin’s fluid, or 96% ethanol. Separate liver samples (approx. 10 g) were collected for toxicological analysis, immediately frozen, and stored at −20 °C for <6 months (after that, liver samples were transferred to the Department of Pharmacology and Toxicology and stored at −80 °C until assayed, <6 months). Owner information was anonymized and, according to the local and EU law, no ethical clearance was necessary to carry out the analyses. For histopathological examination, tissues were dehydrated in graded ethanol, cleared, and embedded in paraffin. Sections 5–7 µm thick were prepared using an MC-2 microtome and stained with hematoxylin and eosin for routine light microscopy. Carbohydrates were demonstrated using the McManus periodic acid–Schiff (PAS) reaction, and hemosiderin was detected with Perls’ Prussian blue staining. Nissl staining was performed to visualize basophilic structures of nerve cells. Histological and histochemical examinations were carried out using a Leica DM2500 microscope (Leica Microsystems, Zürich, Switzerland), and photomicrographs were captured with a Leica DFC450 C digital camera using Leica Application Suite software (version 4.4). Detailed protocols with materials are described in Supplementary File S2 together with relevant references [24,31,32,33,34].