Detection of Anti-Isoniazid and Anti-CYP Antibodies in Patients with Isoniazid-Induced Liver Failure

Abstract

Isoniazid (INH)-induced hepatotoxicity remains one of the most common causes of drug-induced idiosyncratic liver injury and liver failure. This form of liver injury is not believed to be immune-mediated because it is not usually associated with fever or rash, does not recur more rapidly on rechallenge, and previous studies have failed to identify anti-INH antibodies. In this paper we found antibodies present in the sera of 15/19 cases of INH-induced liver failure. Anti-INH antibodies were present in 8; 11 sera had anti-CYP2E1 antibodies, 14 sera had antibodies against CYP2E1 modified by INH, 14 sera had anti-CYP3A4 antibodies, and 10 sera had anti-CYP2C9 antibodies. INH was found to form covalent adducts with CYP2E1, CYP3A4, and CYP2C9. None of these antibodies were detected in sera from INH-treated controls without significant liver injury. The presence of a range of anti-drug and autoantibodies has been observed in other drug-induced liver injury that is presumed to be immune-mediated.

Conclusion

These data provide strong evidence that INH induces an immune response that causes INH-induced liver injury.

Given its unquestioned efficacy, isoniazid (INH) remains the keystone of tuberculosis treatment despite its causing relatively frequent liver injury and liver failure (1–3). The hepatotoxic effects of INH have been thought to be due to the bioactivation of acetylhydrazine (AcHz), a metabolite of INH (4–6). However, this conclusion was based on a rat model of acute toxicity that does not resemble clinical INH-induced liver injury (4, 6). More recently we have demonstrated that INH itself can be bioactivated and bind covalently to the liver proteins of mice in vivo and to human liver microsomes in vitro (7). Covalent binding of the parent drug was observed in rats, but it was much less than in mice, and it appears that mice are a better model for humans, especially humans with the slow acetylator phenotype, than rats. Despite this, the mechanism of INH-induced hepatotoxicity is currently unknown, and there is no good animal model in which mechanistic hypothesis can be tested. Decades ago, it was speculated that the mechanism of liver injury could be a hypersensitivity reaction because it had characteristics such as delay in onset and no simple relationship between the dose and risk of liver injury that are typical of immune-mediated reactions. In addition, there were cases with a fast onset upon rechallenge and cases associated with fever, rash, and eosinophilic infiltrate in the liver (2). INH can clearly induce an immune response because it frequently causes a fever and rash or antinuclear antibodies independent of liver injury, and occasionally provokes an autoimmune reaction similar to lupus (3, 4). However, in most cases of INH-induced liver injury, especially mild injury, it is not associated with allergic features, there is a lack of rapid onset on rechallenge, and anti-INH antibodies have not been detected. The lack of anti-INH antibodies provided an argument against an immune-mediated mechanism for INH-induced liver injury, leading to use of the term “metabolic idiosyncrasy” (8–10). More recently, reconsideration of an immune basis for the majority of idiosyncratic hepatotoxic reactions has led us to reassess the evidence (11). We postulated that mild cases of liver injury may resolve if immune tolerance develops that can eliminate memory T cells. This might prevent a rapid response on rechallenge; only severe cases with failure of immune tolerance would exhibit the hallmark features of an immune-mediated reaction (3). In this report we re-examine whether INH-induced liver injury is associated with evidence of an immune-mediated reaction and anti-drug antibodies.

Patients and Methods

Human Subjects

Upon research ethics board approval, a total of 20 patients undergoing prophylaxis with INH were recruited by the Toronto Western hospital (Toronto, ON). After obtaining informed consent, blood was drawn into heparinized tubes from these 20 patients before the initiation of INH therapy to be used as baseline measurement, and patients were followed every month until they finished INH therapy. None of the 20 patients developed severe hepatotoxicity; five patients out of twenty developed a mild increase in alanine aminotransferase (ALT, 47 – 144 U/L). Serum samples were also obtained from 19 patients enrolled in the Acute Liver Failure Study Group (ALFSG) registry and who were presumed to have INH-induced toxicity leading to encephalopathy and coagulopathy as stipulated by entry criteria for the study. Each patient’s clinical history was reviewed by the site principal investigator and by the study center principal investigator (WML) and were adjudicated as at least probable or higher (>50% likelihood) due to INH (12–14). Results of serum antinuclear antibodies (ANA), antibodies against liver/kidney microsomes (LKM), anti-mitochondrial antibodies (AMA), anti-smooth muscle antibodies (ASMA) and ALT were also available for some patients from the ALFSG database as measured by the respective hospitals.

Method for Detection of Anti-INH antibodies

Lysozyme (L; Sigma, Oakville, ON) was modified with an N-hydroxysuccinimide activated ester of isonicotinic acid (INA-NHS) using a previously described procedure to give lysozyme coupled to INH (L-INH) (7). The activated ester of isonicotinic acid and the reactive metabolite of INH both react with amino groups on proteins and form the same product; therefore, this method should mimic the covalent binding to proteins that occurs in vivo (7). Briefly lysozyme (2.5 mL at 2 mg/mL solution in PBS pH 7.4) was incubated with 10 mg of INA-NHS and reacted for 1 hour at room temperature. The mixture was dialysed using a 1,000 MW cut-off filter. Protein concentration was measured by a bicinchoninic acid kit (Fisher Scientific, Ottawa, ON) and 10 μg of protein/well was loaded on the gel. The protein was separated by electrophoresis (8% SDS-PAGE) and transferred onto a nitrocellulose membrane (Bio-Rad, Mississauga, ON). Human serum was used as the primary antibody diluted in tris-buffered saline with Tween (TBST) pH = 7.4 containing 4% milk. The secondary antibody was anti-human IgG-peroxidase (AbD Serotec, Raleigh, NC). As a positive control, rabbit anti-INH antibody was used to detect INH bound to lysozyme, detected by goat anti-rabbit IgG-peroxidase (Sigma) as described before (7). Bound peroxidase was detected using Supersignal West Pico Chemiluminescent Substrate (Fisher Scientific).

Detection of Anti-CYP Antibodies

A pool of human liver microsomes (HLM) from 50 donors, CYP2E1, CYP3A4, and CYP2C9 were purchased from (BD Biosciences, Missisauga, ON). CYP2E1 was modified by the reactive metabolite of INH in two ways: 1) by reacting CYP2E1 with INA-NHS, which we refer to as CYP2E1-INH, in a similar manner to the modification of lysozyme and 2) by incubation of CYP2E1 (0.5 mg/mL) with INH (500 μM) and a NADPH-generating system (Solutions A and B; BD Biosciences) for 1 hour at 37 °C.

Enzyme-linked immunosorbent assay (ELISA); 96-well microtiter plate (Bethyl Laboratories, Montgomery, TX) was coated with 400 ng of protein overnight. Proteins included HLM, CYP2E1, CYP2E1-INH, CYP3A4, CYP2C9, lysozyme or L-INH. Human serum diluted 1:1000 in TBST containing 4% milk was used as primary antibody and allowed to sit overnight with slight agitation. As secondary antibody, polyclonal goat anti-human IgG-peroxidase (AbD Serotec) was used. Color was developed using 3,3′,5,5′-tetramethylbenzidine (TMB) solution and absorbance was monitored at 450 and 540 nm.

Covalent Binding of INH to CYP Isozymes

INH (100 μM) was incubated with a total protein concentration of 0.5 mg/mL for CYP2E1, CYP2C9, or CYP3A4 for 30 min at 37°C in the presence or absence of an NADPH-generating system. After 30 min, the reaction was stopped by placing the reaction mixture on ice. Protein concentration was measured by BCA kit and 5 μg of protein/lane was loaded a gel. Covalent binding of INH to each CYP isozyme was determined by western blotting using the anti-INH antibody that we have previously used to demonstrate INH binding to rodent liver and human liver microsomes (7).

Statistical Analysis

Statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, CA). Data was analysed using two tailed nonparametric t-test. A p value < 0.05 was considered significant (*p < 0.05; **p < 0.01; ***p < 0.001).

Results

Lysozyme was modified by reacting it with an activated ester of isonicotinic acid to give L-INH (Figure 1); this mimics the in vivo bioactivation and covalent binding of INH to protein and is similar to what we previously used to make an anti-INH antibody that was used for these western blots (7). We also modified other proteins such as human serum albumin and Blue Carrier Immunogenic Protein (Fisher Scientific) with INA-NHS analogous to lysozyme or used a liver homogenate from untreated or INH-treated mice instead of L-INH in order to detect anti-INH antibodies; however, when control sera (sera obtained from patients before the start of INH prophylaxis as described in the Methods) were used as the primary antibody for the western blots, we observed extensive background binding that precluded the use of these proteins for detection of anti-INH antibodies (data not shown). This was in contrast to lysozyme and L-INH in which there was no binding of control sera, and therefore L-INH was used in subsequent experiments. Preliminary experiments using western blots showed the presence of anti-INH antibodies in the serum of two INH-induced liver failure patients, but not in a patient who had been treated with INH and developed mild liver injury (ALT = 144 U/L; Figure 2). However, not all of the patients who had liver failure had anti-INH antibodies. We used ELISA to screen for the presence of anti-INH antibodies in all of the available serum samples. In addition to patients with liver failure, 20 patients undergoing prophylactic INH treatment without severe liver injury were tested; only 5/20 of these patients had a mild increase in ALT (between 47 and 144 U/L which was associated with an increase in peripheral Th17 cells); however, none of these 20 patients had detectable anti-INH antibodies (Figure 3A). Given the positive preliminary results that detected anti-INH antibodies by western blotting, we used ELISA to screen all of the sera that were available from patients who had been diagnosed as having INH-induced liver failure for anti-INH antibodies. Of these 19 serum samples, 8 tested positive for the presence of anti-INH antibodies (Figure 3A). Preincubation of the serum from those 8 patients who had anti-INH antibodies with INH prevented binding, which indicated that binding was specific for INH-modified proteins (Figure 3B). In addition, all sera that were positive for anti-INH antibodies by ELISA were also analyzed for anti-INH antibodies by western blotting which confirmed the ELISA results. Specifically, in all cases incubation of the western blots with sera from patients who tested positive for anti-INH antibodies by ELISA showed a strong band that appeared only on the lanes loaded with INH-modified lysozyme, but not with lysozyme itself, and preincubation of the human serum with INH almost completely eliminated binding indicating specificity of antiserum for INH (Figure 3C). Four serum samples from patients on INH prophylaxis and 4 from patients with liver failure that did not have anti-INH antibodies as determined by ELISA were analyzed by western blotting and there were no bands corresponding to anti-INH antibodies thus confirming the ELISA results (Supplemental Figure 1 = Figure S1). Demographic data from patients who were on INH prophylaxis only and patients with acute liver failure who had taken INH are shown in Table 1 and Table 2. No correlation between the presence of anti-INH antibodies and other patient characteristics were found.

Figure 1.

Western blot of lysozyme (L) or lysozyme modified by INA-NHS (L-INH), which mimics the covalent binding of the reactive metabolite of INH. Detection was with an anti-INH antibody as previously described (7).

Figure 2.

Detection of anti-INH antibodies by western blotting. L = lysozyme and L-INH = INH-modified lysozyme. Serum was diluted 1:400; #1-0 = control serum from a patient at baseline, #1–2 = serum from the same patient 98 days after initiation of therapy (ALT = 144 U/L). ALF-19 and ALF-3 are sera from two patients with INH-induced liver failure.

Figure 3.

Detection of anti-INH antibodies by ELISA and western blotting. A) Baseline sera before initiation of INH therapy (Baseline), sera from patients who were treated with INH but did not develop significant hepatotoxicity (Prophylaxis), sera from patients who developed INH-induced liver failure but did not test positive for anti-INH antibodies (Liver Failure −) and sera from patients who had liver failure and tested positive for anti-INH antibodies (Liver Failure +) were diluted 1:1000. The ELISA plate was coated either with lysozyme (L) or INH-modified lysozyme (L-INH); the ratio of OD absorbance from L-INH/L was used to determine the presence of anti-INH antibodies. B) To determine if the antibodies in the sera from the 8 patients that tested positive for anti-INH antibodies were, in fact, specific for INH, the sera were preincubated with 200 μM INH for 1 hour at 4 °C (+INH) and this blocked the binding. C) The presence of anti-INH antibodies in the remaining 6 serum samples that tested positive for anti-INH antibodies by ELISA were confirmed by western blotting. Either lysozyme (L) or INH-modified lysozyme (L-INH) was loaded on a gel and transferred into a nitrocellulose membrane as described in the Methods section. Poncau S staining is shown as the loading control. Serum was diluted 1:400; + INH = serum was preincubated with 1 mM INH at 4 °C for 1 hour. Values represent Mean ± S.E. Statistically significant from control **p < 0.01.

Table 1.

Demographic data in patients with mild or no liver injury due to INH prophylaxis only.

	Gender	Country of Origin	Age	Days on INH	Peak ALT (U/L)/AST (U/L)/Bilirubin (μmol/L)	Year Enrolled In Study	Concomitant Disease	Other Medication
1	M	Philippines	56	98	144/63/11	2010		Vitamins
2	F	Vietnam	47	231	73/52/10	2010
3	M	Iraq	43	205	59/41/9	2010	Hypertension	Lisinopril
4	M	Canada	20	77	47/36/12	2011		Topical acne product – name unknown
5	M	Canada	19	64	52/28/8	2011		Vitamin C, D
6	F	Philippines	38	56		2010
7	M	Philippines	43	129		2011	Hypertension Gout	Multivitamin
8	F	South Korea	25	196		2011		Multivitamin
9	M	China	24	221		2011
10	M	Ghana	54	210		2011	Hypertension High Cholesterol	Altace, Lipitor, Lipidil, ASA
11	F	Moldova	28	238		2011		Birth Control Pill
12	M	China	29	105		2011	Ankylosing Spondylitis	Naproxen, Multivitamin
13	M	Jordan	59	98		2011	Diabetes Type 2, Asthma, COPD, Osteoporosis	Ventolin Flovent
14	M	Philippines	23	174		2011
15	F	Philippines	22	174		2011
16	M	India	30	28		2011	Migraines	Advil
17	F	Philippines	44	170		2011	Asthma	Advair 250
18	M	Pakistan	41	75		2012	Deviated septum surgery
19	F	Philippines	48	28		2012		Protein powder, Vegetable powder
20	F	Brazil	24	35		2012	Bulimia X 8 years	Vitamin B12

The serum assayed for the patients who had mild liver injury due to INH (n = 5) was when their ALT levels were highest. All the other patients had normal liver function tests before (baseline) and after treatment with INH for the duration of days specified above. M = male; F = female

Table 2.

Demographic data in patients with liver failure.

Patient ID	Gender	Race	Age	Year Enrolled In Study	Medications Taken	Total Dose	Duration	Date Last taken expressed as number of days from enrolment	Concomitant Disease
ALF-1	F	African American	65	2007	INH	300 mg/day	8 mo	−5
					Pyridoxine	50	8 mo	−5
					Motrin
ALF-2	F	African American	63	2001	RMP	300 mg/day	4 mo	−2
					INH	300 mg/day	1 mo	−136
					PZA	2000 mg/day	4 mo	−2
					HCTC	50 mg/day	36 mo	−2
ALF-3	F	African American	21	1998	INH	300 mg/day	5 mo	−16
					acetaminophen
ALF-4	M	White	30	2001	PZA	2000 mg/day	30 days	−8
					Motrin	400 mg/day		−18
					RMP	600 mg/day	30 days	−8
					Multivitamin
					acetaminophen
					INH	300 mg/day	4 mo	−44
ALF-5	F	White	59	2002	Premarin
					Darvocet
					Trazodone
					Paxil
					INH	300 mg/day	2 mo	−19
ALF-6a	F	Other	41	2004	Dilatin	200 mg/day			epilepsy, developmental delay with meningitis
					Multivitamin
					Reglan	10 mg/day		−51
					Protonix	40 mg/day		−51
					Duragesic Patch	25 mg/day		−51
					Colace
					Zoloft	50 mg/day		−51
					Carbamazepine
ALF-7	F	White	39	2006	INH	300 mg/day	6 mo	−25	migraines, irritable bowel syndrome, obesity, cholecystectomy
					Ortho Evra (Patch)	50 mg/day
					Reglan
					Atenolol
ALF-8	F	White	45	2008	INH	300 mg/day	80 days	−14	obesity
					Vitamin B6		80 days	−14
					Paxil	20 mg/day		−31
ALF-9	F	White	66	2008	Advair		12 mo		diabetes
					Vitamin B6	50 mg/day	8 mo	−47
					Tamoxifen	20 mg/day		−4
					INH	300 mg/day	8 mo	−47
					Crestor	20 mg/day	24 mo	−4
					Singulair	10 mg/day	12 mo
					Pyridoxine	50 mg/day	8 mo	−47
					Levothyroxine	137 μg/day	12 mo
					PZA	500 mg/day	8 mo	−47
					Spirvia	18 μg/day	12 mo
					RMP	300 mg/day	8 mo	−47
ALF-10	F	White	31	1998	INH		4 mo
ALF-11	M	White	28	1998	Vitamin B6		6 mo	−84
					INH		6 mo	−84
ALF-12	F	White	47	2004	PZA	250 mg/day	21 days	−5	brain cysts, postcraniotomy seizures, optic neuritis, hypertension
					RMP	600 mg/day	21 days	−5
					Pepcid
					Ethambutol	1200 mg/day	21 days	−5
					Depakote	1000 mg/day		−5
					Tegretol	1200 mg/day	36 mo	−5
					INH	300 mg/day	21 days	−5
					Metoprolol	50 mg/day		−5
					Carbamazepine	100 mg/day
					Dexamethasone	1 mg/day	21 days
ALF-13	F	Other	17	2000	Tylenol	2600 mg/day	1 day	−4
					Amoxicillin	500 mg/day	1 day	−4
					Compazine	10 mg/day	1 day	−4
					INH	300 mg/day	6 mo	−18
ALF-14	F	Asian	27	2001	INH	300 mg/day	6 mo	−7
ALF-15	M	White	22	1998	PZA	1250 mg/day	1 mo	−8
					acetaminophen
					RMP	600 mg/day	1 mo	−8
					Ethambutol	1200 mg/day	1 mo	−8
					INH	300 mg/day	21 days	−16
ALF-16	F	White	35	2003	Ethambutol		21 days	−4
					PZA		21 days	−4
					INH		21 days	−4
					RMP		21 days	−4
ALF-17	F	Asian	37	2003	INH		5 mo
					Amoxicillin		5 days	−25
					Cefzil		X days
					Acyclovir			−8
ALF-18	M	White	53	2007	INH	300 mg/day	90 days	−15	diabetes, high cholesterol, hypertension
					Diovan	160 mg/day	24 mo	−15
					Tenuate sr	75 mg/day	3 mo	−15
					Crestor	20 mg/day	7 mo	−15
ALF-19	M	Native Hawaiian/Other Pacific Islander	51	2010	PZA		3 mo	−5	chronic viral hepatitis B, hypertension
					INH		3 mo	−5
					RMP		3 mo	−5
					Ethambutol		3 mo	−5
					Atenolol	50 mg/day		−1

Bold indicates patients who tested positive for the presence of anti-INH antibodies. ALF = Acute liver failure. M = male; F = female; INH = isoniazid; RMP = rifampicin; PZA = pyrazinamide; mo = months. Concomitant disease were anything from collagen/vascular disease, chronic liver disease, endocrine/diabetes, psychiatric disease, neurological, hypertension, heart disease, renal, pulmonary disease, substance abuse, GI disease, HIV/AIDs, intravenous drug use at any time in past, other, or none.

We looked for antibodies against human liver microsomes (HLM) because in a previous study we found that they bioactivate INH (7); however, as with mouse liver homogenate high background binding was observed with control sera (data not shown). We tested CYP2E1 as an antigen to detect INH-induced antibodies because patients who carry the high activity CYP2E1 c1/c1 genotype appear to have a higher risk of developing liver injury (15). Eleven patients who had liver failure also had anti-CYP2E1 antibodies, but no anti-CYP2E1 antibodies were detected in patients who underwent INH prophylaxis (Figure 4A). Fourteen patients had antibodies against CYP2E1-INH (Figure 4B), and the concentration of these antibodies appeared to be greater than antibodies against INH or CYP2E1 because the average difference in the signal compared to control was significantly greater. This suggests that the protein is part of the epitope recognized by many of the antibodies. We also tested the ability of CYP2E1 that had been incubated with INH and NADPH to detect antibodies not detected by L-INH or CYP2E1 chemically modified by INH. However, there was no difference between CYP2E1 modified in this way and unmodified CYP2E1. This is presumably because, although we can detect covalent binding of INH to CYP2E1, under these conditions, the turnover of substrate and the amount of covalent binding would be much less than with chemical modification of CYP2E1, and it insufficient to allow detection of a difference in binding when compared with native CYP2E1. INH has also been shown to irreversibly inhibit other CYPs such as CYP3A4 and CYP2C9, which suggests that INH covalently binds to these CYPs (16). Incubation of INH with CYP2E1, CYP2C9, or CYP3A4 confirmed that INH does bind to these isozymes, and the binding requires the presence of NADPH (Figure 5) indicating that it is a metabolite that binds. Such binding could lead to antibodies against these proteins and this was also tested and confirmed. Specifically, 14 patients had anti-CYP3A4 antibodies and 10 patients had anti-CYP2C9 antibodies (Figure 4C, D). Patients either had antibodies against one of the tested proteins or they did not; therefore, in Figure 4, these two groups were separated. However, even if the results of the positive and negative patients are combined (indicated as “All Liver Failure” in the figure), there is a clear difference between patients with liver failure and either untreated controls or INH-treated patients without significant liver injury.

Figure 4.

Detection of anti-CYP antibodies in the serum from patients treated with INH. Control sera were from patients before starting INH (Baseline), sera from patients treated with INH but without significant liver injury (Prophylaxis), sera from patients with INH-induced liver failure but without anti-CYP antibodies (Liver Failure (−), sera from patients with INH-induced liver failure that do have anti-CYP antibodies (Liver Failure (+), or sera from all of the patients who had INH-induced liver failure (All Liver Failure). A) autoantibodies against CYP2E1. B) antibodies against INH-modified CYP2E1, C) autoantibodies against CYP3A4. D) autoantibodies against CYP2C9. Values represent Mean ± S.E. Statistically significant from control **p < 0.01, ***p < 0.001.

Figure 5.

In vitro covalent binding of INH to CYP2E1, CYP2C9, or Cyp3A4. Each CYP isozyme was incubated with INH (100 μM) in the presence or absence of an NADPH regenerating system. Rabbit, anti-INH antibody was used to detect covalent of INH to CYP isozyme as previously described (7).

As part of their clinical workup, some of the patients with INH-induced liver failure were tested for the presence of autoantibodies. A low titre of antinuclear antibodies (ANA) was found in 4 out of 11, anti-smooth muscle antibodies (ASMA) in 1 out of 8, anti-mitochondrial antibodies (AMA) in 1 out of 6, and anti-liver/kidney microsome antibody (KLM) in none of the 2 patients tested. The antibody profile for each patient is presented in Table 3. In addition, we looked to see if there was a correlation between patient ALT, bilirubin, or INR and the presence of anti-INH antibodies, but no correlation was observed. Data showing ALT, bilirubin, and INR levels for patients with liver failure are shown in Figure S2.

Table 3.

Pattern of antibodies in the sera from 20 patients with INH-induced liver failure.

Patient ID	Anti-INH	Anti-CYP2E1	Anti-CYP2E1-INH	Anti-CYP3A4	Anti-CYP2C9	ASMA ratio	ANA ratio	AMA ratio	KLM ratio
ALF-1		+	+	+		−	640	−	−
ALF-2	+	+	+	+	+	ND	ND	ND	ND
ALF-3	+		+	+		ND	ND	ND	ND
ALF-4				+		ND	ND	ND	ND
ALF-5						−	−	ND	ND
ALF-6		+	+	+	+	ND	ND	ND	ND
ALF-7	+		+			ND	−	ND	ND
ALF-8						ND	ND	ND	ND
ALF-9	+	+	+	+	+	−	320	ND	ND
ALF-10	+		+	+	+	−	−	−	ND
ALF-11						−	−	−	ND
ALF-12						−	160	−	ND
ALF-13		+	+	+	+	ND	ND	ND	ND
ALF-14	+	+	+	+		20	160	20	ND
ALF-15		+	+	+	+	ND	ND	ND	ND
ALF-16		+	+	+	+	ND	ND	ND	ND
ALF-17		+	+	+	+	ND	−	ND	ND
ALF-18	+	+	+	+	+	−	−	−	−
ALF-19	+	+	+	+	+	ND	ND	ND	ND

Blank or – means that the antibody was not detected, ND means not done. ALF = Acute liver failure. AMA, anti-mitochondrial antibodies; ANA, anti-nuclear antibodies; ASMA, anti-smooth muscle antibodies; LKM, anti-liver/kidney microsome.

Discussion

The mechanism of INH-induced hepatotoxicity remains poorly understood. However, it is generally not considered to be immune-mediated and has been referred to as metabolic idiosyncrasy. One reason for this belief is that, in contrast to the liver injury caused by several other drugs such as halothane, previous studies of INH-induced liver injury failed to detect anti-INH antibodies (8, 9). To the best of our knowledge, this study is the first to find anti-INH and anti-CYP autoantibodies in the serum of patients with INH-induced liver failure. Anti-INH antibodies were found in the serum of 8/19 patients who had INH-induced liver failure (Figure 3), but not in patients treated with INH who had only very mild liver injury or no increase in ALT at all. In the case of other drugs such as halothane that induce idiosyncratic liver injury and are associated with antibodies, a range of antibodies have been observed (17–19). This includes antibodies against drug-modified proteins, anti-CYP antibodies, and other autoantibodies. We demonstrated that INH is bioactivated by and covalently binds to CYP2E1, in the presence of a NADPH regenerating system (Figure 5), which is consistent with the observation that the high activity variant CYP2E1 c1/c1 genotype is associated with more severe liver injury (15). INH has been shown to be an inhibitor of 2C9, 2E1, and 3A4, which suggests that INH is bioactivated by several CYPs, and therefore these modified proteins might also induce antibody formation (20, 21). In addition to CYP2E1, we showed that INH can form covalent adducts to CYP2C9 and CYP3A4 (Figure 5). We also found that 11 patients who had INH-induced liver failure had anti-CYP2E1 antibodies, 14 patients had antibodies against CYP2E1 modified by INH, 14 patients had anti-CYP3A4 antibodies, and 10 patients had anti-CYP2C9 antibodies (Figure 4). Out of 19 patients who had liver failure, 15 (79%) had antibodies against INH, CYP3A4, 2E1, or 2C9, and most patients had antibodies to several native or INH-modified proteins (Table 3). Given the large number of proteins that are modified by the reactive metabolite of INH(7), it is quite possible that if we tested for other antibodies, all of the patients would have one or more antibody against drug-modified or native proteins. INH induces a different spectrum of antibodies in different patients (Table 3), and such heterogeneity is common in idiosyncratic drug reactions (11).

The previous two reports that did not detect anti-INH antibodies only studied patients with mild liver injury (22, 23); we also did not detect such antibodies in patients with mild liver injury (n = 5). In addition, some of the antigens that we tried gave unacceptable background binding so finding the best antigen is important. Our data clearly indicate that most cases of severe INH-induced liver injury associated with liver failure have antibodies that were induced by INH, and the absence of such antibodies in patients without injury or mild injury suggests that INH-induced liver failure is immune-mediated. The presence of anti-INH antibodies also suggests that the reactive metabolite responsible for this immune response came from bioactivation of INH itself and not acetylhydrazine as previously believed. Although the presence of anti-INH antibodies only in patients with INH-induced liver failure suggests an immune-mediated mechanism, we cannot rule out the possibility that the liver failure was instrumental in the induction of these antibodies. In addition, some of those patients who had anti-INH antibodies were also being treated with other drugs that could contribute to the presence of autoantibodies. However, in most cases the timing made INH or INH/pyrazinamide more likely, and there were two clean cases of INH-induced liver failure (patient ID ALF-10 and 14), which were associated with the presence of anti-INH antibodies and autoantibodies similar to most of the other samples (Table 2,3).

There are other data that strongly suggest that INH-induced liver injury is immune-mediated, and further that it is INH rather than acetylhydrazine that is recognized by the immune cells involved in INH-induced liver injury. Specifically, Warrington found a positive lymphocyte transformation test (LTT) when lymphocytes from cases of mild INH-induced liver injury were incubated with INH-modified protein but not with INH itself; however, more severe cases of DILI also had a positive LTT to INH (23, 24). Maria and Victorino also reported a positive LTT test from a patient with INH-induced hepatotoxicity (25). The liver injury caused by other drugs such as halothane and tienilic acid can be associated with a variety of antibodies (17–19). Some of these antibodies are against drug-modified proteins and some are autoantibodies against native proteins, especially the CYP that formed the reactive metabolite, and the pattern varies from patient to patient. INH also commonly induces the production of autoantibodies, sometimes resulting in a lupus-like syndrome (26). This panel of anti-drug and anti-CYP antibodies appear to have sufficient sensitivity and specificity that they could be useful in causality assessment where INH is one of possible causes of severe liver injury. It suggests that anti-CYP antibodies may be present in many other cases of idiosyncratic drug-induced liver injury, although this is less likely for drugs in which the degree of covalent binding to CYPs is insufficient to lead to suicide inhibition of the enzymes.

Although the presence of anti-INH and anti-CYP autoantibodies only in patients with INH-induced liver failure suggests that INH-induced liver injury is immune-mediated, there is no evidence that these antibodies are actually responsible for the liver injury. They could be a result of an immune response to the liver injury or even an attempt to resolve the immune response; many immune cells have Fc antibody receptors, and intravenous immunoglobulin is used to treat immune-mediated idiosyncratic drug reactions such as toxic epidermal necrolysis. However, these antibodies do indicate that severe INH-induced liver injury involves the immune system. The positive LTT indicates the presence of sensitized lymphocytes, which may represent the major mechanism of liver injury. Combining our data with the Warrington data suggests that mild cases of INH-induced liver injury resolve with immune tolerance, and it is only when this immune tolerance fails and the immune response spreads that more severe liver injury occurs.

Supplementary Material

Supp Fig S1-S2

Figure S1: Absence of anti-INH antibodies in samples from patients before starting INH treatment or after treatment with INH in the absence of significant liver injury compared with sera without anti-INH antibodies from patients with INH-induced liver failure. Lysozyme (L) or INH-modified lysozyme (L-INH) was loaded on a western blot, and serum from patients diluted 1:400 was used as primary antibody. #4-0 = serum from a patient at baseline (ALT = 23), #4-5 = serum from the same patient at revisit number 5 from initiation of INH treatment and a small increase in ALT (ALT = 49). Likewise, #10-0 was from a patient before treatment (ALT = 12) and #10-5 was the same patient at revisit number 5 (ALT = 15); #11-0 was before treatment (ALT = 11) and #11-5 was at revisit number 5 (ALT = 24); #27-0 was before treatment (ALT = 15) and #27-1 was at revisit number 1 (ALT = 16). ALF-6, ALF-11, ALF-5, and ALF-16 are samples from patients with INH-induced liver failure who tested negative for anti-INH antibodies by ELISA. In these blots, the contrast was adjusted to a maximum to show the background and illustrate that there were no bands that could be attributed to anti-INH antibodies.

Figure S2: ALT, bilirubin and international normalized ration (INR) for patients who had liver failure due to INH as a function of days after initial hospitalization. In Y-axis, values are expressed as upper limit of normal (ULN); the upper limit of normal for each parameter was considered to be: 40 U/L for ALT, 1.2 mg/dL for bilirubin and 1.5 for INR. A) Patients who had anti-INH antibodies. B) Patients negative for anti-INH antibodies.

Abbreviations

ALFSG

Acute Liver Failure Study Group

ALT

alanine aminotransferase

AMA

anti-mitochondrial antibodies

ANA

anti-nuclear antibodies

ASMA

anti-smooth muscle antibodies

CYP

cytochrome P450

CYP2E1-INH

CYP2E1 chemically modified to mimic covalent binding by the reactive metabolite of INH

ELISA

enzyme-linked immunosorbent assay

HLM

human liver microsomes

INA-NHS

N-hydroxysuccinimide ester of isonicotinic acid

INH

isoniazid

L

lysozyme

L-INH

lysozyme chemically modified to mimic binding by the reactive metabolite of INH

LKM

anti-liver/kidney microsome antibody

LTT

lymphocyte transformation test

TBST

tris-buffered saline with Tween

TBM

3,3′,5,5′-tetramethylbenzidine