Evaluation of patterns of liver toxicity in patients on antiretroviral and anti-tuberculosis drugs: a prospective four arm observational study in ethiopian patients

Objectives

To evaluate the incidence, type, severity and predictors of antiretroviral and/or anti-tuberculosis drugs induced liver injury (DILI).

Methods

A total of 1,060 treatment naive patients were prospectively enrolled into four treatment groups: HIV patients receiving efavirenz based HAART alone (Arm-1); TB-HIV co-infected patients with CD4≤200 cells/μL, receiving concomitant rifampicin based anti-TB and efavirenz based HAART (Arm-2); TB-HIV co-infected patients with CD4>200 cells/μL, receiving anti-TB alone (Arm-3); TB patients taking rifampicin based anti-TB alone (Arm-4). Liver enzyme levels were monitored at baseline, 1st, 2nd, 4th, 8th, 12th and 24th weeks during treatment. CD4 and HIV viral load was measured at baseline, 24th and 48th weeks. Data were analyzed using multivariate Cox Proportional Hazards Model.

Results

A total of 159 patients (15%) developed DILI with severity grades 1, 2, 3 and 4 of 53.5%, 32.7%, 11.3% and 2.5% respectively. The incidence of cholestatic, hepatocellular or mixed pattern was 61%, 15% and 24%, respectively. Incidence of DILI was highest in Arm-2 (24.2%)>Arm-3 (10.8%)>Arm-1 (8.8%)>Arm-4 (2.9%). Concomitant anti-TB-HIV therapy increased the risk of DILI by 10-fold than anti-TB alone (p<0.0001). HIV co-infection increased the risk of anti-TB DILI by 4-fold (p = 0.004). HAART associated DILI was 3-fold higher than anti-TB alone, (p = 0.02). HAART was associated with cholestatic and grade 1 DILI whereas anti-TB therapy was associated with hepatocellular and grade ≥ 2. Treatment type, lower CD4, platelet, hemoglobin, higher serum AST and direct bilirubin levels at baseline were significant DILI predictors. There was no effect of DILI on immunologic recovery or virologic suppression rate of HAART.

Conclusion

HAART associated DILI is mainly cholestatic and mild whereas hepatocellular or mixed pattern with high severity grade is more common in anti-tuberculosis DILI. TB-HIV co-infection, disease severity and concomitant treatment exacerbates the risk of DILI.

Keratin-18 and microRNA-122 complement alanine aminotransferase as novel safety biomarkers for drug-induced liver injury in two human cohorts

BACKGROUND & AIMS

There is a demand for more sensitive, specific and predictive biomarkers for drug-induced liver injury (DILI) than the gold standard used today, alanine aminotransferase (ALT). The aim of this study was to qualify novel DILI biomarkers (keratin-18 markers M65/M30, microRNA-122, glutamate dehydrogenase and alpha-foetoprotein) in human DILI.

METHODS

Levels of the novel biomarkers were measured by enzyme-linked immunosorbent assay or real-time quantitative reverse-transcription PCR (qRT-PCR) in two human DILI cohorts: a human volunteer study with acetaminophen and a human immunodeficiency virus (HIV)/tuberculosis (TB) study.

RESULTS

In the acetaminophen study, serum M65 and microRNA-122 levels were significantly increased at an earlier time point than ALT. Furthermore, the maximal elevation of M65 and microRNA-122 exceeded the increase in ALT. In the HIV/TB study, all the analysed novel biomarkers increased after 1 week of treatment. In contrast to ALT, the novel biomarkers remained stable in a human cohort with exercise-induced muscular injury.

CONCLUSIONS

M65 and microRNA-122 are potential biomarkers of DILI superior to ALT with respect to sensitivity and specificity.

Isoform-specific alanine aminotransferase measurement can distinguish hepatic from extrahepatic injury in humans

Serum alanine aminotransferase (ALT) is used as a clinical marker to detect hepatic damage and hepatoxicity. Two isoforms of ALT have been identified, ALT1 and ALT2, which have identical enzymatic capacities and are detected simultaneously in human serum/plasma using classical clinical chemical assays. Differences exist in the expression patterns of the ALT1 and ALT2 proteins in different organs which suggest that changes in the proportion of ALT1 and ALT2 in plasma may arise and reflect damage to different human organs. However, this has not been previously studied due to the lack of a selective methodology that can quantify both ALT1 and ALT2 isoforms in the total ALT activity normally measured in clinical samples. To the best of our knowledge, our current study reveals for the first time, that under 3 different conditions of liver damage (non-alcoholic fatty liver disease, hepatitis C and during liver surgery) the leakage of ALT1 activity into plasma greatly exceeds that of ALT2, and that the measurement of ALT1 during liver damage is equal to the measurement of total ALT activity. By contrast, during skeletal muscle injury, induced in volunteers by physical exertion, the leakage of ALT2 exceeds that of ALT1 and the proportion of circulating ALT isoforms changes accordingly. The ALT isoform changes occurring in plasma reflect previously demonstrated relative contents of ALT1 and ALT2 activities in human liver and skeletal muscle. These data suggest that assessing the percentage contribution of ALT1 and ALT2 activities to total ALT activity in plasma may distinguish hepatic from extrahepatic injury using the same standard analytical platform.

The development of in vitro culture methods to characterize primary T-cell responses to drugs

Adverse drug reactions represent a major stumbling block to drug development and those with an immune etiology are the most difficult to predict. We have developed an in vitro T-cell priming culture method using peripheral blood from healthy volunteers to assess the allergenic potential of drugs. The drug metabolite nitroso sulfamethoxazole (SMX-NO) was used as a model drug allergen to establish optimum assay conditions. Naive T cells were cocultured with monocyte-derived dendritic cells at a ratio of 25:1 in the presence of the drug for a period of 8 days, to expand the number of drug-responsive T cells. The T cells were then incubated with fresh dendritic cells, and drug and their antigen responsiveness analyzed using readouts for proliferation, cytokine secretion, and cell phenotype. All five volunteers showed dose-dependent proliferation as measured by 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester content and by (3)H-thymidine uptake. CD4 T cells that had divided in the presence of SMX-NO had changed from a naive phenotype (CD45RA+) to a memory phenotype (CD45RO+). These memory T cells expressed the chemokine receptors CCR2, CCR4, and CXCR3 suggesting a mixture of T(H)1 and T(H)2 cells in the responding population, with a propensity for homing to the skin. Drug stimulation was also associated with the secretion of a mixture of T(H)1 cytokines (interferon γ) and T(H)2 cytokines (interleukin [IL]-5 and IL-13) as detected by ELISpot. We are currently developing this approach to investigate the allergenic potential of other drugs, including those where an association between specific human leucocyte antigen alleles and susceptibility to an immunological reaction has been established.

What are the best reference values for a normal serum alanine transaminase activity (ALT)? Impact on the presumed prevalence of drug induced liver injury (DILI)

BACKGROUND

In clinical research, the definition of the upper limit of normal (ULN) is rarely detailed. For alanine transaminase (ALT), there are several definitions of ULN-ALT but no recognized global reference. Furthermore the inter-laboratory variability of results expressed using ULN-ALT is higher than using the actual value of ULN expressed in IU/L. Regulatory agencies still use ULN-ALT for the definition of drug adverse events such as drug induced liver disease (DILI).

METHODS

We applied two extreme definitions of ULN-ALT (26 and 66 IU/L) in two populations with different liver disease risk: 7463 consecutive volunteers representative a low risk population, and 6865 consecutive patients hospitalized in a tertiary referral center. The same assay technique was used for both populations on fresh plasma in the same laboratory.

RESULTS

In the low risk population the liver disease estimates ranged from 0% to 1.99% according to ULN-ALT definition and gender; prevalence of liver disease as defined by Temple's criteria (3×ULN) decreased significantly with increased ULN-ALT threshold and prevalence of liver disease was lower in females compared to males (all P<0.001). In the high risk population the estimates of liver disease prevalence ranged from 0.78% to 15.85%; disease prevalence using both Temple's corollary and Hy's law criteria (3×ULN-ALT and bilirubin >34 μmol/L) decreased significantly with increased ULN-ALT threshold and females compared to males. In the low risk population the two major factors associated with ULN variability were gender and BMI.

CONCLUSION

Artificial statistical modifications of the procedures chosen for the ULN-ALT definition change dramatically the prevalence of DILI estimates. A consensus in liver disease definitions seems mandatory for DILI studies in order to prevent misleading conclusions.

Application of Dolichos biflorus in immunoassay detection of kidney collecting duct biomarkers

Currently there are no biomarkers for detecting collecting duct damage in man. Antibodies to several collecting duct-specific antigens exist but sandwich assays have been difficult to establish due to the need for two different antibodies to the same protein. We hypothesized that a collecting duct-specific lectin could be used in combination with a collecting duct-specific antibody to negate the need for two different antibodies. The collecting duct specificity of selected antibodies (NiCa II 13C2, Pap XI 3C7, HuPaP VII 2B11 and aquaporin 2), was verified by immunohistochemistry. Aquaporin 2 and Pap XI 3C7 were used successfully in setting up assays with the lectin Dolichos biflorus, using the Meso Scale Discovery (MSD) platform. Antigen expression was highest in the papillae of rat and human kidney (corresponding to the greatest density of collecting ducts) and was also present in normal urine. We propose that further qualification and validation would lead to an assay for detecting collecting duct damage in man.

A systems biology approach to understanding elevated serum alanine transaminase levels in a clinical trial with ximelagatran

Ximelagatran was developed for the prevention and treatment of thromboembolic conditions. However, in long-term clinical trials with ximelagatran, the liver injury marker, alanine aminotransferase (ALT) increased in some patients. Analysis of plasma samples from 134 patients was carried out using proteomic and metabolomic platforms, with the aim of finding predictive biomarkers to explain the ALT elevation. Analytes that were changed after ximelagatran treatment included 3-hydroxybutyrate, pyruvic acid, CSF1R, Gc-globulin, L-glutamine, protein S and alanine, etc. Two of these analytes (pyruvic acid and CSF1R) were studied further in human cell cultures in vitro with ximelagatran. A systems biology approach applied in this study proved to be successful in generating new hypotheses for an unknown mechanism of toxicity.

A solution to the 1D NMR alignment problem using an extended generalized fuzzy Hough transform and mode support

This paper approaches the problem of intersample peak correspondence in the context of later applying statistical data analysis techniques to 1D 1H-nuclear magnetic resonance (NMR) data. Any data analysis methodology will fail to produce meaningful results if the analyzed data table is not synchronized, i.e., each analyzed variable frequency (Hz) does not originate from the same chemical source throughout the entire dataset. This is typically the case when dealing with NMR data from biological samples. In this paper, we present a new state of the art for solving this problem using the generalized fuzzy Hough transform (GFHT). This paper describes significant improvements since the method was introduced for NMR datasets of plasma in Csenki et al. (Anal Bioanal Chem 389:875-885, 15) and is now capable of synchronizing peaks from more complex datasets such as urine as well as plasma data. We present a novel way of globally modeling peak shifts using principal component analysis, a new algorithm for calculating the transform and an effective peak detection algorithm. The algorithm is applied to two real metabonomic 1H-NMR datasets and the properties of the method are compared to bucketing. We implicitly prove that GFHT establishes the objectively true correspondence. Desirable features of the GFHT are: (1) intersample peak correspondence even if peaks change order on the frequency axis and (2) the method is symmetric with respect to the samples.

The impact of CYP2E1 on the development of alcoholic liver disease as studied in a transgenic mouse model

BACKGROUND/AIMS

CYP2E1 metabolizes ethanol, generates reactive oxygen species, and is suggested to be important for development of alcoholic liver disease. The present study aims to evaluate the role of CYP2E1 in combination with ethanol for development of alcoholic liver disease using mice transgenic for the human CYP2E1 gene.

METHODS

Changes in hepatic gene expression were monitored in controls and mice transgenic for human CYP2E1, treated with ethanol or isocaloric dextrose intragastrically for 4 weeks, and related to pathology using Affymetrix microarrays and TaqMan RealTime PCR.

RESULTS

Presence of the CYP2E1 transgene increased liver injury and increased expression of stress related genes. Microarray analyses revealed the expression of structural genes, particularly cytokeratin 8 and 18, to be highly related to pathology.

CONCLUSIONS

This in vivo study confirms several findings regarding CYP2E1 and alcohol previously found only in vitro. These results provide in vivo evidence that CYP2E1 overexpression aggravates hepatic injury, and suggest that expression of cytokeratins 8 and 18 can be considered as biomakers for the progression of alcoholic liver disease.

Feature detection and alignment of hyphenated chromatographic-mass spectrometric data. Extraction of pure ion chromatograms using Kalman tracking

In this paper we present a new method, called TracMass, for analyzing data obtained using hyphenated chromatography-mass spectrometry (XC/MS). The method uses a Kalman filter to extract pure, noise-free ion chromatograms by exploiting the latent second order structure in the XC/MS data. TracMass differs from current state-of-the-art methodologies, which extract chromatograms by binning along the m/z axis and further processes the data in various ways, e.g. by baseline correction, component detection algorithm, peak detection, and curve resolution to extract molecular features. The proposed method was validated by analyzing two plasma datasets: one derived from 99 quality control samples where TracMass extracted 8880 Pure Ion Chromatograms (PICs) present in > or =90 of the samples. The second dataset was spiked with two different internal standard mixtures to test differential expression analysis. Here TracMass found 20000 PICs present in 10 samples, all differentially expressed analytes, and also a previously unreported discriminating metabolite. Finding as many PICs as possible is in this context essential to ensure that even small differentiating features are found (if they exist). The resulting data representation from TracMass (PICs) can be used directly for statistical analysis, and the method is fast (approximately 5min/sample), with few adjustable parameters.

Correlation network analysis for data integration and biomarker selection

High-throughput biomolecular profiling techniques such as transcriptomics, proteomics and metabolomics are increasingly being used in in vivo studies to recognize and characterize effects of xenobiotics on organs and systems. Of particular interest are biomarkers of treatment-related effects which are detectable in easily accessible biological fluids such as blood. A fundamental challenge in such biomarker studies is selecting among the plethora of biomolecular changes induced by a compound and revealed by molecular profiling, to identify biomarkers which are exclusively or predominantly due to specific processes. In this work we present a cross-compartment correlation network approach, involving no a priori supervision or design, to integrate proteomic, metabolomic and transcriptomic data for selecting circulating biomarkers. The case study we present is the identification of biomarkers of drug-induced hepatic toxicity effects in a rodent model. Biomolecular profiling of both blood plasma and liver tissue from Wistar Hannover rats administered a toxic compound yielded many hundreds of statistically significant molecular changes. We exploited drug-induced correlations between blood plasma analytes and liver tissue molecules across study animals in order to nominate selected plasma molecules as biomarkers of drug-induced hepatic alterations of lipid metabolism and urea cycle processes.

Analysis of time-related metabolic fluctuations induced by ethionine in the rat

The time-course of metabolic events following response to a model hepatotoxin ethionine (800 mg/kg) was investigated over a 7 day period in rats using high-resolution (1)H NMR spectroscopic analysis of urine and multivariate statistics. Complementary information was obtained by multivariate analysis of (1)H MAS NMR spectra of intact liver and by conventional histopathology and clinical chemistry of blood plasma. (1)H MAS NMR spectra of liver showed toxin-induced lipidosis 24 h postdose consistent with the steatosis observed by histopathology, while hypertaurinuria was suggestive of liver injury. Early biochemical changes in urine included elevation of guanidinoacetate, suggesting impaired methylation reactions. Urinary increases in 5-oxoproline and glycine suggested disruption of the gamma-glutamyl cycle. Signs of ATP depletion together with impairment of the energy metabolism were given from the decreased levels in tricarboxylic acid cycle intermediates, the appearance of ketone bodies in urine, the depletion of hepatic glucose and glycogen, and also hypoglycemia. The observed increase in nicotinuric acid in urine could be an indication of an increase in NAD catabolism, a possible consequence of ATP depletion. Effects on the gut microbiota were suggested by the observed urinary reductions in the microbial metabolites 3-/4-hydroxyphenyl propionic acid, dimethylamine, and tryptamine. At later stages of toxicity, there was evidence of kidney damage, as indicated by the tubular damage observed by histopathology, supported by increased urinary excretion of lactic acid, amino acids, and glucose. These studies have given new insights into mechanisms of ethionine-induced toxicity and show the value of multisystem level data integration in the understanding of experimental models of toxicity or disease.

Proof of principle of a generalized fuzzy Hough transform approach to peak alignment of one-dimensional 1H NMR data

In metabolic profiling, multivariate data analysis techniques are used to interpret one-dimensional (1D) 1H NMR data. Multivariate data analysis techniques require that peaks are characterised by the same variables in every spectrum. This location constraint is essential for correct comparison of the intensities of several NMR spectra. However, variations in physicochemical factors can cause the locations of the peaks to shift. The location prerequisite may thus not be met, and so, to solve this problem, alignment methods have been developed. However, current state-of-the-art algorithms for data alignment cannot resolve the inherent problems encountered when analysing NMR data of biological origin, because they are unable to align peaks when the spatial order of the peaks changes-a commonly occurring phenomenon. In this paper a new algorithm is proposed, based on the Hough transform operating on an image representation of the NMR dataset that is capable of correctly aligning peaks when existing methods fail. The proposed algorithm was compared with current state-of-the-art algorithms operating on a selected plasma dataset to demonstrate its potential. A urine dataset was also processed using the algorithm as a further demonstration. The method is capable of successfully aligning the plasma data but further development is needed to address more challenging applications, for example urine data.

Reliable Profile Detection in Comparative Metabolomics

A strategy for processing of metabolomic GC/MS data is presented. By considering the relationship between quantity and quality of detected profiles, representative data suitable for multiple sample comparisons and metabolite identification was generated. Design of experiments (DOE) and multivariate analysis was used to relate the changes in settings of the hierarchical multivariate curve resolution (H-MCR) method to quantitative and qualitative characteristics of the output data. These characteristics included number of resolved profiles, chromatographic quality in terms of reproducibility between analytical replicates, and spectral quality defined by purity and number of spectra containing structural information. The strategy was exemplified in two datasets: one containing 119 common metabolites, 18 of which were varied according to a DOE protocol; and one consisting of rat urine samples from control rats and rats exposed to a liver toxin. It was shown that the performance of the data processing could be optimized to produce metabolite data of high quality that allowed reliable sample comparisons and metabolite identification. This is a general approach applicable to any type of data processing where the important processing parameters are known and relevant output data characteristics can be defined. The results imply that this type of data quality optimization should be carried out as an integral step of data processing to ensure high quality data for further modeling and biological evaluation. Within metabolomics, this degree of optimization will be of high importance to generate models and extract biomarkers or biomarker patterns of biological or clinical relevance.