i just made love
Application of High-Throughput Sequencing to
Circulating microRNAs Reveals Novel Biomarkers for Drug-Induced Liver Injury
Julian Krauskopf*,1,Florian Caiment*,Sandra M.Claesn*,Kent J.Johnson †,Roscoe L.Warner †,Shelli J.Schomaker ‡,Deborah A.Burt ‡,Jiri Aubrecht ‡,and Jos C.Kleinjans*
detailly*Department of Toxicogenomics,Maastricht University,Maastricht 6200MD,The Netherlands,†Pathology
Department,University of Michigan,Ann Arbor,Michigan 48109and ‡Drug Safety Rearch and Development,Pfizer,Inc.,Groton,Connecticut 06340
latour
1
To whom correspondence should be addresd at Department of Toxicogenomics,Maastricht University,PO Box 616,6200MD Maastricht,The Netherlands.E-mail:j.krauskopf@maastrichtuniversity.nl.
The authors certify that all rearch involving human subjects was done under full compliance with all government policies and the Helsinki Declaration.
ABSTRACT
Drug-induced liver injury (DILI)is a leading cau of acute liver failure and the major reason for withdrawal of drugs from the market.Preclinical evaluation of drug candidates has failed to detect about 40%of potentially hepatotoxic compounds in humans.At the ont of liver injury in humans,currently ud biomarkers have difficulty differentiating vere DILI from mild,and/or predict the outcome of injury for individual subjects.Therefore,new biomarker approaches for predicting and diagnosing DILI in humans are urgently needed.Recently,circulating microRNAs (miRNAs)such as miR-122and miR-192have emerged as promising biomarkers of liver injury in preclinical species and in DILI patients.In this study,we focud on examining global circulating miRNA profiles in rum samples from subjects with liver injury caud by accidental acetaminophen (APAP)overdo.Upon applying next generation high-throughput quencing of small RNA libraries,we identified 36miRNAs,including 3novel miRNA-like small nuclear RNAs,which were enriched in the rum of APAP
overdod subjects.The t comprid miRNAs that are functionally associated with liver-specific biological process and relevant to APAP toxic mechanisms.Although more patients need to be investigated,our study suggests that profiles of circulating miRNAs in human rum might provide ad
muscles
ditional biomarker candidates and possibly mechanistic information relevant to liver injury.
Key words :acetaminophen overdo;APAP;miRNA;snRNA;rum;miR-122
Drug-induced liver injury (DILI)is a rious complication in drug therapy and a primary cau of drug failures during clinical trials and after market introduction (Giacomini et al.,2007).In clinical practice,DILI makes up for 50%of acute liver failure cas (Kaplowitz,2005).During preclinical development,many compounds are screened out,bad on liver injury findings in animal studies (Chang and Schiano,2007).Despite this rigorous screening,it has been reported that up to 40%of hepatotoxic compounds in humans remain undetected in animal studies that utilize conventional histopathology and clinical pathology endpoints (Xu et al.,2004;Zhang et al.,2012).
Conventional biomarkers of DILI relate either to reduced cell integrity detected as leakage of proteins into the circulation (eg,alanine aminotransfera (ALT))or to functional disability of hepatocytes to produce total bilirubin (TBIL)(Amacher et al.,2013).The risk of DILI is mainly assd using Hy’s law
V
C The Author 2014.Published by Oxford University Press on behalf of the Society of Toxicology.All rights rerved.For Permissions,plea e-mail:journals.
268
TOXICOLOGICAL SCIENCES ,143(2),2015,268–276
doi:10.1093/toxsci/kfu232
Advance Access Publication Date:October 29,
大连外语培训学校2014
(Temple,2006)which has been widely ud for diagnosis of liver injury.However,it remains difficult to differentiate between -vere and mild DILI cas and,to predict clinical outcomes. While drug development continues to be slowed or halted due to ALT elevations in clinical trials the increas
of ALT rum activities may only indicate a transient hepatocellular injury that does not progress to vere DILI.Furthermore,significant increas in TBIL levels are obrved only after a vere liver in-jury has already occurred(for review e(Amacher et al.,2013)). Therefore,development of new biomarker approaches comple-menting the conventional paradigm is esntial for improve-ment of diagnosis and management of DILI.
Recently,circulating microRNAs(miRNAs)have emerged as potential biomarkers of organ damage,including the liver (Szabo and Bala,2013).The small noncoding RNAs(snRNA)of about22nucleotides(nt)are involved in posttranscriptional gene regulation(Bartel,2004).Importantly,circulating miRNAs are highly stable in a wide range of biofluids such as blood (Turchinovich et al.,2012).Some miRNAs are tissue-specific (Liang et al.,2007)and some are conrved across preclinical species(Landgraf et al.,2007).Two liver-enriched miRNAs,miR-122and miR-192,have previously been identified as promising biomarkers of liver injury.In blood samples of mice injected with a single do of acetaminophen,the two miRNAs indi-cated a higher nsitivity for DILI in comparison to the classical biomarker ALT(Wang et al.,2009).Subquently,in acetamino-phen-induced acute liver injury patients,the potential u of miR-122and miR-192as biomarkers of human drug-induced liver necrosis was confirmed(Starkey Le
wis et al.,2011). Furthermore,veral studies indicated miR-122and other miRNAs as promising biomarkers of drug-induced acute liver necrosis(Antoine et al.,2013),chronic hepatitis B(Zhang et al., 2012),and hepatocellular carcinoma(Xu et al.,2011).
Previous studies on circulating miRNAs utilized microarray technology or qPCR and were limited to analyzing known and publicly available miRNA quences.Due to vast improvements in quencing technology,high-throughput quencing now prents a more comprehensive,nsitive,and cost-effective approach which enables the quantitative analysis of microRNAs at relatively low expression levels across the whole miRNA ge-nome,as well as the identification of quence variations and novel miRNAs(Morin et al.,2008).Conquently,by utilizing high-throughput quencing we identified33miRNAs,and three novel miRNA-like snRNAs,enriched in the blood of acet-aminophen-overdod patients.
MATERIALS AND METHODS
Sample collection and RNA preparation.The blood samples from healthy subjects and subjects with accidental acetaminophen overdo were collected at the University of Michigan health care system under an approved IRB(2000-005).The lection cri-teria for healthy subjects included normal levels
of ALT,TBIL, aspartate aminotransfera(AST),alkaline phosphata(ALK), gluco(Gluc),blood urea nitrogen(BUN),rum creatinine (SCRE)and creatine kina(CK)and abnce of any liver impair-ment in medical history.The samples from acetaminophen-overdod subjects were collected from leftover diagnostic samples taken during the N-acetyl cysteine(NAC)treatment of APAP poisoning.A final t of24samples included samples from6healthy(3females and3males,average age33years) and6acetaminophen-overdod(3females and3males, average age33years)subjects.A single blood sample from each healthy subject was ud for analysis.A ries of samples (2–6samples)covering a time cour of variable length for each individual subject with APAP overdo was analyzed.Serum samples were recovered from rum parator tubes following centrifugation of whole blood at3000Âg for10min at room temperature.Serum samples were kept at4 C for up to72h before aliquots were frozen atÀ80 C and stored until shipped for analysis.ALT,AST,ALK,TBIL,Gluc,BUN,SCRE,and CK were measured by Siemens Advia1800chemistry autoanalyzer with Siemens chemistry reagents.Glutamate dehydrogena(GLDH) and malate dehydrogena(MDH)were measured in rum on the Siemens ADVIA2400Automated Chemistry System.GLDH was measured using commercially available kits(Randox Labs Ltd,Roche).MDH was determined according to Bergmeyer and Bernt(1983).Serum miRNA was isolated using the miRNeasy Mini Kit(QUIAGEN,Valencia,CA)and the quality and amount of isolated miRNAs was evaluated using the Agilent Small RNA Kit.
High-throughput quencing.The isolated RNA was prepared for quencing by using the TruSeq Small RNA-Seq Preparation Kit (Illumina,Hayward,CA)and then quenced using the Illumina HiSeq2000quencing platform according to the manufac-turer’s protocol(GEO accession:GSE59565).Sequencing reads of 100nt were retrieved in fastq format and were procesd using miRDeep2(version2.0.0.5)(Friedlander et al.,2012).Briefly,the quences were30adapter-trimmed and lected for reads with a minimum length of18nt.Remaining reads were collapd to distinct quences and mapped to the human genome(hg19); bad on this mapping,novel miRNA quences have been pre-dicted.The predicted miRNAs were then added to the miRNA precursor and mature quences which were retrieved from miRBa(relea20)(Griffiths-Jones,2006).Subquently,the miRNA expression was quantified in the24samples.
The expression output was analyzed using the open-source software R(version2.15.2)and Bioconductor(Gentleman et al., 2004).For the detection of differentially expresd miRNAs,the DESeq package(version1.10.1)(Anders and Huber,2010)was ud;after normalization of the quantitative miRNA levels,the negative binominal test was ud to produce a list of differen-tially expresd miRNAs(controls vs first time-points after acetaminophen overdo).MicroRNAs were considered significantly differentially expresd at a p value below.05 (adjusted for multiple testing u
sing Benjamini-Hochberg method as implemented in DESeq).A cond significance crite-rion,to compensate for bias introduced by very low abundant quences,was a minimum mean of20read counts in either the healthy or the overdod samples(Kang et al.,2013).
Tissue enrichment.A publicly available small RNA quencing datat derived from two human individuals across8tissues (lung,kidney,skeletal muscle,heart,pancreas,frontal orbital gyrus,spleen,and liver tissue)(Faghihi et al.,2010)was exploited to reveal possible tissue enrichment of the detected circulating miRNAs.Bad on the expression of the miRNAs in each tissue,determined using miRDeep2,a tissue enrichment score was calculated.This score was defined as the sum of the rum level of a particular miRNA in a tissue divided by the sum of the rum levels of this miRNA in all tissues(van de Bunt et al.,2013).
RT-qPCR validation.For the purpo of validating the quencing results,quantitative PCR analysis was performed for miR-122, miR-483,the two isoforms of miR-23a and the novel miRNA-like molecules pmiR-U2-snRNA and pmiR-chr19-5802.All PCRs
KRAUSKOPF ET AL.|269
were carried out using the miRCURY LNA TM Universal RT miRNA PCR(Exiqon,Vedbaek,Denmark)i
n duplicate according to the manufacturer’s manual v.5.3.The miRNA miR-7i was lected for normalization becau the rum levels of this miRNA had the lowest standard deviation across all24samples. The quantification of each miRNA relative to the miR-7i was calculated using the equation:N¼2ÀD Ct,D Ct¼Ct miRNAÀC miR-7i (Livak and Schmittgen,2001).
so thatRESULTS
Small RNA-quencing
The quencing of RNA samples isolated from rum of6 healthy and6APAP-overdod subjects,yielded on average9.5 million(standard deviation5.4million)unfiltered quencing reads.The relatively low levels of miRNAs in blood caud higher occurrence of adapter-adapter ligations within the quencing reaction(Williams et al.,2013);hence,a relatively high amount of quencing adapters was quenced. Conquently,within the process of filtering many reads were discarded but on average2.1million(standard deviation1.5mil-lion)high-quality reads were obtained.To identify circulating miRNAs in rum of subjects in our study,the quencing reads were aligned to the1872publicly available miRNA precursor quences.On average,20%(standard deviation10%)of the quencing reads could be assigned to known miRNAs.A total of355miRNAs were detected acro
ss all rum samples (Supplementary Table1).The remaining reads were related to other RNA species such as degraded mRNA or unknown small RNA quences.To validate the miRNA levels detected via next-generation quencing,we have performed RT-qPCR for a small subt of miRNAs and samples;in accordance with next-gener-ation quencing data,significantly higher levels of miR-122, miR-483,and the miR-23a isomiR were obrved in the samples from acetaminophen-overdod subjects than in samples from healthy subjects.Furthermore,we were able to validate the prence of the two novel miRNA-like snRNAs pmiR-U2-snRNA and pmiR-chr19-5802as well as their relative higher levels in the samples from the acetaminophen-overdod subjects (Supplementary Fig.1).nontoxic
MicroRNA profiles differentiate healthy and acetaminophen-overdod subjects
To compare profiles of circulating miRNAs across the whole sample t,we calculated the Pearson correlation using levels of all miRNAs detected in each sample(Fig.1).The analysis of all pairwi correlations showed a high correlation among cir-culating miRNA profiles in a group of six healthy subjects(1–6). On the other hand,the majority of miRNA profiles in APAP-overdod subjects measured at the first day after the overdo (7–12),displayed a low correlation with profiles from healthy subjects.Furthermore,the circulating miRNA profiles detected at late time points after acetam
inophen overdo,showed improved correlation with the profiles from healthy subjects indicating a recovery from the APAP poisoning.
Bad on the normalized miRNAs rum levels,a negative binomial test of the6controls(1–6)and the6earliest samples from overdod subjects(7–12)resulted in identification of36 miRNAs with altered rum levels in APAP-overdod subjects that included3newly predicted miRNAs.Of the detected miRNAs,8(miR-107,miR-122,miR-130a,miR-148a,miR-192, miR-22,miR-27b,and miR-30a)were previously found to be enriched in plasma of acetaminophen-overdod mice(Wang et al.,2009).The two most studied miRNA-bad biomarker candidates,miR-122(Fig.2A)and miR-192(Fig.2B),showed low baline levels with little variability in healthy subjects.As expected,both miRNA levels in rum were sharply incread at the ont of acetaminophen poisoning in all6subjects(7–12)fol-lowed by a steady decrea during day1and day2after admis-sion of APAP.The circulating miRNA levels had returned to baline after day3.A comparable time cour pattern was obrved for the majority of the other34miRNAs.An overview of the expression profile can be found in Supplementary Figure2. Association of detected miRNAs with conventional biomarkers
To compare the acetaminophen-induced time cour respon of the36miRNAs identified in our study,with the respon of conventional biomarkers of liver injury,we performed a cluster analysis ba
d on Pearson correlation results(Fig.3).The analy-sis revealed4distinct clusters.The red and black clades formed a cluster that is associated with conventional biomarkers of hepatocyte integrity.In the red clade,a group of12miRNAs had a similar respon pattern in the acetaminophen-overdod patients as ALT,AST as well as an emerging biomarker of liver injury,namely MDH(Schomaker et al.,2013).The cloly related black clade reprented9miRNAs with time cour profiles cor-relating with another emerging biomarker of liver injury,eg GLDH(Schomaker et al.,2013).On the other hand,the blue clade that forms its own distinct cluster included8miRNAs with a time cour profile similar to TBIL that typically asss the functional extraction capacity of liver.The third cluster repre-nted by the green clade,consisted of6miRNAs that demon-strated a clear respon to the acetaminophen overdo that does not correlate with the respon of the conventional bio-markers of DILI.As expected,biomarkers that are typically unaffected after APAP overdo such as CK,ALK,BUN,Gluc,and SCRE formed a single clade(turquois)with only1miRNA. Variation in miRNA mature quences
The majority of the detected miRNAs showed variations in the mature quences in comparison with their reference quence as documented in miRBa.The miRNAs,called isomiRs, were mainly altered in length at the30end of the mature quence.For the majority of the detected miRNAs,the m
ost abundant isomiR was not the same as the canonical quence in miRBa.While obrving the expression of different isomiRs which have originated from the same precursor miRNA,some of the isomiRs were found to be differentially prent in rum and others not.For example,miR-23a could be detected in its canonical form in all samples at relatively low levels with small variability.However,an isomiR that appeared with3nt missing at the30end could be detected at relatively high rum levels after acetaminophen overdo(Fig.4).We have con-firmed this data via RT-qPCR(e Supplementary Fig.1). Supplementary Table2shows the detected miRNAs with docu-mented quences and with their most abundant quences detected in this datat.Most of the detected quences were30 isomiRs.Some were also prent in their canonical form,and one miRNA showed a50variation.Morin et al.(2008)stated that the variations could be related to the variability in DICER or DROSCHA cleavage positions during the maturation of miRNAs and suggested evaluating the most abundant isomiR,instead of the reference,as this may yield the most robust approach for comparing miRNA levels.
MiRNA-like snRNAs
The miRNA tool miRDeep2identified289putatively novel miRNA mature quences.Of the predicted miRNAs,3were
270|TOXICOLOGICAL SCIENCES,2015,Vol.143,No.2
狼爸found with significantly altered rum levels in the acetamino-phen-overdod patients.A BLAT arch (hg19)for one of them pointed to a quence,marked as U2snRNA,which is involved in the major splicesosomal complex.A 24-nt long frag-ment from the U2snRNA appears highly abundant in blood at early time points after acetaminophen overdosing (remarkably high level in females).In Supplementary Figure 3,the number of quencing reads per ba (ba coverage)for the 184nt long U2snRNA quence is shown for subject 9.A fragment from ba 93to 116is highly expresd,whereas the coverage of the remain-ing 162nt is negligible.This confirmed that this fragment is not a by-product from the degradation of U2snRNA.Furthermore,
the
FIG.1.Comparison of circulation miRNA profiles in healthy and APAP-overdod subjects.Analysis of all pairwi comparisons of circulating miRNA profiles was per-formed via Pearson correlation algorithm using levels of miRNA prent in each sample.Low correlation is indicated by blue and high correlation by yellow
color.
FIG.2.Serum levels of liver-associated miRNAs after acetaminophen overdo.The liver associated miRNAs miR-122and miR-192were detected in rum from acet-aminophen-overdod patients.The graph reprents rum levels,bad on normalized quencing reads,of the liver-associated miRNA hsa-miR-122(A)and hsa-miR-192(B)in control and acetaminophen-overdod samples.
KRAUSKOPF ET AL.|271
two putative miRNAs,pmiR-chr1-2130and pmiR-chr19-5802,were detected,showing a noticeable respon to acetaminophen overdo.Neither could be perfectly assigned to any location on the human genome hg19.Therefore,we assume that the miRNA-like molecules originated from regions on the genome
that were not yet documented.The location of the best hit for the quences is shown in Supplementary Table 2,the quen-ces in Supplementary Table 3.
Tissue distribution and biological function of circulating miRNAs The possible tissue origin of the miRNAs that were identified in our study as responsive to APAP overdo,was evaluated by asssi
ng a publicly available small RNA quencing datat across 8tissues (lung,kidney,skeletal muscle,heart,pancreas,frontal orbital gyrus,spleen,and liver)from 2subjects (Faghihi et al.,2010).For each miRNA,a tissue enrichment score was cal-culated (e “Materials and Methods”).In this analysis,a value of >0.5indicated tissue enrichment (van de Bunt et al.,2013).The tissue enrichment scores for all miRNAs are prented in a heatmap in Figure 5.Bad on their enrichment scores,we found liver-enriched miRNAs hsa-miR-122-5p (0.99),hsa-miR-192-5p (0.78),hsa-miR-483-5p (0.71),hsa-miR-194-5p (0.62),and hsa-miR-210-3p (0.58)circulating after acetaminophen overdo.We also detected muscle-enriched miRNAs hsa-miR-193b-5p (0.77),hsa-miR-378c (0.76)and hsa-miR-378a-3p (0.73),frontal orbital gyrus-enriched miRNAs miR-125b-3p (1)and hsa-miR-125b-5p (0.52),as well as pancreas-enriched miRNAs hsa-miR-148a-3p (0.74)and hsa-miR-130b-3p (0.70).Interestingly,our t of miRNAs did not include miRNAs enriched in heart,kidney,lung,and spleen.
To evaluate the liver-specific biological function of circulating miRNAs that responded to APAP overdo,we have analyzed available literature (Table 1,Supplementary Table 4).The majority of identified miRNAs are shown to be either involved in hepatocyte regeneration/proliferation and tumorigenesis or regulation of liver-specific metabolism such as metabolism of lipids,gluco,cytochr
ome P450,and insulin signaling.Additional miRNAs have a role in stress respons such as oxidative stress,endoplasmic reticulum
stress,
FIG.3.Pearson clustering of miRNAs with other biomarkers.Clustering is bad on Pearson correlation that was calculated using rum levels of miRNAs or pro-tein concentration/activity values of classical and emerging biomarkers of DILI across all 24samples.Biomarkers that are not nsitive to APAP-induced liver injury such as CK,Gluc,ALK,creatinine,and BUN were ud as negative control in this
aj是什么意思analysis.
FIG.4.Differential respons of isomiR form of miR-23a to APAP overdo.The figure reprents rum levels,bad on normalized quencing reads of the canonical quence (A)and a 30isomiR (B)of miR-23a.
replay是什么意思272|TOXICOLOGICAL SCIENCES ,2015,Vol.143,No.2
