Quantitative analyses and transcriptomic profiling of circulating messenger RNAs as biomarkers of rat liver injury

Abstract

Serum aminotransferases have been the clinical standard for evaluating liver injury for the past 50-60 years. These tissue enzymes lack specificity, also tracking injury to other tissues. New technologies assessing tissue-specific messenger RNA (mRNA) release into blood should provide greater specificity and permit indirect assessment of gene expression status of injured tissue. To evaluate the potential of circulating mRNAs as biomarkers of liver injury, rats were treated either with hepatotoxic doses of D-(+)-galactosamine (DGAL) or acetaminophen (APAP) or a myotoxic dose of bupivacaine HCl (BPVC). Plasma, serum, and liver samples were obtained from each rat. Serum alanine aminotransferase and aspartate aminotransferase were increased by all three compounds, whereas circulating liver-specific mRNAs were only increased by the hepatotoxicants. With APAP, liver-specific mRNAs were significantly increased in plasma at doses that had no effect on serum aminotransferases or liver histopathology. Characterization of the circulating mRNAs by sucrose density gradient centrifugation revealed that the liver-specific mRNAs were associated with both necrotic debris and microvesicles. DGAL treatment also induced a shift in the size of plasma microvesicles, consistent with active release of microvesicles following liver injury. Finally, gene expression microarray analysis of the plasma following DGAL and APAP treatment revealed chemical-specific profiles. Conclusion: The comparative analysis of circulating liver mRNAs with traditional serum transaminases and histopathology indicated that the circulating liver mRNAs were more specific and more sensitive biomarkers of liver injury. Further, the possibility of identifying chemical-specific transcriptional profiles from circulating mRNAs could open a range of possibilities for identifying the etiology of drug/chemical-induced liver injury. HEPATOLOGY 2010

In the clinic, serum alanine aminotransferase (ALT) is the most frequently used biomarker to detect liver injury.1 Mechanistically, ALT is believed to be released into the circulation following loss of membrane integrity and cell lysis. Upon release, the ALT enzyme has a half-life in blood between 40 and 60 hours across multiple species.2, 3 Use of ALT to assess liver damage has some limitations. ALT is not specific for liver injury, also increasing with injury to other tissues such as skeletal muscle.4, 5 In addition, the generic nature of ALT release following hepatocellular necrosis does not allow the biomarker to distinguish between various etiologies of liver injury.

Cellular release of RNA molecules into the circulation has been shown to occur through multiple mechanisms. Among passive processes, the release of cellular messenger RNA (mRNA) and micro RNA (miRNA) has been shown following necrotic cell death.6, 7 The RNA molecules enter the circulation and are either associated with cellular debris or in naked form.6 Among active processes, mRNA and miRNA molecules have been identified within membrane-encapsulated vesicles released by cells. These include exosomes,8 shedding vesicles,9 and apoptotic blebs.10 Exosomes are small vesicles (40-100 nm) that are formed by inward budding of endosomal membranes.11 The vesicles are packaged within larger intracellular multivesicular bodies that release their contents to the extracellular environment through exocytosis. Shedding vesicles (<200 nm) are released from live cells through direct budding from the plasma membrane,9 whereas apoptotic blebs (100 to >1,000 nm) bud directly from the plasma membrane upon cell death.12 After release from the cell, exosomes, shedding vesicles, and apoptotic blebs circulate in the extracellular space, where most are broken down within minutes due to the display of phosphatidylserine on the external side of the membrane.9, 13 A fraction of the vesicles move by diffusion into the circulation and appear in biological fluids. The function of exosomes and shedding vesicles are believed to be cell-to-cell communication and platforms for multisignaling processes.9, 12 Although exosomes and shedding vesicles are released in healthy individuals, many pathological conditions and cellular perturbations stimulate further release of the particles.9, 12

Circulating RNAs have been shown to be useful biomarkers for multiple clinical endpoints including mortality in acute trauma patients,6 the diagnosis of preeclampsia,14 the diagnosis and monitoring of diabetic retinopathy15 and neuropathy,16 and as diagnostic or prognostic markers for multiple cancers.17-21 Recently, circulating miRNAs were also identified as potential biomarkers for drug-induced liver injury (DILI).22 The use of circulating mRNA as a biomarker offers several advantages. First, it can be readily obtained from patients. Second, amplification technologies such as polymerase chain reaction (PCR) allow highly sensitive and quantitative detection of specific mRNAs. Third, identification of targets of toxicity can be achieved using tissue-specific transcripts. Finally, microarray technologies can be exploited to broadly survey transcriptional changes in biological processes and signaling pathways and develop high-dimensional transcriptional profiles to discriminate among disease states or treatments. This latter characteristic provides promise to identify etiologies of liver injury to the extent that different causative processes (and different drugs) may generate distinct transcriptional profiles.

The objective of this study was to evaluate the specificity and sensitivity of circulating liver-specific mRNAs for assessing liver injury and to characterize the form of the liver mRNAs in the blood. To achieve this objective, hepatotoxicants acetaminophen (APAP) and D-galactosamine (DGAL) and the skeletal muscle toxicant bupivacaine HCl (BPVC) were administered to rats. Liver-specific mRNAs were measured in the plasma by quantitative PCR (qPCR) and compared with serum enzyme levels and liver histology to assess the sensitivity and specificity of the biomarkers. The liver mRNAs in the plasma were further characterized using sucrose density centrifugation together with electron microscopy (EM) and qPCR. Lastly, gene expression microarray analyses of plasma mRNAs isolated from APAP- and DGAL-treated rats were performed to assess the potential of transcriptional profiles for distinguishing among hepatotoxic treatments.

Actb, β-actin; Alb, albumin; ALT, alanine aminotransferase; APAP, acetaminophen; AST, aspartate aminotransferase; BPVC, bupivacaine HCl; DILI, drug-induced liver injury; DGAL, D-(+)-galactosamine; Fgb, fibrinogen B β-polypeptide; Hp, haptoglobin; qPCR, quantitative real-time RT-PCR; RT-PCR, reverse-transcription polymerase chain reaction.

Materials and Methods

qPCR Analyses.

Total RNA was reverse-transcribed according to the manufacturer's protocol using the High Capacity cDNA RT kit (Applied Biosystems, Foster City, CA). The resulting cDNA was amplified using the Taqman Universal PCR master mix (Applied Biosystems) and FAM-MGB probes and primers. The probes and primers were either predesigned (Applied Biosystems) or custom-designed and ordered from Applied Biosystems' custom oligo synthesis service. Gene expression assays targeting the 5′, middle, and/or 3′ regions of albumin (Alb), haptoglobin (Hp), fibrinogen B β-polypeptide (Fgb), and β-actin (Actb) were used (see Supporting Table 1 for gene expression assay information). In the DGAL samples the mRNA region interrogated did not have a significant impact on the qPCR results. Therefore, data from only one primer and probe set for each gene were analyzed for the remaining experiments. For Alb the gene expression assay for the middle region was used and for Fgb and Hp the 5′ assay was analyzed. The Applied Biosystems Prism 7900 HT Sequence Detector was used to generate qPCR measurements. In addition to the biological replicates (n = 7-8/treatment group), a minimum of two technical replicates was performed for each sample and gene. Copy number per mL plasma was calculated based on standard curves generated from plasmid DNA. Plasmid DNA was prepared using complementary DNA (cDNA) clones obtained from Open Biosystems (Huntsville, AL) and purified using the Qiagen HiSpeed Midi Plasmid purification kit (Qiagen, Valencia, CA). Clone information is as follows: Alb (Open Biosystems clone ID No. 7303856); Fgb (Open Biosystems clone ID No. 7371665); Hp (Open Biosystems clone ID No. 7321960); and Actb (Open Biosystems clone ID No. 6920838).

All additional experimental procedures are described in detail in the Supporting Information.

Results

DGAL Induces Hepatocellular Necrosis Concomitant with Increases in Liver-Specific mRNAs and Serum Enzyme Levels.

Histopathologic examination of rats treated with 1,000 mg/kg DGAL revealed moderate panlobular hepatocellular necrosis that was randomly distributed throughout the liver (Fig. 1A,B). The necrosis was observed in all of the treated animals, which were graded histologically as 3 on a scale of 0 to 5. Statistically significant, treatment-related increases were observed for serum ALT and aspartate aminotransferase (AST) levels, with increases of 109- and 81-fold over controls, respectively (Fig. 1C). (See Supporting Table 2 for serum ALT, AST, and bilirubin data.)

Figure 1. DGAL induces widespread hepatocellular necrosis concomitant with elevations in serum enzyme and plasma mRNA levels. Male Sprague-Dawley rats were administered DGAL at 0 (sterile PBS) or 1000 mg/kg intraperitoneally and sacrificed after 24 hours. H&E-stained sections of (A) control and (B) DGAL-treated liver show that DGAL induces moderate panlobular hepatocellular necrosis that was randomly distributed throughout the liver. (C) DGAL induces treatment-related increases in serum ALT and AST levels. Values are mean ± SEM; n = 15 rats per group. ***P < 0.001; Welch's ANOVA F-test. DGAL induces treatment-related increases in circulating (D) Alb, (E) Fgb, and (F) Hp mRNA levels. RNA was isolated from cell-free plasma and subjected to Taqman-based qPCR analysis. Taqman gene expression assays targeting the 5′, middle, and/or 3′ regions were analyzed. Standard curves were generated in all of the assays and absolute quantitation used to determine copy number per mL plasma. Statistically significant treatment-related increases were observed for all regions of all of the genes tested. Values are mean ± SEM; n = 8 rats per group. **P < 0.01; ***P < 0.001; ^aas determined by two-tailed t test; ^bas determined by Welch's ANOVA F-test.

Analysis of cell-free plasma mRNA isolated from rats treated with DGAL by qPCR demonstrated treatment-related changes in Alb, Fgb, and Hp (Fig. 1D-F). Given potential concerns regarding the integrity of RNA circulating in the blood, where RNase levels are known to be high, multiple regions of the RNA from each of the genes were interrogated. Assays directed toward the 5′, mid and 3′ regions of Alb demonstrated fold increases of 78, 68, and 103 times that of control values, respectively. Assays directed against the 5′, mid, and 3′ regions of Fgb demonstrated fold increases of 5.4, 4.2, and 7.1 times that of controls, respectively. Lastly, assays directed against the 5′ and 3′ regions of Hp demonstrated fold increases of 3.5 and 24 over controls, respectively. In all three genes, treatment-related increases were observed in all of the regions tested (Fig. 1D-F). Notably, these increases in circulating Alb, Fgb, and Hp mRNAs occurred despite a significant decrease in the expression of Fgb and Hp mRNAs in the livers of treated animals (Supporting Fig. 1). In addition, a small but statistically insignificant decrease in Alb mRNA was observed in the livers of treated animals.

APAP Induces Centrilobular Necrosis in a Dose- and Time-Dependent Manner.

Rats were administered APAP at 0, 100, 700, and 1,400 mg/kg by gavage and sacrificed 6, 24, and 48 hours after treatment. Hematoxylin and eosin (H&E)-stained liver sections were examined by an accredited pathologist and scored for incidence and severity of hepatocellular necrosis (Fig. 2A,B). No evidence of hepatotoxicity was evident 6 hours after treatment at any of the doses. At 24 hours, hepatotoxicity was observed at 700 and 1,400 mg/kg. Moderate hepatocellular necrosis was observed in the centrilobular region at 700 mg/kg in two of the eight animals examined. At 1,400 mg/kg, extensive centrilobular necrosis was observed in all eight of the treated animals (Fig. 2B). The coagulative necrosis in many cases bridged into the centrilobular regions of adjacent lobules. At 48 hours, one rat from each of the 700 and 1,400 mg/kg treatment groups died. In all of the surviving rats, moderate to moderately severe hepatocellular necrosis was observed. Similar to the 24-hour timepoint, coagulative and bridging necrosis was observed (data not shown). There was no histologic evidence of APAP-related hepatotoxicity at 100 mg/kg at any of the timepoints.

Figure 2. APAP induces centrilobular hepatocellular necrosis in a dose-dependent and time-dependent manner concomitant with elevations in serum enzyme and plasma mRNA levels. Male Sprague-Dawley rats were administered APAP at 0, 100, 700, or 1400 mg/kg by gavage and sacrificed at 6, 24, and 48 hours. H&E-stained sections of 24 hour (A) control, and (B) 1400 mg/kg APAP-treated liver indicate extensive centrilobular necrosis. Similar pathology was observed 48 hours after treatment. No hepatotoxicity was observed at 100 mg/kg APAP at any timepoint, and moderate necrosis was observed in some rats at 700 mg/kg at 24-48 hours (data not shown). APAP induces treatment-related increases in serum ALT and AST levels (C) 24, and (D) 48 hours after treatment. *P < 0.05, Kruskal-Wallis test. No significant changes were observed at the 6-hour timepoint (data not shown). APAP induces dose-dependent increases in plasma Alb, Fgb, and Hp mRNA levels (E) 24, and (F) 48 hours after treatment. (See Fig. 1 legend for experimental details.) *P < 0.05, Tukey-Kramer test. Values are mean ± SEM; n = 8 rats per group.

Serum ALT and AST levels were also assessed to determine hepatotoxicity following APAP administration. No treatment-related increases were observed at 6 hours (data not shown). At 24 hours significant treatment-related changes were observed only at 1400 mg/kg, with ALT and AST levels increased 43- and 103-fold over that of controls, respectively (Fig. 2C). At 48 hours ALT and AST levels were significantly increased at 700 mg/kg (14-fold and 10-fold over controls) and at 1400 mg/kg (37-fold and 33-fold over controls) (Fig. 2D). (See Supporting Table 3 for serum ALT, AST, and bilirubin data.)

APAP Treatment Increased Circulating Alb, Fgb, and Hp mRNA Levels in a Dose- and Time-Dependent Manner.

No treatment-related increases were observed 6 hours after APAP treatment (data not shown). At 24 hours statistically significant increases in circulating mRNA levels were observed for all three liver-specific mRNAs (Fig. 2E). Circulating Fgb and Hp were significantly increased at 100, 700, and 1400 mg/kg APAP, with Fgb levels exhibiting fold increases of 6, 61, and 131. Hp mRNA increased by 39-fold, 230-fold, and 158-fold over controls, respectively. Circulating Alb mRNA was significantly increased by 1900-fold and 875-fold over controls, respectively, at 700 and 1400 mg/kg. Similar to that observed with DGAL, the increase in circulating Alb, Fgb, and Hp mRNAs in the 1,400 mg/kg treatment group occurred despite a significant decrease in the expression of Alb and Hp mRNAs in the livers of treated animals (Supporting Fig. 1). At 48 hours, Hp mRNA levels were increased by 9-fold, 54-fold, and 86-fold over controls, respectively, at 100, 700, and 1400 mg/kg APAP, whereas Alb mRNA levels were increased by 13-fold and 31-fold at 700 and 1400 mg/kg. Circulating Fgb mRNA levels were significantly increased by 5-fold and 13-fold over controls, respectively, at 700 and 1400 mg/kg (Fig. 2F).

Skeletal Muscle Toxicant BPVC Elevates Serum ALT and AST Levels But Not Circulating Alb, Fgb, or Hp mRNA Levels.

To compare the specificity of circulating liver mRNAs with AST and ALT following skeletal muscle injury, rats were treated with BPVC and serum enzymes and circulating liver mRNAs were measured 24 hours after treatment. BPVC treatment induced a modest but statistically significant elevation in serum ALT (1.96×) and AST (3.58×) levels (Fig. 3A; Supporting Table 4). However, plasma Alb, Fgb, and Hp mRNA levels remained unchanged with BPVC treatment (Fig. 3B), demonstrating greater specificity in detecting hepatotoxicity than possible with serum transaminases.

Figure 3. Myotoxic BPVC induces elevations in serum ALT and AST but not circulating liver-specific mRNAs. Sprague-Dawley rats received 0 or 0.5% BPVC in the right and left tibialis anterior muscles and were sacrificed at 24 hours. Histopathologic examination revealed no evidence of hepatotoxicity (data not shown). (A) BPVC induces treatment-related increases in serum ALT and AST levels. ****P < 0.0001, Welch's ANOVA F-test. (B) BPVC treatment does not induce treatment-related increases in circulating liver-specific mRNA levels. (See Fig. 1 legend for experimental details.) No treatment-related increases were observed in any of the mRNAs tested. Values are mean ± SEM; n = 10 rats in the control group and 20 rats in the BPVC group.

Separation of Plasma by Sucrose-Density Gradient Centrifugation Reveals Density-Specific mRNA Distribution and Microvesicle Profiles Following DGAL Treatment.

Under normal conditions, naked mRNAs should be rapidly degraded in blood and not serve as a robust biomarker of liver injury. To examine the form of the circulating liver mRNAs and the mechanism protecting them from degradation, the 14,000g pellets isolated from the plasma of control and DGAL-treated rats were separated by sucrose density gradient centrifugation. EM and qPCR analyses were performed on each fraction to determine the size and state of the microvesicles and any density- and treatment-related variations in mRNA levels. With a few exceptions, the Alb, Fgb, and Actb mRNA were present in control animals in higher amounts in the mid density fractions (1.10-1.18 g/mL) when compared to the low/high density fractions (Fig. 4A-D). For Fgb mRNA, fractions 1.07 and 1.10 g/mL also contained relatively high amounts, whereas for Hp mRNA the low density fractions (1.07-1.10 g/mL) and the 1.21 g/mL fraction contained the highest amounts of mRNA. Differential distribution of the various mRNAs among the density fractions was also observed following treatment with DGAL. For Alb mRNA, treatment-related increases were observed in all fractions, with the middle fractions (1.10-1.18 g/mL) containing 89% of the total mRNA copies. Similarly, Fgb mRNA showed treatment-related increases in all density fractions with the exception of the 1.10 g/mL fraction. The middle fractions also contained the highest treatment-related increases and contained 84% of the total mRNA copies. For Hp mRNA, no treatment-related changes were observed in the low-density fractions, whereas the middle fractions contained the majority of changes and contained 81% of the mRNA copies. Some treatment-related changes in Hp mRNA were also observed in the high-density fractions. The distribution of Actb mRNA was similar to Fgb, with treatment-related differences primarily in the middle and high-density fractions.

Figure 4. Characterization of circulating mRNAs following sucrose density gradient separation. Cell-free plasma from rats treated with PBS or 1,000 mg/kg DGAL were centrifuged at 14,000g for 45 minutes. Pellets were resuspended and separated by sucrose density gradient centrifugation, with 11 1-mL fractions collected and density determined. These fractions were split between two tubes and centrifuged at 110,000g for 1 hour with the pellet from one of the two tubes undergoing RNA isolation and qPCR analysis and the pellet from the second tube analyzed by EM (see Fig. 5). Plasma was pooled from different rats to obtain needed volume for separation. This experiment was conducted twice. Sucrose density gradient separation reveals treatment- and density-specific profiles of circulating mRNAs. RNA was isolated from the pellet and subjected to Taqman-based qPCR analysis for (A) Alb, (B) Fgb, (C) Hp, and (D) Actb. (See Fig. 1 legend for experimental details.) (E) Hepatotoxicity induces increases in plasma microvesicle (MV) diameter in specific sucrose density-separated fractions. Plasma microvesicle size was determined in six of the 11 fractions collected by image analysis of EM images. Fifty microvesicles were counted in each fraction. Values are mean ± SEM. *P < 0.05, Welch's ANOVA F-test. (F) Percentage of microvesicles over 100 nm is dependent on treatment and density of fraction analyzed. These percentages were calculated based on the information used in E.

EM examination of the microvesicle fractions 1, 4, 5, 6, 9, and 11 (corresponding to densities 1.07, 1.11, 1.13, 1.18, 1.24, and 1.26 g/mL) revealed that intact, spherical microvesicles were present in all of the control fractions analyzed (Fig. 5). However, DGAL-treated fractions of intermediate densities (1.11, 1.13, and 1.18 g/mL) contained debris, cell fragments, and misshapen vesicles in addition to intact spherical microvesicles. DGAL-treated fractions of low density (1.07 g/mL) showed minimal cellular debris, whereas the high-density fractions (1.24 and 1.26 g/mL) showed no evidence of debris or fragmentation. DGAL treatment caused an increase in mean microvesicle diameter in certain density fractions, most notably at the ends of the gradient (1.07 and 1.25 g/mL) (Fig. 4E,F). Cellular debris was excluded from the analysis in the assessment of microvesicle size.

Figure 5. EM examination of plasma pellet following sucrose density gradient separation reveals density and treatment-specific characteristics. (See Fig. 4 legend for experimental information.) EM images of fraction 1 (1.07 g/mL), 4 (1.11 g/mL), 5 (1.13 g/mL), 6 (1.18 g/mL), 9 (1.24 g/mL), and 11 (1.26 g/mL) are shown.

In Situ Hybridization of Alb mRNA Following Liver Injury Shows Debris and Microvesicle Association of Transcripts In Vivo.

In situ hybridization following APAP treatment (1,400 mg/kg, 24 hours) showed Alb mRNA in the lumen of the centrilobular vein and was present either in small vesicles/particles or associated with structures that were consistent with cellular debris (Fig. 6).

Figure 6. In situ hybridization of Alb mRNA following liver injury shows debris and microvesicle association of transcripts in vivo. Alb mRNA was detected using in situ hybridization on liver sections from rats treated with 1,400 mg/kg APAP and sacrificed 24 hours later. Sections were hybridized to digoxigenin-labeled probe for Alb and counterstained with nuclear fast red stain. Purple-stained microvesicles and debris indicate presence of Alb mRNA. (A) Low and (B) high magnification images of the same section are shown.

Whole Genome Microarray Analysis Reveals Treatment-Specific Transcriptomic Profiles Following D-GAL and APAP Treatment.

Total RNA was isolated from the 14,000g plasma pellet from rats treated with DGAL or APAP. The RNA was amplified, labeled, and hybridized to whole genome Affymetrix Rat 230_2 arrays. Using a 2-fold change and a false discovery rate of <0.05 as the criteria, 1,374 and 804 probe sets were identified as differentially expressed following APAP and DGAL treatment, respectively; 132 probe sets were shared between the DGAL and APAP treatments. Of the 132 differentially expressed probe sets, only 14 were up-regulated and 24 were down-regulated in both the APAP and DGAL experiments. Consequently, only 38 probe sets were similarly affected following these two treatments (i.e., were altered in the same direction). Table 1 provides information on the mRNAs that had the greatest treatment-related increases following exposure to both APAP and DGAL, APAP alone, or DGAL alone.

Table 1. Treatment-Related Increases in Circulating mRNAs as Detected by Affymetrix Microarray Analysis

		Fold-Change
Gene	Accession No.	APAP	DGAL	Pathway/Function
* Predicted.
Circulating mRNAs Increased in Both APAP and DGAL Experiments
Alb	NM_134326	102.57	83.32	Inflammatory response; acute phase response signaling
Ccl2	NM_031530	79.83	25.35	Cytokine; chemotaxis
Serpina1	NM_022519	24.15	12.61	Inflamm. response; immune cell trafficking
Rbp4	NM_013162	20.44	36.82	Retinol binding; insulin resistance
G1p2*	NM_001106700	17.16	5.78	Interferon induced; ubiquitin-like
Mt2a	XM_001062488	16.40	18.20	Oxidative stress
Alb	NM_134326	11.26	77.08	Inflamm. resp.; acute phase response signaling; endocytosis
Lta4h	NM_001030031	6.27	3.40	Xenobiotic metabolism; arachidonic acid metabolism
Gpnmb	NM_133298	6.10	4.04	Integrin pathway; cell signaling; adhesion
LOC259245	NM_001024248	4.94	11.52	Alpha-2 u globulin, PGCL5
Tuba4a	NM_001007004	4.21	2.24	Alpha tubulin 4a; cancer
Pgrmc1	NM_021766	2.82	3.12	Heme binding; CYP450 metabolism
Prdx5	NM_053610	2.57	2.03	Cell response; modification of H2O2
Top 20 Circulating mRNAs Increased in APAP Experiment Only
Vcl*	NM_001107248	70.40		Integrin signaling; adhesion; migration
Cxcl2	NM_053647	68.29		Cytokine; ephrin receptor signaling
Ptprj	NM_017269	61.43		Hematol. System; cell-cell signaling; adhesion; activation
Hspa1a	NM_031971	40.60		Inflammatory cell recruitment; cell death
Cxcl7	NM_153721	37.96		Cytokine; red blood cell survival
Cxcl12	NM_001033882	30.90		Cytokine; hematopoiesis; migration, apoptosis
Xpo5*	NM_001108789	30.54		Protein transport, binding
Ceacam1	NM_001033860	28.57		Chemotaxis, hematopoiesis, apoptosis, proliferation
Fam183b	XM_213312	26.31		None identified
Slc24a3	XM_001054941	25.57		Ion transport; calcium homeostasis
Kcnab1	NM_017303	24.65		Potassium channel
Pde5a	NM_133584	24.18		Cyclic nucleotide phosphodiesterase
Arhgap28	NC_000018	23.18		Rho GTPase activating protein
RT1-A2	NM_001008829	22.73		None identified
Slc24a3	XM_001054941	22.11		Protein transport; cancer
Galnt1	NM_024373	22.06		N-acetyl galactosaminyl transferase
Srxn1	NM_001047858	21.98		Cell-cell signaling / cellular response
St3gal2	NM_031695	21.94		Amino acid modification
Tmem24	NM_001011996	21.52		Encodes transmembrane protein
Hspa1b	NM_212504	21.07		Inflamm. cell recruitment; cell death
Top 20 Circulating mRNAs Increased in DGAL Experiment Only
Ambp	NM_012901		111.24	Transporter; acute phase response
Mug1	NM_023103		108.00	Acute phase response (proteinase inhibitor)
Cyp2c7	NM_017158		102.92	Xenobiotic metabolism
LOC259246	NM_001033958		96.26	None identified
Fgg	NM_012559		65.02	Hematological system; platelet binding
Ttr	NM_012681		60.61	Thyroid hormone transport, apoptosis, acute phase response
Apoc3	NM_012501		58.09	Lipoprotein metabolism; hyperlipidemia
Vtn	NM_019156		57.99	Cell adhesion; spreading; protein degradation
Apoa1	NM_012738		53.34	HDL component; cholesterol efflux; diabetes; hyperlipidemia, Hypercholesterolemia; amyloidosis
Itih1*	NM_001107291		52.97	Plasma protease inhibitor
Fgl1	NM_172010		52.04	None identified
Kng1	NM_001009628		45.12	Proteolysis, vascularization
Gc	NM_012564		42.82	None identified
Itih3	NM_017351		40.88	Plasma protease inhibitor
Cyp2c13	NM_138514		40.38	Xenobiotic metabolism
Apoh	NM_001009626		39.84	Lipoprotein metabolism; coagulation
Serpinc1	NM_001012027		38.19	Serine protease inhibitor; blood coagulation cascade
Serpina3k	NM_012657		36.42	Serine protease inhibitor
Apoc1	NM_001109996		35.57	Metabolic and Endocrine Disorders; apolipoprotein, hyperlipidemia, insulin resist.
Spp2	NM_053577		34.78	Secreted phosphoprotein (cystatin)

Functional Analysis of Differentially Expressed Genes.

Although several cellular functions were transcriptionally altered in both APAP and DGAL treatments, the majority of pathways and functions were distinctly enriched between the two treatments. Key cellular functions that were altered in both treatments included cell death, hematological system development and function, and molecular transport (Table 2). Transcriptional changes in several mRNAs involved in immune cell trafficking and inflammatory response were noted with both treatments. However, a closer examination of the genes contained within the category revealed distinct differences. Transcriptional changes in immune cell trafficking following APAP treatment involved migration, activation, and movement of leukocytes, neutrophils, granulocytes, lymphocytes, and bone marrow cells. In contrast, there was no transcriptional evidence for lymphocyte or bone marrow cell involvement following DGAL treatment. Another cellular function that showed treatment-related differences was calcium homeostasis. APAP, but not DGAL, induced changes in over 79 genes involved in calcium flux, mobilization, quantity, and transport.

Table 2. Ingenuity Pathway Analysis of APAP and DGAL Microarray Data

APAP Experiment Findings
Key Functions	No. of Genes Differentially Expressed	Key Subcategories (No. of Genes Differentially Expressed)
Cell death	252	Cell death (223) – Necrosis (21); liver cells (18) Apoptosis (194) – inhibition (23); liver cells (14)
Hematological system dev. and function	170	Proliferation (68), hematopoiesis (53), differentiation (45), adhesion (32), chemotaxis (28)
Molecular transport	103	Calcium flux (26), mobilization (35), quantity (47), protein transport (26)
Immune cell trafficking	95	Migration, activation, movement, adhesion, etc.
Cell division	91	Arrest (47)

Canonical Pathways	P	Gene Exp. Ratio	Selected Genes (Fold Increase)
Integrin signaling	7.29E-08	31/203	Itga6 (19); Itgb1(17); Vcl (70); Mapk1 (15)
Virus entry via endocytic pathways	2.13E-06	17/96	Hla-C (23); Itga6 (19); Itgb1 (17); Prkc (4); Src (7)
Thrombin signaling	2.30E-06	28/203	Arhgef12 (12); F2Rl2 (12)
Antigen presentation	2.90E-06	9/39	Hla-C (23); Hla-E1 (9); Canx (5); Calr (3)
Caveolar-mediated endocytosis	3.78E-06	15/182	Alb (103); Hla-C (23); Itga6 (19); Itgb1 (17); Itgav (12)
Fcy receptor-mediated phagocytosis in macrophages and monocytes	4.92E-06	18/104	Lyn (10); Mapk1 (15); Nck2 (8); Src (7), Rac1 (4) Prkcq (3), Canx (5), Calr (3), B2m (3)
Lipid Ag presentation by CD1	1.25E-05	6/20	Canx (5); Pdia3 (5); Psap (20); Calr (3)
Ephrin receptor signaling	1.38E-05	25/195	Adam10 (14); Angpt1 (23); Cxcl12 (31); Grin1 (10); Jak2 (13); Itgb1 (17)
B-cell receptor signaling	4.14E-05	21/155	Akt3 (4); Lyn (10); Fcgr2b (7); Dapp1 (5); Gsk3b (6)

DGAL Experiment Findings
Key Functions	No. of Genes Differentially Expressed	Key Subcategories (No. of Genes Differentially Expressed)
Lipid metabolism*	83	Metabolism (40); transport (17); modification (22); synthesis (23)
Molecular transport*	85	Quantity (49), transport (28); release (25); protein localization (10)
Small molecule biochemistry*	106	Homeostasis (12); secretion (14); lipid production (14); lipid accumulation (16)
Hematological system development and function	107	Activation (32); proliferation (32); adhesion (22); chemotaxis (21); coagulation (18)
Inflammatory response	91	Immune response (43); inflammation (27); cell movement (21); aggregation (17); adhesion (16)
Cell death	125	Apoptosis (104); necrosis (12)

Canonical Pathways	P	Gene Exp. Ratio	Selected Genes (Fold Increase)
* Lipid changes predominate (lipid metabolism, transport, modification, etc.). † These genes are differentially expressed in all four canonical pathways listed.
Acute phase response signaling	1.07E-21	38/178	Alb (83), Ambp (111), Apoa1 (53), Fgg (65), Ttr (61)
Complement system	1.05E-12	14/36	C9 (10); Cfi (25); Cfb (13); Cf1h (14); Serping1 (7)
Coagulation system	8.64E-09	11/37	F2 (14); Fga (28); Fgg (65); Plg (17); Serpinc1 (38)
Linoleic acid metab.	1.54E-07	15/125	Cyp2a2 (15); Cyp2c7 (103); Cyp2c9 (32); Cyp2c13 (40); Cyp3a2 (26); Pla2g4a (7)†
CYP450-mediated xenobiotic metab.	1.90E-07	18/210
Arachidonic acid metab.	9.48E-07	17/226
Fatty acid metab.	4.91E-06	16/192
LPS-IL-mediated inhib of RXR function	6.25E-05	18/205	Apoc1 (36), Apoc2 (23), Cyp2c9 (32), Cyp3a4 (26)
Tryptophan metab.	1.01E-04	15/255	Cyp2c7 (103); Cyp4f12 (13); Aadat (3); Cyp2a2 (15); Mettl7b (9)

The top canonical pathways that were the most significantly affected by APAP treatment were all involved either in the immune response (antigen presentation, receptor-mediated phagocytosis in macrophages and monocytes, lipid antigen presentation by CD1, or B-cell receptor signaling), extracellular interactions (integrin signaling, extracellular matrix effects; ephrin receptor signaling, cell-to-cell communication), or membrane-related changes (virus entry by way of endocytic pathways, caveolar-mediated endocytosis). A review of the mRNAs with the greatest changes following APAP treatment readily fit in this group, with several cytokines, transporters, and cell signaling molecules represented (Tables 1, 2). Treatment-related changes in mRNAs associated with oxidative stress generation, apoptosis induction, and necrosis were also evident and are consistent with APAP-induced hepatotoxicity. No increases in CYP450-related mRNAs were found.

The top canonical pathways that were the most significantly affected by DGAL treatment were all involved in the immune response: acute phase response signaling, complement system, and coagulation system changes. As with APAP, most of the mRNAs with the greatest increases following DGAL treatment fall into these categories, including mRNAs for various apolipoproteins, fibrinogens, and serine protease inhibitors. The top canonical pathway impacted was lipid metabolism, with molecular transport and small molecule biochemistry also scoring high, in large part due to lipid-involving pathways being altered (e.g., metabolism, transport, modification, etc.). These changes are consistent with previous reports of alterations in lipid metabolism and the composition of phospholipid membranes in DGAL-induced hepatotoxicity.23 Canonical pathways involved in various metabolic pathways—linoleic acid, arachidonic acid, fatty acid, and CYP450-mediated metabolism—were significantly affected and are also consistent with DGAL-induced effects.

Discussion

The standard biomarkers for liver injury are elevations in serum aminotransferases. The enzymes are released upon cell lysis. The enzymes are not expressed exclusively in the liver, leading to a lack of potential specificity, and the generic nature of their release provides little information with regard to the mechanism of injury. In this study, circulating liver mRNAs were compared with traditional serum transaminases and liver histology using rat models of hepatotoxicity and myotoxicity. Treatment with hepatotoxic doses of APAP and DGAL significantly increased plasma levels of all three liver-specific mRNAs analyzed (Alb, Hp, and Fgb) and serum ALT and AST. Treatment with myotoxic doses of BPVC did not increase circulating liver mRNAs, whereas serum ALT and AST showed statistically significant increases. The elevation in ALT and AST following muscle injury is consistent with previous rat and human studies4, 5 and underscores the potential lack of specificity of these enzymes. As a result, circulating liver mRNAs have the potential for greater specificity in the diagnosis of liver injury than serum transaminases.

In comparing the sensitivity of the different biomarkers, significant changes in circulating Hp and Fgb mRNA were observed following APAP administration at 24 hours in the 100 mg/kg dose group. All three liver mRNAs were significantly increased in the 700 and 1,400 mg/kg APAP dose groups. In comparison, ALT and AST showed significant elevations only at the 1,400 mg/kg dose. Histologically, only two animals showed liver injury at the 700 mg/kg dose, with all animals affected at 1,400 mg/kg. At 48 hours, significant changes in circulating Hp mRNA were observed at the 100 mg/kg dose, whereas changes in serum transaminases and liver histology were observed only at the 700 and 1,400 mg/kg doses. Thus, changes in circulating liver mRNAs also appeared more sensitive than AST, ALT, or liver histology.

To understand both how the circulating mRNAs were protected from rapid degradation and their potential mechanisms of release, sucrose density gradient centrifugation was used together with qPCR and EM. In untreated control animals, the combined measurements showed that the mRNAs were contained within microvesicles, presumably providing protection from ubiquitous RNases. The liver and “housekeeping” mRNAs were also not distributed uniformly through the density gradient. With the exception of Hp mRNA, the mRNAs in control animals were generally present in higher amounts in the middle density fractions (1.10-1.18 g/mL), which are traditionally identified as exosomal.8, 24 However, based on the respective size differences between exosomes (40-100 nm) and shedding vesicles (<200 nm), density fractions 1.11 and 1.13 g/mL appear to contain a significant percentage of shedding vesicles in control animals, whereas the remaining fractions appear primarily exosomal. Apart from the higher amount of mRNAs in the middle fractions in control animals, each mRNA appeared to have a unique distribution among the various density fractions. The biological significance of the different density distributions of mRNAs in the untreated animals is unclear, but one potential explanation is that the different mRNAs may be selectively packaged into exosomes and shedding vesicles.

In DGAL-treated animals, a similar nonuniform distribution of the mRNAs among the density fractions was observed except that all mRNAs showed treatment-related increases in the middle density fractions. The treatment-related increase in the middle density fractions was primarily due to the presence of cellular debris in these fractions and the apparent nonselective release of mRNAs through cell lysis. Presumably, the association with cellular debris also protects the mRNA from rapid degradation. Interestingly, the Hp mRNA showed treatment-related increases in mRNA only in the middle fractions that correspond to the cellular debris, whereas Alb and Fgb mRNA also showed increases in the low- and high-density fractions. Given that DGAL stimulated the release of larger particles in the low- and high-density fractions, it seems likely that Alb and Fgb mRNAs are actively released in these particles. Additional studies would be required to distinguish whether these particles are shedding vesicles or apoptotic blebs, but an increased release of shedding vesicles has been shown to occur following cellular perturbation.9, 12

Collectively, the unfractionated measurements of liver mRNAs following DGAL and APAP administration and the sucrose density gradient analyses provide an initial picture of the nature of these mRNAs in the circulation with and without liver injury. All liver-specific and “housekeeping” mRNAs were present at detectable levels in control animals, indicating that the active release of these mRNAs in exosomes and shedding vesicles is a physiological process and not dependent on liver injury. In treated animals, the increase in the liver mRNAs in the general circulation preceded pathological changes or increases in serum transaminases with respect to dose and were shown to be present in both microvesicles and cellular debris. This suggests a dose-dependent or injury-dependent shift in the mechanism of release from an active process at low, nontoxic doses to both active and passive processes at cytotoxic doses. Notably, the increase in circulating liver mRNAs occurred despite a significant decrease in the expression of these mRNAs in the livers of treated animals. This supports the finding that the increase in circulating liver mRNAs is through increased release and not through increased transcription.

Previous studies in toxicogenomics have demonstrated that unique transcriptomic profiles can be generated by different hepatotoxicants and grouped based on mechanisms of action when mRNA from the liver is used.25, 26 It may also be possible to detect hepatotoxicant-specific profiles from circulating mRNA. However, the functional and mechanistic interpretation of the profiles is more challenging than those derived from a specific tissue. The mRNAs in the plasma are a complex mix that result from multiple mechanisms of release (i.e., active processes and necrosis), multiple tissues, and even from different locations within an organ (e.g., centrilobular versus periportal injury). Part of this complexity can be seen in the pathway analysis for these two hepatotoxicants, where the majority of differentially expressed mRNAs were related to hematological and immunological functions. Transcriptional alterations of genes related to hematological system function can be expected, given that the liver is a source of many of these proteins and the nonspecific release of mRNAs following necrosis. In contrast, the mRNAs related to immunological functions may be derived from cells in the immune system responding to the necrotic damage within the liver. Previous studies have demonstrated a role for exosomes in immune signaling27 and mRNAs have been detected in mast cell exosomes.8

The mechanistic interpretation of these gene expression profiles could be improved through the purification of liver-specific microvesicles. Protein composition of microvesicles has been demonstrated to be cell-type specific and proteomic studies in mouse and rat hepatocytes have identified liver-specific membrane proteins that could be used in antibody-based capture approaches.28 By enriching for liver-specific microvesicles and performing gene expression microarray analysis on these vesicles, mechanistic interpretation of the profiles could be done without the confounding effects of other tissues and release by necrosis. Such techniques might allow analysis of the liver transcriptome even in healthy individuals without liver injury.

In summary, we found that circulating mRNAs hold potential advantages over traditional biochemical-based biomarkers in assessing liver injury and toxicity. First, the selected liver mRNAs showed greater sensitivity and specificity. Second, because the two hepatotoxicants tested demonstrated distinct transcript profiles, circulating mRNA may be useful in identifying the cause of liver injury. Finally, the finding that microvesicles are actively released by hepatocytes suggests that these vesicles could provide a “virtual biopsy” of the transcriptional state of the liver that could be further exploited in assessing the liver response to perturbations in the absence of overt injury.

Acknowledgements

The authors thank the following individuals for their outstanding efforts and assistance during this project: Otis Lyght (image analysis); Linda Pluta (microarrays), Ann Roberson (graphics), Earl Tewksbury (EM), Gabrielle Wilson (histopathologic examination), Michael Lawton (advice), and Joe Paulauskis (advice).