Potential novel biomarkers for monitoring the fibrogenic process in liver

Introduction 

Fibrosis is characterised by the excess deposition of extracellular matrix (ECM) involving molecular and histological re-arrangement of various types of collagens, proteoglycans, structural glycoproteins and hyaluronan. It is a hallmark of liver cirrhosis and contributes significantly to the deleterious outcome of chronic liver diseases. The deposition of ECM in the space of Disse (i.e., perisinusoidal fibrosis), the generation of (incomplete) subendothelial basement membranes and the strangulation of hepatocytes by the surrounding matrix impair not only the blood flow through the organ, but also the biosynthetic function of hepatocytes and the clearance capability of these and other cell types. The widely used diagnostic ‘gold standard’ of liver biopsy has many drawbacks, in addition to invasiveness, such as sampling error (about 1/50,000th of liver mass is obtained), irreproducible sample quality depending on the length and size of the tissue specimen (coefficient of variation 45–35%) and a histological evaluation strictly dependent on the experience of the pathologist (observer error). Therefore, the development of non-invasive, objective and quantitative serum- or plasma-based biomarkers of fibrogenesis is an important goal, which can be reached by two, principally different, approaches: class I and class II serum fibrosis markers.

Classification of biomarkers of fibrosis 

Class I fibrosis biomarkers are pathophysiologically derived from ECM turnover and/or from changes in the fibrogenic cell types, in particular, hepatic stellate cells (HSCs) and (myo-)fibroblasts [1]. They should reflect the activity of the fibrogenic and/or fibrolytic process and, thus, remodelling of ECM. These biomarkers do not indicate the extent of connective tissue deposition, that is, the stage of fibrotic transition of the organ. Frequently, they are costly laboratory tests and are the result of translation of fibrogenic mechanisms into clinical application. Thus, their selection is driven by hypothesis.

Class II fibrosis biomarkers mostly estimate the degree of fibrosis (extent of ECM deposition). In general, they comprise common clinical–chemical tests (e.g., enzymes, proteins and coagulation factors), which do not necessarily reflect ECM metabolism or fibrogenic cell changes. Their pathobiochemical relation with fibrogenesis is indirect, if at all. Thus, their selection is not driven by hypothesis, but empiric. The markers are standard laboratory tests and are integrated into multiparametric panels.

In general, both types of serum biomarkers follow different pathophysiological concepts. Class I markers inform about ‘what is going on’ (i.e., grade of fibrogenic activity); class II markers indicate ‘where fibrosis is’ (i.e., stage of fibrosis).

Class I fibrosis biomarkers 

These biomarkers are matrix components increasingly expressed by activated HSCs and (myo-)fibroblasts, have a delayed clearance by Kupffer cells or sinusoidal endothelial cells in the liver due to metabolic dysfunction and/or haemodynamic bypasses or are increasingly expressed mediators of fibrogenesis such as transforming growth factor (TGF)-β. Taken together, of the several procollagen and collagen fragments proposed, only the N-terminal propeptide of type III-procollagen (PIIINP) has reached a limited clinical application, but no widespread acceptance [2]. Sensitivities of about 76–78% and specificities of 71–81% have been reported, which can be increased up to 88%, if combined with additional collagen fragment markers. It should be emphasised that PIIINP is not a liver-specific biomarker. Similarly, structural glycoproteins (e.g., undulin and tenascin), biosynthetic (e.g., prolylhydroxylase) or catabolic enzymes (e.g., matrix metalloproteinases) of collagen and other ECM components have not been convincing in the detection, grading and staging of fibrosis (Table 1). In several studies, hyaluronic acid (hyaluronan) turned out to be currently the relatively best class I biomarker of fibrosis, having an area under curve (AUC) of 0.97, sensitivity of 86–100% and specificity of about 88%, in a recent investigation of cirrhosis due to non-alcoholic fatty liver disease [3] and other aetiologies. Since the negative predictive value of hyaluronan at a cut-off value of 60μg/L is much higher (98–100%) than the positive predictive value (61%), the main utility of serum hyaluronan lies in its ability to exclude advanced fibrosis and cirrhosis. Its stimulated synthesis in activated HSC, secretion into the sinusoidal blood stream and a short half-life of 2–9min in the circulation are good suppositions for a valid fibrosis biomarker. Laminin was reported to be a predictor of portal hypertension since significantly elevated concentrations were found under these conditions [4]. TGF-β concentrations in plasma are elevated in and correlate with the severity of liver disease and were suggested as non-invasive biomarker of fibrosis. However, the significant correlation with aspartate-aminotransferase (AST) and alanin-aminotransferase (ALT) activity [5] and the pathobiochemical finding that substantial amounts of TGF-β are localised in hepatocytes and released into the medium if hepatocytes are permeabilised [6] suggest the elevation of TGF-β as a marker of necrosis instead of fibrogenesis.

Table 1. Class I biomarkers of liver fibrogenesis.

		Specimen			Method
		Serum	Urine	Liver biopsy
	Extracellular matrix-related enzymes
	Enzyme
	Prolylhydroxylase	+	−	+	Radioenzymatic, RIA
	Monoamine-oxidase	+	−	(+)	Enzymatic
	Lysyloxidase	+	−	+	RIA
	Lysylhydroxylase	+	−	−	RIA
	Galactosylhydroxylysyl-glucosyltransferase	+	−	+	RIA
	Collagenpeptidase	+	−	+	Enzymatic
	N-Acetyl-β-d-glucosaminidase	+	+	+	Enzymatic
	Collagen fragments and split products
	Type of collagen
	Type I-procollagen
	• N-terminal propeptide (PINP)	+	−	+	ELISA
	• C-terminal propeptide (PICP)	+	−	+	RIA
	Type III-procollagen
	• Intact procollagen	+	−	−	RIA
	• N-terminal propetide (PIIINP)
	• Complete propeptide (Col 1–3)	+	−	−	RIA
	• Globular domain of propeptide (Col-1)	+	−	−	RIA
	Type IV-collagen
	• NC1-fragment [C-terminal] crosslinking domain (PIVP)	+	+	−	ELISA, RIA
	• 7S domain (“7S collagen”)	+	+	−	RIA
	Type VI-collagen	+	+	+	RIA
	Glycoproteins and matrix metalloproteinase (inhibitors)
	Marker
	Laminin, P1-fragment	+	−	−	RIA, EIA
	Undulin	+	−	−	EIA
	Vitronectin	+	−	−	EIA
	Tenascin	+	−	−	ELISA
	YKL-40	+	−	+	RIA/ELISA
	(Pro)matrix metalloproteinase (MMP-2)	+	−	−	ELISA
	Tissue inhibitor of metalloproteinases (TIMP-1, TIMP-2)	+	−	−	ELISA
	sICAM-1 (soluble intercellular adhesion molecule, sCD54)	+	−	−	ELISA
	sVCAM-1 (soluble vascular cell adhesion molecule, sCD106)
	Glycosaminoglycans
	Hyaluronic acid (hyaluronan)	+	−	−	Radioligand assay
	ELISA
	Molecular mediators
	Transforming growth factor-β (TGF-β)	+	−	+	ELISA
	Connective tissue growth factor (CTGF/CCN2)	+	?	+	ELISA

Preliminary studies point to CTGF/CCN2 in serum as an innovative class I biomarker of fibrogenesis [7]. This 38-kD protein is synthesised not only in HSCs, but also in hepatocytes where the expression and secretion are strongly dependent on TGF-β [8]. Accordingly, CTGF expression in fibrotic liver tissue is up-regulated and its concentration in blood elevated if fibrogenesis is occurring. There is a correlation between CTGF levels and fibrogenesis, because the levels decrease in fully developed, end-stage cirrhosis than in fibrosis. The AUCs for fibrosis versus control and cirrhosis versus control were calculated to be 0.955 and 0.887, respectively, the sensitivities 100% and 84%, respectively, the specificities 89% and 85%, respectively [7]. These criteria suggest CTGF as a potentially valuable class I biomarker of active fibrogenesis.

Recently, the glycoprotein YKL-40 (‘chondrex’, molecular mass 40kD), likely a growth factor for fibroblasts and endothelial cells, was shown to be strongly expressed in human liver tissue. In particular, HSCs contain YKL-40 mRNA. Several studies have found elevated YKL-40 concentrations in the sera of patients with liver diseases. Sensitivities and specificities around 80% and an AUC of 0.81 for fibrosis are reported for HCV patients [9], for those with alcoholic liver disease, specificity of 88% and a low sensitivity of 51% were calculated [10]. Serum concentrations of this protein correlated with other ECM products secreted by HSCs and fibroblasts (e.g., PIIINP, hyaluronan, MMP-2 and TIMP-1). It is claimed that YKL-40 concentrations reflect the degree of liver fibrosis, but extensive clinical evaluation is required and other inflammatory diseases as potential conditions of YKL-40 elevations have to be excluded.

Class II fibrosis biomarkers 

This category comprises a rapidly increasing, great variety of biochemical scores and multiparameter combinations (i.e., biomarker panels), which are selected by various statistical models and mathematical algorithms (e.g., multiple logistic regression analysis). They fulfil the most appropriate diagnostic criteria for detection and staging of fibrosis and to a lesser extent for grading of fibrogenesis. In general, the panels consist of rather simple (standard) laboratory tests, which are subject to changes in the serum or plasma of fibrotic and cirrhotic patients (Table 2). Several of the parameters included in the more than 20 scores currently available have no pathophysiological relation to fibrogenesis. Some of them have an indirect relation, and only few parameters can be regarded as being directly related to fibrogenesis. The parameters measured comprise those of necrosis such as ALT and AST, coagulation-dependent tests, transport proteins, bilirubin and some ECM parameters. Frequently, the reduction of platelet counts in cirrhotic patients is included. Most prevalent are the Fibrotest™ and for necro-inflammatory activity the Actitest™ (Biopredictive, Paris, France). They are based on γ-glutamyltransferase (γ-GT), total bilirubin, haptoglobin, α2-macroglobulin, apolipoprotein A1 and for Actitest, additionally, on ALT [11]. The data of Fibrotest and Actitest are calculated with a patented artificial intelligence algorithm to give the measures of fibrosis stage and necro-inflammatory grade (activity), respectively. The Wai-score based on AST, alkaline phosphatase and platelet count [12], the ELF test based on TIMP-1, PIIINP, hyaluronan [13] and the Hepascore based on bilirubin, γ-GT, hyaluronan, α2-macroglobulin, age and gender [14] are further scores with up to now limited clinical application. Fibrotest was recommended to be a better predictor than biopsy staging for HCV complications and death [11]. Recently, Fibrotest™ and Actitest™ were included into biomarkers for the prediction of liver steatosis (Steato-test™), alcoholic steato-hepatitis (ASH-test™) and non-alcoholic steato-hepatitis (NASH-test™) by supplementation with serum cholesterol, triglycerides, glucose (and AST for NASH-test) adjusted for age, gender and body mass index (BMI) [15]. The diagnostic criteria elaborated in a large cohort of patients suggest Steato-test as a simple and non-invasive quantitative measure of liver steatosis and the NASH-test as a useful screening procedure for advanced fibrosis and NASH in patients with various metabolic syndromes [15]. FibroMax™ (Biopredictive) was recently developed as a method of combined calculation of these fibrosis-related tests in a single procedure. Comparative evaluation of class II serum biomarker panels, however, could not highlight their clinical superiority [16]. Since only about 40% of the results were assigned to be correct, a fraction of about 50–70% was inaccurate with regard to the staging of fibrosis severity and a small fraction of results was even incorrect [16]. Thus, the presently suggested multiparameter approaches with class II fibrosis biomarker panels have to be taken with caution in clinical practice. A successful approach to improve the diagnostic accuracy of the panel markers in chronic hepatitis C might be their stepwise combination [17]. By combining the sequential algorithms of APRI, Forns’ index and Fibrotest (Table 2), the diagnostic performance could be significantly improved, resulting in a reduction of the need for liver biopsy by 50–70% [17]. However, it should be emphasised that the combination of individually assessed parameters necessarily creates relative high variance due to the imprecision of each separate measurement [18]. Coefficients of variation range from series to series between 3% and 6% for common clinical–chemical parameters and from 4% to more than 12% for hyaluronan, PIIINP and other matrix parameters. Furthermore and even more important is the lack of standardised assays for many of these parameters, which excludes the general use of cut-offs and algorithms [18].

Table 2. Class II Biomarkers of Liver Fibrogenesis.

	Index	Parameters	Chronic liver disease	Sensitivity (%)	Specificity (%)
	PGAA-index	Prothrombin time, γGT, apolipoprotein A1, α2-macroglobulin	Alcohol	79	89
	Bonacini-index	ALT/AST-ratio, INR, platelet count	HCV	46	98
	Sheth-index Park-index	AST/ALT (De Ritis)	HCV HCV	53 47	100 96
	PGA-index	Prothrombin time, γGT, apolipoprotein A1	Mixed	91	81
	Fortunato-score	Fibronectin, prothrombin time, PCHE, ALT, Mn-SOD, β-NAG	HCV		94
	Fibrotest (fibro-score)	Haptoglobin, α2-macroglobulin, apolipoprotein A1. γGT, bilirubin	HCV HBV	75	85
	Pohl-score	AST/ALT-ratio, platelet count	HCV	41	99
	Actitest	Fibrotest+ALT	HCV
	Forns-index	Age, platelet count, γGT, cholesterol	HCV	94	51
	Wai-index (APRI)	AST, platelet count	HCV	89	75
	Rosenberg-score (ELF-score)	PIIINP, hyaluronan, TIMP-1	Mixed	90	41
	Patel-index (FibroSpect)	Hyaluronan, TIMP-1, α2-macroglobulin	HCV	77	73
	Sud-index (fibrosis probability-index, FPI)	Age, AST, cholesterol, insulin resistance (HOMA), past alcohol intake	HCV	96	44
	Leroy-score	PIIINP, MMP-1	HCV	60	92
	Fibrometer test	Platelet count, prothrombin index, AST, α2-macroglobulin, hyaluronan, urea, age	Mixed	81	84
	Hepascore	Bilirubin, γGT, hyaluronan, α2-macroglobulin, age, gender	HCV	63	89
	Testa-index	Platelet count/spleen diameter-ratio	HCV	78	79
	FIB-4	Platelet count, AST, ALT, age	HCV/HIV	70	74
	FibroIndex	Platelet count, AST, γ-globulin	HCV	38	97

Abbreviations: GGT, γ-glutamyltransferase; PIIINP, N-terminal propeptide of type III-procollagen; TIMP, tissue inhibitors of metalloproteinases; MMP, matrix metalloproteinases; β-NAG, N-acetyl-β-glucosaminidase; AST, aspartate-aminotransferase; ALT, alanine aminotransferase; INR, international normalized ratio.

Developments of innovative biomarkers 

Growing understanding of the pathogenesis of hepatic fibrosis indicates potentially powerful non-invasive (blood) biomarkers of hepatic fibrogenesis and fibrosis (Table 3). CTGF/CCN2 was already mentioned as a pluripotent downstream modulator of TGF-β and was found to be up-regulated by TGF-β in hepatocytes. Although most of CTGF will only have a defined paracrine function in fibrogenic tissue, a certain fraction spills over into the circulation, resulting in elevated serum concentrations during active fibrogenesis [7]. Circulating level of CTGF might be an objective and sensitive readout of ongoing fibrogenesis in necro-inflammatory liver tissue.

Table 3. Future candidate biomarkers of non-invasive diagnosis and follow-up of liver fibrogenesis.

	Biomarker	Specimen	Assay technology	Pathobiochemical basis
	CTGF	Serum	Immunoassay	TGF-β induced expression in and secretion by hepatocytes and hepatic stellate cells
	Fibrocytes	Blood, buffy coat	Flow cytometry of CD34+, CD45+, Coll I+ cells qPCR	Supplementation of local fibroblasts at site of liver injury by bone-marrow-derived fibrocytes
	BMP-7	Serum	Immunoassay	Antagonist of TGF-β, inhibitor of EMT
	G-CSF GM-CSF M-CSF	Blood	Immunoassays	Mobilisation of bone-marrow-derived fibrocytes
	Proteomics	Serum	Mass spectrometry (MS)	Fibrosis-specific serum protein profiles
	Glycomics	Serum	Adaptation of DNA-sequencer/fragment analyzer technology to profiling of desialylated N-linked oligo-saccharides	Fibrosis-specific profiles of desialylated serum protein linked oligo-saccharides (N-glycans)
	Xylosyltransferase (EC 2.4.2.26)	Serum	LC–MS/MS	Key enzyme of the biosynthesis of glycosaminoglycan chains in proteoglycans, e.g., in hepatic stellate cells and hepatocytes

Bone-marrow-derived fibrocytes might offer new approaches not only for the understanding of the pathogenesis, but also for the diagnosis of liver fibrosis. Fibrocytes are circulating progenitor cells (CD34 positive) of haematopoietic origin (CD45 positive) capable of differentiating into diverse mesenchymal cell types [19]. The additional markers of fibrocytes, that is, positivity of type I collagen and the CXCR4 chemokine expression can be used to quantitate this special sub-population of circulating leucocytes applying quantitative polymerase chain reaction (PCR) and/or flow cytometry. The determination of the colony-stimulating factors, M-CSF, G-CSF and GM-CSF, which are increasingly expressed in fibrotic liver tissue and elevated in serum [20], are possibly involved in the mobilisation of fibrocytes from the bone-marrow and their homing in the liver during fibrogenesis. They might be further candidates for diagnostic evaluation.

A new, but currently still controversial aspect of fibrogenesis is epithelial–mesenchymal transition (EMT) of hepatocytes and biliary epithelial cells, respectively, to (myo-)fibroblasts [21]. EMT is governed by the balance of TGF-β (pro-EMT) and its antagonist, that is, BMP-7 (anti-EMT). In addition to its anti-EMT effect, BMP-7 was shown to have anti-apoptotic and anti-inflammatory activities. Thus, the measurement of BMP-7 alone or even in relation to TGF-β in serum might reflect the activity of fibrogenesis and, hence, the velocity of fibrotic organ transition [22].

Xylosyltransferase (XT), a key enzyme of the biosynthesis of glycosaminoglycans in proteoglycans, was shown to have increased activities in the sera of patients with connective tissue diseases. With high-performance liquid chromatography (HPLC)–tandem mass spectrometry, measurements in large cohorts of liver fibrotic patients seem possible [23]. Since HSCs in fibrotic liver tissue (myofibroblasts) have a greatly stimulated proteoglycan synthesis [24], XT activity in serum might be a promising class I biomarker of fibrogenesis.

Further successful developments could emerge from serum proteome profiling [25] and from total serum protein glycomics, that is, the pattern of N-glycans [26]. It was reported that a unique serum proteomic fingerprint is identified in the sera of patients with fibrosis, which enables differentiation between different stages of fibrosis and a prediction of fibrosis and cirrhosis in patients with a chronic hepatitis B infection [25]. Specificities and sensitivities and accuracy of prediction of cirrhosis are about 89%. Similarly, N-glycan profiling can distinguish between compensated cirrhosis from non-cirrhotic chronic liver diseases with sensitivity and specificity of 79% and 86%, respectively [26].

Supplementation of all these laboratory tests by modern high-resolution or molecular imaging analyses would be extremely helpful in the consolidation of objective and valid non-invasive biomarkers of diagnosis and follow-up of fibrogenic (liver) diseases. In conclusion, currently available type I and II serum biomarkers should be used with caution, because neither single nor panel markers fulfil the requirements of an ideal non-invasive biomarker of fibrosis [27], that is, analytical simplicity allowing performance in any laboratory, standardisation of the test system and calibrators allowing comparison between the laboratories over a long period, cost-effectiveness, specificity for the liver and the disease, clear association with the stage of fibrosis or grade of fibrogenesis and independency of the aetiology of the fibrosis. Even the relative best and most extensively evaluated type I (i.e., hyaluronan) and type II (i.e., Fibrotest and Actitest) serum biomarkers do not meet the criteria of an ideal marker. Further detailed insight into the mechanism of liver fibrosis and improvement of analytical techniques will result in new approaches for non-invasive assessment of fibrosis with biochemical or physical means.

In addition, genetic markers linked with the progression rate of fibrosis will become important diagnostic and prognostic tools for patients with liver fibrosis.