0% found this document useful (0 votes)
106 views16 pages

MSB 20188793 PDF

Uploaded by

Andreea Roxana
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
106 views16 pages

MSB 20188793 PDF

Uploaded by

Andreea Roxana
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

Article

Plasma proteome profiling discovers novel proteins


associated with non-alcoholic fatty liver disease
Lili Niu1,2,†, Philipp E Geyer1,2,†, Nicolai J Wewer Albrechtsen1,2,3,4,5 , Lise L Gluud6,7, Alberto Santos1,
Sophia Doll1,2, Peter V Treit2, Jens J Holst3,4 , Filip K Knop4,6,8, Tina Vilsbøll6,8, Anders Junker6,8,
Stephan Sachs9, Kerstin Stemmer9, Timo D Müller9, Matthias H Tschöp9, Susanna M Hofmann10,11,12 &
Matthias Mann1,2,*

Abstract Introduction
Non-alcoholic fatty liver disease (NAFLD) affects 25% of the popu- Non-alcoholic fatty liver disease (NAFLD) is the most common liver
lation and can progress to cirrhosis with limited treatment disease with an estimated prevalence of 25% in the general popula-
options. As the liver secretes most of the blood plasma proteins, tion of Western and Asian countries (Bellentani et al, 2010; Fan
liver disease may affect the plasma proteome. Plasma proteome et al, 2017). NAFLD has become an enormous clinical and economic
profiling of 48 patients with and without cirrhosis or NAFLD burden with annual medical costs of over $100 billion in the United
revealed six statistically significantly changing proteins (ALDOB, States alone and is projected to keep growing in parallel with the
APOM, LGALS3BP, PIGR, VTN, and AFM), two of which are already increasing prevalence of obesity and type 2 diabetes (T2D)
linked to liver disease. Polymeric immunoglobulin receptor (PIGR) (Younossi et al, 2016). Unfortunately, progression is usually asymp-
was significantly elevated in both cohorts by 170% in NAFLD and tomatic and only manifests when patients develop end-stage liver
298% in cirrhosis and was further validated in mouse models. disease, which has limited treatment options. Although there is
Furthermore, a global correlation map of clinical and proteomic promising activity in the development of small-molecule drugs
data strongly associated DPP4, ANPEP, TGFBI, PIGR, and APOE with (Cassidy & Syed, 2016), there is currently no Food and Drug Admin-
NAFLD and cirrhosis. The prominent diabetic drug target DPP4 is istration (FDA)-approved drug with beneficial effects on clinical
an aminopeptidase like ANPEP, ENPEP, and LAP3, all of which are outcomes (Friedman et al, 2018).
up-regulated in the human or mouse data. Furthermore, ANPEP NAFLD is defined as fat accumulation in the liver exceeding 5–
and TGFBI have potential roles in extracellular matrix remodeling 10%, measured either by imaging methods or by liver histology after
in fibrosis. Thus, plasma proteome profiling can identify potential exclusion of other etiologies, for instance, heavy alcohol consump-
biomarkers and drug targets in liver disease. tion and medication-induced steatosis (Kotronen & Yki-Jarvinen,
2008). NAFLD is further categorized histologically into simple steato-
Keywords biomarker discovery; mass spectrometry; NAFLD; NASH; plasma sis and non-alcoholic steatohepatitis (NASH) with or without
proteome profiling fibrosis. Up to 90% of patients with NAFLD have simple steatosis,
Subject Categories Genome-Scale & Integrative Biology; Molecular Biology with relatively benign prognosis and small risk of progression to
of Disease; Post-translational Modifications, Proteolysis & Proteomics advanced fibrosis and liver-related mortality, but an increased risk of
DOI 10.15252/msb.20188793 | Received 12 December 2018 | Revised 27 January cardiovascular events (Dyson et al, 2014; Singh et al, 2015).
2019 | Accepted 28 January 2019 However, 10–30% of NAFLD patients have NASH, a more severe
Mol Syst Biol. (2019) 15: e8793 form of the disease with hepatocellular injury and hepatic

1 Novo Nordisk Foundation Center for Protein Research, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark
2 Department of Proteomics and Signal Transduction, Max Planck Institute of Biochemistry, Martinsried, Germany
3 Department of Biomedical Sciences, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
4 Faculty of Health and Medical Sciences, Novo Nordisk Foundation Center for Basic Metabolic Research, University of Copenhagen, Copenhagen, Denmark
5 Department of Clinical Biochemistry, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
6 Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
7 Gastrounit, Medical Division, Hvidovre Hospital, University of Copenhagen, Hvidovre, Denmark
8 Clinical Metabolic Physiology, Steno Diabetes Center Copenhagen, Gentofte Hospital, Hellerup, Denmark
9 Helmholtz Diabetes Center at Helmholtz Centre Munich & Division of Metabolic Diseases, Institute for Diabetes and Obesity, Technische Universität München, Munich, Germany
10 Institute for Diabetes and Regeneration, Helmholtz Diabetes Center at Helmholtz Zentrum München, German Research Center for Environmental Health (GmbH),
Neuherberg, Germany
11 German Center for Diabetes Research (DZD), Neuherberg, Germany
12 Medizinische Klinik und Poliklinik IV, Klinikum der LMU, München, Germany
*Corresponding author. Tel: +49 89 8578 2557; E-mail: [email protected]

These authors contributed equally to this work

ª 2019 The Authors. Published under the terms of the CC BY 4.0 license Molecular Systems Biology 15: e8793 | 2019 1 of 16
Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

inflammation. NASH has a substantial risk of progression to serious prognostic implications of cirrhosis, developing risk markers
advanced fibrosis and mortality (Singh et al, 2015). Early detection or predictors of cirrhosis development is of utmost importance.
of NASH or fibrosis followed by lifestyle or pharmaceutical interven- As a central secretory organ of the human body, the liver
tion would therefore be an ideal first step toward reducing future produces the majority of plasma proteins with a direct function in
cirrhosis-related and cardiovascular deaths. the circulation, which is why many of the classical biomarkers for
Liver biopsy remains the gold standard for diagnosing NASH liver dysfunction are in this category. By extension, liver disease is
and for staging the severity of fibrosis. However, there are inherent likely to affect the blood plasma proteome. Mass spectrometry (MS)-
limitations of liver biopsy, as it is invasive, associated with compli- based proteomics is the technology of choice to analyze proteins in a
cations such as bleeding, and suffers from sampling bias. Therefore, systematic and systems-wide fashion. It has undergone persistent
the field increasingly employs surrogate non-invasive techniques innovations in terms of sample preparation, instrumentation, acqui-
for the diagnosis of these conditions (Bril & Cusi, 2017). Radio- sition methods, and computational analysis and has contributed to
graphic techniques for the assessment of steatosis and liver stiffness many breakthroughs in basic research over the last decades
include ultrasound, transient elastography (TE, FibroScanTM; Echo- (Aebersold & Mann, 2003, 2016; Altelaar et al, 2013; Richards et al,
sens, Paris, France), multiparametric magnetic resonance imaging 2015). Today, it clearly has the potential to facilitate disease-related
(MRI), magnetic resonance elastography (MRE), and magnetic reso- biomarker discovery in an unbiased and non-hypothesis-driven
nance spectroscopy (MRS). TE with controlled attenuation parame- manner (Geyer et al, 2017). Our group has recently developed an
ter (CAP) accurately quantifies liver steatosis and stiffness in automated, rapid, and robust shotgun proteomics pipeline that
patients with NAFLD (Mikolasevic et al, 2016), but it does not have allows the streamlined analysis of several hundred plasma proteins,
sufficient accuracy to discern between different stages of fibrosis a technology known as “plasma proteome profiling” (Geyer et al,
(Chang et al, 2016). Furthermore, the use of TE in NAFLD has 2016a). It requires only one microliter of plasma and features high
significant limitations in older patients and those with obesity (body reproducibility and low variability. So far we have applied this tech-
mass index (BMI) > 35 kg/m2) or type 2 diabetes as the rate of nology to study the effects of sustained weight loss on the human
unreliable measurements is higher in these populations (Dyson plasma proteome and to rigorously assess the quality of plasma
et al, 2014). Even with a large-size (XL) probe for patients with samples (Geyer et al, 2016b; preprint: Geyer et al, 2018).
obesity, there can still be a discordance in comparison with liver In this study, we employed and augmented plasma proteome
biopsy (Myers et al, 2012) and complementary methods would profiling with a recently introduced data acquisition method termed
surely be beneficial. “BoxCar” which covers the proteome with about tenfold higher
Hepatic injury results in release of specific liver enzymes into the dynamic range (Meier et al, 2018). We successfully applied this
circulation, and these are routinely measured with blood tests. Clas- technology to identify new biomarker candidates for non-alcoholic
sic markers for liver damage include alanine aminotransferase (ALT) fatty liver disease.
and aspartate aminotransferase (AST). Elevated levels of these
enzymes are often combined with patient data such as age and
platelet count in NAFLD severity indices such as the fibrosis-4 (FIB- Results
4) index (Sterling et al, 2006) and with biomarkers in the case of fatty
liver index (Bedogni et al, 2006), NAFLD liver fat score (Kotronen Study design and assessment of the plasma proteome analysis
et al, 2009), APRI, NAFLD fibrosis score (Angulo et al, 2007), and
the Enhanced Liver Fibrosis panel (ELF) (Parkes et al, 2011). Among As the plasma proteomics pipeline has not yet been applied to liver
these tests, ELF is a commercial panel of three markers reflecting the disease, we first set out to study a phenotype for which the effects
process of extracellular matrix remodeling and fibrogenesis. This should be very drastic. Liver cirrhosis is a more severe condition
method has been found to be the most cost-effective non-invasive than NAFLD and a common end-stage of most types of chronic liver
method in identifying patients with advanced liver fibrosis (stages 3 diseases. As the liver has then undergone substantial changes in
and 4) and has been recommended by NICE guideline as a blood test structure and function irrespective of disease etiologies, its analysis
to screen for advanced liver fibrosis in adults (Glen et al, 2016). Most should form a basis for our general understanding of the effects of
of these tests were developed more than a decade ago and comprise liver damage on the plasma proteome profile.
panels of simple clinical and laboratory variables. Furthermore, by We first chose a cohort of ten non-diabetic patients with cirrhosis
their nature, these markers detect relatively late changes in liver (cirrhosis-cohort) to compare them against ten age-, sex-, and BMI-
pathology. Therefore, in the management of NAFLD, there is an matched healthy controls as well as against eight matched individu-
unmet need to develop non-invasive or minimally invasive methods als with T2D and no liver disease (Junker et al, 2015; Fig 1A).
with better sensitivity and selectivity to detect NASH and fibrosis as However, our main goal was to investigate the changes in the
well as to predict the progression of patients at risk. plasma proteome in patients with NAFLD before progression to
After the fibrotic state, which is usually non-symptomatic, liver cirrhosis. As NAFLD is highly associated with obesity and T2D, we
disease can further progress to cirrhosis and convert normal liver included both subtypes for separate comparisons (Junker et al,
architecture into structurally abnormal liver nodules (Anthony et al, 2016; Fig 1A). The first cohort consisted of ten obese patients with
1977). It is characterized by multiple severe physiological condi- NAFLD and normal glucose tolerance (NGT) (NAFLD-cohort 1), and
tions, such as reduction in protein synthesis, abnormalities in the the second included ten patients with both NAFLD and T2D
coagulation system, and portal hypertension. Different insults, for (NAFLD-cohort 2). These cohorts were compared to the matched
instance, viral infection, steatohepatitis, and autoimmune hepatitis, controls without NAFLD. The average duration of diabetes was
can all result in liver cirrhosis (Burt et al, 2015). Because of the 50 months. The cirrhosis and NAFLD studies had a total of 48

2 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors


Lili Niu et al A protein marker panel for NAFLD Molecular Systems Biology

A
C D

B E F

Figure 1. Design and quality control of the human study.


A In total, 48 participants from three sub-studies of either NAFLD or cirrhosis with the indicated numbers of patients were included in this study.
B Fasting plasma was collected and distributed into a 96-well plate for proteomic analysis. Proteins were denaturized, reduced, alkylated, and digested using the
automated plasma proteome profiling pipeline, and purified peptides were analyzed in triplicate measurements in a randomization manner by LC-MS/MS. The
resulting 144 raw files were analyzed together with 168 library files by the MaxQuant and Perseus software programs.
C Numbers of quantified proteins in the triplicate measurements.
D Dynamic range of quantified proteins (LFQ, label-free quantitation values).
E Assessment of study quality by analyzing erythrocyte-specific proteins (red circles) and coagulation markers (blue circles). HBA, HBB, HBD: hemoglobin subunits alpha,
beta, delta; FGA, FGB, FGG: fibrinogen chains alpha, beta, gamma.
F Assessment of quantitation accuracy of the LC-MS/MS instrumentation by the number of proteins with a coefficient of variation (CV) below 30, 20, or 10%,
respectively, within three technical replicates.

participants (24 females) with a mean age of 57 years and a mean measurements and found that 272 proteins had a median CV below
BMI of 28 kg/m2 (Dataset EV1). 20% (Fig 1F).
We improved our previously described plasma proteome profil-
ing pipeline (Geyer et al, 2016a) by deep libraries in combination Plasma proteome profiling of patients with liver cirrhosis
with a new acquisition method termed BoxCar (Meier et al, 2018).
This method results in a tenfold increased dynamic range of peptide We first compared the plasma proteome profiles of patients with
signals due to equalized filling pattern of the Orbitrap mass liver cirrhosis to that of the two control groups (10 matched healthy
analyzer. The library consists of pooled and depleted plasma controls and 8 individuals with only T2D but no liver disease). This
samples—including from patients with liver diseases (see Materials revealed a dramatic shift in the quantitative proteome composition,
and Methods)—that were separated at the peptide level into 24 frac- reflected by 57 significantly differentially abundant proteins (Stu-
tions. This deep plasma library consisted of in total 2,081 proteins. dent’s t-test, P < 0.01, permutation-based FDR < 0.05), of which 31
The study samples were then measured without depletion, but with proteins were up-regulated and 26 down-regulated (Fig 2A).
BoxCar scans and in technical triplicates, resulting in about 150 LC- The down-regulated proteins include several classes with a direct
MS/MS datasets (Fig 1B). We quantified on average 503 proteins function in the bloodstream, and almost all of them are produced in
per individual (684 in total, of which 3 by single peptide) (Fig 1C), the liver. One of the classes contained proteins regulating blood
with MS signals that spanned an abundance range of six orders of coagulation and fibrinolysis, for instance, prothrombin (F2), vitamin
magnitude (Fig 1C and D). To investigate the quality of the study K-dependent protein C and S (PROC, PROS1), alpha-2-antiplasmin
samples in terms of consistency of collection and handling, we used (SERPINF2), antithrombin-III (SERPINC1), kallikrein (KLKB1),
our recently developed quality marker panels for coagulation and heparin cofactor II (SERPIND1), and carboxypeptidase (CPB2)
erythrocyte contamination (Geyer et al, 2016a; preprint: Geyer (Fig 2B). This reflects the central role of the liver in hemostasis, in
et al, 2018). No outliers of these marker indices were found based which most clotting and anticoagulation factors are synthesized by
on global protein abundance profiles, indicating that the samples parenchymal cells that are also involved in the clearance of acti-
were of high quality and that changes in plasma proteins should be vated products (Palta et al, 2014). The prominent 12–30% abun-
due to disease-related pathological disturbances (Fig 1E). We dance changes of these proteins in the plasma proteome (Fig 2C)
further analyzed the reproducibility of the measurements by calcu- are likely to underlie the well-known disturbances of the coagula-
lating the coefficients of variation (CVs) within triplicate tion system associated with cirrhosis and affecting the prolonged

ª 2019 The Authors Molecular Systems Biology 15: e8793 | 2019 3 of 16


Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

A
Figure 2. Reduced protein production and increased immunological
response in cirrhotic liver.
A Volcano plot of statistical significance against log2-fold change between
the cirrhosis group (N = 10) and non-NAFLD group (N = 18), showing
significantly differentially expressed proteins shaded in blue and down-
regulated liver-specific proteins color-coded according to the classification
of Human Protein Atlas (HPA). Significance was defined by independent
two-sample t-test (two-sided) corrected by permutation-based FDR of 0.05.
The percentage of down- and up-regulated “liver-specific” proteins is
indicated.
B Hierarchical clustering of significantly expressed proteins between the
cirrhosis group and non-NAFLD group. Intensities of proteins were log2-
transformed and Z-scored to normalize across individuals. Proteins involved
in different biological processes or belonging to different classes are
B indicated by color.
C Violin plot of mean fold changes for down- and up-regulated proteins. The
fold changes of down-regulated proteins were further calculated separately
for the three protein classes indicated in panel B.

bleeding time and increased thrombosis risk in these patients. A


second class consisted of carrier proteins such as apolipoprotein M
(APOM), apolipoprotein C1 (APOC1), and clusterin (CLU), which
play important roles in cholesterol metabolism and are likely to be
associated with cirrhosis-related dyslipidemia due to reduced liver
biosynthesis capacity (Chrostek et al, 2014). In addition, we found
important changes in proteins associated with hormone and vitamin
transportation such as retinol-binding protein (RBP4), transthyretin
(TTR), and vitamin D-binding protein (GC). A final class of proteins
are components of the complement system such as complement
component 6 (C6), complement factor H-related protein 3 (CFHR3),
and complement C1q tumor necrosis factor-related protein 3
(C1QTNF3) (Fig 2B). Members of the latter two classes were down-
regulated substantially, reflecting severe damage to liver function in
cirrhosis (Dataset EV1).
The majority of the up-regulated proteins are involved in
immune system regulation and inflammation such as complement
component C7 (C7), immunoglobulin chains (immunoglobulin
heavy variable 5–51, immunoglobulin J chain, and immunoglobulin
heavy constant mu), polymeric immunoglobulin receptor (PIGR),
vascular cell adhesion protein 1 (VCAM1), alpha-2 macroglobulin
(A2M), and alpha-1-antitrypsin (SERPINA1). Among these proteins,
A2M is an established marker of liver fibrosis (Naveau et al, 1994)
and already incorporated into biomarker blood tests, for example,
the SteatoTest (Poynard et al, 2005) and FibroTest (Imbert-Bismut
et al, 2001). In our study, A2M showed a 54% up-regulation in
patients with cirrhosis.
To investigate the tissue origin of the up-regulated and down-
regulated proteins, we used the data and annotation of the Human
Protein Atlas (HPA) (Kampf et al, 2014). Of note, 77% of the down-
C regulated proteins are classified as “liver-specific” with different
degrees of enrichment in the liver relative to the rest of the body
(see Materials and Methods). In stark contrast, only one of the 31
up-regulated proteins was annotated as liver-specific. This indicates
that down-regulation of liver proteins is a reflection of impaired
hepatic protein synthesis and secretion in patients with cirrhosis,
whereas up-regulation represents a systemic response of the body to
cirrhosis (Fig 2A and B). We conclude that liver cirrhosis is clearly
reflected in the plasma proteome profiles of patients through
changes in abundance of proteins originated from the diseased

4 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors


Lili Niu et al A protein marker panel for NAFLD Molecular Systems Biology

organ as well as specific biological processes that are directly related LGALS3BP, is likewise associated with fibrosis in diverse tissues, for
to the disease phenotype. instance, kidney, lung, liver, and heart (Li et al, 2014). Its critical
role in fibrosis has led to ongoing studies to develop galectin-3
Plasma proteins highly associated with NAFLD targeted anti-fibrotic drugs. An inhibitor targeting galectin-3, GR-
MD-02, is currently entering phase III trials for NASH and thus may
Having revealed proteome changes in plasma of patients with lead to a first therapy for the treatment of fatty liver disease with
cirrhosis, we next investigated if the same proteins were affected in cirrhosis (Henderson et al, 2006; Harrison et al, 2016). The relation
milder forms of liver disease. Our long-term motivation is to help between galectin-3 and LGALS3BP in the context of fibrogenesis
improve detection of NAFLD, which has proved to be challenging, remains to be investigated.
especially in relation to the identification of the subsets of patients We next asked if any of the NAFLD-associated plasma proteins
who will progress. NAFLD is very heterogeneous and is highly asso- were also altered in the plasma proteome of cirrhosis patients. This
ciated with obesity and T2D. The prevalence of ultrasonographic was the case for three of the six proteins, which changed also in
NAFLD in patients with T2D reaches 70% (Leite et al, 2009), much NAFLD: PIGR (298%), LGALS3BP (170%), and APOM (21%)
higher than the 25% of the general population with this diagnosis (Fig EV1A and B). Interestingly, the level of PIGR was up-regulated
(Bellentani et al, 2010). The presence of T2D is an independent in all three cohorts and 128% higher in the cirrhosis group
predictor of NASH and advanced fibrosis in NAFLD and vice versa compared to NAFLD, consistent with the notion that increased PIGR
(Williams et al, 2013; Bazick et al, 2015). Based on these findings, levels may be a novel marker of disease progression (Fig EV1C).
we decided to analyze our two NAFLD cohorts for common and LGALS3BP is also more strongly up-regulated in cirrhosis. The fact
potentially separate protein markers. that ALDOB, the protein with the largest fold change in NAFLD,
Our analyses showed a 186% increase of the polymeric showed no difference in cirrhosis may be due to the impaired
immunoglobulin receptor (PIGR), a 341% increase of fructose- protein synthesis and secretion that was apparent in the cirrhotic
bisphosphate aldolase B (ALDOB), and a 22% increase of vitro- liver (Fig EV1D).
nectin (VTN) in NAFLD patients with NGT (NAFLD-cohort 1)
compared with healthy controls (P < 0.001, Fig 3A and B). Global correlation map reveals a panel of five plasma proteins
Immunoglobulin heavy constant delta (IGHD) was down-regulated correlated with liver enzymes
in NAFLD patients by 67%; however, it was not significant in an
up-front one-way ANOVA across all five experimental groups and Routine blood tests provide evidence for abnormalities in the
was therefore excluded from this panel (Table EV1). liver, for example, inflammation and hepatocyte cell death. The
Previous studies have found up-regulation of immunoglobulin A most commonly tested liver enzymes are ALT, AST, alkaline
in patients with NAFLD (Inamine & Schnabl, 2018), and this has phosphatase (ALP), and gamma-glutamyl transferase (GGT). Note
been proposed as a biomarker of changes in the gut microbiota. No that levels of these enzymes do not necessarily reflect liver
previous evidence has associated NAFLD with immunoglobulin G function but rather liver damage. To evaluate if the plasma
(IgG) or M (IgM), and none of the included patients had serum concentration of liver enzymes was associated with other
immunoglobulins outside the normal range. In accordance with the proteins, we employed global correlation analysis of the plasma
increased levels of PIGR in patients with NAFLD, we also observed proteome (Wewer Albrechtsen et al, 2018). Filtering for a quanti-
up-regulated immunoglobulin chains in cirrhosis, among which tative data completeness for at least 70%, pairwise correlation of
most are derived from IgA, IgM, and IgG. These indicate an elevated the liver enzymes and the quantified proteins resulted in a data
immunological response in patients with NAFLD and cirrhosis. matrix with 431 protein levels and 21 clinical parameters. Corre-
In the NAFLD-cohort 2, we found statistically significant lating all variables with each other followed by hierarchical clus-
increases in the levels of PIGR by 157%, galectin-3 binding protein tering led to a global correlation map containing about 100,000
(LGALS3BP) by 102%, and AFM by 58% (Fig 3C and D). APOM was Pearson correlation coefficients, with clearly apparent clusters of
significantly decreased by 25%. Like IGHD, IgGFc binding protein co-varying proteins and variables (Fig 4A). Strikingly, one of
(FCGBP) was also statistically different between disease and control these groups contained all the above-mentioned liver enzymes, as
groups but was excluded from the panel due to insignificance in the well as five further proteins quantified by plasma proteome profil-
one-way ANOVA (Table EV1). This associates a panel of six ing (Fig 4B). One of these is PIGR, which we had identified as a
proteins, ALDOB, PIGR, VTN, LGALS3BP, AFM, and APOM, with protein with the potential to discriminate the non-NAFLD and the
NAFLD. At least one of them, PIGR, associated with NAFLD inde- NAFLD groups. The other four are also highly relevant for liver
pendently of the glucose tolerance. disease: APOE (apolipoprotein E), dipeptidyl peptidase-4 (DPP4),
AFM and LGALS3BP have already been suggested as potential TGFBI (transforming growth factor-beta-induced protein ig-h3),
markers for NAFLD (Bell et al, 2010; Wood et al, 2017), providing a and aminopeptidase N (ANPEP). Individually correlating each of
positive control. AFM is a human vitamin E-binding protein in the four liver enzymes with the plasma proteome confirmed
plasma that is primarily expressed in the liver and secreted into the very high statistical significance, such as that of ALT to ANPEP
circulation. It is strongly associated with components of the meta- (Pearson correlation 0.69; P < 10 7; Fig EV2A).
bolic syndrome (Dieplinger & Dieplinger, 2015), NAFLD, and alco- After identifying proteins correlating with the established liver
holic liver disease (ALD) (Bell et al, 2010; Liu et al, 2011; Neuman damage marker ALT by the global correlation map, we asked if we
et al, 2014). LGALS3BP, which was up-regulated in NAFLD and could reproduce these correlations in our very recently published
T2D, is a candidate biomarker for hepatitis C-related fibrosis and study on the effect of bariatric surgery in morbidly obese individuals
cirrhosis (Cheung et al, 2010). Interestingly, galectin-3, the ligand of on the plasma proteome (Wewer Albrechtsen et al, 2018).

ª 2019 The Authors Molecular Systems Biology 15: e8793 | 2019 5 of 16


Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

A C

B D

Figure 3. A panel of proteins strongly associated with NAFLD in human cohorts.


A Volcano plot of statistical significance against log2-fold change between NAFLD (N = 10) and controls (N = 10) in NAFLD subtype 1: NAFLD in normal glucose
tolerance. Significance is controlled by P-value (independent two-sample t-test, two-sided) and minimum fold change (s0 parameter in Perseus) indicated by the
cutoff curve, demonstrating significantly up-regulation of PIGR, ALDOB, and VTN.
B Box-and-whisker plot showing the distribution of mass spectrometric intensity values of three proteins in the first NAFLD cohort with median fold changes. The
yellow line is the median, the top and the bottom of the box represent the upper and lower quartile values of the data and the whiskers represent the upper and
lower limits for considering outliers (Q3+1.5*IQR, Q1-1.5*IQR) where IQR is the interquartile range (Q3–Q1). ***, P < 0.001 (independent two-sample t-test, two-sided).
Number of replicates is defined in panel (A).
C Volcano plot of statistical significance against log2-fold change between NAFLD (N = 8) and controls in NAFLD (N = 10) subtype 2: NAFLD in T2D, showing that AFM,
LGALS3BP, and PIGR are significantly up-regulated and APOM significantly down-regulated.
D Box-and-whisker plot showing the distribution of mass spectrometric intensity values of four proteins in the second NAFLD cohort with median fold changes.
Representation of boxes and whiskers is defined as in panel (B). Number of replicates is defined in panel (C).

Correlating protein levels with clinically determined ALT levels increase was not statistically significant after correction for multiple
across 175 plasma samples confirmed that ALDOB, PIGR, and hypothesis testing (P = 0.02). Studies have shown that increased
ANPEP are statistically significantly correlated, whereas TGFBI is hepatic expression of DPP4 is associated with NAFLD (Balaban
very close to the threshold (Fig EV3; Dataset EV2). et al, 2007; Miyazaki et al, 2012; Itou et al, 2013). Hepatocyte-
The global correlation map also showed a significant association specific DPP4 overexpression in mice increases body fat and
between DPP4 and liver enzymes. Although DPP4 was elevated by promotes hepatic steatosis, suggesting that this association is causa-
29% in NAFLD patients compared to non-NAFLD controls, this tive (Baumeier et al, 2017). Taken together with our proteome

6 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors


Lili Niu et al A protein marker panel for NAFLD Molecular Systems Biology

A B

Figure 4. Global correlation map of the plasma proteome and clinical variables in human cohorts.
A Pairwise correlation of proteins and clinical variables over the 48 study participants, resulting in a matrix of correlation coefficients where each variable is compared
to all others. Variables with a high positive correlation to each other will cluster together in groups of red rectangles (high correlations). Negative correlation is
indicated in blue patches.
B The magnified area highlights a cluster of variables that contains the four main clinical measurements for liver diseases (blue names) as well as five proteins, which
were quantified by plasma proteome profiling (black gene names).

profiling results, levels of soluble DPP4 in plasma may represent an generically be applied across species to determine to which degree
even stronger candidate as an indicator of liver damage. Biologi- the biological responses are similar between them. There are well-
cally, it is a serine protease targeting incretins such as glucagon-like established NAFLD mouse models that are generated by high-fat diet
peptide-1 (GLP-1) and glucose-dependent insulinotropic peptide (HFD), and these are commonly used to study the effect of incretin
(GIP). DPP4 inhibitors are widely used as blood glucose-lowering agonist treatment for diet-induced obesity. We took advantage of a
agents in T2D (Deacon, 2018). DPP4 also plays a role in the degra- NAFLD mouse cohort designed to determine the effects of incretin
dation of extracellular matrix, the imbalance of which is a hallmark agonists—alone and in combination—on improving the metabolic
of liver fibrosis (Itou et al, 2013). phenotype. It consisted of mice with mild and severe NAFLD
Interestingly, the correlation analysis showed a connection induced by high-fat diet for either up to two months or more than
between liver enzymes, TGFBI and ANPEP, which are also involved six months (Fig 5A). The mice with mild NAFLD were treated with
in the remodeling of extracellular matrix (Bieche et al, 2005). vehicle for 21 days, while the mice with severe NAFLD were
However, nothing is known about a possible role of TFGBI and randomly divided into four groups, treated with either vehicle, GIP
ANPEP in liver disease. The up-regulated levels of the extracellular receptor agonist, GLP-1 receptor agonist, or GIP/GLP-1 co-receptor
matrix protein TGFBI that we observed in these patients could agonist for 15 days while maintaining a high-fat diet. These mono
reflect early scar tissue formation in the liver, which in turn may and dual receptor agonists have previously been reported to reduce
induce increased levels of DPP4 and ANPEP. body weight and hepatic steatosis. The unimolecular dual GIP/GLP-
1 receptor agonist has shown superior efficacy relative to GIP or
Investigation of the plasma proteome in a mouse model of GLP-1 receptor agonists in the past (Frias et al, 2017; Jall et al,
high-fat diet-induced NAFLD 2017). Therefore, in our cohorts it is expected to have a larger effect
than the two other treatments, providing four groups of disease
As a MS-based method, plasma proteome profiling does not rely on trajectories of mice with different severity of NAFLD. This study
specific protein epitopes (unlike most antibodies) and can design allows us to investigate the potential of our newly identified

ª 2019 The Authors Molecular Systems Biology 15: e8793 | 2019 7 of 16


Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

A B C

D E F

G H I

J K L

M N O

Figure 5.

8 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors


Lili Niu et al A protein marker panel for NAFLD Molecular Systems Biology

◀ Figure 5. Plasma proteome changes in a HFD-induced NAFLD mouse model.


A Mouse cohort design.
B Box-and-whisker plot showing the distribution of log2-intensity values of body weight across five groups: HFD_1-2 m (N = 6), HFD > 6 m (N = 6), GIP (N = 7),
GLP-1 (N = 8), and GLP-1/GIP (N = 6).
C Volcano plot of statistical significance against log2-fold change between mice on > 6 months of HFD and mice on 1–2 months of HFD. Significance is controlled by
FDR-corrected P-value and minimum log2-fold change of 1 indicated by the blue-dotted line, demonstrating that Saa1, Pigr, Aldob, Lap3, Enpep, and Dpp4 are
significantly up-regulated.
D–O Box-and-whisker plot showing the distribution of log2-intensity values of statistical significantly regulated proteins across five groups. Number of replicates is
defined in panel (B). The yellow line is the median, the top and the bottom of the box represent the upper and lower quartile values of the data and the whiskers
represent the upper and lower limits for considering outliers (Q3+1.5*IQR, Q1-1.5*IQR) where IQR is the interquartile range (Q3–Q1).
Data information: Significance was defined by independent t-test (two-sided) followed by Benjamini–Hochberg correction for multiple hypothesis testing with a
significance level of *P < 0.05, **P < 0.01, and ***P < 0.001.

marker candidates to monitor treatment effects of this co-agonist in insulin resistance (Ghorpade et al, 2018). Interestingly, along with
NAFLD. Dpp4, two other aminopeptidases—Lap3 (leucine aminopeptidase
Mice with severe NAFLD exhibited a 27% increase in body weight 3) and Enpep (glutamyl aminopeptidase)—also increased in
compared to mice with mild NAFLD. Upon treatment with GLP-1 severe NAFLD compared to mild NAFLD, by 273 and 327%,
mono-agonist or GIP/GLP-1 co-agonist, they lost 19% or 24% of body respectively, followed by a decrease upon treatment (Fig 5K and
weight, respectively (Fig 5B). There was no significant weight loss L). Protein–protein correlation analysis of the entire dataset of 82
upon GIP agonist treatment alone. T-test analysis of the quantitative mice in our experimental setup (Dataset EV3) reveals that Lap3,
changes in the plasma proteome between severe and mild NAFLD Enpep, and Me1 (NADP-dependent malic enzyme) correlate with
mice resulted in 71 significant hits—40 up-regulated and 31 down- Pigr, Aldob, and Dpp4 (Fig EV4). This is consistent with our
regulated (Fig 5C–O). Pigr and Aldob, which we had found to be asso- human study, where PIGR, DPP4, and ANPEP co-vary with the
ciated with NAFLD in the human cohorts, were also highly signifi- four classic liver enzymes.
cantly up-regulated upon progression in mice (Fig 5C). Median We also found that Apoa4, a major component of high-
plasma levels of Pigr in severe NAFLD mice were 780% higher than density lipoprotein and chylomicrons, increased by 30% upon
those in mild NAFLD (Fig 5D), an even greater change than in the progressing from mild to severe NAFLD. This was partially
human cohorts. While Pigr levels decreased upon treatment, especially reversed upon GLP-1 receptor agonist and GIP/GLP co-agonist
in the GLP-1 and combination treatment, these changes were not treatment (21 and 31% decrease, respectively; Fig 5M). Adipo-
significant. Like Pigr, plasma levels of Aldob are drastically increased nectin negatively correlates with body weight and its plasma
in NAFLD compared to mild NAFLD (830%, Fig 5E). These increases concentration is reduced in patients with NASH (Balmer et al,
were mostly reversed upon treatment by GLP-1 and GIP/GLP-1 recep- 2010). We found adiponectin levels decreased by 23% in severe
tor agonist (down-regulation by 62 and 72%, respectively). Sord (sor- NAFLD compared to mild NAFLD, followed by increased levels
bitol dehydrogenase), another enzyme involved in fructose in the three treatment groups compared to the non-treatment
metabolism, followed a very similar pattern trajectory (Fig 5F). group (Fig 5N).
The plasma levels of three proteins that are involved in the acute
phase reaction increased strongly upon progression from mild to
severe NAFLD: Saa1 (serum amyloid A-1 protein) by 3,250%, Apcs Discussion
(serum amyloid P-component) by 440%, and Orm2 (alpha-1-acid
glycoprotein 2) by 270% (Fig 5G–I). In a weight loss study in Currently established protocols in clinical practice for the diagnosis
humans, we had previously classified these proteins as members of and follow-up of NAFLD have certain limitations; for instance, they
a plasma protein panel indicating increased systemic inflammation may not be sufficiently sensitive at early disease stages. MS-based
(Geyer et al, 2016b). Apcs and Orm2 are primarily expressed in the proteomics technology holds great potential in generating novel
liver, and Saa1, which is highly up-regulated in response to insights into disease mechanism and discovering new biomarkers.
inflammation and tissue injury, is expressed in both liver and To identify novel proteins associated with NAFLD and to under-
adipose tissue according to the Human Protein Atlas. Given that stand the effect of liver cirrhosis on the plasma proteome, we here
inflammation of visceral adipose tissue contributes to the develop- analyzed plasma samples of 48 participants with our streamlined
ment of insulin resistance and steatohepatitis, increased inflamma- plasma proteomics workflow, enhanced with a novel MS-acquisition
tory factors in the blood in our experimental setup could originate method featuring high sensitivity.
from both liver and adipose tissue, specifically hepatocytes, adipo- Our results revealed clinically interesting proteome changes in
cytes, and resident or recruited macrophages. NAFLD and in cirrhosis, where we found dysregulation of proteins
Plasma levels of Dpp4 increased by 322% in abundance in associated with thrombosis and homeostasis. These findings are the
severe NAFLD (Fig 5J) compared to mild NAFLD. This could be proteomic reflection of the hemostatic complications of liver
due to tissue leakage or proactive secretion mediating tissue disease, in particular increased risk of bleeding and venous throm-
crosstalk via circulating factors. A recent study showed that boembolism in patients with cirrhosis (Yang et al, 2014). Vitamin A
obesity in mice stimulates hepatocytes to synthesize and secrete and D deficiency is a common feature in chronic liver disease. We
Dpp4, which promotes visceral adipose tissue inflammation and found proteins involved in hormone and vitamin transportation to

ª 2019 The Authors Molecular Systems Biology 15: e8793 | 2019 9 of 16


Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

be altered in patients with cirrhosis, for example, retinol-binding from both animal studies and human studies suggests that dietary
protein (RBP4) and transthyretin (TTR). Reduced RBP4 levels are in added fructose intake when consumed in excess is a principal driver
concordance with a previous study (Bahr et al, 2009). Both RBP4 of NAFLD and its deleterious consequences (DiNicolantonio et al,
and TTR are primarily produced in hepatocytes and transport retinol 2017). We also observed increased plasma ALDOB levels in the
to peripheral tissues. In the liver, hepatocytes secret RBP4, to enable mouse NAFLD model. This increase may be a consequence of cell
retinoid storage in the hepatic stellate cells (HSC), accounting for as damage, as hepatocytes inflamed due to excessive fat accumulation
much as 50–60% of the total retinoid present in the body. The insuf- may leak out ALDOB. Since that is visible already in NAFLD, it may
ficient levels of RBP4 observed by proteomics might therefore be more sensitive than the usually measured liver enzymes AST and
contribute to vitamin A deficiency. ALT.
Among the interesting similarities between patients with cirrho- In our analysis, we also correlated all proteomics and clinical
sis and patients with NAFLD was PIGR, a little studied protein variables with each other to identify proteins correlating with
produced in the GI tract and endothelial cells but also in the liver. known markers for liver diseases under pathophysiological
Strikingly, PIGR increased in abundance in all our three cohorts in conditions. In this global correlation analysis, involving close to
line with the severity of liver damage and the up-regulation is most 100,000 individual correlation coefficients, proteins that are tightly
dramatic in cirrhosis. We further validated this novel finding in a co-regulated likely share a common origin or (patho)physiological
mouse model of NAFLD induced by high-fat diet. PIGR is a receptor, pathway. Remarkably, in this unbiased analysis, four clinical
which mediates transcytosis of immunoglobulins from the basolat- markers used in liver disease—ALT, AST, ALP, and GGT—clus-
eral to the apical surface of the epithelia, facilitating the secretion of tered into a single group together with a panel of five proteins of
IgA and IgM. PIGR levels also co-varied with AST, ALT, ALP, and special interest—PIGR, DPP4, ANPEP, TGFBI, and APOE. Among
GGT, four clinical liver markers. The inter-individual variation in these proteins are two enzymes involved in degradation of the
plasma PIGR levels in controls was relatively small, much lower ECM (ANPEP, DPP4) and an ECM protein (TGFBI). This may
than that in the NAFLD and cirrhosis cohorts. This finding was also again reflect the remodeling of the ECM in the liver, which is part
observed in patients with T2D with no liver damage, thus making of the progression to liver fibrosis. Even more interesting, DPP4,
PIGR an interesting biomarker candidate for the inclusion in liver which proteolytically cleaves GLP-1 and GIP, has been reported to
damage tests. increase in both plasma and liver tissue of NAFLD patients
The plasma proteome changed much less in NAFLD than in (Miyazaki et al, 2012; Anoop et al, 2017). DPP4 inhibitors are
cirrhosis and globally the plasma proteome profiles had few signifi- also widely used diabetes drugs worldwide. Interestingly, there
cant outliers, both in the cohort with normal glucose tolerance and are currently clinical trials evaluating the efficacy and safety of
in the T2D cohort. This presumably reflects the resilience and regen- DPP4 inhibitor sitagliptin for the treatment of NAFLD (Fukuhara
erative capacity of the liver and is also in line with the fact that et al, 2014) and in steatosis irrespective of diabetes in NASH
NAFLD or early cirrhosis is often asymptomatic and clinically diffi- (Alam et al, 2018). Taken together with our proteomic data, it
cult to detect. Circulating RBP4 levels are among the changes that would be interesting to further investigate the potential of circulat-
have been controversially discussed in the literature, with some ing DPP4 levels to serve as prognostic marker for both T2D and
studies finding higher levels in NAFLD (Terra et al, 2013) and some NAFLD. The same may apply for the other four proteins in this
reporting unchanged levels (Zhou et al, 2017). This parallels our five-protein panel as they also correlate very well with liver
plasma proteome data, in which RBP4 was clearly significant in enzymes of hepatic damage. Remarkably, we also observed this
cirrhosis but not in the NAFLD sub-groups. close correlation between DPP4 and PIGR in our mouse dataset,
There is increasing awareness of the importance of screening for where we found that two additional aminopeptidases, Lap3 and
the presence of NAFLD, especially in patients with T2D. However, Enpep, similarly were increased under NAFLD. Together, our anal-
this requires markers or marker panels that are specific to NAFLD ysis points to an important relationship between aminopeptidases
and ideally also associated with its progression to fibrosis and cirrho- and NAFLD.
sis. Our plasma proteome profiling experiments of three cohorts In conclusion, in this study we found changes in the plasma
provide a step in this direction. We identified a panel of six proteins proteome of patients with cirrhosis and patients with NAFLD
in the two NAFLD subtypes: three in NAFLD without T2D (PIGR, that are clearly linked to the underlying disease manifestations
ALDOB, and VTN) and four in NAFLD with T2D (PIGR, LGALS3BP, and clinical observations. PIGR and ALDOB are in our panel of
AFM, and APOM). Of these, AFM and LGALS3BP have been reported six proteins significantly associated with NAFLD and were also
as potential markers for NAFLD (Bell et al, 2010; Wood et al, 2017). validated in a mouse NAFLD cohort making them interesting
AFM has been closely linked to metabolic syndrome, insulin resis- candidates for follow-up studies. PIGR is a particularly promising
tance, NAFLD, and alcoholic liver disease (Bell et al, 2010; Liu et al, protein as its association with NAFLD and liver cirrhosis is
2011; Neuman et al, 2014; Kollerits et al, 2017). LGALS3BP has novel, and its levels in plasma are highly correlated with DPP4,
already been used to build multi-component classifiers for the a widely used drug target in the treatment of T2D. This correla-
prediction of NAFLD (Wood et al, 2017) and fibrosis in patients with tion holds true in both human and mouse cohorts. Blocking the
hepatitis C infection (Cheung et al, 2010). The elevated levels of enzymatic function of the other aminopeptidases may have
LGALS3BP and vitronectin (VTN), another extracellular matrix effects on the fibrotic process in NAFLD progression. To evaluate
(ECM) protein, are likely a reflection of remodeling of the ECM in the potential and specificity of the above proteins to be
liver disease. developed or incorporated into liver disease panels, we plan to
ALDOB is a glycolytic enzyme in the fructose catabolism pathway perform plasma proteome profiling on larger and more fine-
that also plays a role in gluconeogenesis and lipogenesis. Evidence grained cohorts.

10 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors


Lili Niu et al A protein marker panel for NAFLD Molecular Systems Biology

Materials and Methods


Reagents and Tools table

Reagent/Resource Reference or Source Identifier or Catalog Number


Experimental Models
Blood plasma samples (H. sapiens) (Junker et al, 2015, 2016) N/A
Blood plasma samples (M. musculus) This study N/A
Chemicals, enzymes and other reagents
Isopropanol Sigma Aldrich/Merck Cat # 67-63-0
Formic acid Sigma Aldrich/Merck Cat # 64-18-6
Acetonitrile Sigma Aldrich/Merck Cat # 75-05-8
Pierce Dimethylsulfoxide (DMSO), LC-MS Thermos Fisher Scientific 85190
Grade
Water, OptimaTM LC/MS Grade Fisher Chemical Cat # W64
25% LC-MS grade ammonia Merck Millipore Cat # 533003
Trifluoroacetic acid Sigma Aldrich/Merck Cat # 76-05-1
Multiple Affinity Removal Column Agilent 5188-5341
Human 6
Seppro Protein Depletion Sigma Aldrich SEP010
Software
MaxQuant (1.5.8.6) https://maxquant.org/ N/A
Perseus (1.5.50) https://maxquant.net/perseus/ N/A
Jupyter Notebook https://jupyter.org/ N/A
Other
Empore SPE SDB-RPS disk Sigma Aldrich/Merck Cat # 66886-U
96-Well Plates Thermo Fisher Cat # AB-1300
Silicone sealing mat, for 96 well Nerbe Plus Cat # 04-090-0000
PCR-plates
Reprosil-Pur Basic C18, 1.9 µm Dr. Maisch Gmbh Cat # r119.b9
PicoFrit self-pack columns Pico FRIT Cat # PF360-75-15-N-5
iST’ sample preparation kit PreOmics GmbH Cat # P.O. 00001
ThermoMixer® Eppendorf Cat # 460-0223
Bravo Automated Liquid Handling Agilent Cat # G5409A
Platform
NanoDropTM One/OneC Microvolume Thermo Fisher Cat # ND-ONEC-W
UV-Vis Spectrophotometer
Concentrator plus Eppendorf Cat # F-45-48-11
EASY-nLC 1200 System
TM
Thermo Fisher Cat # LC140
Q ExactiveTM HF Hybrid Quadrupole- Thermo Fisher Cat # IQLAAEGAAPFALGMBFZ
OrbitrapTM Mass Spectrometer
Q ExactiveTM HF-X Hybrid Quadrupole- Thermo Fisher Cat # 0726042
OrbitrapTM Mass Spectrometer

Ethical approval oral and written informed consent was obtained from all partici-
pants.
The study protocol was approved by the scientific-ethical committee
of the Capital region of Denmark (H-1-2011-082) and registered Study design
with the Danish Data Protection Agency (2011-41-6410) and
ClinicalTrials.gov (reg. no. NCT01492283). The study was conducted Plasma samples, all taken in the fasting state, were derived from two
according to the principles of the Declaration of Helsinki, and previously published studies. The details including corresponding

ª 2019 The Authors Molecular Systems Biology 15: e8793 | 2019 11 of 16


Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

clinical data and laboratory data are in Junker et al (2016, 2015), Methods and Protocols
and the groups are briefly described below. Patients had I) normal
glucose tolerance (NGT) and no liver disease, II) NGT and Plasma sample preparation
NAFLD, III) T2D without liver disease, IV) T2D with NAFLD, and All plasma samples were prepared according to the previously
V) cirrhosis. NAFLD was diagnosed based on histology and graded published plasma proteome profiling pipeline on an automated liquid
according to hepatic fat infiltration: no NAFLD (< 5% fat infiltra- handling system (Agilent Bravo) in a 96-well plate format (Geyer et al,
tion), mild (5–33%), moderate (33–66%), and severe (> 66%). 2016a). In brief, proteins were denatured, reduced, alkylated, and
Non-alcoholic steatohepatitis (NASH) and fibrosis were graded digested and peptides purified on StageTips (Kulak et al, 2014) using
according to the NAFLD activity score. Eight of the 20 patients reagents from the PreOmics “iST” Kit (P.O. 00001, PreOmics GmbH).
with NAFLD also had NASH/fibrosis (all below fibrosis score 5). In detail:
Cirrhosis was diagnosed histologically and clinically, based on 1 Transfer of 5 ll of blood plasma sample into an Eppendorf 96-
signs of decompensation (e.g., ascites). well plate at 4°C on an Agilent Bravo liquid handling system
Individuals with T2D were diagnosed according to World (Plasma plate).
Health Organization criteria. Exclusion criteria included weekly 2 Make a 1:10 dilution by adding 45 ll of SDC reduction and alkyla-
alcohol consumption of more than seven units for women and tion buffer (included in PreOmics “iST” kit) into each well of the
14 units for men, treatment with steatogenic drugs within Plasma plate, mix thoroughly by pipetting 50 times up and down
3 months prior to inclusion, anemia, inflammatory bowel disease, for a volume of 40 ll, and centrifuge the plate up to 300 × g.
gut resection, increased creatinine (> 150 lmol/l), albuminuria, 3 Pipet 20 ll of tenfold diluted plasma into a new 96-well plate
or other chronic diseases. Controls were healthy (and matched (Digestion plate).
with age, BMI, and gender) with no family history of diabetes, 4 Heat the plate at 95°C for 10 min.
signs of liver disease (based on patient history, biochemical 5 Move the Digestion plate to room temperature and cool it down
measurements, and ultrasound assessment), or other chronic for 5 min. Meanwhile, prepare fresh trypsin/LysC mix in 0.05
diseases. (lg/ll) (total volume calculated by 20 ll per sample, 1:100
micrograms of enzyme to micrograms of protein).
Mouse experiments 6 Add 20 ll of trypsin/LysC mix into each well to a final volume
of 40 ll.
To induce substantial obesity and NAFLD, we fed 8-week-old male 7 Heat the Digestion plate at 37°C for 4 h (enzymatic digestion).
C57Bl6/J mice (Charles River, River Laboratories, Wilmington, MA, 8 Quench the reaction by adding 40 ll of the PreOmics washing
USA) with a high-fat, high-sugar diet (HFD) comprising 58% kcal buffer 1, and mix thoroughly by pipetting 20 times up and down.
from fat (D12331; Research Diets, New Brunswick, NJ, USA) for 9 Prepare 96 StageTips on a home-made 3D-printed centrifugation
32 weeks. Another 6 male C57Bl6/J mice were maintained on regu- block (2-plug SDB-RPS material (thickness 0.5  0.05 mm) in
lar chow diet and switched to HFD for 5 weeks at an age of 14-gauge).
34 weeks to induce mild obesity and NAFLD. Mice were single- or 10 Transfer 24 ll of the mixture onto the StageTips. Centrifuge the
double-housed on a 12:12-h light–dark cycle at 22°C with free access StageTip block at 1,500 × g for 15 min.
to water and food. Sample size estimation was done the same as in 11 Add 150 ll of the PreOmics washing buffer 1 to the StageTips,
the human study (see Study design). and centrifuge at 1,500 × g for 15 min.
12 Add 150 ll of the PreOmics washing buffer 2 to the StageTips,
Mouse pharmacology and centrifuge at 1,500 × g for 15 min.
13 Elute the peptides by adding 60 ll of elution buffer (1% ammo-
The synthesis, purification, and characterization of the GLP-1/GIP nia in 80% acetonitrile), centrifuge at 1,500 × g for 15 min, and
co-agonist and the single GLP-1 and GIP mono-agonist controls were collect peptides in PCR tube-stripes (0.2 ml volume).
described previously, and the peptides were used without any 14 Concentrate the peptide mixture by Speed-Vac at 60°C under
further chemical modification or change in formulation (Finan et al, vacuum for 90 min.
2013). Whole-body composition (fat and lean mass) was measured 15 Re-suspend the peptide mixture in 40 ll of buffer A* (2%
with nuclear magnetic resonance technology (EchoMRI, Houston, acetonitrile, 0.1% TFA in ddH2O).
TX, USA). For the treatment study in mice with severe NAFLD, DIO
mice maintained for 32 weeks on HFD and then randomized to High-pressure liquid chromatography and mass spectrometry
either vehicle, GLP-1, GIP, or GLP-1/GIP treatment according to
body weight and body composition. Mice with mild NAFLD were Samples were measured using LC-MS instrumentation consisting of
fed HFD for 5 weeks and then treated with vehicle for the same time an EASY-nLC 1200 system coupled to a nano-electrospray ion
period as the other mouse groups. Compounds were administered source and a Q Exactive HF Orbitrap (all Thermo Fisher Scientific).
in a vehicle of PBS (Gibco) and were given by daily subcutaneous Purified peptides were separated on 40-cm HPLC columns (ID:
injections at a dose of 10 nmol/kg at a volume of 5 ll per g body 75 lm; in-house packed into the tip with ReproSil-Pur C18-AQ
weight for 15 days. The investigators were not blinded to group 1.9 lm resin (Dr. Maisch GmbH)). For each LC-MS/MS analysis,
allocation during the in vivo experiment. All procedures were around 0.5 lg peptides was injected for the 45-min gradients and
approved by the Animal Use and Care Committee of Bavaria, 1 lg for the fractions of the deep plasma dataset.
Germany, in accordance with the Guide for the Care and Use of Additionally, we established very deep plasma proteome
Laboratory Animals. libraries. We pooled samples from all healthy individuals, NAFLD

12 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors


Lili Niu et al A protein marker panel for NAFLD Molecular Systems Biology

patients, and liver cirrhosis patients separately and depleted the 14 normal distribution (downshifted mean by 1.8 standard deviation
highest abundant plasma proteins by serial depletion with a top6 (SD) and scaled SD (0.3) relative to that of proteome abundance
(Multiple Affinity Removal Column Human 6; Agilent) and top14 distribution). This further resulted in a final dataset (Dataset EV1,
depletion kits (Seppro Protein Depletion; Sigma-Aldrich). After tab 4) with which we performed the statistical analysis (except for
digestion, the peptides were separated by the Spider Fractionator global correlation analysis, which we filtered for 70% data
(Kulak et al, 2017) into 24 fractions. This library was combined completeness across all samples instead of one sub-group, resulting
with an additional peptide library that had been established in the in 431 protein groups, without imputation (Dataset EV1, tab 5)).
same way in a separate study (Wewer Albrechtsen et al, 2018). For mouse plasma, DIA raw data were analyzed with Spectro-
Peptides were loaded in buffer A (0.1% formic acid, 5% DMSO (v/ naut Pulsar XTM with an in-house generated spectra library of mouse
v)) and eluted with a linear 35-min gradient of 3–30% of buffer B plasma that contains 8,899 peptides and 1,458 protein groups
(0.1% formic acid, 5% DMSO, 80% (v/v) acetonitrile), followed by a 7- (mouse FASTA UniProt FASTA database version 201806 containing
min increase to 75% of buffer B and a 1-min increase to 98% of buffer 92,096 entries). Spectronaut Pulsar XTM was used with default
B, and a 2-min wash of 98% buffer B at a flow rate of 450 nl/min. settings for DIA data with the decoy generation set to “mutated”.
Column temperature was kept at 60°C by a Peltier element containing
an in-house-developed oven. For human plasma, MS spectra were Bioinformatic analysis
acquired with a Top15 data-dependent MS/MS scan method (topN
method) for the library and with the BoxCar scan method for study Bioinformatic analysis was performed in the Perseus platform
samples (Meier et al, 2018). The target value for the full scan MS spec- (Tyanova et al, 2016) and Python scripts. Two-sample Student’s t-
tra was 3 × 106 charges in the 300–1,650 m/z range with a maximum test was used to determine the significantly changed proteins
injection time of 55 ms and a resolution of 120,000 at m/z 200. Frag- between disease and control groups with a permutation-based FDR
mentation of precursor ions was performed by higher-energy collisional of 0.05. For significant hits, minimal fold changes together with P-
dissociation (HCD) with a normalized collision energy of 27 eV. MS/ values (controlled by the s0 parameter in Perseus) were used with a
MS scans were performed at a resolution of 15,000 at m/z 200 with an permutation-based FDR of 0.05 resulting from an s0 of 0.01. For
automatic gain control (AGC) target value of 5 × 104 and a maximum mouse plasma analysis, we used scipy.stats.ttest_ind to calculate t-
injection time of 25 ms. For mouse plasma, MS spectra were acquired test probabilities for the means of two independent samples. This
with a data-independent acquisition (DIA) method. The DIA-MS was corrected for an FDR of 0.05 by the Benjamini–Hochberg
method consisted of an MS1 scan from 350 to 1,650 m/z range (AGC method. Results were filtered to have both a significant FDR-
target of 3 × 106, maximum injection time of 50 ms) at a resolution of corrected P-value and a minimum log2-fold change of  1.
120,000 and 22 DIA segments (AGC target of 3 × 106, maximum injec- Shapiro–Wilk test was applied to test normality of each individual
tion time of 54 ms) at a resolution of 30,000 (Dataset EV3). Normalized protein across all five groups in the human study, and 75–80% of
stepped collision energy was set to 25, 27.5, 30, with a default charge proteins are normally distributed. Levene’s test was used to assess
state of 2. The acquisition of samples was randomized to avoid bias. the equality of variances for each individual protein calculated for
two groups in each of the NAFLD sub-cohorts. As a result, 95% of
Data analysis the proteins have equal variances. All six proteins reported in this
study associated with NAFLD met the assumptions of the tests.
For human plasma, mass spectrometric raw files were analyzed in Liver-specific proteins were annotated according to the Human
the MaxQuant environment v.1.5.8.6 (Cox & Mann, 2008) employing Protein Atlas, which defines “liver enriched”, “group enriched”, and
the Andromeda search engine (Cox et al, 2011). The MS/MS spectra “liver enhanced” proteins with at least 500% higher mRNA levels in
were searched against the human UniProt FASTA database (version liver compared to all other tissues, at least 500% higher mRNA
201704, 157,510 entries). Enzyme specificity was set to trypsin with levels in a group of 2–7 tissues compared to the rest, and at least
a maximum of 2 missed cleavages, and the search included cysteine 500% higher mRNA levels in the liver compared to average levels in
carbamidomethylation as fixed modification and oxidation on all tissues, respectively.
methionine and N-terminal acetylation as variable modifications Fold changes between conditions were calculated as Condition
with a minimum required peptide length of 7 amino acids. A false A/Condition B -1 for protein label-free (LFQ) intensities. The
discovery rate (FDR) of 0.01 was set at the peptide and protein changes are indicated in percentage (e.g., increased by 62% and
levels. Study samples were analyzed together with our in-house decreased by 27%).
generated matching library and the “match between runs” algorithm
(Nagaraj et al, 2012). Label-free quantitation (LFQ) was performed
with a minimum ratio count of 2 and normalized in a separate group Data availability
from the library raw files (Cox et al, 2014).
After filtering for “reverse”, “only identified by site”, “contami- The datasets in this study are available in the following databases:
nants”, and at least two valid values in any of the three technical
replicates, we came to a dataset containing 684 protein groups • Proteomic dataset for the human cohorts: PRIDE archive
(Dataset EV1, tab 2). We took the median value from the technical PXD011839 (Vizcaino et al, 2016) (https://www.ebi.ac.uk/pride/
triplicates and further filtered the dataset for 70% data completeness archive/).
in at least one experimental group. This resulted in 520 protein • Proteomic dataset for the mouse model: PRIDE archive PXD012056.
groups with 11% missing values (Dataset EV1, tab 3). We then
replaced the missing values by drawing random samples from a Expanded View for this article is available online.

ª 2019 The Authors Molecular Systems Biology 15: e8793 | 2019 13 of 16


Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

Acknowledgements Bahr MJ, Boeker KH, Manns MP, Tietge UJ (2009) Decreased hepatic RBP4
We thank all members of the Proteomics and Signal Transduction Group (Max secretion is correlated with reduced hepatic glucose production but is not
Planck Institute) and the Clinical Proteomics group, in particular Atul Desh- associated with insulin resistance in patients with liver cirrhosis. Clin
mukh for his help and discussions and Igor Paron, Christian Deiml, Korbinian Endocrinol 70: 60 – 65
Mayr, Gaby Sowa, Jeppe Madsen, Martin Rykær, and Simone Schopper for Balaban YH, Korkusuz P, Simsek H, Gokcan H, Gedikoglu G, Pinar A, Hascelik
technical assistance. The work carried out in this project was partially G, Asan E, Hamaloglu E, Tatar G (2007) Dipeptidyl peptidase IV (DDP IV) in
supported by the Challenge Programme MicrobLiver funded by Novo Nordisk NASH patients. Ann Hepatol 6: 242 – 250
Foundation (Grant No. NNF15OC0016692), the Max Planck Society for the Balmer ML, Joneli J, Schoepfer A, Stickel F, Thormann W, Dufour JF (2010)
Advancement of Science, the European Union’s Horizon 2020 research and Significance of serum adiponectin levels in patients with chronic liver
innovation program (Grant Agreement No. 686547: MSmed project), the Novo disease. Clin Sci 119: 431 – 436
Nordisk Foundation for the Clinical Proteomics group (Grant NNF15CC0001), Baumeier C, Schlüter L, Saussenthaler S, Laeger T, Rödiger M, Alaze SA,
the Novo Nordisk Foundation for the Copenhagen Bioscience PhD Program Fritsche L, Häring H-U, Stefan N, Fritsche A, Schwenk RW, Schürmann A
(NNF16CC0020906), and the Deutsche Forschungsgemeinschaft (DFG: (2017) Elevated hepatic DPP4 activity promotes insulin resistance and
SFB1123-A4). non-alcoholic fatty liver disease. Mol Metab 6: 1254 – 1263
Bazick J, Donithan M, Neuschwander-Tetri BA, Kleiner D, Brunt EM, Wilson L,
Author contributions Doo E, Lavine J, Tonascia J, Loomba R (2015) Clinical model for NASH and
LN and PEG designed the study and performed and interpreted the MS-based advanced fibrosis in adult patients with diabetes and NAFLD: guidelines
proteomic analysis of plasma samples of the cohort, wrote the paper, and for referral in NAFLD. Diabetes Care 38: 1347 – 1355
generated the figures. NJWA designed the study and contributed to the analy- Bedogni G, Bellentani S, Miglioli L, Masutti F, Passalacqua M, Castiglione
sis and writing of the paper. PVT performed MS-based proteomic analysis of A, Tiribelli C (2006) The Fatty Liver Index: a simple and accurate
plasma samples and wrote the paper. SS administered co-agonist treatment predictor of hepatic steatosis in the general population. BMC
and performed blood sampling and body weight measurements in mouse Gastroenterol 6: 33
cohorts. JJH, FKK, TV, LLG, and AJ provided patient material and clinical data Bell LN, Theodorakis JL, Vuppalanchi R, Saxena R, Bemis KG, Wang M,
and revised the manuscript. AS and SD revised the manuscript. MHT, KS, SMH, Chalasani N (2010) Serum proteomics and biomarker discovery across the
and TDM designed and oversaw mouse experiment, provided mouse material, spectrum of nonalcoholic fatty liver disease. Hepatology 51: 111 – 120
and revised the manuscript. MM designed and interpreted the MS-based Bellentani S, Scaglioni F, Marino M, Bedogni G (2010) Epidemiology of non-
proteomic analysis of patient plasma, supervised and guided the project, and alcoholic fatty liver disease. Dig Dis 28: 155 – 161
wrote the paper. Bieche I, Asselah T, Laurendeau I, Vidaud D, Degot C, Paradis V, Bedossa P,
Valla DC, Marcellin P, Vidaud M (2005) Molecular profiling of early stage
Conflict of interest liver fibrosis in patients with chronic hepatitis C virus infection. Virology
The authors declare that they have no conflict of interest. 332: 130 – 144
Bril F, Cusi K (2017) Management of nonalcoholic fatty liver disease in
patients with type 2 diabetes: a call to action. Diabetes Care 40: 419 – 430
References Burt AD, Ferrell LD, Hubscher SG (2015) MacSween’s pathology of the liver.
Edinburgh: Churchill Livingstone/Elsevier
Aebersold R, Mann M (2003) Mass spectrometry-based proteomics. Nature Cassidy S, Syed BA (2016) Nonalcoholic steatohepatitis (NASH) drugs market.
422: 198 – 207 Nat Rev Drug Discov 15: 745 – 746
Aebersold R, Mann M (2016) Mass-spectrometric exploration of proteome Chang PE, Goh GB-B, Ngu JH, Tan HK, Tan CK (2016) Clinical applications,
structure and function. Nature 537: 347 – 355 limitations and future role of transient elastography in the
Alam S, Ghosh J, Mustafa G, Kamal M, Ahmad N (2018) Effect of sitagliptin management of liver disease. World J Gastrointest Pharmacol Ther 7:
on hepatic histological activity and fibrosis of nonalcoholic 91 – 106
steatohepatitis patients: a 1-year randomized control trial. Hepat Med Cheung KJ, Libbrecht L, Tilleman K, Deforce D, Colle I, Van Vlierberghe H
10: 23 – 31 (2010) Galectin-3-binding protein: a serological and histological
Altelaar AF, Munoz J, Heck AJ (2013) Next-generation proteomics: towards an assessment in accordance with hepatitis C-related liver fibrosis. Eur J
integrative view of proteome dynamics. Nat Rev Genet 14: 35 – 48 Gastroenterol Hepatol 22: 1066 – 1073
Angulo P, Hui JM, Marchesini G, Bugianesi E, George J, Farrell GC, Enders F, Chrostek L, Supronowicz L, Panasiuk A, Cylwik B, Gruszewska E, Flisiak R
Saksena S, Burt AD, Bida JP, Lindor K, Sanderson SO, Lenzi M, Adams LA, (2014) The effect of the severity of liver cirrhosis on the level of lipids and
Kench J, Therneau TM, Day CP (2007) The NAFLD fibrosis score: a lipoproteins. Clin Exp Med 14: 417 – 421
noninvasive system that identifies liver fibrosis in patients with NAFLD. Cox J, Mann M (2008) MaxQuant enables high peptide identification rates,
Hepatology 45: 846 – 854 individualized p.p.b.-range mass accuracies and proteome-wide protein
Anoop S, Misra A, Bhatt SP, Gulati S, Pandey RM, Mahajan H (2017) High quantification. Nat Biotechnol 26: 1367 – 1372
circulating plasma dipeptidyl peptidase- 4 levels in non-obese Asian Indians Cox J, Neuhauser N, Michalski A, Scheltema RA, Olsen JV, Mann M (2011)
with type 2 diabetes correlate with fasting insulin and LDL-C levels, triceps Andromeda: a peptide search engine integrated into the MaxQuant
skinfolds, total intra-abdominal adipose tissue volume and presence of environment. J Proteome Res 10: 1794 – 1805
diabetes: a case-control study. BMJ Open Diabetes Res Care 5: e000393 Cox J, Hein MY, Luber CA, Paron I, Nagaraj N, Mann M (2014) Accurate
Anthony PP, Ishak KG, Nayak NC, Poulsen HE, Scheuer PJ, Sobin LH (1977) proteome-wide label-free quantification by delayed normalization and
The morphology of cirrhosis: definition, nomenclature, and classification. maximal peptide ratio extraction, termed MaxLFQ. Mol Cell Proteomics 13:
Bull World Health Organ 55: 521 – 540 2513 – 2526

14 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors


Lili Niu et al A protein marker panel for NAFLD Molecular Systems Biology

Deacon CF (2018) A review of dipeptidyl peptidase-4 inhibitors. Hot topics Imbert-Bismut F, Ratziu V, Pieroni L, Charlotte F, Benhamou Y, Poynard T
from randomized controlled trials. Diabetes Obes Metab 20(Suppl 1): (2001) Biochemical markers of liver fibrosis in patients with hepatitis C
34 – 46. virus infection: a prospective study. Lancet 357: 1069 – 1075
Dieplinger H, Dieplinger B (2015) Afamin — A pleiotropic glycoprotein Inamine T, Schnabl B (2018) Immunoglobulin A and liver diseases. J
involved in various disease states. Clin Chim Acta 446: 105 – 110 Gastroenterol 53: 691 – 700
DiNicolantonio JJ, Subramonian AM, O’Keefe JH (2017) Added fructose as a Itou M, Kawaguchi T, Taniguchi E, Sata M (2013) Dipeptidyl peptidase-4: a
principal driver of non-alcoholic fatty liver disease: a public health crisis. key player in chronic liver disease. World J Gastroenterol 19: 2298 – 2306
Open Heart 4: e000631 Jall S, Sachs S, Clemmensen C, Finan B, Neff F, DiMarchi RD, Tschop MH,
Dyson JK, Anstee QM, McPherson S (2014) Non-alcoholic fatty liver disease: a Muller TD, Hofmann SM (2017) Monomeric GLP-1/GIP/glucagon
practical approach to diagnosis and staging. Frontline Gastroenterol 5: triagonism corrects obesity, hepatosteatosis, and dyslipidemia in female
211 – 218 mice. Mol Metab 6: 440 – 446
Fan JG, Kim SU, Wong VW (2017) New trends on obesity and NAFLD in Asia. J Junker AE, Gluud LL, Holst JJ, Knop FK, Vilsboll T (2015) Influence of
Hepatol 67: 862 – 873 gastrointestinal factors on glucose metabolism in patients with cirrhosis. J
Finan B, Ma T, Ottaway N, Muller TD, Habegger KM, Heppner KM, Kirchner H, Gastroenterol Hepatol 30: 1522 – 1528
Holland J, Hembree J, Raver C, Lockie SH, Smiley DL, Gelfanov V, Yang B, Junker AE, Gluud L, Holst JJ, Knop FK, Vilsboll T (2016) Diabetic and
Hofmann S, Bruemmer D, Drucker DJ, Pfluger PT, Perez-Tilve D, Gidda J nondiabetic patients with nonalcoholic fatty liver disease have an
et al (2013) Unimolecular dual incretins maximize metabolic benefits in impaired incretin effect and fasting hyperglucagonaemia. J Intern Med 279:
rodents, monkeys, and humans. Sci Transl Med 5: 209ra151 485 – 493
Frias JP, Bastyr EJ 3rd, Vignati L, Tschop MH, Schmitt C, Owen K, Christensen Kampf C, Mardinoglu A, Fagerberg L, Hallstrom BM, Edlund K, Lundberg E, Ponten
RH, DiMarchi RD (2017) The sustained effects of a dual GIP/GLP-1 receptor F, Nielsen J, Uhlen M (2014) The human liver-specific proteome defined by
agonist, NNC0090-2746, in patients with type 2 diabetes. Cell Metab 26: transcriptomics and antibody-based profiling. FASEB J 28: 2901 – 2914
343 – 352.e342 Kollerits B, Lamina C, Huth C, Marques-Vidal P, Kiechl S, Seppala I, Cooper J,
Friedman SL, Neuschwander-Tetri BA, Rinella M, Sanyal AJ (2018) Hunt SC, Meisinger C, Herder C, Kedenko L, Willeit J, Thorand B,
Mechanisms of NAFLD development and therapeutic strategies. Nat Med Dahnhardt D, Stockl D, Willeit K, Roden M, Rathmann W, Paulweber B,
24: 908 – 922 Peters A et al (2017) Plasma concentrations of afamin are associated with
Fukuhara T, Hyogo H, Ochi H, Fujino H, Kan H, Naeshiro N, Honda Y, Miyaki prevalent and incident type 2 diabetes: a pooled analysis in more than
D, Kawaoka T, Tsuge M, Hiramatsu A, Imamura M, Kawakami Y, Aikata H, 20,000 individuals. Diabetes Care 40: 1386 – 1393
Chayama K (2014) Efficacy and safety of sitagliptin for the treatment of Kotronen A, Yki-Jarvinen H (2008) Fatty liver: a novel component of the
nonalcoholic fatty liver disease with type 2 diabetes mellitus. metabolic syndrome. Arterioscler Thromb Vasc Biol 28: 27 – 38
Hepatogastroenterology 61: 323 – 328 Kotronen A, Peltonen M, Hakkarainen A, Sevastianova K, Bergholm R,
Geyer PE, Kulak Nils A, Pichler G, Holdt Lesca M, Teupser D, Mann M (2016a) Johansson LM, Lundbom N, Rissanen A, Ridderstrale M, Groop L, Orho-
Plasma proteome profiling to assess human health and disease. Cell Syst Melander M, Yki-Jarvinen H (2009) Prediction of non-alcoholic fatty liver
2: 185 – 195 disease and liver fat using metabolic and genetic factors. Gastroenterology
Geyer PE, Wewer Albrechtsen NJ, Tyanova S, Grassl N, Iepsen EW, Lundgren J, 137: 865 – 872
Madsbad S, Holst JJ, Torekov SS, Mann M (2016b) Proteomics reveals the Kulak NA, Pichler G, Paron I, Nagaraj N, Mann M (2014) Minimal,
effects of sustained weight loss on the human plasma proteome. Mol Syst encapsulated proteomic-sample processing applied to copy-number
Biol 12: 901 estimation in eukaryotic cells. Nat Methods 11: 319 – 324
Geyer PE, Holdt Lesca M, Teupser D, Mann M (2017) Revisiting biomarker Kulak NA, Geyer PE, Mann M (2017) Loss-less nano-fractionator for high
discovery by plasma proteomics. Mol Syst Biol 13: 942 sensitivity, high coverage proteomics. Mol Cell Proteomics 16: 694 – 705
Geyer PE, Voytik E, Treit PV, Doll S, Kleinhempel A, Niu L, Mueller JB, Bader J, Leite NC, Salles GF, Araujo AL, Villela-Nogueira CA, Cardoso CR (2009)
Teupser D, Holdt LM, Mann M (2018) Plasma proteome profiling to detect Prevalence and associated factors of non-alcoholic fatty liver disease in
and avoid sample-related biases in biomarker studies. bioRxiv https://doi. patients with type-2 diabetes mellitus. Liver Int 29: 113 – 119
org/10.1101/478305 [PREPRINT] Li LC, Li J, Gao J (2014) Functions of galectin-3 and its role in fibrotic
Ghorpade DS, Ozcan L, Zheng Z, Nicoloro SM, Shen Y, Chen E, Bluher M, diseases. J Pharmacol Exp Ther 351: 336 – 343
Czech MP, Tabas I (2018) Hepatocyte-secreted DPP4 in obesity promotes Liu S-L, Cheng C-C, Chang C-C, Mai F-D, Wang C-C, Lee S-C, Ho A-S, Chen L-
adipose inflammation and insulin resistance. Nature 555: 673 – 677 Y, Chang J (2011) Discovery of serum biomarkers of alcoholic fatty liver in
Glen J, Floros L, Day C, Pryke R (2016) Non-alcoholic fatty liver disease a rodent model: C-reactive protein. J Biomed Sci 18: 52
(NAFLD): summary of NICE guidance. BMJ 354: i4428 Meier F, Geyer PE, Virreira Winter S, Cox J, Mann M (2018) BoxCar acquisition
Harrison SA, Marri SR, Chalasani N, Kohli R, Aronstein W, Thompson GA, Irish method enables single-shot proteomics at a depth of 10,000 proteins in
W, Miles MV, Xanthakos SA, Lawitz E, Noureddin M, Schiano TD, Siddiqui 100 minutes. Nat Methods 15: 440 – 448
M, Sanyal A, Neuschwander-Tetri BA, Traber PG (2016) Randomised Mikolasevic I, Orlic L, Franjic N, Hauser G, Stimac D, Milic S (2016) Transient
clinical study: GR-MD-02, a galectin-3 inhibitor, vs. placebo in patients elastography (FibroScan(®)) with controlled attenuation parameter in the
having non-alcoholic steatohepatitis with advanced fibrosis. Aliment assessment of liver steatosis and fibrosis in patients with nonalcoholic
Pharmacol Ther 44: 1183 – 1198 fatty liver disease - Where do we stand? World J Gastroenterol 22:
Henderson NC, Mackinnon AC, Farnworth SL, Poirier F, Russo FP, Iredale 7236 – 7251
JP, Haslett C, Simpson KJ, Sethi T (2006) Galectin-3 regulates Miyazaki M, Kato M, Tanaka K, Tanaka M, Kohjima M, Nakamura K, Enjoji M,
myofibroblast activation and hepatic fibrosis. Proc Natl Acad Sci USA Nakamuta M, Kotoh K, Takayanagi R (2012) Increased hepatic expression of
103: 5060 – 5065 dipeptidyl peptidase-4 in non-alcoholic fatty liver disease and its

ª 2019 The Authors Molecular Systems Biology 15: e8793 | 2019 15 of 16


Molecular Systems Biology A protein marker panel for NAFLD Lili Niu et al

association with insulin resistance and glucose metabolism. Mol Med Rep 5: circulating levels were higher in nonalcoholic fatty liver disease vs.
729 – 733 histologically normal liver from morbidly obese women. Obesity 21: 170 – 177
Myers RP, Pomier-Layrargues G, Kirsch R, Pollett A, Beaton M, Levstik M, Tyanova S, Temu T, Sinitcyn P, Carlson A, Hein MY, Geiger T, Mann M, Cox J
Duarte-Rojo A, Wong D, Crotty P, Elkashab M (2012) Discordance in (2016) The Perseus computational platform for comprehensive analysis of
fibrosis staging between liver biopsy and transient elastography using the (prote)omics data. Nat Methods 13: 731 – 740
FibroScan XL probe. J Hepatol 56: 564 – 570 Vizcaino JA, Csordas A, del-Toro N, Dianes JA, Griss J, Lavidas I, Mayer G,
Nagaraj N, Kulak NA, Cox J, Neuhauser N, Mayr K, Hoerning O, Vorm O, Perez-Riverol Y, Reisinger F, Ternent T, Xu QW, Wang R, Hermjakob H
Mann M (2012) System-wide perturbation analysis with nearly complete (2016) 2016 update of the PRIDE database and its related tools. Nucleic
coverage of the yeast proteome by single-shot ultra HPLC runs on a Acids Res 44: D447 – D456
bench top Orbitrap. Mol Cell Proteomics 11: M111.013722 Wewer Albrechtsen NJ, Geyer PE, Doll S, Treit PV, Bojsen-Møller KN,
Naveau S, Poynard T, Benattar C, Bedossa P, Chaput JC (1994) Alpha-2- Martinussen C, Jørgensen NB, Torekov SS, Meier F, Niu L, Santos A,
macroglobulin and hepatic fibrosis. Diagnostic interest. Dig Dis Sci 39: 2426 – 2432 Keilhauer EC, Holst JJ, Madsbad S, Mann M (2018) Plasma proteome
Neuman MG, Cohen LB, Nanau RM (2014) Biomarkers in nonalcoholic fatty profiling reveals dynamics of inflammatory and lipid homeostasis markers
liver disease. Canadian J Gastroenterol Hepatol 28: 607 – 618 after Roux-en-Y gastric bypass surgery. Cell Syst 7: 601 – 612.e3
Palta S, Saroa R, Palta A (2014) Overview of the coagulation system. Indian J Williams KH, Shackel NA, Gorrell MD, McLennan SV, Twigg SM (2013)
Anaesth 58: 515 – 523 Diabetes and nonalcoholic Fatty liver disease: a pathogenic duo. Endocr
Parkes J, Guha IN, Roderick P, Harris S, Cross R, Manos MM, Irving W, Rev 34: 84 – 129
Zaitoun A, Wheatley M, Ryder S, Rosenberg W (2011) Enhanced liver Wood GC, Chu X, Argyropoulos G, Benotti P, Rolston D, Mirshahi T, Petrick A,
fibrosis (ELF) test accurately identifies liver fibrosis in patients with Gabrielson J, Carey DJ, DiStefano JK, Still CD, Gerhard GS (2017) A
chronic hepatitis C. J Viral Hepat 18: 23 – 31 multi-component classifier for nonalcoholic fatty liver disease (NAFLD)
Poynard T, Ratziu V, Naveau S, Thabut D, Charlotte F, Messous D, Capron D, based on genomic, proteomic, and phenomic data domains. Sci Rep 7: 43238
Abella A, Massard J, Ngo Y, Munteanu M, Mercadier A, Manns M, Albrecht Yang ZJ, Costa KA, Novelli EM, Smith RE (2014) Venous thromboembolism in
J (2005) The diagnostic value of biomarkers (SteatoTest) for the prediction cirrhosis. Clin Appl Thromb Hemost 20: 169 – 178
of liver steatosis. Comp Hepatol 4: 10 Younossi ZM, Blissett D, Blissett R, Henry L, Stepanova M, Younossi Y, Racila
Richards AL, Merrill AE, Coon JJ (2015) Proteome sequencing goes deep. Curr A, Hunt S, Beckerman R (2016) The economic and clinical burden of
Opin Chem Biol 24: 11 – 17 nonalcoholic fatty liver disease in the United States and Europe.
Singh S, Allen AM, Wang Z, Prokop LJ, Murad MH, Loomba R (2015) Fibrosis Hepatology 64: 1577 – 1586
progression in nonalcoholic fatty liver versus nonalcoholic steatohepatitis: Zhou Z, Chen H, Ju H, Sun M (2017) Circulating retinol binding protein 4
a systematic review and meta-analysis of paired-biopsy studies. Clin levels in nonalcoholic fatty liver disease: a systematic review and meta-
Gastroenterol Hepatol 13: 643 – 654.e649 analysis. Lipids Health Dis 16: 180
Sterling RK, Lissen E, Clumeck N, Sola R, Correa MC, Montaner J, M SS,
Torriani FJ, Dieterich DT, Thomas DL, Messinger D, Nelson M (2006) License: This is an open access article under the
Development of a simple noninvasive index to predict significant fibrosis terms of the Creative Commons Attribution 4.0
in patients with HIV/HCV coinfection. Hepatology 43: 1317 – 1325 License, which permits use, distribution and reproduc-
Terra X, Auguet T, Broch M, Sabench F, Hernandez M, Pastor RM, Quesada IM, tion in any medium, provided the original work is
Luna A, Aguilar C, del Castillo D, Richart C (2013) Retinol binding protein-4 properly cited.

16 of 16 Molecular Systems Biology 15: e8793 | 2019 ª 2019 The Authors

You might also like