Sequential Serum Metabolomic Profiling after Radiofrequency

Subscriber access provided by University Libraries, University of Memphis

Article

Sequential serum metabolomic profiling after radiofrequency ablation of hepatocellular carcinoma reveals different response patterns according to aetiology Corentine Goossens, Pierre Nahon, Laurence Le Moyec, Mohamed N. Triba, Nadia Bouchemal, Roland Amathieu, Nathalie Ganne-Carrié, Marianne Ziol, JeanClaude Trinchet, Nicolas Sellier, Abou Diallo, Olivier Seror, and Philippe Savarin J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.5b01032 • Publication Date (Web): 25 Mar 2016 Downloaded from http://pubs.acs.org on March 28, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 27

Journal of Proteome Research

1 2 3 4 5 6 7 8 9 10 1 12 13 14 15 16 17 18 19 20 21 2 23 24 25 26 27 28 29 30 31 32 3 34 35 36 37 38 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58 59 60 ACS Paragon Plus Environment


Page 2 of 27

1 1 3

2

Title. Sequential serum metabolomic profiling after radiofrequency ablation of 5

4

hepatocellular carcinoma reveals different response patterns according to aetiology 6 7 8

Authors. Corentine Goossens1, Pierre Nahon2,3, Laurence Le Moyec4, Mohamed Nawfal 9 1

10

Triba1, Nadia Bouchemal1, Roland Amathieu1,5, Nathalie Ganne-Carrié2,3, Marianne Ziol6,7, 12 13

Jean-Claude Trinchet2,3,7,+, Nicolas Sellier8, Abou Diallo9, Olivier Seror3,8, Philippe Savarin1 * 15

14

17

16

Departments and institutions: 18 19 1 Université

2

21

20

2 Groupe

23

Paris 13, Sorbonne Paris Cité, CSPBAT, UMR 7244, CNRS, Bobigny, France

hospitalier Paris Seine-Saint-Denis, pôle d’activités cancérologiques spécialisées,

25

24

APHP, Hôpital Jean Verdier, Bondy et Université Paris 13, Sorbonne Paris Cité, UFR SMBH, 26 27

Bobigny, France 28 30

29 3 INSERM

31

U1162, Génomique Fonctionnelle des Tumeurs solides, Université Paris 5, Paris,

3

32

France 35

34

4

36

Université d’Evry Val d’Essonne, UBIAE, INSERM U902, Evry, France

38

37 5 Service

39

d’Anesthésie-Réanimation, GHU PSSD, Hôpital Jean Verdier, Bondy et Université

41

40

Paris 13, Sorbonne Paris Cité, UFR SMBH, Bobigny, France 42 43 6 APHP,

4

service d’Anatomie Pathologique, Hôpital Jean Verdier, Bondy et Université Paris 13,

46

45

Sorbonne Paris Cité, UFR SMBH, Bobigny, France 47 49

48 7 BB-0033-00027.

51

50

Centre de Ressources Biologiques Maladies du foie, Groupe hospitalier

Paris-Seine-Saint-Denis, Bondy, France 52 54

53 8 APHP,

5

service de Radiologie, Hôpital Jean Verdier, Bondy, France

57

56 9 Service

58

d’Information Médicale, GHU PSSD, Hôpital Jean Verdier, Bondy, France

60

59

+ This author deceased during the preparation of the manuscript

ACS Paragon Plus Environment

Page 3 of 27


2 1 3

2

Running title. Aetiology influence on the metabolic profiling of hepatocellular carcinoma 4 6

5

Financial support: Supported by grants from SESAME Ile de France (SESAME NMR 7 8

Biomolecules) and from Association Française pour l’Etude du Foie (AFEF). 9 10 12

1

Corresponding Author: Philippe Savarin, 74, rue Marcel Cachin, 93017 Bobigny, France, +33 14

13

(0)148387323, [email protected] 15 16 17

Word count : 3729 18 19 21

20

Number of figures and tables : 8 2 23 24 26

25

The authors disclose no potential conflicts of interest. 27 28 29 30 31 32 3 34 35 36 37 38 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58 59 60



Page 4 of 27

3 1 3

2

Abstract 4 5 7

6

Radiofrequency ablation (RFA) is commonly performed as a curative approach in patients 8 9

with hepatocellular carcinoma (HCC). However, the risk of tumor recurrence is difficult to 10 12

1

predict due to a lack of reliable clinical and biological markers and identification of new 13 14

biomarkers poses a major challenge for improving prognoses. Metabolomics is a promising 17

16

15

technique that may lead to the identification and characterization of new disease 18 20

19

fingerprints. The objective of the present study was to explore, preoperatively and at various 2

21

time points post-RFA, the metabolic profile of serum samples from HCC patients in order to 23 25

24

identify factors associated with treatment response and recurrence. 27

26

Sequential sera obtained before and after RFA procedures for 120 patients with HCC due to 28 30

29

cirrhosis were investigated using Nuclear Magnetic Resonance (NMR) metabolomics. A 32

31

multilevel orthogonal projection to latent structure (OPLS) analysis was used to discriminate 3 35

34

intra-individual metabolic changes in response to RFA treatment. 37

36

Recurrence-free survival differed depending on the underlying cause of cirrhosis. The 38 40

39

statistical model showed significant differences depending on whether the liver disease had 42

41

a viral or non-viral aetiology before RFA intervention (explained variance of R2Y=0.89 and 43 45

4

predictability of Q2Y=0.34). These profiles were also associated with specific and distinct 47

46

metabolic responses after RFA. 48 49 50 51 53

52

Keywords: Liver cancer; RFA; NMR fingerprinting; metabolomics; OPLS 54 5 56 57 58 59 60


Page 5 of 27


4 1 3

2

Introduction 4 5 6 7 9

8

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer and 1

10

is mainly due to cirrhosis in Western countries1. With the advent of diagnostic imaging 12 14

13

techniques and medical surveillance of patients at risk, diagnosis at an earlier stage can be 16

15

achieved and curative procedures can be implemented2. Among the latter, radiofrequency 17 19

18

ablation (RFA) has gained popularity and has been shown to be highly effective in the 21

20

management of primary hepatic tumors. However, despite current progress in the use of 2 24

23

this technique, the rate of overall recurrence is about 70% at 5 years and this remains one of 26

25

the most challenging aspects of this field3. Prognostication for HCC after curative treatment 27 29

28

is difficult, in part due to the lack of useful biomarkers that would allow for the selection of 31

30

patients at higher risk of tumor recurrence or enable accurate assessment of treatment 32 34

3

response4. 36

35

Metabolomics is recognized as a promising technique in the field of systems biology 37 39

38

that allows the evaluation of global metabolic changes5. Proton nuclear magnetic resonance 41

40

(1H NMR) spectroscopy-based metabolomics can be used to identify and quantify 42 4

43

metabolites within a biological fluid in a single experiment, thus defining a metabolic profile, 46

45

also called a metabolomic fingerprint6. NMR-based metabolomics has been recently used in 47 49

48

the field of liver disease for the detection of early biomarkers and altered metabolic 51

50

pathways7, 8. NMR-based metabolomic studies of HCC have also been performed in sera, 52 54

53

urine, and tissues to elucidate significant changes in tumor progression9-11. Changes in serum 5 56

metabolomic profiles in patients with advanced HCC reflect the distinct activation or 57

60

59

58

impairment of specific biological pathways, in particular, energetic metabolism12. Even more interestingly, inter- and intra-individual analysis of serum changes during follow-up of



Page 6 of 27

5 1 3

2

treated patients may provide essential information related to therapeutic responses and/or 5

4

patterns of recurrence. Such improvements might lead to better stratification of patients 6 8

7

and provide the data needed for implementation of personalized treatment algorithms. 10

9

In the present study, serum samples from patients with stage 0 or A HCC (classified by 1 13

12

Barcelona Clinic Liver Cancer (BCLC) criteria) were investigated using NMR-metabolomics. 14 15

The objective was to explore, preoperatively and at various time points post-RFA, the 16 18

17

metabolic profiles of patients with HCC to identify factors associated with disease recurrence 19 20

and treatment response. 21 2 23 24 25

Patients and Methods 27

26 28 29 31

30

Ethics Statement. The Institutional Review Board of Jean Verdier University Hospital 32 3

approved the protocol and the French Research Delegation Office accepted the creation of a 34 36

35

dedicated bio-collection for the patients who have been included in the protocol. The CNIL 37 38

(National Informatics and Liberty Commission) also approved the creation of both bio39 41

40

collection and database. Written informed consent was obtained directly from each patient 42 43

at inclusion. 4 46

45

Patient selection, Follow-up, and Collection of Serum Samples. Between January 47 48

2002 and December 2012, 711 cirrhotic patients referred to our institution for the 49 51

50

management of HCC by percutaneous ablation were considered. Among them, 349 52 53

underwent monopolar or multipolar RFA and had frozen serum samples available before and 54 56

5

after the procedure. Serum samples were collected, stored and provided by the “centre de 57 58

ressources biologiques maladie du foie”, Bondy, France (BB-0033-00027). The following 59 60

inclusion criteria were considered: 1) biopsy-proven cirrhosis, 2) tumor stage 0 (one nodule


Page 7 of 27


6 1 3

2

smaller than 2 cm) or A (one nodule smaller than 5 cm or a maximum of three nodules 5

4

smaller than 3cm), classified according to BCLC criteria, 3) availability for regular follow-up. 6 8

7

HCC was diagnosed according to BCLC criteria: histological evidence or demonstration of a 10

9

focal lesion more than 1 cm in size with arterial hyper-vascularization and wash-out at the 1 13

12

portal phase by imaging techniques. The decision to treat patients with RFA was made by a 14 15

multidisciplinary 16

team

including

hepatologists,

oncologists,

liver

surgeons,

and

18

17

interventional radiologists according to international recommendations13. Clinical, biological, 19 20

and radiological data for patients were recorded on the day of the first blood collection that 21 23

2

was considered to be the inclusion date. The cause of the underlying liver disease was 24 25

recorded, patients were considered to be in the “viral-HCC” group if they had hepatitis B 26 28

27

virus (HBV) or hepatitis C virus (HCV)-related cirrhosis, and in the “non-viral” group if they 29 30

had alcohol- and/or dysmetabolic-related cirrhosis. 31 3

32

Response was assessed at one month after RFA, every three months for the next two years, 34 35

and then every six months after that. Assessments included a three-phase contrast 36 38

37

computed tomography (CT) scan or magnetic resonance imaging (MRI) examination, and 39 40

alpha-fetoprotein (AFP) serum level measurement. A tumor was considered as completely 41 43

42

ablated if no nodular or irregular enhancement of the liver was visible at the arterial phase. 4 45

Local or distant tumor recurrences were assessed. Local and overall tumor progression-free 46 48

47

survival was computed. 49 50

Serum samples were collected at different times before and after RFA intervention and were 51 53

52

immediately stored at -80°C until NMR spectroscopic analysis. Sequential blood samples 54 5

were selected: 1) within 14 days before RFA procedure (t0), 2) the day after the RFA 58

57

56

procedure (t1) and 3) at the time of treatment response evaluation, ranging from one to 59 60



Page 8 of 27

7 1 3

2

four months after RFA (t2). At this time, liver examination using abdominal CT scan showed 5

4

complete liver necrosis at the location of the tumor without other visible signs of nodules. 6 8

7

Statistical Analysis of Clinical Data. Patient status was specified at the date of the 10

9

most recent follow-up visit before September 2013. Endpoint of interest was recurrence1 13

12

free survival. Patients who underwent liver transplantation after RFA treatment were 14 15

censored from the study. Survival was estimated using the Kaplan-Meier method. Univariate 16 18

17

and multivariate Cox proportional hazards regressions were performed to determine the 19 20

significant clinical and biological parameters for prediction of tumor recurrence-free survival. 21 23

2

All statistical analyses were conducted using SAS software 9.4.1. 25

24

1H-NMR

26

Spectroscopy. For NMR analysis, serum samples were thawed at room

28

27

temperature and 450 μL aliquots were mixed with 50 μL of D 2O. The proton NMR spectra 30

29

were acquired at 20°C on a Bruker AVANCE III 500 spectrometer. For all samples, 1H NMR 31 3

32

spectra were recorded using two pulse sequences (noesygppr1D and a Carr-Purcell34 35

Meiboom-Gill (CPMG) sequence14). The signal was acquired on 64K data points for a 36 38

37

spectral window of 6000 Hz. For each sample, noesygppr1D experiments were collected with 39 40

64 transients and mixing time was 100 ms. The CPMG sequence is frequently used to 41 43

42

suppress broad macromolecular signals according to their short T2 relaxation times. The 4 45

number of transients was 128 and the applied spin-spin relaxation delay was 18 ms. Total 46 48

47

correlation spectroscopy (TOCSY) spectra were acquired (mixing time 80ms, 4K data points, 49 50

32 transients). J-resolved spectra were also acquired with 4K data points and 32 transients. 51 53

52

Resonance assignment was done according to the STOCY and the 2D experiments. Based on 54 5

these results, Chenomx® software and HMDB databank15 were used to confirm metabolite 58

57

56

assignment. Both types of 1D spectra were used to compute the models. All presented 59 60

models were obtained with CPMG spectra except for the one obtained with t0 samples


Page 9 of 27


8 1 3

2

discriminating viral from non-viral patient sera before RFA as a better model was obtain with 5

4

noesygppr1D spectra. 6 8

7

Data Processing. The spectra obtained were processed with NMRPipe software16. 10

9

The spectra were manually phased and a linear baseline correction was applied. The 1 13

12

chemical shifts were referenced using the -glucose signal at 5.23 ppm. The spectral region 14 15

between -1 and 10.5 ppm was divided into 11500 spectral regions of 0.001 ppm width, 16 18

17

called buckets. The spectral region containing residual water (4.5-5.1 ppm) was excluded 19 20

from the analysis. The spectra normalization algorithm employed was the quantile 21 23

2

normalization17. Prior to analysis, all remaining spectral buckets were mean-centered and 24 25

scaled (UV scaling)18. 26 28

27

Multivariate Data Analysis of NMR data. A PCA was first performed to detect any 29 30

group separation based on NMR signal variability. This unsupervised method also enabled 31 3

32

detection and exclusion of any outliers, defined as observations located outside the 95% 34 35

confidence region of the model. A supervised method, the orthogonal projection to latent 36 38

37

structure-discriminant analysis (OPLS-DA) was used. Multilevel OPLS-DA was also applied as 39 40

a paired analysis through the follow-up of one patient, as proposed by Westerhuis et al.19. 43

42

41

Briefly, the between subject variation which is described by the average of observations 4 46

45

from each subject can be calculated and compared to others to evaluate the inter-individual 48

47

variability. This variability can be separated from the within subject variation which is the 49 51

50

difference between the observations and which correspond to the intra-individual 53

52

variability. This pairing provides the advantage of particularly discriminating the metabolic 54 56

5

changes within individuals caused by the intervention and allows for evaluation of the large 58

57

variability between human subjects. For each model, the number of considered patients was 59 60

dependent on the availability of samples t0, t1, or t2 for each patient. PCA, OPLS-DA, and



Page 10 of 27

9 1 3

2

Multilevel OPLS-DA analyses were performed using Simca-P12 (Umetrics, Umea) and in5

4

house Matlab 2012b (Mathworks, Natick, Massachusetts, USA) code based on the Trygg and 6 8

7

Wold method20. The quality parameters of the models, such as the explained variance (R2Y) 10

9

and the predictability of the model (Q2Y) were calculated. Q2Y was calculated by a 7-fold 1 13

12

cross validation procedure and confirmed by exploring the impact of permutations in the 15

14

dataset rows21. The discriminatory power of the model was evaluated by calculating the area 16 18

17

under the receiver operating curve during the cross validation (CV-AUROC). Models were 19 20

validated by permutation tests. The results are presented as a score-plot and a loading plot. 21 23

2

Each point in the score-plot represents the projection of an NMR spectrum (and thus a 24 25

sample from one patient) on the predictive (horizontal axis) and the first orthogonal 26 28

27

component of the model (vertical axis). The loading plot represents the covariance between 29 30

the Y-response matrix and the signal intensity of the various spectral domains. Colors were 31 3

32

also used in the loading plot depending on the correlation between the corresponding 34 35

bucket intensity and Y variable. 36 37 38 39 40

Results 42

41 43 4 46

45

Baseline characteristics of patients. To investigate the metabolic changes in serum 47 49

48

between baseline and post-RFA follow-up, 120 patients fulfilling inclusion criteria were 51

50

selected for this study. The characteristics of patients are shown in table 1 for the whole 52 54

53

cohort and stratified according to the cause of underlying cirrhosis (viral and non-viral). 56

5

Among viral-HCC patients, 11/59 (18.6%) had negative viral replication at time of HCC 57

60

59

58

diagnosis following or during the course on an antiviral treatment. There were significant differences in baseline parameters between the two subgroups for cholesterol and


Page 11 of 27


10 1 3

2

transaminases levels (aspartate aminotransferase (AST) and alanine aminotransferase (ALT)). 5

4

Sequential blood samples pre- and post-RFA were collected on the whole cohort and 6 8

7

included a set of 273 sera. The flow chart of patients and available serum samples is 10

9

presented in figure 1. The data were first analyzed by PCA, leading to the exclusion of 4 1 13

12

outliers. The PCA score is shown in supplementary Figure S1. It can be seen that the two first 15

14

principal components are not correlated to the time of sera collection. A representative 1H 16 18

17

NMR spectrum (CPMG) from the serum of one HCC patient at (a) t2 (from one to four 19 20

months post-ablation time), (b) t1 (one day post-ablation time) and (c) t0 (pre-ablation 21 23

2

period) is presented in figure 2. 24 25

HCC recurrence-free survival: univariate analyses. After a median follow-up time of 26 28

27

3.2 years, 61 (50.8%) patients experienced HCC recurrence (local=8, distant=53, figure 3A). 29 30

When stratified according to the cause of cirrhosis (figure 3B), patients with viral-HCC had a 31 3

32

higher probability of HCC recurrence (HR=1.9, Log-Rank=0.02, supplementary Table S1). 34 35

According to multivariate Cox model, this feature remained an independent predictor of HCC 36 38

37

recurrence when considering the whole population, as did higher gamma-glutamyl 39 40

transferase (GGT) activity (HR=2.6, P=0.003). 41 43

42

Metabolic profile comparisons related to HCC aetiology. In order to provide a more 4 45

in-depth comprehension analysis of the observed difference in the recurrence-free survival 46 48

47

between the two cohorts, metabolomic profiles of non-viral and viral sera were compared. 49 50

An OPLS-DA model was performed on the t0 sera of the whole cohort (118 t0 samples 51 53

52

available). The score plot of the 118 spectra (figure 4A) revealed a clear separation between 54 5

non-viral and viral HCC patients (Q2Y=0.34). Some metabolite variations that differed 58

57

56

between the two HCC aetiology subgroups could be identified in the serum spectra 59 60

according to the loading plot (figure 4B). This model specifically reflects a different metabolic



Page 12 of 27

11 1 3

2

profile in the serum of non-viral versus viral-related HCC patients. It must be emphasized 5

4

that no significant discriminant models could be realized between VHB- and VHC-related 6 8

7

HCC patients. 10

9

Metabolic profiles resulting from the RFA intervention. To further explore and 1 13

12

compare the metabolic changes that occurred as a result of RFA treatment for these two 14 15

cohorts, metabolites in sera collected before and after RFA intervention were studied. The 16 18

17

metabolic signature of the intervention was determined by comparison of the two OPLS-DA 19 20

models for the two subgroups: one comparing serum at t0 and at t1, and the other 21 23

2

comparing serum at t2 and t1. The comparison of these two OPLS-models allows to 24 25

accurately assess the proper effects of the RFA intervention. Considering the high levels of 26 28

27

inter-individual variability observed in the cohort, paired samples were used in multilevel30

29

OPLS-DA models(19). Consequently, four OPLS-DA models were obtained and their loading 31 3

32

plots are presented in figure 5 (score plots are shown in supplementary Figure S2). The t0 vs. 34 35

t1 OPLS-DA models in figure 5A and 5B showed good predictability (Q2Y=0.38 for the 27 36 38

37

patients in the non-viral cohort, and 0.62 for the 34 patients in the viral cohort). For t2 vs. t1 39 40

models, the predictability was 0.4 for the non-viral cohort (figure 5C, 17 patients) and 0.58 41 43

42

for the viral cohort (figure 5D, 27 patients). The four loading plots show that the RFA 4 45

intervention is correlated with an increase of lactate (5), glutamine (9), and 346 48

47

phenylpropionate (22) and a decrease of isoleucine (2), phosphatidylcholine (PC), and 49 50

glycerophosphocholine (GPC) (17). The comparison of loading plots for both sub-groups 51 53

52

(comparing column 1 and column 2 in figure 5) allowed visualization of a different profile for 54 5

viral and non-viral patients and determination of some metabolites, which vary differently 58

57

56

post-RFA intervention depending on the aetiology of cirrhosis (lipids (1), aspartate (13), 59 60

choline (16), glucose (19)). However, the loading plot in figure 5D demonstrates different


Page 13 of 27


12 1 3

2

behavior for glucose and lipids when compared to the three other loading plots. These 5

4

differences seem to indicate that the t2 metabolic state in the viral cohort could be different 6 8

7

from the one observed at t0. In contrast, profiles for the non-viral cohort (loading plots 5A 10

9

and 5C), appear similar. All statically relevant metabolites (p value < 0.05) are reported in 1 13

12

Table 2 and p values are reported in supplementary Table S2. 14 15

Discrimination of t0 from t2 sera for viral-related HCC patients. Profiles of the pre16 18

17

RFA sera at t0 and the sera collected after one to four months post-ablation (t2) were 19 20

compared in order to evaluate the evolution of the metabolic pathways post RFA. 21 23

2

Interestingly, no significant OPLS-DA model discriminating t0 and t2 could be computed in 24 25

the non-viral cohort. However, for viral patients, the t0 and t2 serum spectra could be 26 28

27

discriminated with OPLS-DA analysis (42 patients). Figures 6A and 6B present the paired 29 30

samples score and loading plots. The model demonstrated good predictability of 0.37. Some 31 3

32

discriminating metabolites could be identified and are summarized in figure 6C. The t2 state 35

34

was mainly characterized by an increase of glucose (19), glycerol (21), -methylhistidine (15), 36 38

37

and a decrease of lipids (1), 3-hydroxybutyrate (4), and choline (16) when compared to the 39 40

t0 state. 43

42

41

Metabolomic prediction of HCC recurrence. As these patients were prospectively 4 46

45

followed-up, the recurrence of the tumor after t2 was tested as a driving factor for OPLS-DA. 48

47

Several models were tested according to the time or aetiology but none of them was 49 51

50

statistically relevant. 52 53 54 5 56 57 58 59 60



Page 14 of 27

13 1 3

2

Discussion 4 5 6 7 9

8

Metabolomics technology offers the ability to associate changes in serum metabolic 1

10

profiles with environmental, pathological, or treatment effects. Due to the fact that there is 12 14

13

a pressing need for prognostic biomarkers that can stratify patients according to their risk of 16

15

HCC recurrence, we used metabolomics technology to study potential differences in 17 19

18

metabolic profiles in the sera of HCC patients before and after RFA treatment. In the present 21

20

study, inter- and intra-individual analysis of serum changes during follow-up of RFA-treated 2 24

23

patients highlighted aetiological factors related to therapeutic response. Three strengths 26

25

characterized the present study: i) selection of an homogeneous cohort of 120 patients with 27 29

28

HCC developed in cirrhotic liver and classified according to the cause of liver disease, ii) 31

30

longitudinal assessment of the serum metabolomic changes at various pivotal points of 32 34

3

follow-up, iii) application of statistical analyses exploiting the paired data structure in cross36

35

over multivariate data(19). It was essential to explore and apply the OPLS derivative methods 37 39

38

such as paired data structures because of the large variability between individuals that could 41

40

potentially mask small and subtle effects linked to disease aetiology or treatment. 42 4

43

The present study shows that two-year recurrence-free survival for this prospective 46

45

cohort differed according to the cause of cirrhosis, a parameter that is usually not taken into 47 49

48

account in clinical practice. The effect of viral infection on clinico-pathological features and 51

50

long-term outcome in HCC-treated patients has been the subject of several studies22. Zhou 52 54

53

et al. reported that patients with HBV or HCV infection had a worse 5-year disease-free 5 56

survival when compared to non-viral HCC patients. They also noted no difference in the 557

60

59

58

year survival between the HCV and HBV-positive HCC patients23. In the present cohort, in both subgroups, recurrence was associated with a higher baseline level of serum GGT,


Page 15 of 27


14 1 3

2

independently of the cause of liver disease. As all patients had compensated cirrhosis with 5

4

small HCC, this biological feature might reflect, at least partly, the degree of insulin 6 8

7

resistance related to comorbidities known to be frequent in these populations. This simple 10

9

and non-specific biological feature suggests that the impairment of glucose and lipid 1 13

12

metabolism in liver cancer recurrence after curative treatment could be targeted by 14 15

adjuvant therapies. 16 18

17

Based on these simple clinical observations, serum metabolomic profiling of patients 19 20

was able to provide a more in-depth comprehensive analysis at the metabolic level. When 21 23

2

stratified according to the cause of cirrhosis, patients presented specific metabolomic 25

24

profiles at baseline, as already demonstrated by Qi et al.24. High throughput analyses based 26 28

27

on genomics and/or transcriptomics of liver tumors highlighted differences related to 30

29

aetiology25. Although metabolic pathways are greatly interlinked and interpretation of the 31 3

32

meaning of changes in a particular metabolite is complicated, it appears that the distinct 34 35

metabolites responsible for differences observed in this prospective non-viral- and viral-HCC 36 38

37

cohort revealed significant alterations in lipid metabolism (increase of lipids, N-acetyl 39 40

glycoprotein, choline, and glycerol, and a decrease in GPC in the viral cohort compared to 41 43

42

the non-viral cohort). This feature is consistent with previous metabolic studies that 4 45

demonstrated that HBV and HCV infection perturbed lipid classes such as free fatty acids, 46 48

47

phospholipids, and sphingolipids in cultured hepatoma cells26, 27. In these experiments, an 49 50

increase of choline and phosphatidylcholine was observed in HCV-infection. 51 53

52

The aetiological subgroups also exhibited differences in the evolution of their 54 5

metabolomic profiles post-RFA. First, comparison of sera from different time points allowed 58

57

56

identification of a specific metabolic signature associated with RFA treatment. The main 59 60

metabolites related to the RFA intervention were metabolites associated with energy



Page 16 of 27

15 1 3

2

metabolism alterations, such as lactate and acetate, but also with amino acid metabolism, 5

4

such as isoleucine and glutamine. Several hypothesis could explain this fingerprint: i) the 6 8

7

application of electrical current during RFA treatment causes burns in the liver and produces 10

9

coagulative necrosis which results in parenchymal and tumor cell death 28. ii) the 1 13

12

enhancement of consumption of branched amino acids (BCAA), such as isoleucine, may 14 15

characterize the inflammatory response in liver. Indeed, numerous studies have reported 16 18

17

that the lactate is a major biomarker for tissue hypoxia and necrosis29 and this fact could 19 20

correspond to the inflammatory reaction and could lead to further increases in BCAA 21 23

2

catabolism and glutamine production, consistent with the findings obtained in this study30-32. 24 25

However, through comparison of the profiles of RFA intervention effects, we noticed a 26 28

27

profile quite different for the viral cohort. The metabolic state at t2 appeared to differ from 29 30

the one at t0, contrary to what was observed in the non-viral cohort. Moreover, an OPLS-DA 31 3

32

model discriminating t0 and t2 could be only computed for the viral cohort. This model 34 35

supports once again the idea that there is an evolution in the metabolic state at t2 compared 36 38

37

to the one at t0 for the viral cohort unlike the non-viral cohort. Major changes between t0 39 40

and the t2 profiles involved metabolites implicated in the lipid and glucose metabolic 41 43

42

pathways. It has been reported by Bollard et al.33 that liver regeneration requires 4 45

phospholipids for cell membrane synthesis and very-low-density lipoprotein (VLDL) synthesis 46 48

47

requires the availability of phospholipids, particularly PC. A possible explanation could be 49 50

that the process of regeneration is underway and could affect lipid metabolism. 51 53

52

Modifications in glucose concentration post-RFA could be linked to hyperglycaemia after 54 5

liver resection, as reported by Dungan et al.34. Consequently, for viral-related HCC patients, 58

57

56

metabolic processes in the serum appear to follow different biological pathways than those 59 60

for the non-viral-related HCC patients post-RFA.


Page 17 of 27


16 1 3

2

Discrepancies in metabolomic pathways between non-viral- and viral-HCC were 5

4

furthermore characterized by different patterns of evolution after RFA procedure. 6 8

7

Recurrence-free survival was not identical in the two groups despite similar therapeutic 10

9

managements. This metabolomics investigation demonstrated that biochemical adaptations 1 13

12

to liver injury depend, at least in part, on the aetiology of the disease and this results in 14 15

divergent biochemical responses to RFA treatment. A more in-depth analysis through liver 16 18

17

tissue could provide more specific confirmation for our metabolomic findings. Moreover, to 19 20

improve prognosis concerning tumor recurrence, the study of sera or tissues of a larger 21 23

2

cohort of patients (only with non-viral or viral-HCC) has to be done. Finally, further 24 25

metabolomic investigations focused on HCC and response to curative treatment should be 26 28

27

undertaken in order to implement new paradigms for personalized therapeutic 29 30

management. 31 3

32

Supporting Information Available 34 36

35

Figure S1 : PCA Score plot of the 273 sera 37 39

38

Table S1 : Risk factors for HCC recurrence after RFA procedure 41

40

Figure S2 Multilevel OPLS-DA score plots of the within-subject variation 42 4

43

Table S2 : Discriminant metabolites observed in sera of non-viral and viral HCC patients. 45 46 47 48 50

49

References 51 52 53 54

60

59

58

57

56

5

1. Daoudaki, M.; Fouzas, I., Hepatocellular carcinoma. Wiener medizinische Wochenschrift 2014, 164, (21-22), 450-5. 2. Cartier, V.; Aube, C., Diagnosis of hepatocellular carcinoma. Diagnostic and interventional imaging 2014, 95, (7-8), 709-19.



Page 18 of 27

17 1

60

59

58

57

56

5

54

53

52

51

50

49

48

47

46

45

4

43

42

41

40

39

38

37

36

35

34

3

32

31

30

29

28

27

26

25

24

23

2

21

20

19

18

17

16

15

14

13

12

1

10

9

8

7

6

5

4

3

2

3. Brunello, F.; Cantamessa, A.; Gaia, S.; Carucci, P.; Rolle, E.; Castiglione, A.; Ciccone, G.; Rizzetto, M., Radiofrequency ablation: technical and clinical long-term outcomes for single hepatocellular carcinoma up to 30 mm. Eur J Gastroenterol Hepatol 2013, 25, (7), 842-9. 4. Breuhahn, K.; Gores, G.; Schirmacher, P., Strategies for hepatocellular carcinoma therapy and diagnostics: lessons learned from high throughput and profiling approaches. Hepatology 2011, 53, (6), 2112-21. 5. Nagrath, D.; Caneba, C.; Karedath, T.; Bellance, N., Metabolomics for mitochondrial and cancer studies. Biochim Biophys Acta 2011, 1807, (6), 650-63. 6. Dunn, W. B.; Bailey, N. J.; Johnson, H. E., Measuring the metabolome: current analytical technologies. Analyst 2005, 130, (5), 606-25. 7. Ranjan, P.; Gupta, A.; Kumar, S.; Gowda, G. A.; Ranjan, A.; Sonker, A. A.; Chandra, A.; Ramakant, Detection of new amino acid markers of liver trauma by proton nuclear magnetic resonance spectroscopy. Liver Int 2006, 26, (6), 703-7. 8. Amathieu, R.; Nahon, P.; Triba, M.; Bouchemal, N.; Trinchet, J. C.; Beaugrand, M.; Dhonneur, G.; Le Moyec, L., Metabolomic approach by 1H NMR spectroscopy of serum for the assessment of chronic liver failure in patients with cirrhosis. J Proteome Res 2011, 10, (7), 3239-45. 9. Liu, Y.; Hong, Z.; Tan, G.; Dong, X.; Yang, G.; Zhao, L.; Chen, X.; Zhu, Z.; Lou, Z.; Qian, B.; Zhang, G.; Chai, Y., NMR and LC/MS-based global metabolomics to identify serum biomarkers differentiating hepatocellular carcinoma from liver cirrhosis. International journal of cancer. Journal international du cancer 2014, 135, (3), 658-68. 10. Ladep, N. G.; Dona, A. C.; Lewis, M. R.; Crossey, M. M.; Lemoine, M.; Okeke, E.; Shimakawa, Y.; Duguru, M.; Njai, H. F.; Fye, H. K.; Taal, M.; Chetwood, J.; Kasstan, B.; Khan, S. A.; Garside, D. A.; Wijeyesekera, A.; Thillainayagam, A. V.; Banwat, E.; Thursz, M. R.; Nicholson, J. K.; Njie, R.; Holmes, E.; Taylor-Robinson, S. D., Discovery and validation of urinary metabotypes for the diagnosis of hepatocellular carcinoma in West Africans. Hepatology 2014, 60, (4), 1291-301. 11. Yang, Y.; Li, C.; Nie, X.; Feng, X.; Chen, W.; Yue, Y.; Tang, H.; Deng, F., Metabonomic studies of human hepatocellular carcinoma using high-resolution magic-angle spinning 1H NMR spectroscopy in conjunction with multivariate data analysis. Journal of proteome research 2007, 6, (7), 2605-14. 12. Nahon, P.; Amathieu, R.; Triba, M. N.; Bouchemal, N.; Nault, J. C.; Ziol, M.; Seror, O.; Dhonneur, G.; Trinchet, J. C.; Beaugrand, M.; Le Moyec, L., Identification of serum proton NMR metabolomic fingerprints associated with hepatocellular carcinoma in patients with alcoholic cirrhosis. Clin Cancer Res 2012, 18, (24), 6714-22. 13. Bruix, J.; Sherman, M., Management of hepatocellular carcinoma: an update. Hepatology 2011, 53, (3), 1020-2. 14. Foxall, P. J.; Spraul, M.; Farrant, R. D.; Lindon, L. C.; Neild, G. H.; Nicholson, J. K., 750 MHz 1HNMR spectroscopy of human blood plasma. J Pharm Biomed Anal 1993, 11, (4-5), 267-76. 15. Wishart, D. S.; Tzur, D.; Knox, C.; Eisner, R.; Guo, A. C.; Young, N.; Cheng, D.; Jewell, K.; Arndt, D.; Sawhney, S.; Fung, C.; Nikolai, L.; Lewis, M.; Coutouly, M. A.; Forsythe, I.; Tang, P.; Shrivastava, S.; Jeroncic, K.; Stothard, P.; Amegbey, G.; Block, D.; Hau, D. D.; Wagner, J.; Miniaci, J.; Clements, M.; Gebremedhin, M.; Guo, N.; Zhang, Y.; Duggan, G. E.; Macinnis, G. D.; Weljie, A. M.; Dowlatabadi, R.; Bamforth, F.; Clive, D.; Greiner, R.; Li, L.; Marrie, T.; Sykes, B. D.; Vogel, H. J.; Querengesser, L., HMDB: the Human Metabolome Database. Nucleic Acids Res 2007, 35, (Database issue), D521-6. 16. Delaglio, F.; Grzesiek, S.; Vuister, G. W.; Zhu, G.; Pfeifer, J.; Bax, A., NMRPipe: a multidimensional spectral processing system based on UNIX pipes. Journal of biomolecular NMR 1995, 6, (3), 277-93. 17. Bolstad, B. M.; Irizarry, R. A.; Astrand, M.; Speed, T. P., A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 2003, 19, (2), 185-93. 18. Van den Berg, R. A.; Hoefsloot, H. C.; Westerhuis, J. A.; Smilde, A. K.; van der Werf, M. J., Centering, scaling, and transformations: improving the biological information content of metabolomics data. BMC Genomics 2006, 7, 142.


Page 19 of 27


18 1 3

2

19. Westerhuis, J. A.; van Velzen, E. J.; Hoefsloot, H. C.; Smilde, A. K., Multivariate paired data analysis: multilevel PLSDA versus OPLSDA. Metabolomics 2009, 6, (1), 119-128. 20. Trygg, J.; Wold, S., Orthogonal projections to latent structures (O-PLS). J.Chemometrics 2002, 16, 119-128. 21. Triba, M. N.; Le Moyec, L.; Amathieu, R.; Goossens, C.; Bouchemal, N.; Nahon, P.; Rutledge, D. N.; Savarin, P., PLS/OPLS models in metabolomics: the impact of permutation of dataset rows on the K-fold cross-validation quality parameters. Molecular bioSystems 2015, 11, (1), 13-9. 22. Tanase, A. M.; Dumitrascu, T.; Dima, S.; Grigorie, R.; Marchio, A.; Pineau, P.; Popescu, I., Influence of hepatitis viruses on clinicopathological profiles and long-term outcome in patients undergoing surgery for hepatocellular carcinoma. Hepatobiliary Pancreat Dis Int 2014, 13, (2), 16272. 23. Zhou, Y.; Si, X.; Wu, L.; Su, X.; Li, B.; Zhang, Z., Influence of viral hepatitis status on prognosis in patients undergoing hepatic resection for hepatocellular carcinoma: a meta-analysis of observational studies. World J Surg Oncol 2011, 9, 108. 24. Qi, S.; Tu, Z.; Ouyang, X.; Wang, L.; Peng, W.; Cai, A.; Dai, Y., Comparison of the metabolic profiling of hepatitis B virus-infected cirrhosis and alcoholic cirrhosis patients by using (1) H NMRbased metabonomics. Hepatol Res 42, (7), 677-85. 25. Nault, J. C., Pathogenesis of hepatocellular carcinoma according to aetiology. Best Pract Res Clin Gastroenterol 2014, 28, (5), 937-947. 26. Negro, F.; Sanyal, A. J., Hepatitis C virus, steatosis and lipid abnormalities: clinical and pathogenic data. Liver Int 2009, 29 Suppl 2, 26-37. 27. Roe, B.; Kensicki, E.; Mohney, R.; Hall, W. W., Metabolomic profile of hepatitis C virusinfected hepatocytes. PLoS One 2011, 6, (8), e23641. 28. Higgins, H.; Berger, D. L., RFA for liver tumors: does it really work? Oncologist 2006, 11, (7), 801-8. 29. Bernal, W.; Donaldson, N.; Wyncoll, D.; Wendon, J., Blood lactate as an early predictor of outcome in paracetamol-induced acute liver failure: a cohort study. Lancet 2002, 359, (9306), 55863. 30. Holecek, M., Branched-chain amino acids and ammonia metabolism in liver disease: therapeutic implications. Nutrition 2013, 29, (10), 1186-91. 31. Saxena, V.; Gupta, A.; Nagana Gowda, G. A.; Saxena, R.; Yachha, S. K.; Khetrapal, C. L., 1H NMR spectroscopy for the prediction of therapeutic outcome in patients with fulminant hepatic failure. NMR Biomed 2006, 19, (5), 521-6. 32. Tripathi, P.; Bala, L.; Saxena, R.; Yachha, S. K.; Roy, R.; Khetrapal, C. L., 1H NMR spectroscopic study of blood serum for the assessment of liver function in liver transplant patients. J Gastrointestin Liver Dis 2009, 18, (3), 329-36. 33. Bollard, M. E.; Contel, N. R.; Ebbels, T. M.; Smith, L.; Beckonert, O.; Cantor, G. H.; LehmanMcKeeman, L.; Holmes, E. C.; Lindon, J. C.; Nicholson, J. K.; Keun, H. C., NMR-based metabolic profiling identifies biomarkers of liver regeneration following partial hepatectomy in the rat. J Proteome Res 2010, 9, (1), 59-69. 34. Dungan, K. M.; Braithwaite, S. S.; Preiser, J. C., Stress hyperglycaemia. Lancet 2009, 373, (9677), 1798-807. 51

50

49

48

47

46

45

4

43

42

41

40

39

38

37

36

35

34

3

32

31

30

29

28

27

26

25

24

23

2

21

20

19

18

17

16

15

14

13

12

1

10

9

8

7

6

5

4

52 53 54 5 56 57 58 59 60



Page 20 of 27

19 1 2 3 5

4

Table 1 6 7 8 9 10 1 12

Whole cohort

Non viral-HCC

Viral-HCC

(n=120)

(n=61, 50.9%)

(n=59, 49.1%)

Age (years) *

66 [59.5-73]

65 [60-73]

68 [57-75]

0.96

Male gender †

90 (75%)

46 (75.4%)

44 (74.6%)

0.92

Uninodular HCC †

87 (72.5%)

43 (70.5%)

44 (74.6%)

0.62

Biopsy-proven HCC †

71 (59.2%)

35 (57.4%)

36 (61%)

0.69

Largest HCC nodule (mm)*

24.5 [18-34]

20 [18-35]

25 [18-33]

0.64

6 [4-21]

6 [4-11]

6.5 [4-40]

0.42

1.06 [0.7-1.6]

1.08 [0.7-1.4]

0.9 [0.7-1.4]

0.67

4.3 [3.5-5]

4.4 [3.9-5.1]

3.6 [3.2-4.7]

0.005

HDL (g/L) *

1.3 [1.1-1.6]

1.4 [1.1-1.8]

1.3 [1.2-1.6]

0.4

LDL (g/L) *

2.3 [1.7-2.9]

2.3 [2-2.8]

1.8 [1.6-3.1]

0.47

5.9 [5.1-8]

6.2 [5.2-8.2]

5.7 [5-7.6]

0.20

CRP (g/L) *

4 [4-7]

5 [4-8]

4 [4-5]

0.51

AST (IU/L) *

48 [36-69]

45 [33-61]

56 [37-101]

0.006

ALT (IU/L) *

36 [26-54]

32 [25-41]

44 [30-92]

Sequential Serum Metabolomic Profiling after Radiofrequency

Recommend Documents