SEARCH

SEARCH BY CITATION

Keywords:

  • ‘e-negative’ hepatitis B;
  • genotype D;
  • HBV;
  • Hepatitis B e-antigen (HBeAg)-negative chronic infection;
  • inactive carrier;
  • liver cirrhosis;
  • mutations

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Transparency Declaration
  9. References

Clin Microbiol Infect 2012; 18: E412–E418

Abstract

Hepatitis B e-antigen (HBeAg)-negative chronic HBV infection is highly prevalent in several parts of the world, including India, with the clinical spectrum ranging from inactive carrier (IC) state to chronic ‘e-negative’ hepatitis B (CHB) and culminating in advanced liver disease such as cirrhosis (LC). The present study has for the first time investigated the natural diversity of HBV belonging to genotype D in treatment-naïve Indian patients representing the above phases of HBeAg-negative infection to identify candidate mutations associated with each disease state. Studies of full-length HBV/D sequences revealed that the progressive accumulation and persistence of mutations in basal core promoter, negative regulatory element, Pre-core region, the B- and T-cell epitopes of X protein as well as deletions in the PreS region contribute significantly to disease progression from IC through CHB to LC. In addition, the development of CHB was associated with a significant increase in viral variants characterized by mutations in enhancer II, preS1 promoter, T-cell epitope of core and B-cell epitope region of PreS1. While few of the mutations were previously reported in the context of HBV genotypes B and C, others had not been documented before. Our results thus highlight a distinct pattern of mutation in HBV/D that may help in predicting clinical outcomes of HBeAg-negative infection and have implications for better clinical management of the patients.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Transparency Declaration
  9. References

During the course of chronic HBV infection, HBeAg seroconversion to its antibody (anti-HBe) is known to lead to a favourable outcome in patients and coincides with normalization of liver biochemical tests, loss or decrease of serum HBV DNA, subsidence of hepatic inflammation and clinical remission [1,2]. Patients with such quiescent infection are commonly referred to as ‘inactive carriers’ (IC) [3]. However, active hepatitis may develop in up to one-third of IC without reversion of HBeAg in their serum [4,5]. This phase, called ‘HBeAg-negative chronic hepatitis’ (CHB), is characterized by absence of HBeAg, detectable levels of HBV DNA, raised serum alanine aminotransferase (ALT), and histological findings of continued necroinflammation of the liver [6]. The sequelae of e-negative CHB may include mild to moderate fibrosis, cirrhosis or even hepatocellular carcinoma (HCC) [7]. In recent years several reports have suggested an increasing incidence of HBeAg-negative CHB worldwide, including in India [6,8]. Different mutations arise in the HBV genome during chronic infection, which had been attributed to error-prone viral replication and directional selection by host immunity. There is growing evidence that viral polymorphism is one of the critical factors that influence disease progression and outcome. Of particular importance were the double mutations, A1762T and G1764A, in the basal core promoter (BCP) of HBV and deletions in the preS2 and X regions that were increasingly prevalent with advancement of disease severity [9,10]. However, these studies were mostly carried out on HBV belonging to genotypes B and C that prevail in eastern Asia and the findings may not be generalized to populations that are infected with other genotypes of HBV, given that different genotypes have unique patterns of mutations and distinct propensities for selection of mutants [11]. HBV belonging to genotype D has a wide geographical distribution and predominates in the Mediterranean area, Middle East and India [12]. The present study aims to carry out a genome-wide analysis of HBV of genotype D from treatment-naïve, chronic e-negative Indian patients presenting (i) as IC, (ii) with CHB and (iii) with liver cirrhosis (LC) to assess how the temporal acquisition and persistence of different mutations influence the evolving clinical landscape of e-negative HBV infection.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Transparency Declaration
  9. References

Collection of samples and storage

Sixty treatment-naïve HBV-infected patients, having HBsAg positivity (>6 months) but negative for HBeAg and representing different phases of chronic e-negative infection, were included in the study from the Hepatology Clinic of the Institute of Post Graduate Medical Education and Research, Kolkata, between August 2009 and July 2010. Patients with other viral infections and history of significant alcohol intake were excluded. The IC were characterized by consistently normal levels of ALT (2–45 IU/L) and AST (2–40 IU/L) during four consecutive follow-ups within a period of 1 year prior to enrollment and had low HBV DNA (<104 copies/mL). The e-negative CHB patients were diagnosed on the basis of serum HBV DNA >104 copies/mL, persistent or intermittent elevation of ALT levels, histologically documented moderate or severe necroinflammation and absence of cirrhosis. LC was diagnosed by histological analysis of liver biopsy tissues or by ultrasonography and supplemented with clinical evidence of portal hypertension. From each participant 5 mL blood was collected, and serum was isolated and stored in a −80°C refrigerator until use. The study was approved by the Institutional Ethical Review Committee.

Testing of serological markers and liver enzymes

Serum samples were tested for ALT/AST levels, total protein, albumin, alkaline phosphatase, prothrombin time (INR) and bilirubin as well as for serological markers (HBsAg, HBeAg and anti-HBe) using commercially available kits.

HBV DNA isolation and quantification

HBV DNA was extracted from 200 μL serum using the QIAamp DNA Mini Kit (Qiagen, Valencia, CA, USA). Viral load was determined by real-time PCR using SYBR Green PCR Master mix (Applied Biosystems, Foster City, CA, USA) [13]. The lower limit of detection was 250 copies/mL.

Amplification of full-length HBV genome sequencing and phylogenetic analysis

Full-length HBV DNA (c. 3.2 kb) was amplified by the one-step amplification method described by Gunther et al. [14]. In the case of a low HBV DNA level, a second round nested PCR was performed using two different primer sets, MP1 with R5 and F3 with MP2 (Table 1). The PCR products were sequenced directly with different internal primers (Table 1) using the BigDye terminator v3.1 cycle sequencing kit (Applied Biosystems) on an automated DNA sequencer. Sequence editing and analysis were performed using Seqscape V2.5 software.

Table 1.   Primers used in this study
Primer namePrimer sequencesLocation (nt)a
  1. aNt positions are given according to HBV sequence with accession no. AF121242 obtained from GenBank.

  2. bPrimer used for sequencing.

  3. cPrimer used for nested PCR.

F1b5′-CACAAGAGGACTCTTGGACT-3′1653–1672
F3b,c5′-CGCCTCATTTTGTGGGTCAC-3′2801–2820
F4b5′-CTCAGGCCATGCAGTGGAA-3′3164–3182
F5b5′-GATGTGTCTGCGGCGTTTTA-3′376–395
R1b5′-CCACCTTATGAGTCCAAGG-3′2457–2475
R3b5′-AACTGGAGCCACCAGCAG-3′57–74
R5b,c5′-AAAGCCCAAAAGACCCACAAT-3′996–1016
MP1c5′-GAGCTCTTCTTTTTCACCTCTGCCTAATCA-3′1821–1841
MP2c5′-GAGCTCTTCAAAAAGTTGCATGGTGCTGG-3′1806–1825

HBV sequences thus obtained were compared with representative sequences of ten HBV genotypes (A–J) retrieved from GenBank. Alignments were carried out using CLUSTALX software and a phylogenetic tree was constructed by the neighbour joining (NJ) method using the Kimura 2 parameter model in MEGA software version 5. To confirm the reliability of the phylogenetic tree analysis, bootstrap resampling and reconstruction were carried out 5000 times. Sequence variability was analysed with the help of multiple alignment data.

Statistical analysis

Data were analysed using SPSS (ver10.1) software. The results were presented as median (range) or mean ± standard deviation (SD) as appropriate. One-way ANOVA or the Kruskal–Wallis test were used, as appropriate, for group comparisons of quantitative variables depending on whether the variances of the groups were similar or not. P-trend test and Fisher’s exact test were performed for evaluating the role of a specific mutation with disease progression and to make pairwise group comparisons of mutation frequencies, respectively. For all tests, a p-value <0.05 was considered significant.

Nucleotide sequence accession numbers

The nucleotide sequences of 40 HBV/D isolates are available in the GenBank database (http://www.ncbi.nlm.nih.gov/GenBank/index.html) under accession numbers JN664909JN664948.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Transparency Declaration
  9. References

Clinical, serological and demographic data

Based on clinical, biochemical, serological and histological assessment, and HBV DNA levels, 16 out of 60 HBeAg-negative patients were categorized as IC, 20 as CHB and 24 as LC (Table 2).

Table 2.   Clinical, demographic and biochemical data of different categories of HBeAg-negative patients
 IC (n = 16)CHB (n = 20)LC (n = 24)p value
  1. SD, standard deviation; ALT, alanine aminotransferase; AST, aspartate aminotransferase; INR, international normalized ratio; IU, international unit.

  2. $One-way ANOVA p value.

  3. *Kruskal–Wallis test p value.

Age (year), mean ± SD46.44 ± 13.7931.73 ± 10.6742.73 ± 10.67<0.0388$
Sex, M:F9:513:717:7 
ALT (IU/L), median (range)31 (21–39)47 (38–67)89 (47–238)<0.0001$
AST (IU/L), median (range)30 (24–44)49 (45–95)92 (51–265)<0.0001$
Total protein (g/dL), mean ± SD6.68 ± 0.547.26 ± 0.357.58 ± 0.59<0.0081$
Albumin (g/dL), mean ± SD3.7 ± 0.473.75 ± 0.753.33 ± 0.68<0.3012$
Total bilirubin (mg/dL), median (range)0.89 (0.30–1.30)0.80 (0.30–1.30)1.8 (0.86–3.10)<0.0031$
Alkaline phosphatase (IU/L), median (range)110 (42–407)121 (42–263)187 (78–344)<0.004*
INR, mean ± SD1.11 ± 0.051.18 ± 0.051.35 ± 0.16<0.0001$
HBV DNA (copies/mL), mean ± SD3358 ± 39092.5 × 106 ± 3.5 × 1069.6 × 105 ± 2.6 × 106<0.0001*

HBV genotypes

The phylogenetic analysis of 60 complete HBV sequences obtained from HBeAg-negative patients revealed the presence of three distinct genotypes, D, A and C (Fig. 1). Forty out of 60 HBV isolates (66.66%) belonged to genotype D while 9 (15%) were of genotype A and 11 (18.30%) clustered with genotype C. Subsequent analyses of viral mutation with respect to different disease stages were carried out exclusively on 40 HBV/D isolates, of which 14 strains were from IC and 13 each from CHB and LC.

image

Figure 1.  Phylogenetic tree analysis of full-length sequences of 60 HBV isolates from chronic HBeAg-negative patients along with 50 reference sequences of HBV belonging to different genotypes (A–J) derived from GenBank, including three sequences of HBV from non-human primates. HBV sequences from GenBank are indicated by their genotypes followed by accession numbers and country of origin. The sequences determined in the study are given by the isolate number followed by the clinical status of the patient (IC/CHB/LC) from whom it was derived. The phylogenetic tree was built using the Jukes Cantor and neighbour joining (NJ) method by MEGA software version5 and bootstrap resampling and reconstruction were carried out 5000 times.

Download figure to PowerPoint

Mutational patterns in HBV/D in different phases of e-negative infection

Promoter/regulatory regions  The expression of the HBV genes is controlled by four different promoters, two enhancers (EnhI (nt. 1074–1234) and EnhII (nt. 1636–1741)) and a negative regulatory element (NRE; nt. 1613–1636). The commonly described double mutations in BCP (A1762T and G1764A) that cause reduced level of HBeAg expression were observed in 11 out of 40 (27.5%) HBV/D from HBeAg-negative subjects. These mutations were present in 30.76% of CHB and 53.84% of LC but were absent in IC (Ptrend = 0.002; Table 3). Another BCP mutation, C1773T, was found in CHB (30.76%) and LC (38.46%) but was undetected in IC. Overall, the combination of mutations A1762T+G1764A+C1773T increased significantly during the progressive phases from IC to LC (Ptrend = 0.009). Within EnhII, the frequency of C1694T was found to increase significantly in CHB (46.15%) as compared with IC (7.14%) (PIC vs. CHB = 0.029) although it declined among LC (23.07%; Table 3). An increasing prevalence of the A1635T/G mutation in NRE was also noted in advancing clinical stages of e-negative infection, being 14.28% in IC, 46.15% in CHB and 53.85% in LC (Ptrend = 0.033) (Table 3). In preS1 promoter (SP1; nt. 2682–2887) scattered nucleotide changes were observed in isolates from all three groups but the most noticeable mutation was at position nt. 2732 detected in one IC (7.14%), eight CHB (61.53%) and three LC (23.07%), suggesting its potential importance in the development of CHB (PIC vs. CHB = 0.004). No significant change was observed in other HBV promoters.

Table 3.   Frequencies of mutations across the complete genome of HBV of genotype D derived from different groups of HBeAg-negative patients
Mutation site in HBVClinical categories of patients from whom HBV was derived P trend testFisher exact test
Nucleotide substitution mutation [location]IC, n = 14, (%)CHB, n = 13, (%)LC, n = 13, (%)p1 (IC vs. CHB)*p2 (CHB vs. LC)*p3 (IC vs. LC)*
  1. *Fisher exact test p value; p<0.05 is indicated in bold.

 A1635T/G [NRE]2 (14.28)6 (46.15)7 (53.85) 0.033 0.0820.500 0.037
 C1694T [EnhII]1 (7.14)6 (46.15)3 (23.07)0.317 0.029 0.9520.269
 A1762T [BCP]04 (30.76)7 (53.84) 0.002 0.041 0.214 0.002
 G1764A [BCP]04 (30.76)7 (53.84) 0.002 0.041 0.214 0.002
 A1762T+G1764A04 (30.76)7 (53.84) 0.002 0.041 0.214 0.002
 C1773T [BCP]04 (30.76)5 (38.46) 0.016 0.041 0.500 0.016
 A1762T+G1764A+ C1773T02 (15.38)5 (38.46) 0.009 0.2220.189 0.015
 G1896A [Pre-core]02 (15.38)6 (46.15) 0.003 0.2220.101 0.006
 G1899A [Pre-core]01 (7.69)6 (46.15) 0.002 0.481 0.037 0.006
 G1896A+G1899A004 (30.76) 0.009 1 0.047 0.040
 C2732T [SPI]1 (7.14)8 (61.53)3 (23.07)0.334 0.004 0.9920.269
Amino acid substitution Mutation
 X ORF
  P46S1 (7.14)4 (30.76)6 (46.15) 0.023 0.1400.344 0.028
  T47S1 (7.14)3 (23.07)6 (46.15) 0.020 0.2690.205 0.028
  A102V/T2 (14.28)6 (46.15)7 (53.84) 0.033 0.0820.500 0.037
  xI88F/V2 (14.28)6 (46.15)7 (53.85) 0.033 0.0820.500 0.037
  xK130M04 (30.76)7 (53.84) 0.002 0.041 0.214 0.002
  xV131I04 (30.76)7 (53.84) 0.002 0.041 0.214 0.002
 Core ORF
  c S35A/T05 (38.46)3 (23.07)0.123 0.016 0.8990.098
  c E40D1 (7.14)7 (53.84)4 (30.76)0.166 0.011 0.9450.140
  c S183P01 (7.69)5 (38.46) 0.005 0.4810.080 0.016
 Surface ORF
  Δ PreS01 (7.69)4 (30.76) 0.016 0.4810.139 0.040
  preS1 A39R1 (7.14)9 (69.23)3 (23.07)0.339 0.001 0.9980.269
  preS1 S96A/T1 (7.14)7 (53.84)3 (23.07)0.327 0.011 0.9790.269
  s T118V/R1 (7.14)3 (23.07)4 (30.76) 0.016 0.4810.161 0.041

HBV PreS/S ORF  The amino acid (aa) sequences of HBV surface proteins (PreS1, PreS2 and S) deduced from the nucleotide sequences revealed in-frame deletions, ranging from single to 8 aa in four LC patients (30.76%) and one CHB patient (7.69%), while deletions were conspicuously absent among IC (Ptrend = 0.016). All deletions detected in LC were mapped within the B-cell epitope (aa 120–145) of PreS2 while a single amino acid deletion (aa 94) in PreS1 within overlapping B and T cell epitopes, was observed in an isolate from CHB. The most notable amino acid substitutions observed in PreS1 were A39R and S96A/T, both within B-cell epitope regions, and were found in a significantly high proportion among CHB as compared with IC (PIC vs. CHB < 0.05) (Table 3). Additionally, another mutation, T118V/R, localized within the major hydrophilic region (MHR; aa 100–169) of S was encountered at a high frequency in LC (30.76%), followed by CHB (23.07%) and IC (7.14%) (Ptrend = 0.016) (Table 3).

HBV × ORF  As HBV NRE/EnhII/BCP overlaps partially with the HBx coding sequence, mutations at nucleotides 1635(A[RIGHTWARDS ARROW]T), 1753(T[RIGHTWARDS ARROW]C), 1762(A[RIGHTWARDS ARROW]T) and 1764(G[RIGHTWARDS ARROW]A) induce substitutions in HBX at amino acid positions 88 (I[RIGHTWARDS ARROW]F), 127(I[RIGHTWARDS ARROW]T), 130(K[RIGHTWARDS ARROW]M) and 131(V[RIGHTWARDS ARROW]I) respectively. Concurrently, a significantly higher prevalence of xI88F, and HBx130 + HBx131 double mutations were noted in LC in comparison to IC (Table 3). Three other mutations, namely P46S and T47S within the B-cell epitope region (aa 29–48) and A102V/T in the helper T-cell (TH) epitope (aa 91–105) of X, were observed with advancement of disease severity, being most abundant in HBV from LC, followed by CHB and IC (Ptrend < 0.05 for each; Table 3).

HBV precore (PC)/core ORF  The two mutations, G1896A and G1899A, in the PC region frequently reported in e-negative HBV variants were absent in HBV/D from IC. However, they were noted in 15.38% and 7.69% of CHB, respectively, and in 46.15% of LC (Table 3). Thus the combination of G1896A+G1899A mutation might be a predictive risk factor for the development of cirrhosis from CHB (p 0.047). Variabilities in TH epitopes, the cytotoxic T-cell (CTL) epitope and the B-cell epitope of HBV core protein were high in LC in comparison to IC (Table 4). Of particular importance were substitutions S35A/T and E40D within the TH epitope that had been found to be significantly linked with the development of CHB (Table 3). In addition, the mutation S183P showed an increasing trend over the course of infection from IC to LC (Ptrend = 0.005; Table 3).

Table 4.   Frequencies of mutations in immune epitope regions of core ORF of HBV/D isolates from different clinical categories of HBeAg-negative patients
  Clinical categoriesRegions of HBV core gene
Th epitopesCTL epitopesB-cell epitopes
35–4548–6918–2784–10176–89105–116130–135
IC (n = 14) Frequency (%)3 (21.42)4 (28.57)2 (14.28)4 (28.57)6 (42.85)7 (50.00)6 (42.85)
CHB (n = 13) Frequency (%)6 (46.15)5 (38.46)3 (23.07)3 (23.07)4 (30.76)9 (69.23)3 (23.07)
LC (n = 13) Frequency (%)7 (53.84)7 (53.84)6 (46.15)5 (38.46)6 (46.15)9 (69.23)8 (61.53)

HBV polymerase ORF  The sequences of the HBV polymerase gene (except one) were found to have a wild-type reverse transcriptase (RT) domain, which is the prime target of antiviral therapy. As the PreS/S ORF completely overlaps the polymerase gene, the deletions observed in the PreS1/PreS2 regions of five HBV strains mentioned above correspond to deletions in the non-essential spacer regions of the polymerase.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Transparency Declaration
  9. References

The present study demonstrated that the advancing clinical stages of chronic e-negative infection correlate with a significant increase in the frequencies of mutations in the HBV/D genome. A majority of these mutations were localized in the transcription regulatory regions (BCP/EnhII/NRE/SP1) and in immunologically relevant sites. While some mutations successively increased across the three phases IC, CHB and LC, others were more important with a specific disease phase.

Consistent with previous studies on HBV genotypes B and C [15,16], our results indicated that BCP double mutation, A1762T and G1764A were associated with increased risk of severe liver disease in e-negative HBV/D-infected patients. Another BCP mutation, C1773T, which had been previously implicated in the development of cirrhosis in patients carrying genotypes B and C [16], also showed an increased prevalence with progression of disease in subjects with HBV/D. Interestingly, A1635T/G in NRE, C1694T in EnhII and C2732T in SP1 found to have a significant association with clinical outcome in this study were not documented previously with respect to any genotype. However, a study by Zhu et al. [17] on HBV/C derived from liver biopsy specimens of chronic hepatitis B patients in China reported the presence of mutation at 1635 but it did not correlate with cirrhosis or HCC. The preferential selection of coding mutations within epitope regions of HBV of genotypes B and C during chronic infection had been reported earlier. HBeAg seroconversion was found to be associated with an increased frequency of mutations in CTL directed epitopes [18] while a recent study highlighted an increased number of mutations in CTL epitopes in cirrhosis and HCC groups regardless of their HBeAg status [19]. Although we observed sequence heterogeneity in CTL epitopes of HBV/D, no statistically significant association was found with any disease state. In hepatoma patients of Taiwan, where HBV/B and C prevail, dominant mutations were detected within regions of aa 26–45 in the B-cell epitope and 116–127 in the T-cell epitope in HBx [20]. In contrast, our results indicated that variations in HBx were most pronounced at aa 46 and 47 in the B-cell epitope and at 102 in the TH epitope and occurred most frequently among LC. Our data are in agreement with a past study [21], demonstrating that HBV derived from anti-HBe-positive subjects with active hepatitis harboured a greater number of mutations in the TH epitope of the core protein and a high rate of substitution (E[RIGHTWARDS ARROW]D) was noted at aa 40. The PreS mutations, A2964C and T3116C, had been reported to be associated with increased risk of cirrhosis in HBV/C-infected patients [22]. The mutation A2964C corresponds to substitution at amino acid 39 (K[RIGHTWARDS ARROW]N) within the B-cell epitope of PreS1 in HBV/C. Mutation in this location was found to be significantly associated with the development of CHB from IC in our study population, although the amino acid change involved in HBV/D was different (A39R). The other D-specific mutation in PreS1, S96A/T, also within the B-cell epitope and increasingly present in CHB, has not been reported before. An increased variability at aa 126 in MHR of HBV/B had been reported during the progression of chronic hepatitis B in Taiwanese patients [23]. However, in the present study, the most significant mutation in MHR that was related to severity of e-negative infection was at position 118 (T118V/R). This mutation was also perceived in 6% of the chronic carriers in Morocco [24]. Similar to prior studies on HBeAg-negative patients infected with HBV of genotype B or C [25], our results also suggested that deletion in the preS region of HBV/D was an important predictor of cirrhosis. The only clinically important mutation in HBV/D found outside the epitopic region that increased successively during progression from IC through CHB to LC was S183P in C-terminus of HBV core protein.

The clinical significance of precore mutations G1896A and G1899A remains controversial, as they were linked with severe liver disease in some studies [26–28] while other studies failed to confirm the association [29,30]. Our data indicated that the prevalence of both of these mutations and also their combination was significantly high in LC compared with CHB and IC. Our result was also in accordance with a previous report from northern India where PC mutants of HBV were found to be associated with a more aggressive course of liver disease [8].

Thus the present study has identified mutations in HBV/D associated with different outcomes of chronic HBeAg-negative infection. The most important clinical implication of our findings is the possibility that by tracking these candidate viral genomic markers it may be possible to identify high-risk individuals well before clinical diagnosis of advanced liver disease so that early interventions would greatly benefit these patients and help in decreasing the overall incidence of cirrhosis and eventually HCC. However, a major limitation of our study is a small sample size and the use of a cross-sectional design, making it difficult to ascertain whether these mutations preceded the development of advanced liver diseases. Further large-scale prospective studies are needed to validate these observations.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Transparency Declaration
  9. References

We thank the Indian Council of Medical Research for funding this study through project No. 5/8/7/5/2008/ECD-I.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Transparency Declaration
  9. References