2.1. Ethics statement

All the procedures were in compliance with EU Directive 2010/63/EU for animal experiments. Study design protocols and standard operating procedures (concerning the hamsters and the fish) were approved by the Committee on the Ethics of Animal Experiments at the ICG SB RAS (permit number 42 of 25 May 2018).

2.2. Experimental design

Twenty-four SPF (specific pathogen-free) golden Syrian hamsters were used in this study. All the procedures with hamsters were performed at the SPF animal facility of the ICG SB RAS. For collecting O. felineus metacercariae, a naturally infected freshwater fish (Leuciscus idus) was net-caught in the Ob River near Novosibirsk (Western Siberia, Russia). O. felineus metacercariae were extracted as described previously [13]. C. sinensis and O. viverrini metacercariae were extracted from naturally infected freshwater fish (Seoul, Republic of Korea, and Khon Kaen, Thailand, respectively). The extracted metacercariae were identified under a light microscope.

Two-month-old male hamsters (M. auratus) were randomly subdivided into four groups: uninfected (group 1) and infected orally with 75 metacercariae of O. felineus (2), C. sinensis (3), or O. viverrini (4) by gastric intubation. One and 3 months after the infection initiation (postinfection: p.i.), three hamsters from the uninfected group and three hamsters from each of the other groups were euthanized using carbon dioxide. All the infections were conducted with a difference of approximately 5–7 days between infections with each species of trematode. All hamsters of the same age from different litters were combined into one group, from which hamsters were randomly selected for both the infected and uninfected groups to exclude the batch effect. Thus, the animals were also sacrificed with a difference of 5–7 days.

All the hamsters were examined daily for signs of illness, injury, or abnormal behavior by SPF-trained personnel. Food and water availability and the macroenvironment (temperature, humidity, noise, light intensity, and cleanliness) were also evaluated daily. No unexpected deaths of hamsters were registered during this study. To analyze the infection intensity the EPG score (number of eggs per gram stool) was calculated at 3 months p.i.. No statistically significant differences in EPG score were found among groups. The data indicate that liver flukes infected hamsters with the similar extent.

2.3. Tissue collection and RNA extraction

The liver was quickly excised and frozen in liquid nitrogen and then stored at -80°C until analysis. RNA was extracted from the frozen tissue with the Aurum Total RNA Extraction Kit (Bio-Rad, USA). The total RNA was purified on Agencourt RNAClean XP beads (Beckman Coulter, Germany). The quality and quantity of the total RNA were evaluated on a NanoDrop 2000 spectrophotometer. The quality of samples for RNA-seq was assessed using Agilent 2100 Bioanalyzer and the Total RNA Nano Kit (Agilent Technologies, USA). Only samples with an RNA integrity number greater than 8.0 were chosen for gene expression analysis.

2.4. Library construction and sequencing

Blind coding was used during DNA library preparation. All the DNA libraries were constructed by the same personnel approximately within 1 week in accordance with the standard New England Biolabs protocol. Briefly, polyA-tailed mRNAs were purified from 1 μg of total RNA using the NEBNext Poly(A) mRNA Magnetic Isolation Module. Then, directional cDNA libraries were created by means of the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina. Size selection of DNA fragments was performed on Agencourt AMPure XP beads (Beckman Coulter, USA). Next, PCR enrichment of the adapter-ligated library was conducted (six cycles of PCR). The size and quantity of the library were verified on the Agilent Bioanalyzer, and libraries were subjected to paired-end (2 × 150) sequencing on the Illumina HiSeq 2500 platform (Genewiz LLC, USA).

PCR enrichment of the adapter-ligated library was then performed (six PCR cycles). The size and quality of DNA libraries were verified using the Agilent High Sensitivity DNA Kit. The length of the target fragment in all libraries was ~320 bp.

2.5. Gene expression analysis

On average, ~37 million paired-end reads (range: 32–50 million) were obtained from each cDNA sample by Illumina stranded sequencing. The quality of raw sequencing data was checked using FastQC v.0.11.9, and the sequencing data were preprocessed with the Trimmomatic 0.36 tool to remove adapters and low-quality sequences [14]. The data were mapped to the M. auratus BCM_Maur_2.0 reference genome assembly in STAR version 2.7.10b [15]. To analyze differential gene expression, the Wald test with a threshold of 0.05 from R package DESeq2 (v.1.38.3) was performed [16]. The Benjamini–Hochberg correction for multiple testing was applied to the resulting p values, and the genes with an adjusted P value (P_adj) < 0.05 and whose expression fold changes were greater than 2.0 were designated as DEGs. Weakly expressed genes (the sum of gene expression values for all libraries less than 47) were removed from the expression matrix.

Principal component analysis (PCA) for assessing the degree of library clustering was performed using the expression matrix of all genes after logarithmic transformation [function rlog() with parameter blind = F from R package DESeq2]. R package PCAtools (v.2.10.0) was employed for this analysis [17]

Pathway enrichment analysis was performed using Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Molecular Signatures Database (MsigDB) databases based on mouse genome annotation [R package org.Mm.eg.db (v.3.16.3) for analysis via GO and KEGG, and R package msigdb (v.1.6.0) for analysis via MsigDB]. R package clusterProfiler (v.4.6.2) was used for this analysis [18,19].

Differences in the response of various cell types to infection were determined using R package BisqueRNA (v.1.0.5) and published data from single-cell RNA sequencing (scRNA-seq) analyses of the human liver [20,21].

Identification of the gene clusters manifesting specific patterns across samples was carried out by the likelihood ratio test (LRT) from the DESeq2 R-package and DEGreport (v.1.28.0) R-package [22]. Rlog-transformed normalized counts were employed to identify only DEGs (P_adj < 0.05), then the degPatterns function was executed to find the sets of genes that showed similar expression patterns across sample groups. Groups of genes from clusters were subjected to functional analysis to identify the associated functions.

Differences in the response of liver cell types to infection were evaluated by means of R package BisqueRNA (v.1.0.5) and published scRNA-seq data from the human liver [20,21]. The data distribution was checked for normality by the Shapiro–Wilk test and by construction of distribution histograms [stats (v.4.3.0) R package]. For normally distributed data, significance was tested by ANOVA followed by Tukey’s test for multiple comparisons. For data that were not normally distributed, the Kruskal–Wallis test was performed followed by the pairwise Kruskal–Wallis test with the Benjamini–Hochberg correction for multiple comparisons.

2.6. Histopathological analysis

For this analysis, the liver tissue from two lobes was carefully dissected, placed in 10% formaldehyde in phosphate-buffered saline (BioVitrum, Russia), and processed as described earlier [23–25]. The tissue slices were stained with hematoxylin and eosin by the standard method [23–25] and were examined under a light microscope (Axioskop 2 Plus; Zeiss, Germany).

To assess the severity of liver lesions, a semiquantitative histological analysis of the ratio (morphometry) type [26] was performed on all visual fields of two lobes of each animal. Blind coding was used during histopathology evaluation. The evaluation was conducted by two trained observers in parallel, and then a senior pathologist reviewed and advised on controversial cases. Sections of two liver lobes from each animal were stained with hematoxylin and eosin, after which a morphometric grid of 10 × 10 squares was superimposed on each consecutive field of view (100–150 photographs per animal). The number of squares occupied by such lesions as periductal fibrosis, cholangiofibrosis, infiltration by inflammatory cells, BiliN, and hyperplasia of the bile duct epithelium was determined, and the proportion of the liver area occupied by the lesion (P) was calculated using the formula: P = (X/Y) × 100%, where X is the number of affected squares, and Y is the number of all squares in all fields of view. Each field of view was scored, and a mean score was assigned to the whole tissue of that sample [24–26]. A heat map was constructed using the heatmap.2 (v.2.38) R package.

2.7. Weighted gene coexpression network analysis (WGCNA)

This analysis was carried out to search for clusters of coexpressed genes [27]. It allows to detect gene clusters having similar expression patterns and to calculate a correlation of these clusters with experimental conditions or clinical manifestations [27]. For the analysis, we chose genes for which P_adj < 0.05 in at least one type of infection (9557 genes). The Pearson correlation coefficient was used to assess the correlation. An optimal threshold value was selected between 1 and 20 using the pickSoftThreshold() function with parameters RsquaredCut = 0.8 and networkType = "signed hybrid." A calculated threshold value of 14 was utilized for this study (WGCNA R package version 1.72–5).

The correlation of clusters with various structural liver lesions (periductal fibrosis, inflammation, cholangiofibrosis, and dysplasia) was computed from the semiquantitative histopathological data (stats R package version 4.3.0). Analysis of functional enrichment of clusters was carried out with R packages clusterProfiler and msigdb in the MSigDB database.

2.8. Western blotting

Liver lysates were prepared using RIPA Buffer according to a previously described method [23]. The protein concentration in each lysate was determined with the BCA Protein Assay Kit (Thermo Scientific, USA). Western blot analysis was performed in accordance with the standard procedure [23]. Western blot results were quantified in the Quantity One software (v.4.6.9). The resulting intensity values were then normalized to β-actin. The expression level of the β-actin gene did not change significantly according to our analysis of transcriptomic data. Statistical significance was calculated by the nonparametric Kruskal–Wallis test and the pairwise Wilcoxon test with the Benjamini–Hochberg correction using R package stats (v.4.3.0). Antibodies and dilutions employed in this study were as follows: anti–beta-actin (1:3000; cat. # ab8226, Abcam), anti-MMP9 (1:2000; cat. # ab228402, Abcam), anti-TNFa (1:2000; cat. # MAA133Ra21, CloudClone), anti–α-smooth muscle actin (α-SMA; 1:2000; cat. # ab124964, Abcam), anti-CD68 (1:2000; cat. # ab201340, Abcam), anti-CD163 (1:200; cat. # ab199402, Abcam), anti–E-cadherin (1:3000; cat. # ab76055, Abcam), and anti–IL-10 (1:2000; cat. # ab25073, Abcam) antibodies with respective secondary antibodies: a horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (H+L) antibody (1:7000; cat. # AS014, ABclonal, USA) and an HRP-conjugated rabbit anti-mouse IgG (H+L) antibody (1:40000; cat. # ab97046, Abcam, USA). Quantitative densitometric analyses were performed on digitized images of immunoblots in the Quantity One v.4.6.9 software (Bio-Rad, USA).

2.9. Statistical analysis

The data were subjected to statistical analysis in the STATISTICA 6.0 software (Statsoft, USA). The Shapiro–Wilk test was carried out to check whether the data distribution was normal. The one-way ANOVA F-test with Newman–Keuls post hoc analysis was applied to significant main effects and interactions to evaluate differences between some sets of means.

3.1. Outline of sequencing results

On average, ~37 million paired-end reads (range: 32–50 million) were obtained from each sample by Illumina stranded sequencing (S1 Table). Genome mapping using the STAR software identified 785 million (86.8%) uniquely mapped reads, whereas 39 million (4.3%) reads were mapped more than once.

PCA revealed clustering of samples into principal components 1 and 2 depending on trematode species (Fig 2A). PC1 reflects the response factor to any infection, and the severity of this response was found to increase toward C. sinensis infection. PC2 probably indicates some specific response to O. viverrini infection. Moreover, PCA showed clustering of RNA samples into four separate groups based on liver fluke species. C. sinensis caused the largest shift in PC1, O. viverrini the smallest shift in PC1, and infection with O. felineus occupied an intermediate position. For principal component 2, the largest shift was caused by infection with O. viverrini, and PC2 did not explain any variation in the group of C. sinensis samples.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Sequencing data on liver samples from hamsters infected with O. viverrini, O. felineus, or C. sinensis at 1 and 3 months p.i.

a. PCA of samples; b. Volcano plots of the distribution of DEGs (|log₂FoldChange| > 1) and P_adj < 0.05 [-log₁₀(P_adj) > 1.3]. Genes with reduced expression are highlighted in red, and genes with elevated expression are highlighted in green.

https://doi.org/10.1371/journal.pntd.0012685.g002

A total of 3148 genes were found to be differentially expressed (P_adj< 0.05, |log₂FC| > 1) in C. sinensis–infected hamsters (1884 upregulated and 1264 downregulated) as compared with the uninfected animals (S2 Table). A total of 1464 genes proved to be differentially expressed in O. viverrini–infected hamsters as compared with the uninfected animals (593 upregulated and 871 downregulated) (S3 Table), and 1408 genes were differentially expressed in O. felineus–infected hamsters as compared with the uninfected animals (779 upregulated and 629 downregulated) (S4 Table). Volcano plots of differentially expressed genes (DEGs) between infected and uninfected groups corresponded to one of the time points analyzed (1 or 3 months) are presented in S1 Fig.

3.2 Enrichment analysis of DEGs

Functional gene enrichment analysis revealed that the largest number of cellular pathways was associated with infection by C. sinensis: 329 significantly overrepresented GO biological processes (false discovery rate [FDR] < 0.05), 14 significantly enriched (q-value < 0.05) gene sets (hallmarks) according to the Molecular Signatures Database, and 57 significantly enriched (P_adj < 0.05) KEGG pathways. The smallest number of pathways was associated (in terms of significant enrichment) with O. viverrini infection: 65 significantly overrepresented GO biological processes, eight significantly enriched (q-value < 0.05) MSigDB gene sets (hallmarks), and 25 significantly enriched KEGG pathways. O. felineus infection was associated with 211 significantly over-represented GO biological processes, 15 significantly enriched MSigDB gene sets (hallmarks), and 30 significantly enriched KEGG pathways.

Fig 3A presents 23 significantly enriched gene sets (hallmarks) (q-value < 0.05) according to the Molecular Signatures Database. The MSigDB gene sets most enriched within the group of overexpressed genes—common to all three infections—were the inflammatory response, IL2-STAT5 signaling pathway, and TGF-beta signaling pathway (Fig 3A). Xenobiotic metabolism turned out to be enriched within the group of downregulated genes common to all three infections. Of note, enriched pathways common for O. viverrini and C. sinensis were found that were not affected in O. felineus–infected animals, in particular, genes downregulated in response to ultraviolet radiation. For O. felineus and C. sinensis infections, but not for O. viverrini–infected animals, pathways associated with inflammation were detected, in particular TNF-α signaling via NF-kB, IL6-JAK-STAT3 signaling, epithelial-mesenchymal transition, apoptosis, E2F targets, and angiogenesis. The MSigDB gene sets significantly associated only with C. sinensis infection were Hallmark_KRAS_signaling_up (M5953, genes upregulated by KRAS activation), Hallmark_apical_surface (M5916, genes encoding proteins over-represented on the apical surface), Hallmark_myogenesis (M5909), Hallmark_G2M_checkpoint (M5901), and Hallmark_apical_junction (M5915, components of the apical junction complex).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Enrichment analysis of DEGs using MSigDB gene sets (a) and KEGG signaling pathways (b). CS: C. sinensis, OF: O. felineus, OV: O. viverrini. The size of the dots indicates the Overlap score of the pathway genes, and the color of the dots reflects different q-values (a) and P_adj values (b).

https://doi.org/10.1371/journal.pntd.0012685.g003

Fig 3B presents 32 most significantly enriched KEGG signaling pathways. The KEGG signaling pathways common to the three infections in the subgroup of upregulated DEGs were TGF-beta and PI3K-Akt signaling pathways and Hippo and MAPK signaling. The O. viverrini infection subgroup of DEGs was found to be significantly enriched with signaling pathways regulating pluripotency of stem cells (Fig 3B). C. sinensis infection stimulated Wnt, Relaxin signaling, NF-kB signaling, and JAK-STAT signaling. Pathways common between O. viverrini and C. sinensis infections were found that were not upregulated in O. felineus–infected animals, in particular, RAS and FOXO signaling pathways. O. felineus and C. sinensis infections, but not O. viverrini infection, were associated with pathways related to high levels of inflammation, such as TNFα, chemokine, and P53-signaling pathways, Rap1 and ErbB signaling.

3.3. Gene cluster analysis of DEGs

In-depth clustering analysis revealed specific patterns of DEGs across the groups of liver RNA samples. To identify such genes, we performed the likelihood ratio test, which grouped genes based on their expression patterns across groups of liver RNA samples. Thus, the genes were subdivided into several clusters (gene groups) (Figs 4, S2 and S5 Tables). In particular, cluster 1 (DEGs upregulated in O. viverrini and O. felineus but not in C. sinensis at 1 month p.i.: (114 genes). Cluster 2 (DEGs upregulated in all three trematode infections at 1 month p.i.: 72 genes); cluster 3 (DEGs upregulated only in C. sinensis–infected animals at 1 month p.i.: 52 genes); and cluster 4 (DEGs upregulated only in O. viverrini–infected and C. sinensis–infected animals at 1 months p.i.: 35 genes).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Clustering analysis of DEGs found in the liver of hamsters (M. auratus) infected with one of three species of trematodes (at 1 and 3 months p.i.).

Likelihood ratio (LRT) test revealed gene clusters exhibiting particular patterns across samples. Pathway enrichment analysis was performed using Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Molecular Signatures Database (MsigDB) databases. UN: Uninfected; CS: C. sinensis, OF: O. felineus, OV: O. viverrini.

https://doi.org/10.1371/journal.pntd.0012685.g004

MsigDB pathways that were found to be enriched (FDR < 0.05) within cluster 1 (upregulated at 1 month p.i.) are related to adipogenesis, and Acyl-CoA Hydrolase Activity was an over-represented GO term. The KEGG pathways most enriched within cluster 2 included Ras and Hippo signaling, transcriptional misregulation in cancer, and sphingolipid signaling pathway (Fig 2). MsigDB pathways that were found to be enriched (FDR < 0.05) within this cluster are related to TNF-α signaling via NF-kB and TGF-beta signaling pathway.

In contrast to cluster 2, the MsigDB pathway that was found to be enriched (FDR < 0.05) within cluster 3 (upregulated in C. sinensis only at 1 month p.i.) was myogenesis. It was this module that contained 52 genes whose products are associated with cytoskeleton organisation, NAD biosynthesis via nicotinamide riboside salvage pathway, and protein auto-ADP-ribosylation.

The KEGG pathways most enriched within cluster 4 (upregulated in O. viverrini–infected and C. sinensis–infected hamsters) included fatty acid elongation, biosynthesis of unsaturated fatty acids, propanoate metabolism, and peroxisome proteasome (Fig 2). MsigDB pathways that proved to be enriched (FDR < 0.05) within this cluster are related to bile acid metabolism, fatty acid metabolism, and xenobiotic metabolism.

3.4. Estimation of liver cell type proportions in bulk expression

Cell heterogeneity is of interest in profiling changes in tissue composition associated with pre-cancerous changes. Measures of cell composition can be leveraged to identify cell-specific differential expression from bulk RNAseq data.

Each cell type has its own expression profile and metabolic and signaling pathways and responds to an infectious agent (such as a helminth) in a unique way. On the other hand, it is unknown which pathways in liver cells in general and the responses of which cell types in particular may explain the difference in the carcinogenic properties among the liver flukes. Using the bisqueRNA package for R, interspecies differences in the response of different liver cell types to the trematode infections were determined next.

The bisqueRNA package extracts trends in cellular composition from normalized RNA-seq data by means of only cell type–specific marker genes that have been obtained using single-cell reference data [21]. The logic behind this approach is that the expression of marker genes will directly correlate with the proportion of a relevant cell type, that is, the higher the expression of marker genes, the greater is the contribution of the cell type that owns this marker. This linear covariance can be captured in PCA. Moreover, the more specific a gene’s expression to a given cell type, the more accurately its expression will reflect the contribution of this cell type. The first component from PCA has been shown to capture most of the variance and can provide an estimate of the differences in the response of different cell types to a factor of interest [20].

Using a single-cell–based reference profile [21], 20 cell types were identified using sequencing of 8444 cells and lists of marker genes for each cell type. As a result of applying this package to our transcriptome data, it was revealed that there were significant differences for cholangiocytes, stellate cells, macrophages, endothelial cells, and hepatocytes (S6 Table).

The contribution of stellate cells—whose markers of the activated form are Acta2, Col1a1, Tagln, Col1a2, Col3a1, Sparc, and Rbp1—changed significantly only during infection with C. sinensis compared to the control (pairwise Kruskal–Wallis test with Benjamini–Hochberg correction, P_adj = 0.030) (Fig 5A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Cell-proportion estimates in RNAseq data.

The figure shows proportion estimates for stellate cells (a), cholangiocytes (b), inflammatory macrophages (c), noninflammatory macrophages (d), and hepatocytes of six cell types (e). The arrow indicates the order of arrangement of hepatocyte types from the portal triad (Hep3) to the central vein (Hep2). The data were analyzed using ANOVA (F) or Kruskal-Wallis (H) tests and are presented as box-and-whisker diagram; *p < 0.05; **p < 0.01; ***p < 0.001 as compared to the uninfected (Un) group.

https://doi.org/10.1371/journal.pntd.0012685.g005

C. sinensis infection caused the strongest response of cholangiocytes, which are characterized by overexpression of such genes as Krt7, Krt19, Epcam, Fxyd2, Cldn4, Cldn10, Mmp7, Cxcl1, and Tff2 (Kruskal–Wallis test with Benjamini–Hochberg correction, P_adj = 0.013), whereas no significant changes in cell types were detected during O. viverrini infection (Fig 5B).

The analysis detected two cell types related to macrophages (Fig 5C and 5D). Both cell populations are positive for the Cd68 marker, with inflammatory macrophages featuring overexpression of genes Lyz, Csta, and Cd74, and the noninflammatory macrophages are distinguished by overexpression of e.g. Cd5l, Marco, Vsig4, Ccdc88a, and Lipa. Inflammatory macrophages produce proinflammatory cytokines (IL-1β and TNFα) and are antigen-presenting cells capable of inducing Th1/Th17 cellular responses and killing pathogens. Anti-inflammatory macrophages, on the contrary, produce anti-inflammatory cytokines IL-10, TGF-β, and others and induce Th2/T_reg cellular responses. They respond to extracellular parasites and are associated with resolution of inflammation and regeneration of damaged tissue [28].

This method uncovered a significant response of inflammatory macrophages only to C. sinensis infection (pairwise Kruskal–Wallis test with Benjamini–Hochberg correction, P_adj = 0.013) (Fig 5C and S7 Table), whereas the response of anti-inflammatory macrophages changed significantly for all infections compared to uninfected animals (ANOVA with Tukey’s correction: for C. sinensis: P_adj = 0.00003, O. felineus: P_adj = 0.006, O. viverrini: P_adj = 0.014) (Fig 5D).

The six hepatocyte types in question [21] are characterized by lower proliferative activity and overexpression of albumin (ALB). The difference between hepatocytes that separates them into different cell types is the location of the cell types in the acinus on the central vein–portal triad axis (Fig 5E). Hep3 cells are periportal hepatocytes, while Hep2 are pericentral hepatocytes. Hep1 cells are located closer to periportal hepatocytes, and hepatocytes of the Hep5, Hep4, and Hep6 types are located between Hep3 and Hep1 cells (Fig 5E).

Periportal hepatocytes (Hep3) are characterized by the expression of such marker genes as Hmgcs1, Acss2, Tm7sf2, Tmem97, Msmo1, and Lepr, and pericentral hepatocytes (Hep2) by the expression of Ghr, Aldh6a1, Rcan1, Ar, Plin1, Rprd1b, and others.

Liver fluke infections induced changes in the proportion of pericentral Hep2 cells and proportions of Hep1, Hep5, and Hep6 central hepatocytes but no significant changes were detected in the proportion of periportal Hep3 hepatocytes (Fig 5E). Full statistical comparisons between hepatocyte clusters by ANOVA with Tukey’s test and by the pairwise Kruskal–Wallis test with the Benjamini–Hochberg correction are presented in S7 Table.

3.5. Validation of protein expression in the liver

Western blot analysis of marker proteins was performed on hamsters’ liver lysates at 3 months p.i. (Fig 6A and 6B). α-SMA levels were significantly increased in liver tissue in all three infections (P = 0.0065 for C. sinensis infection, P = 0.0229 for O. felineus infection, and P = 0.0065 for O. viverrini infection), thus confirming activation of stellate cells and the development of fibrosis in liver tissue. Matrix metalloproteinase MMP9 was also found to be upregulated in C. sinensis infection (P = 0.026), and this finding may be related to epithelial–mesenchymal transition, which was found above to be associated with this infection in our functional enrichment analysis of DEGs. CD68, which is a macrophage marker protein, was upregulated by C. sinensis infection and O. felineus infection. Levels of IL10, CD163, and E-cadherin did not change significantly at 3 months p.i.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Levels of marker proteins in the liver of hamsters (M. auratus) with O. felineus (OF), O. viverrini (OV), or C. sinensis (CS) infection at 3 months p.i. as assessed by western blotting.

a. Immunoblots. Representative images are shown. α-SMA: α-smooth muscle actin. b. Results of the densitometry of the immunoblots (A); they are normalized to β-actin. The data are presented as box-and-whisker diagram; *p < 0.05; **p < 0.01 as compared to the uninfected (Un) group (Kruskal-Wallis test).

https://doi.org/10.1371/journal.pntd.0012685.g006

3.6. Liver histopathology

A scoring method of the ratio (morphometry) type [26] was applied to structural liver lesions (S8 Table). The most common structural liver lesion was periductal fibrosis, which was registered in all infections at 1 month p.i. but was most severe in C. sinensis–infected and O. felineus–infected hamsters (Fig 7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Liver histopathology of hamsters uninfected or infected with C. sinensis, O. felineus, or O. viverrini at 1 and 3 months p.i.

BD: bile duct, PF: periductal fibrosis; E: epithelium of bile duct; BiliN: biliary intraepithelial neoplasia; Inf: inflammatory infiltrate. Scale bar = 50 μm in main images, and scale bar = 20 μm in insets.

https://doi.org/10.1371/journal.pntd.0012685.g007

Based on the obtained morphometry data, PCA was performed (Fig 8A), and its findings were presented as a heat map (Fig 8B and S8 Table). The area of cholangiofibrosis enlarged significantly only at 3 months p.i. The area of inflammatory-cell infiltration and bile duct epithelial hyperplasia increased at 1 month p.i. and remained elevated at 3 months p.i. in all infected animals. Notably, epithelial hyperplasia was most pronounced during O. viverrini infection (Fig 8B). Biliary intraepithelial neoplasia (BiliN) increased significantly in all three types of infection at 3 months p.i.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Quantitative histological data and association with coexpressed gene clusters in the liver.

a. PCA based on semiquantitative histological evaluation. A scoring method of the ratio (morphometry) type [26] was applied to data from measurements (S8 Table). This method is based on examination of several fields of view in a tissue (e.g., 10 random 400× fields) for each animal, each field was scored, and a mean score was assigned to the whole tissue of that animal. Each slice of a lobe was analyzed in all fields of view (20–30 fields). Each field of view was divided into 100 squares in Morphometry software. PC: principal component; b. Heatmap of liver morphometry data. Color codes reflect the area of structural damage as a percentage of the total area. *p < 0.05 as compared to the uninfected animals, ^#p < 0.05 as compared to O. viverrini. Un: uninfected; OF: Opisthorchis felineus, OV: O. viverrini, CS: C. sinensis; BiliN: biliary intraepithelial neoplasia. c. Correlation of coexpressed gene modules with structural liver damage (Pearson correlation coefficients greater than 0.3 are shown); d. Correlation of coexpressed gene modules with helminth species (^Pearson coefficient > 0.3; ^^Pearson coefficient > 0.5; ^^^Pearson coefficient > 0.7).

https://doi.org/10.1371/journal.pntd.0012685.g008

PCA (Fig 8A) revealed that the clustering of tissue samples from O. viverrini–infected animals was due to BiliN. Samples from animals infected with C. sinensis and those infected with O. felineus clustered together, and their clustering was more influenced by periductal fibrosis. It is noteworthy that clustering of tissue samples from animals at 1 month p.i. is associated with inflammation, whereas cholangiofibrosis is characteristic of 3 months p.i.

3.7. Weighted gene coexpression network analysis (WGCNA)

This was an analysis of genes whose expression changed significantly during at least one type of infection; nine clusters of coexpressed genes were obtained (Fig 8C and 8D and S9–S17 Tables). Pearson analysis of correlation between the identified clusters of coexpressed genes and the structural liver morphometry data (ST8 Table) revealed cellular signaling pathways associated with some liver lesions (Fig 8C). Pearson correlation coefficients are presented when greater than 0.3.

Inflammation and periductal fibrosis proved to be enhanced in similar ways and were associated with the activation or suppression of similar groups of coexpressed genes (Fig 8C). In particular, cluster 8 (2113 genes) directly correlated with inflammation and periductal fibrosis (Fig 8C and S16 Table). This cluster of genes is enriched with PI3K-AKT-mTOR, TGFβ, and IL6-JAK-STAT3 signaling pathways. Besides, clusters 1 (S9 Table) and 9 (S17 Table) also directly correlate with inflammation and periductal fibrosis and are enriched with signaling pathways such as Hippo, MAPK, ErbB, PI3K-AKT-mTOR, IL6-JAK-STAT3, and TGFβ, consistently with published data [10,11,29,30]. Moreover, an inverse correlation with these structural lesions was observed for clusters 2, 3, 4, and 7 (S10–S12 and S15 Tables). These clusters of genes are enriched with such terms as metabolism of bile acids, fatty acids, and xenobiotics, with the PPAR signaling pathway, and the DNA repair pathway. This outcome may be due to metabolic dysfunction of the liver and bile ducts, which is usually observed in trematodoses.

In contrast, such lesions as cholangiofibrosis and BiliN were associated with a different pattern of gene expression, in particular, gene cluster 1 (TNFα signaling via NF-kB, epithelial-mesenchymal transition, p53 pathway, and TGF-beta signaling) was not correlated with cholangiofibrosis and BiliN. On the other hand, gene cluster 5 (S13 Table) was more associated with cholangiofibrosis and BiliN, but the correlation coefficient was low, just as in cluster 8. Cluster 4, enriched with DNA repair pathway genes, is inversely correlated with BiliN (S12 Table).

Given that we were also interested in differences among the three infections, we examined possible correlations of coexpressed gene modules with the type and duration of infection (Fig 8D). Pearson correlation coefficients are shown when greater than 0.3. Cluster 1 (594 genes, S9 Table), which is enriched with TNFα, NF-kB, TGFβ, p53, and epithelial-mesenchymal transition pathways, inversely correlated with the absence of infection at 1 month p.i. Cluster 7 (585 genes, S15 Table) was inversely correlated with C. sinensis infection and proved to be enriched with fatty acid, bile acid, and xenobiotic metabolism pathways as well as adipogenesis. Cluster 9 (2149 genes, S17 Table) directly correlated with C. sinensis infection and proved to be enriched with IL2-STAT5 and IL6-JAK-STAT3 pathways, epithelial-mesenchymal transition, and inflammatory response pathways.

Clusters 5 and 6 are directly correlated with O. viverrini infection and are enriched with a set of genes regulated by MYC, genes most downregulated in hypoxia-tolerant cancer, and with the pathways of metabolism of fatty acids, bile acids, and xenobiotics as well as oxidative phosphorylation, protein secretion, and others.

Clusters 2 (2699 genes, S10 Table) and 4 (924 genes, S12 Table) directly correlated with the absence of infection and inversely correlated with C. sinensis infection. Cluster 2 (S10 Table) is enriched with DNA repair pathways, fatty acid and xenobiotic metabolism, and adipogenesis, while cluster 4 is enriched with DNA repair pathways (S12 Table).