A Transcriptomic Signature of Mouse Liver Progenitor Cells

Abstract

Liver progenitor cells (LPCs) can proliferate extensively, are able to differentiate into hepatocytes and cholangiocytes, and contribute to liver regeneration. The presence of LPCs, however, often accompanies liver disease and hepatocellular carcinoma (HCC), indicating that they may be a cancer stem cell. Understanding LPC biology and establishing a sensitive, rapid, and reliable method to detect their presence in the liver will assist diagnosis and facilitate monitoring of treatment outcomes in patients with liver pathologies. A transcriptomic meta-analysis of over 400 microarrays was undertaken to compare LPC lines against datasets of muscle and embryonic stem cell lines, embryonic and developed liver (DL), and HCC. Three gene clusters distinguishing LPCs from other liver cell types were identified. Pathways overrepresented in these clusters denote the proliferative nature of LPCs and their association with HCC. Our analysis also revealed 26 novel markers, LPC markers, including Mcm2 and Ltbp3, and eight known LPC markers, including M2pk and Ncam. These markers specified the presence of LPCs in pathological liver tissue by qPCR and correlated with LPC abundance determined using immunohistochemistry. These results showcase the value of global transcript profiling to identify pathways and markers that may be used to detect LPCs in injured or diseased liver.

Figure 1

Principal component analysis (PCA) displays transcriptome clustering sorted by tissue type. Each node represents the average of at least 3 microarray replicates and is color-coded according to tissue type. The three principal components that captured most differences in the datasets are plotted with three different three-dimensional views of the same PCA visualization shown in (a), (b), and (c).

Figure 2

Hierarchical clustering identifies 5 clusters (A, B, C, D, and E) of distinct gene expression data for liver progenitor cells. ANOVA was performed to compare liver progenitor cells to all other cell/tissue type groups. The 8,623 probe sets that displayed expression levels significantly different to LPCs (p < 0.01) were subsequently clustered according to the parameters of Euclidean distance and complete linkage.

Figure 3

Probe set clusters A, C, and D contain pathways that are overrepresented in LPCs and liver tissue. Clusters A and C contain probe sets that are upregulated in liver progenitor cells (LPCs) and cluster D contains those that are upregulated in developed liver whilst being low in LPCs. The Database for Annotation, Visualization and Integrated Discovery online tool was used to identify overrepresented pathways (p < 0.05) within each of these clusters. Pathways are displayed together with the number of genes from the cluster list (counts), the total that belongs to each pathway (pathway total) and corresponding p values.

Figure 4

Distinct expression profiles for liver progenitor cells identify three gene expression groups. (a) Hierarchical clustering of expression profiles that correlate with known liver progenitor cell markers. (b) Close view of the yellow highlighted rows from (a), with particular focus on the LPC and developed liver groupings. Three distinct groups of gene expression patterns (α, β, and γ) are defined.

Figure 5

Four liver injury models display different degrees of pathology and liver progenitor cell (LPC) response. Liver sections from wild-type mice subjected to a choline-deficient, ethionine-supplemented (CDE; (a) and (b)) or 3,5-diethoxy-carbonyl-1,4-dihydrocollidine (DDC; (c) and (d)) diet, or transgenic 178.3 ((e) and (f)) or Met-Kb ((g) and (h)) mice were stained with panCK ((a), (c), (e), and (g)) or H&E ((b), (d), (f), and (h)). Arrows in panels (b) and (d) and (h) indicate LPCs, ductular reactions, and small basophilic cells, respectively. Arrowheads in panels (b) and (d) indicate steatosis and porphyrin accumulations, respectively. Scale bars represent 100 μm.

Figure 6

A liver progenitor cell (LPC) transcriptomic signature can identify liver injury models with LPC induction. qPCR-generated mRNA expression levels of known (Cd24a and Sox9) and identified (group α, Murc; β, Ncam1 and Ltbp3; and γ, Mcm2 and M2pk) LPC markers in four different liver injury models. Data are normalized to gene expression in the appropriate control livers (indicated by the dotted line) and are relative to the Taf4a housekeeping gene. Data represents mean + SEM of 3 separate qPCR assays with significance determined by Student's t-test (^∗ p < 0.05, ^∗∗ p < 0.01, and ^∗∗∗ p < 0.001).

Figure 7

Gene expression of group β and γ genes, Ncam1 and Mcm2, and the known marker Cd24a correlate with panCK staining in vivo. Representative images of (a) control and (b) choline-deficient, ethionine-supplemented (CDE) mouse liver sections stained for panCK. Expressions of (c) Ncam1, (d) Mcm2, and (e) Cd24a relative to Taf4a in a variety of CDE and control liver tissues are plotted against their corresponding panCK positivity values.