Identification of liver cell-type-specific transcriptional portraits under normal physiological condition
In order to characterize the molecular profiles representing the HEPs, HSCs, and ECs under physiological conditions, we utilized three transgenic lines Tg(fabp10a:dsRed), Tg(hand2:EGFP) and Tg(kdrl:Hsa.HRAS-mCherry) which express red (dsRed, mCherry) or green fluorescent proteins (GFP) in the corresponding cell types [14, 20, 21]. Whole livers were dissected from adult zebrafish from each of the transgenic lines used in this study (Fig. 1A). Fluorescent microscopy of liver from the corresponding transgenic lines confirmed the fluorescence observed in the corresponding cell types (Fig. 1B). We prepared cell suspensions and performed FACS according to previously established protocols (See Methods, Supp. Figure 1). The number of RNA-seq reads corresponding to fluorescent reporters specific to each cell-type (Fig. 1B) was strongly enriched in fluorescent-positive samples, which confirmed the purity of FACS isolated samples (Fig. 1C). In order to ascertain the cell-type gene signatures, we performed differential expression comparisons between samples and identified the most enriched genes in each cell type (Fig. 2A, Supp. Table 2). The largest number of cell-specific genes were found in ECs (4553), then in HSCs (380) and in HEPs (126) (Supp. Table 2). These included known cell-specific markers for ECs (sox18 [22], sele [23], flt1 [24]) and HEPs (soat2 [25]) (Fig. 2B). On the other hand, genes related to fatty acid metabolism (fasn [26], fat3b, hmgcra [27], hmgcs1 [28], elovl4a [29]) and cholesterol biosynthesis (cyp51, sc5d, hmgcra, msmo1, nsdhl, hmgcs1, dhcr7) were upregulated in HSCs which are known to contain vitamin A lipid droplets [30] (Supplementary Table 2). Gene ontology (GO) analysis revealed the enrichment of genes related to angiogenesis in ECs, insulin-like growth factor receptor signalling genes and cellular phosphate ion homeostasis in HEPs and lipid transport and metabolism genes in HSCs (Fig. 2C). Taken together, the enrichment of known markers and the relevant GO terms in ECs, HEPs, and HSCs support the identity of the respective cell types.
TAA metabolism is reflected in the transcriptional shift in liver cells
We then sought to determine the transcriptional signatures of early hepatotoxic injury response in each of the three liver cell types. We induced liver injury using TAA at a concentration of 500 mg/kg of body mass. The short term TAA treatment induced mild histological changes with observed inflammation and extracellular lipid droplets (Fig. 1D). We then collected whole livers from TAA-treated Tg(fabp10a:dsRed), TgBAC(hand2:EGFP) and Tg(kdrl:Hsa.HRAS-mCherry) fishes, isolated the corresponding cell types by FACS, and performed RNA-seq.
We evaluated cell-type-specific transcriptional response to TAA activation by looking at the expression of genes related to TAA metabolism and genes activated in response to liver injury and fibrogenesis (Fig. 2D, Supp. Table 3). The increased expression of genes related to cell redox homeostasis such as catalase (cat) [31], cytochromes (cyp2y3, cyp2p6) [32], superoxide dismutase 2 (sod2) [32], glutathione peroxidase 1a (gpx1a) [33] was observed in response to TAA, with the most striking response in ECs. Pro-fibrotic genes [8] including ECM proteins such as collagens (col1a1a, col1a2, col5a2a, col5a1, col6a3), decorin (dcn) as well as metallopeptidase inhibitor 2a (timp2a), integrin alpha V (itgav) and annexin 5b (anxa5b) were specifically upregulated in HSCs, in response to TAA (Fig. 2D).
TAA induces transcriptional reprogramming of hepatic endothelial cells
To provide a global view of the behaviour of correlated gene clusters in three hepatic cell types in response to TAA, we used self-organizing map based tool oposSOM R package [34]. The tool first constructed transcriptional portraits of all the samples, then a second unsupervised reduction step was performed, further reducing dimensionality to overexpression spots representing clusters (A-H, Supp. Table 4) of co-expressed metagenes which are highly expressed in, at minimum, one condition (Fig. 3A, B) [35]. To link overexpression with gene set overrepresentation in a sample- and spot-specific way, we visualized the metagene expression across samples on the heatmap (Fig. 3C) and performed the gene set overrepresentation analysis (Fig. 3D, Supp. Table 5). The gene expression portraits of both control and TAA-treated samples from each of the three cell types revealed that short-term TAA exposure induced strong changes in genome-wide expression landscapes between cell types in physiological state and upon TAA activation (Fig. 1E, F). Interestingly, the most striking changes induced by TAA treatment were observed in ECs (Fig. 1G).
Analysis of the SOM clusters in ECs revealed an increase in expression of genes related to metabolic and redox processes as well as cellular transport (Fig. 3C, D - clusters B and F). We also observed downregulation of genes related to vasculature development as well as activation of immune response in ECs after treatment with TAA (Fig. 3C, D - clusters G and H).
In HEPs, TAA treatment induced an increase in the expression of gene sets associated with regulation of metabolic processes, namely carboxylic acid and hydroxy compound metabolism, as well as intra- and intercellular transport when compared to their control counterparts (Fig. 3C, D - cluster B). In contrast, we observed a decreased expression of gene sets associated with the formation and function of endoplasmic reticulum as well as negative regulation of various growth binding factors (Fig. 3C, D - clusters E and G). We also observed a relative reduction of expression of genes associated with the G2/M cell cycle transition in TAA-treated HEPs (Fig. 3C, D - cluster D).
Modest changes in gene expression were observed in HSCs. Analysis of clusters revealed that upregulated gene sets were associated with extracellular space and structure organization as well as protein hydrolysis (Fig. 3C, D - cluster A), which reflects the known role of HSCs in ECM formation during liver damage response [9]. Conversely, we observed downregulation of genes associated with G2/M cell cycle transition, endoplasmic reticulum, estrogen response and immune activation (Fig. 3C, D - clusters G and H).
Altogether, cell-type-specific transcriptome profile revealed transcriptional response to short term TAA exposure. All of the analyzed cell types were subject to TAA-induced transcriptional shifts, with the highest change observed in ECs. These were hallmarked by decrease of vascular-specific genes and the increase of fatty acid and carbohydrate metabolism genes as well as in immune response-associated genes.
TAA leads to genome-wide changes in chromatin regions enriched in binding sites for transcription factors regulating fatty acid metabolism and angiogenesis
Epigenetics has been acknowledged as an important player in liver fibrosis and regeneration [36–38], with a prospect of the development of epigenetic biomarkers and therapies. To investigate this aspect of liver damage, we ask whether epigenetic changes are involved in the earliest stages of liver fibrosis. To determine whether and to what extent epigenetic landscape in each liver cell type is altered during early stage liver injury, we characterized the changes in chromatin accessibility in HEPs, HSCs, and ECs upon TAA treatment.
We observed that in TAA-treated animals the most significant changes in chromatin state compared to control were observed in ECs, followed by HSCs and HEPs (Fig. 4A, B). ATAC-seq peaks distribution across the genome showed that the highest fraction of peaks (30–40%) was localized in the promoter (+/- 3kb) regions (Fig. 4C, Supp. Table 7). Interestingly, the most significant changes in chromatin accessibility was observed in ECs, with the largest number of upregulated peaks found within the promoters of genes in clusters B (440 peaks) and F (74 peaks) and downregulated peaks in clusters G (120 peaks) and H (113 peaks) (Fig. 5A). The observed changes in chromatin accessibility correlates with changes observed in the transcriptional levels of genes within the corresponding clusters (increase in clusters B and F, and decrease in clusters G and H) (Fig. 4D). On the other hand, modest changes in chromatin accessibility were observed in the other two cell types. In HEPs, the highest change was observed in cluster B (30 up- and 18 downregulated). In HSC, 62 and 7 peaks were upregulated or downregulated in cluster B, respectively and 39 downregulated in cluster H.
To identify potential regulators involved in TAA response in each cell type, we searched for transcription factor (TF) motifs enriched in differentially accessible promoter peaks from SOM cluster genes (Fig. 5B-D, Supp. Table 6). Significant enrichments (p-adjusted < 0.05, Supp. Table 6) were identified predominantly in five tested groups of regions: cluster B upregulated regions in ECs and HSCs, cluster G downregulated regions in ECs and cluster H downregulated regions in ECs and HSCs. In ECs, we observed significant enrichment in motifs of fatty acid metabolism nuclear receptors such as RXR [39], THRB [40], HNF4A [41] and PPARA [39] among peaks upregulated in cluster B. This is in accordance with the result of gene set overrepresentation analysis (Fig. 3D). A drop in chromatin accessibility was observed for ECs peaks located in the promoter of genes from cluster G. TFs motifs identified in this cluster belong to ETS family (ETV2, ERG, SPDEF, ETS1) and Sox family (Sox6, Sox17, Sox3) involved in cell differentiation, migration and proliferation [42–44]. In HSCs, we found enriched motifs of TFs involved in cellular glucose homeostasis such as FOXA3 [45], FOXK1 [46], FOXK2 [47] and cell differentiation such as RARA, TR4, FOXA1, FOXA3 [48]. In cluster H downregulated regions, both in EC and HSC, we also found enriched motifs of ETS family including ETV2, ERG, ELF5, ELF3, ETS1, EHF, SPIB, ELF4. Additionally, in HSCs we found enrichment of ATF4 and Chop motifs, which are known to be involved in response to endoplasmic reticulum stress [49, 50]. Notably, ETS TFs also regulate endothelial function and homeostasis [51]. Altogether, our results show increased chromatin accessibility in the promoter regions of gene clusters associated with fatty acid metabolism, especially in ECs, and decrease of accessibility in clusters related to endothelial homeostasis and inflammatory response.