Isolation of LSECs and KCs using SE-1 and CD11b/c yields highly pure cell preparations
An overview of the transcriptomics and proteomics experiments and purity tests of cells used in the experiments is given in Figure 1. LSECs and KCs were purified by immunomagnetic bead cell separation (MACS) of non-parenchymal liver cell (NPC) suspensions generated from collagenase perfused rat liver, then plated for 0.5h (KCs) or 1h (LSECs) and washed with medium before RNA and protein extraction (Figure 1a). For LSEC, we used the SE-1 monoclonal antibody (43, 44) (Table 1), which targets FcgRIIb2 (45) and has been previously tested for MACS-based purification of rat LSECs (43). The isolated cells were >97% LSECs (i.e. fenestrated endothelial cells), as examined by scanning electron microscopy (SEM), and 96.6% were stabilin-2 positive by immune staining (Figure 1b-d). The few contaminating cells were KCs and stellate cells. A monoclonal antibody to CD11b/c (Table 1), targeting complement receptor 3 (CR3) was used to purify KCs. This yielded 94.9% KCs - contaminating cells were 3.1% LSECs and 1.6% stellate cells (Figure 1b).
Quantitative expression of marker genes used for cross validation of the transcriptomics and proteomics data are listed in Figure 1e. Consistent with SEM and immunocytochemistry analysis of MACS-isolated cells, expression of macrophage and stellate cell markers were low in the LSEC transcriptomes and proteomes, whereas expression of LSEC and stellate cell markers were low in the KC transcriptomes and proteomes.
To check the hepatic intralobular distribution of cells expressing SE-1 (i.e. FcγRIIb2), and CD11b/c, frozen rat liver sections were stained with the same antibodies used for MACS-isolation of cells (Figure 2). The SE-1 antibody showed a strict sinusoidal staining pattern, colocalizing with the LSEC marker stabilin-2 (24, 46, 47) in all sinusoids (Figure 2a, b). Most CD11b/c positive cells were located in the periportal region and showed a different staining pattern than stabilin-2 (Figure 2c).
Global information generated from omics data profiling
In the RNA-seq experiment 10,306 genome features were deemed expressed and included in the subsequent analyses, while in the label-free proteomics experiment 2996 non-redundant protein IDs were deemed expressed and included in the further analyses. Principal component analysis (Figure 3a) segregated the LSEC and KC samples into disparate clusters coherent with the distinct biology of the cells.
Figure 3b illustrates the total number of gene products identified with the respective techniques, and their overlap, in the LSEC and KC groups. The proteome covered 26-27% of the transcriptome. Notably, most proteins (90.8-91.5%) identified in the proteome had valid corresponding mRNA in the transcriptome. To evaluate the coherence between the transcriptome and proteome, we calculated the global Pearson correlation coefficient r using the expression data between the omics datasets for each cell type. The global correlation r value was 0.57 for LSECs, and 0.63 for KCs (Figure 3c) which are in the upper end of the previously reported range of 0.4-0.6 (41, 48) supporting the reliability of the data.
Differentially expressed gene products are key to understanding phenotypic and functional variation between cell types. The results of the differential expression analyses of the RNA-seq data, and the proteomics data are summarized in Figure 3d. We identified 2109 gene products in the transcriptome (20.5%) as significantly differentially expressed (with cutoff of FDR (false discovery rate) ≤ 0.05 and |log2 fold change| ≥ 1) in LSECs and KCs. Similarly, in the proteome, 886 proteins (~30%) were significantly differently expressed in the two cells (with cutoff of FDR ≤ 0.05 and |log2 fold change| ≥1). Despite differences in percentage of differentially expressed gene products in the RNA-seq and proteomics experiments, the log2 fold changes for the unique gene products identified in both datasets showed high correlation (r=0.74 [95% CI: 0.72-0.75]) (Figure 3e), suggesting good congruence between the two techniques.
LSECs and KCs show enrichment of terms reflecting their ontogeny
We used ranked gene lists based on expression level from the RNA-seq experiment as input for gene set enrichment analysis (GSEA) (49, 50) to identify the intrinsic functional characteristics of LSECs and KCs. GSEA showed enrichment of 268 biological processes in LSECs and 121 biological processes in KCs with FDR q-value ≤ 0.05 corresponding to Gene Ontology (GO) terms (51, 52) in Molecular Signatures Database (49, 53) that concur with the generic role of these cells (Additional file 1; Figure 4). Like other endothelial cells, LSECs are involved in development, morphogenesis, patterning and maintenance of blood vessels, and displayed enrichment of gene sets associated with response to vascular endothelial growth factor and regulation of WNT, BMP, and TGFβ signalling pathways. KCs, being macrophages, displayed enrichment of terms related to adaptive and innate immune responses.
Expression of genes associated with endocytic function, cytoskeleton organization, and positive regulators of endocytosis, such as 1-phosphatidylinositol-4-phosphate 5-kinase (Pip5klc), phospholipase D1/2 (Pld2), integrin subunit beta1 (Itgb1), GTPase Hras, clathrin adaptor protein (Dab2), caveolin1 (Cav1), and E3 ligase NEDD4 (Nedd4) were higher in LSECs than in KCs (Additional file 2). Moreover, LSECs showed higher expression of transport-related proteins such as EH domain-containing protein 3 (Ehd3), which is suggested to be involved in transport of stabilin-1-positive vesicles (39), adaptor-related protein complex 1 beta 1 subunit (Ap1b1), and sorting nexin (Snx) 8 and 33, which are associated with vesicular transport (Additional file 2). Interestingly, RNA-seq of LSECs revealed high expression of genes coding for connective tissue components such as Sparc, Col4a1, Col4a2, Egfl7, and Mfge8, indicating a significant role of these cells in extracellular matrix maintenance and remodeling of liver (Additional file 2). Transcription factor Gata4, which is essential for LSEC differentiation (39, 47) was specifically expressed in the LSEC transcriptome (Additional file 2).
Most gene products involved in KC immune functions are also expressed in LSECs
Genes associated with the term immune system processes (GO:0002376) include 2645 annotated objects in the rat genome database (December 13, 2019). Of these, we found 1466 expressed genes in the RNA-seq data, and 554 expressed genes in the label-free proteomics experiments that were associated with the term (Figure 5a; Additional file 3). Both cells expressed numerous immune genes - the majority of which were expressed at low density but more abundant in KCs compared to LSECs. To ascertain the immunological role of expressed genes we performed functional enrichment analysis (DAVID 6.8 (54, 55)) of genes with expression values ³10 RPKM (reads per kilobase of exon model per million mapped reads (56)) separately in the LSEC and KC RNA-seq datasets. The threshold 10 RPKM was set to increase the confidence of the results. The immune terms that were significantly enriched (FDR ≤ 0.05) in KC and LSEC transcriptomes were similar, and each term contained almost similar number of expressed genes in the two cells (Figure 5b).
Both cell types show high expression of scavenger receptors and immune lectins
LSECs and KCs express a variety of SRs, C-type lectins, and TLRs (16, 17, 27-29). We found that both cells expressed many SRs and immune lectin gene products at high densities, of which some were cell type specific (Figure 6a; Additional file 4), providing the capacity of rapid sensing and clearance of various danger molecules. Among these were the macrophage mannose receptor (Mrc1) and macrophage SR-A1 (Msr1) which were abundantly expressed both in the LSEC and KC transcriptomes and proteomes (Figure 6a) and confirmed by immune cytochemistry (Figure 6c). The high-density lipoprotein receptor SR-B1 (Scarb1) was also equally expressed in the rat LSEC and KC transcriptomes, but at low density, and were not identified in the cell proteomes. However, immune labelling experiments validated SR-B1 protein expression in both LSECs and KCs (Figure 6c), in accordance with (57). Of note, CD36, a reliable LSEC marker in human liver (58) was evidently expressed in rat KCs but was very low in rat LSECs (Figure 6a). Same receptor was previously reported to be absent from Sprague Dawley rat LSECs in western blot and immune fluorescence experiments (59).
Stabilin-1 (Stab1) and stabilin-2 (Stab2) were expressed at much higher densities in the LSECs than in KCs (Figure 6a-b). Immune labeling of NPCs (Figure 6c) and frozen rat liver sections (Figure 2) for stabilin-2 confirmed LSEC specific expression and a typical LSEC distribution pattern in all hepatic zones of this protein, in accordance with (60) supporting the use of stabilin-2 as a specific pan-LSEC marker. Furthermore, rat LSECs showed high mRNA and protein expression of Clec4g (LSECtin) and Clec4m (DC-SIGNR) (Figure 6a, b), as was also reported in a study of human LSECs (61), where Clec4g was used as a specific LSEC marker in liver single cell transcriptome studies (62).
Some of the receptors reported in the literature to discriminate KCs from other liver cells, were also expressed in the LSEC transcriptome. These included Marco, Cd5l, Clec4f, Cd163, lgals3, and Cd68 (Figure 6a, b). However, their transcript level in KCs were significantly higher compared to LSECs, and their abundance in the LSEC proteome was low. Immune labeling of NPCs for CD163 (not shown) and CD68 showed staining of KCs only (Figure 6c) and labeling of rat liver sections for CD68 together with the LSEC marker stabilin-2 showed a staining pattern of CD68 that is typical for KCs (Additional file 5), supporting the proteomic results.
Several TLRs were detected in LSECs and KCs transcriptomes (Additional file 3). The abundance of Tlr4, 5, 6, 7, 8, 10, 11, and 12 mRNA was significantly higher (FDR≤ 0.05) in KCs, whereas Tlr2, 3, and 13 were not significantly different. The only TLR identified by proteomics at steady state was TLR3 which was identified in both cells.
Immune regulatory factors expressed by LSECs and KCs
When reviewing genes annotated with cytokine receptor binding (GO:0005126), cytokine receptor activity (GO:0004896), complement activation (GO:0006956), and complement receptor activity (GO:0004875), we identified 209 genes in the transcriptome (out of 551 objects associated under the terms), and 54 proteins in the proteome (Additional file 6). Low protein identification may be due to the fact that these genes are normally expressed at low levels in non-stimulated cells from healthy animals (as analyzed in this study), and many gene products associated with the terms represent secreted proteins, mostly found extracellularly. Thus, the bulk of gene products affiliated with the terms were only detected in the transcriptome, and at low level. Many were also differently expressed in the LSEC and KC transcriptomes (Figure 7a).
Figure 7b-c reflects the complex cytokine milieu of the sinusoids. LSECs showed significantly higher expression of the cytokine receptors Tgfbr3, Il6st, Osmr, Il1r1 and Lifr (Figure 7b) enabling them to sense and respond to the cytokines Tgfb3, Osm, Il1b and Lif in paracrine and autocrine manners. Tgfb3, Osm, and Il18 were more abundantly expressed by KCs (Figure 7c). LSECs also expressed high levels of Ackr3 (Figure 7b) which is involved in scavenging and degradation of chemokines, thus regulating their levels in the hepatic sinusoids. KCs showed significantly higher expression of the cytokine receptors Il6r and Csf3r, and chemokine receptor Cxcr4, (Figure 7b) which allow KCs to respond to Ccl24, Cxcl12, Ccl2, Ccl6, and Ccl7 in an autocrine or paracrine manner (Figure 7c).
The expression of colony stimulating factor receptors Csf1r, Csf2ra and Csf3r were also higher in KCs (Figure 7b). Of these, Csf1r and Csf2ra were detected by proteomics, being significantly higher in KCs (Additional file 6). Interaction of colony stimulating factor receptors with their ligands, e.g. Csf1 and Csf2 which were abundantly expressed in LSECs (Figure 7c), affects KC maturation (63), underlining the importance of LSECs for proper KC function.
The complement system is an important part of the innate immune system. Hepatocytes are major producers of complement proteins, whereas NPCs regulate complement activation (42). Gene products representing complement receptors (Figure 7d), and triggers of complement activation (C1qa, C1qb, C1qc; Figure 7e) were significantly more abundant in the KC transcriptome and proteome datasets, whereas the expression of the C1 inhibitors C1qbp and Serping1 was similar in the two cells (detected only in the transcriptome; Figure 7e).
LSECs express the machinery needed for antigen presentation and lymphocyte activation
A series of studies in mouse models suggest that LSEC cross-presentation of exogenous soluble antigens to naïve T cells is central to maintaining liver immune tolerance (reviewed in (1)). However, there are some controversies (37). As LSECs rapidly dedifferentiate in culture (39, 40) and cells are cultured for several days in lymphocyte stimulation experiments, the in vivo contribution of LSECs in adaptive immunity may be difficult to extrapolate from in vitro experiments. There may also be species differences. We therefore investigated the basal expression of gene products associated with antigen processing and presentation (GO:0019882), and lymphocyte co-stimulation (GO:0031294) in rat LSECs and KCs (Additional file 7). The expression of tap-transporters, immunoproteases, and lysosomal enzymes involved in processing and intracellular traffic of antigens, were similar in the transcriptomes and proteomes of both cells except for Ctse (cathepsin E) and Ctss (cathepsin S) which were significantly more abundant in the KCs (Figure 8a-b). Expression of MHC class II genes was detected in both cells, but significantly higher in KCs (Figure 8c). Concerning co-stimulatory molecules, LSECs expressed significantly higher levels of gene products involved in activation of T-cells (Cav-1, Dpp4, Cd40, Cd274, Cd320, and Efnb1), while KCs showed an abundance of gene products from the B7/CD28 superfamily (Cd80, Cd86, Btla, Icoslg) known to inhibit T-cells (Figure 8d). Btla, Icoslg, and Cd4 were expressed in both cells, but significantly higher in KCs. BTLA (64), and CD4 (65) are also reported in human LSECs.
A minor subset of rat LSECs expresses the pan leukocyte marker CD45
CD45 is reported to be widely expressed in rat LSECs, with high expression in periportally located LSECs, and low expression in mid-zonal LSECs (38, 66). We here report low expression of CD45 in the LSEC transcriptome, and an even lower expression in the LSEC proteome compared to KCs (Figure 9a). In order to explore this further, we did flow cytometry of NPCs, and CD45 and stabilin-2 double immune labelling of rat liver sections (Additional file 8). We did not observe a clear co-localization of CD45 with the LSEC marker stabilin-2 in the sinusoids, suggesting either absence or low expression of CD45 in rat LSECs in general or expression in a small subpopulation of these cells. We then performed multicolor flow cytometry (Figure 9b-f) of rat liver NPCs labeled with antibodies to CD45, SE-1/FcgRIIb2 (specific LSEC marker), and CD31 (pan endothelial cell marker; Additional file 8). NPCs from the 25-45% Percoll gradient interface were used instead of SE-1-MACS-isolated LSECs to eliminate any selection bias. Using strict gating (Additional file 8), we found that 4.0% (±1.06, n=4) of small, low complex, live-gated SE-1+ cells were CD31+CD45+ (Figure 9g), suggesting expression of CD45 in a small subpopulation of LSECs.
LSECs from normal liver have been reported to not express CD31on the cell surface (67) but in our flow cytometry experiments (Fig.9g) this marker was shown to be expressed in 97.4% (±1.80, n=4) of SE-1 positive cells. Immune staining of liver sections showed positive staining in all vasculature, albeit weaker in LSECs than in other endothelia (Additional file 8).