*3.4. In Vitro Activated MFB Share Key Characteristics of In Vivo MFB*

Our scRNASeq analyses revealed a striking heterogeneity of MFB in vivo compared to resting HSCs, indicating a functional diversity of these cells during fibrogenesis in vivo. This prompted us to investigate to which extent in vitro activated MFB would reflect this heterogeneity as well. Ultrapure FACS-sorted mouse HSCs were therefore cultivated for up to 9 days on uncoated plastic dishes, the standard model of HSCs to MFB trans-differentiation in vitro [14]. HSCs and MFB were harvested at baseline, day 1, day 3, and days 7 and 9 for scRNASeq analysis (Figure 4A). On the one hand, scRNASeq analysis revealed that in vitro activated MFB clustered dependent on the time of cultivation and could be differentiated into early (day 1), intermediate (day 3), and late (day 7 and day 9) MFB (Figure 4B,C). On the other hand, in vitro activation induced a remarkable MFB heterogeneity over time, which became apparent at the intermediate (day 3) and the late (days 7 and 9) time-point. A more granular analysis of the scRNASeq data generated from the culture-activated HSCs/MFB revealed sub-clusters at all time-points (Figure 4C), for which the marker gene expression patterns varied (Figure 4D and Supplementary Table S2).

**Figure 4.** Single-cell RNA sequencing analysis of in vitro activated myofibroblasts. (**A**) Schematic overview of the experimental setup. (**B**) t-SNE plot showing all in vitro activated MFB and resting HSCs dependent on their origin. (**C**) t-SNE plot showing all in vitro activated MFB and resting HSCs dependent on cluster. (**D**) Violin plots showing the relative expression of selected marker genes for each cluster. *n* = 4 with an average of 1000 cells per condition.

We then particularly looked at some markers that we had identified from the in vivo data sets (compare to Figures 2 and 3). The expression of α-SMA (*Acta2*) and *S100a6* was found to be upregulated from early to late MFB in vitro, indicating the relevance of *S100a6* expression as a marker of activated MFB and revealing a high concordance with the in vivo data. For collagens, we observed a clear upregulation of *Col1a2* or *Col5a2* over time during MFB maturation in vitro (Figure 4D). On the contrary, chemokine expression tended to decrease in sequential samples during culture-induced MFB activation. The production of the chemokines *Ccl2* and *Cxcl1* was only found in early MFB, while late MFB populations from day 3 to day 9 did not show any expression of these chemokines. *Cxcl12* displayed the highest expression in resting HSCs and early MFB (Figure 4D). While in vivo, MFB from fibrotic livers consisting of heterogeneous subsets, in which either collagens or chemokines were highly expressed, in vitro activated MFB only expressed these marker genes in a time-dependent manner and not simultaneously, indicating in vitro activation reflected important aspects of MFB biology only transiently during cultivation.

### **4. Discussion**

HSCs trans-differentiation to MFB is a key event in the progression from chronic liver injury to liver fibrosis, which occurs as a consequence of chronic hepatic inflammation e.g., following alcoholic or non-alcoholic steatohepatitis. Liver fibrosis is also considered as a cornerstone event in the progression toward liver cirrhosis or hepatocellular carcinoma [15]. Given the wide range of functional contributions of HSCs and MFB for liver physiology and for fibrogenesis [4], a better understanding of the heterogeneity and subpopulations of HSC and MFB may help to identify novel therapeutic targets to treat liver fibrosis. The upregulation of collagens and α-SMA in MFB is a well-established finding [16] and was confirmed by our single-cell based data for MFB in vivo as well as in vitro. While almost all was MFB upregulated collagen production, only half of the cells expressed chemokines. This

was even more striking in vitro, in which chemokine expression appeared early after HSC activation and was down-regulated during later maturation. Activated MFB secrete chemokines capable of recruiting myeloid cells from the circulation, such as e.g., neutrophils via CXCL1 and monocytes via CCL2, which has been linked to exacerbated hepatic inflammation [3]. In fact, HSCs had been reported to induce monocyte chemotaxis via CCL2 upon recognizing danger signals via toll-like receptor 4 [17]. Our scRNASeq data support that this is a feature of HSCs and early activated MFB in vitro, while the existence of MFB sub-populations in fibrotic livers in vivo allows to maintain the production of inflammatory chemokines and cytokines during fibrogenesis. On the other hand, as this feature was missing in our in vitro MFB dataset for the late MFB, we conclude that the use of plastic adherence cultivated MFB may not be a useful tool for analyzing chemokine production by MFB when the cells are being studied after 7 days of culture. However, it needs to be noted that we did not further evaluate other stimuli or cultivation methods as TGF-β or platelet derived growth factor subunit beta (PDGF-β) for the in vitro culture, which could potentially give different results regarding collagen and chemokine expression.

The scRNASeq data sets generated in this work may help to guide future studies on the functional relevance and interactions of HSC/MFB subsets. For instance, we identified S100A6 as a novel marker of activated MFB, following liver fibrosis, which could be confirmed in vitro. The functional involvement of S100A6 for HSCs/MFB signaling, however, requires further evaluation. Data from mouse models of fibrosis indicated that recombinant S100A6 would aggravate fibrogenesis via inducing HSCs proliferation [18]. While S100A6 was consistently found across HSCs/MFB in vivo and in vitro, some MFB properties appear restricted to distinct clusters and/or differentiation states. Cluster MFB II showed characteristics of both leukocytes by expressing CD74, *C3* or SLPI but also expressed various collagens. These cells might represent macrophages that have trans-differentiated into myofibroblasts, which has been described for renal fibrosis previously [11] and could potentially explain parts of the immunologic properties that have often been assigned to HSCs and/or trans-differentiated MFB [4].

Importantly, more work is needed to get a better spatial resolution on the MFB populations in fibrosis. The cluster MFB IV showed some features that had been previously reported for portal fibroblasts [19]. It will be important to confirm the scRNASeq data in other models of liver fibrosis, such as bile duct ligation leading to a preferential expansion of the portal fibroblasts, and to define the exact localization of the MFB sub-clusters in fibrotic livers in vivo. Last, but not least, it will be important to identify HSC and MFB populations in human liver. scRNASeq data from healthy human liver already indicated the existence of HSC clusters [20], and similar analyses from cirrhotic human livers are currently ongoing.

Taken together, our scRNASeq analyses from healthy and fibrotic mouse livers demonstrated a yet unrecognized heterogeneity of HSCs and MFB in vivo, suggesting a concerted interplay of functionally diverse MFB subsets during liver fibrogenesis.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/8/5/503/s1, Table S1: Mean gene expression of HSC and MFB I-IV, Table S2: Mean gene expression of HSC I-V and early, intermediate and late MFB.

**Author Contributions:** F.T. guided the research. O.K., T.P.R., and R.W. designed and performed the experiments. O.K. and J.H. performed the RNA sequencing analysis. F.T., R.W., and O.K. wrote the manuscript. All authors reviewed and approved the manuscript.

**Funding:** This work was supported by the German Research Foundation (DFG; Ta434/5-1 and SFB/TRR57). The study sponsor had no role in the study design or in the collection, analysis, or interpretation of data.

**Conflicts of Interest:** The authors declare no conflict of interest.
