Characterization of the proteome and metabolome of human liver sinusoidal endothelial-like cells derived from induced pluripotent stem cells

The liver is a complex organ composed of several cell types organized hierarchically. Among these, liver sinusoidal endothelial cells (LSECs) are specialized vascular cells known to interact with hepatocytes and hepatic stellate cells (HSCs), and to be involved in the regulation of important hepatic processes in healthy and pathological situations. Protocols for the differentiation of LSECs from human induced pluripotent stem cells, hiPSCs, have been proposed and in-depth analysis by transcriptomic profiling of those cells has been performed. In the present work, an extended analysis of those cells in terms of proteome and metabolome has been implemented. The proteomic analysis confirmed the expression of important endothelial markers and pathways. Among them, the expression of patterns typical of LSECs such as PECAM1, VWF, LYVE1, STAB1 (endothelial markers) , CDH13, CDH5, CLDN5, ICAM1, MCAM-CD146, ICAM2, ESAM (endothelial cytoskeleton) , NOSTRIN, NOS3 (Nitric Oxide endothelial ROS), ESM1, ENG, MMRN2, THBS1, ANGPT2 (angiogenesis), CD93, MRC1 (mannose receptor), CLEC14A (C-type lectin), CD40 (antigen), and ERG (transcription factor) was highlighted . Besides, the pathway analysis revealed the enrichment of the endocytosis, Toll-like receptor, Nod-like receptor, Wnt, Apelin, VEGF, cGMP-PCK, and PPAR related signaling pathways. Other important pathways such as vasopressin regulated water reabsorption, fluid shear stress, relaxin signaling, and renin secretion were also highlighted. At confluence, the metabolome profile appeared consistent with quiescent endothelial cell patterns. The integration of both proteome and metabolome datasets revealed a switch from fatty acid synthesis in undifferentiated hiPSCs to a fatty oxidation in LSECs and activation of the pentose phosphate pathway and polyamine metabolism in hiPSCs-derived LSECs. In conclusion, the comparison between the signature of LSECs differentiated following the protocol described in this work, and data found in the literature confirmed the particular relevance of these cells for future in vitro applications.


Introduction
Liver sinusoidal endothelial cells (LSECs) are the most abundant nonparenchymal cells in the liver, form a very specialized microvasculature, and play an important role in hepatic functions and diseases (De Smedt et al., 2021;Poisson et al., 2017).Those cells possess morphological features specific to the liver and differ from vascular endothelial cells by the lack of basal membrane and by the presence of large pores called fenestrae (Poisson et al., 2017).The fenestrae, clustered in sieve plates, allows bidirectional mass transfer between blood (mixture of arterial and venous content) and hepatocytes (transforming unit, Poisson et al., 2017;Gage et al., 2020).
LSECs also exhibit high endocytotic activity for macromolecular waste products, due to the presence of a high density of endocytic vesicles (Braet andWisse 2002, Poisson et al., 2017).Among other functions, LSECs contribute to the regulation of sinusoidal flow, to the maintenance of hepatic stellate cell quiescence, to liver regeneration, and are involved in hepatic complications such as fibrosis, cirrhosis, and hepatocellular carcinoma (Braet andWisse 2002, Poisson et al., 2017).
The lack of complex in vitro models including mature and functional human Liver Sinusoidal Endothelial Cells (LSECS) remains one of the bottleneck problems for the pharmaceutical industry to provide relatively low-cost toxicological data and predictive information on the effects of xenobiotics in complex hepatic models.In that regard, promising alternatives such as the differentiation and maturation of human-induced pluripotent stem cells (hiPSCs) have been developed.As such, several investigations have led to the differentiation of hepatocyte-like cells (HLCs) from hiPSCs (Si Tayeb et al., 2010, Danoy et al., 2020, Kido et al., 2015).In parallel, hiPSCs have also been used to provide a reliable source for other hepatic cells such as HSCs (Coll et al., 2018) and LSECs (Koui et al., 2017).In this frame, hiPSCs have been differentiated toward an LSECs phenotype in which immunostaining coupled to Western blotting confirmed expression of important LSECs markers such as Stabilin-2 (Koui et al., 2017).
Transcriptome analysis with the nano Cap Analysis Gene Expression (nanoCAGE) method was then performed to further confirm the overall vascular commitment of the cells (upregulation of the APLN, LYVE1, VWF, ESAM, and ANGPT2 genes).The analysis of promoter motif activities highlighted several transcription factors (TFs) of interest during the LSECs differentiation from iPSCs (IRF2, ERG, MEIS2, SPI1, IRF7, WRNIP1, HIC2, NFIX_NFIB, BATF, and PATZ1).Based on this investigation, a regulatory network involving the relevant TFs, their target genes as well as their related signaling pathways was proposed (Danoy et al., 2020).In the present paper, to extend the characterization of the hiPSCs-derived LSECs produced by the differentiation protocol of Koui et al., 2017, their proteome and metabolome profiles were investigated.
The panel of signatures and pathways highlighted by those analysis and their integration were then discussed with datasets from both in vivo and in vitro studies.

hiPSCs culture and differentiation protocol
The hiPSCs used in this study (454E2) were provided by the cell bank of the Riken BioResource Research Center (RIKEN BRC).454E2 hiPSCs were cultured and differentiated following the protocol previously described by Koui et al., (2017).After differentiation, the cells were harvested and stored in liquid nitrogen.When required, the needed number of cells was seeded on plates (1.5 x 10 4 cells/cm 2 ) coated with fibronectin (20 µg/mL, Life Technologies) for 1 hour at 37°C, for further culture and proliferation for 9 days.

Metabolomic analysis
The global metabolomic profiling was performed with culture media (basal & used in culture, culture medium) using gas chromatography coupled to quadrupole mass spectrometry (GC-MS).Before the GC-MS analysis, an extraction step was performed on the culture media, according to the protocol previously described in (Jellali et al., 2018).Samples were analyzed by Agilent 7890B gas chromatograph coupled with an Agilent 5977A mass spectrometer.Separation was achieved on a Rxi-5SilMS column from Restek (30 m with a 10 m Integra-Guard column -ref . The data files obtained were analyzed using the AMDIS software (http://chemdata.nist.gov/mass-spc/amdis/).Peak areas were determined using Masshunter Quantitative Analysis (Agilent) and normalized to ribitol.Metabolite contents are expressed in arbitrary units.The complete protocol is detailed in Supp File 1.

Proteomic analysis
The protocol for the preparation of proteins preparation protocol has been previously described (Danoy et al., 2020, details in Supp. File 1).Briefly, the protein samples were digested using trypsin, and the peptides were separated into 5 fractions.
Liquid chromatography and mass spectrometry analyses were performed on an U3000 RSLC nanoflow-HPLC system coupled to an Orbitrap fusion MS analysis (Thermo Fisher Scientific).The Maxquant software was used to identify and quantify proteins.
The database used was a concatenation of human sequences from the Uniprot-Swissprot database (Uniprot, release 2018-06) and an incremented list of contaminants.False discovery rate (FDR) was kept below 1% on both peptides and proteins.

Data statistical analysis
The multivariate statistical analysis of proteomic and metabolomic data were performed using XLSTAT software and MetaboAnalyst, a web-based platform for metabolomics data analysis (Chong et al., 2018).Supervised Partial Least Squares-Discriminant Analysis (PLS-DA) was carried out to discover significant variations between the groups.The quality of the PLS-DA model was evaluated by the R 2 Y (fitting degree) and Q 2 (prediction parameter) values.The variable importance for the projection (VIP) parameter was used to select variables that had the most significant contribution in discriminating between both groups.VIP is a weighted sum of squares of the PLS weight that indicates the importance of each variable to the whole model.
In parallel, univariate statistics were performed with Student's t-test, and P-value less than 0.05 was considered statistically significant.Variables with significant changes in the groups (P-value < 0.05 and VIP > 1) were selected as discriminating variables.
Finally, MetaboAnalyst and Idep9.1/KEGG were used for pathway enrichment analysis of metabolomic and proteomic data, respectively.

Cell morphology
After differentiation, LSECs were plated and amplified until they reached the confluence.The typical morphologies observed are presented in Fig. 1A-B.Notably, the cells displayed a comparable morphology to their primary counterpart, cultured in similar conditions (Fig. 1C).They formed a confluent, cobblestone-like, monolayer with large nuclei typical of endothelial morphology.

Characterization of the proteome profile of hiPSCs-derived LSECs
The proteome analysis of undifferentiated hiPSCs vs hiPSCs-derived LSECs was performed on 3187 positively identified proteins (Supp.File 2) to assess the changes in protein expression through the differentiation.The comparison led to the identification of 2537 proteins differentially expressed with a p_value below 0.05.
The multivariate analysis allowed to separate hiPSCs-derived LSECs and hiPSCs as shown in Fig. 2A-B.The top 50 proteins differentially expressed in the PLS analysis are shown in the heatmap of Fig. 2C.Among the proteins overexpressed in hiPSCs, FASN (fatty acid synthetase), HSP90AB1, FKBP4 (protein interaction at the estrogen receptor), PSAT1, PHGDH (Amino Acid metabolism), and CEBPZ were identified as discriminant proteins.hiPSCs-derived LSECs were characterized as expressing high levels of CPT1A, HADHA (involved in the fatty acid beta-oxidation), OGDH (TCA), ITGB1, MYH9, ACTN4, FLNB, VIM, THBS1 (tight junction, focal adhesion, regulation of actin), CD93 (receptor for mannose-binding leptin) which confirms the changes in metabolism occurring during the differentiation.
To extract pathways specific to the hiPSCs-derived LSECs, proteins with a fold change above 1.2 / below 0.8 and with the P-value < 0.05 (579 proteins) and all proteins with P-value < 0.05 (2537 respectively) were input in the KEGG pathway database.The complete list of proteins used, and pathways extracted is given in Supp.
The extracted pathways were further arranged using iDEP.91.In that regard, a heatmap of the most discriminating biological processes extracted by the Gene Set Enrichment Analysis which characterized the profiles of hiPSCs and the hiPSCsderived LSECs was created (Fig. 3A).The hierarchical clustering of the important biological processes enriched by the proteins, extracted by Desq2 analysis was also created (Fig. 3B, Supp.File 2 for the list of genes in each pathway).It was found that the profile of hiPSCs was characterized by biological processes such as RNA process, DNA process, and cell cycle while the hiPSCs-derived LSECs' profile was characterized by vesicle transport, sprouting angiogenesis, IL12 production, calciumrelated signaling, and endo/exocytosis processes.
Among those metabolites, 26 were found to be differentially present in both media with a P-value below 0.05 (Fig. 4C).Pathway analysis on those revealed phenomena such as enrichment of the galactose metabolism, glucose-alanine cycle, the fructosemannose degradation, Warburg effect, gluconeogenesis.Various amino acid metabolisms, glycolysis, and pathways related to the metabolism of TCA were also highlighted (Fig. 5A-B, Supp File 3).Enrichment pathway analysis, using KEGG as a database, highlighted the valine-leucine isoleucine biosynthesis, Aminoacyl-tRNA biosynthesis, and Galactose metabolism as the top 3 pathways (Supp File 3).

Discussion
In the present study, the proteomic and metabolic profiles of hiPSCs-derived LSECs were extracted to evaluate the quality of the differentiated cellular model.Those results complete the histologic and transcriptomics analysis previously performed (Danoy et al., 2020).In this previous study, hiPSCs-derived LSECs were shown to exhibit typical patterns of LSECs as compared to their primary counterpart.In addition, the cells were also found to be Stabilin-2 + , Stabilin-1 + , CD144 + , and CD31 + .In Table 2, the signature observed for the proteomic and metabolomic analysis of hiPSCs-  et al., 2010;Poisson et al., 2017) or CD31 + , LYVE-1 + , L-SIGN + , Stabilin-1 + , CD34 -, PROX-1 -cells (Lalor et al., 2006, Poisson et al., 2017).In hiPSCs-derived LSECs, the proteome signature was characterized as CD34 -, CD45  , 2004).Thus, the fact that hiPSCs-derived LSECs co-express PECAM and ICAM1 illustrates that fenestration is probably not achieved in the current cellular model.This is in agreement with previously published results (Koui et al., 2017), and is also widely observed in models of LSECs cultured as a 2D monolayer (Lalor et al., 2006).Finally, a brief comparison between the proteomic signatures from the literature (HUVECs, as a positive control, and fibroblasts, as a negative control) was performed to further confirm the quality of the cellular model.In HUVECs, LSECs markers such as CD36, LYVE1, and STAB2 were not expressed while NRCAM, F11R, CEACAM21, CEACAM19, CADM4, CADM3, CADM1, JAM2, CDON, SELE were found to be expressed, but not in hiPSCs-derived LSECs (Madugundu et al., 2019).
The metabolic signature of LSECs is usually found to be highly variable depending on their state.In their quiescent state, endothelial cells are characterized by intense fatty acid oxidation and moderate glycolysis (Dumas et al., 2020).As migratory phenomena are induced, a switch to the production of lipids for membrane fluidity is usually observed (Dumas et al., 2020).As proliferation is induced, endothelial cells were characterized by intense glycolysis and glutamine breakdown, moderate fatty acid oxidation, oxidative phosphorylation, and serine biosynthesis (Dumas et al., 2020).In Table 3, the metabolomic profile observed in the presented analysis of hiPSCs-derived LSECs was synthesized and compared to the metabolomic profile of quiescent, migratory, and proliferative endothelial cells found in the literature.In the presented cellular model, the profile was found to be typical of beta-oxidation as shown by the low levels of azelaic acid.Besides, consumption of several TCA substrates, which reflects glycolytic activity was observed while no change in the levels of glutamine and serine was found.High levels of PPP intermediates such as sorbitol, xylulose could also be observed.As the analysis was performed on confluent hiPSCs-derived LSECs, the profile observed would illustrate a quiescent-like phenotype rather than a proliferative behavior.
By integrating both the proteomic and metabolomic profiles of hiPSCs-derived Endocytosis is also known to be involved in the regulation of the transport of lipids which is also one of the major functions of LSECs (Hammoutene et al., 2019).
Furthermore, lipids have also been found to promote LSECs proliferation as well as the maintenance of their differentiation (Hang et al., 2012).The present dataset suggests a switch in the lipid metabolism in the differentiation of hiPSCs into LSECs.
At the proteome level, this could be characterized by the overexpression of FASN in hiPSCS (fatty acid synthetase), and of CPT1A and HADHA (fatty acid beta-oxidation) in hiPSCs-derived LSECs which implies a degradation of fatty acid as this would be the case in mature LSECs.Besides, lipid homeostasis was highlighted via PPAR signaling in the proteome dataset while the fatty acid degradation was illustrated by the consumption of azaelic acid in the metabolome dataset.hiPSCs-derived LSECs also exhibited an activation of the pentose phosphate pathway (PPP) via the measured concentrations of xylulose and sorbitol which is consistent with the phenotype observed in healthy endothelial cells' phenotype (Dumas et al., 2020).Activation of the PPP could further be confirmed via its function toward nucleotide synthesis for angiogenic activity (18 hits and 15 hits in nucleotide sugar metabolism and nucleotide excision repair respectively at the proteome level and production of putrescine, a substrate of polyamine at the metabolome level), antioxidant defense (overexpression of GPX1, GPX2, GSTO1, MGST2, MGST3, GSTK1, GSTM3 in the proteome dataset), and nitric oxide synthesis (overexpression of NOS3 in the proteome dataset) (Bierhansl et al., 2017) which further confirms the viability of the presented cellular model.
Nonetheless, this model still presents important limitations toward the complete reproduction of the in vivo phenotype of LSECs.Indeed, fenestration could not be observed in hiPSCs-derived LSECs (Koui et al., 2017) which is a strong impairment toward the function of the cells in fluid exchange and transport of molecules.
Fenestration is known to be nitric oxide-dependent via the endogenous nitric oxide synthase (eNOS)-soluble guanylate cyclase (sGC)-cGMP pathway (DeLeve et al., 2017) for which the overexpression was found in the proteome dataset.However, one of the main sources of NO in the normal liver is known to be mediated via endothelial nitric oxide synthase (eNOS) activation by shear stress (Shah et al., 1997, Poisson et al., 2017), which is not reproduced in the current experimental setup.Subjection to shear stress is known to induce the KLF2 transcription factor (Gracia-Sancho et al., 2019, Poisson et al., 2017) which stimulates the endothelial upregulation of vasodilating agents such as nitric oxide (NO) (Poisson et al., 2017, Parmar et al., 2006).
Although partial activation of shear stress-related markers (in the KEGG pathway) and vasoconstriction/vasodilation markers (via THSB1) could be confirmed, the lack of direct exposition to shear stress remains a major weakness in the current approach.
Finally, further confirmation of the differentiation protocol is required with different hiPSCs lines, as biological variability has been observed (Volpato et al., 2020;Ortmann et al., 2020).

Conclusions
In this study, the analysis of the proteome and metabolome dataset of hiPSCsderived LSECs was presented to assess the quality of this differentiated cellular model.
The cell signature could be confirmed as characteristic of LSECs, expressing typical vascular and endothelial markers in addition to LSEC specific proteins while important liver and LSEC patterns were also observed.Specifically, vital functions such as lipid metabolism, the pentose phosphate pathway, and the production of polyamine could be confirmed.In terms of future perspectives, the study of the hiPSCs-derived LSECs in conditions reproducing their in vivo microenvironment could help to further characterize the cells via exposition to shear stress and to other liver cell types in coculture.

derived
LSECs was compared to other LSECs signatures found in the literature.In details, LSECs are typically characterized by a specific signature including markers such as VEGFR3 + LSECs, characterization of the cellular model was further performed.Pattern recognition receptors and components of inflammatory response usually expressed in LSECs include endocytosis receptors, Toll-like receptors, and Nod-like receptors(DeLeve et al., 2017).The enrichment of both Toll-like receptors and Nod-like receptors signaling (12 and 33 hits with P-value < 0.05 resp.) was confirmed in the present analysis.Regarding endocytosis, known to be one of the major functions of LSECs in the normal liver, it could be highlighted as illustrated by the KEGG endocytosis pathway (76 hits with P-value <0.05), and the C-type lectin receptor signaling (15 hits, P-value <0.05).Also, the LSECs endocytose is known to involve the scavenger receptor (SR), the mannose receptor, and the FC gamma receptor Ilb2.Notably, the main known SRs are SR-H/Stabilin 1 and SR-H/Stabilin 2(Poisson et al. 2017;DeLeve et al., 2017).The presence of STAB2 was also confirmed by western blot in the current cellular model(Danoy et al., 2020;Koui et al 2017) while mannose was detected in the metabolome as a significantly modulated molecule which further confirms the endocytosis function of the presented cellular model.

Figure 1 :
Figure 1: Morphology of hiPSCs-derived LSECs on Day 1 (A) and upon reaching confluence on Day 7 (B).Morphology of primary human LSECs upon reaching confluence on Day 5 (C).

Figure 3 :
Figure 3: Important biological processes extracted with iDEP.91 in the comparison between hiPSCs-derived LSECs and hiPSCs proteome profile.Corresponding heatmap of the parametric gene set enrichment analysis in which red denotes
(Poisson et al., 2017)se receptors), and NOS3 (nitric oxide endothelial ROS) as shown in Table1.Notably, and VEGFR3 and L-SIGN were not detected in the proteome signature of hiPSCs-derived LSECs and major mature LSECs markers including CD4 and CD32, and immature markers such as CD34, 1F10(Poisson et al., 2017), CD4, CD32, CD34, and 1F10 proteins were not expressed in this cellular model either.Concerning CD31, it can be noted that this adhesion molecule, which is expressed in several endothelial cells, including in LSECs, is known to disappear in the latter with the establishment of fenestration (De Leve et a. + , VE Cadherin + (CDH5), VWF + , VEGFR2 + , and completed by the expression of CD146 (marker common in several types of endothelial cells), CLEC14A (C type lectin),

Table 1 :
Signature of proteins overexpressed in hiPSCs-derived LSECs in the comparison with hiPSCs.Comparison with data found in the literature.

Table 2 :
Selected pathways related to the function of LSECs, and associated protein extracted with a p_value below 0,05 in the present analysis.

Table 3 :
Dumas et al., 2020.etabolic profile of the hiPSCs-derived LSECs analyzed in the present analysis with profiles defined in the literature.Data reproduced from the review ofDumas et al., 2020.

Table 1 :
Signature of proteins overexpressed in hiPSCs-derived LSECs in the comparison with hiPSCs.Comparison with data found in the literature.

Table 2 :
Selected pathways related to the function of LSECs, and associated protein extracted with a p_value below 0,05 in the present analysis.

Table 3 :
Dumas et al., 2020.etabolic profile of the hiPSCs-derived LSECs analyzed in the present analysis with profiles defined in the literature.Data reproduced from the review ofDumas et al., 2020.