Expression of genes involved in drug metabolism differs between perfusable 3D liver tissue and conventional 2D‐cultured hepatocellular carcinoma cells

Tubular 3D liver tissue with enhanced capillary‐like structures branching from a large main channel is potentially useful for drug discovery because the perfusable main channel and capillary‐like structures enable mass transfer into and out from the tissue. Tubular liver tissue is comprised of the hepatocellular carcinoma cell line HepG2, human umbilical vein endothelial cells (HUVECs), and mesenchymal stem cells (MSCs), using a perfusion device functioning as the interface for an external pump. This study aimed to compare the expression of genes involved in drug metabolism between 2D‐cultured hepatocellular carcinoma cells and 3D‐cultured tubular liver tissue. Gene expression profiles of 2D‐cultured cells and tubular liver tissue were compared using RNA sequencing. Multidimensional scaling analysis revealed that culture dimensionality had a more prominent effect on gene expression profiles than perfusion conditions. More specifically, genes involved in drug metabolism such as CYP2D6, CYP2E1, NNMT, and SLC28A1 were slightly upregulated in the 3D cultures, while certain genes such as ALDH1B1, ALDH1A2, and SULT1E1 were downregulated. These results indicate that gene expression profiles are largely influenced by culture dimensionality and are potentially useful to researchers intending to switch from 2D culture to 3D culture of hepatocellular carcinoma or other tissue types.

Tubular 3D liver tissue with enhanced capillary-like structures branching from a large main channel is potentially useful for drug discovery because the perfusable main channel and capillary-like structures enable mass transfer into and out from the tissue. Tubular liver tissue is comprised of the hepatocellular carcinoma cell line HepG2, human umbilical vein endothelial cells (HUVECs), and mesenchymal stem cells (MSCs), using a perfusion device functioning as the interface for an external pump. This study aimed to compare the expression of genes involved in drug metabolism between 2D-cultured hepatocellular carcinoma cells and 3D-cultured tubular liver tissue. Gene expression profiles of 2D-cultured cells and tubular liver tissue were compared using RNA sequencing. Multidimensional scaling analysis revealed that culture dimensionality had a more prominent effect on gene expression profiles than perfusion conditions. More specifically, genes involved in drug metabolism such as CYP2D6, CYP2E1, NNMT, and SLC28A1 were slightly upregulated in the 3D cultures, while certain genes such as ALDH1B1, ALDH1A2, and SULT1E1 were downregulated. These results indicate that gene expression profiles are largely influenced by culture dimensionality and are potentially useful to researchers intending to switch from 2D culture to 3D culture of hepatocellular carcinoma or other tissue types.
Various 3D liver-like cultures comprising hepatocellular carcinoma cells, such as spheroids [1,2], cell-laden hydrogel [3,4], and organoids [5,6], have been developed. Drug discovery studies have increasingly focused on 3D liver-like tissue cultures using cell culture plates owing to their various advantages over conventional 2D cultures of hepatocellular carcinoma cells, including increased cellular functions such as albumin secretion and xenobiotic metabolism and increased sensitivity to hepatotoxic compounds. To compensate for the lack of a perfusable vascular network in 3D liver-like tissues, we previously developed a tubular 3D liver-like culture, called tubular liver tissue, with enhanced capillary-like structures branching from a large main channel [7], upon combining a perfusion device [8,9] with a collagen gel populated with the hepatocellular carcinoma cell line HepG2, human umbilical vein endothelial cells (HUVECs), and mesenchymal stem cells (MSCs). The perfusable main channel and capillary-like structures facilitate the mass transfer of not only oxygen and nutrients but also test substances and their metabolites both into and from the tissue, which can then be sampled, thereby facilitating their applications in drug discovery studies. However, differences in drugmetabolizing gene expression between tubular liver tissue and the conventional 2D-cultured HepG2 remain unknown. Herein, we constructed a tubular liver tissue and used RNA sequencing (RNA-seq) to compare its gene expression profiles under perfused and nonperfused conditions, with that of 2D-cultured HepG2 cells, HUVECs, and MSCs mixed at the same ratio as that used for generating the liver tissue (Fig. 1). We mainly focused on genes involved in drug metabolism, categorized as phase I, II, III, and nuclear receptor genes. Phase I enzymes are involved in oxidation, reduction, and hydrolysis of xenobiotics. Phase II enzymes are involved in conjugation reactions. Phase III enzymes are usually related to transportation of xenobiotics [10]. Since drug metabolism in vivo mainly involves these three phases, research in the field of drug discovery largely targets them. In addition, a set of nuclear receptors are also investigated because it is known that they control hepatic metabolism and hepatotoxicity [11]. Our results will potentially benefit researchers intending to switch from 2D-cultured hepatocellular carcinoma cells or other 3D liver-like tissues to tubular liver tissue.

Materials and methods
Construction of the tubular liver tissue Tubular liver tissue was constructed using a perfusion device, as previously described (Fig. 1A,B) [7]. Briefly, the device was filled with a collagen matrix (IAC-50; Koken In the perfused group, the tubular liver tissue was cultured under perfusion with medium, using a tube pump. Total RNA was extracted from tissue detached from the device and used for RNA-seq. (B) In the nonperfused group, the tubular liver tissue was submerged in the medium and statically cultured. Total RNA was extracted from the tissue detached from the device and subjected to RNA-seq. (C) In the 2D-cultured group, hepatocellular carcinoma cell line HepG2, HUVECs, and MSCs were cultured in cell culture dishes. The cells were dissociated, mixed at the same ratio as tubular liver tissues, and used for RNAseq after total RNA extraction.
Co., Tokyo, Japan) populated with HepG2 cells (3 9 10 7 cells/mL), HUVECs (3 9 10 7 cells/mL), and MSCs (0.5 9 10 7 cells/mL). The ratio of each cell type was determined based on the literature [5,12]. The needle previously set in the device was extracted. Then, HUVECs were infused into the resultant tunnel to form the main channel. Subsequently, we connected the device to an external pump for perfusion (Fig. 1A) and immersed the device in culture medium for nonperfusion (Fig. 1B). The culture medium was composed of 1 : 1 mix of HepG2 and HUVEC media, which were individually used for the culture of cell populations in 2D conditions.

RNA-seq analysis
Total RNA was extracted from 3D-and 2D-cultured perfused and nonperfused cells using NucleoSpin RNA (Macherey Nagel GmbH & Co. KG, Duren, Germany) and sequenced using NovaSeq 6000 (Illumina Inc., San Diego, CA, USA) with biological duplicates. Read data were processed and analyzed using STAR (2.7.1a) [13], RSEM (1.3.1) [14], and edgeR (3.28.1) [15,16] with the hg38 reference genome and gene annotation Ensembl GRCh38. The read count data were normalized using the trimmed mean of M values. Gene ontology (GO) enrichment analysis was performed using the "database for annotation, visualisation and integrated discovery" (DAVID) [17,18] via RDAVIDWebService (1.20.0) [19] and executed on R [20]. The enriched GO terms (adjusted P-value < 0.01) obtained from a category GOTERM_BP_FAT of DAVID were classified into 4 clusters by clustering analysis based on semantic similarity computation [21]. The semantic similarities between all the pairs of the enriched GO terms were calculated using GOSemSim [22]. Then, hierarchical clustering according to the calculated similarities was performed using the hclust function of R.

Results and Discussion
Culture dimensionality had a prominent effect on gene expression To assess overall differences among the three groups, the data were analyzed through multidimensional scaling ( Fig. 2A). As expected, data points were clustered in accordance with the culture conditions in a plane of the first two dimensions, indicating that variations of the samples in each group were small enough to evaluate differences among groups and that these conditions had robust effects on gene expression profiles. Thereafter, pairwise group comparisons were carried out (Fig. 2B); 792 (upregulated: 381, downregulated: 411), 5974 (upregulated: 3437, downregulated: 2537), and 6016 (upregulated: 3598, downregulated: 2418) differentially expressed genes [DEGs; |log 2 FC| ≥ 1 and false discovery rate (FDR) < 0.05] were identified between perfused and nonperfused cells, perfused and 2D-cultured cells, and nonperfused and 2D-cultured cells, respectively, suggesting that culture dimensionality (i.e., tubular liver tissue vs 2D) has a more prominent effect on gene expression than perfusion, even though previous studies have reported that perfusion retains tissue viability and functions. This finding was further confirmed through clustering analysis. As shown in the heat map ( Fig. 2C), the perfused and nonperfused groups were closely clustered, compared to 2D-cultured groups. GO enrichment analysis was also performed to investigate the differences between the perfused and 2D-cultured groups in detail (Tables 1 and 2). The enriched GO terms were classified by clustering analysis based on semantic similarity computation. Consequently, it was revealed that genes involved in extracellular matrix organization, blood circulation, ion transmembrane transport, and vascular formation were enriched in the perfused condition (Table 1), whereas those involved in cell proliferation (Table 2) were enriched in 2D culture condition. This observation indicates that the perfused tubular tissue is physiologically more relevant than the 2D-cultured cells. Drug-metabolizing genes were basically upregulated in the tubular liver tissue We evaluated the expression levels of drug-metabolizing genes classified into phase I, phase II, and       phase III, and nuclear receptor genes (Figs 2B and 3, Table 3-6) as previously described [2,10,11]. For the phase I genes, mean log 2 fold change (FC) values were 0.4 and 0.5 in the perfused and the nonperfused groups (Fig. 3A, Table 3), respectively, indicating that the phase I genes, particularly CYP2D6 and CYP2E1, in both groups were slightly upregulated on an average; CYP2D6 and CYP2E1 encoding cytochrome P450 members are involved in metabolism of 20% and 2% of known drugs, respectively [23]. Some phase I genes were downregulated, including aldehyde dehydrogenases such as ALDH1A2, which encodes a retinal dehydrogenase with retinaldehyde metabolizing potential but only very low activity with acetaldehyde and propanal, and ALDH1B1, which detoxifies alcohol-derived acetaldehyde [24]. The downregulated genes might be related to cell proliferation since the 2D-cultured cells were at a highly proliferative state as revealed by GO enrichment analysis (Table 4). For phase II genes, the trend was more neutral; mean log 2 FC values were 0.0 for both perfused and the nonperfused groups (Fig. 3B, Table 4). These values may have been negatively biased owing to an extremely low log 2 FC value (À9.1) of SULT1E1 which encodes a sulfotransferase catalyzing estradiol and estrone sulfation. Among the upregulated phase II genes, prominent drug-metabolizing genes were identified, including SULT2B1 which catalyzes pregnenolone and dehydroepiandrosterone sulfation, and NNMT which encodes nicotinamide N-methyltransferase involved in the biotransformation of numerous drugs and xenobiotic compounds [24]. Among phase III genes, mean log 2 FC values were 0.5 both in perfused and in nonperfused groups (Fig. 3C, Table 5), wherein SLC28A1 and AQP1 were particularly upregulated. SLC28A1 encodes a sodium-dependent and pyrimidine-selective transporter involved in uridine, cytidine, thymidine, and nucleoside-derived drug transport, whereas AQP1 encodes an aquaporin channel [24]. Among nuclear receptor genes, mean log 2 FC values were 0.9 and 1.0 in the perfused and nonperfused groups, respectively, indicating higher expression levels than those of phase I-III genes (Fig. 3D, Table 6). In particular, VDR encoding the vitamin D receptor and NR1H4 encoding the bile acid receptor, both associated with hepatotoxicity and metabolism [11], were significantly upregulated. Together, drug-metabolizing gene expression profiles were similar between perfused and nonperfused groups. Furthermore, these genes were basically upregulated in the tubular liver compared with 2Dcultured cells, although it should be noted that some genes were downregulated and also that an interaction among different cell types was excluded in the 2D-culture condition for simplicity. It is assumed that the upregulation of drug-metabolizing genes was due to the physiologically relevant culture conditions, such as ECM, cell-cell interactions, and blood flow, of the tubular liver tissue (Fig. 4). The metabolism and hepatotoxicity of many drugs, such as midazolam, bufuralol, acetaminophen, and diclofenac, are assessed more accurately by using static 3D culture conditions (spheroid and cell-laden ECM) rather          than 2D conditions [2,5]. Considering that the perfused tubular liver tissue is more viable (4.6-fold RNA levels) and more feasible for drug injection into the tissue and metabolite collection than the nonperfused tissue (equivalent to static 3D culture), the perfusable tubular liver tissue might be a promising experimental model for drug discovery studies.
Considering the utility of the perfusable tubular liver tissue, the difference with conventional 2D-cultured cells should be considered.

Conclusions
In conclusion, herein we compared the gene expression profiles of three groups: perfused and nonperfused tubular liver tissues and conventional 2D-cultured hepatocellular carcinoma cells. Although all three groups displayed adequate differences among one another to be clearly clustered, differences resulting from culture dimensionality (i.e., tubular liver tissue or 2D-cultured cells) were particularly significant. Furthermore, assessment of drug-metabolizing gene expression revealed that expression patterns differed between tubular liver tissues and 2D-cultured cells, being upregulated in the tubular liver tissue on average. The present results are potentially relevant to researchers using perfusable tubular liver tissues.