Successful drug discovery and development is a very challenging and complex process that consists of five stages. The first stage, called the prediscovery stage, covers basic research, which aims to understand the mechanisms of different diseases and finally identify potential target molecules. During the second step, named the drug discovery stage, therapeutic agents, which affect the investigated disease by interacting with target proteins, have been developed. The third stage consists of preclinical research on the most promising therapeutic agents via in vitro and in vivo models. The fourth stage includes three phases of clinical trials, which are performed on humans to evaluate the efficacy and, most importantly, the safety of the therapeutic drug. The fifth stage involves reviewing the results of clinical research by the Food and Drug Administration (FDA) or other regulatory organs and finally making decisions about the approval of the new medicament on the market. This last step also includes postmarket monitoring of potential side effects and adverse events of the released drug (Mohs and Greig, 2017; Kandi and Vadakedath, 2023; Singh et al., 2023). The whole procedure is very expensive and time-consuming because it lasts, on average, 10–15 years (Sun et al., 2022). Moreover, it is worth noting that the success rate of the introduction of the new medicament on the market is relatively low. According to the statistical resources, currently, the total number of new drugs in the R&D pipeline worldwide is 22,825. However, approximately 50 new substances are released annually into the market. For example, in 2023, the FDA approved only 55 new medicaments on the market (Statista, 2025; de la Torre and Albericio, 2024; Shareef et al., 2024).

Before potential therapeutic drugs are tested in human volunteers, preclinical research is absolutely indispensable. In general, during this study, the efficacy and safety of the analyzed therapeutic medicaments were investigated. More precisely, in in vitro and in vivo models, research on the toxicology, pharmacodynamics, and pharmacokinetics of drug candidates has been performed (Honek, 2017). Notably, studies on drug toxicity, especially drug hepatotoxicity, are extremely important. The metabolism of drugs and xenobiotics occurs in the liver. When the tested drug (or its metabolites) causes hepatotoxicity, abnormal biotransformation of drugs and other adverse reactions occur, which leads to hepatic dysfunction called drug-induced liver injury (DILI) and, in many cases, acute liver failure (ALF). Therefore, the ability to predict drug hepatotoxicity in the early stage of development is highly important (Fisher et al., 2015; Segovia-Zafra et al., 2021). Thus, another vital aspect is the choice of appropriate in vitro and in vivo models to study hepatotoxicity. Currently, human and animal cell-based in vitro and animal in vivo models are used in drug toxicity research. However, the literature reports poor predictability of animal models due to many discrepancies between human and animal species, for example, in proteins engaged in drug metabolism (Heinonen, 2015).

To study drug hepatotoxicity, in vitro cell-based models should meet certain conditions. First, they should be characterized by membrane polarity because such a configuration enables the proper functionality of drug transporters to be maintained. Moreover, they should demonstrate the activity of phase I and phase II drug-metabolizing enzymes, which are responsible for xenobiotic biotransformation (Han et al., 2019). On the basis of the abovementioned features, the most suitable cells for studying drug hepatoxicity are human liver parenchymal cells - hepatocytes. However, the following obstacles, such as poor availability, lack of the capacity to proliferate, and quick dedifferentiation ex vivo with the loss of the ability to perform liver-specific functions, limit their usage in preclinical studies (Rowe et al., 2013). Therefore, many attempts to preserve the functionality of human hepatocytes in vitro have been undertaken, including cocultures with nonparenchymal cells, 3D cultures, additives to the culture medium (e.g., growth factors) or the utilization of different growth substrata, which enhance cell adhesion. Nevertheless, as the availability of human hepatocytes is limited, scientists have focused on alternative sources of cells, such as cell lines derived from different primary liver cancers (hepatomas): hepatoblastomas and hepatocellular carcinomas (HCCs), or hepatocyte-like cells derived from stem cells, which could constitute a model of human liver parenchymal cells (Gómez-Lechón et al., 2014; Ruoß et al., 2020; Pluta et al., 2021).

Currently, cancerous liver-derived cell lines are widely used in drug hepatotoxicity research. The most commonly used cell line is HepG2; however, the number of available cell lines is enormous. The other widely used cell lines are HepaRG, Huh7 or Hep3B (Donato et al., 2013). Notably, the functionality of different human tumor liver cell lines is not comparable to that of liver parenchymal cells. For example, the HepG2 cell line derived from human hepatoblastoma (thus far mistakenly thought to originate from human HCC) was isolated in 1975 and is characterized by a lack (e.g., CYP1A2, CYP2C9) or low (e.g., CYP3A4, UGT1A1) expression of phase I- and phase II-metabolizing enzymes. Only the phase II enzymes SULT1A1 and SULT2A1 are expressed at levels similar to those of human hepatocytes (Aden et al., 1979; López-Terrada et al., 2009; Wiśniewski et al., 2016). Moreover, HepG2 cells and its clonal derivative - C3A, similar to HCC cell lines (e.g., SNU449, SNU475, Hep3B, and Huh7), have an unfunctional urea cycle, which results from low or undetectable expression of genes and proteins involved in this metabolic process. Interestingly, scientific data revealed that this is a common feature of HCC tumors. Therefore, it is suggested that HepG2 cells should be used for testing anticancer drugs instead of evaluating the potential hepatotoxicity of newly developed drugs (Pluta et al., 2020; Missiaen et al., 2022; Jakubowska et al., 2024). Another cancer cell line, Hep3B, is derived from primary HCC and is widely known as an in vitro model for investigating the mechanism of hepatotoxicity caused by acetaminophen (APAP). In contrast to HepG2 cells, Hep3B cells are positive for hepatitis B virus (HBV) and have tumorigenic potential. Moreover, they are characterized by the presence of fewer proteins specific for human hepatocytes and different responses to analyzed drugs than HepG2 cells. A possible explanation for this phenomenon is the different origins of the 2 cell lines (derived from hepatoblastoma and HCC, respectively), which results in different phenotypes: Hep3B cells present more features characteristic of mesenchymal cells (fibroblasts), whereas HepG2 cells present more features characteristic of hepatocytes (Manov et al., 2002; Slany et al., 2010; Qiu et al., 2015). In turn, the HuH7 cell line is widely used in research concerning hepatitis C virus (HCV) infection because of its high susceptibility to this pathogen. Moreover, this HCC cell line has also been utilized in drug cytotoxicity studies; however, reports in the literature have shown that the mRNA levels of phase I- and phase II-metabolizing enzymes are extremely low compared with those in human hepatocytes (with the exception of the CYP1A1 gene) (Nakabayashi et al., 1982; Choi et al., 2009). Nevertheless, the most promising cell line for studying potential drug hepatotoxicity is HepaRG cells. These cells are derived from HCC and are able to exhibit two different cell phenotypes—hepatocyte-like or biliary-like. Importantly, the majority of HepaRG cells described in the literature express phase I- and phase II-metabolizing enzymes (e.g., CYP3A4, CYP1A2, CYP2C9, and CYP2E1) at levels more comparable to those in human hepatocytes than those in the HepG2 cell line (Gripon et al., 2002; Guillouzo et al., 2007).

As previously mentioned, different hepatic cell lines utilized in drug hepatoxicity assays may significantly differ from human liver parenchymal cells and from each other. Therefore, the aim of our research was to demonstrate that before a hepatic cell line is selected for preclinical studies, it is necessary to investigate and compare the expression of genes and proteins, which are crucial in assessing the hepatotoxicity of the analyzed drug. In this report, we compared the expression of selected relevant liver-related genes in HepG2, C3A, SNU449, and SNU475 cell lines using RT-qPCR analysis. Additionally, we analyzed the differences in their phenotype and the ability to perform liver-specific functions. Finally, we investigated the expression of liver-related proteins, especially those engaged in drug metabolism, and compared with results obtained for isolated human hepatocytes.

In this study we compared several hepatoblastoma (C3A, HepG2) and HCC (SNU449, SNU475) cell lines, in terms of their behavior during culture and the differences between their phenotypes, including morphology and in vitro growth rate, as well as liver-related gene expression at the protein and mRNA levels. The comparison revealed noticeable differences between cell lines either within cells of the same cancer origin or between different types of cancer (hepatoblastoma or HCC). Notably, these cell lines, which are derived from various liver tumors, exhibit varying levels of invasiveness and distinct responses to anticancer treatments. Moreover, all tested cancer cell lines differed significantly from isolated human hepatocytes, which generally presented much greater gene expression levels.

Among all the tested cell lines, C3A had the fastest growth, with the shortest doubling time. In contrast, after 4 days in culture, there were ca. 6 times fewer SNU475 cells, suggesting that no doubling occurred during this time. Moreover, compared with other cell lines, SNU475 cells presented only slightly lower cell viability (80% vs 90% of living cells). In general, hepatoblastomas grow faster in culture than do HCCs (Figures 1, 2) and this is what determines their greater availability. Additionally, hepatoblastoma cultures presented greater cell metabolic activity than HCC cultures did, as measured via the MTT test, with the exception of SNU475 in the case when the absorbance was calculated per number of living cells. This finding might indicate that this cell line, despite its very low proliferation rate, is the most metabolically active among the tested cells (Figures 4A,B). Another physiological difference observed between these two cancer types is the tendency of hepatoblastomas to spontaneously form aggregates (so-called islands) whereas HCCs exhibit a more flattened phenotype (stronger adherence) (Figure 3). It is worth noting, that in preparation for hepato- or cytotoxicity assays, spheroids or organoids, which recapitulate the 3D architecture, heterogeneity, and cell functions of primary tissues, may be preferable forms of cell culture.

One of the most important hepatocyte functions, enabling the estimation of liver-specific cell functionality, is albumin production (Zakrzewska et al., 2017). In this context, C3A cells performed best and hepatoblastomas outperformed HCC cell lines. As shown by the results of the quantitative ELISA, SNU449 secreted minimal amounts of albumin into the culture medium, whereas in the SNU475 cultured medium, albumin production was not detected (Figure 5). Interestingly, in hepatoblastomas albumin synthesis increased during 7 days of culture, but in SNU449, it decreased. When the ability of the examined cells to form AV was tested, it appeared that hepatoblastomas formed these structures efficiently, yet the HCC cell lines were incapable of doing so (Figures 6A,B). This observation is important because the polarization of hepatocyte membranes enables intracellular transport, which is necessary for xenobiotic processing and detoxication. Notably, that according to our unpublished observations, isolated human hepatocytes dedifferentiate so quickly ex vivo that they are also unable to form bile canaliculi precursors.

Using western blot, a semiquantitative method for estimation of protein production, we checked the absence or presence, in the tested cell lines and hepatocytes, of the following proteins: albumin (ALB), which is produced exclusively by hepatocytes; arginase 1 (ARG1), one of the key proteins of the urea cycle; and carbamoyl-phosphate synthase 1 (CPS1), the major mitochondrial urea cycle enzyme. Moreover, we analyzed the expression levels of alpha-fetoprotein (AFP), which is specific for immature hepatocytes and the tumor cells; human nuclear factor 4 alpha (HNF4α) and human nuclear factor 1 alpha (HNF1α), which are transcription factors that regulate the expression of several liver-specific genes. We also verified the expression of the cytochrome P450 isoenzymes CYP3A4, CYP2E1, and CYP2C8, which are involved in the metabolism of xenobiotics (Figure 7). The profile of the expression of these proteins differs significantly between cancer types. The presence of albumin was confirmed only in HepG2 and C3A cells, as well as in human hepatocytes and liver cell isolate. As expected, AFP was not found in the HH or LCI but was also not present in the tested HCC cells and was found only in the analyzed hepatoblastomas. The production of urea cycle enzymes ARG1 and CPS1 was below detection levels in both cancers which is in accordance with the observations published by other authors (Missiaen et al., 2022). Interestingly, the protein band corresponding to ARG1 is visible in the HH and LCI preparations, whereas this band assigned to CPS1 is only visible in the HH preparations. The major liver regulatory proteins HNF4α and HNF1α were detected only in cultured HH and in both HepG2 and C3A hepatoblastoma cells. HNF4α, but not HNF1α, was also detected in LCI, but these proteins were not detected in SNU475 and SNU449 HCC cells. Similarly, the main liver cytochrome P450 isoenzyme, CYP3A4, was present in the HH, LCI, HepG2, and C3A preparations but not in the SNU preparations. However, the other two isoenzymes, CYP2E1 and CYP2C8, were found only in the HH.

At the mRNA level, the expression of the following liver-specific genes was tested via RT-qPCR: ALB, ARG1, CPS1, HNF4α, HNF1α, CYP3A4, CYP2E1, CYP2C8, GLUL, and PXR. The levels of transcripts were quantitatively measured in the C3A, HepG2, SNU449, and SNU475 cell lines and compared to the gene expression in isolated human hepatocytes (Figure 8). Generally, the results for ALB mirror the gene expression observed at the protein level. Compared with those of HH, the ARG1, CPS1, HNF4α, and HNF1α expression levels in tumor cell lines are minimal, if at all. The observations are also similar for all tested P450 isoenzymes and for PXR, the pregnane X receptor, which is a nuclear receptor of foreign toxic substances that upregulates the expression of proteins involved in the detoxification and clearance of these substances from the body. In summary, the RT-qPCR results revealed that the expression of the tested genes in tumor cells of both origins did not approach the levels observed in HH with the exceptions of ALB and GLUL. GLUL encodes glutamine synthetase, an enzyme involved in ammonia detoxification, but its function is not linked to the urea cycle. In this case, the transcript was detected at very low levels in C3A, HepG2, and SNU449 cells but not in SNU475 and HH. This finding is unexpected since its activity is usually observed in a subset of pericentral hepatocytes, and again, it might result from rapid HH dedifferentiation in vitro.

Despite their imperfections in recapitulating hepatocyte functionalities, the described cell lines, especially HepG2 and C3A, are widely used in basic research and drug testing. In contrast to HepG2 and its clonal derivative C3A, which is derived from hepatoblastoma (López-Terrada et al., 2009; Arzumanian et al., 2021), the SNU449 and SNU475 cell lines belong to the mesenchymal-like subgroups of HCC (Oz et al., 2021; Nguyen et al., 2022). Both SNU cell lines are commonly used as models of cancer cells to study the mechanisms of tumor development, biological activities, and pathways leading to cell death (Azbazdar et al., 2023; Suo et al., 2023; Baris and Demir, 2024). They are also commonly used to evaluate anticancer targets and the potential of experimental compounds (Lee et al., 2020; Sanaei et al., 2021; Abdou et al., 2023), drug nanocarriers (Xu et al., 2023; Pieretti et al., 2024), new nonpharmacological technologies (Saadawy et al., 2023), and well-known drug resistance mechanisms (Son et al., 2021).

The differences in response to the toxicity of anticancer drugs arise from the heterogeneity of individual cancer cell types. SNU475 cells are described in the literature as a model of a more aggressive cancer with higher migratory capability than less invasive HepG2, with SNU475 cells being more mesenchymal-like and HepG2 cells being more hepatoblast/epithelial-like (Caruso et al., 2019; Akbari et al., 2021; Nguyen et al., 2022). For example, in a study on how cytoskeletal dynamics regulate stromal invasion behavior in different liver cancer subtypes, the authors noted that cytoskeletal disruptions notably hinder highly migratory SNU475 cells, but have a less significant effect on more proliferative HepG2 cells (Nguyen et al., 2022). In a study by (Akbari et al., 2021) the authors showed that LGR5 expression—a marker important for liver regeneration and stem cell maintenance—was detected in hepatoblast-like lines such as HepG2, but was low or absent in mesenchymal-like lines, including SNU449 and SNU475. Activation of the canonical Wnt pathway significantly increased LGR5 expression in HepG2, but not in SNU449 or SNU475.

SNU449 and SNU475 cells exhibit different responses to chemotherapy. Consistent with the findings of (Mok et al., 2006), SNU449 demonstrated greater resistance to standard chemotherapeutic agents, such as paclitaxel, docetaxel, doxorubicin, and the most effective patupilone, than other HCC cell lines, including SNU475. Similarly (Sanaei et al., 2021), reported that SNU449 showed greater resistance to sodium butyrate than SNU475 (IC₅₀: 6.976 µM vs 4.175 µM). Although apoptosis levels induced by both compounds were higher in SNU475, statistical differences between the cell lines were not assessed. In contrast, both lines exhibited similar sensitivity to 5′-fluoro-2′-deoxycytidine. In line with this, no significant differences were observed between SNU449 and SNU475 in response to plasma-activated air-driven water mist (Saadawy et al., 2023). On the other hand, in a study by (Son et al., 2021), HepG2 was the most sensitive to sorafenib, while SNU475 was the most resistant, likely due to differing expression levels of FGL1, a protein associated with sorafenib response. Interestingly, in response to ZnO nanoparticles, SNU449 cells were more sensitive than HepG2, as indicated by significantly lower IC₅₀ values. Additionally, ZnO-induced cell death differed: apoptosis predominated in HepG2, while necroptosis occurred in SNU449 (Pieretti et al., 2024).

In summary, the response to chemotherapy differs among HepG2, C3A, SNU449, and SNU475 cells, which may be due, among other factors, to a different repertoire of active genes responsible for drug metabolism, including cytochromes, and to differences in cell physiology, such as growth behavior, albumin production and the ability to maintain membrane polarity (AV formation). In our study, we demonstrated, at both the protein and transcript levels, different levels of expression of the selected cytochrome P450 isoforms, namely, CYP2E1, CYP2C8, and CYP3A4, which are activated by different xenobiotics. Additionally, the PXR gene, which plays important roles in disposing of endogenous and xenobiotic compounds and maintaining metabolic homeostasis, was differentially expressed in the tested cell lines. Importantly, these levels were much greater in HH. Similarly, these differences were observed for hepatocyte nuclear factors (HNF1α and HNF4α), genes that regulate the synthesis of major liver-related proteins. Finally, they liver-relevant proteins, such as ALB (major liver-produced protein responsible for the binding of toxic compounds), ARG1 and CPS1 (enzymes involved in the detoxication of ammonium cations via the urea cycle), are present in different amounts in the tested cells. Considering the above factors, careful selection should be performed to choose the proper cell line to test particular xenobiotic-driven hepatotoxicity. Moreover, these results indicate that the SNU475 and SNU449 cell lines in particular might not be perfectly suited for drug testing. Nevertheless, HH constitute the best cell source for drug testing, and no alternative cell source can replace the functionality of these cells. On the other hand, HH are not widely available, quickly dedifferentiate ex vivo and do not guarantee reproducibility, because originate from patients with variable genetic backgrounds.

The main limitation of the presented study is the lack of functional testing of the investigated genes. Additionally, the inducibility of various cytochrome P450 isoforms was not analyzed. The study was also restricted to four cell lines, and the cultures were maintained in a monolayer. This experimental design was chosen to simplify the study as much as possible and to minimize the influence of multiple variables on the results. According to our assumptions, this approach allowed us to reduce interpretative errors resulting from the high genetic background variability—both among the studied cell lines and primary hepatocytes derived from different donors—and to present, in an intentionally simplified manner, conclusions regarding the variable suitability of cell lines of different origins for cytotoxicity studies.