The application of PHHs is considered the gold standard method for toxicological and pharmacological study. These cells more closely resemble the in vivo phenotype as they recapitulate the specific human liver function and metabolism. The harvest of cells from patients suffering from NAFLD has improved the precision of studies on candidate drug metabolism (Ganji et al., 2015[34]; Schwartz et al., 2020[98]). However, the shortage of PHHs and the reliance on patient surgery are the main reasons that using human primary hepatocytes is not very common. Moreover, the donor-dependent variability of PHHs leads to significant variation and low reproducibility in experimental results (Zeilinger et al., 2016[145]; Bell et al., 2018[8]).

PHHs are terminally differentiated liver parenchymal cells but under in vitro conditions show phenotypic instability, limited culture duration, and proliferation. Nonetheless, when PHHs were exposed to the free fatty acids (FFAs) demonstrated an increase in lipid accumulation and endoplasmic reticulum (ER) stress. The steatosis induced by FFAs was significantly reduced after treatment with glucagon-like peptide 1 (GLP-1) (Wobser et al., 2009[135]). This study displays a new role for GLP-1 analog in hindering steatosis progression. Furthermore, Hepatic Stellate cell (HSC) activation was done when these cells were exposed to the conditioned media harvested from FFAs-treated PHH. This finding indicates that lipid accumulation in hepatocytes caused HSC activation, enhanced their apoptosis resistance, and induced the expression of the profibrotic genes including transforming growth factor-beta (TGF-β), and matrix-metalloproteinase-2 (MMP-2), tissue inhibitor of metalloproteinase-1 (TIMP-1), tissue inhibitor of metalloproteinase-2 (TIMP-2), nuclear-factor κB (NF-κB)-dependent monocyte chemoattractant protein-1 (MCP-1) in HSCs.

In summary, these data display the pathophysiological link between liver steatosis and fibrosis progression. As a result, despite the advantages of PHHs to investigate a range of NAFLD key events, the bottleneck is that these cells are not widely available to generate reproducible in vitro NAFLD models for high-throughput assessments, which is essential for commercialization of the approach.

Recently, hepatocyte-like cells (HLCs) differentiated from human pluripotent stem cells (hPSCs), i.e. embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) (Malik and Rao, 2013[71]). Human ESCs are considered a gold standard and distinguished pluripotent stem cells however, their isolation from human embryos and the possibility of teratoma formation create serious ethical concerns for their use in research and therapy. Human iPSCs are promising alternatives for ESCs derived from somatic cells, particularly the fibroblast, through transcriptional reprogramming. These pluripotent stem cells have the same potential as ESCs and can similarly differentiate into HLCs (Touboul et al., 2010[118]).

To date, both sources of human PSCs are considered powerful tools in cutting-edge study and very limited in therapy. Many groups have generated HLCs that somehow express liver-specific genes and proteins, secret albumin (ALB), produce urea, and store glycogen, as well as demonstrated Cytochromes P450 (CYPs) enzyme activity and inducibility (Zahmatkesh et al., 2021[142]). However, there is no protocol to produce fully mature hepatocytes which are functionally equivalent to PHHs from these PSCs. Indeed, the HLCs derived from ESC or iPSC, express the immature hepatocyte marker, alpha-fetoprotein (AFP), secrete less ALB, and exhibit significantly decreased CYPs activity compared to PHHs (Bukong et al., 2012[10]; Yi et al., 2012[138]; Schwartz et al., 2014[99]).

To induce further maturation and metabolic functionality of the HLCs, a variety of approaches, particularly 3D cultivation and co-culture have been used, however, the issue remained (Roy-Chowdhury et al., 2017[92]). Nonetheless, as an alternative to PHHs, these cells can be used for some investigations or limited therapy including patient-derived HLCs generation for drug screening, assessments of inter-individual variability in drug response, or hepatotoxicity (Godoy et al., 2013[37]). Indeed, hiPSC derived from patients with hereditary metabolic diseases can be utilized to model diseases in laboratory conditions. The iPSC technology combination with genome editing technologies, e.g., clustered regularly interspaced short palindromic repeats (CRISPR) and CRISPR-associated protein-9 (Cas9), provides further potential for disease modeling and therapeutic applications (Lyu et al., 2018[69]).

Holmgren and colleagues (2020[47]) used differentiated iPSCs into the HLCs to generate a NAFLD model in which in response to lipid accumulation and ER stress the inflammatory marker such as tumor necrosis factor α (TNFα) was upregulated. Ouchi et al. (2019[80]) developed an hiPSC-derived model of hepatic steatosis where human pluripotent stem cells are co-differentiated into stromal and epithelial lineages whereupon exposure to FFA shows hepatocyte ballooning, fibrosis phenotype, and an increase in inflammation markers. Despite the advantage of PSCs for disease modeling, so far, there are some limitations to their use due to insufficient functionality, incomplete hepatic differentiation that limits the reproducibility of experiments, and difficulties to produce large volumes of these cells (Ouchi et al., 2019[80]).

Human hepatic cell lines are immortalized cells, either derived from liver tumors (e.g., HepaRG, HepG2, Hep3B, SK-Hep-1, and HuH7) or produced through genetically manipulated primary hepatocytes (e.g., SV40 Large T, and hTERT) (Wilkening et al., 2003[134]; Andersson et al., 2012[2]; Samanez et al., 2012[96]). The immortalized cell lines have the wide availability and ability of unlimited proliferation, which makes them more cost-effective compared to PHHs and HLCs and an appropriate cell source for large-scale usages with high reproducibility (Ramboer et al., 2014[89]).

However, human hepatic cell lines are limited by lower functionality, especially in CYP450 activity compared to PHHs (Guo et al., 2011[42]). Also, some mutations in the immortalized cell lines may significantly alter the cell phenomena compared to normal hepatocytes in the human body. Therefore, the selection of appropriate cell lines for each type of experiment should be considered intelligently. For example, HepG2 cells that are widely used for NAFLD in vitro modeling, are homozygous for the PNPLA3 I148M mutation, which is involved in several key metabolic functions (Gunn et al., 2017[41]).

Huh-7 cells have effective CYP3A4 activity and have been used to study the mechanism of chemicals that improve steatosis and reveal new NAFLD pathogenesis mechanisms (Sivertsson et al., 2010[104]). Khamphaya and colleagues (2018[57]) used the Huh-7 cell line to investigate the mechanism of liver regeneration impairment that occurs in NAFLD. They demonstrated, similar to the rat model and liver biopsy of patients with NAFLD, a decrease in type II inositol 1,4,5-trisphosphate receptor (ITPR2) may account for the impaired liver regeneration in the fat-loaded Huh7 cells.

HepaRG cells are a human bipotent progenitor cell line able to differentiate into two different types of liver cells, cholangiocyte-like, and hepatocyte-like cells, that may be promising surrogates for PHHs. HepaRG exhibits drug-metabolizing enzymes and drug transporters expression patterns similar to PHHs, so that is becoming an interesting tool for hepatic disease modeling (Lübberstedt et al., 2011[68]; Andersson et al., 2012[2]).

Tolosa and colleagues (2016[116]) exposed HepaRG cells to 28 compounds that have been reported to induce steatosis for mechanistic studying of drug-induced liver steatosis. Lipid content, ROS production, mitochondrial membrane potential alternation, and lipid metabolism gene expression were assayed to investigate the mechanisms involved in drug-induced steatosis. The finding of this study has clinical relevance because most effects of the drug were 100-fold under the concentrations of the therapeutic plasmatic concentration. These results indicate the value of HepaRG cells as a cell-based assay system for studying the mechanisms involved in drug-induced steatosis.

Overall, using cell lines for disease modeling facilitates standardized protocols, reproducible, and scalable studies. However, a major drawback in using cell lines is their tumor phenotype and altered metabolic functions which limits true comparison with the human condition.

The interaction of non-parenchymal cells (NPCs) with hepatoblasts and hepatocytes is critical during liver development, normal physiological state, cellular response to injury, progression of diseases, or regeneration. It has been demonstrated that the addition of NPCs to in vitro cultured hepatocyte systems prevents their dedifferentiation, and improves hepatic function and metabolic response to drug treatment (Hasmall et al., 2000[45]; Wheeler et al., 2014[133]).

Today various types of primary, stem cell-derived, and immortalized NPCs including liver sinusoidal endothelial cells (LSECs), Kupffer cells (KCs), and hepatic stellate cells (HSCs) are available. Just like PHHs, primary NPCs are isolated from the liver, by enzymatic perfusion (Damm et al., 2013[20]; Pfeiffer et al., 2015[83]; Werner et al., 2015[131]). Due to the high cost and instability of the primary cells, several cell lines or alternative cell types have been used in co-cultures with hepatocytes to provide signals similar to the ones provided by primary NPCs. In this section, we will introduce the proper cell composition for optimizing the microenvironment in liver microtissues for disease modeling.

ECM is the main parameter in the hepatic microenvironment and has a crucial role in maintaining hepatocyte functionality and polarization in healthy liver, as well as critical in cellular behavior under disease initiation and progression. It has been demonstrated that in tissue engineering, using an ECM-like scaffold is an insightful approach as it recapitulates the native tissue microenvironment better. To date, a variety of synthetic and natural compounds with different mechanical and biochemical properties have been used to fabricate the artificial 2D substrate or 3D scaffold for liver tissue or organ equivalent construct bioengineering. In this section, we will discuss recreating the appropriate scaffolds/matrices for optimizing the microenvironment in the generation of liver microtissues for disease modeling.

For many years, cell culture has been done in flasks, Petri dishes, and plates, or more recently in bioreactors. In these traditionally produced culture conditions cell-microenvironment interaction, which determines cell function and phenotype and influences cell response to stimuli, is not fully controlled (Sengupta et al., 2014[100]). Lately, the microfluidic approach incorporates various knowledge, including biochemistry, engineering, physics, and biology, for developing a novel technique in cell culture on a micro-scale (Young and Beebe, 2010[140]).

Over the last decade, microfluidic-based cell culture platforms became more attractive in biological research applications, such as drug toxicity assay, differentiation of the stem cell, and biological processes (Dittrich and Manz, 2006[24]; Wan et al., 2011[124]; Zervantonakis et al., 2011[146]). The main advantage of the microfluidic platform for cell culture includes mimicking in vivo cellular microenvironments by precise control of spatial and temporal factors, such as soluble factors gradient, nutrients and oxygen supply, metabolites elimination, and shear stress (Sung et al., 2010[111]; Marimuthu and Kim, 2011[72]; Zervantonakis et al., 2011[146]). Since the in vivo hepatocytes' microenvironment is highly perfusing, microfluidic-based cell culture is more appropriate for providing physiologically relevant conditions and preserving the phenotype and functions of the liver cells. However, liver chip throughput is relatively low, which is not appropriate for high throughput and fast industrial applications.

The integration of biosensors offsets the drawbacks of liver chip devices in terms of throughput. In addition, real-time monitoring rather than endpoint analysis would help to better understand the dynamic changes in disease progression (Asif et al., 2021[4]). Furthermore, organ-on-a-chip-based systems can be engineered to design fine tissue architecture. Moreover, a small volume of microfluidic devices is another advantage of this system. It is demonstrated if hepatocytes culture in a microfluidic channel, even if without flow, it remained functional for 21 days in the presence of sufficient nutrients, due to the accumulation of endogenous growth factors including EGF, HGF, and IGF (Choi et al., 2020[15]).

In addition, as was previously mentioned in the review, the sourcing of human hepatocytes is associated with significant challenges. So far, there are many robust protocols for hPSCs differentiation into hepatocytes, but these protocols require a combination of expensive appliances and reagents. Furthermore, the resulting hepatocytes exhibit lower levels of enzyme activity compared to adult hepatocytes (Duan et al., 2010[28]). Microfluidic platforms enhance differentiation efficiency and decrease the cost of differentiation protocols by using fewer reagents (Giobbe et al., 2015[36]).

Moreover, communication between different organs has a critical role in disease progression and drug-induced hepatotoxicity. Therefore, it is necessary to study the interaction between multiple organs and their particular role in the development and progression of the disease. In this regard, connecting a liver chip with other organ chips, which are known as “body-on-a-chip” or “human-on-a-chip” platforms, is emerging as a valuable technology for a better understanding of the complex mechanisms of such disease.

Up to now, a lot of studies by the 2D monoculture of PHHs, various human hepatic cell lines (Huh-7, HepG2, HepaRG), and HLCs then exposing these cells to FFA showed some key features in NAFLD, such as steatosis, ER stress, and oxidative stress (Anthérieu et al., 2011[3]; Sharma et al., 2011[101], Graffmann et al., 2016[39]). However, tracking multiple intercellular interactions in NAFLD required the co-culture of these liver parenchymal cells with the other NPCs presented in the tissue.

Feaver and colleagues (2016[31]) engineered an in vitro NAFLD model in a contact-less co-culture fashion in which the PHHs were cultured at the bottom of a transwell whereas human KCs and primary human HSCs were cultured at the top of the transwell. The system integrated hemodynamic-like flow with the media containing elevated FFA, insulin, and glucose levels to induce a NASH phenotype. Their results had some similarities with NASH patients in terms of lipids accumulation, mitochondrial dysfunction, increased oxidative stress, inflammation, and fibrotic markers in the in vitro model. However, in this transwell model, the applied non-parenchymal cells have no physical interactions with PHHs.

In another study, a monoculture of LX2 or their co-culture with Huh-7, in both contact-less or contact fashions, were exposed to FFA for 24 hours. FFA induces intracellular lipid accumulation in all three experimental groups, but HSC activation was significantly increased only in contact co-culture conditions (Barbero-Becerra et al., 2015[5]). This result showed that the activation of HSC is independent of the accumulation of FFA but requires cell-to-cell interactions with injured hepatocytes.

Incorporating other liver cell types into NAFLD models can effectively overcome several disadvantages of monoculture systems, but dimensionality and medium flow are still missing. 3D cell culture could better represent the in vivo situation, thus proving more pathophysiological or physiological relevance. Also, 3D co-culture systems can be utilized for investigating the more complex features of NASH, such as inflammation and liver fibrosis.

Kozyra and colleagues (2018[61]) developed a 3D human hepatic spheroid model to mimic steatotic conditions. PHHs from different donors were cultured as a 3D spheroid and supplemented with pathophysiological concentrations of FFA, carbohydrates, and insulin to mimic steatotic conditions and insulin resistance over three weeks. They demonstrated the accumulation of lipids in hepatocytes and the upregulation of lipogenic and insulin-resistant genes. Pingitore and colleagues (2019[85]) generated a model of 3D spheroids consisting of hepatocyte and HSC cell lines (HepG2 and LX-2, respectively). The spheroids showed lipid accumulation and fibrotic features upon exposure to FFA. In addition, they also showed rescued lipid accumulation by treating the spheroids with Elafibranor or Liraglutide which are in clinical trials for NASH treatment.

In addition to fibrosis, inflammation is also a critical event in NAFLD. Suurmond and colleagues (2019[112]) developed an in vitro human liver spheroid for NAFLD by co-culturing HepG2, HUVECs, and Kupffer cells in GelMA and evaluated the disease progression from steatosis to NASH. They demonstrated that upon exposing this microtissue to FFA, KCs could be activated and secrete pro-inflammatory cytokines, as well as higher levels of cellular stress. They concluded that KCs require for creating more reliable in vitro human liver models of NAFLD.

Derivation of organoids directly from patients with NASH provides a cell source for personalized disease modeling and fundamental NAFLD/NASH biology and facilitates high throughput screening for drug development.

Ouchi and colleagues (2019[80]) developed a method for a human liver organoid generation where human pluripotent stem cells co-differentiated into epithelial and stromal lineages and then incorporated in matrigel and exposed to retinoic acid in a specific maturation medium. The resulting organoid was composed of multi-cellular including hepatocyte-, stellate-, and Kupffer-like cells whose transcriptome was comparable to native liver tissues. Treatment of these multicellular liver organoids, using increasing doses of FFAs displayed a gradual increase in intracellular lipids accumulation. This organoid recapitulates the stepwise and progressive nature of steatohepatitis and secretes inflammatory cytokines such as IL-6, IL-8, and TNF-α. Prolonged exposure of the organoids with FFAs leads to hepatocyte ballooning as well as upregulation of vimentin and α-SMA. Moreover, upon FFA treatment, the storage rate of vitamin A in the stellate-like cell has declined, indicating that these cells are activated. Additionally, measuring the liver organoid's stiffness by atomic force microscopy (AFM) confirmed the fibrosis.

Gurevich and colleagues (2020[43]) generated liver organoids throughout the co-culture of iPSC-derived hepatocyte-, stellate-, and Kupffer-like cells from NASH donors on collagen-coated plates. They demonstrated that the hepatocytes derived from NASH donors showed accumulation of lipids and recapitulated in vivo characteristics, even without FFA supplementation. McCarron and colleagues (2021[74]) generated organoids from the liver cells of patients suffering from NASH and normal livers as controls. The NASH patient-derived organoids exhibited significant up-regulation in their metabolic and pro-inflammatory pathways and showed all typical features of NASH, including reduced albumin secretion, increased apoptosis sensitivity, and lipid accumulation.

Due to the capability to mimic in vivo physiological parameters, microfluidic-based cell culture improves the metabolic activity and functionality of hepatocytes ensuring that disease progression and drug metabolization are more similar to the in vivo condition (Sivaraman et al., 2005[103]). Moreover, the liver is not a lone organ involved in NAFLD, other organs such as adipose tissue and the gut have an important role in the disease initiation and progression. In the past decades, liver-on-a-chip and multiorgan-on-a-chip have been powerful tools for drug screening, liver disease modeling, and studying multiple organ interactions.

Gori and colleagues (2016[38]) developed an advanced microfluidic perfused liver model in which HepG2 was cultured in parallel microchannels that mimic the endothelial-parenchymal interface of a liver sinusoid, allowing the diffusion of nutrients and removal of the wastes similar to the hepatic microvasculature, but actual endothelial cells were not employed in this system. In this microfluidic device, HepG2 cells were perfused in a culture medium supplemented by oleic and palmitic acid, for 24 h and 48 h. They demonstrated a gradual and lower accumulation of intracellular lipids, increased liver cell viability, and minimal oxidative stress in microfluidic dynamic compared with static cultures. Therefore, the chronic condition of in vivo steatosis is more closely replicated in the microfluidic condition. Hence, this system provides a more reliable model than the static 2D cultures for studying NAFLD pathogenesis.

In addition, Kostrzewski and colleagues (2017[59]) developed a 3D NAFLD perfused platform by using PHH cells. The liver-on-a-chip platform composes of 12 isolated bioreactors in which fluid is recirculated by a micro pump. The channels were coated with a collagen scaffold in which the 3D microtissue could form. After 14 days of exposure to FFA, the hepatocytes showed intracellular lipid accumulation, upregulation of adipokines, and genes associated with NAFLD. Similar to patients suffering from NAFLD, the metabolic activity of the hepatocytes was significantly decreased. On the other hand, when the NAFLD model was treated with anti-steatotic compounds the hepatocytes showed reduced intracellular lipid content compared to untreated controls. However, these platforms focused only on hepatocyte cells, ignoring the importance of their interaction with NPCs, so that unable to reconstruct the suitable liver microenvironment for NAFLD modeling.

Cho and colleagues (2021[14]) developed an advanced liver-on-a-chip platform composed of four cell types, including HepaRG, HUVEC, primary Kupffer cells, and primary stellate cells in GelMA to achieve a human model of hepatic fibrosis derived from NAFLD. They showed accumulation of lipids in hepatocytes (steatosis), neovascularization in endothelial cells, the inflammatory response by activated Kupffer cells (steatohepatitis), and extracellular matrix deposition by activated stellate cells (fibrosis). In this model, the use of stellate cells caused an increase in inflammatory responses and fibrosis markers upregulation. It is suggested that the application of all necessary primary cells or cell lines provide a suitable condition for producing liver microtissue. However, in most differentiation approaches for NAFLD modeling, PSCs only differentiate into HPLCs, and the other essential cell types such as pro-fibrotic and/or inflammatory cells are lacking (Mun et al., 2019[78]).

Lee and Sung (2018[65]) designed a “gut-liver-on-a-chip” platform in which FFAs absorption through the gut layer and then the accumulation of lipids in hepatocytes was demonstrated. This gut-liver chip has two layers that, by a porous membrane, were separated. Gut cells were cultured (Caco-2) on the top layer, and liver cells (HepG2) were cultured on the bottom layer. When butyrate, an anti-steatotic compound that enhances the gut barrier function is administered to the gut compartment, a significant decrease in lipid accumulation in the liver cell is observed. Whereas applying TNF-α by compromising the barrier function of the gut increased hepatic steatosis. This platform had the potential to recapitulate the absorption and lipid accumulation in the gut and liver; moreover, it was a more relevant in vitro model of hepatic steatosis than the conventional model of culture.

Slaughter and colleagues (2021[105]) presented a human-on-a-chip model consisting of human liver and adipose tissue, to evaluate the metabolic factors that contribute to NAFLD initiation and progression. In this study, the indirect effects of adipocytes on hepatocytes upon the progression of NAFLD were validated.

Gather together, in such novel organ-on-a-chip models, it is possible to interplay between the liver and other organs to address the multi-tissue etiology of NAFLD. However, because of high complexity, high cost, and limitations to scalability, these organ-on-a-chip models were not widely used in laboratories.

3D bioprinting could be used to create 3D liver constructs with high accuracy and specific spatial organization. Various bioinks, such as synthetic, natural, and hybrid polymers have been used in 3D bio-printing. Type I collagen has been widely used as a hepatocytes scaffold, but it is less suitable for 3D bio-printing due to its low viscosity and slow rate of polymerization. An improved hybrid of type I collagen and hyaluronic acid yielded a simple, printable bioink that allows PHHs to become viable over two weeks in the resultant construct and respond to drug treatment (Mazzocchi et al., 2018[73]).

Decellularized liver ECM-derived hydrogel with photo-cross-link capability was created as a bioink to provide specific signals (Ma et al., 2016[70]). However, as far as we know, only one 3D bio-printing approach was used for NAFLD in vitro modeling (Kizawa et al., 2017[58]). In this study, cells were collected from genetically obese Zucker rats with NAFLD, and the construction of bio-printed liver tissue was carried out according to the Kenzan methods (Moldovan et al., 2016[77]). On day 23, staining the bio-printed liver structure with Oil-red O showed high lipid content, which indicates that NAFLD conditions can be sustained and monitored in the long term.

Moreover, 3D bio-printing technology can also improve liver-on-a-chip systems by overcoming limitations in providing various cell types and ECM for spatial heterogeneity (Lee and Cho, 2016[64]).

The authors declare that they have no conflict of interest.

NA drafted the manuscript and contributed substantially to this manuscript. NR and RG reviewed the draft and prepared the figures. TA critically contributed to reviewing the manuscript. MV and AP developed the concept, contributed to drafting and critical review, and finally approved the manuscript.

The current article has been written as a part of the comprehensive literature review for conceptualizing a series of research projects on NAFLD modeling which is financially supported by grants received from Royan Institute for Stem Cell Biology and Technology, ACECR (Grant No. 99000166) and Shahid Beheshti University of Medical Sciences (Grant No. 43004196), Tehran, Iran. We would like to express our appreciation to our colleagues at Royan institute, Liver research group, particularly Dr. Zahra Farzaneh for their support.