hiPSC-derived macrophages improve drug sensitivity and selectivity in a macrophage-incorporating organoid culture model

Accurate simulation of different cell type interactions is crucial for physiological and precise in vitro drug testing. Human tissue-resident macrophages are critical for modulating disease conditions and drug-induced injuries in various tissues; however, their limited availability has hindered their use in in vitro modeling. Therefore, this study aimed to create macrophage-containing organoid co-culture models by directly incorporating human-induced pluripotent stem cell (hiPSC)-derived pre-macrophages into organoid and scaffold cell models. The fully differentiated cells in these organoids exhibited functional characteristics of tissue-resident macrophages with enriched pan-macrophage markers and the potential for M1/M2 subtype specialization upon cytokine stimulation. In a hepatic organoid model, the integrated macrophages replicated typical intrinsic properties, including cytokine release, polarization, and phagocytosis, and the co-culture model was more responsive to drug-induced liver injury than a macrophage-free model. Furthermore, alveolar organoid models containing these hiPSC-derived macrophages also showed increased drug and chemical sensitivity to pulmonary toxicants. Moreover, 3D adipocyte scaffold models incorporating macrophages effectively simulated in vivo insulin resistance observed in adipose tissue and showed improved insulin sensitivity on exposure to anti-diabetic drugs. Overall, the findings demonstrated that incorporating hiPSC-derived macrophages into organoid culture models resulted in more physiological and sensitive in vitro drug evaluation and screening systems.


Introduction
Macrophages are innate immune cells that mediate immune responses against external pathogens or substances [1].Specifically, tissue-resident macrophages, derived from yolk sac precursor cells or the bone marrow, are crucial for maintaining tissue homeostasis and initiating immune responses during injury or infection [2].Liver tissue-resident macrophages (Kupffer cells) are the largest tissueresident macrophage population that contribute to drug-induced liver injury (DILI) by secreting proinflammatory cytokines [3][4][5][6].Consequently, Kupffer cells are essential for modeling DILI and various liver diseases.Although previous studies have established a co-culture system involving primary hepatocytes and Kupffer cells [7,8], acquiring primary Kupffer cells from healthy individuals is challenging owing to their low yield and high cost.As an alternative, some studies have utilized stable cell lines derived from a patient with an acute monocytic leukemia, such as the immortalized monocyte cell line THP-1 that differentiates into macrophage-like cells.However, THP-1 cells are only partially suitable for toxicological assessments because they exhibit low expression of surface receptors involved in phagocytosis, cytokine release, and drug detoxification [9][10][11].In the lungs, the alveolar epithelium includes type I and II alveolar epithelial cells and macrophages, which interact with each other and play key roles in gas exchange, secreting pulmonary surfactants, and engulfing foreign particles.In drug-induced pulmonary toxicity, alveolar macrophages infiltrate the alveolar space and initiate tissue damage by inducing inflammation [12].Typically, pulmonary toxicity assessments mainly rely on in vivo animal (rodent) studies that partially simulate human conditions [13,14] mainly because of their distinct anatomy, but are also extremely timeconsuming, costly, and completely unsuitable for (high-throughput) screening.Several in vitro models, including mono-and complex co-cultures, have been proposed to evaluate drug toxicity after inhalation but have limitations because most do not represent the complex physiological tissue containing different cell types or use cancer-derived cells that have often lost physiological relevant properties [15][16][17].Adipose tissue is a well-recognized endocrine organ that influences both glucose and lipid metabolism by releasing adipokines [18][19][20][21].Given the prevalence of insulin resistance in individuals with type 2 diabetes, in vitro models that accurately mimic this physiological condition are lacking.Adipose tissue macrophages, which majorly contribute to the induction of insulin resistance, interfere with insulin signaling by secreting inflammatory cytokines [22][23][24][25][26][27].A previous study reported that 3D co-culture models of human adipose-derived mesenchymal stem cells (ADMSCs) mimic insulin resistance, allowing investigation of the effects of anti-obesity or anti-diabetic drugs [28].This study aimed to investigate whether human-induced pluripotent stem cell-derived macrophages (hiPSC-MPs) can be effectively used to improve in vitro model systems of liver, lung, and adipocytes.These cells offer advantages over primary cells by avoiding non-physiological changes often observed in stable cell lines like THP-1 or RAW264.7 cells, providing unlimited access that is challenging with primary cells from patients.Furthermore, we aimed to generate 3D models incorporating hiPSC-MPs that can differentiate into tissue-resident-like macrophages in the organoid with typical functional characteristics, including immune responses, to serve as advanced models for assessing drug safety and efficacy.Additionally, we demonstrated how our 3D models, incorporating adipose tissue derived from ADMSCs and hiPSC-MPs, effectively mimic insulin resistance, with the aim of providing valuable tools for in vitro drug screening to alleviate insulin resistance.

Ethics statement
All experiments involving human embryonic stem cells and hiPSCs were approved by the Public Institutional Bioethics Committee (approval numbers: P01-202102-41-001 and P01-201805-41-001) designated by the Ministry of Health and Welfare and Korea Centers for Disease Control and Prevention.All experiments were conducted following approved guidelines.

Co-culture of hiPSC-derived macrophages with HepG2 cells
HepG2 cells (ATCC, Manassas, VA, USA) were cultured in 100 cm 2 cell culture dishes in DMEM with high glucose and pyruvate content (Gibco) supplemented with 10% FBS (Gibco).A Transwell system (Corning, Tewksbury, MA, USA) was used to coculture HepG2 and ipreMPs.HepG2 cells were seeded in the lower compartment of the Transwell.After 1 d, ipreMPs were seeded on Transwell inserts and cultured for 72 h.The culture medium used Advanced DMEM/F12 supplemented with 1X GlutaMAX.The ratio of HepG2 cells to ipreMPs was 2:1 [7].

Fabrication of 3D ADMSC bead-type scaffold containing hiPSC-derived pre-macrophage model
Alginate bead scaffolds were prepared as previously described [33].The hydrogels were constructed using a 3D cell-printing system that included 20 mg ml −1 alginate, 0.5 mg ml −1 gelatin, and 0.5 mg ml −1 type I collagen (Sigma-Aldrich).The hydrogels were extruded through a 21 gauge needle into a 1% CaCl solution containing 5% FBS for 5 min to complete the alginate hydrogel crosslinking.Alginate beads containing ADMSC and ipreMPs were printed and cultured in the ADMSC medium for 12 d.The DM contained 100 g ml −1 insulin, 2 mM 1-methyl-3isobutylmethylxanthine, 4 µM dexamethasone, and 40 µM indomethacin for 8 d.The DM was then replaced with a medium containing 40 µg ml −1 insulin for 4 d.After the cells were fully differentiated, the medium was changed to an insulin-free medium containing 10% FBS, 1% mixture of 100 g ml −1 penicillin, and 100 g ml −1 streptomycin for 2 d.

Real-time quantitative PCR
Total RNA, extracted with TRIzol reagent (Thermo Fisher Scientific), underwent reverse transcription using GoScript™ Reverse Transcription Mix (Promega).Subsequent qPCR utilized GoTaq ® qPCR Master Mix (Promega) on a StepOnePlus Real-Time PCR System (Applied Biosystems).PCR results are presented as the relative fold change compared to controls, normalized to the geometric mean of glyceraldehyde-3-phosphate dehydrogenase (GAPDH).∆Ct values were calculated as the difference between Ct values for GAPDH and the target gene.The ∆Ct value of control cells served as the reference (∆Ct).Relative gene expression levels were determined using the formula 2-∆∆Ct.Primers used are listed in table S1.

Flow cytometry analysis
Cells were dissociated using Type LE (Gibco) and fixed with 4% formaldehyde.The samples were incubated with fluorescent-labeled primary antibodies for 10 min on ice and washed twice with 0.5% BSA (Sigma-Aldrich) in DPBS.Flow cytometry was performed using CytoFLEX (Beckman Coulter, Brea, California, USA), and the data was processed using Kaluza analysis software (Beckman Coulter).Positive gates were defined using the isotype controls.The antibodies used in the experiments are listed in table S2.

Immunohistochemistry
For immunofluorescence, the cells were fixed in 4% formaldehyde (Sigma-Aldrich) for 30 min at room temperature.Organoids were isolated from the Matrigel using a Cell Recovery solution (Corning), fixed in 4% formaldehyde for 30 min at RT, and washed with DPBST.For blocking, samples were resuspended in cold organoid washing buffer (OWB) containing 0.1% Triton X-100 and 0.2% BSA in DPBS and incubated for 30 min at 4 • C. The samples were then incubated overnight on a shaker with primary antibody-containing OWB at 4 • C. The fluorescein-conjugated secondary antibodycontaining OWB was added and incubated overnight at 4 • C. Three additional wash cycles were performed, followed by nuclear staining with 4 ′ -6-diamidino-2phenylindole (DAPI; Sigma-Aldrich) and mounting with ProLongTM Glass Antifade Mountant (Thermo Fisher Scientific).Images were captured using an Olympus FV3000 confocal microscope (Olympus) at 10× magnification.The antibodies used in the experiments are listed in table S2.
2.10.RNA-sequencing analysis mRNA libraries were generated using a TruSeq Stranded mRNA LT Sample Prep Kit, followed by sequencing on an Illumina NovaSeq 6000 platform according to the manufacturer's instructions.Differentially expressed genes and principal components were calculated using the var and prcomp functions in the stats (v3.6.1)R package.Euclidean distance measurements and the complete agglomeration method were applied for the hierarchical clustering.

Cytotoxicity assay
Cell viability assays utilized the CellTiter-Glo ® Cell Viability Assay (Promega) per the manufacturer's cytotoxicity assay instructions.The chemicals and drugs evaluated for cytotoxicity testing assays are listed in the supplementary information (see tables S3-5).HepG2 cells were seeded in 24-well Transwell plates (Corning) and treated with drugs after 24 h, while ipreMPs were seeded into Transwell inserts.After 24 h, cytotoxicity tests on HepG2 cells were conducted with CellTiter-Glo reagent, and luminescent signals were recorded using a Glomax (Promega).In 3D models, organoids were cultured in Ultra-lowattachment 96-well plates.Compounds, prepared in a basal medium at 0.1% DMSO concentration, were tested after 24 h, and cytotoxicity was assessed with CellTiter-Glo reagent.Obtained values, normalized to the vehicle control, were used to visualize the dose-response relationship via non-linear regression curves (GraphPad Prism 9, GraphPad Software).

Phagocytosis, enzyme-linked immunosorbent (ELISA), and lipid droplet formation assay
To evaluate the phagocytic function of hiPSC-derived macrophages (iMPs), Fluoresbrite ® Polychromatic Red Microspheres 1.0 µm (Polyscience) were used as a tracer of phagocytosis.For quantification of secreted cytokines and chemokines, commercially available ELISA kits were purchased and used for analysis (see table S6).BODIPY 493/503 (Invitrogen) was used for the lipid droplet formation assay.All assays were performed according to the manufacturer's instructions.

Western blot analysis and plasma membrane extraction
Cell lysates were treated with PRO-PREP protein extraction solution (iNtRON Biotechnology).Protein samples were loaded onto 4%-12% NuPage Bis-Tris Mini Gels (Invitrogen), transferred to a polyvinylidene difluoride membrane (Amersham Biosciences), and blocked with casein-blocking buffer solution (Sigma-Aldrich) for 15 min.Incubation with primary antibodies occurred at 4 • C overnight, followed by anti-rabbit secondary antibody (Thermo Fisher Scientific) incubation.Immunoreactive bands were visualized with SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) using chemiluminescence (BIORAD Inc.).Protein density analysis was performed with ImageJ software (National Institutes of Health).

Glucose uptake assay in the 3D co-culture model of ADMSC-derived adipocytes with ipreMPs
To determine glucose uptake, 3D bead-type scaffolds with monocultured or co-cultured cells were differentiated into adipocytes in 24-well tissue plates.Cellular glucose uptake was assessed using 2-NBDG, a fluorescent glucose analog, and an Abcam glucose uptake assay kit.After maturation, scaffold dissolution in a 10 mM EDTA solution, and cell suspensions were centrifuged, mixed, and incubated with HBSS, 2%v/w BSA, 80 µM 2-NBDG, and/or insulin.Following incubation, free 2-NBDG was washed out, and retained fluorescence was measured using a BIORAD fluorescence plate reader.Insulindependent 2-DG uptake levels were determined with a colorimetric glucose uptake assay kit (Abcam).

Statistical analysis
Data are presented as mean ± standard deviation (SEM) from two or three independent experiments performed in triplicate unless otherwise indicated.Graph visualization and statistical analyses were performed using GraphPad Prism version 9. The pvalues were calculated using ANOVA or multiple comparison t-test.

Generation and characterization of ipreMPs and iMPs
First, ipreMPs were generated as a source of tissue-resident macrophages according to a previously reported protocol [34].hiPSCs differentiated into MPs in a stepwise manner, including primitive streak and Hb induction, EMP differentiation, preMP specification, and MP maturation (figure 1(A)).To verify the differentiation process of macrophages, we measured the expression of lineage-specific markers.For each of the three stages (EMP, preMP, and MP), the expression of stage-specific markers in the differentiated cells significantly increased compared to undifferentiated hiPSCs (figure S1).CD45 + CX3CR1 + CD14 − cells emerged on day 14 of differentiation, giving rise to CD45 high CX3CR1 high CD14 + ipreMPs by day 25   [35].Thus, we examined whether the iMPs were polarized into distinct subtypes according to specific cytokine stimuli.In response to lipopolysaccharide (LPS) stimulation of M1 polarization, iMPs specifically increased proinflammatory cytokine secretion, such as interleukin (IL)-6, tumor necrosis factor-alpha (TNF-α), and IL-12 (figure 1(G), upper panel).In contrast, IL-4 exposure for M2 polarization stimulated the secretion of antiinflammatory cytokines such as IL-10, transforming growth factor-beta (TGF-β1), and IL-13 (figure 1(G), lower panel).Furthermore, polarized iMPs showed an increased expression of M1 (CD80 and CD86) and M2 (CD163 and CD206) surface markers (figure 1(H)) after stimulation with LPS or IL-4.These results suggest that iMPs were successfully generated and their in vivo functions were replicated.

Derivation of Kupffer-like cells from ipreMPs via co-culture of an in vitro hepatic cell model
Subsequently, we investigated whether ipreMPs could differentiate into liver-resident macrophages (Kupffer cells) by co-culturing them with the human hepatocellular carcinoma cell line HepG2 (figure 2(A)).
The expression of V-set and immunoglobulin domain containing 4 (VSIG4), macrophage receptor with collagenous structure (MACRO), CD169, vascular cell adhesion protein 1 (VCAM1), and CD163 is enriched in Kupffer cells in the human liver [36][37][38].Coculture of ipreMPs with HepG2 stimulated the differentiation of Kupffer-like cells (KLCs) with significantly higher expression of VSIG4, MARCO, CD169, and VCAM1 compared to a monoculture of ipreMPs, whereas the expressions of EMR1, CD68, CD16, and CD163 were either reduced or hardly effected (figure 2(B)).Flow cytometric analysis revealed that KLCs predominantly expressed VCAM1, CD169, and CD163 (figure 2(C)).Kupffer cells also polarize into M1 or M2 subtypes depending on the pathological conditions [39].Therefore, KLC polarization into the M1 and M2 subtypes after LPS or IL-4 exposure was investigated, and the results demonstrated that KLCs were polarized into M1 (secreting proinflammatory cytokines IL-6, TNF-α, and IL-12 after LPS exposure) and M2 (secreting anti-inflammatory cytokines IL-10, IL-13, and TGF-β1 after IL-4 exposure) (figure 2(D)).These findings demonstrate that ipreMPs can be efficiently specialized in KLCs when co-cultured with HepG2 cells.Transcriptomic profiles also confirmed the efficient specialization of iMPs into KLCs.Heat map analysis showed that Kupffer cell marker genes, such as MARCO, CD5 molecule-like (CD5L), and CD68, were predominantly expressed in KLCs compared to iMPs and THP-1-derived macrophages (THP-1-MPs) (figure 2(E), left panel).Additionally, the expression of both inflammatory and non-inflammatory macrophage marker genes was higher in KLCs than in THP-1-MPs and iMPs (figure 2(E), middle and right panels).Therefore, transcriptome analysis indicated that KLCs were more responsive to immune stimulation than THP-1-MPs.To determine the effect of coculture of macrophages on adverse drug reactions, we examined the drug responsiveness of HepG2 cells in the presence or absence of macrophages against nine DILI drugs (figure S3(A)).As a result of exposure to each hepatotoxic drug, HepG2 co-cultured with ipr-eMPs (HepG2-MP) showed significantly greater cytotoxicity than monocultures of HepG2 in most cases (figures S3(B) and 3(B), left panels).Notably, the liver toxicity of penicillamine was only detectable in the presence of macrophages.Trovafloxacin and troglitazone showed severe cytotoxicity in the presence of LPS stimulation.In contrast, the cytotoxic response of non-toxic structural analog drugs did not significantly increase (figure S3(C), right panels).In conclusion, specialized ipreMPs co-cultured with HepG2 cells and KLCs exhibited the characteristics of Kupffer cells, and this co-culture system could be a useful tool for assessing drug safety in vitro.

3D hHO-MP models for in vitro liver toxicity testing
A recent study demonstrated that hiPSC-derived hepatic organoids (hHOs) improved drug metabolic functions compared to HepG2 cells in an in vitro model for toxicity testing [31].To develop a more physiological and feasible model for toxicity testing than HepG2-MP co-culture, we designed 3D hHOs with incorporated macrophages (hHO-MP model).
The cells were dissociated from hHEOs at the proliferative progenitor stage before maturation into functional hepatic organoids and seeded with ipreMPs in 96-well ULA plates to generate 3D spherical structures (figure 3(A)).These 3D cell aggregates were further differentiated into functional hHO-MP models for 15 d according to a previously published protocol [31].In the hHO-MP models, the expression of hepatic marker genes, including ALB, AAT, and MRP2, was significantly higher than that in hHOs but not HNF4A (figure 3(B), upper panels).Notably, the expression of cytochrome P450 (CYP) genes, such as CYP2C19, CYP3A4, and CYP2D6, essential for metabolizing clinically used drugs, significantly increased in the hHO-MP models (figure 3(B), lower panels).Increased expression of Kupffer cell markers, such as CD206, CLEC4F, MARCO, and VSIG4, demonstrated that the ipreMPs further differentiated into specialized intrahepatic macrophages in a 3D .The distribution of CD68-and EMR1-positive cells and increased uptake of fluorescent beads also suggested that ipreMPs differentiated in functional intrahepatic macrophages (figures 3(D) and (E)).Furthermore, the hHO-MP models showed increased secretion of IL-6/TNF-α or IL-13/TGF-β1 in response to LPS or IL-4 stimulation, indicating that ipreMPs polarized into M1or M2-like subtypes (figure 3(F)).These results suggest that ipreMPs differentiated into functional MPs within the 3D structure and serve as liver-resident macrophages.To determine the effect of macrophages on drug responsiveness, we evaluated drug sensitivity and specificity in hHO-MP models for hepatotoxic drugs and their non-toxic structural analogs.Six DILI drugs, including acetaminophen, Dpenicillamine, and triptolide, known as immune cellmediated hepatotoxic drugs [38][39][40], were used to investigate the drug sensitivity of hHO-MP models.
In general, the organoids were more sensitive compared to HepG2, but more importantly and consistent with the results obtained in the HepG2-MP co-culture, the cytotoxicity of the tested drugs was detected at significantly lower concentrations in the hHO-MP group compared with that in the hHO group (figures 3(G) and (H)).Drug specificity for DILI drugs was also investigated by comparing the drug response in hHO-MP models to their non-toxic structural analogs.The hHO-MP model was more responsive to hepatotoxic drugs than the hHO model (figure 3(H), upper panel), whereas cytotoxicity to the non-toxic analogs such as levofloxacin, rosiglitazone, and fluconazole was not observed in either group and thus, was unaffected by the presence of macrophages (figure 3(H), lower panel).Exposure to hepatotoxic drugs increased the dose-dependent secretion of the proinflammatory cytokine IL-1β in the hHO-MP models but not in the absence of macrophages (figures S4(A) and (B)).In contrast to IL-1β, IL-6 was upregulated in the presence of macrophages regardless of drug exposure, and a concentrationdependent decrease in IL-6 secretion was observed only in the hHO-MP models exposed to hepatotoxic drugs (figures S4(C) and (D)).These results indicate that hHO-MP models are more sensitive to hepatotoxic drugs than macrophage-free hepatic organoids, probably mediated by the induction of inflammatory cytokine secretion, such as IL-1β, by the macrophages.

3D hLO-MP models comprising alveolar epithelial cells and lung-resident macrophages for in vitro pulmonary toxicity testing
We further investigated whether ipreMPs could develop into lung-resident macrophages within the alveolar lung organoids derived from hiPSCs (hLOs).
To generate a 3D lung organoid model incorporating macrophages (hLO-MP model), we first generated hLOs with abundant alveolar epithelial type 2 cells according to a recently reported protocol [32].The hLOs cultured in Matrigel droplets (D-hLOs) were dissociated into single cells, aggregated with ipreMPs in ULA plates, and then the mixture of cells selforganized into 3D structures and were further cultured for 5 d (figure 4(A)).The expressions of lungspecific marker genes, including SFTPC, MUC5AC, TP63, ABCA3, NKX2.1, and ACE2, were predominantly upregulated in the hLO-MP models compared to that in the D-hLOs (figure 4(B)).Notably, the hLO-MP models had significantly higher transcript levels of alveolar epithelial cell type 2 markers (SFTPC and ABCA3) and the goblet cell marker MUC5AC than the hLOs model (figure 4(B)).Increased expression of tissue-resident macrophage marker genes, such as CD68, CD169, CD14, AND CD206, indicated the presence of macrophages in the hLO-MP models (figure 4(C)).Additionally, immunofluorescence staining demonstrated that CD68-and EMR1positive macrophages were incorporated into the 3D structure of alveolar epithelial cells (figure 4(D)).
In the latex bead uptake assay, more fluorochromeconjugated latex beads accumulated in the hLO-MP models compared to the hLO models, suggesting that macrophages have phagocytic functions in the 3D spheroidal structure of alveolar epithelial cells (figure 4(E)).To determine the influence of macrophages on toxicants, we also evaluated the cytotoxic effects of pulmonary toxic drugs and chemicals with different concentrations on hLOs in the presence or absence of macrophages.Cytotoxicity to disinfectants used in humidifiers such as polyhexamethylene guanidine (PHMG) and chloromethylisothiazolinone/methylisothiazolinone (CMIT/MIT), as well as lung-toxic drugs such as bleomycin and amiodarone, was significantly more pronounced in the hLO-MP group than that in the hLOs group (figure 4(F)).Comparable to the results for heptic organoids, exposure to pulmonary toxicants increased the secretion of IL-1β in the hLO-MP group and this effect was dependent on the presence of macrophages (figure 4(H)).Additionally, we observed an elevated IL-6 secretion level in organoid-containing macrophages, which decreased when exposed to toxicants (figure 4(G)).In general, these results were highly consistent with those from the hHO-MP models.Notably, the hLO-MP models were more sensitive to pulmonary toxicants, suggesting that hLO-MP models can be used as a more physiologically relevant and sensitive model system for in vitro toxicity testing in different tissues.

3D AD-MP models comprising ADMSC-derived adipocytes (AD) and adipose tissue-resident macrophages for evaluating insulin-resistance drug effects
Macrophages in white adipose tissue produce inflammatory cytokines that regulate insulin resistance status [40].To determine whether iMPs could serve as functional mediators in the insulin resistance model, we investigated the impact of ipreMPs on the differentiation of ADMSC and their role in the onset of insulin resistance.To establish a 3D co-culture model of ADMSC-derived adipocytes with ipreMPs (AD-MP model), we employed alginate beads as a scaffold previously optimized for the ADMSC proliferation and differentiation and adipose cell-macrophage coculture [33,41].Figure 5(A) shows the 3D AD-MP model generated and differentiated in alginate scaffolds according to a previously described protocol [33].
After 12 d of differentiation, the increased expression of adipogenic markers with abundant lipid droplet formation throughout the 3D structure indicated the efficient differentiation of ADMSCs into mature adipocytes, even in the presence of macrophages (figures 5(B) and (C)).Differentiated adipocytes exhibited robust cell viability in alginate beads in the ADMSC-derived pre-adipocytes (pre-AD), AD, and AD-MP models.(figure 5(C)).Additionally, CD68 and CD163 expressions confirmed the specialization of tissue-resident macrophages of ipreMPs in the AD-MP models (figure 5(D)).Macrophage incorporation significantly increased the chemokine levels, including IL-8, CCL2, and CXCL1, which are closely associated with the onset of insulin resistance (figures 5(F) and (G)).The AD-MP models also showed a marked reduction in GLUT4 expression and attenuated AKT phosphorylation (figure 5(E)).Consequently, macrophage-stimulated inflammatory-related signals decreased GLUT4 expression by inhibiting the AKT signaling pathway in the AD-MP model, subsequently attenuating insulin sensitivity (figure 5(H)).Glucose uptake results using 2-NBDG showed that insulin sensitivity was significantly reduced in the AD-MP models (figure 5(H)).In the AD-MP Models, no significant improvement in glucose uptake was observed even after ADMSC differentiation, regardless of exposure to insulin (figure 5(H)).This is in contrast to the AD models, which show a significant increase in glucose uptake upon insulin stimulation.These results indicate that ipreMPs are indispensable for the onset of insulin resistance in ADMSC-derived adipocytes through proinflammatory cytokine-mediated suppression of the AKT pathway.In the next step, we investigated whether AD-MP models could be used for screening insulin resistance-mitigating compounds.Four commercially used drugs, metformin (a biguanide), GW9662 (a PPARγ antagonist), pioglitazone, and rosiglitazone (a PPARγ agonist), were evaluated in the AD-MP models to determine their effects on restoring insulin sensitivity.After 24 h, GW9662, pioglitazone, and rosiglitazone significantly improved glucose uptake to normal levels, but not metformin (figure 5(H)).Finally, CXCL1 secretion was also reduced by all four drugs; however, IL-8 and CCL2 concentrations were only inhibited in the rosiglitazone-treated AD-MP models (figure 5(I)), consistent with previous reports [19,42].These results suggest that including iPSC-derived macrophages in ADMSC could be useful for establishing a more predictive model for studying drug effects on insulin resistance.

Discussion
This study reports the use of hiPSC-derived macrophages as tissue-resident macrophages for improved in vitro drug efficacy and safety evaluation.Remarkable progress has been made in the development of cell-based assays by adapting 3D cell culture technology, including the more recent organoid and organ-on-a-chip approaches, to improve the predictive power of in vitro models of human disease and target tissue toxicity [43].However, the lack of non-parenchymal cells, particularly tissueresident macrophages, in the cellular model system has often limited our understanding of the paracrine and inflammation-related signaling-mediated cellular responses that contribute to the onset of disease and drug-induced injury.Recent evidence suggests that toxicant-induced tissue damage is exacerbated by cross-talk between macrophages and parenchymal cells directly affected by various toxicants [44].This led to the development of an advanced cell culture model system replicating in vivo macrophageparenchymal cell interactions involved in tissue injury progression.Currently, various methods have been reported for the successful development of macrophage-incorporating organoid culture models, ranging from 3D spheroids to scaffolded systems, using different macrophage cell sources [45,46].In particular, recently developed macrophageincorporating intestinal organoid culture models provided a framework to facilitate the development of hPSC-derived organoid systems bearing resident macrophages.One such approach involves the generation of hiPSC-derived human colonic organoids (HCOs) with macrophages, where the codeveloping hemogenic endothelial (HE)-like cells yield functional macrophages that resemble resident human fetal intestinal macrophages [47].These HCO macrophages acquired a transcriptional signature reminiscent of human fetal intestinal and colonic resident macrophages and exhibited functional behaviors in terms of modulating cytokine secretion in response to inflammatory stimuli as well as phagocytosis of pathogenic bacteria [47].Despite the fact that this study provides a compelling model for studying colonic parenchymaresident macrophage interactions, its application is restricted to colonic organoids since macrophages can only be generated during HCO development induced by specific signaling manipulations.In another intestinal model, hPSC-derived macrophages S Jo et al generated to mimic fetal development were successfully incorporated into intestinal organoids by migration during in vitro culture [48].Macrophages incorporated into the intestinal organoids recapitulated aspects of microanatomical localization and acquired the similar transcriptomic profiles observed in developing intestinal macrophages.Although the functional aspects of incorporated macrophages within intestinal organoids were not comprehensively determined in vitro, they significantly contributed to the regulation of metabolism and growth of the developing intestinal organoids in transplanted mice [48].
Our hiPSC-based approach has several advantages over recent methods for integrating macrophages into organoid systems.Notably, mixing the partially differentiated cell is simpler, and it allows self-organization and congruent final differentiation of the two cell types.Additionally, combining the two cell populations in one organoid better represents the physiological conditions compared to a co-cultured transwell system.This could allow researchers to reduce the costly and time-consuming steps required in traditional co-culture methods and improve scaleup for use in high-throughput assays.Furthermore, co-cultivation of macrophages and organoids derived from the same hiPSC line may help in preventing unanticipated background immune reactions caused by donor mismatch and offer the possibility to generate organoids from patients to analyze diseaserelated differences in drug responses.In this study, we generated tissue-resident macrophages directly from hiPSCs and then incorporated them into organoids and scaffolded cell models.In our proposed 3D models, macrophages are integrated through the direct addition of ipreMPs to proliferating organoids at intermediate differentiation stages.Dissociated hHEO and hLO constituent cells mixed with ipr-eMPs self-organized in ULA plates and differentiated under DM.iMPs demonstrated intrinsic phagocytic functions and the potential to polarize into M1 and M2 subpopulations in vitro upon external cytokine stimulation, with highly expressed pan-macrophage markers.This intrinsic property of iMPs to respond to external stimuli persisted even after they were specialized as resident macrophages incorporated in the 3D models.Furthermore, the incorporation of ipreMPs had a positive effect on the maturation of liver and lung organoid models (figures 3(B) and 4(B)).A previous study showed that iPSC-derived primitive macrophages were able to specialize into tissue-resident macrophage-like cells in response to organ-specific cues from co-cultured iPSC-derived neurons in vitro and neighboring tissues where they were transplanted in vivo [49].Co-culture of hiPSC-derived macrophages and hepatocytes also had synergistic effects on the upregulation of marker genes in both cell types during cultivation [45].On the other hand, several studies have shown that pro-inflammatory signals secreted by macrophages, such as IL-6 and TNF-α, promote the expansion of hepatocytes and enable long-term culture in vitro [50,51].Together with our results, these observations suggest that the crosstalk between the co-cultured developing organoids and the iMPs, which may serve as tissueor cell-specific cues for maturation, is critical to facilitate the specification of both organoids and macrophages.
The modulatory effect of macrophages in regulating drug-metabolizing enzymes and acute-phase proteins via the release of inflammation-related cytokines and chemokines leads to altered drug and chemical susceptibility in co-cultured hepatocytes [8,45,52].Consistent with previous reports, our initial evaluation of the HepG2-ipreMP co-culture model showed that ipreMPs significantly enhanced the cytotoxic effects of nine DILI drugs in HepG2 cells during LPS stimulation (figures S3(B) and (C)).This increased drug susceptibility for the DILI drugs was also observed in the hHO-MP model, in which the ipreMPs were directly incorporated during differentiation (figures 3(G) and (H)).Notably, our donor-matched hHO-MP model showed increased drug susceptibility to all tested drugs without LPS stimulation, unlike in previous reports [45].This requires further investigation to elucidate the relevance between the differences in drug responsiveness in hepatocytes and the background immune response of co-cultured macrophages from different sources (cell lines and primary cells) that might also represent distinct differentiation status with different genetic backgrounds.Increasing evidence suggests that certain proinflammatory cytokines, including IL-1β, IL-6, and TNF-α, secreted by activated macrophages, exacerbate the adverse effects of toxicants on co-cultured hepatocytes [24,45,53].On activation of macrophages through LPS exposure, IL-6 secretion was significantly upregulated in the hHO-MP model regardless of drug exposure but decreased with the concentration of DILI drugs (figures S4(C) and (D)).Unlike IL-6, IL-1β secretion was much higher upon exposure to DILI drugs than LPS (figures S4(A) and (B)).Furthermore, no significant increase in the level of IL-1β release was observed in the hHO-MP model exposed to nontoxic structural analogs such as levofloxacin, rosiglitazone, and fluconazole (figure S4(B)).In contrast, the elevated IL-6 secretion level was maintained independent of exposure to non-toxic structural analogs (figure S4(D)).A significant increase in TNF-α release was also observed in hHO-MPs upon LPS stimulation; however, the concentrationdependent TNF-α release patterns varied from drug to drug (data not shown).These results suggest that activated macrophages may sense toxic drug metabolites and/or associated subtle microenvironmental changes and subsequently upregulate IL-1β release, leading to an enhanced cytotoxic effect in the hHO-MP model.This is further supported by the results from the hLO-MP models exposed to pulmonary toxicants: IL-1β secretion, but not IL-6, was significantly increased upon exposure to toxicants in the model system (figure 4).Although IL-1β release was greatly increased in the absence of macrophages, IL-1β release was dependent on exposure to toxicants, consistent with the results in hHO-MPs (figures 4(G) and (H)).In contrast to IL-1β release, IL-6 levels decreased in a concentration-dependent manner after exposure to toxicants, as shown in the lungs of bleomycin-treated mice [45,46].
Adipose tissue macrophages play a key role in obesity-induced insulin resistance by mediating inflammatory-related signaling pathways and inducing macrophage recruitment and phenotype switching [26,54].Several in vitro models of adipocyte-macrophage interactions have been developed to mimic insulin resistance in adipose tissue observed in vivo [33,41].More recently, cocultures of human adipocytes and macrophages derived from rodent or human monocytes in 3D alginate scaffolds have successfully recapitulated obesity-induced insulin resistance to investigate the effects of anti-diabetic drugs [26].ADMSC scaffolds seeded with ipreMPs also differentiated into functional adipose-like tissue despite the presence of macrophages.Notably, the ipreMPs incorporated into the 3D structure acted as functional tissueresident macrophage-like cells without the addition of any specific growth factors, supplements, or culture medium for macrophage maturation (figures 3(D)-(F), 4(C)-(E), and figure 5(D)).In the presence of ipreMPs, ADMSCs were efficiently differentiated into mature adipocytes containing a multitude of lipid droplets, although the expression of adipocyte markers slightly decreased (figures 5(B) and (C)).Consistent with a previous report [41], ipr-eMPs inhibited GLUT4 expression via AKT phosphorylation at 1% concentration in the AD-MP model (figure 5(E)), significantly suppressing insulin sensitivity (figure 5(H)).Subsequent evaluation of T2DM drugs suggests the applicability of AD-MP models for screening insulin resistance-alleviating compounds as a potential in vitro model.Exposure to PPARγ modulators, including GW9662, pioglitazone, and rosiglitazone, significantly improved glucose uptake in the AD-MP model (figure 5(H)).Pioglitazone and rosiglitazone restored the insulin sensitivity in AP-MP models to the level observed in the absence of macrophages (figure 5(H)).These two PPARγ agonists are known to improve insulin resistance in animal models of T2DM by restoring the decreased GLUT4 expression in adipose tissue [55][56][57].In contrast, no significant change in glucose uptake was observed in the AD-MP models after exposure to metformin (figure 5(H)).Previous studies suggest that the glucose-lowering effects of metformin in T2DM are secondary to decreased hepatic glucose synthesis and increased skeletal muscle glucose disposal [58][59][60].Furthermore, the molecular mechanism underlying the action of metformin remains poorly understood.Therefore, the efficacy of metformin in improving insulin resistance has not been confirmed in AD-MP models.Our AD-MP models had limitations in recapitulating in vivo organ-to-organ connectivity and immune cell recruitment involved in obesity-induced insulin resistance.However, despite these limitations, this model may still allow efficient screening of compounds that directly improve insulin sensitization in adipose tissues.

Conclusion
In conclusion, this study established various macrophage-incorporating organoid culture models using ipreMPs, confirmed that the macrophages in each tissue differentiated into tissue-resident-like functional macrophages, and improved the applicability of organoid models in a drug evaluation system.Notably, a drug assessment liver model containing immune cells was developed to accurately mimic the physiological environment during drug-induced hepatotoxicity, including immune cell-mediated exacerbation of hepatotoxicity.Previous studies on drug testing for lung diseases were mainly conducted in vivo using rodents, are time-consuming and costeffective, heavily discussed because of animal welfare issues, and limited because of interspecies differences.Similarly, most studies on insulin resistance and diabetes treatment-related drug testing relied heavily on in vivo experiments.Therefore, this study is significant because of the in vitro experiments that provide a more human-relevant and controlled environment.Moreover, developing in vitro models that mimic the in vivo environment has become increasingly important, considering the increasing ethical and regulatory challenges associated with animal experiments.Although continuous improvement of such models is necessary, they currently offer a significant opportunity to conduct various studies on drug action mechanisms in vitro.

Figure 3 .
Figure 3. of 3D hHO-MP models for in vitro liver toxicity testing.(A) Schematic of 3D hHO-MP model culture system.(B) Relative expression of hepatocyte marker genes in the hHEOs, hHOs, and hHO-MP models.Data are presented as mean ± SEM (n = 3) and compared to hHEOs.(C) Relative expression of liver tissue-resident macrophage marker genes in the hHOs and hHO-MP models.Data are presented as mean ± SEM (n = 3) and compared to hHOs.(D) IF images of hepatocyte (AAT) and macrophage (CD68 and EMR1) markers after differentiation.Cell nuclei were stained with DAPI.(E) Phagocytosis assay via uptake of fluorescent latex beads.(F) Secretion level of proinflammatory (IL-6 and TNF-α) and antiinflammatory (IL-13 and TGF-β) cytokines after LPS or IL-4 stimulation in hHO-MP models.Data are presented as mean ± SEM (n = 3) and compared to control.(G) Toxicity assessment of hepatotoxic drugs and non-toxic structural analogs in hHOs and hHO-MP models.Data are represented as mean ± SEM (n = 3).Scale bars: 100 µm.ns, not significant, * p < 0.1, * * p < 0.01, * * * p < 0.001, * * * * p < 0.0001.

Figure 4 .
Figure 4. Establishment of 3D hLO-MP models for in vitro pulmonary toxicity testing.(A) Schematic of 3D hLO-MP model culture system.(B) Relative expression level of alveolar epithelial marker genes in the D-hLOs, hLO, and hLO-MP models.Data are presented as mean ± SEM (n = 3) and compared with the D-hLOs model.(C) Relative expression of alveolar tissue-resident macrophage marker genes in the hLO and hLO-MP models.Data are presented as mean ± SEM (n = 3) and compared to the hLO models.(D) IF images of the alveolar epithelial cell (SFTPB) and macrophage (CD68 and EMR1) markers in the hLO-MP models.(E) Phagocytosis assay via uptake of fluorescent latex beads.(F) Toxicity assessment of lung toxic drugs in the hLO and hLO-MP models.Data are presented as mean ± SEM (n = 3) and compared with the hLO model.(G) IL-6 secretion levels in the hLOs and hLO-MP models treated with pulmonary toxicants for 24 h.(H) IL-1β secretion levels in the hLOs and hLO-MP models treated with pulmonary toxicants for 24 h.Scale bars: 100 µm.ns, not significant, * p < 0.1, * * p < 0.01, * * * p < 0.001, * * * * p < 0.0001.