Automatic classification of experimental models in biomedical literature to support searching for alternative methods to animal experiments.

Current animal protection laws require replacement of animal experiments with alternative methods, whenever such methods are suitable to reach the intended scientific objective. However, searching for alternative methods in the scientific literature is a time-consuming task that requires careful screening of an enormously large number of experimental biomedical publications. The identification of potentially relevant methods, e.g. organ or cell culture models, or computer simulations, can be supported with text mining tools specifically built for this purpose. Such tools are trained (or fine tuned) on relevant data sets labeled by human experts. We developed the GoldHamster corpus, composed of 1,600 PubMed (Medline) articles (titles and abstracts), in which we manually identified the used experimental model according to a set of eight labels, namely: "in vivo", "organs", "primary cells", "immortal cell lines", "invertebrates", "humans", "in silico" and "other" (models). We recruited 13 annotators with expertise in the biomedical domain and assigned each article to two individuals. Four additional rounds of annotation aimed at improving the quality of the annotations with disagreements in the first round. Furthermore, we conducted various machine learning experiments based on supervised learning to evaluate the corpus for our classification task. We obtained more than 7,000 document-level annotations for the above labels. After the first round of annotation, the inter-annotator agreement (kappa coefficient) varied among labels, and ranged from 0.42 (for "others") to 0.82 (for "invertebrates"), with an overall score of 0.62. All disagreements were resolved in the subsequent rounds of annotation. The best-performing machine learning experiment used the PubMedBERT pre-trained model with fine-tuning to our corpus, which gained an overall f-score of 0.83. We obtained a corpus with high agreement for all labels, and our evaluation demonstrated that our corpus is suitable for training reliable predictive models for automatic classification of biomedical literature according to the used experimental models. Our SMAFIRA - "Smart feature-based interactive" - search tool ( https://smafira.bf3r.de ) will employ this classifier for supporting the retrieval of alternative methods to animal experiments. The corpus is available for download ( https://doi.org/10.5281/zenodo.7152295 ), as well as the source code ( https://github.com/mariananeves/goldhamster ) and the model ( https://huggingface.co/SMAFIRA/goldhamster ).

Shedding light into the black box: Advances in in vitro systems for studying implantation.

Implantation represents a critical step for embryonic development and pregnancy. Its success depends on the complex interplay between a receptive endometrium and a competent embryo. Implantation-related events remain hardly accessible, making implantation a true "black box" in developmental biology. Improved in vitro models are becoming useful experimental tools, as they are considerably more accessible than in vivo models, easier to manipulate, and permit the use of human cells or tissues, thus increasing the translational value of the studies. In this Review, we briefly summarize the relevant cell types and structures involved into the process of implantation, in order to outline which compartments are indispensable for creating the perfect in vitro model. We also critically address advantages and limitations of available models and assess their application potential. Moreover, we examine the chances and challenges brought by the latest approaches to recapitulate the endometrial compartment, as well as by peri-implantational embryoids.

Metabolism of remimazolam in primary human hepatocytes during continuous long-term infusion in a 3-D bioreactor system.

BACKGROUND: Remimazolam is an ultra-short acting benzodiazepine under development for procedural sedation and general anesthesia. It is hydrolyzed by CES1 to an inactive metabolite (CNS7054). PURPOSE: In this study, the effect of continuous remimazolam exposure on its metabolism and on CES1 expression was investigated in a dynamic 3-D bioreactor culture model inoculated with primary human hepatocytes. METHODS: Remimazolam was continuously infused into bioreactors for 5 days at a final concentration of 3,000 ng/ml (6.8 µM). In parallel, 2-D cultures were run with cells from the same donors, but with discontinuous exposure to remimazolam. RESULTS: Daily measurement of clinical chemistry parameters (glucose, lactate, urea, ammonia, and liver enzymes) in culture supernatants indicated no noxious effect of remimazolam on hepatocyte integrity as compared to untreated controls. Concentrations of remimazolam reached steady-state values of around 250 ng/ml within 8 hours in 3-D bioreactors whereas in 2-D cultures remimazolam concentrations declined to almost zero within the same time frame. Levels of CNS7054 showed an inverse time-course reaching average values of 1,350 ng/ml in perfused 3-D bioreactors resp. 2,800 ng/ml in static 2-D cultures. Analysis of mRNA expression levels of CES1 indicated no changes in gene expression over the culture period. CONCLUSION: The results indicated a stable metabolism of remimazolam during 5 days of continuous exposure to clinically relevant concentrations of the drug. Moreover, there was no evidence for a harmful effect of remimazolam exposure on the integrity and metabolic activity of in vitro cultivated primary human hepatocytes.

Microscale 3D Liver Bioreactor for In Vitro Hepatotoxicity Testing under Perfusion Conditions.

The accurate prediction of hepatotoxicity demands validated human in vitro models that can close the gap between preclinical animal studies and clinical trials. In this study we investigated the response of primary human liver cells to toxic drug exposure in a perfused microscale 3D liver bioreactor. The cellularized bioreactors were treated with 5, 10, or 30 mM acetaminophen (APAP) used as a reference substance. Lactate production significantly decreased upon treatment with 30 mM APAP (p < 0.05) and ammonia release significantly increased in bioreactors treated with 10 or 30 mM APAP (p < 0.0001), indicating APAP-induced dose-dependent toxicity. The release of prostaglandin E2 showed a significant increase at 30 mM APAP (p < 0.05), suggesting an inflammatory reaction towards enhanced cellular stress. The expression of genes involved in drug metabolism, antioxidant reactions, urea synthesis, and apoptosis was differentially influenced by APAP exposure. Histological examinations revealed that primary human liver cells in untreated control bioreactors were reorganized in tissue-like cell aggregates. These aggregates were partly disintegrated upon APAP treatment, lacking expression of hepatocyte-specific proteins and transporters. In conclusion, our results validate the suitability of the microscale 3D liver bioreactor to detect hepatotoxic effects of drugs in vitro under perfusion conditions.

Hepatic differentiation of human iPSCs in different 3D models: A comparative study.

Human induced pluripotent stem cells (hiPSCs) are a promising source from which to derive distinct somatic cell types for in vitro or clinical use. Existent protocols for hepatic differentiation of hiPSCs are primarily based on 2D cultivation of the cells. In the present study, the authors investigated the generation of hiPSC-derived hepatocyte-like cells using two different 3D culture systems: A 3D scaffold-free microspheroid culture system and a 3D hollow-fiber perfusion bioreactor. The differentiation outcome in these 3D systems was compared with that in conventional 2D cultures, using primary human hepatocytes as a control. The evaluation was made based on specific mRNA expression, protein secretion, antigen expression and metabolic activity. The expression of α-fetoprotein was lower, while cytochrome P450 1A2 or 3A4 activities were higher in the 3D culture systems as compared with the 2D differentiation system. Cells differentiated in the 3D bioreactor showed an increased expression of albumin and hepatocyte nuclear factor 4α, as well as secretion of α-1-antitrypsin as compared with the 2D differentiation system, suggesting a higher degree of maturation. In contrast, the 3D scaffold-free microspheroid culture provides an easy and robust method to generate spheroids of a defined size for screening applications, while the bioreactor culture model provides an instrument for complex investigations under physiological-like conditions. In conclusion, the present study introduces two 3D culture systems for stem cell derived hepatic differentiation each demonstrating advantages for individual applications as well as benefits in comparison with 2D cultures.

Effects of Co-Culture Media on Hepatic Differentiation of hiPSC with or without HUVEC Co-Culture.

The derivation of hepatocytes from human induced pluripotent stem cells (hiPSC) is of great interest for applications in pharmacological research. However, full maturation of hiPSC-derived hepatocytes has not yet been achieved in vitro. To improve hepatic differentiation, co-cultivation of hiPSC with human umbilical vein endothelial cells (HUVEC) during hepatic differentiation was investigated in this study. In the first step, different culture media variations based on hepatocyte culture medium (HCM) were tested in HUVEC mono-cultures to establish a suitable culture medium for co-culture experiments. Based on the results, two media variants were selected to differentiate hiPSC-derived definitive endodermal (DE) cells into mature hepatocytes with or without HUVEC addition. DE cells differentiated in mono-cultures in the presence of those media variants showed a significant increase (p < 0.05) in secretion of α-fetoprotein and in activities of cytochrome P450 (CYP) isoenzymes CYP2B6 and CYP3A4 as compared with cells differentiated in unmodified HCM used as control. Co-cultivation with HUVEC did not further improve the differentiation outcome. Thus, it can be concluded that the effect of the used medium outweighed the effect of HUVEC co-culture, emphasizing the importance of the culture medium composition for hiPSC differentiation.

The Cell-Surface N-Glycome of Human Embryonic Stem Cells and Differentiated Hepatic Cells thereof.

Human embryonic stem cells (hESCs) are pluripotent stem cells that offer a wide range of applications in regenerative medicine. In addition, they have been proposed as an appropriate alternative source of hepatocytes. In this work, hESCs were differentiated into definitive endodermal cells (DECs), followed by maturation into hepatocyte-like cells (HLCs). Their cell-surface N-glycome was profiled and also compared with that of primary human hepatocytes (PHHs). Undifferentiated hESCs contained large amounts of high-mannose N-glycans. In contrast, complex-type N-glycans such as asialylated or monosialylated biantennary and triantennary N-glycans were dominant in HLCs, and fully galactosylated structures were significantly more abundant than in undifferentiated hESCs. The cell-surface N-glycosylation of PHHs was more biologically processed than that of HLCs, with bisialylated biantennary and trisialylated triantennary structures predominant. This is the first report of the cell surface N-glycome of PHHs and of HLCs being directly generated from hESCs without embryoid body formation.

Global Transcriptional Response of Human Liver Cells to Ethanol Stress of Different Strength Reveals Hormetic Behavior.

BACKGROUND: The liver is the major site for alcohol metabolism in the body and therefore the primary target organ for ethanol (EtOH)-induced toxicity. In this study, we investigated the in vitro response of human liver cells to different EtOH concentrations in a perfused bioartificial liver device that mimics the complex architecture of the natural organ. METHODS: Primary human liver cells were cultured in the bioartificial liver device and treated for 24 hours with medium containing 150 mM (low), 300 mM (medium), or 600 mM (high) EtOH, while a control culture was kept untreated. Gene expression patterns for each EtOH concentration were monitored using Affymetrix Human Gene 1.0 ST Gene chips. Scaled expression profiles of differentially expressed genes (DEGs) were clustered using Fuzzy c-means algorithm. In addition, functional classification methods, KEGG pathway mapping and also a machine learning approach (Random Forest) were utilized. RESULTS: A number of 966 (150 mM EtOH), 1,334 (300 mM EtOH), or 4,132 (600 mM EtOH) genes were found to be differentially expressed. Dose-response relationships of the identified clusters of co-expressed genes showed a monotonic, threshold, or nonmonotonic (hormetic) behavior. Functional classification of DEGs revealed that low or medium EtOH concentrations operate adaptation processes, while alterations observed for the high EtOH concentration reflect the response to cellular damage. The genes displaying a hormetic response were functionally characterized by overrepresented "cellular ketone metabolism" and "carboxylic acid metabolism." Altered expression of the genes BAHD1 and H3F3B was identified as sufficient to classify the samples according to the applied EtOH doses. CONCLUSIONS: Different pathways of metabolic and epigenetic regulation are affected by EtOH exposition and partly undergo hormetic regulation in the bioartificial liver device. Gene expression changes observed at high EtOH concentrations reflect in some aspects the situation of alcoholic hepatitis in humans.

Self-assembled 3D spheroids and hollow-fibre bioreactors improve MSC-derived hepatocyte-like cell maturation in vitro.

3D cultures of human stem cell-derived hepatocyte-like cells (HLCs) have emerged as promising models for short- and long-term maintenance of hepatocyte phenotype in vitro cultures by better resembling the in vivo environment of the liver and consequently increase the translational value of the resulting data. In this study, the first stage of hepatic differentiation of human neonatal mesenchymal stem cells (hnMSCs) was performed in 2D monolayer cultures for 17 days. The second stage was performed by either maintaining cells in 2D cultures for an extra 10 days, as control, or alternatively cultured in 3D as self-assembled spheroids or in multicompartment membrane bioreactor system. All systems enabled hnMSC differentiation into HLCs as shown by positive immune staining of hepatic markers CK-18, HNF-4α, albumin, the hepatic transporters OATP-C and MRP-2 as well as drug-metabolizing enzymes like CYP1A2 and CYP3A4. Similarly, all models also displayed relevant glucose, phase I and phase II metabolism, the ability to produce albumin and to convert ammonia into urea. However, EROD activity and urea production were increased in both 3D systems. Moreover, the spheroids revealed higher bupropion conversion, whereas bioreactor showed increased albumin production and capacity to biotransform diclofenac. Additionally, diclofenac resulted in an IC50 value of 1.51 ± 0.05 and 0.98 ± 0.03 in 2D and spheroid cultures, respectively. These data suggest that the 3D models tested improved HLC maturation showing a relevant biotransformation capacity and thus provide more appropriate reliable models for mechanistic studies and more predictive systems for in vitro toxicology applications.

Hepatic Differentiation of Human Induced Pluripotent Stem Cells in a Perfused Three-Dimensional Multicompartment Bioreactor.

The hepatic differentiation of human induced pluripotent stem cells (hiPSC) holds great potential for application in regenerative medicine, pharmacological drug screening, and toxicity testing. However, full maturation of hiPSC into functional hepatocytes has not yet been achieved. In this study, we investigated the potential of a dynamic three-dimensional (3D) hollow fiber membrane bioreactor technology to improve the hepatic differentiation of hiPSC in comparison to static two-dimensional (2D) cultures. A total of 100 × 10(6) hiPSC were seeded into each 3D bioreactor (n = 3). Differentiation into definitive endoderm (DE) was induced by adding activin A, Wnt3a, and sodium butyrate to the culture medium. For further maturation, hepatocyte growth factor and oncostatin M were added. The same differentiation protocol was applied to hiPSC maintained in 2D cultures. Secretion of alpha-fetoprotein (AFP), a marker for DE, was significantly (p < 0.05) higher in 2D cultures, while secretion of albumin, a typical characteristic for mature hepatocytes, was higher after hepatic differentiation of hiPSC in 3D bioreactors. Functional analysis of multiple cytochrome P450 (CYP) isoenzymes showed activity of CYP1A2, CYP2B6, and CYP3A4 in both groups, although at a lower level compared to primary human hepatocytes (PHH). CYP2B6 activities were significantly (p < 0.05) higher in 3D bioreactors compared with 2D cultures, which is in line with results from gene expression. Immunofluorescence staining showed that the majority of cells was positive for albumin, cytokeratin 18 (CK18), and hepatocyte nuclear factor 4-alpha (HNF4A) at the end of the differentiation process. In addition, cytokeratin 19 (CK19) staining revealed the formation of bile duct-like structures in 3D bioreactors similar to native liver tissue. The results indicate a better maturation of hiPSC in the 3D bioreactor system compared to 2D cultures and emphasize the potential of dynamic 3D culture systems in stem cell differentiation approaches for improved formation of differentiated tissue structures.

Cell sources for in vitro human liver cell culture models.

In vitro liver cell culture models are gaining increasing importance in pharmacological and toxicological research. The source of cells used is critical for the relevance and the predictive value of such models. Primary human hepatocytes (PHH) are currently considered to be the gold standard for hepatic in vitro culture models, since they directly reflect the specific metabolism and functionality of the human liver; however, the scarcity and difficult logistics of PHH have driven researchers to explore alternative cell sources, including liver cell lines and pluripotent stem cells. Liver cell lines generated from hepatomas or by genetic manipulation are widely used due to their good availability, but they are generally altered in certain metabolic functions. For the past few years, adult and pluripotent stem cells have been attracting increasing attention, due their ability to proliferate and to differentiate into hepatocyte-like cells in vitro However, controlling the differentiation of these cells is still a challenge. This review gives an overview of the major human cell sources under investigation for in vitro liver cell culture models, including primary human liver cells, liver cell lines, and stem cells. The promises and challenges of different cell types are discussed with a focus on the complex 2D and 3D culture approaches under investigation for improving liver cell functionality in vitro Finally, the specific application options of individual cell sources in pharmacological research or disease modeling are described.

In Vitro Model for Hepatotoxicity Studies Based on Primary Human Hepatocyte Cultivation in a Perfused 3D Bioreactor System.

Accurate prediction of the potential hepatotoxic nature of new pharmaceuticals remains highly challenging. Therefore, novel in vitro models with improved external validity are needed to investigate hepatic metabolism and timely identify any toxicity of drugs in humans. In this study, we examined the effects of diclofenac, as a model substance with a known risk of hepatotoxicity in vivo, in a dynamic multi-compartment bioreactor using primary human liver cells. Biotransformation pathways of the drug and possible effects on metabolic activities, morphology and cell transcriptome were evaluated. Formation rates of diclofenac metabolites were relatively stable over the application period of seven days in bioreactors exposed to 300 µM diclofenac (300 µM bioreactors (300 µM BR)), while in bioreactors exposed to 1000 µM diclofenac (1000 µM BR) metabolite concentrations declined drastically. The biochemical data showed a significant decrease in lactate production and for the higher dose a significant increase in ammonia secretion, indicating a dose-dependent effect of diclofenac application. The microarray analyses performed revealed a stable hepatic phenotype of the cells over time and the observed transcriptional changes were in line with functional readouts of the system. In conclusion, the data highlight the suitability of the bioreactor technology for studying the hepatotoxicity of drugs in vitro.

Periodic harvesting of embryonic stem cells from a hollow-fiber membrane based four-compartment bioreactor.

Different types of stem cells have been investigated for applications in drug screening and toxicity testing. In order to provide sufficient numbers of cells for such in vitro applications a scale-up of stem cell culture is necessary. Bioreactors for dynamic three-dimensional (3D) culture of growing cells offer the option for culturing large amounts of stem cells at high densities in a closed system. We describe a method for periodic harvesting of pluripotent stem cells (PSC) during expansion in a perfused 3D hollow-fiber membrane bioreactor, using mouse embryonic stem cells (mESC) as a model cell line. A number of 100 × 10(6) mESC were seeded in bioreactors in the presence of mouse embryonic fibroblasts (MEF) as feeder cells. Over a cultivation interval of nine days cells were harvested by trypsin perfusion and mechanical agitation every second to third culture day. A mean of 380 × 10(6) mESC could be removed with every harvest. Subsequent to harvesting, cells continued growing in the bioreactor, as determined by increasing glucose consumption and lactate production. Immunocytochemical staining and mRNA expression analysis of markers for pluripotency and the three germ layers showed a similar expression of most markers in the harvested cells and in mESC control cultures. In conclusion, successful expansion and harvesting of viable mESC from bioreactor cultures with preservation of sterility was shown. The present study is the first one showing the feasibility of periodic harvesting of adherent cells from a continuously perfused four-compartment bioreactor including further cultivation of remaining cells.

Feasibility study of an active wound dressing based on hollow fiber membranes in a porcine wound model.

INVESTIGATION: A novel active wound dressing (AWD) concept based on a microporous hollow fiber membrane network was investigated in an animal model. It provides a local solution-perfused environment for regenerative cell nutrition, wound irrigation, debris removal, electrolyte balancing, pH regulation, and topical antibiosis. The device is capable of supplying soluble factors, as tested experimentally for the recombinant human growth and differentiation factor-5 (rhGDF-5). METHODS: Following in vitro studies for rhGDF-5 using primary human keratinocytes and dermal fibroblasts, we employed a porcine partial thickness wound model with five distinct wounds on each back of n=8 pigs. Four wound groups were perfused differently over 9 days and compared with a negative control wound without perfusion: (1) 1% trehalose solution, pH 5.5; (2) rhGDF-5 (150 ng/ml) in 1% trehalose solution, pH 5.5; (3) nutrition solution; and (4) rhGDF-5 (150 ng/ml) in nutrition solution with 1% trehalose, pH 5.5. RESULTS: Promoted wound healing was observed within group 1 and more pronounced within group 2. Groups 3 and 4, with nutrition solution, showed significant adverse effects on wound healing (p<0.05). CONCLUSIONS: The investigated AWD concept appears to be an interesting therapeutic tool to study further wound healing support. Additionally, topical application of rhGDF-5 could be promising.

Serum-free culture of primary human hepatocytes in a miniaturized hollow-fibre membrane bioreactor for pharmacological in vitro studies.

Primary human hepatocytes represent an important cell source for in vitro investigation of hepatic drug metabolism and disposition. In this study, a multi-compartment capillary membrane-based bioreactor technology for three-dimensional (3D) perfusion culture was further developed and miniaturized to a volume of less than 0.5 ml to reduce demand for cells. The miniaturized bioreactor was composed of two capillary layers, each made of alternately arranged oxygen and medium capillaries serving as a 3D culture for the cells. Metabolic activity and stability of primary human hepatocytes was studied in this bioreactor in the presence of 2.5% fetal calf serum (FCS) under serum-free conditions over a culture period of 10 days. The miniaturized bioreactor showed functions comparable to previously reported data for larger variants. Glucose and lactate metabolism, urea production, albumin synthesis and release of intracellular enzymes (AST, ALT, GLDH) showed no significant differences between serum-free and serum-supplemented bioreactors. Activities of human-relevant cytochrome P450 (CYP) isoenzymes (CYP1A2, CYP3A4/5, CYP2C9, CYP2D6, CYP2B6) analyzed by determination of product formation rates from selective probe substrates were also comparable in both groups. Gene expression analysis showed moderately higher expression in the majority of CYP enzymes, transport proteins and enzymes of Phase II metabolism in the serum-free bioreactors compared to those maintained with FCS. In conclusion, the miniaturized bioreactor maintained stable function over the investigated period and thus provides a suitable system for pharmacological studies on primary human hepatocytes under defined serum-free conditions.

HepaRG human hepatic cell line utility as a surrogate for primary human hepatocytes in drug metabolism assessment in vitro.

INTRODUCTION: Primary human hepatocytes are considered as a highly predictive in vitro model for preclinical drug metabolism studies. Due to the limited availability of human liver tissue for cell isolation, there is a need of alternative cell sources for pharmaceutical research. METHODS: In this study, the metabolic activity and long-term stability of the human hepatoma cell line HepaRG were investigated in comparison to primary human hepatocytes (pHH). Hepatocyte-specific parameters (albumin and urea synthesis, galactose and sorbitol elimination) and the activity of human-relevant cytochrome P450 (CYP) enzymes (CYP1A2, CYP2C9, CYP2C19, CYP2D6 and CYP3A4) were assayed in both groups over a period of 14 days subsequently to a two week culture period in differentiated state in case of the HepaRG cells, and compared with those of cryopreserved hepatocytes in suspension. In addition, the inducibility of CYP enzymes and the intrinsic clearances of eleven reference drugs were determined. RESULTS: The results show overall stable metabolic activity of HepaRG cells over the monitored time period. Higher albumin production and galactose/sorbitol elimination rates were observed compared with pHH, while urea production was not detected. CYP enzyme-dependent drug metabolic capacities were shown to be stable over the cultivation time in HepaRG cells and were comparable or even higher (CYP2C9, CYP2D6, CYP3A4) than in pHH, whereas commercially available hepatocytes showed a different pattern The intrinsic clearance rates of reference drugs and enzyme induction of most CYP enzymes were similar in HepaRG cells and pHH. CYP1A2 activity was highly inducible in HepaRG by ß-naphthoflavone. DISCUSSION: In conclusion, the results from this study indicate that HepaRG cells could provide a suitable alternative to pHH in pharmaceutical research and development for metabolism studies such as CYP induction or sub-chronic to chronic hepatotoxicity studies.

Optimization of a serum-free culture medium for mouse embryonic stem cells using design of experiments (DoE) methodology.

The in vitro culture behaviour of embryonic stem cells (ESC) is strongly influenced by the culture conditions. Current culture media for expansion of ESC contain some undefined substances. Considering potential clinical translation work with such cells, the use of defined media is desirable. We have used Design of Experiments (DoE) methods to investigate the composition of a serum-free chemically defined culture medium for expansion of mouse embryonic stem cells (mESC). Factor screening analysis according to Plackett-Burman revealed that insulin and leukaemia inhibitory factor (LIF) had a significant positive influence on the proliferation activity of the cells, while zinc and L: -cysteine reduced the cell growth. Further analysis using minimum run resolution IV (MinRes IV) design indicates that following factor adjustment LIF becomes the main factor for the survival and proliferation of mESC. In conclusion, DoE screening assays are applicable to develop and to refine culture media for stem cells and could also be employed to optimize culture media for human embryonic stem cells (hESC).

Interwoven four-compartment capillary membrane technology for three-dimensional perfusion with decentralized mass exchange to scale up embryonic stem cell culture.

We describe hollow fiber-based three-dimensional (3D) dynamic perfusion bioreactor technology for embryonic stem cells (ESC) which is scalable for laboratory and potentially clinical translation applications. We added 2 more compartments to the typical 2-compartment devices, namely an additional media capillary compartment for countercurrent 'arteriovenous' flow and an oxygenation capillary compartment. Each capillary membrane compartment can be perfused independently. Interweaving the 3 capillary systems to form repetitive units allows bioreactor scalability by multiplying the capillary units and provides decentralized media perfusion while enhancing mass exchange and reducing gradient distances from decimeters to more physiologic lengths of <1 mm. The exterior of the resulting membrane network, the cell compartment, is used as a physically active scaffold for cell aggregation; adjusting intercapillary distances enables control of the size of cell aggregates. To demonstrate the technology, mouse ESC (mESC) were cultured in 8- or 800-ml cell compartment bioreactors. We were able to confirm the hypothesis that this bioreactor enables mESC expansion qualitatively comparable to that obtained with Petri dishes, but on a larger scale. To test this, we compared the growth of 129/SVEV mESC in static two-dimensional Petri dishes with that in 3D perfusion bioreactors. We then tested the feasibility of scaling up the culture. In an 800-ml prototype, we cultured approximately 5 x 10(9) cells, replacing up to 800 conventional 100-mm Petri dishes. Teratoma formation studies in mice confirmed protein expression and gene expression results with regard to maintaining 'stemness' markers during cell expansion.