A Planar Culture Model of Human Absorptive Enterocytes Reveals Metformin Increases Fatty Acid Oxidation and Export.

BACKGROUND & AIMS: Fatty acid oxidation by absorptive enterocytes has been linked to the pathophysiology of type 2 diabetes, obesity, and dyslipidemia. Caco-2 and organoids have been used to study dietary lipid-handling processes including fatty acid oxidation, but are limited in physiological relevance or preclude simultaneous apical and basal access. Here, we developed a high-throughput planar human absorptive enterocyte monolayer system for investigating lipid handling, and then evaluated the role of fatty acid oxidation in fatty acid export, using etomoxir, C75, and the antidiabetic drug metformin.
METHODS: Single-cell RNA-sequencing, transcriptomics, and lineage trajectory was performed on primary human jejunum. In vivo absorptive enterocyte maturational states informed conditions used to differentiate human intestinal stem cells (ISCs) that mimic in vivo absorptive enterocyte maturation. The system was scaled for high-throughput drug screening. Fatty acid oxidation was modulated pharmacologically and BODIPY (Thermo Fisher Scientific, Waltham, MA) (B)-labeled fatty acids were used to evaluate fatty acid handling via fluorescence and thin-layer chromatography.
RESULTS: Single-cell RNA-sequencing shows increasing expression of lipid-handling genes as absorptive enterocytes mature. Culture conditions promote ISC differentiation into confluent absorptive enterocyte monolayers. Fatty acid-handling gene expression mimics in vivo maturational states. The fatty acid oxidation inhibitor etomoxir decreased apical-to-basolateral export of medium-chain B-C12 and long-chain B-C16 fatty acids, whereas the CPT1 agonist C75 and the antidiabetic drug metformin increased apical-to-basolateral export. Short-chain B-C5 was unaffected by fatty acid oxidation inhibition and diffused through absorptive enterocytes.
CONCLUSIONS: Primary human ISCs in culture undergo programmed maturation. Absorptive enterocyte monolayers show in vivo maturational states and lipid-handling gene expression profiles. Absorptive enterocytes create strong epithelial barriers in 96-Transwell format. Fatty acid export is proportional to fatty acid oxidation. Metformin enhances fatty acid oxidation and increases basolateral fatty acid export, supporting an intestine-specific role.

Single-cell transcriptomics on human enterocytes characterized lipid-handling gene expression patterns that were mimicked in culture. Candidate drugs were tested on a high-throughput 96-well, scale-up system, showing that metformin increases fatty acid oxidation and fatty acid export.

See editorial on page 407.

The small intestinal (SI) epithelium is a selective barrier that serves as the point of entry for essential micronutrients and macronutrients to meet energy demands and preserve general homeostatic functions. Lipids are the most energy-dense of the macronutrients and are absorbed by the SI epithelium. The majority of dietary lipids are triglycerides and are broken down by the stomach and intestinal lumen into fatty acids and monoglycerides. Fatty acids then are taken up by absorptive enterocytes, the predominant cell type of the SI epithelium. Lipids then can be metabolized and used by absorptive enterocytes for cellular functions such as energy production, membrane synthesis, and storage as lipid droplets, or distributed to the body by the well-accepted lipoprotein–lymphatic system and by the less-appreciated fatty acid–portal vein pathway.3, 4, 5

Investigating uptake, metabolism, and export of dietary fatty acids, here collectively called fatty acid handling, in vitro is challenging because of the historical lack of physiologically relevant culture models. Because metabolic disorders such as dyslipidemia, diabetes, and obesity,, are increasing,, there is strong interest in evaluating how genetics,, and environmental factors, such as alterations in gut microbiota12, 13, 14, 15 and eating behaviors,,, are associated with fatty acid–handling mechanisms, which by nature are complex to study in human beings or animal models. Limitations of 3-dimensional organoid cultures, ethical considerations of human research, and inadequacies of animal models compound the challenges and limit scientific progress toward solutions for these and other metabolic diseases. In this regard, an absorptive enterocyte cell culture platform using primary human cells and coupled with simple and sensitive detection of fatty acids and their metabolites would represent a significant improvement and address many of the limitations of existing in vitro culture models.

Traditionally, cell culture models of human SI epithelium largely have relied on cancer or transimmortalized mouse intestinal cell lines (ie, Caco-2, m-ICcl2, and so forth), which retain properties consistent with undifferentiated states.18, 19, 20 Organoid culture models have become popular alternatives because they typically are derived from primary intestinal epithelial stem cells (ISCs) and can differentiate into the main mature lineages of the differentiated gut epithelium., Organoids are small (diameter, ∼100–1000 μm) spherical structures cultured in thick hydrogels. They are composed of polarized cells in which the enclosed apical or luminal surface precludes application of fatty acids to the physiologically relevant surface to mimic dietary absorption. Organoids can be cultured with the apical surface facing outward (apical-out organoids); however, this causes the basal surface to be enclosed and inaccessible, preventing sampling of exported or metabolized lipids. Collectively, these factors limit interpretations, reduce throughput, and prohibit analyses necessary to accurately assess fatty acid handling across the absorptive enterocyte monolayer.

Conventional methods to detect fatty acids has relied on using radioisotope- and heavy isotope–labeled fatty acid analogs.24, 25, 26, 27, 28, 29, 30 These isotope-labeled fatty acids are thought to behave similarly to native fatty acids in absorptive, metabolic, and export processes24, 25, 26, 27, 28, 29, 30; however, special safety precautions and sophisticated downstream analytics (eg, mass spectrophotometry) are required, limiting access of these assays to many investigators. Fluorescently labeled fatty acids represent an attractive alternative because they do not require special handling, are sensitive, commercially available, and can be detected using a variety of common instruments and methods (eg, plate readers, microscopes, thin layer chromatography [TLC]). For example, BODIPY (Thermo Fisher Scientific, Waltham, MA) (B) is a brightly fluorescent fluorophore, and B-fatty acid analogs have been shown to mimic endogenous fatty acid metabolism and transport, making it an effective tracer for fatty acids in lipid-handling studies.31, 32, 33, 34, 35 Importantly, unlike isotopically labeled fatty acids, fluorescently labeled fatty acids also permit imaging of fatty acid handling in live and fixed cells.,

Our group recently developed methods for culturing and indefinitely expanding primary human ISCs as 2-dimensional (2D) monolayers. ISCs cultured this way then can be transferred to Transwell (Corning, Corning, NY) inserts, cultured to confluence, and then terminally differentiated into absorptive and secretory lineages found in vivo., Monolayers cultured on permeable Transwell inserts are in contact with apical and basal reservoirs where factors, drugs, and metabolites easily can be added or sampled throughout an experiment. Unlike organoids, monolayers grow as planar sheets rather than spheres suspended in thick hydrogels, allowing use of imaging systems that are in common use in basic science laboratories, robotic drug screening, and validation platforms.

Our study here had 2 primary goals: the first was to develop and validate a new culture system to study fatty acid handling, and the second was to use the system to show utility for evaluating the impact that a set of drugs has on fatty acid oxidation and export of fatty acid metabolites. We took an approach that first defined the baseline transcriptomic state of relevant lipid-handling genes and then tailored culture conditions to mimic the in vivo lipid-handling gene profiles. Readouts for fatty acid handling are designed for practicality, sensitivity, and high-throughput applications. Using this new system, we pharmacologically inhibited and potentiated fatty acid oxidation and observed changes in fatty acid export that informed a hypothesis that fatty acid oxidation increases export of medium- and long-chain fatty acid metabolites across the basolateral membrane. This was tested and the findings showed new biological insights into the role of fatty acid oxidation on export of fatty acids with implications for understanding blood-glucose regulation and appetite control.

Results

Single-Cell Transcriptomics of Jejunal and Ileal Human Mucosa Define Early, Intermediate, and Mature Nutrient-Handling Enterocytes

First, we sought to characterize the baseline transcriptomic profiles of human absorptive enterocytes in vivo to guide the development of an in vitro model of human fatty acid handling. The distal SI (jejunum and ileum) represents the majority of absorptive epithelium in the human SI; however, lipid-handling transcriptional profiles of jejunal and ileal absorptive lineages have not been fully described at the single-cell level. Endoscopic biopsy specimens have enabled single-cell transcriptomics to be performed on human duodenal and ileal mucosa; however, single-cell transcriptomics only recently have been reported for the jejunum,, with limited characterization of absorptive function., To further define absorptive function of the distal SI, single-cell RNA-sequencing (scRNAseq) was performed on primary jejunal and ileal epithelium isolated from a healthy organ donor. To identify the distinct cell types captured in the data set, dimensional reduction was performed, showing that the cell populations represented the major reported cell lineages in the human SI epithelium (Figure 1A). Most cells in the data set clustered separately from the minority secretory lineages (enteroendocrine, goblet, bestrophin 4 (BEST4)+, and tuft). This main cluster comprised cells consistent with ISCs, transit-amplifying cells, and absorptive enterocytes (Figure 1A). The high viability, quality, and capture of the full complement of major SI epithelial lineages provided a strong foundation for transcriptomic characterization.

Figure 1

Single-cell transcriptomics of healthy human absorptive epithelium. (A–D) Unbiased Leiden clustering of primary jejunum and ileum. (A) Identification of cell types. (B) Inferred cell-cycle state based on expression of previously categorized G1, G2M, and S phase–associated genes. (C) ISC score based on expression of established ISC genes LGR5, OLFM4, ASCL2, SMOC2, and SOX9. (D) Absorptive enterocyte score based on expression of established absorptive enterocyte genes APOA1, APOA4, APOC3, ALDOB, and SEPP1. (E and F) Gene expression of ISCs, and absorptive enterocyte populations identified in panel A. (E) Expression of top 20 ISC genes previously identified in human ileum. ∗LGR5 was not within the top 20 ISC genes previously identified in human ileum, but it was included for reference. (F) Expression of top 20 absorptive enterocyte genes previously identified in human ileum. (G) Bottom right: Vectors calculated based on a dynamic model of RNA velocity showing likely cell transitions of ISCs, early absorptive enterocytes, intermediate absorptive enterocytes, and mature absorptive enterocytes. Top left: Schematic showing likely positions of identified ISCs and absorptive enterocyte populations along the crypt–villus axis. Note that secretory and ileal lineages were removed from the analysis to focus on the relationship between ISCs and maturing of jejunal absorptive enterocytes. (H) A latent time value between 0 and 1 was assigned to each cell to order cells along the trajectory modeled by RNA velocity. A latent time of 0 means that the cell has yet to experience any differentiation in the modeled trajectory whereas a latent time of 1 means that the cell has progressed completely through the modeled differentiation pathway. (I) Identified human jejunal absorptive enterocytes were ordered based on latent time (x-axis) and DEGs of identified human ISCs and absorptive enterocyte populations were plotted on the y-axis. ∗LGR5 was not among the top 20 ISC genes identified in human Ileum. (J) UMAPs showing expression of select DEGs from the population of cells along the ISC to mature absorptive enterocyte differentiation axis. (K) In situ hybridization of ISCs and early absorptive enterocyte DEGs OLFM4 and DMBT1, respectively (left). In situ hybridization of intermediate absorptive enterocyte and mature absorptive enterocyte DEGs FABP1 and APOA4, respectively (right). AE, absorptive enterocyte; DAPI, 4′,6-diamidino-2-phenylindole; mRNA, messenger RNA; UMAP, uniform manifold approximation and projection.

In human duodenum and ileum, absorptive enterocytes are subcategorized as early, intermediate, and mature based on maturation state., Genes associated with these 3 absorptive enterocyte subsets also were observed in jejunal and ileal Leiden clustering (Figure 1A). Jejunal and ileal absorptive enterocytes clustered together, showing a high degree of transcriptomic similarity between these cells. Because Leiden-cluster boundaries appear artificially binary, each cell was interrogated independently using a computational score comprising curated gene sets that would predict cell lineages for each Leiden cluster. Three different curated gene sets from prior studies were used to identify different cell-cycle stages, ISCs,44, 47, 48 and absorptive enterocytes (Figure 1B–D). The cell-cycle score showed that most cycling cells are associated strongly at one end of the cluster, which is consistent with cells showing the strongest ISC score (Figure 1B and C). By contrast, cells with the highest absorptive enterocyte score were associated strongly with the opposite end of the same cluster (Figure 1D). Twenty ISC signature genes that were reported to be the most highly expressed genes in human ileal ISCs also were highest in jejunal ISCs (Figure 1E). These ISC-associated genes gradually decreased in jejunal absorptive enterocyte populations consistent with differentiation (Figure 1E). Conversely, 20 of the most highly expressed absorptive enterocyte signature genes identified in human ileal mature absorptive enterocytes were lowest in jejunal ISCs. These absorptive enterocyte-associated genes gradually increased in jejunal absorptive enterocyte populations as they differentiated (Figure 1F). Together, these data suggest jejunal absorptive enterocytes follow a similar maturation trajectory as duodenal and ileal cells from other reports.,,

Computational Lineage Trajectory Analyses Show Human Absorptive Enterocytes Progress Through Maturation Stages With Increasing Fatty Acid-Handling Gene Expression

A recent study of murine jejunal villi described an absorptive enterocyte maturation program wherein spatially distinct zones along the villus were used to define absorptive enterocyte functions (eg, antimicrobial functions toward the base, lipoprotein secretion toward the tip). Whether human absorptive enterocytes also progress through maturation states corresponding to distinct functions along the length of the villus is unknown. Progressive expression of mature absorptive enterocyte markers in our jejunal data set (Figure 1F) led to the hypothesis that human absorptive enterocyte clusters correspond to lineages with distinct gene expression profiles of maturation along the villus. To test this hypothesis, we used a computational framework based on the rate of change of spliced and unspliced messenger RNA ratios (ie, RNA velocity) and a differentiation-specific RNA velocity-based metric (ie, latent time) (Figure 1G and H). Human jejunal cells were used in the analysis to be consistent with previous mouse studies. Vectors calculated based on the solved dynamic model of RNA velocity predict a trajectory of ISCs gradually differentiating to mature absorptive enterocytes (Figure 1G, arrows). Next, a latent time value between 0 and 1 was assigned to each cell to order cells along the trajectory modeled by the RNA-velocity vectors (Figure 1H). A latent time of 0 means that the cell has yet to enter the modeled trajectory, whereas a latent time of 1 means that the cell has progressed completely through the modeled differentiation pathway. Through combining RNA velocity and latent time, we show a single path of likely cell transitions from ISCs into early absorptive enterocytes, followed by intermediate absorptive enterocytes, and culminating with mature absorptive enterocytes (Figure 1G and H).

To visualize the number of differentially expressed genes (DEGs) between these maturation states (ie, early absorptive enterocyte, intermediate absorptive enterocyte, and mature absorptive enterocyte), the 1537 identified DEGs from these populations were plotted against all jejunal ISCs, early absorptive enterocytes, intermediate absorptive enterocytes, and mature absorptive enterocytes as ordered by the calculated latent time (Figure 1I, Supplementary Table 1). ISCs and mature absorptive enterocytes had the largest amount of DEGs (732 and 592, respectively), with early absorptive enterocytes and intermediate absorptive enterocytes only having 106 and 107 DEGs, respectively (Supplementary Table 3). The pattern of gene expression showed that genes highest in ISCs gradually turn off as the cells mature (Figure 1I). Conversely, mature absorptive enterocyte genes begin to turn on in the early absorptive enterocyte and intermediate absorptive enterocyte states (Figure 1I).

In situ hybridization was performed on jejunal tissue sections from the same donor to validate the predicted computational trajectory by locating marker genes for each maturation stage to points along the villus (Figure 1J and K). Consistent with our predictions, OLFM4 (ISC marker) localized to the crypt base, DMBT1 (antimicrobial function) was found in the upper crypt and lower villus, FABP1 (lipid chaperone) was enriched in the midvillus region, and APOA4 (chylomicron assembly) was enriched at the villus tip (Figure 1K). Consistent with findings in mice, human absorptive enterocytes appear to perform distinct functions as they differentiate along the crypt–villus axis.

The cell maturation analyses were further refined next by using curated gene sets specific for fatty acid handling (Figure 2). Expression of genes associated with chylomicron assembly (APOA1, APOA4, APOB) were enriched in intermediate absorptive enterocyte and mature absorptive enterocyte populations, while distinct subsets of genes involved in regulating fatty acid transport (CD36, SLC27A1, SLC27A5) and fatty acid oxidation (CPT1A, PPARG, PPARGC1A) were enriched differentially in each maturation state, suggesting discrete regulation of these lipid-handling mechanisms along the villi (Figure 2). Together, these findings support distinct transcriptional states and associated lipid-handling mechanisms with each absorptive enterocyte maturation stage along the crypt–villus axis in the human jejunum.

Figure 2

Expression of lipid-handling genes in in vivo ISCs and human absorptive enterocyte populations. Genes were selected from Gene Ontology (GO) terms relating to lipid metabolic processes. Fatty acid binding protein genes were selected manually and not derived from GO terms (bottom left). The following genes were not detected in our data set and were excluded: DGAT2L7P, MIR29B1, MIR30C1, MIR548P, APOA2, PLIN4, ABCD2, GRM1, NTSR1, SLC17A6, SLC17A7, SLC1A6, ABCD2, and TWIST1.

Characterization of Culture Model by scRNAseq Shows Differentiation of Human Jejunal ISCs Generates Highly Pure Monolayers of Absorptive Enterocytes

In prior work, our group developed platforms that promote robust, long-term expansion of human ISCs. When cultured on Transwell culture systems, ISCs can be induced to differentiate into absorptive enterocytes with strong barrier function and used for transport studies, (Table 1). The Transwell membrane allows for application and retrieval of fatty acids and their metabolites from the apical and basal reservoirs, respectively. Both are crucial parameters not available in organoids that enable mechanistic studies of fatty acid handling in vitro.

Table 1

Tissue Culture Media Conditions

Reagent	Stock solution	EM	DM
L-WRN–conditioned medium		50%
Advanced DMEM/F12	1×	50%	100%
Glutamax	100×	1×	1×
HEPES	1 mol/L	10 mmol/L	10 mmol/L
Primocin	50 mg/mL	50 ug/mL	50 ug/mL
N-acetyl-cysteine	1 mol/L	1 mmol/L	1 mmol/L
Murine EGF	250 ug/mL	50 ng/mL	50 ng/mL
Nicotinamide	1 mol/L	10 mmol/L
B27	50×	1×
Gastrin	1 mg/mL	10 nmol/L
PGE2	1 mmol/L	10 nmol/L
A8301	5 mmol/L		500 nmol/L
SB202190	30 mmol/L	3 umol/L
Y27632	10 mmol/L	10 umol/L

EGF, epidermal growth factor; L-WRN, L-cell derived Wnt3a, R spondin 3, and Noggin; PGE2, prostaglandin E2.

To define cell phenotypes and characterize lineage purity at single-cell resolution, scRNAseq was performed on ISCs under expansion conditions (Figure 2) and absorptive enterocyte monolayers after 5 days of differentiation (Figure 3A). Inhibition of Notch signaling is required for specification of the secretory lineage., Because we initiated differentiation by removal of ISC growth factors without addition of a Notch inhibitor, we hypothesized that ISCs would follow an absorptive rather than secretory differentiation trajectory. Leiden clustering of ISC expansion and differentiation conditions showed 2 distinct clusters unique to each media formulation (Figure 3B). Forty percent of cells in the ISC cluster were predicted by computational methods to be in either S or G2/M phase, whereas more than 99% of cells in absorptive enterocyte differentiation conditions were predicted to be in G0/G1 phase (Figure 2C).

Figure 3

Single-cell transcriptomics of 2D ISCs and absorptive enterocyte cultures. (A) Schematic of isolation, expansion, and differentiation of ISCs from the jejunum of a healthy organ donor. (B) Unbiased Leiden clustering of collagen-grown ISCs (green) and ISCs differentiated for 2 days (orange). (C) Inferred cell-cycle state based on expression of previously categorized G1, G2M, and S phase–associated genes. (D) Left: ISC score based on expression of established ISC genes LGR5, OLFM4, ASCL2, SMOC2, and SOX9. Right: Absorptive enterocyte score based on expression of established absorptive enterocyte genes APOA1, APOA4, APOC3, ALDOB, and SEPP1. (E) Expression of secretory lineage markers MUC2, CHGB, TRPM5, and BEST4. AE, absorptive enterocyte; BEST4, bestrophin 4; CHGB, chromagranin B; Duo, duodenum; EEC, enteroendocrine cell; Ile, ileum; Jej, jejunum; MUC2, mucin2; TRPM5, Transient Receptor Potential Cation Channel Subfamily M Member 5; UMAP, uniform manifold approximation and projection.

Computational cell-cycle predictions cannot distinguish between cells in G1 phase or cells that have left the cell cycle; thus, scoring of ISC and absorptive enterocyte gene profiles was performed to assess if differentiation conditions were sufficient to confer terminal differentiation toward an absorptive fate (Figure 2D). ISC and absorptive enterocyte scores were generated for cells grown in in vitro monolayers using curated gene sets that identify either ISCs or absorptive enterocytes (Figure 3D). This scoring showed that in vitro monolayers had higher expression of ISC genes in expansion conditions and higher expression of absorptive enterocyte genes in differentiation conditions (Figure 3D). Very rare cells expressing markers of classic secretory lineages (goblet cells, enteroendocrine cells, tuft cells, BEST4+ cells) were observed (Figure 3E). These data show that culture conditions promoted ISC differentiation toward an absorptive cell fate and indicate the potential to model physiological absorptive enterocyte differentiation in vitro.

Culture Methods Developed in 96-Transwell Format Drive Time-Dependent ISC Differentiation and Absorptive Enterocyte Maturation Described by Transcriptomic States

Absorptive enterocyte planar monolayers were scaled to a 96-well Transwell format for high-throughput applications. Transepithelial electrical resistance (TEER) was used to monitor barrier integrity over 15 days of ISC expansion (4 days) and absorptive enterocyte differentiation (12 days) (Figure 4A). As expected, ISC conditions produced a stable and low TEER, while differentiation media (DM) produced an immediate and progressive increase in TEER that peaked at 7 days of differentiation (Figure 4A). The progressive increase in TEER suggests changes in gene expression patterns consistent with mature absorptive enterocytes. To confirm this at the transcriptomic level, bulk RNAseq was performed on absorptive enterocytes in differentiation conditions for 0, 2, 5, 7, 10, and 11 days (Figure 4B). Principal component analysis showed tight agreement between technical replicates and showed rapid and large transcriptomic changes after just 2 days of differentiation (Figure 4B). Bulk RNAseq of absorptive enterocyte monolayers showed a trajectory of progressive transcriptional changes through time (Figure 4B). Top in vivo DEGs (Figure 4C) were expressed at corresponding maturation time points in vitro (Figure 4D), suggesting an intrinsic program of progressive absorptive enterocyte maturation that begins upon removal of ISC niche growth factors.

Figure 4

RNAseq shows a time-dependent absorptive enterocyte maturation program in jejunal ISCs in vitro. (A) TEER of jejunal ISCs (n = 12 Transwell cultures). ISCs were maintained in expansion media for 3 days and then switched to differentiation media for 12 days. TEER was monitored daily. Red triangles indicate time points at which cells were lysed for bulk RNAseq. A Kolmogorov–Smirnov test was performed comparing TEER between day 0 (D0) and D1–D7. ∗∗∗∗P < .0001. (B) Principal component analysis of sequenced transcriptomes. (C) Expression of the top 30 DEGs in ISCs, early absorptive enterocytes, intermediate absorptive enterocytes, and mature absorptive enterocytes. (D) In vitro expression of the top 30 in vivo ISC, early absorptive enterocyte, intermediate absorptive enterocyte, and mature absorptive enterocyte DEGs. SNHG19, SNHG25, and NEAT1 were not detected in our in vitro data set and were excluded. ∗LGR5 was not among the top 30 DEGs in ISCs but it was included for reference.

To further determine the optimal time points to investigate fatty acid handling (ie, fatty acid transmembrane transport, fatty acid oxidation, lipid droplet formation, chylomicron secretion, and triglyceride metabolic processes), the expression of key genes in these processes was evaluated. There was a general increase in fatty acid–handling genes as cells matured both in vivo and in vitro (Figure 5A). Linear correlation was used to quantitatively describe the extent of similarity between in vivo and in vitro fatty acid–handling gene expression during absorptive enterocyte maturation (Figure 5A and B). Pairwise comparisons of mean lipid-handling gene expression values showed that absorptive enterocytes on day 7 are more similar to intermediate and mature absorptive enterocytes than ISCs and early absorptive enterocytes (Figure 5A and B). Thus, TEER and expression profiles of key fatty acid handling genes were used as criteria to inform the optimal window to investigate fatty acid handling within our model system. Based on these criteria, 7 days of differentiation was the only time point to show both high TEER and close similarity to in vivo mature absorptive enterocyte gene expression of fatty acid–handling genes. For this reason, we chose to assay fatty acid handling in differentiated absorptive enterocytes after 7 days of directed differentiation.

Figure 5

RNAseq shows a time-dependent lipid-handling program in jejunal ISCs in vitro. (A) Expression of a curated set of fatty acid–handling genes in vivo (left) and in vitro (right). Genes were curated from the genes in Figure 2. (B) Matrix showing linear regression plots for lipid-handling genes. Each dot represents the mean expression for a lipid-handling gene. The x-axis represents the mean (n = 3 samples per time point) expression from in vitro bulk RNAseq data. The y-axis represents the mean expression for each cluster from in vivo scRNAseq data. The dotted line shows the line that was used to calculate residuals and is drawn with slope = 1 (ie, mean expression of in vitro differentiation bulk RNAseq data perfectly matches in vivo scRNAseq mean expression for each cluster). Red dots indicate the lowest residual sum of squares for each row (ie, lowest residual for each time point sampled in the in vitro differentiation experiment). (C) Residual plot of key lipid-handling genes. FA, fatty acid.

High-Throughput Planar Absorptive Enterocyte Cultures Produce Robust Epithelial Barriers and Preserve Fatty Acid–Handling Functions Detected by Fluorescent Fatty Acid Conjugates

High TEER and transcriptomic data suggest that absorptive enterocyte monolayers could serve as an effective in vitro model for human absorptive enterocyte lipid handling. Strong barrier function is required for accurate interpretation of fatty acid handling by absorptive enterocytes because a leaky barrier would allow fatty acids to bypass the absorptive enterocyte monolayer. As a means to provide sensitive readouts for fatty acid handling that is accessible to most laboratories, we adopted fluorescent analogs that can be detected easily by a fluorescent plate reader and TLC. To validate barrier function, 10 kilodaltons of dextran conjugated to Alexa Fluor 647 (Dextran-647), a nonabsorbable fluorescent polysaccharide, was applied to the apical monolayer surface, followed by quantification of basolateral fluorescence at 2, 4, and 6 hours of apical exposure (Figure 6A). There were nearly undetectable levels of Dextran-647 (<0.2 pmol) in the basal reservoir at any time point, whereas Transwells without cells allowed more than 30 pmol to diffuse through and reached an equilibrated concentration within approximately 4 hours. These data show strong barrier function and that significant epithelial barrier defects can be detected readily by a Dextran-647 signal in the basal reservoir as early as 2 hours after exposure.

Figure 6

Application of fluorescent polysaccharide and fatty acid show integral barrier and fatty acid handling in vitro. (A) Left: Schematic showing application of Dextran-647 to the apical surface of absorptive enterocytes or empty Transwell cultures. Right: Subsequent sample collection and fluorescence quantification via plate reader. ISCs were differentiated for 7 days before application of 1 umol/L Dextran-647. A total of 50 uL media from the basal reservoir was removed at 2, 4, and 6 hours from empty Transwell inserts or absorptive enterocyte cultures and replaced with fresh differentiation media; Bonferroni multiple comparisons test. ∗∗∗∗P < .0001. (B) Schematic of B-C12 application to absorptive enterocyte Transwell cultures and subsequent retrieval from apical, cellular, and basal reservoirs. (C) TLC (polar) of basal media from absorptive enterocyte cultures after 6 hours of incubation with B-C12 (left). Retention factor (RF) of B-C5, B-C12, and C16 standards were used to generate a standard curve to infer the chain length of fatty acids whose RF values do not correspond to B–fatty acid standards used (right). (D and E) TLC of apical, cellular, and basal reservoirs of absorptive enterocyte Transwell cultures after 6 hours of B-C12 application using polar and protonating solvents. A previous study using TLC to identify B-lipid species with the same solvents used in this study reported NFBs that do not correspond to B-lipids used. ∗Location of B-C4, B-C6, and B-C8 were inferred from panel C. Experiments were performed in triplicate (n = 3 absorptive enterocyte Transwell cultures). AE, absorptive enterocyte; B, BODIPY; B-C4, B-C6, B-C8, B-C12, B-C16, BODIPY–fatty acids; FA, fatty acid; TAG, triglyceride.

BODIPY (B) is a bright fluorophore that has a different fluorescent signature than Dextran-647, facilitating separate detection of these 2 molecules by plate readers. B–fatty acid analogs have been used extensively for lipid trafficking and metabolism studies in vitro and in vivo and are considered to be handled accurately by absorption, metabolism, and export mechanisms.58, 59, 60, 61, 62 Thus, to characterize fatty acid–handling properties of absorptive enterocyte monolayers, we applied the medium-chain fatty acid analog B-C12 to the apical surface of absorptive enterocyte cultures for 6 hours (Figure 4B). Apical (input) and basal (output) media were collected, and absorptive enterocytes were lysed after 6 hours of apical exposure (Figure 4B). TLC was used to detect and quantify input of B–fatty acids as well as metabolized and exported B–fatty acid products. A combination of polar and protonating solvents were used to resolve the complex mixture of B–fatty acids and metabolites (Figure 6D and E). B-labeled fatty acid standards were applied to the TLC plate in a separate lane and unknown fatty acid species of different carbon chain lengths were extrapolated (Figure 6C). These methods were scaled for the 96-well Transwell system using less than 170 μL media or cell lysate. A previous study using TLC to identify B-lipid species with the same solvents used in this study reported naturally fluorescent bands (NFBs) that do not correspond to B-lipids (Figure 6D and E). TLC showed clear separation of key fatty acid species in all 3 reservoirs (ie, apical, cellular, basal), indicating robust sensitivity of this approach for detection of a broad range of fatty acid species (Figure 6D).

To evaluate fatty acid handling by absorptive enterocyte monolayers, B-C12 was incubated with absorptive enterocyte monolayers for 6 hours followed by TLC analysis to identify the B–fatty acid or B-metabolites in each reservoir (ie, apical, cellular, basal). TLC showed that most of the B-labeled species in the apical reservoir were medium-chain fatty acids (B-C12, B-C8, and B-C6) (Figure 6D and E). A short-chain species consistent with B-C4 also was detected in the apical reservoir at a lower level (Figure 6D and E). In the intracellular reservoir, the largest species were triglycerides, indicating robust fatty acid esterification, and phospholipids indicative of B-C12 being metabolized and incorporated into the lipid bilayers. In the cell lysates, fatty acids were some of the lowest B-lipid species, suggesting dynamic metabolism, diffusion, and mobilization. In the basal reservoir, fatty acids were the predominant B-lipid species, with a smaller fraction consisting of triglycerides, suggesting basal export of fatty acids and chylomicrons (Figure 6D and E). Together these data show robust detection of input B-fatty acids, a broad range of the derivative metabolites, and their relative distributions in each reservoir as the B-lipids are processed by absorptive enterocytes.

Inhibiting Fatty Acid Oxidation in Absorptive Enterocyte Monolayers Decreases Basolateral Export of Oxidized Fatty Acid Species

We next explored the utility of the platform for investigating small-molecule perturbations on fatty acid oxidation. Etomoxir, a carnitine palmitoyl transferase 1 (CPT1) inhibitor, was used to inhibit fatty acid oxidation. When CPT1 is inhibited, fatty acids cannot be imported into the mitochondria where fatty acid oxidation normally catabolizes longer-chain fatty acids to smaller-chain fatty acids. CPT1A is expressed robustly in primary human absorptive enterocytes both in vivo and in vitro on day 7 of differentiation (Figure 5A). Etomoxir was used at a concentration of 100 umol/L because of recent reports64, 65, 66 of off-target effects at high concentrations (>100 umol/L). After pretreatment with etomoxir, a variety of different-chain-length B–fatty acids (short-chain fatty acid, B-C5; medium-chain fatty acid, B-C12; or long-chain fatty acid, B-C16) were applied to the apical reservoir to mimic postprandial absorptive enterocyte exposure to fatty acids in vivo. Media from the basal reservoir was taken at 2, 4, and 6 hours after application of fluorescent fatty acids (Figure 7A). The impact of etomoxir on basal export of fatty acids and fatty acid metabolites was evaluated in real time by total fluorescence detection by plate reader, and TLC was used to identify BODIPY-labeled metabolites using small volumes of media from the 96-well Transwell platform (Figure 8A–C).

Figure 7

Etomoxir reduces fatty acid export. (A) Left: Schematic showing basal reservoir sampling after application of B–-fatty acids to the apical surface of absorptive enterocyte Transwell cultures and subsequent quantification of basal fluorescence via plate reader. Right: Quantification of basal fluorescence from B-C5–, B-C12–, and B-C16–treated absorptive enterocyte cultures treated with vehicle (dimethyl sulfoxide), or etomoxir. Bonferroni multiple comparisons test. ∗∗∗∗P < .0001. A total of 50 uL basal media was collected and quantified by plate reader at 2, 4, and 6 hours after fatty acid application. (B) Quantification of Dextran-647 in the basal reservoir of absorptive enterocyte monolayers from basal samples of fatty acid–handling screens taken at 2, 4, and 6 hours. Bonferroni multiple comparison test. ∗∗∗∗P < .0001. Experiments were performed in triplicate (n = 3 absorptive enterocyte Transwell cultures). ANOVA, analysis of variance; B-Cx, BODIPY lipids.

Figure 8

Inhibiting fatty acid oxidation reduces fatty acid export. (A–C) TLC of basal-reservoir media collected 6 hours after vehicle or etomoxir and B-C5, B-C12, or B-C16 application. A previous study using TLC to identify B-lipid species with the same solvents used in this study reported NFBs that do not correspond to B-lipid standards used. Approximate locations of NFBs are labeled. ∗Location of B-C4, B-C6, and B-C8 were inferred from Figure 6C. (B and C) Right: Quantification of the fluorescence contributed by the B-C6 band divided by the cumulative fluorescence of all bands in each TLC lane. Unpaired t test. ∗∗∗∗P < .0001. (D and E) Cellular-reservoir fluorescence of B-C12– or B-C16–treated absorptive enterocyte Transwell cultures treated with vehicle (dimethyl sulfoxide [DMSO]) or etomoxir. Unpaired t test. ∗∗P < .01, ∗∗∗∗P < .0001. Scale bars: 100 um. (F) Apical-reservoir fluorescence of B-C12– or B-C16–treated absorptive enterocyte Transwell cultures treated with vehicle (DMSO) or etomoxir. Unpaired t test. ∗P < .05, ∗∗∗∗P < .0001. (G) Schematic showing proposed mechanism of reduced B–fatty acid export by etomoxir. B-fatty acids were applied at a concentration of 20 umol/L. Etomoxir was applied at a concentration of 100 umol/L. Experiments were performed in triplicate (n = 3 absorptive enterocyte Transwell cultures). B-Cx, fluorescence coming from all B-lipids in basal, apical, or cellular reservoirs; Eto, etomoxir; FA, fatty acid; FAO, fatty acid oxidation; TAG, triglyceride; Veh, vehicle.

Strong barrier function through the duration of B–fatty acid and etomoxir treatment was confirmed by lack of Dextran-647 in the basal reservoir (Figure 7B). When B-C5 was the apical input fatty acid, basal fluorescence did not significantly change when fatty acid oxidation was inhibited (Figure 7A). By contrast, when B-C12 and B-C16 were the apical input fatty acids, fatty acid oxidation inhibition significantly reduced basal-reservoir fluorescence (Figure 7A). TLC showed that the most abundant lipid species in the basal reservoir of B-C5–treated cultures was B-C5 (Figure 8A), suggesting that this fatty acid diffused passively through absorptive enterocyte monolayers because fatty acids become increasingly water soluble as chain length decreases. For control cultures (vehicle) treated with B-C12 and B-C16, the most predominant lipid species corresponded to B-C6 (Figure 8B and C), indicating catabolism of B-C12 and B-C16.

After treatment of B-C12 and B-C16 cultures with the fatty acid oxidation inhibitor etomoxir, B-C6 decreased significantly (Figure 8B and C). In B-C12–treated cultures, etomoxir caused the apical input fatty acid (B-C12) to be the predominant lipid species in the basal reservoir (Figure 8B). In cultures exposed to apical B-C16, etomoxir treatment did not result in B-C16 being the predominant lipid species in the basal reservoir (Figure 8C). These findings suggest that long-chain fatty acids (>B-C12) are less amenable to passive diffusion than short-chain (B-C5) and medium-chain (B-C12) fatty acids. Together these data indicate that fatty acids of shorter chain length (B-C5, B-C6) are more amenable to basal export by absorptive enterocytes than fatty acids of longer chain length (B-C12, B-C16), which appear to require catabolism via fatty acid oxidation to generate a shorter fatty acid derivative (B-C6) that then can be readily exported or diffuse to the basal reservoir.

Although total basal-reservoir fluorescence and TLC show that etomoxir reduces fatty acid export in B-C12 and B-C16 cultures, these results cannot rule out the possibility that etomoxir reduces basal-reservoir fluorescence by inhibiting apical fatty acid import, thus resulting in less intracellular fatty acid available to undergo fatty acid oxidation. To test whether etomoxir reduced apical fatty acid import, fluorescence from the apical reservoir of B-C12– and B-C16–treated cultures was quantified after 6 hours of etomoxir treatment (Figure 8F). If etomoxir reduced fatty acid import, there would be more total B–fatty acid fluorescence in the apical reservoir compared with control, however, the data showed the opposite, that etomoxir significantly reduced apical–reservoir fluorescence in cultures exposed to B-C12 and B-C16 (Figure 8F). These results support the conclusion that etomoxir does not reduce apical fatty acid uptake.

The finding that etomoxir significantly reduced both apical and basal reservoir fluorescence was somewhat surprising and raised the hypothesis that impaired fatty acid oxidation resulted in accumulation of B–fatty acids in the cellular reservoir. To test this, absorptive enterocyte monolayers were treated with B-C12 or B-C16 along with vehicle or etomoxir. After 6 hours of apical exposure, absorptive enterocyte monolayers were imaged to quantify cellular fluorescence (Figure 8D and E). Etomoxir significantly increased cellular reservoir fluorescence in B-C12/B-C16–treated cultures, further supporting the conclusion that impaired fatty acid oxidation reduced fatty acid export (Figure 8D and E). Together, these data support the conclusion that medium- and long-chain fatty acids are catabolized by fatty acid oxidation to smaller-chain fatty acids that can be exported as free fatty acids (Figure 8G).

Metformin and C75 Potentiate Fatty Acid Oxidation in Absorptive Enterocyte Monolayers and Increase Basolateral Export of Oxidized Fatty Acids

To further support a role for fatty acid oxidation in regulating basal export of B–fatty acids, it was hypothesized that treating absorptive enterocyte monolayers with drugs that augment fatty acid oxidation would result in increased fatty acid export. To test this hypothesis, absorptive enterocyte monolayers were treated independently with metformin and C75, drugs that show fatty acid oxidation augmentation in other cell types., Metformin, commonly known as an antidiabetic drug, potentiates fatty acid oxidation by preventing formation of the CPT1 inhibitor malonyl-CoA and C75, a weight loss–inducing drug, potentiates fatty acid oxidation by increasing CPT1 activity (Figure 9F). After metformin or C75 exposure, B-C12 or B-C16 was applied to the apical reservoir to mimic postprandial fatty acid exposure. Media from the basal reservoir was collected at 2-hour intervals for 6 hours and total fluorescence was quantified to measure fatty acid export. Metformin and C75 significantly increased fluorescence in the basal reservoir of B-C12– and B-C16–treated monolayers (Figure 9A). Strong barrier function through the duration of B–fatty acid and agonist treatment was confirmed by lack of Dextran-647 in the basal reservoir (Figure 9B). TLC was performed to determine the lipid species exported to the basal reservoir of B-C12 and B-C16 cultures (Figure 9C and D). A lipid species corresponding to B-C6 was the most abundant B–fatty acid species in the basal reservoir of B-C12– and B-C16–treated cultures (Figure 9C and D), indicating that B-C6 is the primary B–fatty acid metabolite exported.

Figure 9

Enhancing fatty acid oxidation augments fatty acid export. (A) Quantification of basal-reservoir fluorescence from B-C12– or B-C16–treated absorptive enterocyte cultures treated with vehicle (dimethyl sulfoxide [DMSO]), C75, or metformin (Met). Bonferroni multiple comparisons test. ∗P < .05, ∗∗∗P < .001, ∗∗∗∗P < .0001. (B) Quantification of Dextran-647 in the basal reservoir of absorptive enterocyte monolayers from basal samples of fatty acid–handling screens taken at 2, 4, and 6 hours. Bonferroni multiple comparison test.∗∗∗∗P < .0001. Experiments were performed in triplicate (n = 3 absorptive enterocyte Transwell cultures). (C and D) Left: TLC of basal-reservoir media from B-C12– or B-C16–treated absorptive enterocyte cultures treated with vehicle, C75, or metformin. Right: Quantification of the fluorescence contributed by the B-C6 band divided by the cumulative fluorescence of all bands in each TLC lane. Bonferroni multiple comparisons test. ∗P < .05, ∗∗P < .01. (E) Basal-reservoir fluorescence of B-C12–treated cultures treated with either vehicle, C75, metformin, C75 and etomoxir, or metformin and etomoxir. Fluorescence was measured 6 hours after B–fatty acid application. Bonferroni multiple comparisons test. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001. (F) Proposed mechanism of enhanced fatty acid export by C75 and metformin. Concentrations of C75, metformin, and etomoxir were 40 umol/L, 3 mmol/L, and 100 umol/L, respectively. A previous study using TLC to identify B-lipid species with the same solvents used in this study reported NFBs that do not correspond to B-lipid standards used. ∗Location of B-C4, B-C6, and B-C8 were inferred from the experiment in Figure 10. Experiments were performed in triplicate (n = 3 absorptive enterocyte Transwell cultures). ACC, acetyl-CoA carboxylase; AMPK, AMP-activated protein kinase; B-Cx, BODIPY-lipids; FA, fatty acids; FAO, fatty acid oxidation; Met, metformin; TAG, triglyceride; Veh, vehicle.

Figure 10

( Titles above plots indicate the gate from where the cells came (ie, the previous density plot). (A) Initial doublet discriminator, with exclusion of likely red blood cells and immune cells. (B) Forward-scatter–based doublet discriminator. (C) Back-scatter–only doublet discriminator. (D) Forward-scatter width-based doublet discriminator. (E) Final gate used to distinguish live cells based on negative gating for Annexin V-APC. (F and G) Violin plots showing distributions of quality control (QC) parameters including the number of total reads per cell, number of genes counted in each cell, and the percentage of mitochondrial reads per cell. (F) Prefiltering distribution of QC parameters. (G) Postfiltering distribution of QC parameters. (H) Region of cells overlaid on UMAP. BSC, backscatter; FSC, forward scatter; Jej, jejunum; Ile, ileum; UMAP, uniform manifold approximation and projection.

To probe whether the B–fatty acid export effects were dependent on fatty acid oxidation, absorptive enterocyte monolayers exposed to B-C12 or B-C16 were co-treated with each of the fatty acid oxidation potentiators, C75 or metformin, and the fatty acid oxidation inhibitor, etomoxir. After 6 hours of B–fatty acid exposure and drug treatment, basal reservoir fluorescence was quantified. Etomoxir blocked the ability of C75 and metformin to increase basal reservoir fluorescence in cultures exposed to B-C12 and B-C16 (Figure 9E), supporting the conclusion that metformin and C75 function in a fatty acid oxidation–dependent manner to increase basal export of oxidized long- and medium-chain fatty acid metabolites (Figure 9F).

Discussion

Once born from an ISC, progenitor cells transition through a number of lineage states during their 7-day lifespan in vivo. Mouse studies suggest the absorptive lineage transitions through functional maturation states classified as early, intermediate, and mature phases as they migrate up the villus axis. These functional maturation states are associated with different biological roles related to antimicrobial functions early on and progress into nutrient handling function toward their late and terminal maturation stage. Here, we show by single-cell transcriptomics that human jejunal absorptive enterocytes in vivo generally share a similar maturation defined by early, intermediate, and late maturation phases. Importantly, these states correlate with discrete lipid-handling gene profiles as absorptive enterocytes mature. Although our study focused on lipid-handling genes and mechanisms, the transcriptomic data sets will be useful to evaluate aspects of nutrient handling (ie, carbohydrate, protein, and vitamin metabolism) in primary human absorptive enterocytes.

The absorptive lineage is the default pathway taken by progenitor cells unless the master regulator of secretory lineage fate, Atoh1, is expressed. Chromatin states across the genome are associated with hardwiring of some lineage-specification programs,, and dynamic chromatin states have been described as cells move through ISC to secretory and absorptive lineage states., Data presented here show that removal of ISC maintenance factors from ISC monolayers promote a stereotypical absorptive enterocyte lineage program in an epithelial-autonomous manner, suggesting human absorptive enterocytes are hardwired to undergo progressive maturation over time. Because our in vitro absorptive enterocyte monolayer system generally mimics in vivo absorptive enterocyte maturation and lifespan, it is highly suited to define the intrinsic nature of chromatin dynamics through human absorptive enterocyte maturation and is amenable to testing how extrinsic influences such as dietary factors or the microbiome might influence chromatin states and associated gene expression.

The majority of dietary lipids consist of long-chain fatty acids. Long-chain fatty acids undergo intracellular esterification to triglycerides, are packaged into chylomicrons, exported through the basal membrane, and distributed through the lymphatic system., Previous studies have shown that short-chain fatty acids and medium-chain fatty acids can bypass the chylomicron-lymphatic pathway and pass unesterified into the portal vein. Because long-chain fatty acids can be oxidized to short-chain fatty acids and medium-chain fatty acids via fatty acid oxidation, we hypothesized that fatty acids derived from fatty acid oxidation of long-chain fatty acids could be exported through the basal membrane as free fatty acids. Using our culture system, we show that smaller-chain fatty acids generated from fatty acid oxidation of long-chain fatty acids can be exported unesterified across the basal membrane of absorptive enterocytes. We show a dependency on fatty acid oxidation in regulating basal export of long-chain fatty acid–derived fatty acids because their basal export was decreased by inhibiting fatty acid oxidation with etomoxir and increased by enhancing fatty acid oxidation with C75 and metformin. This novel pathway of long-chain fatty acid–derived fatty acid export might explain why patients with abetalipoproteinemia show distribution of the majority of dietary fatty acids yet are unable to secrete chylomicrons.

Basal export of free fatty acids regulated by fatty acid oxidation could be involved in other physiological responses to dietary lipids, disease etiologies, and pharmaceuticals that target fatty acid oxidation pathways. Our findings show that metformin, a satiety-inducing/glucose-lowering drug, enhances basal fatty acid export. Stimulation of free fatty acid receptors, which are restricted to the basal surface of enteroendocrine cells immediately adjacent to absorptive enterocytes, stimulate secretion of glucose-lowering glucagon-like peptide 1 (GLP-1) and satiety-inducing peptide YY (PYY) gut hormones. Reports have shown that metformin elicits intestinal GLP-1 secretion by enteroendocrine cells and that this effect significantly contributes to systemic glucose lowering. This raises the possibility that the absorptive enterocyte–enteroendocrine cell axis uses fatty acid oxidation to augment basal fatty acid export, which may be a mechanism underlying some of the efficacy of metformin.

Importantly, exposure of metformin to immortalized enteroendocrine cell lines fails to stimulate GLP-1 secretion, suggesting metformin does not act directly on enteroendocrine cells to stimulate GLP-1 secretion but rather through a more complex mechanism. In light of our results showing metformin increases fatty acid export, metformin may act to increase GLP-1 secretion via an absorptive enterocyte–fatty acid oxidation–enteroendocrine cell axis. In this scenario, apically localized medium-chain fatty acids and long-chain fatty acids are absorbed by absorptive enterocytes, catabolized to shorter-chain fatty acid species that can passively diffuse through the basolateral membrane, where they interact with nearby free fatty acid receptors on enteroendocrine cells to stimulate GLP-1 release. Further development of our absorptive enterocyte culture system to support co-culture of enteroendocrine cell and absorptive enterocytes with loss and gain of function for key fatty acid oxidation genes will be required to test this hypothesis.

Aside from basic science applications, our high-throughput, 96-well, absorptive enterocyte culture platform is highly suitable for drug screening and validation. Compared with standard 12-well Transwell plates, scaling the platform to 96-wells increases the plate form factor by 8 while simultaneously requiring approximately 8 times fewer cells. This substantially increases the number of biological and technical replicates that can be performed per plate and reduces plate-to-plate variability when performing experiments to evaluate therapeutic indices for a drug. Real-time quantification of fatty acid mobilization through the epithelial barrier by plate reader allows for kinetic studies, while the planar format of these cultures facilitates simultaneous high-content microscopic readouts. Although the absorptive enterocyte monolayer surface and reservoir sample volumes are small, we show that sufficient material for RNA sequencing and highly sensitive detection of B-conjugated fatty acids and metabolites by plate reader and TLC remains. The scalability, physiological relevance, and sensitivity of our platform to detect changes in fatty acid handling could facilitate the discovery of treatments for metabolic disorders impacted by intestinal fatty acid handling such as obesity, dyslipidemia, and diabetes.

Materials and Methods

Donor Selection

Human donor intestines were accepted and received from HonorBridge (formerly Carolina Donor Services, Durham, NC) based on the following donor acceptance criteria: age 65 years or younger; brain-dead only (as opposed to donation after cardiac death); and negative for human immunodeficiency virus, hepatitis, rapid plasma region (RPR) test for syphilis, tuberculosis, or Coronavirus Disease 2019. Tissue from a 29-year-old Caucasian man was used for single-cell dissociation and scRNAseq. Tissue from a 51-year-old African American man was used for tissue culture, in situ hybridization, scRNAseq of collagen-grown ISCs and absorptive enterocyte monolayers, and bulk RNAseq. Human donors had no history of bowel surgery, severe abdominal injury, cancer, or chemotherapy. Increased-risk donors (ie, history of incarceration or intravenous drug use) were accepted, provided negative infectious disease results were obtained. In addition, donor cases in which the pancreas was placed for transplant were excluded given that pancreatic transplants require removal of proximal small intestinal tissue.

Organ Resection and Single-Cell Dissociation

Whole human intestines were transported to University of North Carolina Chapel Hill in ice-cold University of Wisconsin Solution, with tissue dissection beginning within 8 hours of cross-clamping. First, fat and connective tissue were trimmed from the donated organs and intestines were subdivided into 6 regions after measurement. For the small intestine, the proximal 20 cm was deemed duodenum. Jejunum and ileum were determined through an even split of the remaining small intestine. Two 3 × 3 cm2 resections were isolated from the center of jejunum and ileum for dissociation.

Resections were incubated in 10 mmol/L N-acetyl-cysteine at room temperature for 30 minutes to remove mucus, then tissue was moved to ice-cold isolation buffer, which consisted of 5.6 mmol/L Na2HPO4, 8.0 mmol/L KH2PO4, 96.2 mmol/L NaCl, 1.6 mmol/L KCl, 43.4 mmol/L sucrose, 54.9 mmol/L D-sorbitol, and 100 umol/L Y27632, and then washed several times by gently inverting the tubes. Tissues then were incubated in Isolation Buffer with 2 mmol/L EDTA and 0.5 mmol/L dithiothreitol, and then shaken vigorously to remove crypts. Shakes were repeated several times, checking for crypts and/or villi each time. High-yield small intestinal shakes were pooled to approximate 1:1 villus-to-crypt tissue by cell mass. Crypts and villi were dissociated to single cells using 4 mg/mL protease VIII in Dulbecco's phosphate-buffered saline (DPBS) + Y27632 on ice for approximately 45 minutes with trituration via a P1000 micropipette every 10 minutes. Dissociation was checked on a light microscope and then clumps were removed using filtration.

Cell Sorting, Library Preparation, and Sequencing

Single cells were washed with DPBS + Y27632 and then resuspended in Advanced Dulbecco’s modified Eagle medium (DMEM)/F12 + 1% bovine serum albumin + Y-27632. Annexin V-Allophycoallenine (APC) (1:100) was added for live/dead staining and 1 TotalSeq Anti-Human Hashtag Antibody (Biolegend, San Diego, CA) per region to allow for tracking all 6 regions with a single library preparation. Cells were washed with Advanced DMEM/F12 + 1% bovine serum albumin + Y27632 and then resuspended in the same solution for sorting on a Sony Cell Sorter SH800Z (Sony, Tokyo, Japan). Cells were gated using forward and backward scatter and Annexin V to enrich for live single epithelial cells (Figure 10A–E). Annexin V staining showed high viability (95.7%) of single cells, with 1788 having passed quality filtering (Figure 10E). Twenty-five thousand cells were collected from each separate region, and then all regions were combined before sequencing. Library preparation was performed with the Chromium Next GEM Single Cell 3’ GEM, Library & Gel Bead Kit v3.1 (10× Genomics, Pleasanton, CA). Sequencing was performed on an Illumina NextSeq 500 (Illumina, San Diego, CA).

scRNAseq Data Processing

After sequencing, reads were aligned to reference transcriptome GRCh38 with the 10× Cell Ranger pipeline (10× Genomics, Pleasanton, CA). Mapped reads were filtered and counted by barcode and Unique Molecular Identifier (UMI) and then transformed into an AnnData object using the Python (Beaverton, OR) implementation of scanpy (v1.7.2). Annotations for cell-cycle phase were added after previously published methods. The number of genes, number of UMIs, and percentage of mitochondrial expression for all cells in each sample were visualized and used to identify thresholds for high-quality cells to include in further analysis (Figure 10F and G, Table 2). After filtering, read counts were log-transformed and normalized to the median read depth of the data set. For both single-cell sequencing experiments no batch correction was performed because each data set was analyzed separately. Variability owing to read count, percentage of mitochondrial reads, and cell-cycle phase were regressed out by simple linear regression. Highly variably genes were identified using Seurat (New York, NY). A total of 2585 and 4186 highly variable genes were used for principal component analysis for the in vivo and in vitro scRNAseq data sets, respectively. Counts for each gene were scaled to have a mean of zero and unit variance. A k-nearest neighbor (kNN) graph was constructed with 10 neighbors and was used to calculate Leiden clusters for both in vivo and in vitro scRNAseq data sets (in vivo clustering parameters: Leiden resolution = 0.5, num_neighbors = 10, num_pcs = 40; in vitro clustering parameters: Leiden resolution = 0.1, num_neighbors = 10, num_pcs = 15) and Epithelial Cell Adhesion Molecule-negative (EPCAM-) cells were removed from the in vivo data set. PAGA (Munich, Germany) was used to initialize UMAP (Ottawa, Ontario) embeddings of Leiden clusters.,

Table 2

Quality Control Parameters for scRNAseq

	In vivo scRNAseq	In vitro scRNAseq
Minimum genes	>500	>1000
Percentage of mitochondrial reads	<75%	<50%
Minimum counts	>3000	>5000
Maximum counts	<50,000	<80,000

Regional hashtag deconvolution followed published methods for both scRNAseq data sets. Briefly, raw hashtag read counts were normalized using centered log ratio transformation followed by k-medoid clustering (k = 6 medoids). Hashtag noise distributions were determined by removing the cluster with the highest expression of a specific hashtag, then a negative binomial distribution was fit to the data of the remaining cells. Cells were considered positive for a hashtag if counts for the specific hashtag were greater than the distribution’s 99th percentile (P < .01) threshold. Cells positive for multiple hashtags were excluded as likely doublets. Cells called for colon and duodenum were removed from the data set. For in vitro data, only hashtags 1 and 3 were kept for analysis.

RNA Velocity

Velocyto (Cambridge, MA) 0.17.16 was used to generate the initial loom file and scvelo (Munich, Germany) 0.2.4 was used for all integration of spliced/unspliced loom file integrations with the processed anndata (Munich, Germany)object and all subsequent trajectory analysis.,, Briefly, to ensure connectedness of the data set, clusters that corresponded to stem cells and the different stages of enterocyte differentiation were used for trajectory analysis. All other clusters were removed from the data set and the remaining clusters were reprocessed and reclustered. The resulting data set was integrated with the spliced and unspliced read counts and 4100 highly variable genes were kept for fitting to the RNA velocity dynamic model. For moment calculation, num_pcs = 40 and num_neighbors = 50. Function arguments for calculating RNA velocity vectors and latent time were default values, except for the specification of the dynamic model in scvelo.tl.velocity.

Differential Gene Expression for scRNAseq

Differential expression analysis was performed using the de.test.wald function in the Python implementation of diffxpy (version 0.7.4). From the output data, significant DEGs (q < .05) were attributed to the cluster with the highest mean value for each gene.

Linear Correlation Analysis

Read count data from both data sets were normalized to the same value and then log-transformed. The mean expression values for each lipid-handling gene were calculated per scRNAseq cluster and per group of in vitro differentiation bulk RNAseq. Pairwise comparisons of mean gene expression values for each in vivo differentiation state and each in vitro differentiation state were made (Figure 5B). A line was drawn representing a perfect correlation of in vitro bulk RNAseq expression to in vivo scRNAseq expression (Figure 5B). Residuals for each lipid-handling gene were calculated based on the deviation from this line. The residual sum of squares thus represents a quantitative measure of the overall similarity of expression of lipid-handling genes between each in vivo differentiation state and each in vitro differentiation state, with a lower value indicating a better fit to the line describing a perfect correlation.

2D Collagen Scaffold Preparation

2D collagen scaffolds were made by diluting ice-cold type I rat tail collagen (354236; Corning, Corning, NY) to 1 mg/mL in ice-cold neutralization buffer according to an established protocol. Briefly, 6-well tissue-culture plates (3516; Corning) were incubated at 37ºC for 1 hour before 1 mL diluted rat-tail collagen was poured into each well. After collagen was poured into each well of a 6-well plate, plates were tilted to completely coat the bottom of each well. Plates then were incubated at 37ºC for 1 hour before being overlaid with 3 mL room-temperature DPBS. Collagen scaffolds were kept overlaid with DPBS at room temperature until seeded with ISCs.

2D Collagen ISC Culture and Passaging

Primary human jejunal crypts of Lieberkühn were isolated and cultured on 2D collagen scaffolds according to an established protocol. Briefly, after isolation, crypts were suspended in expansion media (EM) with 10 mmol/L Y27632 at a density of approximately 1667 crypts/mL. DPBS was removed from 6-well collagen-coated plates and immediately seeded with 3 mL suspended crypts. Media was removed the following day and replaced with fresh EM without Y27632. Media was changed every other day.

Six-well collagen ISC cultures were passaged when they became approximately 70% confluent according to an established protocol. Briefly, ISCs containing collagen scaffolds were detached from the tissue culture plate by applying a 1-mL pipette tip to the side of the well and moving the tip around the entire perimeter. Once the collagen scaffold detached from the well, it was transferred to a 5-mL canonical using a 1-mL pipette. Next, 1 mL EM and 10 mmol/L Y27632 were added to the 5-mL conical containing the collagen scaffold and ISCs. To digest the collagen scaffold, 100 uL collagenase type IV was added to the conical and incubated at 37ºC for 10 minutes. Once the collagen was dissolved, ISCs were pelleted at 800 × g for 3 minutes. The supernatant was removed and ISCs were resuspended in 5 mL DPBS. ISCs were pelleted and supernatant was removed as before. To dissociate ISC monolayers, ISCs were resuspended in 150 uL TrypLE Express (Thermo Fisher Scientific, Waltham, MA) containing 10 mmol/L Y27632 and incubated at 37ºC for 5 minutes. After incubation, ISC monolayers were triturated by pipetting up and down 10 times against the bottom of the tube with a 1-mL pipette. Note that ISC monolayers were not dissociated to single cells. After trituration, ISCs were pelleted, and supernatant was aspirated as before. ISCs were resuspended in 9 mL EM with 10 mmol/L Y27632 and seeded in three 6-well collagen scaffolds at 3 mL/well. Passaging was performed until the desired number of cells was obtained.

Transwell Preparation, ISC Seeding, Expansion, and Differentiation

Transwells were prepared according to an established protocol. Briefly, the apical surface of 12-well (3460; Corning) or 96-well (7369; Corning) Transwell inserts was overlaid with ice-cold 1% Growth Factor Reduced Matrigel (Corning, Corning, NY) diluted in ice-cold DPBS. Transwell plates were left incubating at 37ºC in 5% CO2 overnight. Inserts were rinsed by replacing the 1% Matrigel with DPBS the following day. After Transwell preparation, 2D collagen ISC monolayers were dissociated as previously described and ISCs were suspended in EM containing 10 mmol/L Y27632 at approximately 280,000 cells/mL. Suspended ISCs were seeded on the apical surface of Transwell inserts (500 uL for 12-well Transwells, 150 uL for 96-well Transwells). EM was added to the basal reservoir at the time of seeding (1.5 mL for 12-well Transwells, 240 uL for 96-well Transwells). Apical and basal media was replaced the day after seeding with fresh EM without Y27632. To induce differentiation, EM in apical and basal reservoirs was replaced with DM to initiate differentiation.

Adapting 2D Transwell Cultures for 96-Well Transwell Plates

For scaling to 96-well Transwell cultures, 25 mL media reservoirs and a 12-channel pipette were used for transferring 1% Matrigel, DPBS, and media to 96-well Transwell plates. Ninety-six–well Transwell inserts were covered with a hydrophobic porous film (12-631; Genesee Scientific, San Diego, CA) to minimize well-to-well variability caused by evaporation while maintaining gas exchange. For inducing differentiation in 96-well Transwell cultures, EM in the apical and basal surface was replaced with DM (apical, 100 uL; basal, 240 uL).

Single-Cell Dissociation of Transwell Epithelial Monolayers

Media was removed from ISCs expanded on collagen and from absorptive enterocytes on 12-well Transwell inserts and washed with 1× DPBS. A total of 1.5 mL of 3 mmol/L EDTA in phosphate-buffered saline was applied to ISCs expanded on collagen and on apical and basal reservoirs of 12-well Transwell cultures until most cells were lifted off the collagen or Transwell surface as determined by visual inspection every 2 minutes under an inverted microscope. EDTA was aspirated using a P1000 micropipette and redistributed over the Transwell surface to facilitate detachment of cells from the Transwell at 2-minute intervals until the majority of cells had detached. After the cells detached, 1.5 mL DPBS was applied to each well and rinsed by aspirating and re-applying the EDTA/DPBS. Cells then were aspirated and pelleted by centrifugation at 500 × g for 5 minutes. The supernatant was removed and replaced with 1 mL of 4 mg/mL cold protease in DPBS. Cells in cold protease were incubated on ice and pipetted semivigorously 10 times every 2 minutes before visual examination of dissociation under an inverted light microscope. This was repeated until all cells were singlets. After single-cell dissociation, the cold protease was quenched with Advanced DMEM/F12 + 1% fetal bovine serum. Cells were pelleted as described earlier and resuspended in Advanced DMEM/F12 + 1% fetal bovine serum.

Bulk RNA Sequencing Preparation, Processing, and Analysis

To investigate the dynamic changes in gene expression as ISCs differentiate into absorptive enterocytes in vitro, RNAseq was performed on human intestinal epithelial monolayers immediately before seeding onto Transwells (day 0) and on days 2, 5, 7, 10, and 11 of differentiation on Transwells. Three samples were collected from each time point and RNA was extracted using the RNAqueous-Micro Total RNA Isolation Kit according to the manufacturer’s protocols and stored at −80°C. RNA quality was assessed before library preparation by using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) to determine the RNA integrity number. After confirmation that each sample had an RNA integrity number of at least 8, integrated fluidic circuits for gene expression and genotyping analysis were prepared using the Advanta RNA-Seq NGS Library Prep Kit (Fluidigm, South San Francisco, CA) for the Fluidigm Juno and sequenced with the Fluidigm Biomark HD system. Gene level expression was obtained through pseudo-alignment of reads to human genome GRCh38 using Kallisto (Berkeley, CA). Expression values for plotting were obtained by Trimmed Mean of M-values (TMM) normalization across all samples using EdgeR (Parkville, Australia) package.90, 91 Sequencing data from each time point were combined and then normalized to the data set median and log transformed. Principal component analysis was performed with scikitlearn (v0.24.0, Rocquencourt, France).

TEER

Barrier integrity of 96-well Transwell cultures was monitored by quantifying TEER using EVOM or EVOM3 (World Precision Instruments, Sarasota, FL) TEER meters in conjunction with STX100C96 electrodes (World Precision Instruments).

Basal Fluorescence Quantification

Etomoxir (100 umol/L), 40 umol/L C75, 3 mmol/L metformin, or vehicle (dimethyl sulfoxide) were suspended in 100 uL DM and applied to the apical surface of absorptive enterocyte Transwell cultures for 1 hour. After incubation, apical media was replaced with 100 uL DM containing drugs or vehicle, 1 umol/L dextran, Alexa Fluor 647 (Dextran-647), and 20 umol/L of either BODIPY-FL C5, BODIPY-FL C12, or BODIPY-FL C16 (Figures 7B and 9B). Basal media (50 uL) was collected at 2, 4, and 6 hours and replaced with an equivalent volume of DM. Quantification of basal fluorescence was performed using a CLARIOstar Plus Microplate Reader (BMG Labtech Inc., Ortenburg, Germany). Fluorescence arbitrary units were converted to picomoles using a standard curve.

TLC

For the assessment of fatty acid handling via TLC (Figure 4), BODIPY-FL C12 was applied to the apical surface of absorptive enterocyte Transwell cultures. Apical and basal media were collected after 6 hours. Cells were released from Transwell inserts via application of 100 uL 1× TrypLE Express containing 10 mmol/L Y27632. Cells were lysed by undergoing 3 freeze-thaw cycles. Lipid extracts were generated from apical and basal media along with cell lysates using the Bligh and Dyer method. Lipid extracts were resuspended in 100% ethanol and spotted on silica gel TLC plates. Lipid species were separated by placing spotted silica gel plates in a glass chamber containing either a polar (chloroform/ethanol/triethylamine/water, 30:34:30:8 mL) or protonating (petroleum ether/ethyl ether/acetic acid, 45:5:0.5 mL) solvent.

The following BODIPY analogs were used as standards to identify lipid species: BODIPY-FL C5 (B-C5), BODIPY-FL C12 (B-C12), BODIPY-FL C16 (B-C16), BODIPY 493/503 (B), and β-BODIPY-FL C12-1-Hexadecanoyl-sn-Glycero-3-Phosphocholine (HPC) (phospholipids). 18:1-18:1-C11 TopFluor TG (Avanti Polar Lipids, Birmingham, AL) is a triglyceride conjugated to a fluorophore with similar properties (chemical structure and excitation/emission) as BODIPY and was used to infer the location of triglycerides on TLC plates owing to BODIPY–triglyceride analogs not being commercially available at the time of this publication. Only basal media samples were collected and assessed via TLC from the experiments in Figures 5 and 6. After incubation in solvent, spotted silica gel plates were dried and scanned on an iBright FL 1000 Imager (Invitrogen, Waltham, MA) to detect fluorescent lipid species. Excitation and emission channels were set to 455–485 and 508–557, respectively. A previous study using TLC to identify B-lipid species with the same solvents used in this study reported NFBs that do not correspond to B-lipids.

The chain length of unknown B-labeled fatty acids that appeared in the basal media of absorptive enterocyte cultures was inferred by generating a standard curve of the retention factors of B-FA standards and their chain length. The retention factor of the unknown B-labeled fatty acids was used to interpolate the chain length using the standard curve mentioned previously (Figure 10). The 3 unknown B-FAs corresponded to B-C4, B-C6, and B-C8 (Figure 10).

Microscopy of Primary Human Jejunal Tissue

Post-mortem human jejunal tissue was fixed by incubating in 10% neutral buffered formalin overnight at 4ºC. Tissue was rinsed 3 times in water 3 times the following day and stored in 70% ethanol until embedded. Tissue was embedded in paraffin wax. Glass slides were prepared by sectioning embedded tissue. In situ hybridization was performed using RNAscope probes according to the recommended protocol (Table 3). Images were taken using an LSM710 confocal microscope (Zeiss, Jena, Germany).

Table 3

Reagents Used

Reagents	Company	Catalog number
L-WRN cells	ATCC (Manassas, VA)	CRL-3276
Advanced DMEM/F12	Thermo Fisher (Waltham, MA)	12634010
Glutamax, 100×	Thermo Fisher (Waltham, MA)	35050061
1 mol/L HEPES buffer	Corning (Corning, NY)	25-060-CI
Primocin	VWR (Radnor, PA)	MSPP-ANTPM2
N-acetyl-cysteine	MilliporeSigma (Burlington, MA)	A9165
EGF, murine	Peprotech (Rocky Hill, NJ)	315-09
Nicotinamide	MilliporeSigma (Burlington, MA)	N0636-100G
B27	Thermo Fisher (Waltham, MA)	12587-010
Gastrin	MilliporeSigma (Burlington, MA)	G9145-1MG
PGE2	Peprotech (Rocky Hill, NJ)	3632464
A 83-01	MilliporeSigma (Burlington, MA)	SML0788-5MG
SB202190	Peprotech (Rocky Hill, NJ)	1523072
DPBS	Thermo Fisher (Waltham, MA)	14190-144
Matrigel	Corning (Corning, NY)	354230
Y27632	Selleckchem (Houston, TX)	S6390
Etomoxir	MilliporeSigma (Burlington, MA)	E1905-5MG
C75	Cayman Chemical (Ann Arbor, MI)	10005270
Metformin	Cayman Chemical (Ann Arbor, MI)	13118
Dimethyl sulfoxide	MilliporeSigma (Burlington, MA)	D2650-5X5ML
Glyoxal solution 40 wt % in H2O	MilliporeSigma (Burlington, MA)	128465
TrypLE Express	Thermo Fisher (Waltham, MA)	12605010
Dextran 647	Thermo Fisher (Waltham, MA)	D22914
a-BODIPY 530/550-C12-HPC	Thermo Fisher (Waltham, MA)	D3792
BODIPY FL C16	Thermo Fisher (Waltham, MA)	D-3821
BODIPY FL C12	Thermo Fisher (Waltham, MA)	D-3822
BODIPY-FL C5	Thermo Fisher (Waltham, MA)	D-3834
BODIPY 493/503	Thermo Fisher (Waltham, MA)	D-3922
18:1-18:1-C11 TopFluor TG	Avanti Polar Lipids (Birmingham, AL)	810298C-1mg
Na2HPO4	MilliporeSigma (Burlington, MA)	S7907
KH2PO4	MilliporeSigma (Burlington, MA)	P5655
NaCl	MilliporeSigma (Burlington, MA)	S5886
KCl	MilliporeSigma (Burlington, MA)	P5405
Sucrose	Thermo Fisher (Waltham, MA)	BP 220-1
D-sorbitol	Thermo Fisher (Waltham, MA)	BP439-500
EDTA	Corning (Corning, NY)	46-034-Cl
DTT	Thermo Fisher (Waltham, MA)	BP172-5
Protease VIII	MilliporeSigma (Burlington, MA)	P5380
Bovine serum albumin	Thermo Fisher (Waltham, MA)	BP1600-1
TotalSeq-B0251 anti-human hashtag 1 antibody	BioLegend (San Diego, CA)	394631
TotalSeq-B0251 anti-human hashtag 2 antibody	BioLegend (San Diego, CA)	394633
TotalSeq-B0251 anti-human hashtag 3 sntibody	BioLegend (San Diego, CA)	394635
TotalSeq-B0251 anti-human hashtag 4 antibody	BioLegend (San Diego, CA)	394637
TotalSeq-B0251 anti-human hashtag 5 antibody	BioLegend (San Diego, CA)	394639
TotalSeq-B0251 anti-human hashtag 6 antibody	BioLegend (San Diego, CA)	394641
Annexin V-APC	BioLegend (San Diego, CA)	640920
RNAqueous-Micro Total RNA Isolation Kit	Thermo Fisher (Waltham, MA)	AM1931
Chromium Next GE M Single-Cell 3’ GEM, Library and Gel Bead Kit v3.1	10x Genomics (Pleasanton, CA)	PN-100012
RNAscope Probe - Hs-FABP1-C3	Advanced Cell Diagnostics (Newark, CA)	534801-C3
RNAscope Probe - Hs-APOA4	Advanced Cell Diagnostics (Newark, CA)	857841
RNAscope Probe - Hs-OLFM4-C2	Advanced Cell Diagnostics (Newark, CA)	311041-C2
RNAscope Probe - Hs-DMBT1	Advanced Cell Diagnostics (Newark, CA)	478711

DTT, dithiothreitol; EGF, epidermal growth factor; HPC, 1-Hexadecanoyl-sn-Glycero-3-Phosphocholine L-WRN, L-cell derived Wnt3a, R spondin 3, and Noggin; PGE2, prostaaglandin E2.

Microscopy of 2D Absorptive Enterocyte Monolayers

Absorptive enterocyte monolayers were treated with vehicle or 20 umol/L etomoxir and exposed apically to B-C12 or B-C16 for 6 hours. After 6 hours of apical B-C12 or B-C16 exposure, absorptive enterocyte monolayers were rinsed with fresh DM and fixed in 40% glyoxal solution for 20 minutes. After fixation, absorptive enterocyte monolayers were rinsed with 1× DPBS. Next, the Transwell membrane containing the fixed absorptive enterocyte monolayers were removed from the Transwell inserts and placed on glass slides. Absorptive enterocyte monolayers then were overlaid with mounting media and covered with a glass coverslip. Fluorescent images used for quantification were taken at 20× magnification on an Olympus IX2-UCB (Olympus, Shinjuku City, Tokyo, Japan) microscope. Representative fluorescent images of absorptive enterocyte monolayers in Figure 5 were taken at 40× magnification on a Keyence BZ-X810 microscope (Keyance, Osaka, Japan).

Code and Data Availability

Sequencing data sets will be available on the NCBI Gene Expression Omnibus under accession number GSE186583. Python scripts showing the main parts of our analysis will be available on GitHub.