A basolateral macrophage layer changes what intestinal organoid infection models can claim

What the work claims

Goertz, Cipelli, Buettner and colleagues (Goertz performed the experiments with Cipelli; the study was designed and interpreted by the senior author team) built a co-culture in which a differentiated monolayer derived from human colon organoids sits on a 1-micrometer-pore Transwell membrane, and THP-1 macrophage-like cells are attached to the underside of that membrane, on the basolateral side¹. The geometry matters: it reproduces the anatomical arrangement of the lamina propria, where resident macrophages sit beneath the epithelium, separated from the lumen by a single cell layer and a basement membrane.

The central claim has two parts. First, the platform itself: macrophages can be positioned basolaterally without disrupting epithelial barrier integrity, as measured by transepithelial electrical resistance (TEER). Second, the functional result: when the macrophages are activated with interferon-gamma (IFN-gamma, 20 ng/mL) and lipopolysaccharide (LPS, 100 ng/mL) for 48 hours before infection, apical challenge with Listeria monocytogenes or Salmonella Typhimurium at a multiplicity of infection of 10 produces a pronounced reduction in intracellular bacterial burden at 24 hours, together with reduced LDH-release cytotoxicity, compared with epithelial monolayers alone. Under basal conditions, PMA-differentiated but unactivated THP-1 cells reduced Listeria burden but did not significantly affect Salmonella¹.

This is a methods paper with a mechanistic observation attached, and it should be weighted as such: the platform is the durable contribution, the transcriptomics are hypothesis-generating.

How it works

The experimental sequence is worth understanding precisely, because the details determine what a screen built on this system would actually measure. Colon organoids from a single healthy donor (ethics permit 3082-2016, Hannover Medical School) were dissociated and seeded onto Matrigel-coated Transwells, expanded for six days, then switched to differentiation medium for 48 hours. Differentiation reduced the stem cell marker LGR5 and raised the enterocyte marker ALPI and goblet marker MUC2, and TEER climbed through day 8. THP-1 cells, a human monocytic leukemia line, were differentiated into macrophage-like cells with phorbol 12-myristate 13-acetate (PMA) for 48 to 72 hours, then seeded onto the inverted insert so they attached to the basolateral membrane surface¹.

Infection is applied apically, as in the gut lumen, and intracellular bacteria are quantified with a gentamicin protection assay, which kills extracellular bacteria and leaves only the invaded population to be enumerated as colony-forming units. The activation step is the hinge of the paper: resting macrophage-like cells cleared Listeria but not Salmonella; IFN-gamma/LPS-activated cells cleared both and reduced epithelial death¹. The readout is therefore not a static property of the tissue model. It is a function of the immune cell's state.

Bulk RNA sequencing of the co-cultures (three replicates per group) found a coordinated reduction in HLA-D and broader MHC class II-associated gene expression, altered S100A8/S100A9 (the two subunits of the antimicrobial complex calprotectin), and an epithelial lineage shift: ALPI down, MUC2 up, with LGR5 and OLFM4 moving in opposite directions. The authors are careful here, and the care is load-bearing: bulk RNA from a co-culture cannot assign a transcript to a cell type, so these changes may reflect altered gene expression, altered cell composition, or both¹.

Where a skeptic should push

The single most load-bearing assumption is that THP-1 cells, activated by a blunt inflammatory stimulus, stand in for intestinal lamina propria macrophages. They do not, and the authors say so plainly in their limitations. THP-1 is a leukemia-derived line with a monocyte-like baseline; tissue-resident gut macrophages are phenotypically distinct, tolerogenic, and positioned by different cues. Every transcriptional and functional readout in this paper is therefore a measurement of THP-1 biology layered onto colon epithelium, not of human mucosal immunity. That is not a disqualification for a platform paper, but it caps the generality of any claim built on it.

Second, the generalization base is one donor. A single healthy colon biopsy, passaged into organoids, under one differentiation protocol. The epithelial response to activated macrophages (the ALPI/MUC2 shift, the LGR5/OLFM4 divergence) is a property of this donor's epithelium meeting this cell line under IFN-gamma and LPS, and the paper itself flags that the pronounced changes were observed only under activated, infected conditions and should be read as context-dependent¹.

Third, sample sizes are honest and small. The basal-condition Salmonella experiment was a single experiment with three replicates. The activated-condition burden data are from at least three independent experiments; cytotoxicity from two. Bulk RNA-seq at n = 3 per group can establish signatures, not mechanisms. The reduced HLA-D signature is explicitly hypothetical in the paper: it may reflect modulation of epithelial inflammatory state, or it may reflect which cells contribute RNA. Demonstrated: positioning works, barrier holds, activated macrophage-like cells reduce bacterial burden and cytotoxicity. Asserted: the transcriptional program behind it, and its relevance to human mucosa¹.

What this means for organoid-based infection and inflammation drug screens

For the foundry subject, organoid models of human organs and the drug discovery built on them, the non-obvious implication is this: the moment you add an immune compartment to an organoid assay, the immune cell's activation state becomes a hidden experimental variable that can flip your endpoint. In this paper the same epithelial-monolayer-plus-macrophage system gave a pathogen-specific effect at baseline (Listeria restricted, Salmonella not) and a broad effect after activation. A compound screen run in the basal state and a screen run in the activated state would optimize for different biology and could rank the same candidates in opposite orders. Any screening program that adopts co-culture models needs to treat immune state as a controlled factor, on the plate map and in the analysis plan, not as an afterthought.

The opportunity is real. The geometry is screening-compatible: apical infection, basolateral sampling, TEER as a continuous barrier readout, LDH as a toxicity readout, flow cytometry for both compartments. A co-culture like this can be run in the multiwell format that drug discovery actually uses, and it closes a genuine gap in intestinal models, which have no resident immune cells by default. It is also a better substrate than epithelial-only organoids for testing immunomodulatory adjuvants, barrier-protective compounds, and antimicrobial combinations where efficacy is partly immune-mediated.

The threat is subtler and worth stating bluntly. If co-cultured cell lines become the default immune compartment because they are cheap and reproducible, the field will build screening cascades that are internally consistent and biologically narrow: they will select for compounds that modulate THP-1 responses to IFN-gamma and LPS, then discover in vivo that lamina propria macrophages do not behave that way. Reproducibility across plates is not validity across biology. The authors' own roadmap, primary or patient-matched macrophages and more donors, is exactly the right one, and it is also the expensive one, which is precisely why it will be skipped unless reviewers and program leads demand it.

The bottom line

Established: a basolateral macrophage-like compartment can be added to colon organoid monolayers without breaking the barrier, and activated macrophages restrict both Listeria and Salmonella while reducing epithelial death. Hypothesis: the HLA-D, calprotectin, and epithelial-lineage transcriptional shifts mediate or accompany this protection. What would confirm it: patient-matched primary macrophages, multiple donors, and single-cell or spatial resolution to assign transcripts to cell types. What would break it: if primary lamina propria macrophages fail to reproduce the restriction under physiologically reachable activation states, or if the epithelial lineage shift proves to be a cell-composition artifact of bulk sequencing. For drug discovery, the platform is adoptable now for mechanism-of-infection studies, but any screen built on it should treat macrophage state and donor identity as first-class variables, because the model's answer changes when they do.