Background

Monocytic myeloid cells are a highly plastic immune cell subset, capable of directing a wide variety of immune and tissue modeling responses depending on their activation state.1 As early responders to tissue injury or infection they direct both initial inflammation and post-inflammatory phases characterized by tissue-healing and fibrosis, making their functions in health and disease manifold. Cancers use various mechanisms to induce favorable phenotypes in myeloid cells, pushing this highly plastic cell type towards an immunoinhibitory phenotype often characterized and labeled as tumor-associated macrophages (TAMs).2 3 TAMs with immunoinhibitory features have also been linked to poor prognosis in colorectal cancer (CRC) cohorts, while detection of macrophage inflammation markers seemed to correlate with better prognosis.4 Thus, targeting TAMs and understanding their response to treatment will be relevant for the development of effective therapeutic strategies.3 5

Patient-derived tumor organoids (PDOs) appear exceptionally promising to investigate the interaction between TAMs or TAM-like cells and cancer cells of individual patients ex vivo, in a sufficiently scalable manner.6 At the same time, the field of macrophage biology has been hampered by conflicting results originating from mouse models, making a case for the development of preclinical human models to better understand the role of myeloid cells in the human tumor microenvironment (TME).7 PDOs were developed during the past decade and have major advantages compared with traditional cell line systems, specifically a high similarity to human cancer as it presents in vivo, including preserved cellular heterogeneity8 and highly stable epigenetic profiles.9 While such PDOs have enabled predicting responses to certain chemotherapeutics as well as radiotherapy, responses to a number of standard-of-care therapeutics were not well recapitulated.10–13 This is likely due to the fact that these PDOs lack cells contributing to the TME, especially immune cells within the experimental set-ups. Some studies have attempted to supplement cells from the immune compartment, investigating T-cell co-culture with PDOs.14 15 Studies exploring the interaction of CRC-PDOs and immune cells have so far often disregarded other elements of the TME such as cancer-associated fibroblasts (CAFs).16 17 We have recently shown, that CAFs are essential for the maintenance of PDO differentiation and a transcriptomic profile similar to phenotypes seen in the TME, also allowing for the omission of multiple additives in the PDO culture medium, which are necessary for PDO growth in monoculture, but are likely to interfere with multiple assay readouts.18 We therefore hypothesize that CAFs are similarly essential for modeling treatment responses of TAM-like cells in a reconstituted primary CRC culture set-up.

Here we generated ex vivo organotypic co-culture assays to study the effects of treatments on monocyte-derived myeloid cells in CRC, reconstituting essential components of the complex microenvironment through the addition of CAFs. An oncolytic influenza A virus (O-IAV), previously generated in our laboratory,19 20 served as a model treatment with strong expected immunostimulatory effects on macrophages. We compared virotherapy to standard-of-care chemotherapeutics, oxaliplatin and 5-fluorouracil (5-FU), which are only loosely associated with anticancer immunity and appear to have both immuno-stimulatory and immuno-suppressive properties.21–23 Importantly, due to practical and ethical restrictions, we only have very limited knowledge of chemotherapy-induced changes in monocytic myeloid cells during the acute phase within the first 24 hours after treatment. Our approach identified differential responses of monocytes and macrophages in the co-culture with primary CAFs. We revealed a strict functional dependency of macrophage phagocytosis on the presence of CAFs in the triple-culture system and characterized unexpected immunostimulatory properties of chemotherapy-induced cell death on both TAM-like cells and CAFs. Our findings re-define the role of monocyte-derived macrophages and CAFs in the TME and underline the utility of the here established culture systems for precision medicine and drug screening approaches.

MethodsCell lines, viruses and chemotherapeutics

Vero African green monkey kidney cells were cultured in OptiPro SFM and were purchased from the American Type Culture Collection (ATCC).

O-IAV-green fluorescent protein (GFP) and O-IAV viruses were generated based on H1N1-NS1116-GFP previously generated in our laboratory,20 with O-IAV viruses expressing the A/Chicken/Kurgan/05/2005 hemagglutinin (GenBank Accession: DQ449632) instead of A/Puerto Rico/8/34 hemagglutinin. The avian polybasic cleavage site was modified to a trypsin site (TETR/GLF), reversing pathogenicity.24 Titration of viruses was performed using a tissue culture infective dose 50% (TCID50) assay.

Oxaliplatin and 5-FU were kindly provided by the internal pharmacy of the Vienna General Hospital.

Catalog numbers and manufacturers of all materials used in this study are available in the online supplemental table 1.

Isolation of primary monocytes and macrophages

Whole blood of young, healthy, age and sex matched volunteers as well as patients with CRC was collected in VACUETTE EDTA tubes. Monocytes were isolated using the EasySep Direct Human Monocyte Isolation Kit.

For macrophage generation monocytes were attached to cell culture plates for 2 hours in Roswell Park Memorial Institute (RPMI)-1640. Adherent monocytes were further incubated in RPMI-1640 supplemented with 10% fetal calf serum and 100 ng/mL macrophage colony stimulating factor (M-CSF) for 7 days, yielding macrophages.

Primary fibroblast culture

CAFs were cultured as previously described.18

Generation of primary CRC organoids

CRC organoids were cultured as previously described.6 18 The source material for organoid generation was CRC tissue dissected from the colon or metastatic liver obtained intraoperatively from the resected tissue. Tissue was cooled on ice for less than 15 min until dissection. Patients were stable during the operation. The pathological diagnosis was colon adenocarcinoma. Tissues were subjected to freezing in a Corning CoolCell system at −80°C in CELLBANKER-2 (ASMBIO) overnight and further stored in liquid N2 for 1–12 months until processing for organoid and fibroblast culture. Samples were included in further experiments if organoid growth was achieved.

Infection of organoids

For Geltrex submerged infection of organoids, domes containing organoids were incubated with 5×102 TCID50/initially seeded cell for 1 hour at 37°C in a 5% CO2 atmosphere under slight agitation and subsequently incubated in culture medium. Domes were subjected to dissociation by 30 min incubation in 2.5 U/mL dispase. Single-cell suspension was achieved by incubation in TrypLE. Readout of infectivity was performed by flow cytometry for cells expressing the viral GFP tag.

For quantification of viral titers, organoids were infected in a Geltrex submerged culture as described above and incubated for 24 hours in the presence of 1 µg/mL L-1-tosylamide-2-phenylethyl chloromethyl ketone-treated trypsin. Supernatants were collected and infective virus titer was quantified by TCID50.

PDO viability

PDOs were seeded on day 0 and subjected to media change on days 2, 4 and 5. Treatments were started with seeding, on day 4, on day 5, or completely omitted for untreated control conditions. For O-IAV, cultures were infected using 5×102 TCID50/initially seeded cell for 1 hour and subsequently incubated in ENAS. For chemotherapy, cultures were incubated in ENAS in the presence and absence of 20 µM oxaliplatin and/or 0.5 mM 5-FU. Cell viability was determined by CellTiter-Glo 3D according to the manufacturer’s instructions. Luminescence was determined using Varioskan LUX.

Triple culture

Fibroblast-conditioned medium was generated by a 3-day low-volume culture of CAFs in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffered advanced DMEM/F12 containing GlutaMAX (1×) and 1% fetal bovine serum (FBS) (basal medium). Primary monocytes from volunteer donors and CAFs were detached using TrypLE and organoids were released from Geltrex using dispase. The cells were embedded in Geltrex domes at a ratio of 20:10:1 (CAFs:monocytes:organoid-cells) for all triple culture assays (for some controls one or more cell types were omitted while keeping the same cell number per culture; if cell types are omitted, detailed description of present cell types is indicated within the figures underneath the subplot). For the set-up, the domes were cultured in basal medium, fibroblast conditioned medium or ENAS for 72 hours. For infection, domes were cultured in basal medium for 3 days, infected using 5×102 TCID50/initially seeded cell for 1 hour and cultured in basal medium for a further 24 hours. For chemotherapy, domes were incubated in a basal medium containing 20 µM oxaliplatin and/or 0.5 mM 5-FU for 24 hours. All domes were then subjected to dispase and TrypLE incubation and the resulting single-cell suspension was either subjected to single-cell transcriptomics as described below or stained with fluorescence-labeled antibodies (online supplemental table 2) and analyzed by flow cytometry.

Triple culture readouts

For H&E and immunofluorescence (IF) stainings extracellular matrix (ECM) domes were fixed in 1% paraformaldehyde for 1 hour and embedded in agarose. Agarose-embedded domes were fixed a second time in 1% paraformaldehyde for 1 hour, dehydrated and embedded in paraffin. 3 µm sections were applied on SuperFrost Plus slides. Slides were heated to 56°C for 12 hours and subsequently rehydrated. H&E staining was performed as previously described.25 For IF, slides were subjected to antigen retrieval at 121°C under high pressure in 10 mM sodium citrate with 0.05% Tween 20 at pH=6. Slides were washed with TRIS-buffered saline solution (TBS), permeabilized in 0.2% Triton X-100 for 5 min, washed again in TBS with 0.05% Tween 20 (TBS-T), and blocked with 10% goat serum, 1% bovine serum albumin, 0.05% Tween 20 and 0.3M Glycin for 60 min. Primary staining was performed using an anti-vimentin antibody at 1:200 dilution and an anti-CD45 antibody at 1:100 dilution at 4°C in a humid chamber overnight. Slides were washed with TBS-T and incubated with secondary antibodies against pan-keratin at 1:100 concentration and Alexa Fluor 546 conjugated anti-rabbit IgG at 1:400 concentration as well, as Alexa Fluor 647 conjugated anti-mouse IgG at 1:400 concentration. Slides were washed with TBS-T and finally stained with 4’,6-diamidino-2-phenylindole, dihydrochloride (DAPI) at 1:1,000 dilution before a final wash with TBS-T and mounting in Fluoromount-G. Slides were imaged as described below.

Supernatants from treated cultures were shock-frozen. Aliquots were thawed and subjected to LEGENDplex analysis according to the manufacturer’s instructions. A customized kit measuring interleukin-6 (IL-6), IL-1β, tumour necrosis factor (TNF) and IL-10 was used. The assay was acquired using the DxFlex flow cytometer. Flow cytometry data was converted to cytokine concentrations by use of LEGENDplex analysis software according to the manufacturer’s instructions.

Sample preparation for single-cell RNA-sequencing

For the CAF-monocyte/macrophage experiment co-cultures at a 2:1 CAF:monocytic cell ratio and single cultures of CAFs or macrophages with equivalent seeding densities were cultured for 24 hours in basal medium. Cells were subsequently counted and untreated CAFs pooled with untreated macrophages or freshly isolated monocytes, and further processed using the 10x Genomics workflow as described below.

Triple cultures were set up and treated as described above. Single cells were stained with DAPI and live DAPI-negative cells were sort purified and used for single-cell sequencing.

Single-cell RNA-sequencing

Single-cell RNA-sequencing (scRNA-seq) was performed using the 10x Genomics Chromium Next GEM 3’ v3.1 (CAF+mono/macro experiment) or 5’ v2 (triple culture experiment) kit and run on a Chromium Single Cell Controller. Complementary DNA recovery, amplification and gene expression libraries were prepared according to the manufacturer’s protocol. Quality control was performed using TapeStation (Agilent) and Qubit Fluorometer (Thermo Fisher). Libraries were sequenced on a NovaSeq SP flow cell (Illumina) with 50 bp read length in paired-end mode.

The scRNA-seq data set was processed, explored and visualized using the Trailmaker (https://app.trailmaker.parsebiosciences.com/) pipeline (Parse Biosciences, formerly Biomage-hosted community instance of Cellenics). Count matrices for individual experiments (CAFs+mono/macro and triple culture) were separately uploaded onto Trailmaker. Employing the distribution of mitochondrial read percentages in each sample, a threshold was established at two median absolute deviations above the median. Cells surpassing this threshold in mitochondrial read percentages were eliminated (mitochondrial content cut-off range: 6.8–12%). This step led to the removal of 2.75–8.65% of cells, depending on the sample. The rlm method of the MASS R package (V.7.3–56) was used to adjust a robust linear model to the relationship between the number of genes with at least one count and the number of uni-molecular identifiers (UMIs) of each droplet.26 A fitted model (with a tolerance level of 1 − alpha, where alpha is one divided by the number of droplets in each sample) was used to predict the number of genes for each droplet and droplets outside the boundaries of the prediction interval were filtered out (alpha value cut-off range: 8.40E-05–6.4E-04, 0.1–0.74% of cells filtered out). The doublets rate was calculated for each sample using the scDblFinder R package (V.1.11.3)27 leading to the removal of 6.06–13.71% of cells. Sample-specific thresholds for all samples are reported in online supplemental table 3. After processing, 1,458–10,334 high-quality cells were kept with overall filtering rates of 12.87–17.07%. The data set underwent log-normalization, followed by the selection of the top 2,000 highly variable genes based on the variance stabilizing transformation method. Subsequently, principal component analysis was conducted, using the top 25 principal components (PCs) for the CAF+mono/macro experiment, 22 PCs for the triple culture experiment and 15 PCs for the re-clustered CAFs from the CAF+mono experiment. These components collectively explained 76–90% of the total variance of experiments. Integration of these components was performed using the Harmony R package.28 Clustering was carried out via Seurat’s implementation of the Louvain method, and cluster-specific marker genes were identified by comparing cells within each cluster against all other cells using the presto package implementation of the Wilcoxon rank-sum test.29 Finally, visualization of the results was achieved through a Uniform Manifold Approximation and Projection (UMAP) embedding.

Differential expression analysis

Differential expression analysis was performed using Trailmaker analysis tool and batch differential expression method. Within a selected cell type, different samples were compared with respective control samples. The resulting tables with deregulated genes (DE) are included in online supplemental table 4. Genes were sorted based on the logFC values (descending for upregulated and ascending for downregulated genes). For further analysis, top 25/100/200 genes with equal or more than 0.5 average expression values were selected, and all the non-coding RNAs and pseudogenes were filtered out. The resulting list of genes was used for the Gene Set Enrichment Analysis (GSEA) using the Enrichr tool.30–32

Flow cytometry and microscopy

All cells were incubated in 40% human AB serum for 5 min before staining and subsequently stained for 30 min. Antibodies used for each experiment are referenced in online supplemental table 2. Cells were acquired using the Gallios G flow cytometer and Kaluza Gallios software or the DxFlex flow cytometer and CytExpert software. For imaging flow cytometry, cells were acquired using the ImageStreamX Mark II and the IDEAS software. Compensation was performed using single stainings of respective cell types treated with 5% Tween 20 for 5 min for viability stains, as well as single stainings of UltraComp eBeads compensation beads with respective antibody stains. Gating, compensation matrix calculation and data visualization was performed in Kaluza Analysis software.

Microscopic images of living cultures were taken with the Axiovert 40 CFL microscope, and the AxioCam ICC 3. H&E and IF staining was imaged using the Vectra Polaris Automated Quantitative Pathology Imaging System.

Statistical analysis

Differences between the two groups were analyzed by Student’s t-test. Differences between three or more groups were analyzed by two-way analysis of variance and subsequent Tukey’s multiple comparisons. Statistical computing as well as scatter plot visualization was performed in GraphPad Prism software.

ResultsCAFs instruct monocytes to gain a TAM-like phenotype

Classification of TAM subtypes is a frequently revisited topic in the field of tumor biology. Current subclassifications drift away from a separation of TAMs expressing an M1 or M2 signature, based on findings that marker genes of both are frequently found on the same myeloid cell in the TME of several cancer types.33 More recent classifications of TAM subsets in CRC are designated based on the expression of markers including FCN1, SPP1 and C1QC.34 35 However, studies to identify their developmental origin or assigning functional properties to these subsets in humans are limited. To test whether monocytes or macrophages are differentially influenced by a co-culture with primary CAFs, we isolated circulating monocytes from peripheral blood, as an ex vivo surrogate for tumor-naïve monocytic cells infiltrating the TME. In parallel, we examined the effects of CAF co-culture on mature macrophages, as a proxy for tissue-resident macrophages entering a tumor. Monocultures of CAFs, monocytes and macrophages, as well as co-cultures of either monocytes or macrophages with CAFs for 24 hours were subjected to scRNA-seq (figure 1A). Using UMAP analysis of cells pooled from all conditions we identified two primary clusters: FCER1G+ monocytic myeloid cells (including monocytes and macrophages) and COL1A2+ CAFs (figure 1B), along with minuscule clusters of KLRC2+ natural killer cells (cells) and CCL19+ dendritic cells (online supplemental figure S1A, online supplemental table 5). The presence of these small clusters is a likely consequence of impurities from the negative selection of CD14+ cells from the human blood sample. Our data revealed that CAFs significantly influenced TAM marker expression in monocytes but not in macrophages (figure 1C, online supplemental figure S1B). On co-culture, monocytes upregulated the expression of differentiation and activation markers, including CD14, FCN1, CD86, as well as markers associated with an immunosuppressive phenotype like MRC1 (which encodes CD206) and CD163. In our data set, only differentiated macrophages expressed SPP1 while C1QC was expressed by a small fraction of macrophages, regardless of CAF presence (figure 1C). To gain a deeper understanding of the phenotype induced by CAFs in activated monocytes, we selected the top 200 upregulated genes and conducted GSEA. The analysis revealed a highly significant enrichment for genes contributing to an inflammatory response in monocytes, including signals associated with TNF-α, IL-2/STAT5, KRAS, IL-6/JAK/STAT3, and inflammatory responses (figure 1D). In contrast, enrichments for genes contributing to lipid and lipoprotein metabolism were highly significant among the top 200 downregulated genes in monocytes (online supplemental figure S1C). Focusing on the secreted proteins of CAF-instructed monocytes, we detected a strong expression of chemokines, including CXCL1, PPBP (CXCL7), CXCL5, CXCL3, CXCL2, CCL2, and CCL7 (figure 1E). Likewise, we observed an upregulation of genes in CAFs that gave rise to similar pathway enrichments as in monocytes (TNF-α, IL-6/JAK/STAT3) and the induction of the same chemokines (CXCL6, CXCL1, CXCL3, CXCL2, CCL2), suggestive of feedback loops to create chemo-attractive gradients towards myeloid cells (online supplemental figure S1D,E). The analysis of our data set in respect to inflammatory cancer-associated fibroblast (iCAF) and myofibroblastic cancer-associated fibroblast (myCAF) signatures36 revealed that CAFs exhibit plasticity, shifting from an myCAF to an iCAF phenotype in the presence of monocytes (figure 1F). This plasticity, influenced by monocyte-CAF interactions, may contribute to the induction of the inflammatory CAF phenotype in the tumor periphery, a phenomenon observed across various cancer types.37 More detailed analysis of the CAF subset in our scRNA-seq data revealed a high degree of CAF heterogeneity (online supplemental figure S2A–D), with monocyte interaction significantly altering the proportions of specific CAF clusters (online supplemental figure S2C,D), where clusters 2, 3 and 9 were proportionally more present in co-cultured monocytes and characterized by genes contributing to oxidative phosphorylation, epithelial to mesenchymal transition (EMT), TNF signals and other key protumorigenic signaling pathways. In contrast, clusters 4, 8 and 12 were mainly present in CAFs cultures without monocytes and were characterized by increased expression of genes associated with cell proliferation (online supplemental figure S2D–F). This suggests that monocyte co-culture leads to proliferation arrest and a switch to more protumorigenic and inflammatory transcriptional programs in the CAF population.

Figure 1

Single-cell transcriptomics of CAF co-cultures with monocytic myeloid cells reveal CAF-induced TAM-like phenotypes. (A) Schematic representation of the experimental set-up. Cultures were incubated in the basal medium. (B) Visualization of the data set (n=29,449 single cells) using Uniform Manifold Approximation and Projection based on the sample type and cell type features. Overlaid expression of the markers FCER1G and COL1A2 (inside the box). (C) Violin plots of monocytic differentiation and TAM markers expression. Data is represented based on the Z-score of expression, sorted by the sample type, each dot represents one cell. (D) Gene Set Enrichment Analysis using Enrichr (based on MSigDB 2020 Database) of the top 200 upregulated genes in monocytes co-cultured with CAFs for 24 hours. (E) Dot plots showing average expression of chosen pro-inflammatory chemokines upregulated in monocytes co-cultured with CAFs for 24 hours. (F) Dot plots showing average expression of iCAF and myCAF genes expressed in CAFs cultured alone and co-cultured with monocytes for 24 hours. CAF, cancer-associated fibroblast; DC, dendritic cell; iCAF, inflammatory cancer associated fibroblast; myCAF, myofibroblastic cancer associated fibroblast; NK, natural killer; TAM, tumor-associated macrophage.

In summary, our findings indicate that primary CAF interactions with monocytes, but not differentiated macrophages, induce wide-spread phenotypic changes in both cell types, including the upregulation of markers associated with immature macrophages and numerous pro-inflammatory signals and cytokines known to be associated with TAMs in vivo.

CAFs critically drive a functional TAM-like phenotype in CRC organoid triple cultures

Our co-culture of CAFs with monocytes revealed a hitherto unknown influence of human CAFs on the phenotypic appearance of tumor naïve monocytes. However, to further extend the in vitro co-culture to more closely reflect the cellular complexity and relationships within the tumor ecosystem, we generated co-cultures of primary CRC organoids with CAFs and monocyte-derived TAM-like cells. As signals from CAFs contributed to the generation of TAM-like phenotypes, we first assessed the significance of CAF presence in 3D PDO-monocyte organotypic co-cultures. Primary cells were cultured in basal medium submerged ECM for 3 days, comprising CRC PDOs, monocytes and CAFs in various combinations (figure 2A). In the absence of CAFs, monocytes co-cultured with PDOs exhibited round morphology (figure 2B). The addition of CAFs induced a visible change in the morphology of the monocytes (figure 2B). To further characterize epithelial and mesenchymal compartments of the co-cultures, we used histology and IF staining and observed a high density of CAFs within the ECM scaffold resulting in the clustering of CAFs into a well-defined mesenchymal compartment, resembling characteristics seen in the patient tumors (figure 2C,D). Monocytic myeloid cells were found in the mesenchymal compartment, as well as in contact with the outer layer of epithelial tumor cells (figure 2C).

Figure 2

Monocyte differentiation to TAM-like monocytic myeloid cells in three-dimensional co-cultures with CRC organoids depends on the physical presence of CAFs. (A) Schematic representation of the experimental set-up. (B) Immunofluorescence (IF) staining of FFPE sections of a CAF, monocyte and CRC organoid triple culture 72 hours after seeding and culture in basal medium, as well as imaging unstained cultures. Upper images depict IF (DAPI, blue; CD45, red; PanCK, green). Lower images depict the brightfield of live cultures. Monocytes in focus, depicting their morphology in the respective culture conditions. Culture conditions and scale bar are indicated below. (C) Immunofluorescence staining of FFPE sections of a co-culture set-up of monocytes, PDOs and CAFs cultured for 72 hours in basal medium (DAPI, blue; vimentin, yellow; CD45, red; PanCK, green). (D) H&E staining of FFPE sections of a co-culture set-up of monocytes, PDOs and CAFs cultured for 72 hours in basal medium as well as a representative CRC tumor. (E) Flow cytometry quantification of CD206, CD163, CD86, HLA-DR, PD-L1, SIRPα and 4-1BBL median fluorescence intensity (MFI) on macrophages on indicated co-cultures in various media for 3 days. ns p>0.05; *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. CAF, cancer-associated fibroblast; CRC, colorectal cancer; DAPI, 4',6-diamidino-2-phenylindole, dihydrochloride; ECM, extracellular matrix; FFPE, formaldehyde-fixed paraphin embedded; PDO, patient-derived organoid; TAM, tumor-associated macrophage.

Next, we analyzed phenotypic markers of myeloid cells in organotypic co-cultures of PDOs and monocytes in the presence or absence of CAFs via flow cytometry. To study the influence of culture media on the TAM-like phenotype, cultures were grown (1) in the “gold-standard” ENAS medium, which is routinely used in PDO-immune co-cultures but due to complex additives (online supplemental table 6) may interfere with immune cell interactions, (2) in advanced DMEM/F12, described as basal medium, which should have little modulating effects on myeloid cells, or (3) in CAF-conditioned basal medium. Compared with the culture of monocytes alone, the addition of CAFs to the culture-induced upregulation of TAM phenotype markers CD206, CD163 and PD-L1 as well as co-stimulating molecules CD86 and 4-1BBL (figure 2E). SIRPα was downregulated by CAF presence and HLA-DR expression was found unchanged (figure 2E). Effects on all markers remained unchanged after the addition of PDOs, completing the triple culture together with CAFs and monocytes. The addition of PDOs to monocyte cultures alone enhanced PD-L1 expression, but had no significant effect on other markers (figure 2E). Importantly, culture of PDOs, CAFs and monocytes in ENAS medium had an overall suppressive effect on the upregulation of most phenotypic markers as compared with basal medium (figure 2E). Of note, only the expression of 4-1BBL was upregulated by ENAS (figure 2E). Additionally, basic markers of monocytic myeloid lineage and function, CD14 and CD64, were downregulated in triple cultures incubated in ENAS medium (online supplemental figure S3). Additional controls revealed, that in the absence of CAFs ENAS medium induced upregulation of CD86, HLA-DR and PD-L1, downregulation of 4-1BBL and no changes in CD206, CD163 and SIRPα as compared with monocyte culture in conditioned medium (online supplemental figure S4).

Interestingly, the addition of a CAF-conditioned medium to a co-culture of PDOs and monocytes was not able to stimulate marker gene expression (with the exception of HLA-DR and SIRPα, which were found downregulated). However, the stimulatory effect was less pronounced as compared to co-culture with CAFs. This is suggestive of a requirement for direct cell-to-cell contacts or dynamic cytokine communication with CAFs to induce this quantitative difference in marker gene expression on myeloid cells (figure 2E). In accordance with our scRNA-seq data, the addition of CAFs lead to an upregulation of CD14 as well as CD64 expression (online supplemental figure S3), demonstrating the role of CAFs in the maintenance of the highly plastic monocytic cell identity.

Taken together our results demonstrate that primary CAF-PDO-monocyte co-cultures recapitulate many parameters characteristic of TAMs. Our human, primary triple-culture identified CAFs to orchestrate instructive signals leading to incomplete macrophage differentiation. Moreover, CAFs induce morphological changes while upregulating monocyte markers. Thus, direct and indirect communication with CAFs maintains their plastic nature and confers their susceptibility to receiving signals that implement a TAM-like immunosuppressive program.

Single-cell transcriptome sequencing reveals specific therapy-response phenotypes in the triple culture model

We next tested our set-up in response to oxaliplatin in combination with 5-FU, which are used in clinical practice as first-line chemotherapy for CRC. Chemotherapeutic dosage was chosen according to maximal plasma concentrations determined in pharmacokinetic studies of patient treatments.38 39 Due to the known strong immunogenic effects of viruses on monocytic myeloid cells, we used an O-IAV as an additional treatment for reference. First, we tested the capacity of this oncolytic virus to infect tumor organoids. Indeed, O-IAV was able to infect tumor organoids by passive diffusion into the ECM gel (online supplemental figure S5). We further tested its capacity to infect other cell types in a 3-day triple culture. We could observe infection of monocytic cells and PDOs, but not of CAFs within 48 hours post treatment (online supplemental figure S6). To examine the effects of chemotherapy and O-IAV on the phenotypes of the cells in our triple culture model, we performed single-cell transcriptomic analysis. The cultures were treated for 24 hours and then subjected to single-cell sequencing (figure 3A). The UMAP projection of 9,528 sequenced, good quality cells revealed that the clustering was driven by cell types rather than the treatment regimen (figure 3B). However, when we re-clustered only TAM-like cells they displayed a distinct phenotype according to treatment (figure 3C). Moreover, our data indicates that different treatments induced specific polarization patterns in TAM-like cells. Unlike the double cultures of CAFs and monocytes (figure 1C), the triple culture induced polarization of myeloid cells towards C1QC+ TAM like cells, indicating a possible role of the colorectal cell compartment in shaping this macrophage phenotype (figure 3D). While SPP1+ macrophages were largely absent in the triple culture setting (online supplemental figure S7A), we observed high expression of MRC1 (CD206), CD86, and CD163 markers in mock-treated cells, which decreased with treatments, particularly after treatment with chemotherapy (figure 3D). Following treatment with O-IAV, we observed a strong upregulation of canonical type I and II interferon (IFN) signaling, leading to the upregulation of genes including ISG15, MX1, IFIT1, and OAS1 in all three cell types (online supplemental figure S7B). When comparing the strong inflammatory phenotype induced by oncolytic virotherapy with the phenotype induced by chemotherapy in TAM-like cells, we identified common signals related to “Epithelial Mesenchymal Transition”, “Inflammatory Response”, and “TNF-α Signaling via NF-κB” with high significance. This finding suggests that chemotherapy-induced phenotypes in TAM-like cells have an immunomodulatory aspect in our triple-culture setting (figure 3E). Cluster analysis uncovered high plasticity and strong polarization of the TAM-like cells, manifested by a reduction in myeloid cell heterogeneity under both treatments. Oncolytic virotherapy leads to a proportional increase of clusters 4 and 7, resembling inflammatory clusters with a prominent IFN and TNF-α signature (online supplemental figure S8A–D). In contrast, chemotherapy treatment of TAM-like cells resulted in a proportional increase of clusters 0 and 1, characterized by the expression of specific genes downstream of TNF-α signaling and genes involved in the epithelial-to-mesenchymal transition and cell cycle (online supplemental figure S8). Notably, the quality of the response to O-IAV and chemotherapy was more similar in CAFs as compared with TAM-like cells, with a strong induction of TNF signaling (figure 3E, online supplemental figures S7C and S9), putting further emphasis on the specificity and importance of the CAF compartment in the therapy response.

Figure 3

Single-cell transcriptomics of triple cultures enables the definition of distinct myeloid cell states and associated RNA expression profiles induced in response to treatment with chemotherapy or O-IAV. (A) Experimental set-up. Co-cultures were incubated in the basal medium. (B) Visualization of the data set (n=9,528 single cells) using UMAP projection based on the sample type and cell type features. (C) UMAP projection of re-clustered TAM. (D) Violin plots of monocytic differentiation and TAM marker expression. Data is represented based on the Z-score of expression, sorted by the sample type, each dot represents one cell. (E) Venn diagrams (visualized using Venny V.2.0) and Gene Set Enrichment Analysis (GSEA) using Enrichr (based on MsigDB 2020 database) of the common therapy response signatures based on a comparison of the top 100 (TAM) and top 100 (CAFs) upregulated genes. (F) Dot plots showing average expression of ligands and receptors upregulated by O-IAV and chemotherapy represented in all three cell types. CAF, cancer-associated fibroblast; CRC, colorectal cancer; O-IAV, oncolytic influenza A virus; PDO, patient-derived organoid; scRNA-seq, single-cell RNA sequencing; TAM, tumor-associated macrophage; UMAP, Uniform Manifold Approximation and Projection; 5-FU, 5-fluorouracil.

As the induction of the immunosuppressive program in monocytes by CAFs required a combination of direct cell-to-cell contact and secreted factors, we next focused on the regulation of genes encoding secreted factors. The inclusion of three cell types in the culture system allowed us to recapitulate essential cell-to-cell communication axes within the CRC TME. We observed the induction of inflammatory signals in the CRC PDOs, including upregulation of IL-6, CXCL6, IL-1B, and TNFAIP6 expression, in response to both treatments (online supplemental figure S7D). Additionally, both treatments induced an upregulation of IL-6 by CAFs and TAM-like cells, along with the expression of the IL-6 receptors, IL-6R and IL-6ST (Gp130), in a subset of cells in each cell type (figure 3F). Another strongly upregulated signal by both treatments was IL-1 signaling. IL-1A and IL-1B were found to be expressed by CAFs and TAM-like cells, with CAFs being recipients of the signal, as reflected by their expression of IL-1R1 (figure 3F). Interestingly, O-IAV treatment additionally induced strong expression of IL-1RN in TAM-like cells, encoding the IL-1 receptor antagonist protein, suggestive of an IL-1-inhibitory feedback loop to fine-tune the response (figure 3F). Additionally, we identified another communication axis in the triple culture: the FGF2-FGFR2 interaction, with CAFs expressing both FGF2 and FGFR2, being both senders and receivers of this autocrine stimulation, previously implicated in CRC therapy resistance.40 41 Finally, we observed several treatment-specific effects on the level of cell–cell communication, including clinically relevant signals coming from IL-33-IL-1RAP, CCL8-CCR1, and IL-4l1-AHR interactions induced by the oncolytic virus treatment, and LIF-LIFR, MET-HGF, IL-18-IL-18R1 triggered by chemotherapy (figure 3F).42–47

In summary, our data demonstrate that the complex primary CRC organoid triple cultures model multiple molecular aspects of treatment responses. We provide evidence of an immunogenic signal generated by chemotherapy, which, at least in part, counteracted the CAF-induced, immunosuppressive program in TAM-like cells. This is in line with the observed onset of inflammatory signaling circuits.

Chemotherapy treatment counteracts the CAF-induced immunosuppressive program in TAM-like cells

Our single-cell sequencing data revealed treatment-dependent effects on TAM-like cells and CAFs in our triple culture model. We next sought to validate the decreased expression of the immunosuppressive markers CD206 (MRC1), CD163 and CD86 in TAM-like cells on the protein level. We treated cultures from four patients with distinct CRC (figure 4A) with the indicated therapeutics for 24 hours and assessed the expression of immunosuppressive markers in TAM-like cells using flow cytometry (figure 4B). In line with our observations on mRNA level, both chemotherapy and virotherapy downregulated the immunosuppressive markers CD206 and CD163 (figure 4C). However, we observed a trend of the O-IAV to upregulate the expression of the T-cell-activating ligand CD86, whereas chemotherapy suppressed CD86 expression (figure 4C).

Figure 4

Chemotherapeutics and oncolytic influenza A viruses induce repolarization of tumor-associated macrophage-like cells in primary triple culture assays. (A) Table of patient characteristics. (B) Experimental set-up. Cultures incubated in basal medium. (C) Quantification of CD206, CD163 and CD86 median fluorescence intensity (MFI) normalized to the corresponding patient’s MFI from the untreated condition, measured on macrophages from treated triple cultures. (D) Quantification of cytokine concentrations (ng/mL) in the supernatants collected from C, determined via LEGENDplex assay. Upper detection limits are indicated by a dashed line. ns p>0.05; *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. All color-coded scatter-plot circles represent measurements from individual donors. CAF, cancer-associated fibroblast; IL, interleukin; O-IAV, oncolytic influenza A virus; PDO, patient-derived organoid; 5-FU, 5-fluorouracil; TNF, tumour necrosis factor.

In order to validate that the observed effects remain relevant in myeloid cells from the peripheral blood of patients with CRC with a current tumor burden, the above assays were performed with monocytes extracted from the peripheral blood of patients with CRC (online supplemental figure S10). We were able to observe similar deregulation of the TAM markers CD206, CD163 and CD86 as observed with healthy donor monocytes (online supplemental figure S10). Furthermore, we observed higher susceptibility to treatment-induced cell death in TAM-like cells derived from patients with CRC compared with healthy donors (online supplemental figure S11).

Our single-cell transcriptomics data showed a treatment-dependent upregulation of target genes downstream of the IL-6 and IL-1β signaling cascade in both CAFs and TAM-like cells (figure 3). Besides these pro-inflammatory signals, we detected gene activity downstream of the TNF pathway as well as upregulation of STAT3-dependent genes, which could be attributed to signaling downstream of IL-6 and/or IL-10. Following this notion, we probed the supernatants collected from the treated co-cultures to test levels of these cytokines and additional clinically relevant cytokines.48 Both treatments induced the secretion of pro-inflammatory cytokines IL-1β, IL-6 and TNF in patients with primary tumor 1 and 2 (figure 4D). IL-6 and TNF were also induced by both treatments in patient 3, while only the O-IAV induced TNF secretion in patient 4 (figure 4D). The anti-inflammatory cytokine IL-10 was only significantly induced by oncolytic virotherapy in patient 3 (figure 4D).

Taken together, we observed significant downregulation of immunoinhibitory markers by both treatments. Moreover, we observed a patient-dependent increase in the secretion of inflammatory cytokines, demonstrating that our triple culture model faithfully recapitulates the complexity of cell intrinsic and extrinsic immunomodulatory regulators in response to treatments