Ricin

Pooled, genome-wide CRISPR deletion screens were performed in three cell lines: K562 stably expressing SFFV-Cas9-BFP, Ramos cells lentivirally infected with SFFV-Cas9-BFP, and U937 cells lentivirally infected with EF1a-Cas9-Blast34 (link). The library was synthesized, cloned and lentivirally infected into cells as previously described20 (link). Briefly, the parent vector for the libraries was derived from a pSico lentiviral vector which expresses GFP and a puromycin-resistance cassette separated by a T2A sequence45 (link)58 (link); we replaced GFP with mCherry to make the final parent vector, pMCB320. Sublibraries were PCR-amplified from pooled-oligo chips (CustomArray, Agilent), digested with BstXI and BlpI restriction enzymes, and ligated into BstXI/BlpI-cut pMCB320 using T4 ligase. Libraries and vectors will be made available via Addgene. Three days after infection, cells were placed under puromycin selection (0.7 μg ml⁻¹, Sigma) for an additional 3 days after infection, then split at time 0. Throughout the screen, the pooled libraries were maintained at 1,000 cells per guide or a total of ∼250 million cells in large spinner flasks. K562 and U937 were grown for ∼2 weeks, and Ramos cells were growth for ∼3 weeks due to their slower division time. Genomic DNA was extracted following Qiagen's Blood Maxi Kit, and the guide composition was sequenced and compared to the plasmid library using casTLE20 (link) version 1.0 available at https://bitbucket.org/dmorgens/castle. Briefly, casTLE compares each set of gene-targeting guides to the negative controls, selecting the most likely maximum effect size which explains the distribution of targeting guides. It then determines the significance of this maximum effect by permuting the results20 (link). Both safe-targeting and non-targeting controls were used for this analysis. For the ricin sensitivity screen, cells were treated with ricin toxin (Vector Labs) at 0.25 ng ml⁻¹ for 24 h, ricin was removed and then cells were allowed to recover to normal doubling rate. This treatment occurred four times over 2 weeks.

The CRISPRi activity score dataset was obtained from Horlbeck et al. (2016) (link). CRISPRi and CRISPRa ricin tiling data was obtained from Gilbert et al. (2014) (link). CRISPRa activity scores were generated as previously described for the CRISPRi activity dataset, using data from 9 published and unpublished screening datasets. Hit genes were selected using the formula |effect size Z-score x log₁₀ p-value| ≥ 20 in any screen, and the phenotypes for sgRNAs targeting each gene were extracted from the screen in which the gene was a hit and normalized to the mean of the top 3 sgRNAs by absolute value. All datasets are included here as Supplementary file 1.

The gene expression profiles of pro‐inflammatory cytokines/chemokines in bronchoalveolar lavage fluids (BALFs) at 0, 4, 8, 12, 24, 48, and 72 h post‐ricin challenge were examined with multiplex cytokine quantification assays. A total of 12 major cytokines/chemokines were highly upregulated in BALFs from 4 to 72 h post‐ricin challenge (Figure 2A), suggesting the occurrence of an inflammatory cytokine storm during DAD progression. Multiple inflammatory features were also observed in parallel histological examinations (Figure S3A,B). In detail, the abundance of IL‐6 and CXCL1 was significantly upregulated at 8 h and sustained at high levels from 8 through 48 h. GM‐CSF and CXCL2 were robustly induced at 12 h and gradually decreased at 24 and 48 h, although they remained much higher than at 0 h. The two classic proinflammatory mediators Il‐1β and TNF‐α showed a moderate increase from 4 to 48 h, and a significant increase at 72 h. Overall, ricin‐induced DAD expression patterns were characteristic of a cytokine storm, with an excess release of multiple pro‐inflammatory cytokines/chemokines, and asynchronous changes in released cytokines/chemokines denoting diverse inflammatory responses in different lung cell populations during DAD progression.
To further investigate the production sources of these cytokines/chemokines, we analyzed their gene expression profiles using UMAP plots (Figure 2B) and discrete cell populations (Figure S4A). Fibroblasts and neutrophils were the main cell populations that had higher expressed levels of these cytokines/chemokines. After ricin exposure, fibroblasts served as the major source of CXCL1, which acted as the chemoattract for mobile neutrophils (Figure 2C). We next assessed the recruitment and trafficking mechanisms of immune cells by analyzing the expression of chemokines and their corresponding receptors based on scRNA‐seq data (Figure S4B). Several potential mediators, including Cxcl1 (to Cxcr1 and Cxcr2), Cxcl9 (to Cxcr3), and Cxcl10 (to Cxcr3) exhibited high‐levels of gene expression in fibroblasts and might be involved in neutrophil recruitment.
In summary, fibroblasts and neutrophils represent the major source of multiple pro‐inflammatory cytokines/chemokines and play an important role in promoting inflammation and a cytokine storm during DAD progression.

To investigate lung cellular responses during DAD progression, mice were challenged with aerosolized ricin to induce DAD syndrome, as described in our previous study.^[4] We then performed 10× genomic scRNA‐seq on the dissociated pulmonary cells at 0 (homeostatic control), 8, 24, and 48 h post‐ricin challenge, followed by integration analysis using Seurat (Figure 1A). As a result, 36,810 cells were obtained and a uniform manifold approximation and projection (UMAP) of the cell landscape was created (Figure 1B), indicating an overrepresentation of lung cells. Based on highly expressed classical genotypic markers, these lung cells were classified into nine major populations (Figure 1C): neutrophils (the predominant cell population); monocyte‐lineage cells (monocyte/macrophage/dendritic cell; Mono/Mac/DC); natural killer cells (NKC); T cells; B cells; vascular endothelial cells (VEC); fibroblasts (Fib); cluster_1 epithelial cells (Epi_1); and cluster_2 epithelial cells (Epi_2).
The changing proportions of these cell subpopulations over time were assessed (Figure 1D). A global reorganization of lung immune cell populations began 8 h post‐ricin exposure, including an increase of neutrophil and monocyte‐lineage cells, and a decrease of lymphocyte‐lineage cells, including T cells and B cells (Figure 1D). This was confirmed by fluorescence‐activated cell sorting (FACS) (Figure S1) and immunofluorescence imaging (Figure S2A,B). A gradual and sequential reorganization of neutrophils, VEC, and Epi_2 was also observed, indicating a continual change in these cells during DAD progression (Figure 1D).
The obtained cells yielded a total of 2132 differentially expressed genes (DEGs), which were classified into seven major clusters based on their gene expression patterns at 8, 24, and 48 h (Figure 1E). Gene ontology (GO) enrichment analysis indicated these seven clusters had distinct biological processes (Figure 1F). The DEGs in cluster_1 were distributed in epithelial cells (Epi_1 and Epi_2), and mainly enriched in the biological processes of wound healing, epithelial cell migration, and cell shape regulation. The DEGs in cluster_2, mainly belonging to VEC, were involved in cell adhesion and binding. The 260 fibroblast‐related DEGs in cluster_6 were significantly enriched in chemotaxis and inflammatory response. These findings demonstrate the temporal change in lung structural cells during DAD progression.

The spatial histopathological patterns on the STomics slice (including lung parenchyma regions in lower, middle, and upper lobe areas) at 24 h (Figure 6) were analyzed. The slice could be divided into three distinct regions based on the histopathological structures (Figure 6A): the lower lobe represented the terminal bronchial region (Area_A), the middle lobe represented the bronchiole region (area_B), and the upper lobe represented the principal bronchial region (Area_C). A collagen deposition region was distributed mostly in area_A, indicating that area_A was the main area of ricin‐induced DAD and underwent the wound healing progression. In contrast, no obvious pathological changes were observed in area_B or area_C (Figure 6A). These histopathological findings agree with the unbiased spot clustering results in Figure 5B.
The spatial enrichment profile of immune cells and activated lung structural cells in the STomics slice at 24 h (Figure 6B) was then assessed. AEpi was detected mostly in area_B, while AFib was exclusively enriched in area_C. AEndo was distributed in area_A and area_C, while neutrophil was enriched in area_C. A functional enrichment analysis revealed different biological progressions among the three areas (Figure 6C,D). Area_A was responsible for physiological reactions (such as hypoxia‐response for adaption of low oxygen content) and area_C was characterized by activated immune response and immune cell infiltration, while area_B reflected an intermediate state between area_A and area_C. Furthermore, neutrophil recruitment happened in the upper lobe areas close to the bronchi regions. Collectively, scRNA‐seq identified the major activated structural cell population AFib (Figure 3A) and STomics indicated its location in the upper lobe close to the bronchi regions (Figure 6B).
A spatial trajectory using both spatial information and gene expression profiles was then constructed. A natural progression of cluster_0 cells toward cluster_2 cells in area_C (Figure 6E,F) was observed. The root cells of the trajectory showed significant upregulation of collagen deposition markers (such as Saa3, Fth1, Mgp, and Col3a1), while the end cells of the trajectory displayed significant upregulation of transcription factors (such as Egr1, Egr4, Atf4, and Junb), chemokines (such as Cxcl2, Cxcl9), and transcriptional co‐activators (such as Ifrd1) (Figure 6G). These results are consistent with the observations from scRNA‐seq data (Figure 4A–D), again suggesting that AFib served as the major trigger for neutrophil recruitment in the upper lobe.
Based on the above combined analysis of scRNA‐seq and STomics data, we speculated that AFib boosted a rapid and robust immune response via recruitment of neutrophils, and then the recruited neutrophils produced many more proinflammatory mediators to further induce airway inflammation outbreak in the upper lobe, an area we termed “inflamed niche.”