Acidovorax temperans exposure accelerates tumor development in an autochthonous LUAD mouse model

Previous research developed a mutant K-ras and Tp53-driven LUAD mouse model (KP) under the Cre-lox system, resulting in endogenous tumor development reflective of human LUAD [19, 20]. Our group identified Acidovorax temperans as enriched in lung cancer which led us to hypothesize that A. temperans may play a functional role in lung cancer development [13]. To determine if repeated bacterial exposure, as experienced in chronic tobacco smoking, would result in increased tumor growth, we administrated six biweekly intranasal instillations of PBS (sham) or A. temperans in KP mice following Ad-cre instillation (Fig. 1A).

Fig. 1: Acidovorax temperans accelerates tumor growth in a mouse model of lung adenocarcinoma.

A Experimental timeline including bacterial dosage schedule. B MRI images of sham (1X PBS) (top) and A. temperans (bottom) instilled mice at 13 weeks post Adcre. C MRI quantification of tumor volume in sham (n = 14) and A. temperans (n = 13) instilled mice. D Quantification of lung weight in sham (n = 8) and A. temperans (n = 7) instilled mice. E H&E stained images of sham (top) and A. temperans (bottom) instilled mice lungs at 13 weeks post Adcre. F H&E quantification of tumor area in sham (n = 5) and A. temperans (n = 5) instilled mice. Data presented as mean ± SEM, ***p < 0.001, ****p < 0.0001.

Using non-invasive magnetic resonance imaging (MRI), we measured tumor development at 9- and 13-weeks post instillation (p.i.) of Ad-cre and then sacrificed mice at 14 weeks. We found that at 9 weeks p.i., tumor nodules were only present in A. temperans mice (Supplementary Fig. S1A), therefore we focused on the tumor state at 13 weeks p.i. when tumors were visible in both groups. A. temperans instilled mice had visibly larger nodules compared to those instilled with sham by MRI and quantification demonstrated an increase in tumor volume (Fig. 1B, C). Consistent with these results, total lung weight was also increased in A. temperans mice (Fig. 1D). Tumor area as determined by H&E histology was significantly increased and more high-grade lesions were found in the A. temperans mice (Fig. 1E, F, Supplementary Fig. S1B). Taken together, these results revealed that repeated exposure to A. temperans could accelerate lung tumor development in the presence of oncogenic K-ras and Tp53 mutations.

We next asked if the accelerated tumor growth we observed could result from A. temperans persistence in lungs. We instilled sham or A. temperans into mice and homogenized lung tissue at Days 1, 5, and 9 post bacterial instillation for colony plating and enumeration, which revealed a large number of colonies in A. temperans mice on Day 1 only (Supplementary Fig. S1C). In contrast, bacterial colony number was comparable between sham mice at each time point and between A. temperans mice on Days 5 and 9. We identified a total of seven genera, dominated by Lactobacillus and Streptococcus spp., while Acidovorax colonies were only found in the A. temperans mice, and only on Day 1 post instillation (Supplementary Fig. S1D, E). These results indicate A. temperans is short-lived in the lungs and is unlikely to colonize this tissue.

Immune cell infiltration within the tumor microenvironment is altered by A. temperans

To identify possible mechanisms of accelerated tumor development in A. temperans instilled mice, we collected lung tissues from mice at 10 and 14 weeks p.i. and performed bulk RNA-sequencing (Fig. 1A). Pathway enrichment using both GSEA and IPA platforms indicated most pathways upregulated in A. temperans mice, regardless of timepoint, were related to immune function (Fig. 2A). We then used xCell to deconvolute the RNA-sequencing data and predict immune cell infiltration in these mice [21]. These results showed that sham and A. temperans instilled mice clustered separately, regardless of timepoint (Fig. 2B). Overall, we found proinflammatory cells such as macrophages, dendritic cells, neutrophils, and plasma cells highly enriched in A. temperans mice, with effector CD4+ T cells and myeloid cells higher at 10 weeks compared to 14 weeks (Fig. 2C). These results suggest that repeated A. temperans instillation alters the immune compartment of the tumor microenvironment, dramatically increasing the number of proinflammatory cells.

Fig. 2: Immune activation differentiates Sham and A. temperans instilled mice.

A GSEA (left) and IPA (right) pathway enrichment in A. temperans (n = 21) versus sham (n = 18) instilled mice across timepoints. B Normalized xCell predicted immune cell infiltration in both sham and A. temperans instilled mice. C Quantification of xCell predicted cell infiltration comparing sham week 10 (n = 4), sham week 14 (n = 14), A. temperans week 10 (n = 15), and A. temperans week 14 (n = 6) instilled mice. Data presented as median value plus quartiles for boxplots, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Previous studies have demonstrated an important role for a proinflammatory tumor microenvironment in KP LUAD development, with dendritic cells [22], macrophages [23], neutrophils [24,25,26,27], and T cells [11, 28] all implicated in the etiology of this animal model. Considering the overlap of these cell types with those involved in bacterial response, supported by their enrichment in A. temperans mice in our RNA-seq results, we hypothesized that bacterial exposure may accelerate tumor growth by altering the immune microenvironment.

To test this hypothesis, we dissociated lung tissue from four sham and four A. temperans instilled KP mice and isolated the CD45+ fraction by FACS (Fig. 1A). We then performed droplet-capture single cell RNA-sequencing (scRNA-seq), which returned 25,477 total CD45+ cells after filtering. These cells divided into 11 major cell types: monocytes, macrophages, monocyte-derived dendritic cells (MoDC), alveolar macrophages (AMs), conventional dendritic cells (cDC), plasmacytoid dendritic cells (pDC), neutrophils, B cells, plasma cells, NK cells, and T cells (Fig. 3A). Cell types largely overlapped between sham and A. temperans mice, although cell proportions showed largest relative shifts from monocyte-high in sham to increased macrophages, AMs, neutrophils, and T cells in A. temperans mice (Fig. 3A, B). We identified these clusters through three methods, by comparison to the ImmGen database (Fig. 3C) [29], canonical gene markers (Fig. 3D), and top differentially expressed genes (Fig. 3E, Supplementary Table S1). Collectively, these results demonstrate A. temperans alters the immune compartment of the tumor microenvironment in KP mice.

Fig. 3: A. temperans alters the immune microenvironment in lung adenocarcinoma.

A UMAP plots of scRNA-seq cell clusters (left), immune cell types (center), and treatment groups (right). B Barplot of the relative abundance for each cell type by individual mouse with sham (S, n = 4) or A. temperans (At, n = 4) instillation. C Heatmap of ImmGen-based cell type identification. Color scale indicates positive Spearman’s correlation coefficient. D Violin plots for marker genes associated with the different cell types. E Heatmap of the top five differentially expressed genes for each cell type.

Lung macrophages upregulate MHC class II in response to A. temperans

Myeloid cells are the first cells to respond to bacterial lung infections and often secrete proinflammatory cytokines linked to tumor development, leading us to first characterize this compartment. We identified 13 subclusters within the scRNA-seq dataset corresponding to monocyte, macrophages, and dendritic cells (MoMaDCs) (Fig. 4A, B). Overall, we identified two clusters of naïve monocytes (Cd14+, Fcgr3-), three activated monocyte clusters (Act Mono; Cd14+, Fcgr3+), one cycling monocyte cluster (Mki67+, Stmn1+, Top2a+), three macrophage clusters (Cd68+), and four DC clusters (Syngr2+) (Fig. 4C, Supplemental Table S3). Within the macrophages, we identified two clusters of tumor-associated macrophages (TAMs; Fcgr2b+, Ccl4+, Trem2+) [30] and one enriched in complement genes (C4b+, Cfp+, C1qb+). Within the DCs, we identified MoDCs (Ccl5+, Ccr7+, Fscn1+), conventional DCs clusters cDC1 (Clec9a+, Itgae+, Xcr1+) [31] and cDC2 (Mgl2+, Irf4+), and plasmacytoid DCs (Bst2+, Pacsin1+, Siglech+) (Fig. 3C, Supplementary Table S2) [32]. Cell type identification was confirmed by comparison to the ImmGen database (Fig. 4D). We then asked if developmental trajectory followed the conventional route from circulating monocytes to macrophages and DCs. First removing the cDC and pDC clusters as these cell types are not monocyte derived, this analysis revealed that the TAMs and MoDCs were the latest in pseudotime (Fig. 4E, F).

Fig. 4: TAMs are expanded and upregulate MHC II in response to A. temperans.

A UMAP plots of monocytes, macrophages, and dendritic cells (MoMaDCs) cell types (left) and treatment groups (right). B Barplot of the relative abundance for each cell subtype by individual mouse with sham (S, n = 4) or A. temperans (At, n = 4) instillation. C Dotplot of marker genes for each cell type. D Heatmap of ImmGen-based cell type identification. Color scale indicates positive Spearman’s correlation coefficient. E, F UMAP plots of (E) trajectory analysis and (F) pseudotime projection of monocyte-derived cells. G Single-sample GSEA (ssGSEA) heatmap of average normalized enrichment scores for both TAM clusters divided by treatment group. H Comparison of average expression of each MHC II component gene (H2-Aa, -Ab1, -DMa, -DMb1, -DMb2, -Eb1, -Eb2, -Oa, -Ob) by treatment for each cell type. Data presented as median value plus quartiles for boxplots, n.s. not significant, *p < 0.05, ****p < 0.0001.

Macrophages have traditionally been classified as M1 or M2, with classically activated M1 macrophages inducing inflammation against pathogens and tumor cells while M2 macrophages are immunosuppressive. Neither TAM cluster showed expression of M1 or M2 macrophage markers (Supplementary Fig. S2A). To better understand differences in their function, we examined changes in TAM gene expression with single-sample gene set enrichment analysis (ssGSEA) [33]. These clusters were differentiated by cholesterol esterification upregulation in TAM-1 in both treatment groups while TNF signaling was specifically upregulated in TAM-1 in response to A. temperans (Fig. 4G). Both TAM-1 and TAM-2 clusters were highly enriched for both MHC class I and II antigen presentation in response to A. temperans instillation. To determine if MHC upregulation was consistent among monocytes and DCs, we calculated signature scores for all MHC class I and II genes. This revealed A. temperans caused upregulation of MHC I in macrophages and DCs, but only macrophages displayed consistent upregulation of MHC II, with TAM-2 cells having the greatest relative increase (Fig. 4H, Supplementary Fig. S2B, C). Similarly, analysis of the alveolar macrophage (AM) compartment revealed low expression of M1/M2 genes while MHC I and II were both broadly upregulated across AMs in response to A. temperans (Supplementary Fig. S3, Supplementary Table S3). Together, these results indicate that bacterial exposure induces a broad MHC II response across lung macrophages, potentially contributing to increased tumor growth through CD4+ T cell activation.

A. temperans increases tumor-associated neutrophils which express antimicrobial gene programs

Neutrophils are the most abundant immune cell type in human NSCLC and KP mice [26, 34] and have dual function in cancer development and infection response, particularly secretion of proinflammatory cytokines, ROS production, and immunosuppression [35]. These factors implicate neutrophils as essential mediators of early tumor development; therefore, we asked how A. temperans altered expression and function of neutrophils in KP mice.

We identified four clusters of neutrophils from the scRNA-seq data, with sham mice having greater proportions of clusters C1 and C2 while A. temperans instilled mice greatly increased the cell numbers in clusters C0 and C3 (Fig. 5A, B). We then confirmed an increase in total neutrophils in A. temperans instilled mice by immunohistochemistry staining of Ly-6G (Fig. 5C). Markers associated with circulating neutrophils (Sellhi, CD62L; Cxcr4lo) were higher in clusters C1 and C2 while markers linked to increased effector function (Icam1), immunosuppression (Cd274+, PD-L1), and tumor-promotion (Siglecf) were higher in clusters C0 and C3; these expression patterns generally corresponded with treatment group (Fig. 5D, Supplementary Table S4) [36, 37]. To verify this change in marker gene expression from control to bacterial exposure, we performed trajectory analysis on the scRNA-seq clusters. This revealed C1 was earliest and C3 latest in pseudotime (Fig. 5E), consistent with expression of these marker genes.

Fig. 5: Tumor-associated neutrophils are associated with increased anti-bacterial function.

A UMAP plots of neutrophil clusters (left) and treatment groups (right). B Barplot of the relative abundance for each cluster by individual mouse with sham (S, n = 4) or A. temperans (At, n = 4) instillation. C Representative immunohistochemistry images of neutrophil populations (Ly-6G) in sham (left) and A. temperans (right) instilled mice. Scale bar 100 µm. D Density plots of marker gene expression. E UMAP plots of trajectory analysis (top) and pseudotime (bottom) of neutrophil clusters. F ssGSEA heatmap of average normalized enrichment scores for each neutrophil cluster. G Gene signature scores for tumor associated neutrophils (TANs). TAN signature from accession number GSE118245 [40]. Boxplots indicate median and quartile scores. ****p < 0.0001.

To further understand transcriptional differences between sham and A. temperans-associated neutrophils, we performed ssGSEA, which suggested C2 represented the most immature cell state, with high enrichment scores for cell proliferation in bone marrow, immature neutrophils, and cholesterol catabolism, important for neutrophil development and release from the bone marrow (Fig. 5F) [38]. C1 pathways were enriched for neutrophil extravasation while activated, effector functions such migration, chemotaxis, and bacterial response and killing, were primarily associated with C0 and C3, underlined by an LPS response which activated cytokine production.

We then asked if these changes reflected gene expression profiles of immunosuppressive neutrophils in cancer. Tumor-associated neutrophils (TANs) are required for tumor progression and metastasis [39], and comparison of a TAN gene signature revealed C3 had the highest expression of this signature (Fig. 5G) [40]. Collectively, these results demonstrate that dysbiosis response is a key programming event for tumor-associated neutrophils, suggesting that these neutrophils, while being responsible for clearing bacteria from the lungs, alter the tumor microenvironment.

A. temperans robustly induces TH17 polarization

Having demonstrated large proinflammatory changes in the myeloid compartment driven by MHC upregulation in MoMaDCs, we then examined if these changes were reflected in T cells, as well. We identified a total of 12T cell types and sham mice had higher proportions of naïve T cells while A. temperans mice showed greater CD4+ effector populations (Fig. 6A, B, Supplementary Table S5), which we confirmed by immunofluorescence (Fig. 6C). These effector populations included follicular helper T cells (Tfh; Cd200+, Izumo1r+, Slamf6+) [41], Tregs (Foxp3+, Ikzf2+, Ctla4+), Th17 (Il17a+, Tmem176a+, Tmem176b+) [42], and Th1 (Ccr2+, Ifng+). We found three clusters of cells which did not express either Cd4 or Cd8a (Fig. 6D), which corresponded to γδ T cells (Tcrg-C1+, Trdc+), double negative (DN) naïve (Ccr7+, Lef1+, Sell+, Tcf7+), and a DN Treg-like population (Areg+, Gata3+, Il1rl1+) [43].

Fig. 6: T cells are terminally polarized to a Th17 phenotype.

A UMAP plots of T cell subtypes (left) and treatment groups (right). B Barplot of the relative abundance for each cell subtype by individual mouse with sham (S, n = 4) or A. temperans (At, n = 4) instillation. C Representative immunofluorescence images of T cell populations in sham (top) and A. temperans (bottom) instilled mice. Blue DAPI, green CD3, red CD4, scale bar 50 µm. D Dotplot of marker genes for each cell type. E Expression level of tissue residency marker genes by cell type. F Expression level of tissue residency markers by cell type and treatment group. UMAP plots of trajectory analysis by (G) cell type and (H) treatment group. I UMAP plots of pseudotime projection of T cells. n.s. not significant, * p < 0.05, *** p < 0.001, **** p < 0.0001.

T cells from A. temperans mice also showed greater expression of tissue residency markers Cxcr6 and Itgae (CD103) in CD4+ and CD8+ T cells (Fig. 6E, F). As tissue residency is associated with effector function, we then asked if these effector CD4+ cells represented a terminal cell state. Trajectory analysis which revealed naïve CD8 and DN cells were earliest in pseudotime, while Th1 and Th17 cells were latest (Fig. 6G – I). The similarity of marker genes between the Th1 and Th17 cells, combined with high expression of Cxcr6 in Th1 and Th17 cells, suggested that effector CD4+ T cells acquire a tissue residency phenotype prior to polarization. In support of this hypothesis, A. temperans T cells were consistently later in pseudotime than sham (Fig. 6H), suggesting that bacterial exposure is a key factor for establishing CD4+ T cell lung residency and subsequent TH1/TH17 polarization.

A. temperans induces specific IL-17 and broad IFN-γ response in T cells

Previous data examining murine colonic effector T cells suggested that T cell phenotype was shaped by response to specific pathogens [44]. We asked if the T cell polarization induced by A. temperans was specific to this species or was consistent with more general microbial dysbiosis. We first performed bulk TCR-seq from lung tissues, which showed a sharp decrease in TCR diversity in A. temperans mice (Supplementary Fig. S4A, B), due to bacterial-driven hyperextension of specific clonotypes (Supplementary Fig. S4C–E). We then compared bacterially induced T cell gene signatures from mice infected with either Citrobacter rodentium (TH17 response) or Salmonella enterica serovar Typhimurium (IFN-γ response), both Gram-negative species [44]. Expression of the C. rodentium signature was predominantly found in our Th1, Th17, and γδ T clusters (Supplementary Fig. S5A, B). Although the Salmonella Typhimurium signature was also highest in Th1 and Th17 clusters, we observed consistently high expression throughout our dataset, but upregulated in A. temperans mice compared to sham overall (Supplementary Fig. S5C, D). These data suggest that the Th17 cell cluster we observe is not specific to A. temperans; however, the combination of general IFN-γ and specific TH17 polarization may represent a specific inflammatory response to this species.

Based on the widespread expression of the Salmonella Typhimurium gene signature, we asked if both Il17a and Ifng were upregulated in response to A. temperans. Our results showed Il17a and its transcription factor Rorc were largely restricted to Th17 and γδ T cells while Ifng and its transcription factor Stat4 were highly expressed in non-naïve T cells (Supplementary Fig. S6A, B). Overall, most T cell subtypes expressed Ifng, with nearly half of Th1 and a third of Th17 cells positive for this transcript and expression was elevated in response to A. temperans (Supplementary Fig. S6C–E). Within Th17 cells, a subpopulation was double positive for Il17a and Ifng (Supplementary Fig. S6F, G), a highly inflammatory cell state increased in smokers [45]. These data suggest that A. temperans alters the immune microenvironment through multiple signaling pathways which culminate in Il17a + /Ifng + T cells to greatly increase inflammation.

A conserved gene signature in T17 cells is predictive of poor survival in LUAD

In addition to Th17 cells, approximately 40% of γδ T cells also expressed Il17a (Fig. 6D). Examining every T cell cluster revealed that most Il17a+ cells, regardless of cell type, were from A. temperans mice (Fig. 7A). However, given the high percentage of Il17a+ cells in Th17 and γδ T clusters, we hypothesized that gene expression may be similar in both clusters which could then identify a gene set important for IL-17 polarization in pan T cell subtypes (T17). To test this, we combined these two clusters and calculated differentially expressed genes against all other T cells. We then calculated a T17 gene signature score from the expression of each of the genes upregulated in both clusters. The resulting T17 signature was highest in Th17 and γδ T cells, and specifically upregulated in Il17a+ cells compared to Il17a- cells in both Th17 and γδ T clusters (Fig. 7B, C, Supplementary Table S6). Interestingly, this signature score was higher in A. temperans mice in Th17 but not γδ T cells (Fig. 7D). We also observed upregulation in effector CD8+ T cells from A. temperans mice, which suggests repeated exposure to A. temperans induces a T17 polarization in multiple T cell subtypes.

Fig. 7: A pan T17 gene signature is predictive of poor prognosis in human LUAD.

A Percent of cells Il17a positive per T cell subtype and treatment group. B UMAP projection of a pan T17 cell gene signature within the T cell compartment. Expression of a pan T17 cell gene signature in Il17a negative (absent) and positive (expressed) cells within the (C) Th17 and (D) γδ T clusters. E Expression of a pan T17 cell gene signature in all T cell clusters split by treatment group. F Correlation of genus-level Acidovorax abundance and the pan T17 gene signature score within the TCGA LUAD dataset [13]. G Kaplan-Meier curves for survival property within four human LUAD cohorts – GSE30219, GSE31210, GSE50081, and TCGA [74,75,76,77]. Boxplots indicate median and quartile scores. * p < 0.05, ** p < 0.01, **** p < 0.0001.

We then asked if this T17 gene signature was important in human lung cancer. Leveraging the metatranscriptomics data that we had previously generated using TCGA LUAD [13], we examined the association of the T17 signature with Acidovorax exposure in these patients. The T17 signature score was weakly, but positively, correlated with Acidovorax abundance (Fig. 7F), suggesting microbial dysbiosis may also influence T17 polarization in human LUAD. Next, we asked if high expression of the T17 signature was predictive of patient survival. We stratified patients by low or high expression of the T17 signature score in four cohorts of LUAD patients, including TCGA. This stratification revealed high expression of the T17 gene signature was a poor prognostic in LUAD for overall survival (Fig. 7G). These results suggest T17 polarization, regardless of T cell receptor subtype, accelerates tumor development and results in worse survival in patients.

Cell-cell signaling switches from IL-1β driven to broad proinflammatory activation in response to A. temperans

We then investigated cell-cell communication to determine the potential mechanism of A. temperans-mediated LUAD progression. Examination of cytokines within all cell types revealed immune cell-of-origin for multiple cytokines previously implicated in KP mouse etiology: IL-1β (neutrophils), IL-23 (neutrophils), and IL-17 (T cells) (Fig. 8A). In contrast to previous results, IL-22 was not detected, and AMs were not a major source of IL-1β or IL-23 [11], suggesting that introduction of an external bacterial species dramatically alters immune cell signaling in KP mice.

Fig. 8: Cell-cell communication demonstrates A. temperans induces robust signaling between neutrophils, macrophages, and T cells.

A Dotplot of cytokine expression by cell type. B Heatmap depicting differential interactions as calculated by expression of ligand in outgoing cell type and its cognate receptor in the incoming cell type, with total number (left) and interaction strength (right). Cell-cell interactions enriched in sham mice are indicated by blue and those enriched in A. temperans mice indicated by pink. C Cross-referenced incoming and outgoing interaction strength for each cell type in sham (left) and A. temperans (right) mice. D Heatmap of overall signaling patterns by cell signaling pathway for each cell type in sham (left) and A. temperans (right) mice. E Chord diagrams showing pathways significantly enriched in Sham (top) and A. temperans mice (bottom). Chord width indicates aggregate expression of the ligand and receptor, arrow indicates direction from sender to receiver population, outer rings indicate sender cell type, inner rings indicate receiver cell type, and links are colored by interaction pairing. Individual chords are colored by secreting cell type and arrows indicate receptor cell type.

We then compared cell-cell communication by examining combined changes in expression of ligand-receptor pairs in sham and A. temperans mice. Examination of the total signaling strength (combined incoming and outgoing signal) revealed a specific increase in number of interactions among neutrophils, AMs, macrophages, and T cells in response to A. temperans (Fig. 8B, C, Supplementary Tables S7,