Myeloid cells

We identified 214,960 and 160,921 myeloid cells in the tumour and B/H respectively, characterised by the global expression of LYZ and SPI1. We first identified in both data-sets a population of monocytes (expressing CD14, IL1B, FCN1, CXCL8, S100A8, and S100A9) and a broad population of macrophages (expressing CD14, CD163, CD68, MARCO and MRC1) (Figure 1 D, Supp. figure 2). From here we subclustered Mɸ to anti-inflammatory Mɸ (expressing C1QB, C1QC, APOE, MRC1, MARCO) and alveolar Mɸ (expressing FABP4, MCEMP1 and PPARG). We mainly detected alveolar Mɸ in the B/H samples whereas the tumour samples were significantly enriched in monocyte-derived anti-inflammatory Mɸ, as previously described1 . Anti-inflammatory Mɸ are involved in the resolution of inflammation and suppress the immunity against tumour cells, thus enabling the tumour microenvironment to promote cancer progression. We also noted an enhanced heterogeneity of anti-inflammatory Mɸ in the tumour, with several additional Mɸ states being uniquely present in tumour samples. Namely, we observed a population of cancer-associated macrophage-like cells (CAMLs) found to co-express epithelial (KRT18, KRT19, EPCAM) and macrophage markers (LYZ, SPI2, CD14, CD163, C1QC, APOE). Additionally, STAB1+ Mɸ, identified by high expression of scavenger receptor stabilin1 (STAB1), were also located in the tumour environment. High expression of STAB1 in Mɸ has been associated with poor prognosis in advanced cancers2 . Finally, a population of cycling anti-inflammatory Mɸ was exclusively identified in tumour samples, characterised by the expression of MKI67 and TOP2A.

Dendritic cells (DC) are professional antigen-presenting cells that were characterised by expression of CLEC10A, CLEC4A, and CD1C. Further subclustering of DCs revealed monocyte-derived DC2 (mo-DC2, expressing CD14, CD163, MRC1) 3 and conventional DC2 (cDC2) that lacked CD14 expression. Mo-DCs were more abundant in the tumour samples compared to background, consistent with an increased monocyte infiltration and differentiation. In B/H samples, populations of cycling mo-DC2 and cycling cDC2 cells were also identified (expressing MKI67 and TOP2A).

Clusters of cells that had high expression of LYZ and SPI1 and no expression of DC or monocyte/macrophage markers (e.g. CD14 and CD163, respectively) were defined as immature myeloid cells.

T cells

T lymphocytes play a central role in engaging the immune system in fighting cancer. We collected 124,459 and 105,127 T cells in the tumour and B/H samples respectively. T cell populations (expressing TRAC, CD3D, CD3E, CD3G) across both data-sets included non-cytotoxic, cytotoxic T cells, naive T cells (expressing CD40LG) and downregulated T cells, corresponding to states with low or no expression of genes defining the exhausted, naive, or cytotoxic state, but with moderate expression of the global CD3 markers. The non-cytotoxic state includes moderate expression of CD4 and these cells play an active role in the adaptive immune system through the release of specific cytokines to control and polarise the immune response. In contrast, the cytotoxic T cells are programmed to recognise specific antigens and destroy pathogens. We identified two populations of cytotoxic T cells that displayed a graded expression of the cytotoxicity markers GZMA, GZMB, GZMK, IFNG, as well as NKG7 and PRF14 .

Across these cytotoxic and non-cytotoxic states we were also able to assess the degree of T cell exhaustion. In the tumour sample, we identified two populations of exhausted T cells, again with the varying degree of expression of cytotoxic genes leading us to classify these as exhausted cytotoxic T cells and exhausted T cells respectively. This exhausted state (expressing TIGIT, CTLA4, PDCD1, LAG3), indicates a reduced capacity to excrete cytokines, inhibiting the growth and activity of other immune cells. We further observed the tumour-specific exhausted states expressing CXCL13. Complementary studies5 have indicated that immunosuppressive environments can lead to upregulation of CXCL13 in T cell populations, consistent with our observation of CXCL13 upregulation in the exhausted T cell clusters of the tumour samples. We also identified a cycling state of exhausted cytotoxic T cells (additional expression of MKI67 and TOP2A) in both samples. 

Specifically in tumours we observed the regulatory T-cells (Tregs, expressing FOXP3, CTLA4, TNFRSF18) 6 . FOXP3 is a transcription factor essential for the development and inhibitory function of Tregs. The suppressive activity of Treg involves the coordinate activation of CTLA4 and TNFRSF18 and repression of IL2 and IFNG. Finally, in the B/H populations we identified a modest population of 5539

NK cells

We identified 36,046 and 80,703 NK cells in the tumour and B/H environments respectively, where in both samples two distinct populations of NK cells with high and low cytotoxicity were identified according to clear differences in the expression of NKG7, PRF1, GZMA, GZMB, GZMM, GZMK, KLRB1. In the B/H sample the higher cytotoxic state represented the majority of NK cells (71,833 out of 80,703 cells), while in tumour samples there was an even distribution of higher and lower cytotoxic NK cells. Consistent with this expansion of less cytotoxic NK cells, it was recently described that tumour cells are able to reprogram NKs inducing a resting phenotype which, in turn, promotes tumour metastases8 .

B cells

We identified 84,232 and 7,916 B cells in the tumour and B/H samples respectively. Common to both samples, mature B cells were defined based on their expression of global CD79A, CD79B, and IGHM markers. Varying expression of JCHAIN, IGHG1, IGKC, IGHA1, XBP1, and MZB1 was used to identify plasma B and immature plasma B cells, whereas cycling plasma B cell showed substantial upregulation of MKI67 and TOP2A. Unique to the tumour sample, we also identified a population of LYZ+ B cells, with upregulated LYZ expression common to the myeloid cell states, a set of Downregulated B cells which downregulated most of B cells markers (e.g. CD79A, CD79B, MS4A1) and TNF+ B cells that co-expressed B cell and T cell specific genes (MS4A1, CD79A, CD3D, CD3E, CD3G) in addition to a high level of TNF.

In B/H we also identified a small population of 526 NKB cells, identified through the shared expression of global NK (NKG7, PRF1) and B cell (IGHM, CD19, and MS4A1) markers9 , in addition to ID2 (a key transcription factor for NK development). NKB cells are reported to have unique immunity features which distinguish them from B and T cells, including the specific excretion of IL12 and IL18 as a means of eradicating microbial infection10 .

Mast cells

We identified two distinct populations of mast cells in both the tumour and B/H samples (5,255 cells and 4,387 cells respectively). Mast cells were identified through the high expression of HDC, as well as MITF and GATA2, key transcription factors regulating mast cell identity and responsiveness to external factors such as antigenic stimulation11. The high expression of MKI67 and TOP2A also allowed us to resolve a second population of cycling mast cells. 

Non-immune cells

The non-immune component of the TME has been implicated in resistance to multiple types of cancer therapy. In tumour and B/H data-sets we identified 36,025 and 23,572 epithelial cells respectively (expressing EPCAM, CDH1, KRT18, and KRT19) that were further subclustered to ciliated cells (expressing FOXJ1, and RFX2), alveolar type 2 (AT2) cells (expressing SFTPB, MUC1, SFTPC), cycling AT2 cells (additionally expressing MKI67 and TOP2A) and two distinct clusters of epithelial cells present in tumours, namely atypical epithelial cells which downregulated epithelial markers and transitioning epithelial cells which upregulated myeloid markers. In B/H, but not in tumour, we identified a population of club cells (expressing SCGB1A1). A separate population of cycling epithelial cells (based on high MKI67 expression) were identified in the tumour, and annotated as a distinct phenotype based on their separation in UMAP-space.

Common to both tumour and B/H we identified respectively 304 and 4,105 lymphatic endothelial cells (expressing endothelial markers CLDN5, PECAM1 as well as lymphatic endothelium associated genes RAMP2, CCL21, LYVE1). Next, we identified fibroblasts (high expression of COL1A1, COL1A2, SPARC, and DCN) as well as activated adventitial fibroblasts (co-expressing SERPINF1 and THY1 and α smooth muscle actin, ACTA2). 

Consistent with a recent report12, we did not identify AT1 cells in either tumour or B/H tissues

Evaluating cell-type annotations following label transfer

To test consistency in cell-type annotation performed separately in tumour and B/H, we performed reference-query mapping from tumour to B/H using scArches13 (see Methods). The cell type proportions obtained by scVI integration across samples and label transfer correlated closely with separate dataset proportions (Pearson correlation coefficient 0.89). However, we observed discrepancy between the relative abundance of DCs and monocytes. Considering that monocytes can differentiate to DCs under inflammatory conditions these two populations can have similar transcriptional signatures. To test whether monocytes (from separate annotations) assigned to DCs through label transfer express putative DC markers, we plotted a set of known monocyte and DC genes and examined their expression across relevant populations. Our analysis showed that monocytes assigned to be DCs following label transfer do not express putative DC genes such as CDC1, CLEC4A, CLEC10A confirming the correctness of separate annotations (Suppl Figure 2A-C). 

https://www.nature.com/articles/s41467-024-48700-8

https://static-content.springer.com/esm/art%3A10.1038%2Fs41467-024-48700-8/MediaObjects/41467_2024_48700_MOESM1_ESM.pdf