The histological growth patterns in liver metastases from colorectal cancer display differences in lymphoid, myeloid, and mesenchymal cells

Abstract

Colorectal liver metastases grow following different histologic growth patterns (HGPs), classified as desmoplastic and nondesmoplastic (dHGP, non-dHGP), being the latter associated with worst prognosis. This study aimed to investigate the tumor microenvironment (TME) between HGPs supporting different survival. Multiplexed immunohistochemical staining was performed with the Opal7 system in a 100-patients cohort to evaluate the tumor-liver interface with three different cell panels: lymphoid, myeloid, and carcinoma-associated fibroblasts. Differences between HGPs were assessed by Mann-Whitney U test with Pratt correction and Holm-Bonferroni multitest adjustment. Cytotoxic T-cells were more abundant in tumoral areas of dHGP, while non-dHGP had higher macrophages infiltration, Th2, CD163⁺, and Calprotectin⁺ cells as well as higher pSMAD2 expression. Regarding carcinoma-associated fibroblasts, several subsets expressing COL1A1 were enriched in dHGP, while αSMA^low_single cells were present at higher densities in non-dHGP. Interestingly, Calprotectin⁺ cells confer better prognoses in non-dHGP, identifying a subgroup of good outcome patients that unexpectedly also show an enrichment in other myeloid cells. In summary, our results illustrate different TME landscapes with respect to HGPs. dHGP presents a higher degree of immunocompetence, higher amounts of Collagen 1 as well as lesser presence of myeloid cell populations, features that might be influencing on the better prognosis of encapsulated metastases.

Keywords: capsule; desmoplasia; hepatic metastases; histologic growth pattern; microenvironment.

FIGURE 1

Characterization of lymphoid cells and subsets in desmoplastic and nondesmoplastic liver metastases. (A) Box‐plot graphs (mean values and interquartile range) of cell densities (cells/mm²) for the different lymphoid cell markers used and measured in the different compartments assessed. To compare differences, we used Mann–Whitney U test with a Pratt correction for zeros and ties plus Bonferroni adjustment for multiple testing. Asterisks denoted statistical significance, *p < 0.05. (B) As above, box‐plot graphs for main lymphoid subsets. (C) Representative images of CD8 staining for an encapsulating metastasis (dHGP; left image) and nondesmoplastic metastasis (non‐dHGP, right image). As shown in the left image, CD8 cells were more abundant on the dHGP metastasis, both in stromal regions, basically in the fibrous capsule but also inside the tumoral compartment, allocated over the tumor cells. On the contrary, on the right image (non‐dHGP metastasis), CD8 were mainly retained in the liver and in the tumor liver interface (TLI). As segmentation markers, PanCK, in pink, for Pan‐Cytokeratin, and HSA, in cyan, for hepatic‐specific antigen. (D) CD8‐to‐CD4 ratios in the different compartments (left panel). In addition, we computed the ratio CD8^tumor/CD8^stroma (right panel), a metric that estimates the tumor exclusion of the CD8 infiltrates, being higher in the non‐dHGP metastases (p = 0.0055; Mann–Whitney U test with Pratt correction for ties). (E) We assessed the protein expression of CCR4, a cytokine receptor present in Th2 lymphocytes. Nonencapsulating metastases displayed higher densities of CCR4⁺ cells in the tumoral areas of the TMA cores (tumor cells + stromal areas; p = 0.023, Mann–Whitney U test with correction for ties). CCR4 was measured by conventional IHC.

FIGURE 2

Characterization of myeloid cells and subsets in desmoplastic and nondesmoplastic liver metastases. (A) Box‐plot graphs (mean values and interquartile range) of cell densities (cells/mm²) for the different macrophage and myeloid cell markers used and measured in the different compartments assessed. Compartments were determined using morphology and segmentation markers (PanCK, hepatic‐specific antigen, and nuclei staining). (B) As above, box‐plot graphs for polarized macrophages, myeloid nonmacrophages (CD68⁻CD163⁺) and Calprotectin_single cells. (C) SIA, signature of immune activation, ratio of CD8 cells to different immunosuppressive myeloid cells, either M2 macrophages, myeloid nonmacrophage, or Calprotectin⁺ cells. These metrics are considered a surrogate marker for the antitumoral status versus the immunesuppressive environment. (D) representative images of the different macrophage populations assessed. Left panel, for desmoplastic metastases and right panel for nondesmoplastic metastases. To compare differences in A, B, and C, we used Mann–Whitney U test with the Pratt correction for zeros and ties plus Bonferroni and/or false discovery rate adjustment for multiple testing. When Bonferroni adjustment was to astringent we applied FDR for multitesting correction. Asterisks denoted statistical significance, *p < 0.05; **p < 0.001; ***p < 0.0001 after Bonferroni correction. # denoted FDR < 0.05.

FIGURE 3

Kaplan–Meier survival analyses: (A) different behavior of Calprotectin_single enriched samples (Stroma) between encapsulated liver metastases (desmoplastic HGP; left panel) and nonencapsulated liver metastases (nondesmoplastic HGP, middle panel). However, when taking into consideration nondesmoplastic metastases, the high density of Calprotectin was associated with good prognosis only in a subset of patients enriched in macrophages (Log Rank p = 0.031; A, right panel) or also enriched in myeloid nonmacrophage cells (B, middle panel; Log Rank p = 0.008). In low‐enriched macrophages or myeloid cells metastases, Calprotectin_high cases did not confer better outcome (B, left and right panels). We used the mean density value for CD68 or myeloid nonmacrophage cells as a cut off in the entire cohort of metastases. Similar results have been obtained using third‐party data in sarcoma (C; left panel) and in head and neck cancer (D, left panel) high expression values of Calprotectin (considered here as the mean expression of S100A8 and S100A9 genes obtained in RNAseq public databases) provided a better prognosis than low expression in tumors enriched in macrophages.

FIGURE 4

Characterization of carcinoma‐associated fibroblasts and subsets in desmoplastic and nondesmoplastic liver metastases. (A) Box‐Plot graphs (mean values and interquartile range) of cell densities (cells/mm²) for the different CAFs cell markers used and measured in the different compartments assessed. Compartments were determined using morphology and segmentation markers (PanCK, hepatic‐specific antigen, and nuclei staining). (B) Among all the possible combinations of markers used to characterize the CAFs, we selected those that had a representation of at least a fraction equal to or greater than 15% of the total fibroblasts in any of the samples. The densities of the twelve selected were further assessed in the different compartments. To compare differences, we used Mann–Whitney U test with the Pratt correction for zeros and ties plus Bonferroni and/or false discovery rate adjustment for multiple testing. When Bonferroni adjustment was to astringent we applied FDR for multitesting correction. Asterisks denoted statistical significance, *p < 0.05; **p < 0.001; ***p < 0.0001 after Bonferroni correction. # denoted FDR < 0.05.

FIGURE 5

αSMA^low_single cells express high values of a iCAF signature. We used single cell transcriptomic data from 4397 CAFs isolated from six different colorectal cancer metastases. We visualized the tSNE plot for ACTA2 expression (A), showing a wide range of gene expression among all the CAFs. Thus, we computed a ssGSEA score for an iCAF and myCAF signatures (Supporting Information) and calculated the correlation of these ssGSEA scores with the ACTA2 expression. As shown in (B) (top panel), the inverse correlation between ACTA2 and the iCAF scores indicates that CAFs with high ACTA2 expression displayed low iCAF signature values (Spearmen correlation p = 8.65e−71) and just the opposite for a myCAF signature (bottom panel). When we separated the expression of ACTA2 into tertiles, the first tertile (lowest expression of ACTA2) was the one with the highest expression of the iCAF signature in a statistically significant way in relation to the other two tertiles. Finally, we depicted in (D) the tSNE plots for the iCAF and myCAF ssGSEA scores for the 4397 CAFs and the expression of some iCAF and myCAF classical genes. In gray, we represented values equal to zero, in orange scale the nonzero values up to percentile p66 and in blue scale the values above percentile p66.

FIGURE 6

Spearman correlation analysis (blue‐to‐red heatmaps; bar scale −1 to 1 illustrates correlation coefficients) in dHGP and non‐dHGP CRCLM considering the cell subsets evaluated in both total tissue excluding liver (A) or in the stroma only (B). Corresponding yellow‐to‐blue heatmaps illustrate the p values for each Spearman correlation. Heatmaps were made using the website https://software.broadinstitute.org/morpheus/.

FIGURE 7

Tumor microenvironment landscapes found in each histologic growth pattern. Two different TME scenarios are depicted depending on the HGP, where most nonencapsulated metastases display a highly immunosuppressive microenvironment while the better outcome of encapsulated metastases might be associated by the presence of cytotoxic T‐cells over the tumor nests and the restraining capabilities of an extracellular matrix enriched in Collagen 1. The dual role of Calprotectin, which seems to have a value dependent on HGP, is noteworthy (created with BioRender).