Stromal gene signatures in large-B-cell lymphomas

Abstract

Background: The addition of rituximab to combination chemotherapy with cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP), or R-CHOP, has significantly improved the survival of patients with diffuse large-B-cell lymphoma. Whether gene-expression signatures correlate with survival after treatment of diffuse large-B-cell lymphoma is unclear.

Methods: We profiled gene expression in pretreatment biopsy specimens from 181 patients with diffuse large-B-cell lymphoma who received CHOP and 233 patients with this disease who received R-CHOP. A multivariate gene-expression-based survival-predictor model derived from a training group was tested in a validation group.

Results: A multivariate model created from three gene-expression signatures--termed "germinal-center B-cell," "stromal-1," and "stromal-2"--predicted survival both in patients who received CHOP and patients who received R-CHOP. The prognostically favorable stromal-1 signature reflected extracellular-matrix deposition and histiocytic infiltration. By contrast, the prognostically unfavorable stromal-2 signature reflected tumor blood-vessel density.

Conclusions: Survival after treatment of diffuse large-B-cell lymphoma is influenced by differences in immune cells, fibrosis, and angiogenesis in the tumor microenvironment.

Figure 1.. Gene-Expression Predictors of Survival among Patients with Diffuse Large-B-Cell Lymphoma Treated with R-CHOP.

Kaplan–Meier estimates of progression-free and overall survival are shown. Panel A shows that patients with germinal-center B-cell–like diffuse large-B-cell lymphoma had a higher probability of progression-free survival (left) and overall survival (right) than patients with activated B-cell–like diffuse large-B-cell lymphoma. Panel B shows a gene-expression–based predictor of survival among patients with diffuse large-B-cell lymphoma treated with R-CHOP. Kaplan–Meier estimates of progression-free survival (left) and overall survival (right) are based on a multivariate model derived from the germinal-center B-cell, stromal-1, and stromal-2 gene-expression signatures. Survival-predictor scores derived from this model were used to rank the cases of lymphoma, which were then divided into quartile groups as indicated. R-CHOP denotes rituximab plus combination chemotherapy with cyclophosphamide, doxorubicin, vincristine, and prednisone.

Figure 2.. Cellular Derivation of Prognostic Gene-Expression Signatures in Diffuse Large-B-Cell Lymphoma.

Panel A shows the relative gene expression of the stromal-1, stromal-2, and germinal-center B-cell signatures in CD19+ malignant and CD19− nonmalignant subpopulations of cells isolated from three biopsy specimens from patients with diffuse large-B-cell lymphoma (DLBCL). Stromal-1 and stromal-2 signature genes were more highly expressed in the nonmalignant cells, whereas the germinal-center B-cell signature genes were more highly expressed in the malignant cells. The log₂ ratios of gene-expression levels in the CD19− subpopulation to those in the CD19+ subpopulation are depicted according to the color scale shown. Panel B shows the relationship of the stromal-1 signature to gene-expression signatures derived from normal cells. Gene-set enrichment analysis established a relationship between two signatures that are expressed in cells and tissues of mesenchymal origin (normal mesenchyme-1 and mesenchyme-2 signatures) and a monocyte signature, which is expressed more highly in normal blood monocytes than in B cells, T cells, and natural killer cells. No relationship was observed between the stromal-1 signature and a pan–T-cell signature, which is expressed more highly in T cells than in B cells, natural killer cells, and monocytes. The relative levels of gene expression within each sample are depicted according to the color scale shown. R-CHOP denotes rituximab plus combination chemotherapy with cyclophosphamide, doxorubicin, vincristine, and prednisone.

Figure 3.. Expression of Stromal-1 and Stromal-2 Signature Genes in Biopsy Samples from Patients with Diffuse Large-B-Cell Lymphoma.

The relative levels of gene expression within each sample are depicted according to the color scale shown, with each row representing a different signature gene and each column representing a different biopsy sample. The stromal-1 and stromal-2 signature averages are shown for each patient. The stromal score is the component of the multivariate survival model contributed by the difference between the stromal-2 and stromal-1 signature averages. The samples were ranked within each diffuse large-B-cell lymphoma subtype according to the survival-predictor score, which was generated by the survival model and incorporates both the stromal score and the germinal-center B-cell signature (not shown). Representative signature genes are shown. R-CHOP denotes rituximab plus combination chemotherapy with cyclophosphamide, doxorubicin, vincristine, and prednisone.

Figure 4.. Immunohistochemical Analysis of Biopsy Specimens from Patients with Diffuse Large-B-Cell Lymphoma.

Panel A shows that fibronectin, a stromal-1 signature component, was localized prominently in fibrous strands running between malignant cells in the biopsy specimens. Panels B, C, and D show matrix metalloproteinase 9 (MMP9), secreted protein acidic cysteine-rich (SPARC), and connective-tissue growth factor (CTGF), respectively. Each of these components of the stromal-1 signature was localized primarily in histiocytic cells infiltrating the biopsy specimens. Panels E and F show colocalization of SPARC and CTGF with the myelomonocytic marker CD68. Panel G shows Kaplan–Meier estimates of overall survival according to the level of expression of SPARC. Panels H and I show tumor blood-vessel density according to immunohistochemical analysis of CD34+ endothelial cells in biopsy specimens. Representative cases with low or high blood-vessel density (CD34+ cells per μm) are shown. Panel J shows the correlation between tumor blood-vessel density and the stromal score.