Development of patient-derived lymphomoids with preserved tumor architecture for lymphoma therapy screening

Abstract

The efficacy of anti-cancer therapies depends on the genomic composition of the tumor, its microenvironment, spatial organization, and intra-tumor heterogeneity. B-cell lymphomas are a heterogeneous group of tumors emerging from B-cells at different stages of differentiation and exhibiting tumor-specific interactions with the tumor microenvironment. Thus, the effect of drug treatments can be influenced by the tumor composition and functional interactions among immune cells. Here, we develop a platform to maintain small fragments of human lymphoma tissue in culture for several days, and use them to test response to small molecules. We collect 27 patient samples representative of different lymphoma subtypes, and establish ex vivo tissue fragments that retain histological, cellular, and molecular characteristics of the original tissue, here referred to as lymphomoids. Using lymphomoids, we test sensitivity to several clinically approved drugs in parallel and examine tissue remodeling upon treatment. Moreover, when this information is available, we show that the effect of the inhibitors observed in lymphomoids is consistent with the patients' response in the clinic. Thus, lymphomoids represent an innovative ex vivo model to assess the effect of anti-cancer therapies while preserving the tissue structure and its components.

Fig. 1. A system to maintain lymphoma biopsy ex vivo.

a Schematic of the procedure to process and test sensitivity to therapies for murine and human samples (created with Adobe Illustrator). b Hematoxylin and eosin staining of a vavP-Bcl2 mouse spleen and a representative lymphomoid derived from the same spleen (upper images), and B220 + KI67 immunofluorescence of 3 independent lymphomoids (lower images) from 2 vavP-Bcl2 mice (49 weeks old and 58 weeks old). Scale bar 100 μm. c Representative images of 5-color immunofluorescence staining for the indicated markers of a 49 weeks old vavP-Bcl2 mouse spleen and a lymphomoid derived from the same spleen at two different resolutions and a digitalized image for the lymphomoid as a magnified cropped detail. d Quantification of different cell types based on fluorescence signal at different time points (the data points represent different lymphomoids and sections from 3 independent vavP-Bcl2 mice aged 47, 49, and 58 weeks old; spleen n = 3, lymphomoid sections n = 17 for 1 week, n = 7 for 3 weeks, n = 6 for 5 weeks). Data are presented as mean values +/− SD. e Quantification of fluorescence signal to detect proliferating B-cells at different time points (3 independent vavP-Bcl2 mice aged 47, 49, and 58 weeks old; lymphomoid sections n = 17 for 1 week, n = 7 for 3 weeks, n = 6 for 5 weeks). Data are shown in box and whiskers (min to max) with all data points, where the center represents the median, and were analyzed by Kruskal–Wallis followed by Dunn’s posthoc test with correction for multiple comparisons using hypothesis testing. f Quantification of different cell types by flow cytometry of vavP-Bcl2 mouse lymphomoids at different time points (39 weeks old mouse; n = 4 lymphomoids for each time point; the data is plotted as the mean and error, which represents the SD). g–i UMAP projection of single cells RNA-sequencing cells color coded for cells expressing CD79a or CD3 genes (g), color coded by the origin of the samples (2 independent vavP-Bcl2 spleens from mice aged 43 and 47 weeks old, and n = 4 lymphomoids 1 week, and n = 2 lymphomoids 2 weeks) (h), color coded based on the different cell clusters found in the analysis (i). Source data are provided as a Source data file.

Fig. 2. Characterization of human lymphomoids obtained from fresh tumor biopsies.

a Graphical summary of all the cases collected in this study. b Representative images of multicolor immunofluorescence staining of indicated markers of three human lymphoma biopsies (upper panel) and their respective lymphomoids (lower panel). Scale bar 50 μm. c Spearman’s correlation coefficient (two-sided) analyses of untreated lymphomoids vs original tumor biopsies on the different markers analyzed by multiplex IHF (n = 17 B-cell lymphomas; one tumor was excluded from this analysis since it was CD20 negative). The data is shown as a percentage and was normalized to the total number of nuclei detected in each section. The graphs show the regression line and the confidence interval at 95%. d Spearman’s correlation coefficient (two-sided) analyses of untreated lymphomoids vs original tumor biopsies on proliferating CD20+ cells analyzed by multiplex IHF (n = 17 B-cell lymphomas). The data is shown as percentage and was normalized to the total number of CD20+ cells in each section. The graph shows the regression line and the confidence interval at 95%. e Similarity index comparisons in the indicated conditions (N inter = 171, N intra = 19 for both lymphomoid-lymphomoid, and lymphomoid-tumor, originated from a total of n = 399 lymphomoid image acquisitions and n = 85 tumor image acquisitions). The data are shown as boxplots where the upper and lower hinges represent the 25th and 75th percentiles and the center is the median, and were analyzed using the Wilcoxon rank-sum test two-sided with multiple hypothesis correction using the Benjamini–Hochberg procedure. f H&E staining of patient 8 DLBCL biopsy. Tissue features are highlighted with lines of different colors. g, h Representation of 10X Genomics Visium spots color-coded by the tumor score (g) and based on unbiased clustering (h). i Heatmap of cell type enrichment in each spot of the biopsy determined using Bayesprism deconvolution. Only top 4 differentially enriched cell types in each cluster are shown. The cell state proportions are z-scale normalized for each cell type. The dendrogram shows the distances between the different clusters. j H&E staining of six lymphomoids maintained in culture for 3 days. Tissue features are highlighted with lines of different colors. Scale bar 500 μm. k, l Representation of 10X Genomics Visium spots color-coded by the tumor score (k), based on cluster scores obtained in the original tissue (l-top), and pie-charts summarizing the cluster composition of each lymphomoid (l-bottom). Source data are provided as a Source data file.

Fig. 3. Sensitivity to targeted therapies using patient-derived lymphomoids.

a Representative images of immunofluorescence staining for KI67 and CD20 in lymphomoids derived from 4 different biopsies. Scale bar 50 μm. b, c Weighted ratios of proliferating CD20+ cells (b) and total CD20+ cells (c) compared to untreated controls. Each treatment is color-coded and the p-values were calculated using non-parametric ANOVA followed by Dunn’s test (two-sided) with p-value adjustment using Bonferroni’s correction. d H&E and tumor score of ibrutinib and idelalisib-treated lymphomoids. e UMAP projection of 10X Genomics Visium spots colored by sample (left) and by cluster (right). f Heatmap of differentially expressed genes in each cluster and sample obtained in the 10X Genomics Visium experiment. g Comparison of cell type enrichment in ibrutinib (top) and idelalisib (bottom) treated lymphomoids versus control. Source data are provided as a Source data file.

Fig. 4. Matched results obtained in the clinic and on lymphomoids in seven patients.

a–f PET-CT before and after treatment, representative image and quantification (of the CD20 and KI67 immunofluorescence signal of the lymphomoids derived from patient 6 (a, ctrl n = 16; ibrutinib n = 9), patient 14 (b, ctrl n = 18; ibrutinib n = 8), patient 11 (c, ctrl n = 14; lenalidomide n = 7), patient 17 (d, ctrl n = 21; ibrutinib n = 33, lenalidomide n = 10), patient 18 (e, ctrl n = 44, ibrutinib n = 12), patient 13 (f, ctrl n = 14; ibrutinib n = 6, lenalidomide n = 12), and patient 24 (g, ctrl n = 9; lenalidomide n = 9). The data are shown as boxplots where the upper and lower hinges represent the 25th and 75th percentiles and the center is the median, and were analyzed by either two-sided Wilcoxon rank-sum test or Kruskal–Wallis followed by Dunn’s posthoc test. CMR complete metabolic response, NMR no metabolic response, PR partial response. Source data are provided as a Source data file.