Multiplatform Profiling Characterizes Functional Networks in Genomically Stable and Instable Chronic Lymphocytic Leukemia

Background: Genomically instable (GI) chronic lymphocytic leukemia (CLL) is characterized by frequent alterations in DNA-damage response (DDR) genes (e.g. TP53,ATM) and related pathways. Conversely, pathogenic networks in CLL cases which maintain genomic integrity and operate with a functional DDR remain incompletely described.

Methods: Molecular profiling was conducted on CD19 sorted samples derived from patients registered on the CLL8 study (1st-line, FC vs. FCR) for gene expression (GEP)(n=337, Exon 1.0 ST arrays, Affymetrix), copy number aberrations (CNAs) (n=309, SNP Arrays 6.0, Affymetrix) and mutation analyses/signature projections (n=171, whole exome sequencing, Illumina). FISH, IGHV and TP53 mutation analysis was conducted at trial enrolment.

Results: Unsupervised consensus clustering (k=2-6) on variably expressed genes (SD>0.5) was used for class discovery. Two small, but highly differentiated clusters were identified, characterized through NRIP1 and EBF1/tri12. GSEA also segregated the remaining samples into four major clusters showing signatures of inflammation (I) and without inflammation (NI). These clusters were further segregated into GI-CLL clusters with increased "DNA-repair" or clusters with "epithelial-mesenchymal transition"-like signatures (EMT-L). Variability for del(17p)/TP53 mutation was found across clusters (p<0.01), with a low incidence observed in (I/NI)EMT-L (28.3%) vs. (I/NI)GI (71.7%). GISTIC identified CNA enrichment (FDR≤0.25) and potential target genes in affected regions impacting genomic stability and inflammation. (I/NI)GI had gains covering MYC (8q24.21), XPO1/REL (2p16.1) and (NI)GI showed losses involving KNSTRN and BUB1B (15q15.1). Mutational signatures implicated in pathogenic processes in cancer (COSMIC database) showed low activations for signatures indicating defective MMR or DSB repair in (I)EMT-L compared to (NI)GI, especially in IGHV mutated cases. Low p53 and phospho-p53 protein levels indicated a dampened DDR in (I/NI)EMT-L. Losses of 11q22.1-q22.2 and losses exceeding cytoband 13q21.1 were identified in (I/NI)GI. Since affected genes (e.g. YAP1, MMP, protocadherins) are closely linked with EMT in solid tumors, these regions seem indispensable for genomically stable CLL with EMT-like networks. Due to the high frequency of alterations associated with genomic instability and inverse TP53 and ZEB1 GEP, we hypothesized that p53 activation inhibits EMT-like networks in CLL. Cell lines showed miR-200c induction and decrease of target EMT transcription factors ZEB1/TWIST1 after irradiation, resembling the p53-miR-200c-ZEB1 mediated EMT-suppression in solid tumors. To further study these observations we performed mass spectrometry proteomics analysis on two syngeneic murine models; Eμ-myc [C57BL/6J-TgN(Ighmyc)22Bri/J] hemizygous and Eμ-TCL1 [C57BL/6J-TgN(IghTCL1)22Bri/J] hemizygous and compared their CD19-MACS isolated malignant B cells to normal splenic B cells from wildtype mice. Proteome profiles from Eµ-MYC/Eµ-TCL1 mouse samples mirrored GSEA findings in patient samples showing DDR-induction and concurrent downregulation of EMT associated signatures, including inflammation and hypoxia. EMT-like CLL showed transcriptional signatures indicating inflammation in the majority of cases, increased motility, NOTCH-/TGFb-signaling and HIF1a upregulation; all central characteristics observed for EMT. Excluding cases with TP53 defect, (NI)GI showed PFS rates of 17% at 5 years vs. 47% in (I)EMT-L ((NI)GI: median PFS 29.8 vs. (I)EMT-L: 39.5 months, HR:1.83 (95%CI 1.12-3.01), p=0.02) when treated with FC. Addition of rituximab improved outcome only in (NI)GI showing PFS rates of 44% at 5 years, in contrast to 45% at 5 years in (I)EMT-L ((NI)GI: median PFS 58.3 vs. (I)EMT-L: 52.4 months, HR:1.07 (95%CI 0.65-1.74), p=0.797). (I)EMT-L cases therefore lack an increase of efficacy for the addition of rituximab.

Conclusion: We identify distinct biologic subtypes in CLL, characterized by either genomic instability or EMT-like networks which show heterogeneous extent of inflammation and a differential response to treatment. Maintenance of genomic stability and dampened DDR seem critical for pathogenic networks emerging in EMT-like CLL due to the reciprocal inhibition of underlying processes and provide a potential entry point for rational therapeutic intervention.