Novel Richter Syndrome Xenograft Models to Study Genetic Architecture, Biology, and Therapy Responses

Richter syndrome is the transformation of chronic lymphocytic leukemia (CLL)/small lymphocytic lymphoma into an aggressive lymphoma, most commonly a diffuse large B-cell lymphoma (DLBCL). Occurring in approximately 2%–10% of patients with CLL, Richter syndrome is highly aggressive, often refractory to treatment, with a poor outcome of approximately 8–14 months (1–3). Approximately 80% of cases are clonally related to the underlying CLL, while the remaining 20% of patients have a clonally unrelated DLBCL, with a better prognosis similar to that of de novo DLBCL. Phenotypically, Richter syndrome cells express CD20 and CD23, while CD5 is present in a fraction of cases. Reports in the last decade have identified clinical (advanced Rai stage and chemotherapy regimen), biological (CD38 and CD49 expression), and genetic (stereotyped B-cell receptor genes, del17p13, CD38 and LRP4 polymorphisms) features associated with increased risk of Richter syndrome development (1, 4–6). In addition, mutations or abnormalities in specific genes, such as CDKN2A, TP53, MYC, and NOTCH1, appear to be important in the pathogenesis of the disease (7, 8). Despite significant advances in treatment options for patients with CLL, Richter syndrome remains a devastating end stage complication, representing an unmet need for patients with CLL. The lack of Richter syndrome cell lines or murine models, along with the reduced availability of primary samples, have limited understanding of the genetic and molecular mechanisms driving or contributing to the pathogenesis of this disease and the development of novel effective therapies.

Here we report the first establishment of two Richter syndrome patient-derived xenograft (RS-PDX) models, which maintain a close relationship with the primary Richter syndrome clone, offering novel opportunities to highlight pathogenic mechanisms and to uncover signaling pathways of therapeutic interest.

Patient samples were obtained at Weill Cornell Medicine after written inform consent in accordance with Institutional guidelines and the Declaration of Helsinki. The Institutional Review Board approved the study (IRB Weill Cornell Medicine; Protocol: #0107004999A008). The referring physicians (J.N. Allan and R.R. Furman) provided clinical features of patients' samples. Richter syndrome cells were purified using Ficoll-Hypaque (Sigma-Aldrich) from peripheral blood (in the case of RS1316) or lymph node (in the case of RS9737) of two patients with Richter syndrome presenting with typical morphology and immunophenotype. Lymph node was disrupted and single-cell suspension obtained. Richter syndrome cells were phenotypically characterized and analyzed by flow cytometry using a FACSCantoII cytometer (BD Biosciences). Antibodies used for staining were: CD5-FITC and CD19-APC (Milteny Biotec), CD21-FITC, CD23-PE and CD45-PerCP (eBioscience), CD20-FITC and CD49d-PE (BD Biosciences), and CD38-Alexa488 (AbD Serotec).

Phenotypic features were also analyzed by IHC staining. All staining was performed using the Bond Polymer Refine Detection kit and the auto stainer Bond III from Leica Microsystems. Antibodies used for staining were: CD5 (clone 4C7, #PA0168), CD20 (clone MJ1; #PA0906), and CD23 (clone 1B12; #PA0169) from Leica Biosystems; PAX-5 (clone 1B12; #610863) from BD Biosciences and Ki-67 (clone MIB-1; #M7240) from Dako. EBV status was checked by Epstein-Barr encoding region (EBER) in situ hybridization and by Western blotting using an anti-EBNA2 antibody (clone PE2, Leica Microsystems).

A total of 2 × 10⁷ primary Richter syndrome cells from peripheral blood or lymph nodes were resuspended in Matrigel (BD Biosciences) and subcutaneously injected (double flank) in 8-week old NOD/SCID/g chain^−/− (NSG) immunocompromised mice and left to engraft. Tumor masses were then collected, partially disrupted, and tumor cells reinjected as a single-cell suspension in Matrigel. These steps were repeated several times to obtain stable models of Richter syndrome. For the intravenous model, 10⁷ Richter syndrome cells purified from tumor masses were resuspended in PBS and injected in the tail vein of NSG mice. Disease engraftment was monitored by MRI using the Aspect M2 Compact high-performance magnetic resonance imaging system (Aspect Imaging). When presenting clear signs of disease (i.e., tumor volume >1, 5 mm³, splenomegaly or cachexia), mice were euthanized and peripheral blood and organs (kidneys, spleen, bone marrow, lung, liver, and brain) collected and disease spread evaluated by flow cytometry, using anti-human antibodies (CD45PerCPCy5.5/CD19APC/CD20FITC). Sectioned portions of each organs were formalin-fixed and paraffin-embedded and localization of tumor cells confirmed by IHC staining with an anti-CD20 antibody (Novocastra).

Engraftment of Richter syndrome cells was obtained while T. Vaisitti and S. Deaglio were Visiting Scientists at Weill Cornell Medicine (New York, NY). The Institutional Animal care and Use Committee approved all the experiments involving mice. For the establishment and propagation of the models, mice were treated following the European guidelines and with the approval of the Italian Ministry of Health (Authorization #12/2016-PR; protocol #CC652.5).

Somatic rearrangements of primary and PDX samples were analyzed by next-generation sequencing (NGS) based on LymphoTrack IGHV Leader Somatic Hypermutation Assay Panel – MiSeq (Invivoscribe) with small modification. Briefly, VDJ sequencing libraries were amplified by a single multiplex PCR reaction with barcoded primers, followed by library purification through AMPure beads and library quantification by Qubit/Tapestation. Libraries were then pooled and loaded into Miseq for NGS sequencing (600 cycles). For NGS data analysis, the LymphoTrack IGH SHM MiSeq(r) Software provided by Invivoscribe to identify VDJ clonality and leukemic VDJ sequencing was used.

Genomic DNA from each sample was sheared and used for the construction of paired end sequencing library. The exome was captured using the SureSelect Human All Exon V6 (Agilent) and sequenced using Illumina Hiseq4000. We used Genome_GPS v3.0.2 (formerly named TREAT; ref. 9) as a comprehensive secondary analysis pipeline for exome sequencing data at Mayo Clinic. For details, see Supplementary Materials and Methods.

Mutations were validated and followed-up overtime using semiconductor-sequencing technology (IonTorrent PGM), as per manufacturer's protocol and fully described in Supplementary Materials and Methods (10).

Approximately 500 ng of total RNA was subjected to the RNA-seq library preparation using the TruSeq RNA Sample Prep Kit (Illumina). Sequencing reads were filtered out for low quality sequences and trimmed of low quality bases by using FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/). Mapping to hg19 genome and gene quantification was performed by using TopHat software (https://ccb.jhu.edu/software/tophat/index.shtml) and Cufflinks (http://cole-trapnell-lab.github.io/cufflinks/). Downstream analyses were performed on genes showing RPKM (Reads Per Kilobase Million mapped reads) >1 at least in one sample. The P value threshold adopted for the analysis was P < 0.01, and, when comparing PDXs with their parental primary samples, sequences with a minimum fold change <1 in absolute value (|logFc|<1) were considered nondifferentially expressed. Expression analyses were performed as described in Supplementary Materials and Methods (11).

For details regarding pathway analysis by qRT-PCR and Western blotting, see Supplementary Materials and Methods.

Neoplastic cells from lymph node or peripheral blood biopsies of two pathologically diagnosed patients with Richter syndrome patients, RS9737 and RS1316, respectively, were subcutaneously injected in NSG mice. Both patients had a previous history of CLL, had received treatment for Richter syndrome at the time of xenografting, and both died 8 months after diagnosis, as summarized in the clinical vignette in Fig. 1A. Tumor cells from both patients were EBV-negative, as determined by in situ hybridization for EBER (performed on primary samples) and by Western blotting using an anti-EBNA2 antibody (performed on PDX samples; Fig. 1A; Supplementary Fig. S1A and S1B).

After the first engraftment, which occurred in 3 weeks for RS9737 and in 24 weeks for RS1316, tumor masses were reimplanted for a minimum of 10 passages (Supplementary Fig. S2A and S2B). Stabilized RS9737 model takes 21 ± 4 days to reach a subcutaneous mass of approximately 1.5 cm³ after injecting 5 × 10⁶ cells, while RS1316 takes 40 ± 5 days (Fig. 1B). After in vivo stabilization using the subcutaneous approach, both models were also injected intravenously in NSG mice to model the human disease (Supplementary Fig. S2A and S2B; refs. 3, 12). MRI, flow cytometry, and IHC analyses documented extensive and diffuse involvement of spleen, bone marrow, peripheral blood, and extranodal organs, such as brain, lung, and liver upon injection of 10⁷ cells after approximately 4 weeks in the case of RS9737 and approximately 8 weeks in the case of RS1316 (Fig. 1B; Supplementary Fig. S3).

PDX-derived Richter syndrome cells were positive for B-cell markers (CD19, CD21, and CD23), while CD5 was expressed only by RS1316, in agreement with the phenotype of primary Richter syndrome sample. CD38 expression levels were comparable with those of the primary Richter syndrome tumors (Supplementary Fig. S4). IHC analyses confirmed robust expression of CD20, CD23, and of Pax5, in accordance with a DLBCL diagnosis. Primary and PDX tumor cells were highly proliferative, as highlighted by Ki-67 staining (Fig. 1C). Both patients and RS-PDX carried unmutated IGHV genes, IGHV3-7 in the case of RS9737, while case RS1316 carried IGHV3-21 (Supplementary Table S1).

To characterize the genomic architecture of the two RS-PDX models, we first performed targeted deep sequencing of a panel of 28 genes, which are recurrently mutated in CLL (see full list in Supplementary Table S2; Fig. 2A and B; ref. 13). Subsequently, whole-exome sequencing (WES) was used to confirm data, to calculate copy number abnormalities, and to look for additional mutations. Primary cells and PDXs at different in vivo passages were analyzed.

For case RS9737, targeted sequencing was performed on the primary and on cells from passages (P) 1, 3, 5, 7, and 10, while WES was performed on cells from the primary, P1 and P3. RS9737 was characterized by a complex karyotype with more than 20 chromosomal abnormalities, including biallelic deletion of 9p21 (CDKN2A) and monoallelic deletions of 11q23 (ATM) and 17p (TP53) (Table 1). Looking at the chromosomal abnormalities, all lesions present in the primary tumor remained in the PDX, with the sole gain of 19p13 (Table 1). Targeted sequencing revealed that RS9737 had clonal frameshift mutations in NOTCH1 (p.2514fs4) and the residual copy of TP53 (p.H214fs), along with nonsynonymous mutations in KRAS (p.L19F and p.T58I), EGR2 (p.H384N and p.D411Y), and SETD2 (p.G889V), the latter associated with a chemoresistant phenotype in acute leukemias (14). The BTK p.E88K mutation, present at the subclonal level in the primary tumor, was lost in the PDX, which, however, acquired a second BTK mutation in the same PH domain (p.E96G; Fig. 2A; Supplementary Fig. S5; Supplementary Table S3). Besides confirming these data, WES identified additional variants of potential interest, including in genes previously reported as mutated in CLL, such as IRF4 and EGFR (Supplementary Table S4; refs. 15, 16). mRNA sequencing results showed that the primary patient sample and the PDX share 88% of the transcriptome (Fig. 2A), confirming the stability of the model. Pathway analysis of differentially modulated genes highlighted the upregulation of a proliferation signature after the mouse passage, along with increased activation of NOTCH1 pathway. Consistently, Western blot analysis and RT-PCR of key target genes highlighted activation of BCR, NFκB and NOTCH signaling pathways (Fig. 3A–C; Supplementary Fig. S6A–S6C).

For case RS1316, targeted sequencing was performed on primary and on cells from P1, P2, P5, P7, and P8, while WES on cells from the primary, P1 and P2. Karyotype analyses revealed that both primary and PDX were characterized by trisomy 12 among other abnormalities, but the PDX acquired deletion of 6p22-p25 and gain of 8q22-qter (Table 1). Gains of 8q22 have been associated with chemoresistance and metastasis in breast cancer, possibly through activation of MTDH (17). Putative driver mutations shared by the primary and PDX cells included KRAS (p.G13C) and NOTCH2 (p.N1516S). Interestingly, the primary tumor showed subclonal NOTCH1 mutation (p.2514fs4), lost in the PDX, which, however, became clonal for a MED12 (p.G44R) mutation, which could impact on the NOTCH pathway and has recently found to be mutually exclusive with NOTCH1 mutations (Fig. 2B; Supplementary Fig. S5; Table S3; ref. 18). Primary RS1316 cells contained the BTK mutation associated to ibrutinib resistance (p.C481S), which became clonal in the PDX. The subclonal TP53 mutation was lost in the PDX, while the TBL1XR1 p.H127P mutation found in approximately 1% of primary cells became near clonal in the PDX (Fig. 2B; Supplementary Table S3). WES identified several additional mutations, including those in MYC, PIK3C2G, AKT3, which were stable in the primary versus PDX, highlighting the importance of the B-cell receptor (BCR) pathway (Supplementary Table S5). mRNAseq showed approximately 80% of concordance between primary and PDX (Fig. 2B). After passages through the mouse, the PDX appeared to overexpress genes linked to proliferation and metabolism, with a concomitant decrease in expression of genes controlling apoptosis and immune responses and genes negatively regulating NFκB.

In line with these data, pathway analysis showed increased constitutive activation of the NFκB pathway, with low activation of the BCR and NOTCH pathways (Fig. 3B and C; Supplementary Fig. S6B–S6D).

Recent experience with different cancer models has shown that PDX can be highly informative models to study genomic architecture of the disease and hierarchy of mutations. Furthermore, PDX can be successfully exploited to build preclinical platforms for target identification and validation of novel therapies or therapeutic combinations (19). In the case of Richter syndrome this is even more important, as there are no available cell lines or animal models. In addition, as Richter syndrome is a relatively rare complication of CLL, often diagnosed by lymph node biopsy, there is very limited availability of viable primary samples. This work reports for the first time the establishment of two PDX models of Richter syndrome: importantly these models maintain the defining features of the original tumor, including malignant phenotypes, genomic architecture and biomolecular signatures. Being highly stable and relatively easy to use and expand, RS9737 and RS1316 offer the unique opportunity to (i) gain relevant insights into the molecular and genetic drivers of Richter syndrome pathogenesis, (ii) uncover potential actionable candidates, and (iii) explore the efficacy of novel therapies.

At present, functional studies can be performed using viable cells freshly purified from the animals. However, experiments are ongoing to adapt these cells for in vitro growth, either alone or with supportive cells from the microenvironment, as well as to genetically modify them with the CRISPR/Cas9 technology, to address critical functional genomic issues. These studies will be instrumental to understand the function of distinct genes in transformation and progression of Richter syndrome. Our initial characterization shows constitutive activation of different pathways, such as BCR, NOTCH1, and NFκB, all critically involved in the control of CLL biology and, presumably, in Richter syndrome. We also highlight the presence of mutations in selected genes, most of them involved in cell-cycle control, proliferation, chromatin remodeling and epigenetic modification, suggesting that they may have a role in the pathogenetic network of the disease (Supplementary Fig. S7). The genetic architecture of both clones was highly complex with clonal mutations in oncogenes that could potentially be drivers of the transformation, such as EGR2, KRAS, SETD2, and MED12. It is interesting to note that both PDX samples were stabilized from patients that had been treated with ibrutinib and harbored BTK mutations, even if in different domains. While this data does not clarify the role of this kinase in transformation (20), their expansion in the PDX suggests that they conferred some advantage to the neoplastic cell. An alternative hypothesis is that the engrafted clone is actually the result of a two-tiered selection process: the first operated by ibrutinib in the patient, the second operated by the tumor microenvironment in the mouse. However, it is difficult to address these hypotheses due to the lack of primary cells before ibrutinib treatment.

RNA sequencing data offered a unique view of the transcriptome of both models and the possibility to compare expression patterns in cells from the primary and PDX. A common conclusion is that, while expression levels of great majority of genes were conserved, proliferation genes were upregulated in the PDX, along with pathways modulated by the environment, such as NOTCH1. This observation suggests that loss of the human environment with the concomitant presence of a severely immunodeficient host may accelerate growth of the neoplastic cell, offering cross-reactivity in major receptor–ligand systems.

Clinical and experimental evidence indicate that Richter syndrome cells are poorly responsive to drugs that are typically highly effective for patients with CLL, such as Ibrutinib or idelalisib (21), suggesting that novel therapeutic opportunities should be explored on the basis of the genetic and expression profile. In line with this concept, preliminary results obtained with a selective NFκB inhibitor, IT-901, confirm that these cells are, to a considerable extent, sensitive to this drug, which interrupts a critical growth circuit (22).

Finally, additional PDX stabilization is ongoing with the aim of generating a panel of different Richter syndrome models to be used for therapeutic development and proof-of-concept validation of effective strategies.

No potential conflicts of interest were disclosed.

Conception and design:T. Vaisitti, E. Braggio, J.N. Allan, F. Arruga, S. Serra, W. Tam, R.R. Furman, S. Deaglio

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.):T. Vaisitti, A. Zamò, W. Tam, A. Chadburn, R.R. Furman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis):T. Vaisitti, E. Braggio, J.N. Allan, W. Tam, R.R. Furman, S. Deaglio

Writing, review, and/or revision of the manuscript:T. Vaisitti, E. Braggio, J.N. Allan, W. Tam, A. Chadburn, R.R. Furman, S. Deaglio

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):W. Tam, A. Chadburn, R.R. Furman, S. Deaglio

Study supervision:J.N. Allan, S. Deaglio

Other (performed experiments):T. Vaisitti

Other (contributed to perform experiments):F. Arruga

The authors thank Katiuscia Gizzi (Italian Institute for Genomic Medicine) for excellent technical support. This work was supported by the Italian Institute for Genomic Medicine Institutional Funds (to S. Deaglio and T. Vaisitti), by the Associazione Italiana per la Ricerca sul Cancro AIRC (IG-17314 to S. Deaglio), by the Italian Ministry of Health (GR-2011-02346826 to S. Deaglio and GR-2011-02349282 to T. Vaisitti), and by the Ministry of Education, University and Research – MIUR project “Dipartimenti di Eccellenza 2018 – 2022” (to S. Deaglio on behalf of Department of Medical Sciences).