CRISPR genome editing of murine hematopoietic stem cells to create Npm1-Alk causes ALK⁺ lymphoma after transplantation

Fusion oncogenes from recurrent chromosomal rearrangements drive many hematologic malignancies.¹  Traditional modeling of these lesions in mice involves retroviral introduction of human fusion oncogenes to murine hematopoietic progenitors and subsequent syngeneic transplantation. Alternatively, the oncogenes can be engineered into the germline under lineage-specific promoters. For example, introduction of BCR-ABL,^2,3  derived from chronic myeloid leukemia (CML), to progenitors causes a fatal myeloproliferative disease similar to accelerated-phase CML but does not capture the prolonged chronic phase of human CML.⁴ NPM1-ALK, derived from anaplastic lymphoma kinase (ALK)⁺ anaplastic large-cell lymphoma (ALCL) is another classic fusion that drove breakthrough animal modeling for T-cell lymphomas.⁵  First described in 1985, ALK⁺ ALCL is a T-cell non-Hodgkin lymphoma constituting ∼2% of lymphoma diagnoses overall; it is seen most commonly in adolescents and young adults.⁶  It is the most frequent pediatric mature T-cell malignancy. Pathologically, ALK⁺ ALCL cells have an irregular blastic morphology, including large nuclei and scant cytoplasm. The disease is diagnosed by expression of the characteristic ALK protein and CD30, but T-cell surface markers are variably lost, leaving many cases with a null-cell phenotype.^7-9  Previous use of NPM1-ALK in murine lymphoma models demonstrated the gene’s potent transforming properties but did not fully recapitulate the phenotype of human ALK⁺ ALCL. Retroviral infection of bone marrow progenitor cells with NPM1-ALK led to B-cell lymphomas after transplantation to recipients.¹⁰  Transgenic engineering of NPM1-ALK into the murine genome under Vav or CD2 promoters also resulted in B lymphomas.^11,12  Chiarle et al achieved a T-cell phenotype by placing NPM1-ALK under a CD4 promoter, but animals developed immature T-cell lymphomas more similar to lymphoblastic lymphoma, and leaky expression drove B-lineage plasma-cell neoplasms in about a third.¹³  CRISPR/Cas9 genome editing recently resulted in breakthrough modeling of ALK⁺ lung cancer through generation of Eml4-Alk, homologous to the human fusion oncogene, through introduction of genomic breakpoints.¹⁴  Here, we successfully adapted these techniques to modeling hematologic malignancy for the first time by generating Npm1-Alk in the genome of transplantable hematopoietic stem cells (HSCs).

Additional detailed information is provided in supplemental Methods.

HSCs were isolated from fetal livers at embryonic day 14.5 and grown in RPMI 1640 supplemented with 10% fetal bovine serum, 10% WEHI supernatant, interleukin-3 (IL-3) (10 μg/mL), IL-6 (10 μg/mL), and stem cell factor (10 μg/mL). To prevent differentiation, HSCs were maintained in culture for <72 hours. Splenocytes were cultured in RPMI 1640 supplemented with 20% fetal bovine serum.

Individual fetal livers from wild-type C57BL/6 mice with 2 × 10⁵ HSCs each were divided to undergo CRISPR vs mock transfection (Figure 1B). We used Lipofectamine 3000 transfection reagent, with or without 3 μg pX330, with single guide RNAs targeting Npm1 and Alk (sgNA; Figure 1A)¹⁴  over 2 days. After overnight incubation in fresh media, transfected HSCs were injected via the tail vein into sublethally irradiated (4.5 G × 2, separated by 2 hours) syngeneic wild-type C57BL/6 recipients. All primary recipients were 9 weeks old at the time of transplantation. For secondary recipients, 10⁶ splenocytes from the primary recipients were transplanted into irradiated wild-type C57BL/6 mice via the tail vein. These mice were also 9 weeks old at the time of transplantation.

Formalin-fixed paraffin-embedded tissue sections, produced per standard protocols, were used to make H&E-stained and immunohistochemistry (IHC) pathology slides. Primary IHC antibodies were anti-ALK (Abcam, ab180607), anti-CD30 (Invitrogen, MA512532), anti-CD3 (Abcam, ab16669), anti-CD4 (Invitrogen, 14-9766-82), and anti-CD8a (Invitrogen, 14-0195-80). After secondary antibody and counterstaining, all slides were subject to expert hematopathology review (F.V.).

Splenocytes were collected, passaged transiently in vitro, and stained with the following antibodies: B220 (BD Bioscience, 552772), IgM (BD Bioscience, 553437), and Thy1.2 (eBioscience, 12-0902-82). Viable cells were gated through phycoerythrin staining, and stained samples were run on the Attune NxT analyzer. Data were then analyzed using FlowJo software (TreeStar, Inc.).

Amplification and sequencing of (TCRB/IGH/IGKL/TCRAD/TCRG) CDR3 were performed with immunoSEQ (Adaptive Biotechnologies, Seattle, WA). The clonal repertoire is determined using the Shannon entropy-based clonality index, which ranges from 0 to 1, with index values < 0.5 being polyclonal.^13,

The Npm1 and Alk genes on murine chromosomes 11 and 17, respectively, contain intronic breakpoints that can create an in-frame fusion Npm1-Alk oncogene closely analogous to human NPM1-ALK (Figure 1A; supplemental Figure 1A).¹⁴  We transfected 3T3 cells with sgNA vector encoding Cas9 and single guide RNAs targeting these breakpoints and confirmed generation of Npm1-Alk (supplemental Figure 1B-C), including sequence confirmation of an expected 1.5-kb Npm1-Alk PCR band (supplemental Figure 1D). CRISPR-generated Npm1-Alk transforms IL-3–dependent Ba/F3 cells^15,16  but has not previously been compared with human NPM1-ALK. We infected the cytokine-dependent murine pro-B cell line FL5.12 with vector, Npm1-Alk, or NPM1-ALK with GFP coexpression and cycled them through IL-3 withdrawal and rescue, resulting in Npm1-Alk enrichment similar to NPM1-ALK (supplemental Figure 1E). Both the human and mouse alleles transformed the FL5.12 cells to proliferation in cytokine-free media. Npm1-Alk–transformed cells are highly sensitive to clinical ALK kinase inhibitors (supplemental Figure 1F-G), but differences in the ALK kinase domains resulted in differential sensitivity patterns, including relative crizotinib and ceritinib sensitivity but alectinib resistance.

We next divided C57BL/6 murine HSCs from a single fetal liver for mock vs sgNA transfection and confirmed detection of Npm1-Alk in the sgNA-transfected cells (supplemental Figure 2A-B), including Sanger sequencing confirmation of the fusion gene from the polymerase chain reaction (PCR) product (data not shown; same result as supplemental Figure 1D). We repeated paired sgNA/mock transfections of HSCs isolated from individual fetal livers and transplanted them into wild-type C57BL/6 recipients subjected to sublethal irradiation (Figure 1B). No recipients died during the initial 3-4 weeks following transplantation, indicating successful engraftment of transplanted HSCs. All were unremarkable until posttransplant day 270, when an animal transplanted with sgNA-modified HSCs (sgNA-1) became moribund and was euthanized in parallel with its paired mock control. Pathologic analysis showed that, although the mouse transplanted with mock-transfected HSCs was phenotypically normal, the mouse transplanted with CRISPR-transfected HSCs had lymphoma completely effacing the normal architecture of the spleen (Figure 1C). In addition, lymphoma cells invaded multiple organs of the CRISPR-modified HSC recipient, including lung, liver (Figure 1D), and kidney (data not shown). Thereafter, the remaining mice were euthanized, revealing lymphoma in 3 of 4 mice (overall) transplanted with sgNA-modified HSCs (supplemental Figure 2C-D). Like sgNA-1, mouse sgNA-2 had lymphoma involvement of nonhematologic organs (lung and liver, supplemental Figure 2D), whereas disease in mouse sgNA-4 was confined morphologically to the spleen. An additional mouse transplanted with CRISPR-transfected HSCs (sgNA-3) never became moribund and was phenotypically normal when sacrificed a year after transplantation (data not shown). All of the mock-transfected paired control animals were phenotypically normal (Figure 1C-D; others not shown). Fluorescence-activated cell sorting analysis of splenocytes from diseased and control animals showed an increased presence of T-lineage lymphocytes based on Thy1.2 cell surface expression in animals sgNA-1 and sgNA-4, whereas sgNA-2 had decreased numbers of T cells (supplemental Figure 2E). This mouse instead had a prominent population of B220⁺ IgM⁻ cells, consistent with a B-cell lymphoma phenotype in this animal. We next assessed expression of ALK and CD30, pathologic hallmarks of human ALK⁺ ALCL. Quantitative reverse-transcription PCR (qRT-PCR) was used to assess expression of the murine Alk kinase domain in diseased recipients, normalized to corresponding tissues from each animal's paired mock control. The lymphoma-bearing animals had 10× to 10 000× expression in bone marrow, spleen, and other affected organs (Figure 1E). Expression of Tnfrsf8, which encodes murine CD30, was similarly elevated in sgNA-1 and sgNA-4 but not in sgNA-2. Therefore, we carried out IHC evaluation for these markers, which showed, consistent with the quantitative PCR (qPCR) results, high ALK protein in the spleens of all of the diseased animals, including the nuclear staining pattern that is distinctive of human NPM1-ALK (Figure 1F; micrographs taken with 60× objective, 100× objective insets for ALK; full-size 100× micrographs are provided as supplemental Figure 2F).^17,18  sgNA-1 and sgNA-4 showed strong cell surface staining for CD30, whereas sgNA-2, like human ALK⁺ large B-cell lymphoma, was CD30⁻.^19,20  We used reverse-transcription PCR primers spanning the Npm1-Alk fusion to confirm its detection in tumor-affected tissues from transplanted mice (supplemental Figure 3A). Generation of Npm1-Alk in HSCs leads to ALK⁺ lymphomas in transplant recipients after a long latency.

We further characterized the surface phenotype of these lymphomas, focusing on the T-cell markers CD3, CD4, and CD8, which are highly variable in human ALK⁺ ALCL; they are commonly absent, leading to a null-cell phenotype that requires T-cell receptor (TCR) clonality studies to confirm T-cell origin.^21,22  Both animals with suspected T-cell lymphomas, sgNA-1 and sgNA-4, had scattered positivity for CD3, and sgNA-1 also had patches in the spleen that were strongly positive for CD4 and CD8 (Figure 1G; high- and low-magnification views in supplemental Figure 4). sgNA-4 was negative for CD4 and CD8, whereas sgNA-2 was negative for all 3 T-cell markers, as expected. To further clarify the disease phenotype of sgNA-1, we assessed CD4 and CD8 in tumors invasive to other organs, which occurred in this animal but not in sgNA-4. The perivascular lymphomas that arose in the lung of this animal were strongly CD4⁺ and CD8⁺ (supplemental Figure 3B, see “Discussion”). In sum, pathologic analyses show variations in morphology and cell surface markers, potentially consistent with multiple lymphomas, having arisen simultaneously, which could be assessed through TCR clonality studies.

We next tested transplantability of primary tumors to secondary recipients through tail vein injection of 10⁶ splenocytes from primary recipients to wild-type mice after 4-Gy irradiation. Available cell numbers permitted 12 secondary transplantations, 5 each from sgNA-1 and sgNA-4 (T-cell lymphomas) and 2 from sgNA-2 (B cell). Disease onset in all secondary recipients is grouped in Figure 2A, showing lymphomas in 9 of 12 mice beginning by day 180. When the first mouse became moribund, we analyzed peripheral blood from all secondary recipients for expression of Alk and Tnfrsf8 using qPCR (supplemental Figure 5A). Seven of 8 animals that had elevated ALK expression at this time point subsequently developed disease (red boxes). Pathologic assessment showed lymphomas similar to primary recipients infiltrating spleen and other organs (Figure 2B, and data not shown), including 1 mouse with nodal lymphoma (supplemental Figure 5B). Affected animals had enlarged livers and spleens compared with controls sacrificed in parallel (supplemental Figure 5C). Tumors in T-lymphoma recipients showed ALK, CD30, and CD3 IHC staining similar to primary recipients (Figure 2C; supplemental Figure 3C). Diseased secondary recipients transplanted with T lymphomas, like primary recipients, had overexpression of Alk and Tnfrsf8, whereas only Alk was elevated in a diseased secondary recipient of tumor cells from sgNA-2 (Figure 2D). Npm1-Alk was confirmed by reverse-transcription PCR in transplanted mice (supplemental Figure 5D). Protein expression of ALK and its established downstream effectors were also seen in transplanted mice but not wild-type splenocytes via western blotting performed on unselected splenocytes (supplemental Figure 5E). These cells had undergone cryopreservation before this analysis was performed and showed a banding pattern in response to a murine-specific ALK antibody, consistent with partial degradation of NPM1-ALK, as reported previously for the human protein.²³  A high-molecular-weight ALK band also was detected, consistent with perinuclear aggregate NPM1-ALK, which is well described.^24,25  The wild-type control splenocytes did not show any reactivity to the murine ALK antibody. Tumor cells from spleens isolated from sgNA4-R2, sgNA1-00, sgNA1-L1, and sgNA4-L2 proliferated transiently and were highly sensitive to ALK inhibitors, with the sensitivity pattern mirroring Npm1-Alk–transformed FL5.12 cells (Figure 2E; supplemental Figure 5F).

We next assessed TCR clonality of the primary recipients that developed T-cell lymphomas and 2 secondary recipients transplanted with tumor cells from each of these primaries using GeneScan.^26,27  The diversity of β-chain rearrangements is plotted as pie charts (Figure 2F; supplemental Figure 5G). Primary recipients were polyclonal but had significantly decreased clonal diversity compared with mock controls. Additionally, clones identical to the primary recipients were found when the analysis was performed on splenocytes from diseased secondary recipients. Therefore, these lymphomas derive from postthymic T cells but were not monoclonal, consistent with multiple separate transformations having occurred during disease latency in T cells containing the Npm1-Alk translocation.

In sum, we demonstrate transplantable ALK-driven lymphomas resulting from creation of Npm1-Alk by CRISPR genome editing of murine HSCs. The long latency is consistent with natural progression to disease onset after introduction of a single genetic lesion formed from endogenous murine loci. Npm1-Alk⁺ transplanted HSCs could only have contained 1 copy of the fusion because Npm1 is an essential gene, and its other copy must remain intact.^18,28  Therefore, these cells accurately represent the initiating genotype of human ALK⁺ ALCL and resulted in a phenotype with similarities to and differences from human disease. Specifically, T-cell lymphomas had a large-cell morphology, had strong expression of cell surface CD30, and showed TCR rearrangements indicating progression through key steps of thymic maturation. Tumors were oligoclonal compared with mock controls but not monoclonal like human ALCL. Different clones that arose appeared to have various cell surface phenotypes, with differential expression of CD3, CD4, and CD8 among tumor cells, including within the same diseased spleen, illustrated especially by the primary recipient sgNA-1. Tumors invading other organs in this animal were CD4⁺ CD8⁺, a phenotype that is rarely seen in human ALK ALCL, in which expression of CD4 alone or neither marker are most common.^9,29  Animals did not develop a primarily nodal lymphoma as is seen in most human patients; instead, disease involved the spleen, with variable spread to liver, lung, and kidney.

Regardless, the proof-of-principle established here is widely adaptable to additional hematologic malignancies and may help to generate preclinical models with increased fidelity. Piganeau et al used a transcription activator-like effector nuclease approach to generate t(2;5)(p23;q35)/NPM1-ALK in cultured human cells through generation of double-stranded DNA breaks, but this approach was not used to model the disease.³⁰  Utility of our model in preclinical studies is limited by long disease latency in secondary recipients (minimum 180 days) and the phenotypic variability outlined above. Serial passaging of these tumors through additional recipients may or may not reveal isolated clones that result in tumor phenotypes with high fidelity for human ALK⁺ ALCL to fuel preclinical treatment experiments. Tumor cells from primary and secondary recipients proliferated transiently in vitro in media not supplemented with cytokines, but they did not become long-term cell lines. This also could change after passages through additional recipients moving forward or with the use of T-cell stimulatory cytokines.

Early animal-modeling efforts with NPM1-ALK (discussed in “Introduction”) left the cell of origin for ALK⁺ ALCL unclear, and it still has not been definitively revealed. Recently, Malcolm et al shed key light on this question, showing that NPM1-ALK⁺ thymic precursors use the fusion oncogene to bypass the β-selection checkpoint before moving to the periphery to establish systemic lymphomas.³¹  By engineering the endogenous translocation into a subset of HSCs, our approach may have allowed recapitulation of this process, as opposed to earlier transgenic approaches placing NPM1-ALK expression under the control of specific T-lineage promoters. Additional evidence also exists for NPM1-ALK arising in early stem or progenitor cells. Trümper et al showed in 1998 that 14 of 29 healthy individuals (48%) had NPM1-ALK detectable by high-sensitivity PCR in peripheral blood,³²  and even umbilical cord blood from healthy newborns showed evidence of the fusion in ∼2%³³  (although this was not linked to later onset of ALK⁺ ALCL in either study). The polyclonal phenotype that we observed is consistent with multiple individual clones having undergone transformation separately in the mice that developed T lymphomas. Therefore, these primary diseases were molecularly more heterogeneous than human ALK⁺ ALCL.

The authors thank the Sheila and David Fuente Graduate Program in Cancer Biology at the University of Miami and are extremely grateful to Andrea Ventura (Memorial Sloan Kettering Cancer Center, New York, NY) and Danilo Maddalo (Novartis) for providing the sgNpm-Alk plasmid. The authors thank Mingjiang Xu (University of Miami) for developing the pathology core and helping with the slides for H&E staining and IHC, and as well as the laboratory of Feng-Chun Yang (University of Miami) for extensive use of the microscope camera.

This work was supported by National Institutes of Health, National Cancer Institute grant 1R01CA190696-01 (J.H.S.).

S.S.R. was a PhD candidate at the University of Miami, and this work is submitted in partial fulfillment of the requirement for a PhD.

Contribution: S.S.R. and A.D.A. performed qPCR and clonality analysis; L.L. transplanted mice and performed fluorescence-activated cell sorting and IHC; S.S.R., M.F.K., K.K., and N.K.A. performed IHC staining and analyzed the results; F.V. analyzed the IHC and pathology slides; S.S.R. and J.H.S. designed the experiments, analyzed the results, and wrote the manuscript; and J.H.S. provided funding for this study.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jonathan H. Schatz, Batchelor Children’s Research Institute, University of Miami, 1580 NW 10th Ave, Room 419, Miami, FL 33137; e-mail: jschatz@med.miami.edu.