Whole Genome Sequencing of Spontaneously Occurring Rat Natural Killer Large Granular Lymphocyte Leukemia Identifies JAK1 Somatic Activating Mutation

Large granular lymphocyte (LGL) leukemia arises spontaneously in elderly Fischer (F344) rats. This rodent model has been shown to emulate many aspects of the natural killer (NK) variant of human LGL leukemia. Previous transplantation of leukemic material into young F344 rats resulted in several strains of rat NK (RNK) primary leukemic cells. One strain, RNK-16, was adapted into the RNK-16 cell line and established as an aggressive NK-LGL leukemia model. Whole genome sequencing of the RNK-16 cell line identified 255,838 locations where the RNK16 had an alternate allele that was different from F344, including a mutation in Jak1. Functional studies showed Jak1 Y1034C to be a somatic activating mutation that mediated increased STAT signaling, as assessed by phosphoprotein levels. Sanger sequencing of Jak1 in RNK-1, -3, -7, and -16 found only RNK-16 to harbor the Y1034C Jak1 mutation. In vivo studies revealed that rats engrafted with RNK-16 primary material developed leukemia more rapidly than those engrafted with RNK-1, -3, and -7. Additionally, ex vivo RNK-16 spleen cells from leukemic rats exhibited increased STAT1, STAT3, and STAT5 phosphorylation compared to other RNK strains. Therefore, we report and characterize a novel gain-of-function Jak1 mutation in a spontaneous LGL leukemia model that results in increased downstream STAT signaling.


Introduction
Large granular lymphocyte (LGL) leukemia is a lymphoproliferative neoplasm characterized by uncontrolled clonal expansion of LGLs of T-cell or natural killer (NK) cell origin, which causes anemia, neutropenia, and splenomegaly [1]. The World Health Organization classified LGL leukemia into three subtypes: T-cell LGL leukemia, chronic lymphoproliferative disorder of NK cells, and aggressive NK-cell leukemia [2]. The aggressive NK-cell leukemia is distinct from its indolent counterparts, affecting younger individuals with a higher prevalence in Asia and South America [3,4]. While the etiology of LGL leukemia is not well understood, LGL leukemia is characterized by multiple aberrant signaling pathways, including the Janus kinase/signal transducers and activators of the transcription (JAK/STAT) signaling pathway [1,5].
The JAK/STAT pathway relays extracellular signals from cytokines and other factors to intracellular factors to control nuclear transcription. This pathway is essential for various intracellular processes, including apoptosis, immune response, and differentiation. JAKs are non-receptor tyrosine kinases that interact with the intracellular tail of membrane-bound receptor complexes. JAK proteins consist of four domains; an N-terminal FERM (F for 4.1 protein, E for ezrin, R for radixin and M for moesin) domain, an SH2 domain, a pseudokinase domain, and a C-terminal protein tyrosine kinase (PTK) domain. Within the PTK domain lies an activation loop housing two highly conserved adjacent tyrosine residues. JAK signaling is well studied, and mutations in the JAK family are clearly implicated in the pathogenesis of hematological disorders [6,7]. JAK1 mutations have been reported in various hematological malignancies, including 7-27% of T-cell acute lymphoblastic leukemia (ALL) [8][9][10][11][12], 8-12% of T-cell prolymphocytic leukemia [13,14], 1.5% of B-ALL [9], and a small percentage of acute myeloid leukemia [15]. Subsequently, JAK1 mutations have been linked to solid tumor cancers, including Hepatitis B-associated hepatocellular carcinoma [16] and gynecological tumors [17].
Fischer 344 rats are an inbred strain of rats that were used for decades as a primary in vivo model for cancer and toxicology studies [24]. Large granular lymphocyte (LGL) leukemia spontaneously occurs in 30-50% of Fischer 344 rats over 18 months of age [25][26][27]. Presentation of the disease in rats include anemia, splenomegaly, and leukemic cell infiltration of the spleen, liver, and lungs. [28]. Morphologically, these leukemic cells resemble LGLs with cytoplasmic granules. Functional studies using cytotoxicity and cell marker assays characterized these leukemic cells as being of NK cell origin and labelled as rat natural killer (RNK) cells [29]. Transplantation of primary RNK material induces leukemia, and has been utilized to establish 15 RNK strains. Of all strains, RNK-16 was characterized as a highly cytotoxic variant [29,30] and adapted into the RNK-16 cell line [31,32]. While high prevalence of disease and slight variations in features differ from the human counterpart, the morphological, functional, and clinical similarities established RNK-16 as a model of the human aggressive form of natural killer LGL leukemia (NK-LGL leukemia) [28,33,34].
The goal of the present study was to further characterize the RNK-16 cell line through whole genome sequencing (WGS). The results will aid interpretation of future LGL leukemia therapy response studies by our lab and others who use this model. Whole genome sequences from the RNK-16 cell line were compared to the F344 genome to reveal variants in several oncogenes. This led to the discovery of a novel Jak1 Y1034C mutation, which was also validated in ex vivo primary spleen material from leukemic rats. This allowed us to identify the mutated JAK1 as the putative leukemic driver JAK1, as well as mutations that occurred during adaptation to culture. Functional assays showed that Jak1 Y1034C increased downstream STAT1, STAT3, and STAT5 phosphorylation in both the cell line and the primary RNK material. Characterization of this mutation may have a larger impact in future studies as more JAK1 mutations are uncovered in other diseases.

Identification and Validation of Mutations in the RNK-16 Cell Line and RNK-16 Primary Spleen Material
Whole genome sequencing (WGS) was successfully performed on the RNK-16 cell line in an effort to identify variants that contribute to NK cell leukemogenesis ( Figure 1).

Figure 1.
Diagram of rat natural killer (RNK) material. Aging F344 rats naturally develop LGL leukemia. Leukemic cells were transplanted back into young F344 rats to produce several strains of RNK [27]. RNK-16 cells were adapted into a cell line. Transplantation of RNK-1, -3, -7, and -16 into young F344 rats generated ex vivo spleen samples, and those samples in bold were used in these studies.
After filtering for quality control parameters and removing variants also seen in F344 genome, we were left with 255,838 variants (Table S1). Variants were annotated for their impact on the canonical transcripts using SnpEff. Restriction of genes with variants of a Phred quality score of over 20 and labelled as high or moderate impact resulted in 498 genes. Using the Homologene database, these 498 genes were mapped to 457 human genes, of which 157 did not have the same names (Table  S2). Those with matched gene names were queried in the COSMIC Cancer Gene Census, which classifies genes into two tiers based on the extent of evidence that variants in those genes drive oncogenic transformation [35]. Eight genes with mutations in RNK-16 were labelled as having oncogenic transformative potential, including Brd3, Ddr2, Fat1, Jak1, Ncor2 Nono, Stag, and Thrap3 (Table 1). Table 1. Cancer Gene Census (CGC) query revealed genes mutated in the RNK-16 cell line that are implicated in cancer. Tier levels are assigned by CGC and indicate a depth of evidence of mutations in those genes having oncogenic transformative potential. Variants in Ddr2, Fat1, Jak1, Ncor2, and Stag were confirmed using Sanger sequencing in the RNK-16 cell line. Only the mutation in Jak1 was confirmed in the RNK-16 primary spleen cells. Leukemic cells were transplanted back into young F344 rats to produce several strains of RNK [27]. RNK-16 cells were adapted into a cell line. Transplantation of RNK-1, -3, -7, and -16 into young F344 rats generated ex vivo spleen samples, and those samples in bold were used in these studies.

RNK-16 Primary Spleen Cells
After filtering for quality control parameters and removing variants also seen in F344 genome, we were left with 255,838 variants (Table S1). Variants were annotated for their impact on the canonical transcripts using SnpEff. Restriction of genes with variants of a Phred quality score of over 20 and labelled as high or moderate impact resulted in 498 genes. Using the Homologene database, these 498 genes were mapped to 457 human genes, of which 157 did not have the same names (Table S2). Those with matched gene names were queried in the COSMIC Cancer Gene Census, which classifies genes into two tiers based on the extent of evidence that variants in those genes drive oncogenic transformation [35]. Eight genes with mutations in RNK-16 were labelled as having oncogenic transformative potential, including Brd3, Ddr2, Fat1, Jak1, Ncor2 Nono, Stag, and Thrap3 (Table 1). Table 1. Cancer Gene Census (CGC) query revealed genes mutated in the RNK-16 cell line that are implicated in cancer. Tier levels are assigned by CGC and indicate a depth of evidence of mutations in those genes having oncogenic transformative potential. Variants in Ddr2, Fat1, Jak1, Ncor2, and Stag were confirmed using Sanger sequencing in the RNK-16 cell line. Only the mutation in Jak1 was confirmed in the RNK-16 primary spleen cells.

RNK-16 Primary Spleen Cells
Brd3 Sequencing strategies were designed to validate the variants by Sanger sequencing in the RNK-16 cell line. Out of the eight genes, variants were confirmed in five, including Ddr2, Fat1, Jak1, Ncor2, and Stag. As the RNK-16 cell line was adapted to an in vitro culture from ex vivo RNK-16 material, validation of these mutations was necessary to demonstrate that mutations were not the result of cell line transformation. Sanger sequencing of ex vivo RNK-16 spleen cells (Figure 1) confirmed only the Jak1 variant (Figure 2A), indicating that the adaptation of RNK-16 primary cells into the RNK-16 cell line may have resulted in the gain of mutations in Ddr2, Fat1, Ncor2, and Stag, but not Jak1. JAK1 is a tyrosine kinase involved in cytokine signaling that results in the downstream activation of STATs. This Jak1 mutation is found in the PTK domain and is located within the activation loop that contains two tyrosine phosphorylation sites, 1033 and 1034 ( Figure 2B). Results of WGS, as well as Sanger validation of the cell line and primary splenocyte material, are summarized in Table 1. Sequencing strategies were designed to validate the variants by Sanger sequencing in the RNK-16 cell line. Out of the eight genes, variants were confirmed in five, including Ddr2, Fat1, Jak1, Ncor2, and Stag. As the RNK-16 cell line was adapted to an in vitro culture from ex vivo RNK-16 material, validation of these mutations was necessary to demonstrate that mutations were not the result of cell line transformation. Sanger sequencing of ex vivo RNK-16 spleen cells ( Figure 1) confirmed only the Jak1 variant (Figure 2A), indicating that the adaptation of RNK-16 primary cells into the RNK-16 cell line may have resulted in the gain of mutations in Ddr2, Fat1, Ncor2, and Stag, but not Jak1. JAK1 is a tyrosine kinase involved in cytokine signaling that results in the downstream activation of STATs. This Jak1 mutation is found in the PTK domain and is located within the activation loop that contains two tyrosine phosphorylation sites, 1033 and 1034 ( Figure 2B). Results of WGS, as well as Sanger validation of the cell line and primary splenocyte material, are summarized in Table 1.

JAK1 Mutation Increased Downstream STAT Signaling
To examine the effects of Y1034C Jak1 mutation on protein function, we transiently expressed wild-type (WT) Jak1 or Jak1 Y1034C in HEK293-FT cells. Western blotting demonstrated constitutive phosphorylation of downstream STAT targets, including STAT1, STAT3, and STAT5 ( Figure 3). The epitope of the phosphorylated JAK1 antibody recognizes the dual tyrosine phosphorylation in the activation loop (1033/1034), which is altered in the mutant; therefore, it was unable to recognize the mutant phosphorylated protein, even though mutant overexpression was confirmed by the total JAK1 antibody. The total JAK1 protein levels were unchanged, indicating that mutation did not increase production of total JAK1. Low levels of all phosphoprotein targets were observed when transfected with WT Jak1, drastically different from the large increase of phosphorylation with the transfection of Jak1 Y1034C. Thus, insertion of Jak1 Y1034C induced the constitutive activation of STAT signaling, independent of exogenous cytokine signaling.

JAK1 Mutation Increased Downstream STAT Signaling
To examine the effects of Y1034C Jak1 mutation on protein function, we transiently expressed wild-type (WT) Jak1 or Jak1 Y1034C in HEK293-FT cells. Western blotting demonstrated constitutive phosphorylation of downstream STAT targets, including STAT1, STAT3, and STAT5 ( Figure 3). The epitope of the phosphorylated JAK1 antibody recognizes the dual tyrosine phosphorylation in the activation loop (1033/1034), which is altered in the mutant; therefore, it was unable to recognize the mutant phosphorylated protein, even though mutant overexpression was confirmed by the total JAK1 antibody. The total JAK1 protein levels were unchanged, indicating that mutation did not increase production of total JAK1. Low levels of all phosphoprotein targets were observed when transfected with WT Jak1, drastically different from the large increase of phosphorylation with the transfection of Jak1 Y1034C. Thus, insertion of Jak1 Y1034C induced the constitutive activation of STAT signaling, independent of exogenous cytokine signaling.

In Vivo and Ex Vivo Characterization of Primary Rat Natural Killer Material Shows an Aggressive Disease Course and Increased STAT Signaling With Mutant JAK1
Previous characterization of primary ex vivo RNK material reported RNK-16 to have high cytotoxicity compared to RNK-1 and RNK-7 [29,30]. Characterization of RNK-3 was not previously published, although all of the RNK material was isolated together. Additionally, our Sanger sequencing of ex vivo RNK-1, -3, -7, and -16 spleen material detected Jak1 Y1034C only in RNK-16 ( Figure 4A). To further characterize these RNK strains, one million viable primary RNK-1, -3, -7, and -16 cells were injected intraperitoneally (i.p.) into young F344 rats to study leukemia onset of RNK-16 compared to other strains. Rats engrafted with RNK-16 showed significantly shorter time to disease onset, as determined by absolute white blood count and weight loss ( Figure 4B). The average time to disease onset for RNK-16 was 17.7 days, less than half the time for rats injected with RNK-1 to develop leukemia and less than a quarter of the time for rats transplanted with RNK-3 and RNK-7 to develop disease. Protein extracts from spleen cells harvested from leukemic rats were run on western blots and probed for downstream STAT phosphorylation ( Figure 4C). RNK-16 spleen cells had increased STAT1 and STAT3 phosphorylation relative to all other RNKs, as well as similar levels of STAT5 phosphorylation to RNK-7. Taken together, these studies provide molecular and biochemical data to complement the previously published functional characterization of RNK primary material.  Previous characterization of primary ex vivo RNK material reported RNK-16 to have high cytotoxicity compared to RNK-1 and RNK-7 [29,30]. Characterization of RNK-3 was not previously published, although all of the RNK material was isolated together. Additionally, our Sanger sequencing of ex vivo RNK-1, -3, -7, and -16 spleen material detected Jak1 Y1034C only in RNK-16 ( Figure 4A). To further characterize these RNK strains, one million viable primary RNK-1, -3, -7, and -16 cells were injected intraperitoneally (i.p.) into young F344 rats to study leukemia onset of RNK-16 compared to other strains. Rats engrafted with RNK-16 showed significantly shorter time to disease onset, as determined by absolute white blood count and weight loss ( Figure 4B). The average time to disease onset for RNK-16 was 17.7 days, less than half the time for rats injected with RNK-1 to develop leukemia and less than a quarter of the time for rats transplanted with RNK-3 and RNK-7 to develop disease. Protein extracts from spleen cells harvested from leukemic rats were run on western blots and probed for downstream STAT phosphorylation ( Figure 4C). RNK-16 spleen cells had increased STAT1 and STAT3 phosphorylation relative to all other RNKs, as well as similar levels of STAT5 phosphorylation to RNK-7. Taken together, these studies provide molecular and biochemical data to complement the previously published functional characterization of RNK primary material.

Discussion
This study discovered and characterized a Jak1 Y1034C mutation found in a naturally occurring F344 rat NK-LGL leukemia model. RNK-16, characterized as highly cytotoxic [30], was established as an in vitro cell line [32] and model for aggressive NK-LGL leukemia [34]. Whole genome sequencing in the RNK-16 cell line revealed variants in eight oncogenes. Sanger sequencing of the RNK-16 cell line and ex vivo RNK-16 spleen material confirmed a single variant, Jak1 Y1034C, in all tested material. Lack of validation of other variants found in the RNK-16 cell line in the ex vivo RNK-16 spleen material indicates that mutations may have been acquired as a result of in vitro cell culture expansion [36].
Aberrant activation of JAK/STAT signaling is a hallmark of LGL leukemia. Increased activation of this pathway results from increased cytokine signaling [4,37] and somatic activating mutations in STAT3 [19,22,38] and STAT5B [19,20,39], further implicating the oncogenic role of this pathway in LGL leukemia. WGS on a small cohort of NK-LGL leukemia patients did not reveal mutations in JAK1 (manuscript under review). While JAK1 mutations have not been reported in LGL leukemia patients, mutations in JAK1 have been identified in several hematological malignancies [9][10][11]13,15,40]. The majority of mutations occur in the pseudokinase domain, and have been characterized to have oncogenic potential [41]; patients with such mutations have poor prognoses [40].
The activation loop within the C-terminal PTK domain includes two adjacent tyrosine phosphorylation sites. Mutation of the first tyrosine to an alanine abrogated kinase activation, indicating that the first tyrosine is necessary for JAK1 activation. An alanine substitution to the second tyrosine allowed kinase activation, showing that activation is not dependent upon the second tyrosine [42]. We detected a mutation at Y1034, which is the second tyrosine within the activation loop. While not necessary for kinase activation, mutations at this location may affect enzymatic activity [43], as well as the specificity and strength of interactions with substrates, thereby leading to altered downstream signaling. Random mutagenesis experiments on human JAK1 have shown that the specific substitution of cysteine for the second tyrosine, analogous to the mutation that we detected in rat JAK1, drives clonal outgrowth and increases JAK/STAT signaling [41].

Discussion
This study discovered and characterized a Jak1 Y1034C mutation found in a naturally occurring F344 rat NK-LGL leukemia model. RNK-16, characterized as highly cytotoxic [30], was established as an in vitro cell line [32] and model for aggressive NK-LGL leukemia [34]. Whole genome sequencing in the RNK-16 cell line revealed variants in eight oncogenes. Sanger sequencing of the RNK-16 cell line and ex vivo RNK-16 spleen material confirmed a single variant, Jak1 Y1034C, in all tested material. Lack of validation of other variants found in the RNK-16 cell line in the ex vivo RNK-16 spleen material indicates that mutations may have been acquired as a result of in vitro cell culture expansion [36].
Aberrant activation of JAK/STAT signaling is a hallmark of LGL leukemia. Increased activation of this pathway results from increased cytokine signaling [4,37] and somatic activating mutations in STAT3 [19,22,38] and STAT5B [19,20,39], further implicating the oncogenic role of this pathway in LGL leukemia. WGS on a small cohort of NK-LGL leukemia patients did not reveal mutations in JAK1 (manuscript under review). While JAK1 mutations have not been reported in LGL leukemia patients, mutations in JAK1 have been identified in several hematological malignancies [9][10][11]13,15,40]. The majority of mutations occur in the pseudokinase domain, and have been characterized to have oncogenic potential [41]; patients with such mutations have poor prognoses [40].
The activation loop within the C-terminal PTK domain includes two adjacent tyrosine phosphorylation sites. Mutation of the first tyrosine to an alanine abrogated kinase activation, indicating that the first tyrosine is necessary for JAK1 activation. An alanine substitution to the second tyrosine allowed kinase activation, showing that activation is not dependent upon the second tyrosine [42]. We detected a mutation at Y1034, which is the second tyrosine within the activation loop. While not necessary for kinase activation, mutations at this location may affect enzymatic activity [43], as well as the specificity and strength of interactions with substrates, thereby leading to altered downstream signaling. Random mutagenesis experiments on human JAK1 have shown that the specific substitution of cysteine for the second tyrosine, analogous to the mutation that we detected in rat JAK1, drives clonal outgrowth and increases JAK/STAT signaling [41].
Here we demonstrated that Jak1 Y1034C is a naturally occurring somatic mutation found in leukemic rats that was shown to elevate the phosphorylated levels of STAT1, STAT3 and STAT5. We observed the largest fold change of phosphorylated STAT5 by mutant relative to WT JAK1, suggesting STAT5 may be more readily phosphorylated by mutant JAK1 [42,44]. Mutation of JAK1 did not increase total JAK1 protein. While we were unable to assess the phosphorylation of JAK1 itself, presumably due to the overlap of mutation location and antibody epitope, increased phosphorylation of JAK1 was shown when the substitution of cysteine for the second tyrosine was performed in human JAK1 [42]. Additionally, future mass spectrometry-based analyses could be insightful in this regard.
In vivo studies involved the transplantation of primary RNK-1, -3, -7, and -16 material into young F344 rats. Previous characterization of these strains specified high cytotoxicity of RNK-16 compared to other strains, and our study enhanced their findings by examining time to disease onset, JAK1 mutation status, and phosphorylated STAT protein levels. We found that rats transplanted with RNK-16 have significantly more rapid disease development. The Jak1 Y1034C mutation was detected only in ex vivo RNK-16 spleen material, which also exhibited increased protein levels of phosphorylated STAT1, STAT3 and STAT5. High cytotoxicity of RNK-16 compared to other RNK strains may be in part explained by the Jak1 Y1034C mutation, which increases downstream STAT signaling and may drive more aggressive leukemogenesis. Interestingly, we observed increased phosphorylated STAT5 in ex vivo RNK-7 material, suggesting that other factors aside from a Jak1 mutation may promote STAT5 phosphorylation. Differing genetic, epigenetic, and signaling alterations may explain differences between the various RNK strains. Performing WGS on other RNK primary material may reveal other variants in addition to the Jak1 Y1034C mutation that may contribute to the pathogenesis of rat NK-LGL leukemia.

Acquisition of Sequencing Data
Paired-end Illumina short read sequences were downloaded for the F344/NCrl sample from the European Nucleotide Archive (ENA; Run: ERR224448). The RNK-16 cell line was kindly provided by Drs. Craig Reynolds and Howard Young (National Cancer Institute, Bethesda, MD, USA). DNA was isolated by AnaPrep (BioChain Institute, Newark, CA, USA) and quantified by Qubit (Thermo Fisher (Waltham, MA, USA). A sequencing library was prepared with a Nextera (San Diego, CA, USA) kit, using half the recommended DNA input and paired-end sequenced on the NextSeq 500 (Illumina (San Diego, CA, USA)) at the Center for Public Health Genomics at the University of Virginia.

Alignments
Illumina short-read sequences from the RNK-16 cell line and the F344 sample were aligned to the Rattus norvegicus reference sequence (rn6) with BWA mem [45], using the default parameters. Putative PCR duplicates were flagged using SAMBLASTER [46], which was also used to add MC and MQ tags to all the output paired-end alignments, and separate the split-reads, the discordant pairs, and the unmapped sequences from the resulting output (Table S3). The SAM outputs were converted to BAM format and sorted by chromosomal coordinates using SAMTools [47]. All BAM files for the same samples were merged using SAMTools, and BAMreport (https://github.com/aakrosh/BAMreport) was then used to generate the alignment statistics and metrics.

CNV Map
Evidence of GC bias were found in both sequence datasets (Table S3), and the same diagnostic plots also revealed evidence of large copy number aberrations in the RNK-16 sample that could confound variant calls. From there, a copy number variant map was created to define the non-uniform ploidy variation in the RNK-16 sample by binning the aligned fragments into intervals of 10 kb mappable bases. Then, a locally weighted linear regression modeled the relationship between GC content and coverage in those intervals. This LOWESS fit was then used to scale the coverage in each bin, based on the mean coverage in bins with the same GC content. Circular binary segmentation (CBS) [48] identified bins that shared the same copy number state, and a Gaussian mixture model was used to genotype the segments and calculate their copy number state.

Identification of Variants
FreeBayes [49] identified the variants in the samples, requiring that the allele fraction in at least one sample be 0.1 and supported by two reads. The CNV map was provided as an input, and required that the Phred scaled probability of a polymorphism at the loci be greater than 5. The following additional options were used: -pooled-discrete, -pooled-continuous, -genotypequalities, -report-genotype-likelihood-max and -allele-balance-priors-off.

Annotation and Filtering of Variants
Variants where the F344/NCrl sample was homozygous for the reference were labelled, and an alternate allele in RNK-16 was observed as somatic, using filters implemented similar to SpeedSeq [50]. To start, genotype likelihoods were extracted from the output of FreeBayes. For RNK-16, we retrieved the best likelihood to not be the reference, and for F344/NCrl, the best likelihood was to be the reference. We required that (a) both the retrieved likelihoods be above a minimum threshold, and (b) we filtered to remove any low-frequency variants in RNK-16 that were also present in the F344/NCrl sample at low frequencies. We then used SnpEff [51] to annotate these labeled variants, using dbSnp and the variant calls from Hermsen et al. [52]. This left us with 255,838 positions, with evidence of an alternate allele in RNK-16 (Table S1).

Gene Enrichment Analyses
Genes with variants were filtered for a Z score of >20 and those labelled with a medium or high impact. Homolog analysis of these gene variants was performed using DAVID (Database for Annotation, Visualization, and Integrated Discovery), resulting in a list of 458 human genes, of which 157 did not have the same name (Table S2). Variant genes were queried in the CGC (Cancer Gene Census) to determine which were implicated in cancers.

Sanger Sequencing to Validate Whole Genome Sequencing Results
Sequencing primers from Eurofins for eight oncogenes identified in COSMIC were designed using Primerblast [53] (Table S4). Bulk DNA from RNK-1, -3, -7, and -16 was sent to Eurofins for Sanger sequencing to determine oncogene mutation status.

JAK1 Mutagenesis
A full-length clone of the rat JAK1 sequence (XM_006238453.1) in the pCMV3 vector was purchased from Sinobiological (Cat. RG81640-UT), as was the pCMV3 empty vector (Cat. CV011) and pCMV3-GFP (Green Florescent Protein) control vector (Cat. CV026). A JAK1 Y1034C mutation was introduced into the rat JAK1 pCMV3 using an In-Fusion HD Site-directed Mutagenesis kit (Takara (Kusatsu, Shiga Prefecture, Japan), Cat. 638909), with primers acquired from Eurofins (F: 5' AGACCGATAAGGAGTACTACAGT 3', R: 5' GGAGTACTACACAGTCAAGGATGAC 3'). Primers were designed to overlap by 15 bp at the 5', ends with a GC content of 40-50%. The mutant DNA was transformed into Stellar Competent cells, and plasmid DNA was isolated. Insertion of only the desired Y1034C mutation and presence of a full-length JAK1 were confirmed by Sanger sequencing.

Animal Study
Animal experimentation was performed according to protocols approved by the Institutional Animal Care and Use Committee at the University of Virginia (Protocol # 4014-05-17). Male F344 rats of approximately 5 weeks of age were obtained from Charles River Laboratory. After one week of conditioning to our vivarium, one million viable RNK-1, RNK-3, RNK-7, or RNK-16 primary cells were transplanted intraperitoneally into F344 rats in triplicate for all groups, except for RNK-1, where there were not enough viable cells to engraft three animals. Peripheral blood samples were collected weekly to monitor leukemia development after RNK cell engraftment. Rats losing 20% body weight most likely developed NK-LGL leukemia and were euthanized. At necropsy, rat spleens were harvested. Splenocytes were isolated using s Ficoll-Hypaque density gradient centrifugation and cryopreserved in liquid nitrogen for future use. Rat blood was collected using Becton Dickinson (BD) (Franklin Lakes, NJ, USA) microcontainers with K2EDTA (BD 365974). Blood smears were prepared using a standard technique, and staining was done using a VWR (Radnor, PA, USA) Hematology Quick stain kit (VWR#10143-224). Slides were examined (Olympus (Shinjuku, Tokyo, Japan) BX51) using immersion oil. Frozen splenocytes were taken from liquid nitrogen, and protein harvested for western blotting analysis, as described above.

Statistical Analysis
Statistical analyses for all in vitro systems utilized unpaired Student's t-test, with p < 0.05 considered statistically significant. Analyses were performed in both Microsoft Excel and GraphPad Prism.

Conclusions
In summary, this work presents WGS of a spontaneously occurring NK-LGL leukemia model, leading to the discovery and characterization of a somatic activating Jak1 Y1034C mutation. The activating mutant may be a useful tool for biochemical studies of JAK/STAT signaling. This work reinforces the critical contributions of JAK/STAT pathway activation in hematologic malignancies, especially those of T-and NK-cell origin. Additionally, RNK-16 engraftment represents an in vivo model of a leukemia driven by mutationally activated JAK1, which may have utility for the evaluation and development of JAK1 inhibitors.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2072-6694/12/1/126/s1. Table S1: Alignment statistics for the RNK-16 and F344/NCrl samples, Table S2: Complete list of variants with an alternate allele in the RNK-16 cell line compared to F344/NCrl, Table S3: List of human homolog genes mapped to rat genes, with variants in RNK-16 cell line filtered by Q score and impact factor, Table S4

Funding:
LGL leukemia research in the Loughran lab is supported by the National Cancer Institute of the National Institutes of Health under award number R01CA178393 and P30CA044579 (to T.P.L.). The content is solely the responsibility of the authors, and does not necessarily represent the official views of the National Institutes of Health. Additional support was provided by the Bess Family Charitable Fund, the LGL Leukemia Foundation, Dr. Charles and Katharine Hutton Tweedy, William J. Branch, and a generous anonymous donor.