NOD mice exhibit major defects in the earliest stages of T cell development in the thymus. Genome-wide genetic and transcriptome analyses were used to investigate the origins and consequences of an early T cell developmental checkpoint breakthrough in Rag1-deficient NOD mice. Quantitative trait locus analysis mapped the presence of checkpoint breakthrough cells to several known NOD diabetes susceptibility regions, particularly insulin-dependent diabetes susceptibility genes (Idd)9/11 on chromosome 4, suggesting common genetic origins for T cell defects affecting this trait and autoimmunity. Genome-wide RNA deep-sequencing of NOD and B6 Rag1-deficient thymocytes revealed the effects of genetic background prior to breakthrough, as well as the cellular consequences of the breakthrough. Transcriptome comparison between the two strains showed enrichment in differentially expressed signal transduction genes, prominently tyrosine kinase and actin-binding genes, in accord with their divergent sensitivities to activating signals. Emerging NOD breakthrough cells aberrantly expressed both stem cell–associated proto-oncogenes, such as Lmo2, Hhex, Lyl1, and Kit, which are normally repressed at the commitment checkpoint, and post–β-selection checkpoint genes, including Cd2 and Cd5. Coexpression of genes characteristic of multipotent progenitors and more mature T cells persists in the expanding population of thymocytes and in the thymic leukemias that emerge with age in these mice. These results show that Rag1-deficient NOD thymocytes have T cell defects that can collapse regulatory boundaries at two early T cell checkpoints, which may predispose them to both leukemia and autoimmunity.

All T cells arise from a small pool of multipotent progenitors, which undergo proliferation and tightly controlled developmental programming induced and sustained by interactions with the thymic environment rich in Notch ligands and cytokines (1, 2). Thymic precursors are CD4 and CD8 double-negative (DN), and they are defined by sequential changes in surface Kit (CD117), CD25, and CD44 into stages: DN1 (or early T cell progenitor, ETP; KithiCD44hiCD25), DN2 (KithiCD44hiCD25+), DN3 (KitloCD44loCD25+), and DN4 (KitloCD44loCD25lo) (3, 4). A TCR-independent commitment checkpoint has recently been identified within the DN2 stage, marked by decreased Kit expression from the multipotent DN2a stage to the committed DN2b stage (5) and dependent on Bcl11b (6, 7). DN2a cells, similar to more immature DN1/ETP cells, undergo proliferative expansion and express legacy stem/progenitor genes such as Kit, Lyl1, Tal1, Gfi1b, and Sfpi1 (PU.1). These features decline sharply in the DN2b/DN3 stages when T cell specification genes are strongly activated, proliferation slows, and efficient TCR gene rearrangement begins (5, 8). Thus, the T cell commitment checkpoint divides the TCR-negative stages of development into two phases: phase I, wherein cells proliferate and retain alternative lineage potential, and phase II, which prepares committed DN3 cells for the first TCR-dependent checkpoint, β-selection (9). Normally, only cells that successfully rearrange a TCRβ and assemble a signaling pre-TCR complex are permitted to pass through the β-selection checkpoint to DN4 and proliferate. These cells then become CD4+CD8+ double-positive (DP), express TCRαβ, and undergo positive and negative selection (10, 11). Rag-deficient T cells are blocked at β-selection and do not generate DP cells.

The NOD mouse is a model of T cell–mediated autoimmune type 1 diabetes, with >20 genetic regions associated with diabetes susceptibility, many of which specifically affect T cell activities (1215). Most T lineages have been implicated in autoimmune susceptibility or resistance in NOD mice, including CD4, CD8, NKT, regulatory T, and γδT cells (16), raising the possibility that all NOD T cells may share fundamental abnormalities that contribute to loss of self-tolerance and might be traceable to their common progenitors (8). We previously reported defects in the earliest stages of T cell development in both wild-type (WT) and TCR-deficient NOD mice. Precursor cells from WT NOD mice exhibit impaired αβT cell development and enhanced γδT cell generation apparently arising from the ETP-DN2 stages (17). Additionally, the NOD genetic background severely compromises β-selection checkpoint control: despite the absence of TCR expression in NOD.SCID and NOD.Rag1−/− (NOD.Rag) mice, aberrant breakthrough DP thymocytes spontaneously appear in all young adult animals (18). These TCR-negative DP cells do not fully mimic normal β-selected cells, as they still express receptors characteristic of earlier DN stages, such as Kit, IL7Rα, and CD25, which are normally turned off by the DP stage (18). Furthermore, older NOD.SCID and NOD.Rag mice develop thymic tumors at high frequency whereas mice of other strains with these immunodeficiencies do not (1821). This suggests a possible link between early T cell checkpoint control and tumor suppression that may be jointly defective in the NOD genetic background.

We used genome-wide genetic and transcriptome analytical methods to investigate the source and consequences of the NOD.Rag thymocyte checkpoint defect. First, we found quantitative trait loci (QTLs) for this trait, all within several of known diabetes susceptibility regions mapped in WT NOD mice. A major QTL localized within the insulin-dependent diabetes susceptibility gene (Idd)9/11 region of chromosome (chr)4 was confirmed using congenic mice. Additionally, genome-wide transcriptome analyses revealed distinct differences in gene expression between thymocytes from NOD.Rag and B6.Rag control mice. The genes differentially expressed between the two strains were enriched for those encoding signaling proteins, suggesting aberrant signal transduction as a possible precondition for breakthrough. Furthermore, newly emergent NOD.Rag breakthrough cells fail to terminate gene expression programs from earlier stages: they coexpress phase I stem/progenitor genes along with T cell–specific genes characteristic of phase II and post–β-selection stages. This mixed gene expression profile foreshadows the phenotype of thymic tumors found in older mice of this strain, which share characteristics with classes of human early type acute T cell lymphoblastic leukemia (T-ALL), suggesting that primary defects in early T cell checkpoint control underlie some forms of T-ALL.

B6.129S7-Rag1tm1Mom/J (B6.Rag) and NOD.Cg-Rag1tm1Mom Prf1tm1Sdz/SzJ (NOD.Rag) (The Jackson Laboratory) and NOD.B10Idd9 line 905 (14) (Taconic Farms) mice were bred and maintained in the Caltech Laboratory Animal Facility using autoclaved cages, food, and water. All animal protocols were reviewed and approved by the Animal Care and Use Committee of the California Institute of Technology.

For the QTL analysis B6.Rag and NOD.Rag mice were crossed and intercrossed for F2 or backcrossed to NOD.Rag for N2 progeny. Thymocytes from 12 to 14 wk progeny were phenotyped by flow cytometric analysis. DNA was extracted from tail tips of 150 N2 and 30 F2 cross mice and 150 polymorphic single nucleotide polymorphisms (SNPs) were genotyped by The Jackson Laboratory Genetic Services. QTL analysis was carried out using the R/qtl program (22) and p values were obtained from genome-wide significance test using 5000 permutations (23). Congenic NOD.B10Idd9 Rag mice were created by crossing NOD.Rag and NOD.B10Idd9 mice and repeated backcrossing until the knockout Rag1 gene and the B10Idd9 region were homozygous, as determined by PCR analysis.

Freshly isolated thymocytes were either stained immediately for flow cytometetric analysis or cultured on OP9-delta-like (DL)1 or OP9-DL4 cells with 5 ng/ml IL-7, as previously described (17). For cell stimulations, thymocytes were cultured for 1 h in RPMI 1640 supplemented with 10% FBS (Life Technologies) before treatment with PMA. Cells were immediately fixed in 1.5% formaldehyde in PBS at 37°C and permeabilized by slow addition of ice-cold methanol to a final concentration of 90%. Cells were incubated on ice for 30 min, washed with PBS plus 0.5% BSA, and incubated with either phospho-p42/p44 (Erk1/2)-Alexa Fluor 647 Abs or isotype controls (Cell Signaling Technology, Danvers, MA) before washing and flow cytometric analysis.

T thymocytes were FACS sorted to obtain CD25+CD4 cells from NOD.Rag mice at 4 wk age (prebreakthrough) and 7 wk (at the time of first breakthrough), and age-matched B6.Rag mice for RNA extraction. mRNA purification and cDNA library building were performed as described (24). Sequencing was done using Illumina high-throughput genome analyzer IIx sequencers at the California Institute of Technology’s Jacobs Genetics and Genomics Laboratory, and data have been deposited in National Center for Biotechnology Information’s Gene Expression Omnibus (25) and are accessible through Gene Expression Omnibus series accession no. GSE40688 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE40688). Computational analysis was carried out using software developed at the California Institute of Technology (26, 27) as well as Mathematica and R. For analysis, 38-bp, single-read raw sequence reads were trimmed to 32 bp and mapped onto the mouse genome build NCBI37/mm9 using Bowtie (bowtie-0.12.1; http://bowtie-bio.sourceforge.net/index.shtml) with setting “-v 2 -k 11 -m 10 -t–best–strata”. The mappable data were then processed by the ERANGE v. 3.3 RNA-Seq analysis program to obtain the expression level in reads per kilobase per million reads (RPKM) for each gene (26). We obtained 13–17 × 106 mappable reads per sample for analysis. Based on a pairwise comparison between samples, B4 values were divided by 1.27 to remove a scaling artifact. Genes were required to have at least one sample with 32 mappable reads (based on a significance level of 0.05, with a Bonferonni adjustment for 12,000 comparisons) and have an annotated transcript name, leaving 11,098 named genes for further analysis. For each gene, comparisons of expression in the four samples were calculated as differences on a log2 scale, after replacing any zeros by 0.01 as an approximate limit of detection. The significance probabilities (p values) were based on the assumption of Poisson distributions of counts, under the null hypothesis of equal rates, allowing for the difference in the number of megabases sequenced (and mapped). Three-fold change is typically the more stringent requirement and was used to determine the differentially expressed genes. The statistical significance requirement screens out some genes with low expression whose 3-fold differences might be attributable to sampling variation. The design does not permit assessing extra-Poisson variation, which would be more important at larger expression levels, where the 3-fold criterion rather than the statistical comparison is the dominant concern.

To profile the connection between different populations, hierarchical clustering was carried out for each selected subgroup of genes from B6.Rag and NOD.Rag samples or WT B6 ETP to DP populations (24). Individual mRNA data for the selected genes were first normalized by the corresponding geometric mean (GeoMean) of comparing populations, and then one-dimensional hierarchical clustering (along genes) was performed on the results after log2 transformation. Euclidean distance and Ward linkage were used (MATLAB 7.10.0). Clusters were visualized as heat maps.

Quantitative PCR was carried out based on selected RNA-Seq results using FACS-sorted thymocyte samples. RNA was extracted and reverse-transcribed and quantitative PCRs were carried out using SYBR GreenER (Invitrogen) in a GeneAmp 7900HT sequence detection system (Applied Biosystems), as previously described (17). Relative expression was calculated for each gene by the ΔCT (change in cycling threshold) method, normalized to β-actin expression. Primer sequences were as previously published (6, 28), plus: Zap70, forward, 5′-TGTCCTCCTGAGATGTATGCAC-3′, reverse, 5′-ATAGTTCCGCATACGTTGTTCC-3′; Ptcra, forward, 5′-CTGGCTCCACCCATCACACT-3′, reverse, 5′-TGCCATTGCCAGCTGAGA-3′; Notch1, forward, 5′-CCACTGTGAACTGCCCTATGT-3′, reverse, 5′-TTGTTTCCTGGACAGTCATCC-3′; Heyl, forward, 5′-AAGCTGGAGAAAGCTGAGGTC-3′, reverse, 5′-CCAATACTCCGGAAGTCAACA-3′; Kit, forward, 5′-ACTTCGCCTGACCAGATTAAA-3′, reverse, 5′-CGTACGTCAGGATTTCTGGTT-3′; Hhex, forward, 5′-ACTACACGCACGCCCTACTC-3′, reverse, 5′-GTCGTTGGAGAACCTCACTTG-3′; Epha2, forward, 5′-CAGGAAGGCTACGAGAAGGTC-3′, reverse, 5′-CAGGGTATGCTCTGGACACTC-3′; Mllt4, forward, 5′-GACTGGACAGTGACAGGGTGT-3′, reverse, 5′-CAGTATCAGTTCAGGCCCAGT-3′; Dapk1, forward, 5′-CATCACCCTGCATGAGGTCTA-3′, reverse, 5′-TGCCTCCTCTTCAGTCAGAGA-3′; Vegfa, forward, 5′-CTCCGAAACCATGAACTTTCT-3′, reverse, 5′-ATGGGACTTCTGCTCTCCTTC-3′; Tnfrsf9, forward, 5′-GCTGGCCCTGATCTTCATTA-3′, reverse, 5′-ATCGGCAGCTACAAGCATCT -3′. The Tcf7 primers (28) yielded lower values than previously published (5); however, the relative efficiency was consistent between samples in these experiments.

We previously reported that aberrant Kit+ DP breakthrough cells spontaneously appear in the thymuses of all male and female NOD.Rag mice by 8 wk age, whereas B6.Rag and (B6 × NOD)F1.Rag thymocytes arrest normally in DN3 at the β-selection checkpoint (18). To map the genetic basis of the T cell breakthrough, we tested progeny from a (B6 × NOD)F2.Rag intercross and an N2 backcross with NOD.Rag as the recurrent parent. Thymocytes from individual parental and cross progeny were phenotyped by flow cytometry at 12–14 wk age for two traits: 1) the percentage of CD4+ (including DP) cells, which is indicative of illegitimate progression past the β-selection checkpoint, and 2) the percentage of Kit+ cells, indicative of a failure to downregulate a key phase I stem/progenitor gene (see Fig. 1E for a model of developmental stages). As shown in Fig. 1A, B6.Rag and (B6 × NOD)F1.Rag mice at 12 wk age exhibited very few CD4+ or Kit+ cells, whereas most thymocytes from NOD.Rag mice were CD4+ and Kit+. Results for individual progeny differed; for example, no. 87 exhibited a non-breakthrough phenotype similar to B6.Rag, and no. 86 showed Kit+CD4+ breakthrough cells like NOD.Rag mice. The F2 intercross yielded a low percentage of progeny with elevated percentages of CD4+ and/or Kit+ cells (<10%), whereas ∼35% of N2 backcross progeny exhibited T cell breakthrough. Despite indicating differing developmental defects, the traits are not independent, as shown in the strong correlation (r2 = 0.82) between the log-transformed percentages of CD4+ and Kit+ cells for individual progeny (Fig. 1B).

FIGURE 1.

Genome-wide QTL analysis of the NOD.Rag early T cell breakthrough trait shows mapping to known diabetes susceptibility regions. (A) Flow cytometric data showing representative percentage CD4+ versus percentage CD8+ and percentage Kit+ versus percentage CD25+ phenotyping plots for thymocytes from control parental B6.Rag, NOD.Rag, and F1.Rag mice as well as two individual (NOD × B6)F2.Rag cross mice, at 14 wk age. (B) Scatterplot showing the correlation between the two measured thymocyte phenotypic traits: percentage CD4+ and percentage Kit+ cells. (C) Genome-wide QTL scan for percentage CD4+ and percentage Kit+ thymocytes, using the combined N2 plus F2 crosses. LOD scores are plotted for percentage CD4+ cells (black) and percentage Kit+ cells (gray) for all 19 autosomal chromosomes. Black line indicates a LOD score of 3. (D) Expanded plots of LOD scores for the chromosomes with the highest peaks in the combined cross, chr4 and chr17, for the two phenotypes, percentage CD4+ and percentage Kit+ thymocytes. Approximate regions of known diabetes susceptibility Idd loci are also noted. (E) Diagram of key early T cell developmental stages (based on Ref. 9) and possible pathways (dashed and solid arrows) leading to pro– T cell checkpoint breakthrough and T-ALL in NOD.Rag mice.

FIGURE 1.

Genome-wide QTL analysis of the NOD.Rag early T cell breakthrough trait shows mapping to known diabetes susceptibility regions. (A) Flow cytometric data showing representative percentage CD4+ versus percentage CD8+ and percentage Kit+ versus percentage CD25+ phenotyping plots for thymocytes from control parental B6.Rag, NOD.Rag, and F1.Rag mice as well as two individual (NOD × B6)F2.Rag cross mice, at 14 wk age. (B) Scatterplot showing the correlation between the two measured thymocyte phenotypic traits: percentage CD4+ and percentage Kit+ cells. (C) Genome-wide QTL scan for percentage CD4+ and percentage Kit+ thymocytes, using the combined N2 plus F2 crosses. LOD scores are plotted for percentage CD4+ cells (black) and percentage Kit+ cells (gray) for all 19 autosomal chromosomes. Black line indicates a LOD score of 3. (D) Expanded plots of LOD scores for the chromosomes with the highest peaks in the combined cross, chr4 and chr17, for the two phenotypes, percentage CD4+ and percentage Kit+ thymocytes. Approximate regions of known diabetes susceptibility Idd loci are also noted. (E) Diagram of key early T cell developmental stages (based on Ref. 9) and possible pathways (dashed and solid arrows) leading to pro– T cell checkpoint breakthrough and T-ALL in NOD.Rag mice.

Close modal

For the QTL analysis, 150 N2 and 30 F2 phenotyped progeny were genotyped for 150 polymorphic SNPs distributed over the 19 autosomal chromosomes. Genome-wide one-dimensional QTL scans were performed using R/qtl on data from the N2 and F2 crosses alone and combined (22). Peak QTL logarithm of the odds (LOD) scores for the two phenotypes and the N2 and F2 crosses, individually and combined, are shown in Table I, and genome-wide QTL scans for the combined cross data are shown in Fig. 1C. A highly significant QTL (LOD > 12) was detected on distal chr4 for both percentage CD4+ and percentage Kit+ in both the N2 and the combined cross (Fig. 1C, 1D). Weaker QTLs were detected on chr17 for both phenotypes and N2 and combined crosses (LOD > 3 for percentage Kit+) (Fig. 1C, 1D), and on chr13, in the N2 cross only (LOD > 2.5). A scan using the F2 cross alone revealed only a suggestive locus on chr11. Remarkably, all four of these QTLs map to known NOD Idd QTL regions. The chr4 QTL lies within the Idd9/11 region (14, 29, 30), the chr17 QTL encompasses the region containing Idd1, Idd16, Idd23, and Idd24 (3133) (Fig. 1D), the chr13 QTL region overlaps with Idd14 (34), and the chr11 QTL region overlaps with Idd4 (29). These results allow the possibility that the same genes could be involved in the loss of checkpoint control and susceptibility to autoimmune diabetes in NOD mice.

Table I.
Peak regions from a genome-wide QTL scan of N2, F2, and combined N2 plus F2 crosses for percentage CD4+ and percentage Kit+ thymocytes
CrossTraitChrPosition (cM)aPeak (Mb)bLODcp Valued
N2 %CD4+ 91.6 142.8 12.49 <<0.01 
13 40.3 76.2 2.85 <0.05 
17 19.2 40.2 2.69 <0.1 
%Kit+ 91.6 142.8 18.97 <<0.01 
13 40.3 76.2 2.59 <0.1 
17 19.2 40.2 3.2 <0.05 
F2 %CD4+ 11 42.3 70.3 2.85 <0.5 
%Kit+ 11 31 60 3.46 <0.2 
Combined N2 plus F2 %CD4+ 91.6 142.8 13.67 <<0.01 
17 19.2 40.2 2.6 <0.025 
%Kit+ 91.6 142.8 21.13 <<0.01 
17 19.2 40.2 3.87 <0.01 
CrossTraitChrPosition (cM)aPeak (Mb)bLODcp Valued
N2 %CD4+ 91.6 142.8 12.49 <<0.01 
13 40.3 76.2 2.85 <0.05 
17 19.2 40.2 2.69 <0.1 
%Kit+ 91.6 142.8 18.97 <<0.01 
13 40.3 76.2 2.59 <0.1 
17 19.2 40.2 3.2 <0.05 
F2 %CD4+ 11 42.3 70.3 2.85 <0.5 
%Kit+ 11 31 60 3.46 <0.2 
Combined N2 plus F2 %CD4+ 91.6 142.8 13.67 <<0.01 
17 19.2 40.2 2.6 <0.025 
%Kit+ 91.6 142.8 21.13 <<0.01 
17 19.2 40.2 3.87 <0.01 
a

Genetic map position in cM, calculated from cross genotypes.

b

Peak locations in NCBI37/mm9 mouse assembly (Mb) obtained from peak SNP location.

c

Log10 likelihood ratio (LOD) calculated with R/qtl using a simple model for the N2 and F2 crosses and adding cross to the model for the combined N2 plus F2 cross data; only LOD scores >2.5 are shown.

d

The p values are based on genome-wide significance tests for LOD thresholds (5000 permutations) for each cross and phenotype.

Control of percentage CD4+ and percentage Kit+ mapped to the same genetic regions, and therefore they arise from linked or identical genetic sources and are distinct from the Cd4 (chr6) or Kit (chr5) genes themselves. The aberrant upregulation of CD4 occurs in cells that already express high levels of Kit (18), and results in Fig. 1B support this: no progeny were high for %CD4+ and low for %Kit+ (right bottom quadrant), whereas a few progeny were high for %Kit+ and low for %CD4+ (left top quadrant). Also, QTL LOD scores for %Kit+ were typically higher than for %CD4 (Fig. 1C, 1D, Table I), indicating that Kit may be a more sensitive measure of breakthrough. These cells express CD25 and IL7Rα, as well as Kit (18), and their abnormal coexpression with CD4 (and CD8) suggests that the breakthrough arises from specific genes on the NOD genetic background that cause bypass of both the T cell commitment and β-selection checkpoints, as shown in Fig. 1E (and discussed further below).

To confirm the presence of a gene or genes in the NOD Idd9/11 chr4 QTL region controlling the breakthrough trait, NOD.B10Idd9 congenic mice (line 905) (14) were bred to NOD.Rag to generate a NOD.B10Idd9Rag congenic mouse line. The congenic chr4 region contains at least three distinct diabetes resistance alleles from B10 mice, Idd9.1, Idd9.2, and Idd9.3 (Fig. 2A), as well as a possible Idd11 locus (30). The NOD.B10Idd9Rag mice showed significantly lower percentages of CD4+ and DP breakthrough cells than did NOD.Rag mice (Fig. 2B, 2C), demonstrating that the chr4 congenic region includes at least one gene involved in susceptibility to thymocyte breakthrough. Other genetic regions also contribute, as NOD.B10Idd9Rag congenic thymocytes exhibited more breakthrough cells than did B6.Rag (Fig. 2B, 2C), in accord with the QTL analysis.

FIGURE 2.

NOD.B10Idd9Rag congenic mouse strain analysis of the chr4 Idd9 QTL peak region. (A) Diagram of the chr4 congenic region from NOD.B10Idd9 (L. Wicker, Cambridge, U.K.), which was crossed onto NOD.Rag congenic mice. Indicated are the genomic regions derived from B10Idd9 (black) and NOD (white) and the recombination region (gray), as well as the Mappair markers used in the congenic cross and the QTL peak (arrow). Approximate locations for Idd9.1, 9.2, and 9.3 are shown (from T1Dbase.org). (B) Flow cytometric analysis of B6.Rag, NOD.Rag, and four NOD.B10Idd9Rag congenic mice at 12 wk age. (C) Summary of percentage CD4, percentage DP, and percentage Kit+ cells for individual NOD.B10Idd9Rag (NOD.chr4Rag) congenic mice in comparison with parental B6.Rag, NOD.Rag, and F1.Rag mice, all at 12–16 wk old. A p value < 0.0001 is from t tests comparing data from NOD.Rag and NOD.B10Idd9Rag mice.

FIGURE 2.

NOD.B10Idd9Rag congenic mouse strain analysis of the chr4 Idd9 QTL peak region. (A) Diagram of the chr4 congenic region from NOD.B10Idd9 (L. Wicker, Cambridge, U.K.), which was crossed onto NOD.Rag congenic mice. Indicated are the genomic regions derived from B10Idd9 (black) and NOD (white) and the recombination region (gray), as well as the Mappair markers used in the congenic cross and the QTL peak (arrow). Approximate locations for Idd9.1, 9.2, and 9.3 are shown (from T1Dbase.org). (B) Flow cytometric analysis of B6.Rag, NOD.Rag, and four NOD.B10Idd9Rag congenic mice at 12 wk age. (C) Summary of percentage CD4, percentage DP, and percentage Kit+ cells for individual NOD.B10Idd9Rag (NOD.chr4Rag) congenic mice in comparison with parental B6.Rag, NOD.Rag, and F1.Rag mice, all at 12–16 wk old. A p value < 0.0001 is from t tests comparing data from NOD.Rag and NOD.B10Idd9Rag mice.

Close modal

Insights into the breakthrough program should emerge from both determining underlying strain-specific gene expression differences that precede the loss of checkpoint control, as well as investigating the earliest postbreakthrough changes. To first evaluate differences between NOD.Rag and B6.Rag thymocytes that restrict checkpoint control loss to the NOD background, we carried out a genome-wide RNA deep sequencing analysis (RNA-Seq) (26), comparing sorted CD4CD25+ DN thymocytes from NOD.Rag mice at 4 (N4, before breakthrough) and 7 wk (N7, early breakthrough), with corresponding non-breakthrough cells from B6.Rag mice at 4 (B4) and 7 (B7) wks. In each sample, the sorted cells were predominantly pre–β-selection DN3 cells and they were phenotypically similar (Fig. 3A).

FIGURE 3.

RNA-Seq transcriptome analysis comparing CD25+ DN thymocytes from B6.Rag and NOD.Rag mice. (A) Gates used for sorting phenotypically similar CD4CD25+ cells for RNA purification: B6.Rag at 4 and 7 wk (B4, B7) and NOD.Rag at 4 and 7 wk (N4, N7). (B) Difference versus mean of log2 expression of N4/N7 and B4/B7, including 11,098 genes expressed at >32 reads in at least one sample. Horizontal gray lines show 3- and 4-fold differences. Genes differing by >3-fold are shown in dark circles, <3-fold by gray circles. (C) Heatmap for hierarchical clustering of 412 genes differing by >3-fold between N(4 and 7) versus B(4 and 7). Data are shown as log2(RPKM/GeoMean). (D) Hierarchical cluster heatmaps showing the normal (WT B6) developmentally regulated expression (log2(RPKM/GeoMean)) of differentially expressed genes that are higher (top) and lower (bottom) in N(4 and 7) compared with B(4 and 7) using our previously published data (24). (E and F) RNA-Seq tracks showing sequence read histograms for expressed genes of interest in QTL regions from B6.Rag and NOD.Rag CD25+ DN cells at 4 and 7 wk age. Gene tracks are shown using the University of California at Santa Cruz browser, and the data are mapped onto the mouse genome build NCBI37/mm9. (E) Genes located in the chr4 major QTL region, including differentially expressed genes, Epha2, Padi3, Tnfrsf9 (4-1BB), Ctnnbip1 (Icat), a differentially spliced gene, Hdac1 (differentially spliced exons indicated with arrows), and Lck, a nondifferentially expressed gene for reference. (F) Differentially expressed genes in the chr17 QTL region, including Vegfa, Notch4, Mllt4 (Afadin), and Ddr1, and in the chr13 QTL region, Dapk1.

FIGURE 3.

RNA-Seq transcriptome analysis comparing CD25+ DN thymocytes from B6.Rag and NOD.Rag mice. (A) Gates used for sorting phenotypically similar CD4CD25+ cells for RNA purification: B6.Rag at 4 and 7 wk (B4, B7) and NOD.Rag at 4 and 7 wk (N4, N7). (B) Difference versus mean of log2 expression of N4/N7 and B4/B7, including 11,098 genes expressed at >32 reads in at least one sample. Horizontal gray lines show 3- and 4-fold differences. Genes differing by >3-fold are shown in dark circles, <3-fold by gray circles. (C) Heatmap for hierarchical clustering of 412 genes differing by >3-fold between N(4 and 7) versus B(4 and 7). Data are shown as log2(RPKM/GeoMean). (D) Hierarchical cluster heatmaps showing the normal (WT B6) developmentally regulated expression (log2(RPKM/GeoMean)) of differentially expressed genes that are higher (top) and lower (bottom) in N(4 and 7) compared with B(4 and 7) using our previously published data (24). (E and F) RNA-Seq tracks showing sequence read histograms for expressed genes of interest in QTL regions from B6.Rag and NOD.Rag CD25+ DN cells at 4 and 7 wk age. Gene tracks are shown using the University of California at Santa Cruz browser, and the data are mapped onto the mouse genome build NCBI37/mm9. (E) Genes located in the chr4 major QTL region, including differentially expressed genes, Epha2, Padi3, Tnfrsf9 (4-1BB), Ctnnbip1 (Icat), a differentially spliced gene, Hdac1 (differentially spliced exons indicated with arrows), and Lck, a nondifferentially expressed gene for reference. (F) Differentially expressed genes in the chr17 QTL region, including Vegfa, Notch4, Mllt4 (Afadin), and Ddr1, and in the chr13 QTL region, Dapk1.

Close modal

Of >11,000 expressed genes, 412 genes were differentially expressed by at least 3-fold between the two strains (excluding 10 genes more prominently differing between N7 and N4, see below) (Fig. 3B, 3C, Supplemental Table I). These genes are normally expressed in at least one stage of T cell development, from ETP to DP, based on our data from WT B6 mice (24), as shown in heatmaps in Fig. 3D. Genes with strain-specific differences in expression, at higher and lower levels in NOD cells, exhibit highly varied patterns of expression across these stages of T cell development, in contrast to what we find with genes specific to N7 postbreakthrough cells (see below). Overall, 34% of the strain-dependent differentially expressed genes have peak expression in phase I (23% ETP, 11% DN2a), 40% in phase II (19% DN2b, 21% DN3), and 26% in post–β-selection DP cells.

Gene ontology analysis of the differentially expressed genes using DAVID bioinformatics resources 6.7 (35) showed three types of genes were overrepresented: MHC (H2), signaling, and actin-binding protein genes (Table II). Enrichment of H2 genes is not unexpected owing to their expression diversity between mouse strains, although no role is known for them in early T cell development. Protein kinase genes, including tyrosine and serine/threonine kinases, were enriched, as were adherens junction protein genes, which include actin-binding, tyrosine kinase, and calmodulin-binding protein genes. All of these protein classes play critical roles in modulating signal transduction activity in T cells (36). Notably, all 11 of the differentially expressed tyrosine kinases were higher in NOD.Rag cells, suggesting that the two strains likely exhibit differences in signaling activity. This may not only contribute to checkpoint failure, but also promote the thymic lymphomas that develop in these mice, as tyrosine kinases are common oncogenes and therapeutic targets in leukemia and other malignancies (37).

Table II.
Gene ontology analysis on >3-fold differentially expressed genes between NOD.Rag and B6.Rag CD25+ DN cells showing major enriched categories of genes
GOTERM CategoryaGenesbFold Enrichmentp ValueBenjamini
MHC protein complex H2-K1,H2-Q2, H2-M6-PS, H2-Q10, C920025E04rik, H2-OA, H2-BL, H2-Q1, H2-T10, H2-AB1, H2-Q7,H2-T3 (all chr17) 9.8 6.5 × 10−7 1.6 × 10−4 
Actin binding Lima1, Ccdc88a, Mybpc2, Tln2, Inppl1,Myo7a, Evl, Spire2,Tpm3, Coro2b,Fmn2, Syne2, Mtap1a,Capg,Tmod4,Eps8l1,Mylk, Myh10 (chr11) 3.5 1.8 × 10−5 7.9 × 10−3 
Protein tyrosine kinase activity Tyro3, Ddr1 (chr17), Fgfr1, Ptk2, Ltk, Ptpn3, Ptk2b, Hck, Txk, Ephb4, Epha2 (chr4) 3.7 7.7 × 10−4 1.1 × 10−1 
Adherens junction Ptk2, Lima1, Itga6, Ptk2b, Tln2, Lmo7, Evl, Mllt4 (chr17), Arhgap26 4.3 1.1 × 10−3 6.8 × 10−2 
Protein kinase activity Tyro3, Fgfr1, Alpk1, Ltk, Ptpn3, Hck, Ulk4, Ephb4, Epha2 (chr4), Dapk1 (chr13), Rps6kl1, Ddr1 (chr17), Ptk2, Ptk2b, Camk1, Stk39,Camk2B,Txk, Grk5,Camk2a,Mylk 2.0 4.3 × 10−3 1.9 × 10−1 
Calmodulin binding Adcy1, Myo7a,Camk1,Camk2b, Camk2a,Mylk, Myh10 (chr11),Dapk1 (chr13) 3.9 4.5 × 10−3 1.8 × 10−1 
GOTERM CategoryaGenesbFold Enrichmentp ValueBenjamini
MHC protein complex H2-K1,H2-Q2, H2-M6-PS, H2-Q10, C920025E04rik, H2-OA, H2-BL, H2-Q1, H2-T10, H2-AB1, H2-Q7,H2-T3 (all chr17) 9.8 6.5 × 10−7 1.6 × 10−4 
Actin binding Lima1, Ccdc88a, Mybpc2, Tln2, Inppl1,Myo7a, Evl, Spire2,Tpm3, Coro2b,Fmn2, Syne2, Mtap1a,Capg,Tmod4,Eps8l1,Mylk, Myh10 (chr11) 3.5 1.8 × 10−5 7.9 × 10−3 
Protein tyrosine kinase activity Tyro3, Ddr1 (chr17), Fgfr1, Ptk2, Ltk, Ptpn3, Ptk2b, Hck, Txk, Ephb4, Epha2 (chr4) 3.7 7.7 × 10−4 1.1 × 10−1 
Adherens junction Ptk2, Lima1, Itga6, Ptk2b, Tln2, Lmo7, Evl, Mllt4 (chr17), Arhgap26 4.3 1.1 × 10−3 6.8 × 10−2 
Protein kinase activity Tyro3, Fgfr1, Alpk1, Ltk, Ptpn3, Hck, Ulk4, Ephb4, Epha2 (chr4), Dapk1 (chr13), Rps6kl1, Ddr1 (chr17), Ptk2, Ptk2b, Camk1, Stk39,Camk2B,Txk, Grk5,Camk2a,Mylk 2.0 4.3 × 10−3 1.9 × 10−1 
Calmodulin binding Adcy1, Myo7a,Camk1,Camk2b, Camk2a,Mylk, Myh10 (chr11),Dapk1 (chr13) 3.9 4.5 × 10−3 1.8 × 10−1 
a

GOTERM: Gene Ontology Term.

b

Genes expressed at higher amounts in NOD.Rag cells are indicated in bold type. Genes located within a mapped QTL region in this study are noted (chr).

Several genes differentially expressed between NOD.Rag and B6.Rag are located within and near the chr4 QTL peak region (Supplemental Table I). RNA-Seq tracks for several of these genes are shown in Fig. 3E, including Epha2, an ephrin tyrosine kinase receptor strongly expressed in both NOD samples; Padi3, a peptidylarginine deiminase; Tnfrsf9 (CD137, 4-1BB, an Idd9.3 candidate gene) (14); Ctnnbip1 (ICAT), a Wnt pathway inhibitor; and Hdac1, a histone deacetylase, which is differentially spliced. These genes contrast with the stably expressed chr4 region gene, Lck. Furthermore, a number of poorly defined transcription units are differentially expressed between the two strains, including at least three in a region near the chr4 peak that is predicted to encode multiple KRAB-zinc finger proteins (unpublished observation). Differentially expressed genes of interest were also found within other QTL regions, including Vegfa; Notch4; Mllt4 (Afadin), a Ras-associated adherens junction gene; and Ddr1, discoidin domain tyrosine kinase receptor, found in the chr17 QTL region, plus a tumor suppressor gene, death-associated protein kinase, Dapk1, located within the chr13 QTL region and expressed in B6.Rag but not in NOD.Rag cells (Fig. 3F).

The enrichment of differentially expressed signal transduction pathway genes suggests that inappropriate signaling might underlie the NOD.Rag checkpoint breakthrough. Thymocytes from sex-matched 4- to 5-wk-old prebreakthrough NOD.Rag and B6.Rag mice were assayed for differences in responses to the protein kinase C activator PMA, in short- and long-term assays. Thymocytes from both strains used for these experiments were phenotypically similar and >90% were DN3 stage cells (18). Phosphorylation of Erk1/2, mediators of the Ras/MAPK pathway, was measured by p-Erk1/2 intracellular staining of thymocytes from mice stimulated with PMA (Fig. 4A). NOD.Rag populations exhibited cells with consistently lower levels of pErk than B6.Rag in response to PMA stimulation, over all PMA concentrations tested, as shown by plotting mean fluorescence intensity (MFI) ratios between pairs of matched cells (NOD:B6), over multiple experiments, which are almost uniformly below an expected equal value of 1 (red line) (Fig. 4B). These results are in agreement with a recent report showing differences in Erk1/2 phosphorylation between B6 and NOD TCR-transgenic T cells (38).

FIGURE 4.

NOD.Rag and B6.Rag thymocytes exhibit differential responses to PMA stimulation. (A) Representative flow cytometric histogram showing intracellular Erk1/2 phosphorylation (pErk1/2) in freshly isolated thymocytes (>90% DN3 cells) from prebreakthrough NOD.Rag and B6.Rag mice treated for 2 min with PMA. (B) Summary plot of the NOD/B6 ratios of p-Erk1/2 (using geometric MFI from intracellular staining) for thymocytes from age- and sex-matched NOD.Rag and B6.Rag mice treated with 1–25 ng/l PMA. The MFI was consistently lower for NOD.Rag cells than for B6.Rag cells as shown by ratios <1 (equal expression, indicated with a red line). Data included are combined from five independent experiments. (C) Responses of thymocytes from 6-wk-old NOD.Rag (dashed lines) and B6.Rag (solid lines), placed in coculture with OP9-DL1 cells and stimulated with graded doses of PMA, and calcium ionophore A23187 (CI; doses of 0, 22, 43, and 87 nM as indicated) and assayed at day 7 for cell number (left graph) and percentage CD4+ cells (right graph). Results are representative of three independent experiments.

FIGURE 4.

NOD.Rag and B6.Rag thymocytes exhibit differential responses to PMA stimulation. (A) Representative flow cytometric histogram showing intracellular Erk1/2 phosphorylation (pErk1/2) in freshly isolated thymocytes (>90% DN3 cells) from prebreakthrough NOD.Rag and B6.Rag mice treated for 2 min with PMA. (B) Summary plot of the NOD/B6 ratios of p-Erk1/2 (using geometric MFI from intracellular staining) for thymocytes from age- and sex-matched NOD.Rag and B6.Rag mice treated with 1–25 ng/l PMA. The MFI was consistently lower for NOD.Rag cells than for B6.Rag cells as shown by ratios <1 (equal expression, indicated with a red line). Data included are combined from five independent experiments. (C) Responses of thymocytes from 6-wk-old NOD.Rag (dashed lines) and B6.Rag (solid lines), placed in coculture with OP9-DL1 cells and stimulated with graded doses of PMA, and calcium ionophore A23187 (CI; doses of 0, 22, 43, and 87 nM as indicated) and assayed at day 7 for cell number (left graph) and percentage CD4+ cells (right graph). Results are representative of three independent experiments.

Close modal

PMA with calcium ionophore can mimic β-selection–promoting signals in vitro to drive Rag-deficient thymocytes to generate DP cells in long-term cultures. Therefore, NOD.Rag and B6.Rag thymocytes were also tested for developmental responses to these signals in coculture with OP9 stromal cells expressing Notch ligand, DL1 (OP9-DL1). OP9-DL1 coculture supports T cell development through the β-selection checkpoint in thymic progenitors from WT B6 and NOD mice (17, 39), but not from Rag-deficient mice unless an artificial TCR signal is delivered. Cells were treated with graded doses of PMA and calcium ionophore and analyzed after 7 d. NOD.Rag cells (dashed lines) were more sensitive to PMA than B6.Rag cells (solid lines) in this assay, as seen by the shift in dose-response curves for both total cell numbers (Fig. 4C, left), and the percentage of cells that had progressed through the β-selection checkpoint, as measured by percentage CD4+ cells (including DP) (Fig. 4C, right). These results demonstrate consistent differences between the two strains in the responses of their thymocytes to a specific stimulus.

NOD.Rag DP breakthrough cells only partially resemble post–β-selection cells. They express surface CD4, CD8, and CD2 and initiate TCR-Cα and Bcl-xL transcription, like normal or induced DP cells, while also retaining characteristics of pre–β-selection DN cells such as surface CD25, KIT, and IL-7Rα, and Spib and Hes1 transcription (18). To determine whether these were isolated cases or indicators of more general program derangement, the transcriptome data were analyzed for the earliest breakthrough-specific gene changes by comparing NOD.Rag cells (N7) with prebreakthrough NOD.Rag (N4) and normal B6.Rag (B4, B7) cells. Sorted N7 cells excluded breakthrough cells expressing surface CD4 but included their precursors (Fig. 3A).

Very few genes exhibited a >3-fold difference between N7 and N4, and almost all were higher in N7 (Fig. 5A). This finding is consistent with the emergence of a subpopulation of abnormal cells, expressing novel genes; any downregulated genes are likely to be masked by the dominant presence of conventional DN3 cells. Fifty-two breakthrough-specific differentially expressed genes were identified, including only those genes that distinguished N7 from all other samples, N4, B4, and B7, by at least 3-fold (Fig. 5B, Supplemental Table II). To determine whether the emergent population more closely resembles pre– or post–β-selected cells, we used our RNA-Seq data for WT B6 DN1 to DP cells (24) as a reference for the developmentally regulated expression of the N7-specific genes, and their patterns of expression are shown in a heatmap in Fig. 5C. Strikingly, although the breakthrough phenotype was originally identified by expression of post–β-selection markers CD2, CD4, and CD8, most of the genes preferentially expressed by N7 are normal phase I genes. Of the differentially expressed genes, 71% exhibited peak expression in phase I (48% in DN1/ETP and 23% in DN2a), 21% in phase II (15% in DN2b and 6% in DN3), and only 8% in DP. RNA-Seq tracks for some key phase I genes, Lmo2, Hhex, Lyl1, Kit, and Mef2c (9), as well as a post–β-selection gene, Cd2, are shown in Fig. 5D. An additional phase I gene, Dlk1 (delta-like 1; Pref1), was expressed only in N7, as a previously unreported truncated splice form (arrow). Dlk1 supports proliferation in hematopoietic progenitors, maintains cells in an undifferentiated state, and is overexpressed in some leukemias (40).

FIGURE 5.

RNA-Seq transcriptome analysis of NOD.Rag breakthrough cells. (A) Difference versus mean of log2 expression of NOD.Rag 7 wk (N7, at breakthrough) versus 4 wk (N4, prebreakthrough). Horizontal gray lines show 3- and 4-fold differences. Genes differing by > 3-fold are shown in dark circles, <3-fold by gray circles. (B) Heatmap for hierarchical clustering of 52 genes differing by >3-fold between N7 versus all other samples. Data are shown as log2(RPKM/GeoMean). (C) Hierarchical cluster heatmap showing the WT B6 expression profile (log2[RPKM/GeoMean]) of 52 genes differentially expressed between N7 and other samples using data from ETP to DP developmental stages (24). (D) RNA-Seq gene tracks from B6.Rag and NOD.Rag CD25+ DN cells at 4 and 7 wk age showing a selection of genes of interest expressed >3-fold higher in N7 in comparison with N4, B4, and B7, including phase I genes, Lmo2, Hhex, Lyl1 Kit, and Mef2c, post–β-selection gene, Cd2, and a differentially expressed and spliced Notch-related gene, Dlk1 (lack of expression of 3′ exons is indicated with an arrow).

FIGURE 5.

RNA-Seq transcriptome analysis of NOD.Rag breakthrough cells. (A) Difference versus mean of log2 expression of NOD.Rag 7 wk (N7, at breakthrough) versus 4 wk (N4, prebreakthrough). Horizontal gray lines show 3- and 4-fold differences. Genes differing by > 3-fold are shown in dark circles, <3-fold by gray circles. (B) Heatmap for hierarchical clustering of 52 genes differing by >3-fold between N7 versus all other samples. Data are shown as log2(RPKM/GeoMean). (C) Hierarchical cluster heatmap showing the WT B6 expression profile (log2[RPKM/GeoMean]) of 52 genes differentially expressed between N7 and other samples using data from ETP to DP developmental stages (24). (D) RNA-Seq gene tracks from B6.Rag and NOD.Rag CD25+ DN cells at 4 and 7 wk age showing a selection of genes of interest expressed >3-fold higher in N7 in comparison with N4, B4, and B7, including phase I genes, Lmo2, Hhex, Lyl1 Kit, and Mef2c, post–β-selection gene, Cd2, and a differentially expressed and spliced Notch-related gene, Dlk1 (lack of expression of 3′ exons is indicated with an arrow).

Close modal

The prominent expression of phase I stem/progenitor genes suggests that the N7 emergent population either failed to silence or re-established the early T cell program, which normally occurs at the DN2 commitment checkpoint. At the same time, the emerging cells also appear to bypass the β-selection checkpoint, as shown by expression of Cd2 and Cd5, which are normally turned on in DN4 (http://www.immgen.org). Whereas the assayed cells were selected to be negative for surface CD4, Cd4 gene expression was detectably upregulated (1.9-fold higher in N7 than in N4 cells). These results show that the emerging NOD.Rag breakthrough cells have greatly disordered developmental programming and fail to passage or arrest properly at the early T cell checkpoints (Fig. 1E).

These features of the transcriptome analysis suggest a possible linkage between T cell breakthrough and the high incidence of thymic lymphomas that appear in NOD.Rag mice (18, 20, 21). NOD.Rag thymic cell numbers increase dramatically with age whereas B6.Rag thymocyte cell numbers change very little or decline over time (Fig. 6A). Beginning at 20 wk, NOD.Rag mice exhibiting thymic lymphomas appeared, as previously reported (18, 21). Additionally, when DN thymocytes from young NOD.Rag and B6.Rag mice were cultured with Notch ligands and IL-7, both expanded but only NOD.Rag DN cells violated the β-selection checkpoint and upregulated CD4 (Fig. 6B). Furthermore, NOD.Rag cells isolated at 9–12 wk age, after breakthrough but before lymphoma formation, expanded continuously in cultures in the presence of Notch ligand and IL-7. Whereas sorted NOD.Rag DP cells from three individuals expanded 6.7 ± 3.2-fold (SEM) during 5 d, and only generated more DP cells, sorted CD25+ DN cells from the same mice expanded 3.1 ± 1.0 and generated both DP cells and a small population of DN cells (Fig. 6C), suggesting that cells within the DN population may be a source of illegitimate differentiation. These thymocytes failed to survive when cultured in the absence of Notch ligands or IL-7 (data not shown), indicating that they are committed to the T lineage.

FIGURE 6.

Changes in thymocyte cell number, in vitro potential, and gene expression with age and lymphoma. (A) Plot of thymocyte cell numbers with age showing the increase in cell numbers with age in NOD.Rag but not B6.Rag mice. (B) DN cells from 6–wk-old prebreakthrough NOD.Rag mice cultured with OP9-DL1 stroma plus IL-7 exhibit upregulation of CD4 whereas B6.Rag cells do not express CD4. (C) Sorted CD25+ DN and DP cells from a 12-wk-old NOD.Rag mouse proliferate and differentiate when cultured on OP9-DL4 stroma supplemented with 5 ng/ml IL-7 for 3 and 16 d before FACS analysis for expression of CD4 and CD8. Data are representative of two experiments. (DG) Real-time quantitative PCR expression of some developmentally regulated genes in sorted B6.Rag CD25+ DN cells, from 4, 6, and 10 wk age (brown bars), sorted NOD.Rag prelymphoma CD25+ DN, CD4+, and DP cells, from 4, 6, 10, 13, and 16 wk age (green bars), unsorted cells from five independent NOD.Rag lymphomas (pink bars), and representative WT NOD DN1/ETP, DN2a, DN2b, DN3a, and DP cells for reference (blue bars). Relative gene expression levels are shown for representatives of (D) key T cell genes, (E) Notch and target genes, (F) phase I (ETP/DN2a) genes, and (G) differentially expressed genes from the transcriptome analysis.

FIGURE 6.

Changes in thymocyte cell number, in vitro potential, and gene expression with age and lymphoma. (A) Plot of thymocyte cell numbers with age showing the increase in cell numbers with age in NOD.Rag but not B6.Rag mice. (B) DN cells from 6–wk-old prebreakthrough NOD.Rag mice cultured with OP9-DL1 stroma plus IL-7 exhibit upregulation of CD4 whereas B6.Rag cells do not express CD4. (C) Sorted CD25+ DN and DP cells from a 12-wk-old NOD.Rag mouse proliferate and differentiate when cultured on OP9-DL4 stroma supplemented with 5 ng/ml IL-7 for 3 and 16 d before FACS analysis for expression of CD4 and CD8. Data are representative of two experiments. (DG) Real-time quantitative PCR expression of some developmentally regulated genes in sorted B6.Rag CD25+ DN cells, from 4, 6, and 10 wk age (brown bars), sorted NOD.Rag prelymphoma CD25+ DN, CD4+, and DP cells, from 4, 6, 10, 13, and 16 wk age (green bars), unsorted cells from five independent NOD.Rag lymphomas (pink bars), and representative WT NOD DN1/ETP, DN2a, DN2b, DN3a, and DP cells for reference (blue bars). Relative gene expression levels are shown for representatives of (D) key T cell genes, (E) Notch and target genes, (F) phase I (ETP/DN2a) genes, and (G) differentially expressed genes from the transcriptome analysis.

Close modal

If gene expression abnormalities in NOD.Rag breakthrough cells favored their differentiation into lymphoma cells, then at least some of their aberrant features should be preserved in the lymphomas that develop a few months after the breakthrough. To test this prediction, we used real-time quantitative PCR on populations of NOD.Rag cells to track changes in gene expression related to age and lymphoma status. We tested the persistence of representative genes from various developmental stages, as well as a few genes that showed differential expression between NOD.Rag and B6.Rag DN cells (Fig. 6D–G), using RNA from the following sorted thymocyte populations: CD25+ DN cells from 4- to 10-wk-old B6.Rag (arrested at the β-selection checkpoint, brown); CD25+ DN cells from 4- to 6-wk-old NOD.Rag mice (arrested at the β-selection checkpoint, before overt breakthrough, dark green); DN, CD4+, and DP cells from postbreakthrough 10- to 16-wk-old NOD.Rag mice (light green); unsorted thymic lymphoma cells from >4-mo-old NOD.Rag mice (pink); and ETP, DN2a, DN2b, DN3a, and DP cells from WT NOD thymuses for comparison (blue).

NOD.Rag thymocytes obtained from all ages, including CD4+ and DP breakthrough cells and thymic lymphomas, expressed critical T cell genes, including Bcl11b, Tcf7, CD3e, and Zap70 (Fig. 6D). Bcl11b expression in all NOD.Rag samples is particularly significant, as it is critical for phase I gene repression and T cell differentiation past the DN2a stage (6, 7). Bcl11b was reported to be a haploinsufficient tumor suppressor in human T-ALL (41, 42), and expression of Bcl11b, as well as CD3e, was indeed lower by 2- to 3-fold in breakthrough and lymphoma cells in comparison with B6.Rag and NOD WT DN3a/DP controls. A key TCR signaling gene, Zap70, was consistently expressed at higher levels in NOD.Rag cells than in B6.Rag cells, and Il7ra, a critical early T cell cytokine receptor, was sustained at high levels in almost all populations. Continuing expression of Il7ra in all breakthrough cell stages and in thymic lymphoma cells supports the requirement for IL-7 for cell proliferation in early T cells, and this pathway has been reported to be critical in human early T-ALL (43).

Notch signaling is required for early but not later stages of T cell development (44), and constitutive Notch activation is one of the most potent and common oncogenic factors in T-ALL (45). Expression of Notch1, and target genes Dtx1, Hes1, Heyl, and Ptcra, was sustained in all stages, including in lymphoma cells (Fig. 6E), implying sustained Notch signaling.

Among phase I genes (Fig. 6F), Kit expression was sustained at high levels in NOD.Rag cells at all ages, as well as in lymphoma cells. Lmo2 was upregulated >100-fold between 4 and 6 wk in NOD.Rag cells, declined in breakthrough CD4+ and DP cells, but was still expressed in all lymphoma cells. Hhex was also elevated in most NOD.Rag cells, including all lymphomas. Lyl1 and Sfpi1 levels varied between different NOD.Rag cell populations, but, similar to Lmo2 and Hhex, both genes were still expressed in all thymic lymphoma cells at levels higher than in WT NOD DP cells. These phase I genes are known proto-oncogenes and are likely to be major contributors to the thymocyte expansion and lymphomas found in NOD.Rag mice.

The lymphomas did not passively maintain gene expression patterns from DN and breakthrough cells, as shown in Fig. 6G for representative differentially expressed genes, with each showing distinctive expression patterns. Of particular interest, the NOD-specific high-level expression of Epha2 (in the chr4 QTL peak) began before breakthrough and was maintained in all stages, including all lymphomas. Quantitative PCR analysis also showed Epha2 to be differentially expressed between WT NOD and B6 cells, at all DN stages, and in DP and γδT cells (Supplemental Fig. 1A). Higher NOD.Rag cell expression of Mllt4 (Afadin, in chr17 QTL peak) was also confirmed, but expression levels declined with age. Expression of the tumor suppressor kinase, Dapk1 (chr13 QTL peak), was lower in NOD.Rag cells, although some expression persisted in the lymphomas. WT NOD cells also showed a lower expression of Dapk1 in all stages of development in comparison with B6 cells, especially in the DN2a and DP stages where it is undetectable (Supplemental Fig. 1B). Expression of Vegfa and Tnfrsf9 declined with age, but Vegfa was consistently expressed in the lymphomas whereas Tnfrsf9 was variable. Overall, the lymphoma cells maintained many of the unique gene expression patterns found in NOD.Rag N7 cells, especially expression of phase I proto-oncogenes, suggesting that origins of the transformed cells may lie in the early breakthrough cells which fail to shut off the genes that sustain the pleuripotency and proliferation of progenitor T cells (Fig. 1E).

Thymic lymphomas generated in mice with an NOD genetic background preserve many gene expression features of the early breakthrough cells, which themselves emerge as the result of an earlier dysfunction. From the transcriptome analysis, the dominant feature of breakthrough cells emerging spontaneously in the NOD.Rag thymus is the collapse of two regulatory boundaries: one that normally distinguishes precommitment ETP and DN2a stage cells from committed DN3 cells, with the other operating at the β-selection checkpoint that requires a TCR signal for DN3 to DP progression (Fig. 1E). The breakthrough cells and the later-arising lymphoma cells share a highly abnormal coexpression of legacy progenitor cell genes, including proto-oncogenes, normally restricted to uncommitted ETP and DN2a cells, along with T cell genes characteristic of later stage cells, before and after TCR-dependent β-selection. Unlike normal Rag-deficient cells, breakthrough thymocytes undergo long-term expansion and aberrant differentiation in vitro, and thymic cell numbers increase with age in vivo, even before thymic lymphomas appear. Thus, the breakthrough cells may include a population of preleukemic cells, which can provide cellular targets of oncogenic transformation.

One striking feature of the T cell commitment checkpoint in the DN2 stage is the brief period of overlap in phase I and T cell identity genes during the transition from dominant expression to silencing of legacy stem cell genes after the onset of expression of T cell–specific genes (5, 9). The most notable feature of our current transcriptome and quantitative PCR data are the apparent failure to terminate the phase I ETP-DN2 progenitor/stem cell program in early breakthrough cells despite upregulation of T cell–specific genes including Bcl11b, which is involved in their silencing (6), making the breakthrough cells appear to be retained in a DN2-like state. Expression of genes such as Cd2, Cd5, and Cd4 indicate a bypass of the later β-selection checkpoint as well, possibly due to an aberrant pre–TCR-like signal. It is also possible that failure of the first checkpoint might prevent the normal establishment of the second checkpoint, making the DN2 stage the critical point of breakthough vulnerability. Aberrant signal transduction at one or both checkpoints as a promoter of breakthrough is suggested by the transcriptome data showing a marked overrepresentation of signaling genes, especially actin-binding proteins and tyrosine kinases, among differentially expressed genes between NOD.Rag and B6.Rag thymocytes. Actin/cytoskeletal and signal transduction pathways are interrelated in TCR signaling (36), and alterations in expression of key genes may trigger inappropriate signaling at the checkpoints. Evidence that activated NOD CD4+ cells exhibit a higher overall level of phospho-tyrosine than B6 cells (46) supports a potential role for at least some of the 11 differentially overexpressed tyrosine kinases in NOD.Rag cells, and we show in vitro evidence for subtle but consistent strain-specific differences in responses to PMA between cells from the two strains. Furthermore, consistent differences in Erk1/2 phosphorylation have also been reported for immature and mature NOD and B6 TCR-transgenic T cells (38).

The NOD.Rag mixed gene expression signature of early breakthrough T cells and thymic lymphoma cells that emerge at high frequency after 4–5 mo is shared with certain human acute T-ALL, especially of the early LYL1-LMO2 T-ALL subgroup (47), as well as the recently described ETP-ALL, whose gene signatures share features with normal ETP and myeloid cells (48). ETP-ALL also shares similarities with both T cell and myeloid stem cells, and mutations in cytokine (Flt3 and IL-7R) and Ras signaling pathways are prominent (43). In common with ETP-ALL, NOD.Rag breakthrough cells express genes indicative of their early T cell programming, for example, Bcl11b, Ptcra, and Il7r, in addition to stem/progenitor genes such as Lmo2, Hhex, Lyl1, Kit, and Sfpi1, which also likely contribute to the myeloid potential of WT NOD and B6 ETP/DN2a cells (5). The prominent overexpression of Lmo2 in the NOD.Rag breakthrough cells is noteworthy, as DN cells overexpressing transgenic Lmo2 develop into self-renewing thymocytes, with striking gene expression similarities to NOD.Rag breakthrough cells, followed by T-ALL 8 mo later (49). The NOD.Rag breakthrough and thymic lymphoma development may be only partially attributable to the enhanced expression of Lmo2, as thymic tumors appear much more rapidly in NOD.Rag mice, beginning 2 mo after breakthrough cells first appear, possibly due to the elevated expression of tyrosine kinases, which are commonly oncogenic (37).

Genes within the identified QTL regions control propensity for breakthrough in the NOD genetic background and they mapped to known autoimmune diabetes susceptibility regions. The strong chr4 QTL peak in the Idd9/11 region (14) clearly includes at least one recessive NOD allele needed to promote efficient breakthrough, as shown by the reduced breakthrough in congenic NOD.B10Idd9Rag mice. Possible breakthrough-enhancing QTL regions were also detected in the Idd16/23/1/24 region on chr17, Idd14 on chr13, and Idd4 on chr11. Although all of the QTL regions remain too large to determine whether any of the same genes control both autoimmunity and the early T cell breakthrough, the close mapping of the two different T cell–related phenotypes is highly suggestive and requires further analysis to pinpoint source genes. The B10 Idd9 congenic region, which protects against the breakthrough trait, is reported to protect NOD mice against type 1 diabetes after lymphocytic infiltration in the pancreas, suggesting control over the balance between pathogenic and protective T cell responses (14). If the differentially expressed genes found within the QTL regions are also used in establishing self-tolerance, then the autoimmunity promoting T cell defects may be traceable to the earliest common T cell progenitors, even before TCR rearrangement occurs.

To date, animal models of T-ALL have typically depended on artificial gene expression alterations such as ablation of tumor suppressor genes or transgenic overexpression of oncogenes (50). The results reported in this study suggest that the NOD.Rag mouse may be a useful unmanipulated model for studying oncogenic progression from a preleukemic state, which arises spontaneously and predictably from defects in early T cell checkpoint control. Moreover, our genetic analysis suggests a potential linkage to autoimmune diabetes in WT NOD mice, possibly related to an intrinsic defect in thymic αβT cell development, which also arises from ETP-DN2 stages (17). Human epidemiological studies have linked T-ALL and autoimmune type 1 diabetes, both of which are predominantly pediatric diseases (51, 52). Furthermore, genes and molecular pathways involved in lymphocyte immune checkpoints and tumor suppression overlap, and the consequences of loss of checkpoint control, which sets thresholds for appropriate proliferation and survival, can be key in both autoimmunity and in lymphoid cancers (53). Thus, our results could extend alterations in T cell selection and functions back to the earliest stages of their development, and they raise the possibility that genes involved in an early commitment checkpoint breakthrough may contribute to defects in lymphocyte homeostasis and both T cell leukemia and autoimmune disease.

We thank Brian Williams, Justine Chia, Avni Gandhi, and Sagar Damle (California Institute of Technology) for technical and bioinformatics assistance; Donna Walls and Weidong Zhang (The Jackson Laboratory) for genotyping and preliminary statistical analysis; J.C. Zuñiga-Pflücker (University of Toronto) for providing OP9-DL1 and OP9-DL4 cells; L.S. Wicker (Cambridge University) for NOD.B10Idd9 congenic mice; Rochelle Diamond, Diane Perez, and Pat Koen from the California Institute of Technology Flow Cytometry and Cell Sorting Facility; Scott Washburn and Natasha Bouey for animal care; and Ali Mortazavi and members of the Rothenberg Laboratory for helpful suggestions.

This work was supported by National Institutes of Health Grant AI64590 (to M.A.Y.); California Institute of Technology Summer Undergraduate Research Fellowships (to C.Y.L.); the Albert Billings Ruddock Professorship (to E.V.R.); the Louis A. Garfinkle Memorial Laboratory Fund; and the Al Sherman Foundation.

The sequences presented in this article have been submitted to the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE40688) under accession number GSE40688.

The online version of this article contains supplemental material.

Abbreviations used in this article:

chr

chromosome

DL

delta-like

DN

double-negative

DP

double-positive

ETP

early T cell progenitor

GeoMean

geometric mean

Idd

insulin-dependent diabetes susceptibility gene

LOD

logarithm of the odds

MFI

mean fluorescence intensity

QTL

quantitative trait locus

RPKM

read per kilobase per million reads

SNP

single nucleotide polymorphism

T-ALL

T cell acute lymphoblastic leukemia

WT

wild-type.

1
Petrie
H. T.
,
Zúñiga-Pflücker
J. C.
.
2007
.
Zoned out: functional mapping of stromal signaling microenvironments in the thymus.
Annu. Rev. Immunol.
25
:
649
679
.
2
Love
P. E.
,
Bhandoola
A.
.
2011
.
Signal integration and crosstalk during thymocyte migration and emigration.
Nat. Rev. Immunol.
11
:
469
477
.
3
Godfrey
D. I.
,
Kennedy
J.
,
Suda
T.
,
Zlotnik
A.
.
1993
.
A developmental pathway involving four phenotypically and functionally distinct subsets of CD3CD4CD8 triple-negative adult mouse thymocytes defined by CD44 and CD25 expression.
J. Immunol.
150
:
4244
4252
.
4
Porritt
H. E.
,
Rumfelt
L. L.
,
Tabrizifard
S.
,
Schmitt
T. M.
,
Zúñiga-Pflücker
J. C.
,
Petrie
H. T.
.
2004
.
Heterogeneity among DN1 prothymocytes reveals multiple progenitors with different capacities to generate T cell and non-T cell lineages.
Immunity
20
:
735
745
.
5
Yui
M. A.
,
Feng
N.
,
Rothenberg
E. V.
.
2010
.
Fine-scale staging of T cell lineage commitment in adult mouse thymus.
J. Immunol.
185
:
284
293
.
6
Li
L.
,
Leid
M.
,
Rothenberg
E. V.
.
2010
.
An early T cell lineage commitment checkpoint dependent on the transcription factor Bcl11b.
Science
329
:
89
93
.
7
Ikawa
T.
,
Hirose
S.
,
Masuda
K.
,
Kakugawa
K.
,
Satoh
R.
,
Shibano-Satoh
A.
,
Kominami
R.
,
Katsura
Y.
,
Kawamoto
H.
.
2010
.
An essential developmental checkpoint for production of the T cell lineage.
Science
329
:
93
96
.
8
Rothenberg
E. V.
,
Moore
J. E.
,
Yui
M. A.
.
2008
.
Launching the T-cell-lineage developmental programme.
Nat. Rev. Immunol.
8
:
9
21
.
9
Rothenberg
E. V.
,
Zhang
J.
,
Li
L.
.
2010
.
Multilayered specification of the T-cell lineage fate.
Immunol. Rev.
238
:
150
168
.
10
Hoffman
E. S.
,
Passoni
L.
,
Crompton
T.
,
Leu
T. M.
,
Schatz
D. G.
,
Koff
A.
,
Owen
M. J.
,
Hayday
A. C.
.
1996
.
Productive T-cell receptor β-chain gene rearrangement: coincident regulation of cell cycle and clonality during development in vivo.
Genes Dev.
10
:
948
962
.
11
Borowski
C.
,
Martin
C.
,
Gounari
F.
,
Haughn
L.
,
Aifantis
I.
,
Grassi
F.
,
von Boehmer
H.
.
2002
.
On the brink of becoming a T cell.
Curr. Opin. Immunol.
14
:
200
206
.
12
Anderson
M. S.
,
Bluestone
J. A.
.
2005
.
The NOD mouse: a model of immune dysregulation.
Annu. Rev. Immunol.
23
:
447
485
.
13
Delovitch
T. L.
,
Singh
B.
.
1997
.
The nonobese diabetic mouse as a model of autoimmune diabetes: immune dysregulation gets the NOD.
Immunity
7
:
727
738
.
14
Lyons
P. A.
,
Hancock
W. W.
,
Denny
P.
,
Lord
C. J.
,
Hill
N. J.
,
Armitage
N.
,
Siegmund
T.
,
Todd
J. A.
,
Phillips
M. S.
,
Hess
J. F.
, et al
.
2000
.
The NOD Idd9 genetic interval influences the pathogenicity of insulitis and contains molecular variants of Cd30, Tnfr2, and Cd137.
Immunity
13
:
107
115
.
15
Serreze
D. V.
,
Leiter
E. H.
.
2001
.
Genes and cellular requirements for autoimmune diabetes susceptibility in nonobese diabetic mice.
Curr. Dir. Autoimmun.
4
:
31
67
.
16
Todd
J. A.
,
Wicker
L. S.
.
2001
.
Genetic protection from the inflammatory disease type 1 diabetes in humans and animal models.
Immunity
15
:
387
395
.
17
Feng
N.
,
Vegh
P.
,
Rothenberg
E. V.
,
Yui
M. A.
.
2011
.
Lineage divergence at the first TCR dependent checkpoint: preferential γδ and impaired αβ T cell development in nonobese diabetic mice
.
J. Immunol.
186:
826
837
.
18
Yui
M. A.
,
Rothenberg
E. V.
.
2004
.
Deranged early T cell development in immunodeficient strains of nonobese diabetic mice.
J. Immunol.
173
:
5381
5391
.
19
Prochazka
M.
,
Gaskins
H. R.
,
Shultz
L. D.
,
Leiter
E. H.
.
1992
.
The nonobese diabetic scid mouse: model for spontaneous thymomagenesis associated with immunodeficiency.
Proc. Natl. Acad. Sci. USA
89
:
3290
3294
.
20
Shultz
L. D.
,
Lang
P. A.
,
Christianson
S. W.
,
Gott
B.
,
Lyons
B.
,
Umeda
S.
,
Leiter
E.
,
Hesselton
R.
,
Wagar
E. J.
,
Leif
J. H.
, et al
.
2000
.
NOD/LtSz-Rag1null mice: an immunodeficient and radioresistant model for engraftment of human hematolymphoid cells, HIV infection, and adoptive transfer of NOD mouse diabetogenic T cells.
J. Immunol.
164
:
2496
2507
.
21
Chiu
P. P.
,
Ivakine
E.
,
Mortin-Toth
S.
,
Danska
J. S.
.
2002
.
Susceptibility to lymphoid neoplasia in immunodeficient strains of nonobese diabetic mice.
Cancer Res.
62
:
5828
5834
.
22
Broman
K. W.
,
Wu
H.
,
Sen
S.
,
Churchill
G. A.
.
2003
.
R/qtl: QTL mapping in experimental crosses.
Bioinformatics
19
:
889
890
.
23
Kruglyak
L.
,
Lander
E. S.
.
1995
.
A nonparametric approach for mapping quantitative trait loci.
Genetics
139
:
1421
1428
.
24
Zhang
J. A.
,
Mortazavi
A.
,
Williams
B. A.
,
Wold
B. J.
,
Rothenberg
E. V.
.
2012
.
Dynamic transformations of genome-wide epigenetic marking and transcriptional control establish T cell identity.
Cell
149
:
467
482
.
25
Edgar
R.
,
Domrachev
M.
,
Lash
A. E.
.
2002
.
Gene Expression Omnibus: NCBI gene expression and hybridization array data repository.
Nucleic Acids Res.
30
:
207
210
.
26
Mortazavi
A.
,
Williams
B. A.
,
McCue
K.
,
Schaeffer
L.
,
Wold
B.
.
2008
.
Mapping and quantifying mammalian transcriptomes by RNA-Seq.
Nat. Methods
5
:
621
628
.
27
Pepke
S.
,
Wold
B.
,
Mortazavi
A.
.
2009
.
Computation for ChIP-seq and RNA-seq studies.
Nat. Methods
6
(
Suppl.
):
S22
S32
.
28
David-Fung
E. S.
,
Butler
R.
,
Buzi
G.
,
Yui
M. A.
,
Diamond
R. A.
,
Anderson
M. K.
,
Rowen
L.
,
Rothenberg
E. V.
.
2009
.
Transcription factor expression dynamics of early T-lymphocyte specification and commitment.
Dev. Biol.
325
:
444
467
.
29
Ghosh
S.
,
Palmer
S. M.
,
Rodrigues
N. R.
,
Cordell
H. J.
,
Hearne
C. M.
,
Cornall
R. J.
,
Prins
J. B.
,
McShane
P.
,
Lathrop
G. M.
,
Peterson
L. B.
, et al
.
1993
.
Polygenic control of autoimmune diabetes in nonobese diabetic mice.
Nat. Genet.
4
:
404
409
.
30
Brodnicki
T. C.
,
McClive
P.
,
Couper
S.
,
Morahan
G.
.
2000
.
Localization of Idd11 using NOD congenic mouse strains: elimination of Slc9a1 as a candidate gene.
Immunogenetics
51
:
37
41
.
31
Inoue
K.
,
Ikegami
H.
,
Fujisawa
T.
,
Noso
S.
,
Nojima
K.
,
Babaya
N.
,
Itoi-Babaya
M.
,
Makimo
S.
,
Ogihara
T.
.
2004
.
Allelic variation in class I K gene as candidate for a second component of MHC-linked susceptibility to type 1 diabetes in non-obese diabetic mice.
Diabetologia
47
:
739
747
.
32
Boulard
O.
,
Damotte
D.
,
Deruytter
N.
,
Fluteau
G.
,
Carnaud
C.
,
Garchon
H. J.
.
2002
.
An interval tightly linked to but distinct from the H2 complex controls both overt diabetes (Idd16) and chronic experimental autoimmune thyroiditis (Ceat1) in nonobese diabetic mice.
Diabetes
51
:
2141
2147
.
33
Hiromine
Y.
,
Fujisawa
T.
,
Noso
S.
,
Babaya
N.
,
Kawabata
Y.
,
Ikegami
H.
.
2008
.
Congenic mapping of the MHC-linked susceptibility to type 1 diabetes in the NOD mouse: at least two genes contribute to the Idd16 effect.
Ann. N. Y. Acad. Sci.
1150
:
90
92
.
34
McAleer
M. A.
,
Reifsnyder
P.
,
Palmer
S. M.
,
Prochazka
M.
,
Love
J. M.
,
Copeman
J. B.
,
Powell
E. E.
,
Rodrigues
N. R.
,
Prins
J. B.
,
Serreze
D. V.
, et al
.
1995
.
Crosses of NOD mice with the related NON strain: a polygenic model for IDDM.
Diabetes
44
:
1186
1195
.
35
Huang
W.
,
Sherman
B. T.
,
Lempicki
R. A.
.
2009
.
Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources.
Nat. Protoc.
4
:
44
57
.
36
Burkhardt
J. K.
,
Carrizosa
E.
,
Shaffer
M. H.
.
2008
.
The actin cytoskeleton in T cell activation.
Annu. Rev. Immunol.
26
:
233
259
.
37
Wadleigh
M.
,
DeAngelo
D. J.
,
Griffin
J. D.
,
Stone
R. M.
.
2005
.
After chronic myelogenous leukemia: tyrosine kinase inhibitors in other hematologic malignancies.
Blood
105
:
22
30
.
38
Mingueneau
M.
,
Jiang
W.
,
Feuerer
M.
,
Mathis
D.
,
Benoist
C.
.
2012
.
Thymic negative selection is functional in NOD mice.
J. Exp. Med.
209
:
623
637
.
39
Schmitt
T. M.
,
Zúñiga-Pflücker
J. C.
.
2002
.
Induction of T cell development from hematopoietic progenitor cells by delta-like-1 in vitro.
Immunity
17
:
749
756
.
40
Moore
K. A.
,
Pytowski
B.
,
Witte
L.
,
Hicklin
D.
,
Lemischka
I. R.
.
1997
.
Hematopoietic activity of a stromal cell transmembrane protein containing epidermal growth factor-like repeat motifs.
Proc. Natl. Acad. Sci. USA
94
:
4011
4016
.
41
De Keersmaecker
K.
,
Real
P. J.
,
Gatta
G. D.
,
Palomero
T.
,
Sulis
M. L.
,
Tosello
V.
,
Van Vlierberghe
P.
,
Barnes
K.
,
Castillo
M.
,
Sole
X.
, et al
.
2010
.
The TLX1 oncogene drives aneuploidy in T cell transformation.
Nat. Med.
16
:
1321
1327
.
42
Gutierrez
A.
,
Kentsis
A.
,
Sanda
T.
,
Holmfeldt
L.
,
Chen
S. C.
,
Zhang
J.
,
Protopopov
A.
,
Chin
L.
,
Dahlberg
S. E.
,
Neuberg
D. S.
, et al
.
2011
.
The BCL11B tumor suppressor is mutated across the major molecular subtypes of T-cell acute lymphoblastic leukemia.
Blood
118
:
4169
4173
.
43
Zhang
J.
,
Ding
L.
,
Holmfeldt
L.
,
Wu
G.
,
Heatley
S. L.
,
Payne-Turner
D.
,
Easton
J.
,
Chen
X.
,
Wang
J.
,
Rusch
M.
, et al
.
2012
.
The genetic basis of early T-cell precursor acute lymphoblastic leukaemia.
Nature
481
:
157
163
.
44
Wolfer
A.
,
Bakker
T.
,
Wilson
A.
,
Nicolas
M.
,
Ioannidis
V.
,
Littman
D. R.
,
Lee
P. P.
,
Wilson
C. B.
,
Held
W.
,
MacDonald
H. R.
,
Radtke
F.
.
2001
.
Inactivation of Notch 1 in immature thymocytes does not perturb CD4 or CD8T cell development.
Nat. Immunol.
2
:
235
241
.
45
Weng
A. P.
,
Ferrando
A. A.
,
Lee
W.
,
Morris
J. P.
 IV
,
Silverman
L. B.
,
Sanchez-Irizarry
C.
,
Blacklow
S. C.
,
Look
A. T.
,
Aster
J. C.
.
2004
.
Activating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia.
Science
306
:
269
271
.
46
Iwai
L. K.
,
Benoist
C.
,
Mathis
D.
,
White
F. M.
.
2010
.
Quantitative phosphoproteomic analysis of T cell receptor signaling in diabetes prone and resistant mice.
J. Proteome Res.
9
:
3135
3145
.
47
Ferrando
A. A.
,
Neuberg
D. S.
,
Staunton
J.
,
Loh
M. L.
,
Huard
C.
,
Raimondi
S. C.
,
Behm
F. G.
,
Pui
C. H.
,
Downing
J. R.
,
Gilliland
D. G.
, et al
.
2002
.
Gene expression signatures define novel oncogenic pathways in T cell acute lymphoblastic leukemia.
Cancer Cell
1
:
75
87
.
48
Coustan-Smith
E.
,
Mullighan
C. G.
,
Onciu
M.
,
Behm
F. G.
,
Raimondi
S. C.
,
Pei
D.
,
Cheng
C.
,
Su
X.
,
Rubnitz
J. E.
,
Basso
G.
, et al
.
2009
.
Early T-cell precursor leukaemia: a subtype of very high-risk acute lymphoblastic leukaemia.
Lancet Oncol.
10
:
147
156
.
49
McCormack
M. P.
,
Young
L. F.
,
Vasudevan
S.
,
de Graaf
C. A.
,
Codrington
R.
,
Rabbitts
T. H.
,
Jane
S. M.
,
Curtis
D. J.
.
2010
.
The Lmo2 oncogene initiates leukemia in mice by inducing thymocyte self-renewal.
Science
327
:
879
883
.
50
Aifantis
I.
,
Raetz
E.
,
Buonamici
S.
.
2008
.
Molecular pathogenesis of T-cell leukaemia and lymphoma.
Nat. Rev. Immunol.
8
:
380
390
.
51
Shu
X.
,
Ji
J.
,
Li
X.
,
Sundquist
J.
,
Sundquist
K.
,
Hemminki
K.
.
2010
.
Cancer risk among patients hospitalized for type 1 diabetes mellitus: a population-based cohort study in Sweden.
Diabet. Med.
27
:
791
797
.
52
Manda
S. O.
,
Feltbower
R. G.
,
Gilthorpe
M. S.
.
2009
.
Investigating spatio-temporal similarities in the epidemiology of childhood leukaemia and diabetes.
Eur. J. Epidemiol.
24
:
743
752
.
53
Goodnow
C. C.
2007
.
Multistep pathogenesis of autoimmune disease.
Cell
130
:
25
35
.

The authors have no financial conflicts of interest.