Genomic landscape of TCRαβ and TCRγδ T-large granular lymphocyte leukemia

Abstract

Large granular lymphocyte (LGL) leukemia comprises a group of rare lymphoproliferative disorders whose molecular landscape is incompletely defined. We leveraged paired whole-exome and transcriptome sequencing in the largest LGL leukemia cohort to date, which included 105 patients (93 T-cell receptor αβ [TCRαβ] T-LGL and 12 TCRγδ T-LGL). Seventy-six mutations were observed in 3 or more patients in the cohort, and out of those, STAT3, KMT2D, PIK3R1, TTN, EYS, and SULF1 mutations were shared between both subtypes. We identified ARHGAP25, ABCC9, PCDHA11, SULF1, SLC6A15, DDX59, DNMT3A, FAS, KDM6A, KMT2D, PIK3R1, STAT3, STAT5B, TET2, and TNFAIP3 as recurrently mutated putative drivers using an unbiased driver analysis approach leveraging our whole-exome cohort. Hotspot mutations in STAT3, PIK3R1, and FAS were detected, whereas truncating mutations in epigenetic modifying enzymes such as KMT2D and TET2 were observed. Moreover, STAT3 mutations co-occurred with mutations in chromatin and epigenetic modifying genes, especially KMT2D and SETD1B (P < .01 and P < .05, respectively). STAT3 was mutated in 50.5% of the patients. Most common Y640F STAT3 mutation was associated with lower absolute neutrophil count values, and N647I mutation was associated with lower hemoglobin values. Somatic activating mutations (Q160P, D170Y, L287F) in the STAT3 coiled-coil domain were characterized. STAT3-mutant patients exhibited increased mutational burden and enrichment of a mutational signature associated with increased spontaneous deamination of 5-methylcytosine. Finally, gene expression analysis revealed enrichment of interferon-γ signaling and decreased phosphatidylinositol 3-kinase-Akt signaling for STAT3-mutant patients. These findings highlight the clinical and molecular heterogeneity of this rare disorder.

Graphical abstract

Figure 1.

Mutational characteristics of T and GD LGL leukemia subtypes. (A-B) Nonsilent mutation burden and affected gene counts in T and GD subtypes. (C) Euler diagram of genes that are mutated in at least 3 patients, sectioned by LGL subtypes. Genes that are mutated in 4 or more patients and are shared between T and GD-LGL (blue area) are labeled. (D-E) Single nucleotide variant class in the cohort and percent C>T mutation comparison in T and GD subtypes. (F-G) Microsatellite instability scores and SBS1 signature contributions in T and GD subtypes.

Figure 2.

Oncoplot of putative drivers in LGL leukemia. (A) Seven different driver analysis tools identified 15 gene variants with putative driver roles present in at least 3 LGL leukemia samples. Each row represents 1 of the 15 putative drivers, whereas the columns represent individual patients. LGL disease cohorts are color coded at the top. Type of alteration in putative driver genes is indicated by the colored boxes, which are defined under “Alterations,” and gray indicates WT. The top bar graph shows the number of amplification and deletion events. Right bar graph shows the distribution of variants for a given gene across the 2 LGL cohorts. Bottom annotations represent clinical phenotypes of individual patients, with color definitions found in the rightmost annotation. ANC values above 1.5k/µl and HGB values above 12 are annotated as “Normal,” and anything below is “Low.” Treatment indicates if patients were on LGL treatment (methotrexate, cyclophosphamide, or cyclosporine) at the time of sample acquisition. Unknown values are colored in gray. (B) Kyoto Encyclopedia of Genes and Genomes pathway analysis of somatic mutations observed in the cohort using SLAPenrich. The size of the nodes represents the number of genes in the cohort observed in the pathway. The color of the nodes represents the percent of those genes that uniquely belong in each pathway. The thickness of the edges represents the Jaccard index between the nodes.

Figure 3.

Lollipop plot of KMT2D, FAS, and PIK3R1 genes. Lollipop plot of (A) KMT2D, (B) FAS, and (C) PIK3R1 mutations detected in the 105-patient LGL leukemia cohort. Annotations for individual domains are described at the bottom. Domain information was curated from simple modular architecture research tool and conserved domain database.

Figure 4.

Significant co-occurrence of STAT3 with KMT2D or SETD1B mutations. (A) Comutation plot of all nonsilent mutations observed in the 105-patient LGL leukemia cohort. Numbers in brackets after gene names represent the number of mutations in this gene in the cohort. Significant pairs of genes were detected by pairwise Fisher’s exact test. (B-C) VAF density plots were used to infer clones or “clusters.” Each dot represents individual somatic variants observed in a given patient. The color of the dots represents inferred clusters. Colored annotations indicate different types of STAT3 mutations. (B) Four representative KMT2D VAF density plots. The first 2 demonstrate inclusion of both STAT3 and KMT2D in the same predicted clone. The following 2 plots demonstrate 2 mutations associating with 2 different clones. (C) VAF density plot of all patients harboring SETD1B mutations in the cohort.

Figure 5.

Novel activating STAT3 mutations in the coiled-coil domain and clinical phenotype of patients harboring specific STAT3 somatic mutations. (A) Lollipop plot of STAT3 somatic mutations detected in the 105-patient LGL leukemia cohort. Annotations for individual domains are described at the bottom. (B) STAT3 expression vectors, Q160P, L287F, D170Y, and Y640F as well as empty vector and WT STAT3 were transfected into HEK293 cells along with STAT3 luciferase reporter. Shown is a quantification of the relative ratio between firefly STAT3 responsive element against cytomegalovirus-controlled renilla luciferase. Two independent experiments were performed. Shown is 1 representative experiment. (C) Western blot of phospho-Y705 STAT3, total STAT3, mCherry, and vinculin from parallel whole-cell lysates obtained from transfection of empty vector, WT, and mutant STAT3 constructs as in panel B. Two independent experiments were performed. Shown is 1 representative experiment. (D-E) Plots of hematologic parameters ANC and HGB segregated by STAT3 WT, Y640F, D661Y, and N647I mutations. This includes 5 additional patients with N647I mutations identified using Sanger sequencing outside of the original 105-patient cohort. For panel D, Dunnett’s test with WT as a comparison control, P = .045, P = .059, and P = .853 for Y640F, D661Y, and N647I mutations, respectively. For panel E, Dunnett’s test with WT as a comparison control, P = .999, P = .074, and P = .047 for Y640F, D661Y, and N647I mutations, respectively. P values are indicated as follows: *P < .05; **P < .01; ***P < .001.

Figure 6.

Increased mutational burden in STAT3-mutated patients. (A) Boxplot of nonsilent mutational burden comparing STAT3 WT against mutant patients. P value from Wilcoxon t test. (B) Plot of the mutational burden against age for WT and mutant STAT3 patients with linear regression and colored annotation representing STAT3 mutation status. (C) Plot of mutation burden against CD3⁺ CD57⁺ fraction for T-LGL patients with linear regression and colored annotation representing STAT3 mutation status. (D) Top recurrent somatic mutations observed in STAT-mutated and WT groups. The left oncoplot shows patients with STAT3 or STAT5B mutations (n = 56), and the right oncoplot shows STAT WT (n = 49) patients. The top annotation indicates the LGL subtype. Patients with mutations in the chromatin-modifying enzyme gene list (supplemental Table 9) are annotated in the second row in gray. Three chromatin-modifying genes are highlighted in green font. The 10 genes that are most frequently mutated in WT patients are shown in the top half, followed by the 10 genes most frequently mutated in STAT3- and STAT5B-mutated patients in the bottom half. Annotation of “Alterations” indicates the type of mutation affecting the gene. SBS1 signature explains the degree of intensity of the COSMIC SBS1 signature present for each patient.

Figure 7.

Transcriptomic and protein expression comparison of STAT3-mutated patients against STAT3-WT patients. (A-B) RNA-seq analysis of T-LGL leukemia patients. (A) Volcano plot of DESeq2 output of T-LGL STAT3 mutant vs STAT3 WT. Genes with positive log₂ fold changes are more highly expressed in STAT3 mutants, and genes with negative log₂ fold changes are more highly expressed in STAT3 WT. (B) Gene set enrichment analysis of the nodes from the functional modular network analysis. Positive enrichment scores indicate pathways that are positively enriched in STAT3-mutant patients. Negative enrichment scores indicate pathways enriched in STAT3-WT patients. (C) RT-qPCR comparison of mRNA levels of ZBTB46, AKT3, and PDGFB from CD8⁺-isolated normal controls (n = 3; ZBTB46 n = 2 due to signal below the limit of detection in 1 sample), WT (n = 6; ZBTB46 n = 5 due to signal below the limit of detection in 1 sample), and STAT3-mutant LGL patient samples (n = 5). Analysis of variance (ANOVA) P = .028, .014, and .018 for ZBTB46, AKT3, and PDGFB, respectively. (D) Western blotting of pS473 AKT, total AKT, phospho-p44/42 MAPK, total p44/42 MAPK, pS9 GSK3β, and total GSK3β from CD8⁺-isolated normal controls (n = 3) compared with WT and STAT3-mutant LGL patient samples (n = 5). STAT3 mutation type and LGL registry ID are indicated. (E) RT-qPCR comparison of mRNA levels of STAT3, STAT1, and PDGFRB from CD8⁺-isolated normal controls, WT, and STAT3-mutant LGL patients. ANOVA P = .049, .963, and .026 for STAT3, STAT1, and PDGFRB, respectively. (F) Western blotting of pSTAT3, total STAT3, pSTAT1, total STAT1, total PDGFRβ, and vinculin loading control from CD8⁺-isolated normal controls compared with WT and STAT3-mutant LGL patient samples. STAT3 mutation type and LGL registry ID are indicated. Welch’s or Brown-Forsythe ANOVA was used for all RT-qPCR analyses depending on the distribution of the data. Unpaired t with Welch’s correction as a post hoc test. P values are indicated as follows: *P < .05; **P < .01.

Figure 7.

Transcriptomic and protein expression comparison of STAT3-mutated patients against STAT3-WT patients. (A-B) RNA-seq analysis of T-LGL leukemia patients. (A) Volcano plot of DESeq2 output of T-LGL STAT3 mutant vs STAT3 WT. Genes with positive log₂ fold changes are more highly expressed in STAT3 mutants, and genes with negative log₂ fold changes are more highly expressed in STAT3 WT. (B) Gene set enrichment analysis of the nodes from the functional modular network analysis. Positive enrichment scores indicate pathways that are positively enriched in STAT3-mutant patients. Negative enrichment scores indicate pathways enriched in STAT3-WT patients. (C) RT-qPCR comparison of mRNA levels of ZBTB46, AKT3, and PDGFB from CD8⁺-isolated normal controls (n = 3; ZBTB46 n = 2 due to signal below the limit of detection in 1 sample), WT (n = 6; ZBTB46 n = 5 due to signal below the limit of detection in 1 sample), and STAT3-mutant LGL patient samples (n = 5). Analysis of variance (ANOVA) P = .028, .014, and .018 for ZBTB46, AKT3, and PDGFB, respectively. (D) Western blotting of pS473 AKT, total AKT, phospho-p44/42 MAPK, total p44/42 MAPK, pS9 GSK3β, and total GSK3β from CD8⁺-isolated normal controls (n = 3) compared with WT and STAT3-mutant LGL patient samples (n = 5). STAT3 mutation type and LGL registry ID are indicated. (E) RT-qPCR comparison of mRNA levels of STAT3, STAT1, and PDGFRB from CD8⁺-isolated normal controls, WT, and STAT3-mutant LGL patients. ANOVA P = .049, .963, and .026 for STAT3, STAT1, and PDGFRB, respectively. (F) Western blotting of pSTAT3, total STAT3, pSTAT1, total STAT1, total PDGFRβ, and vinculin loading control from CD8⁺-isolated normal controls compared with WT and STAT3-mutant LGL patient samples. STAT3 mutation type and LGL registry ID are indicated. Welch’s or Brown-Forsythe ANOVA was used for all RT-qPCR analyses depending on the distribution of the data. Unpaired t with Welch’s correction as a post hoc test. P values are indicated as follows: *P < .05; **P < .01.