HSP and CD279 gene expression as candidate biomarkers in symptomatic LGLL patients Authors: Giovanna Talarico, Andrea Franceschini, Alessandro Raveane, Paolo Falvo, Saveria Mazzara, Federica Melle, Giovanna Motta, Stefania Orecchioni, Annamaria Tenore, Giuliana Gregato, Claudia Poletti, Roberto Chiarle, Stefano Pileri, Patrizia Mancuso, Francesco Bertolini Published: 2024 ABSTRACT The clinical presentation of T-cell large granular lymphocytic leukemia (T-LGLL) is extremely variable: 30% of patients have neutropenia with no associated symptoms, others present with bacterial infections and sepsis may occur. Tools to predict patient outcome are lacking. Stemming from preliminary results obtained by single cell-RNAseq we investigated by qPCR HSP and IFIT gene families in 27 LGLL patients (23T-LGLL and 4 NK-LGLL), including 11 with neutropenia and/or thrombocytopenia and 16 asymptomatic for the disease. HSP90AA1 and HSPA1B, among HSP family and CD279 exhibited a significantly higher expression in CD3 + CD57 + sorted cells of symptomatic LGLL patients compared to asymptomatic patients and healthy controls. Also, monocytes derived from symptomatic LGLL patients expressed high levels of CCL3, CCL4 and CCL5 mRNA and of IL-1β, IL-6, TNF, and PD-L1 mRNA, thus confirming a pro-inflammatory cytokine profile reminiscent of a non-classical phenotype. Overall, these data provide a rationale for considering HSP and CD279 genes as potential biomarkers for distinguishing symptomatic LGLL patients from asymptomatic ones, emphasizing the importance of further research to explore their implications for targeted therapy development.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12672-024-01657-y. Introduction T-cell large granular lymphocytic leukemia (T-LGLL) is a rare lymphoproliferative disorder, and based on the fifth edition of the World Health Organization (WHO) Classification of Haematolymphoid Tumours (Lymphoid Neoplasms), T-LGLL is classified within the mature T-cell and NK-cell neoplasms section [1–3]. T-LGLL is believed to be a stepwise process, but the etiology of this disease is still unknown. It has been hypothesized that T-LGL expansion is driven by chronic antigen exposure, likely a self or viral peptide, even though the peptide(s) involved in initial expansion has(have) not been discovered [4–6]. Leukemic lymphocytes have dysregulation in apoptosis and a secondary event is necessary to constitutively upregulate the cellular survival pathway leading to the prevalence of a clonal T-cell population [7–13]. Recently, constitutive activation of signalling proteins involved in survival pathways like Janus/STAT (JAK) has been demonstrated in T-LGLL patients, and STAT3 and STAT5b mutations are recognized as the most common gain-of-function genetic lesions up to now, and are identified and recognised as the hallmark of this disorder [14–16]. STAT3 mutations (usually Y640F and D661Y) have been described in 30–40% of CD8 + Tα/β-LGLL and Tγ/δ-LGLLs characterized by a symptomatic disease often associated with neutropenia [17, 18], and reduced overall survival, as compared to the other LGLL cases [19, 20]. By contrast, STAT5b mutations are mostly associated with the indolent CD4 + Tα/β-LGLL and Tγ/δ-LGLL forms (incidence approximately 30–60%) [21–25] or with the uncommon aggressive variant of CD8 + Tα/β-LGLL [26]. The diagnosis of T-LGLL is confirmed by the presence of large granular lymphocytes, phenotypically similar to terminally differentiated effector memory T cells, with a surface CD3 + CD57 + immunophenotype and the monoclonal rearrangement of the T-cell receptor (TCR) gene. Clonality of T-LGL is essential to categorize this lymphoproliferation as a leukemia and to distinguish it from autoimmune or infectious disease. Clonality of T-LGL can be assessed by molecular studies and TCRγ gene analysis can be used for both αβ and γδ T cell lineages, as this gene is rearranged at an early stage of T cell development. To confirm clonality, flow cytometry analysis assessing Vβ TCR cell surface protein repertoire can also be used [27]. However current panels of monoclonal antibodies include about 75% of the Vβ spectrum, and this assay, even if highly specific, is less sensitive than the PCR tests [28, 29]. Most cases of T-LGLL express the αβ TCR and have a surface CD3 + /CD8 + /CD57 + immunophenotype. Less frequently the T-LGL are CD4 + , while the co-expression of CD8 and CD4 is rare. A small subset of cases expresses the γδ TCR. The clinical presentation of T-LGLL is extremely variable: about one-third of patients are asymptomatic at the time of diagnosis with incidental finding of neutropenia, and with no associated symptoms or infections. On the other hand, some patients may exhibit symptoms related to neutropenia with fever due to bacterial infections, and, in serious cases, sepsis may occur. Splenomegaly is reported at 20–50% frequency, and recurrent infection is seen in 15–39% of patients [30]. Anemia is also a frequent symptom seen in patients, and 10–30% of anemic patients are transfusion dependent. Thrombocytopenia is not commonly observed. Treatment with immunosuppressive drugs, mainly cyclosporine and methotrexate, is indicated for patients with symptomatic anemia, reduced neutrophil counts with recurrent infections, or autoimmune conditions such as rheumatoid arthritis (RA). There is no treatment indication for patients with an absolute neutrophil count (ANC) above 0.5 × 109/L and no associated symptoms. Once the diagnosis is established, patients are monitored for symptoms and disease progression regardless of treatment status. In order to investigate the immune landscape in T-LGLL patients, we performed single-cell RNA sequencing (scRNA-seq) on peripheral blood mononucleated cell (PBMCs) of 6 patients affected by T-LGLL and 2 Healthy Controls (HCs), and confirmed the preliminary results of scRNA-seq in 27 patients affected by LGLL (23 T-LGLL and 4 NK-LGLL) using qRT-PCR. To better characterize the patients undergoing scRNA-seq (6 patients and 2 HCs), we investigated tumour genetic characteristics, using a customized capture-based target sequencing panel, including coding exons and splice site of 111 genes recurrently mutated in T-cell lymphomas, Moreover, on a cohort of 9 out of 27 patients, we analysed by the Cytoscan HD chromosomal microarray recurrent chromosomal aberrations as CNA and LOH. Results ScRNA-seq identifies a cluster of activated monocytes enriched in patients with symptomatic disease We performed scRNA-seq on Peripheral Blood Mononucleated Cells (PBMCs) of 6 patients affected by T-LGLL and 2 HCs (Fig. 1).Fig. 1Single cell profiling of T-LGLL patients. a UMAP classification and manual cell type annotation. b Cell type distribution in patients groups (HD: controls, AD: Asymptomatic, SyD: Symptomatic, SeD: Symptomatic patients after secondary diseases). c Violin plots reflecting gene expression in the various cell types Single cell profiling of T-LGLL patients. a UMAP classification and manual cell type annotation. b Cell type distribution in patients groups (HD: controls, AD: Asymptomatic, SyD: Symptomatic, SeD: Symptomatic patients after secondary diseases). c Violin plots reflecting gene expression in the various cell types Four patients had symptomatic T-LGLL: two were under treatment due to severe neutropenia and anemia (Symptomatic Disease, SyD); two patients developed T-LGLL after treatment for B-NHL, but no treatment for T-LGLL was administered (Secondary Disease, SeD) and two patients had T-LGLL but no associated neutropenia, anemia or thrombocytopenia (Asymptomatic Disease, AD). At the time of scRNA-seq analysis, one patient with symptomatic LGLL developed after treatment for B-NHL was discarded (pt3, Table 1), because of a low number of cells after the sequencing analysis.Table 1Laboratory and clinical features of patients with LGLLPatientSexAge (at diagnosis)Clinical presentationOther associated diseaseT or NK PhenotypeLGLL TreatmentClinical courseApplication in the study1F54NeutropeniaNFTMTXCR after treatmentscRNA-seq, RT-PCR, CS, SM2M83NeutropeniaNFTCTXCR after treatmentscRNA-seq, SM3F74NeutropeniaSMZLTNFIndolent diseasescRNA-seq, RT-PCR, SM4M60NeutropeniaSMZLTNFIndolent diseasescRNA-seq, CS, SM5M69Moderate LymphocytosisNFT and TCR gamma/delta + NFIndolent diseasescRNA-seq, RT-PCR, CS, SM6F73Moderate LymphocytosisNFT and TCR gamma/delta + NFIndolent diseasescRNA-seq, SM7M50Neutropenia and anemiaNFTNFIndolent diseaseRT-PCR8M57NeutropeniaMGUS IgGTNFIndolent diseaseRT-PCR9M47Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR, CS10F53Moderate LymphocytosisNFT and TCR gamma/delta + NFIndolent diseaseRT-PCR, CS11F52NeutropeniaAITDTNFIndolent diseaseRT-PCR, CS12F56NeutropeniaNFTNFIndolent diseaseRT-PCR13F46NeutropeniaNFTNFIndolent diseaseRT-PCR14F28Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR15F52Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR16M53Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR, CS17M42Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR18F68Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR, CS19F68Moderate LymphocytosisMBCLTNFIndolent diseaseRT-PCR, CS20F48LymphopeniaRATNFPancitopeniaRT-PCR21F51NeutropeniaNFTNFNeutropeniaRT-PCR22F65Neutropenia an AnemiaMZBLTCTXDeath for T-LGLLRT-PCR23F62B-LNHB-NHLTNFIndolent diseaseRT-PCR24F52LeucopeniaNFNKNFLeucopeniaRT-PCR25M75Routinary ExaminationAITDNKNFIndolent diseaseRT-PCR26M82Lymphocytosis and MGUS IgGNFNKNFIndolent diseaseRT-PCR27F28Routinary ExaminationNFNKNFIndolent diseaseRT-PCRM, male; F, femaleSMZL: Splenic marginal zone lymphoma; MGUS, Monoclonal gammopathy of undetermined significance; AITD, Autoimmune thyroid diseases; RA, Rheumatoid Arthritis; NF, not found; MBCL, Monocytoid B-cell lymphoma; MZBL, Marginal zone B-cell lymphomas; NHL, B-cell non-Hodgkin lymphoma; MTX, Methotrexate; CTX, Cyclophosphamide; CR, Complete response to treatment; CS, CytoScan; SM, STAT mutation Table 1Laboratory and clinical features of patients with LGLLPatientSexAge (at diagnosis)Clinical presentationOther associated diseaseT or NK PhenotypeLGLL TreatmentClinical courseApplication in the study1F54NeutropeniaNFTMTXCR after treatmentscRNA-seq, RT-PCR, CS, SM2M83NeutropeniaNFTCTXCR after treatmentscRNA-seq, SM3F74NeutropeniaSMZLTNFIndolent diseasescRNA-seq, RT-PCR, SM4M60NeutropeniaSMZLTNFIndolent diseasescRNA-seq, CS, SM5M69Moderate LymphocytosisNFT and TCR gamma/delta + NFIndolent diseasescRNA-seq, RT-PCR, CS, SM6F73Moderate LymphocytosisNFT and TCR gamma/delta + NFIndolent diseasescRNA-seq, SM7M50Neutropenia and anemiaNFTNFIndolent diseaseRT-PCR8M57NeutropeniaMGUS IgGTNFIndolent diseaseRT-PCR9M47Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR, CS10F53Moderate LymphocytosisNFT and TCR gamma/delta + NFIndolent diseaseRT-PCR, CS11F52NeutropeniaAITDTNFIndolent diseaseRT-PCR, CS12F56NeutropeniaNFTNFIndolent diseaseRT-PCR13F46NeutropeniaNFTNFIndolent diseaseRT-PCR14F28Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR15F52Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR16M53Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR, CS17M42Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR18F68Moderate LymphocytosisNFTNFIndolent diseaseRT-PCR, CS19F68Moderate LymphocytosisMBCLTNFIndolent diseaseRT-PCR, CS20F48LymphopeniaRATNFPancitopeniaRT-PCR21F51NeutropeniaNFTNFNeutropeniaRT-PCR22F65Neutropenia an AnemiaMZBLTCTXDeath for T-LGLLRT-PCR23F62B-LNHB-NHLTNFIndolent diseaseRT-PCR24F52LeucopeniaNFNKNFLeucopeniaRT-PCR25M75Routinary ExaminationAITDNKNFIndolent diseaseRT-PCR26M82Lymphocytosis and MGUS IgGNFNKNFIndolent diseaseRT-PCR27F28Routinary ExaminationNFNKNFIndolent diseaseRT-PCRM, male; F, femaleSMZL: Splenic marginal zone lymphoma; MGUS, Monoclonal gammopathy of undetermined significance; AITD, Autoimmune thyroid diseases; RA, Rheumatoid Arthritis; NF, not found; MBCL, Monocytoid B-cell lymphoma; MZBL, Marginal zone B-cell lymphomas; NHL, B-cell non-Hodgkin lymphoma; MTX, Methotrexate; CTX, Cyclophosphamide; CR, Complete response to treatment; CS, CytoScan; SM, STAT mutation Laboratory and clinical features of patients with LGLL M, male; F, female SMZL: Splenic marginal zone lymphoma; MGUS, Monoclonal gammopathy of undetermined significance; AITD, Autoimmune thyroid diseases; RA, Rheumatoid Arthritis; NF, not found; MBCL, Monocytoid B-cell lymphoma; MZBL, Marginal zone B-cell lymphomas; NHL, B-cell non-Hodgkin lymphoma; MTX, Methotrexate; CTX, Cyclophosphamide; CR, Complete response to treatment; CS, CytoScan; SM, STAT mutation A graph-based global-level clustering approach applied on uniform manifold approximation projection (UMAP) revealed 13 different homogenous groups of cells (Fig. 1a, Fig. S1.) Each group was named as a cell population using both automatic and canonical cell-type gene marker assignments (Fig. 1a and c). Three different clusters of monocytes, two of which expressed CD14 (i.e. classical) and one CD16 appeared well separated from all the other clusters and we reported their proportions in each of the sub-groups studied (Fig. 1b). Interestingly we observed that only one of the two CD14 clusters was found enriched in patients with symptomatic T-LGLL undergoing treatment (CD14-Mono-2), while the other CD14 cluster (CD14-Mono-1) was represented in equal proportion among HCs and asymptomatic T-LGLL patients. We could identify several differentially expressed genes by the CD14-Mono-2 cluster, when compared against all the other clusters (supp. Table S1). IL-1B, TNF alfa, CCL3 and CCL4 were the most overexpressed cytokines in the CD14-Mono-2 cluster. CD16 and CCR5 were not expressed in CD14-Mono-1 and CD14-Mono-2 clusters further confirming the classical nature of these monocyte clusters [31]. We also identified various clusters of T Cells (e.g. CD4, CD8, TNK), Memory B cells, Dendritic Cells and Platelets (Fig. 1), but at this level of resolution none of them appeared strongly over represented in a specific group of patients (with the only exception of a specific group of CD8 T-cells over-represented in our secondary T-LGLL patient and that we are going to better describe in lower-level CD8 clustering analysis. High Resolution scRNA-seq of CD8 and NK cells identify a cluster of CD8 + cells with HSP over-expression in symptomatic T-LGLL patients To elucidate the transcriptional profile of CD8 + T and NK cells and identify the predominant cell subsets in symptomatic patients, we re-clustered these cells (Fig. 2, Fig. S2, supp. Table S2).Fig. 2Single cell profiling of T-LGLL patients (CD8 and NK subclustering). a UMAP classification and manual cell type annotation. b Violin plots reflecting gene expression in the various cell types. c cell type distribution in patient groups (HD: controls, AD: Asymptomatic, SyD: Symptomatic, SeD: Symptomatic patients after secondary diseases) Single cell profiling of T-LGLL patients (CD8 and NK subclustering). a UMAP classification and manual cell type annotation. b Violin plots reflecting gene expression in the various cell types. c cell type distribution in patient groups (HD: controls, AD: Asymptomatic, SyD: Symptomatic, SeD: Symptomatic patients after secondary diseases) A first cluster that we named CD8 HSP was exclusively present in all three T-LGLL symptomatic patient’s cells and no HCs or asymptomatic patient’s cells. Genes of four different HSP families were found strongly over-expressed in the cluster: Hsp90 family (HSP90AA1, HSP90AB1), Hsp70 family (HSPA1B, HSPA1A), Hsp40 (DNAJB6, DNAJA1) and Hsp10 family (HSPE1). A second cluster that we named CD8 IFIT populated almost all the CD8 + cell pool of the symptomatic patient who developed T-LGLL after B-NHL treatment, while cells of the other two symptomatic patients undergoing treatment harboured a lower amount (around 7–8%) of this population. In this cluster the IFIT gene family (IFIT1, IFIT2 and IFIT3) [36] were strongly over-expressed together with ISG15 and OASL interferon-induced genes and PMAIP1 apoptotic gene. The other 10 clusters involved groups of CD8 + Tcells or NK cells that are also found in controls and asymptomatic T-LGLL patients. In particular, we distinguished a cluster mainly composed of gamma-delta cells in asymptomatic patients based on the overexpression of gamma-delta TCR genes. Furthermore, we could clearly recognize a cluster of CD8 central memory T cells characterized by the over-expression of CCR7 and SELL genes and an Effector Memory cluster where CD27 and CD44 markers were expressed [43]. High resolution scRNA-seq of CD4 and monocytes Finally, we re-clustered CD4 and monocytes cells at a higher resolution (Fig. S3 and Fig. S4, supp. Table S3). We grouped CD4 cells in four different clusters classified as CD4 Naive, CD4 Memory, CD4 KLRB1 High and CD4 Effector based on the expression of the four markers (SELL, CCR7, GZMB, KLRB1) [44, 45]. The CD4 Effector cluster was populated mainly by the cells of patient who developed T-LGLL after treatment for B-NHL. Among its differentially expressed genes, we found NKG7, GNLY, GZMB and GZMH (cytolytic molecules), as well as CCL4 and CCL5 chemokines and immunoregulatory gene CST7 [46]. The Epithelial to Mesenchymal transition driver ZEB2 and FGFBP2 growth factor were also strongly over-expressed [47]. Among monocytes, we identified several clusters similar to the peripheral blood monocyte clusters of healthy subjects described in the single cell studies by Rigamonti et al. and Mulder et al. [48, 49]. Our CD14-Mono2 cluster shares 9 of its top 40 genes with the cMo_infl_3 cluster in Rigamonti et al.’s study [48] which they found to be prognostic of improved survival among gastric-cancer patients. CD14-Mono2 also shares 11 of the top 40 genes with cluster 15 of the Mulder et al. single-cell meta-analysis study [49]. Our second cluster that was populated by cells of HCs and asymptomatic patients share similarities to their cMo_infl_1 cluster. In this cluster AP-1 transcription factor complex members FOS and FOSB [50] were overexpressed together with ABHD5, RALGAPA1 and S100A8 genes [51, 52]. Looking at the non-classical monocyte clusters our CD16 + + cluster 4 (similar to the ncMo cluster of Rigamonti et al.) strongly expressed the kinase inhibitor CDKN1C (found to negatively regulate cell proliferation) together with MS4A7 and RHOC (which instead promotes cell proliferation) [53–55]. Our intermediate monocyte cluster 5 (similar to iMo cluster in Rigamonti et al.) contained human leukocyte antigens genes corresponding to MHC class 2 together with CTSS and CD74 genes, involved in the MHC class 2 pathway [56–58]. Cluster 7 (similar to the cMo MPA cluster in Rigamonti et al.) contained mostly platelet-related genes such as PPBP, PF4, CAVIN2 and TUBB1 [59, 60]. Apart from our CD14-Mono2 cluster, none of the other monocyte clusters appeared to be enriched in symptomatic T-LGLL patients. Gene expression analysis via qRT-PCR unveils distinct profiles comparing symptomatic and asymptomatic LGLL-Derived T Cells with T-cells of healthy controls A qPCR approach was developed and studied in 27 patients affected by LGLL and 12 HCs to validate scRNA-seq data and identify candidate genes capable of discriminating between symptomatic and asymptomatic LGLL patients. We defined as symptomatic patients with neutropenia (< 1500 neutrophils/uL), and/or anemia (< 12 gr/dl), and/or thrombocytopenia (< 140.000 plt/yL9) and asymptomatic those patients with no neutropenia, no anemia, and no thrombocytopenia. A total of 27 LGLL patients (23 T-LGLL and 4 NK-LGLL) and 12 HCs were included in the study. The clinical characteristics of patients are summarized in Table 1. Within the HSP family, HSP90AA1, HSPA1B and HSPA1A were analysed by qPCR and normalized using the b-actin (ACTB) gene as housekeeping. Among these, HSP90AA1 and HSPA1B exhibited significantly higher expression in symptomatic LGLL patients compared to both asymptomatic patients and HCs (Fig. 3A, B). The expression of HSPA1A showed a trend towards increased expression in symptomatic patients compared to asymptomatic ones (Fig. 3C).Fig. 3A–G qPCR showing the expression of HSP90AA1, HSPA1B, HSPA1A, IFIT2, IFIT3, OASL and PDCD1 on CD3 + CD57 + (T-LGLL patients) or CD3− CD56 + CD57 + (NK-LGLL patients) classified as symptomatic (n = 11) and asymptomatic (n = 16). From healthy control (n = 12) we sorted CD3 + CD57 + cells used as control. Statistical analysis: one-way ANOVA (* < 0.05; **p < 0.01) A–G qPCR showing the expression of HSP90AA1, HSPA1B, HSPA1A, IFIT2, IFIT3, OASL and PDCD1 on CD3 + CD57 + (T-LGLL patients) or CD3− CD56 + CD57 + (NK-LGLL patients) classified as symptomatic (n = 11) and asymptomatic (n = 16). From healthy control (n = 12) we sorted CD3 + CD57 + cells used as control. Statistical analysis: one-way ANOVA (* < 0.05; **p < 0.01) Regarding IFIT gene family, IFIT2 and IFIT3 were analysed. Both genes demonstrated a trend of increased expression in symptomatic LGLL patients compared to asymptomatic patients and HCs (Fig. 3D, E). OASL interferon-induced gene did not exhibit any significant differences in gene expression levels among the studied groups (Fig. 3F), while OASL does not appear to be significantly associated with LGLL symptomatology based on gene expression levels. We also examined the expression of CD279 on T and NK sorted T cells. Symptomatic LLGL patients expressed high level of CD279 compared to asymptomatic patients (Fig. 3G). This finding suggests that T cells in LGLL symptomatic patients are likely to be in an exhausted phenotype. Monocytes phenotype in symptomatic patients show a pro-inflammatory profile Since monocytes have been demonstrated to contribute to leukemic cell survival in the majority of T-LGLL subsets, we delved into exploring various genes associated with immune response in sorted monocytes (CD14 + , CD3- cells). We demonstrated by qPCR that monocytes derived from symptomatic LGLL patients express high levels, but not significantly, of CCL3, CCL4 and CCL5 compared to both asymptomatic patients and HCs (Fig. 4 A-C). Moreover, these monocytes produced high levels of IL-1β, IL-6, TNF, and CD274 mRNA, supporting a pro-inflammatory cytokine profile reminiscent of a non-classical phenotype (Fig. 4D–G).Fig. 4A–G mRNA levels of CCL3, CCL4, CCL5, IL-1β, IL-6, TNF and CD274 (qPCR) in monocytes of symptomatic (n = 11) and asymptomatic (n = 16) patients. Monocytes of healthy control (HC) (n = 12) were used as control. Statistical analysis: one-way ANOVA (*p < 0.05; **p < 0.01) A–G mRNA levels of CCL3, CCL4, CCL5, IL-1β, IL-6, TNF and CD274 (qPCR) in monocytes of symptomatic (n = 11) and asymptomatic (n = 16) patients. Monocytes of healthy control (HC) (n = 12) were used as control. Statistical analysis: one-way ANOVA (*p < 0.05; **p < 0.01) Remarkably, our analysis unveiled a lack of statistically significant disparities in the expression levels of HSP genes within monocytes obtained from symptomatic patients when compared with those sourced from asymptomatic ones (p > 0.05) (Fig. 5A–C), unlike the significance observed on lymphocyte cells of LGLL patients.Fig. 5A–C mRNA levels of HSP90AA1, HSPA1B and HSPA1A (qPCR) in monocytes (defined as CD3- CD14 +) of symptomatic (n = 11) and asymptomatic (n = 16) patients. Monocytes of healthy control (n = 12) were used as control. Statistical analysis: one-way ANOVA (**p < 0.01, ***p < 0.001) A–C mRNA levels of HSP90AA1, HSPA1B and HSPA1A (qPCR) in monocytes (defined as CD3- CD14 +) of symptomatic (n = 11) and asymptomatic (n = 16) patients. Monocytes of healthy control (n = 12) were used as control. Statistical analysis: one-way ANOVA (**p < 0.01, ***p < 0.001) Conversely, a significant increase of HSP gene expression was observed specifically in monocytes derived from HCs, in contrast to both symptomatic and asymptomatic patients. This intriguing revelation underscores a distinctive linkage of these genes with the pathogenesis of LGLL. Furthermore, our discoveries suggest an involvement of HSP genes in the dynamics of LGLL development and progression, shedding new light on the intricate molecular mechanisms underlying this disease. Mutational landscape To better dissect the genetic characteristics of the tumors of patients undergoing sc-mRNA-seq (6 patients and 2 HCs), tumour genomic DNA (gDNA) from each sample was extracted and analysed using a capture-based target sequencing panel, including coding exons and splice site of 111 genes recurrently mutated in T-cell malignancies. Even though our samples were a mixture of different cell types, we expected to be able to detect several somatic T-LGL mutations (even if at low allele frequency). For this reason, our approach was different from other recent works [14, 63], where sequencing had been performed on samples with higher leukemic fraction and on saliva as control. We first compared our mutations with those found in Cheon et al [63], looking at the precise position correspondence at nucleotide level. We identified STAT3–N647I missense mutation in a primary symptomatic LGL alfa-beta patient and mutation D661Y in one of our LGL gamma-delta asymptomatic patient. Both mutations occured in the SH2 protein domain of STAT3. This domain is essential for the STAT3 dimer formation that is required to mediate its translocation into the nucleus [63]. A mutation in the splice region of the FAS gene of a primary T-LGLL patient was also shared with a patient in the Cheon study and is also annotated in cosmic to be found on a patient of Primary Cutaneous Marginal Zone Lymphoma (cosmic id: COSV58239513). Then we looked broadly at all other mutations by applying the following filters:Exclude synonymous mutations and mutations on intron/non-coding regionsExclude mutations having a reported population frequency greater than 0.001 (using gnomeAD 2.1 exome data)Exclude mutations having an allele frequency higher than 0.45 (i.e. given our relatively low leukemic fraction we can consider them germline) Exclude synonymous mutations and mutations on intron/non-coding regions Exclude synonymous mutations and mutations on intron/non-coding regions Exclude mutations having a reported population frequency greater than 0.001 (using gnomeAD 2.1 exome data) Exclude mutations having a reported population frequency greater than 0.001 (using gnomeAD 2.1 exome data) Exclude mutations having an allele frequency higher than 0.45 (i.e. given our relatively low leukemic fraction we can consider them germline) Exclude mutations having an allele frequency higher than 0.45 (i.e. given our relatively low leukemic fraction we can consider them germline) In this way, we found 13 mutations on 9 different genes. Of these genes, 7 are also found mutated in three or more patients of Cheon’s cohort (Fig. 6).Fig. 6Oncoplot with the mutational profiles of our T-LGLL patients (A). In addition, the CBC parameters of the patients are reported (B). We show only the genes found mutated in at least three of the patients of the Cheon et al. study. Legend: H (High), N (Normal), L (Low) Oncoplot with the mutational profiles of our T-LGLL patients (A). In addition, the CBC parameters of the patients are reported (B). We show only the genes found mutated in at least three of the patients of the Cheon et al. study. Legend: H (High), N (Normal), L (Low) As in the Cheon study, we found mutations on the epigenetic modifiers’ genes KMT2D, TET2 and STAT3 interacting gene TNFAIP3. Surprisingly KMT2D harbours the same in-frame deletion on 2 of our 6 patient cohort. For the unique patient who developed T-LGLL after B-LNH treatment we also found a novel mutation in STAT3—SH2 domain (E616G), that in COSMIC database is reported in DLBC- and NK-T cell lymphoma samples (cosmic id: COSV52883602). Cytoscan analysis for CAN and LOH Nine out of 27 patients were analysed by the Cytoscan HD for chromosomal microarray recurrent chromosomal aberrations as CNA and LOH. No specific T-LGLL chromosomal aberration or mosaicism were repeatedly detected in our cohort of patients. Discussion T-LGLL, usually classified as a malignant condition due to the presence of monoclonal cytotoxic T cells, is at the crosswalk of neoplasia, chronic inflammation and autoimmune disease [64]; in fact, T-LGLL pathogenesis has some unique characteristics that are not completely understood in their mechanistic causes. A number of somatic mutations including STAT3, STAT5b, and TNFalpha-induced protein 3 have been reported in about half of T-LGLL patients; hence these tumor genetic characteristics cannot always explain the genesis of the disease. Patients with T-LGLL may be at increased risk of secondary primary tumors, and in particularly of B-cell malignancies and of developing autoimmune disorders such as rheumatoid arthritis, systemic lupus erythematosus and autoimmune hemolytic anemia. It is thought that this may be due to the abnormal activation of the immune system in T-LGLL. In this context, chronic antigenic stimulation is considered a hallmark of T-LGLL. It involves dysregulation of apoptosis, most likely because of constitutive activation of several survival pathways including, among others, Jak/Stat, MapK, Stat3, IL-15, phosphatidylinositol 3-kinase–Akt, Ras–Raf-1, MEK1, and NF-κB [26, 63–65]. In the present manuscript, based on preliminary results obtained by sc-RNAseq on 6 patients affected by T-LGLL and 2 HCs, we report the results of the molecular investigations of HSP and IFIT gene family in 27 LGLL untreated patients (23 T-LGLL and 4 NK-LGLL), including 11 patients with clinical manifestations of disease (neutropenia and or thrombocytopenia) and 16 patients with asymptomatic disease. A limitation of our study is the low number of samples undergoing single-cell analysis. However, the preliminary results were confirmed by qPCR on a larger cohort of untreated patients with different clinical manifestation of the disease. By means of sc-RNA investigations we identified two distinct clusters of CD8 cells, designated as CD8 HSP and CD8 IFIT, which exhibit differential gene expression patterns in T-LGLL patients. The CD8 HSP cluster was found to be the dominant population amongst the symptomatic T-LGLL patients’ T cells, with minimal representation in HCs and asymptomatic patients’ T-cells. Notably, this cluster exhibits marked upregulation of multiple HSP genes including Hsp90 family (HSP90AA1 HSP90AB1), Hsp70 family (HSPA1B, HSPA1A), Hsp40 (DNAJB6, DNAJA1) and Hsp10 family (HSPE1). HSPs, a large family of evolutionarily preserved molecular chaperones with pivotal roles in cell survival and cell replication [66], are constitutively expressed by external stresses, including heat, ischemia, heavy metals, ecotoxicity, and hypoglycemia. HSP have been characterized to counteract apoptosis [32, 33] and (only recently) to also contribute to DNA repair [34], are overexpressed in the blood of patients affected by Acute Lymphoblastic Leukemia and Acute Myeloid Leukemia (ALL and AML) as well as in several leukemia cell lines [35]. During the years, several HSP inhibitors have been developed and tested in different clinical trials [7, 8] for various types of cancers. As to the best of our knowledge, HSPs involvement in the pathogenesis of symptomatic T-LGLL has not been described before. We developed a qPCR approach to validate these results on the T-cells of a larger cohort of patients (n = 27). Within the HSP family, HSP90AA1 and HSPA1B exhibited significantly higher expression in symptomatic LGLL patients compared to both asymptomatic patients and HCs. The expression of HSPA1A showed a trend towards increased expression in symptomatic patients compared to asymptomatic ones. Recent studies have implicated IFIT genes as prognostic markers in several cancers [37–40] such as glioblastoma, hepatocellular carcinoma, breast and pancreatic cancers. Besides, in a recent study, IFIT2 has been found overexpressed in Chronic Myeloid Leukemia cells [41], where IFIT genes have been found to reduce the cell-migration activity affecting the epithelial-mesenchymal transition (EMT) pathway [36], while IFIT2 has been characterized to promote pro-apoptotic activity [42]. Overall, these findings suggest that HSP and IFIT genes may serve as potential biomarkers for distinguishing symptomatic LGLL patients from asymptomatic ones, while OASL does not appear to be significantly associated with LGLL symptomatology based on gene expression levels. The immune checkpoint CD279 (PD-1) is expressed in T-cells upon activation and remains expressed along their exhaustion state [61]. In healthy subjects, PD-1 triggering by the cognate ligand PD-L1 protects from autoimmune-related events whereas in the tumor context its triggering is detrimental causing deactivation/inhibition of T cell response [62]. Therefore, we examined the expression of CD279 on T and NK sorted T cells. Symptomatic LLGL patients expressed high level of CD279 compared to asymptomatic patients and may correlate with a more aggressive disease course, reflecting a dysfunctional immune response. However, more studies are necessary to examine the exact role of CD279 in T-LGLL, aiming to clarify its significance in disease progression and potential therapeutic interventions. Interestingly, its ligand PD-L1 (CD274) is overexpressed on sorted monocytes of symptomatic patients when compared to both asymptomatic patients and HCs, suggesting the presence of an immunosuppressive microenvironment status promoting the expansion of T-LGL clones. Regarding the myeloid compartment in T-LGLL patients, by sc-RNA we found that one of the two CD14 + peripheral blood monocyte cell clusters was enriched in patients with symptomatic T-LGLL. Interestingly, when testing HSP and IFIT genes on sorted monocytes, we discovered a specific correlation between the expression of HSP and T cells of symptomatic LGLL patients, hinting at the unique association of these genes with LGLL pathogenesis. Moreover, HSP genes in sorted monocytes were significantly downregulated in T-LGLL patients when compared to healthy controls, but upregulated in T sorted cells of patients when compared to healthy controls. This finding suggests a distinctive association between these genes and the pathogenesis of LGLL. One can speculate that this imbalance in HSP gene expression between monocytes and T cells might promote T-LGLL clonal expansion. We could identify several differentially expressed genes in this cluster: IL-1ΒB, TNF alfa, CCL3 and CCL4 were the most overexpressed cytokines in the CD14-Mono-2 cluster. We demonstrated by qPCR that monocytes derived from symptomatic LGLL patients express high levels of CCL3, CCL4 and CCL5 compared to both asymptomatic patients and HCs. Moreover, these monocytes produced high levels of IL-1β, IL-6, TNF, and PD-L1, suggesting that monocytes isolated from symptomatic patients exhibit a pro-inflammatory cytokine profile reminiscent of a non-classical phenotype. The production of IL-1β cytokine is probably stimulated by the canonical inflammasome initiated by the NLRP3 pattern recognition receptor, also found overexpressed [67, 68]. IER3 overexpression, induced by TNF overexpression mediated by NFKBIA, has been previously characterized to confer cell resistance to TNF-induced apoptosis [69, 70]. A similar role is accomplished by HSPA1A and AIF1 proteins (both overexpressed) [71]. Different evidence in the literature supports the hypothesis that IL-1ΒB and TNF alfa induce the production of CCL3 and CCL4 chemokines via NFKBIA mediation [72–75]. The activation of such immune response could arise by TLR4 stimulation, found mildly overexpressed together with TLR2. As in Vincenzetto et al [76], we found IL-6 strongly overexpressed among our symptomatic patients’ monocytes. A high IL-6 level can contribute to STAT3 activation, as described in several works [77, 78]. This supports the hypothesis of monocytes being one of the main sources of IL-6, acting in trans on other cell types through binding to the soluble form of IL-6R. Drugs that inhibit IL-6 action such as Tocilizumab and Siltuximab may represent a valid approach in refractory LGLL that could deserve a proper clinical investigation. When investigated for DNA aberrations, our patient population showed a mutational landscape not significantly different from those already reported in the literature. We found 47 mutations in 23 different genes; among these, 11 were also found mutated in three or more patients reported in another cohort [63]. As in that manuscript, we found mutations on the epigenetic modifiers’ genes KMT2D, TET2 and STAT3-interacting gene TNFAIP3. Notably, KMT2D harboured the same in frame deletion in two of our patients. We also found a mutation in a STAT3–SH2 domain (E616G), that in the COSMIC database is reported in DLBCL and NK-T cell lymphoma but not in T-LGLL patients. We also report here for the first time an investigation of DNA CNA and CN-LOH in T-LGLL patients by means of the Cytoscan approach [79]. Regions of CNA > 400 Kb and CN-LOH > 4 Mb identified by ChAS software were denoted as true aberrations. Aberrations that were identified were described according to ISCN 2018 in the Results section. No recurrent alterations have been identified so far, but we are investigating more T-LGLL patients and will report a larger series of T-LGLL DNA CNA and CN-LOH aberrations as soon as results are available. Conclusions Taken together, these findings provide novel insights into the molecular heterogeneity of T-LGLL and may have important implications for the development of targeted therapies involving HSP and IFIT genes and their role in T-LGLL prognosis. Future studies should aim to further elucidate the functional relevance of these differentially expressed genes and their potential role in disease pathogenesis. Along a similar line we will investigate their potential as predictive/prognostic biomarkers in T-LGLL. Materials and methods Clinical and laboratory features of patients evaluated with sc-mRNA analysis All patients evaluated in the study signed an appropriate informed consent. Patients with symptomatic T-LGLL undergoing treatment (pt#1 and pt#2) At the time of diagnosis both symptomatic patients presented with severe neutropenia (N 400/uL) and severe anemia (Hb < 8gr/dl). The phenotype of pathological T cells was CD3 + , CD8 + , CD57 + , CD5- and CD7 weakly positive. Molecular analysis of TCR rearrangements by PCR and flow cytometry confirmed the monoclonal nature of the cells. Patient#1 was treated with Metotrexate 5mgx2/day for more than 8 years since diagnosis and showed a very good response. For this patient, at the time of sampling for sc-mRNA analysis, absolute neutrophil count was 2100/ul, Hb12.9 gr/dl. Circulating monoclonal T-LGL were 174/uL, STAT3 mutation was detected, while Stat5B was wild type. Patient#2 before treatment was transfusion-dependent for severe anemia, and he was treated with CTX 50 mg/day for 18 months, with a very good response to treatment. For this patient, at the time of sampling for sc-mRNA analysis, neutrophil count was 3130/ul, Hb 14.6 gr/dl. Circulating monoclonal T-LGL were 167/uL. Wild type STAT3 or STA5b were detected. Patients with symptomatic TLGLL developed T-after treatment for B-LNH (T-LGLL, pt#3 and pt#4) We performed sc-mRNA on two patients, but one patient (pt#3) was discarded due to a low number of cells after the sequencing analysis. Patient#4 was treated for Marginal Zone B—NHL with R-Bendamustine and 4 months after the end of treatment, during a routine blood count, severe neutropenia was detected (absolute neutrophil count: 660/uL). Peripheral blood immunophenotype by flow cytometry confirmed the presence of T -lymphocytes CD3 + , CD57 + , CD4-, CD8 + , CD5 weakly positive, CD7 + , CD2 + , and TCR clonality of CD3 + CD8 + cells was detected by flow-cytometry. STAT3 mutation was detected, while Stat5B was wild type. At the time of sampling for sc-mRNA, circulating monoclonal T-LGL were 255/ul. Patients with asymptomatic T-LGLL (pt#5 and pt#6) At the time of diagnosis, pt#5 had moderate lymphocytosis, but no anemia or thrombocytopenia were detected. Peripheral blood immunophenotype by flow cytometry confirmed the presence of gamma/delta T -lymphocytes CD3 + , CD4-, CD8-, CD5 weakly positive, CD7-, CD2 + . Molecular analysis of TCR rearrangements by PCR confirmed the monoclonal nature of the cells; wild type STAT3 or STA5b were detected. At the time of sampling for sc-mRNA, circulating monoclonal T-LGL were 739/ul. At the time of diagnosis, pt#6 had moderate lymphocytosis (absolute lymphocyte count 7700 uL) but no associated anemia, no thrombocytopenia. Peripheral blood immunophenotype by flow cytometry confirmed the presence of gamma/delta T -lymphocytes CD3 + , CD57 + , CD4-, CD8-, CD5-, CD7-, CD2 + . Molecular analysis of TCR rearrangements by PCR confirmed the monoclonal nature of the cells. STAT 3 was mutated, STAT5b was wild type. At the time of sampling for sc-mRNA, circulating monoclonal T-LGL were 3927/ul. For both patients, no treatment was ever required (median follow-up 5 years). Clinical presentation of patients undergoing RT-PCR The clinical characteristics of 27 patients affected by LGLL who underwent RT-PCR confirmation are summarized in Table 1. Flow cytometry analysis Blood samples were obtained from 27 T-LGLL patients for diagnostic purpose. Flow cytometry was performed on whole blood, with a stain, lyse and wash procedure and samples were collected on a Beckman Coulter Navios-EX 10 Colors Instrument (Beckman Coulter) and since 2021 on DxFlex 13 Colors instrument (Beckman Coulter). Data were analysed using Kaluza software (Beckman Coulter). Flow cytometry analysis of the TCR Vβ repertoire TCR Vβ repertoires of patients and healthy donors were determined using flow cytometry with the IOTest Beta Mark TCR Repertoire kit (Beckman Coulter), coupled with anti-human CD3 Alexa Fluor 750 monoclonal antibody (MoAb, 5uL neat, Beckman Coulter, clone UCHT1, Catalog A66329), anti-human CD4 APC MoAb, (5uL neat, Beckman Coulter, clone 13B8.2, CATOLOG IM2468) and CD8 PC7 MoAb, 5uL neat (Beckman Coulter, clone SFCI21Thy2D3, Catalog 6,607,102). Data acquisition was performed on a Beckman Coulter Navios 10 Colors Instrument and since 2021 on DxFlex 13 Colors instrument. Data were analysed using Kaluza software (Beckman Coulter). PCR analysis of T cell receptor–γ gene rearrangements. Evidence of T-cell clonality was evaluated using T-Cell Receptor Gamma Gene Rearrangement Assay 2.0 (Invivoscribe) according to the EuroClonality/BIOMED-2 guidelines [80]. This is a routine method that includes polymerase chain reaction (PCR) analysis of T-cell receptor-γ (TCR-γ) gene rearrangements followed by Genescan analysis and the results were analysed using Gene Mapper v5.0 software. Sc-mRNA sample preparation For sc-mRNA sequencing, peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll-Hypaque density gradient centrifugation, and isolated PBMCs were cryopreserved in liquid nitrogen according to standard protocols until use. After thawing of cryopreserved PBMC, to remove dead cells for sc-mRNAseq preparation, we took advantage of Dead Cell Removal Kit (Miltenyi Biotech, 130–030–101) following manufacturer’s instruction. 10 k alive cells were then processed with 10 × Technology using V3.1 kit (10 × Genomics), following manufacturer’s instructions. Finally, library integrity was assessed by Hypersensitive (HS) Bioanalyzer run (Agilent). Sc-mRNA data processing and quality control After quality control, we integrated a dataset of 18,139 cells (a mean of 2,267 per sample) and 22,120 genes consisting of seven patients (one was discarded due to a low number of cells after the sequencing analysis). No confounding factors (e.g. batch effect, cell cycle bias) were detected. Raw FASTQ files were converted to gene-cell count matrices employing CellRanger v3 software package. As references, the human genome GRCh38 and GENCODE (release 32) were used and the steps were embedded in a Snakemake pipeline [81]. Quality controls included removal of cells having less than 500 and more than 50,000 transcript, less than 250 genes, a complexity score (log10 genes / transcript) lower than 0.80 and more than 15% mitochondrial transcript, as described in Carpen et al. 2022 [81] and Falvo et al. 2021 [82]. Doublets were detected and removed using scDblFinder package [83] (Fig. S5 and S6). No large differences were found in the cell cycle score when evaluated with Seurat v4 and by using genes downloaded from the Ensembldb R package [84]. Therefore, after evaluation of different approaches we opted for the sctransform-based normalization [85] regressed out for the percentage of mitochondrial transcript as the best method. In order to clean our cells for potential haemoglobin cell contamination, we checked for haemoglobin genes expression counts and filtered out the cells where these genes represented more than 5% of the counts (Fig. S7). Uniform Manifold Approximation and Projection (UMAP) method in Seurat v4 was computed using the first 30 Principal Components (PC) of the merged gene-cell count matrix. The Elbow method detected 15 as the best number of PCs to use for the clustering; these were used for the graph-based clustering approach in Seurat v4. Finally, the best granularity was evaluated by visual inspection and the cells were plotted in a two-dimensional space. Cell assignation of scRNA-seq data Different manual and automatic approaches were used for assigning a specific cell population to the retrieved clusters. First, we explored the automatic assignment of SingleR [86] results using four different databases [87–90]; second, we investigated differential expressing genes in each cluster, retrieved with FindMarker function of Seurat v4 as indicated below, on PanglaoDB [91]; finally we manually explored the known-marker genes for each of the clusters. The clusters were then named accordingly. Sub-clustering analysis Monocytes were filtered out by selecting cells that fall on clusters classified as “Monocytes” in the first cell assignment. Subsequently, a graph-based re-clustering of these subset was performed after a PCA on normalised cell counts in Seurat v4, as indicated above and as performed in Carpen et al. 2022 [81]. The best granularity for Monocyte sub-clustering was evaluated by visual inspection and set as 0.4. Lymphocyte sub-clustering was performed as indicated above by filtering cells classified as T and NK cells in the cell assignment of the whole immune landscape. Furthermore, cells with an expression of CD4 and CD8A/CD8B/FCGR3A higher than zero were retained for the CD4 and CD8/NK re-clustering, respectively. The most-efficient granularity was set as 0.3 for CD4 and 0.1 for CD8/NK. The name of the sub-clusters was assigned according to the expression of known-gene markers taken from the literature and evaluation of differentially expressed genes (DEG) after removal of mitochondrial and ribosomal genes. Sorting of T and monocyte cell subsets PBMCs previously cryopreserved were thawed and then stained to isolate leukemic and monocyte cells by fluorescence-activated cell sorting (FACS ARIA xxx, BD Biosciences). In detail, PBMC of patients affected by T-LGLL were sorted as CD3 + CD57 + , while PBMC of patients with NK-LGLL patients were sorted as CD3- CD56 + CD57 + . For all 12 healthy donors we isolated the counterpart CD3 + CD57 + . The following antibodies were used: anti-CD3-AF750, anti-CD14-Pc5.5, anti-CD45-KO, anti-CD56-AF700, anti-CD57-FITC (all Beckman Coulter). Monocytes were sorted for all samples as CD14 + . Differential gene expression between sorted cell populations was analysed using reverse transcription real-time PCR (qPCR), described below. Analysis of differential gene expression by reverse transcription real-time PCR (qPCR) Total RNA was extracted using the Quick-RNA Microprep Kit (Zymo Research) according to the manufacturer’s protocol and subsequently quantified by NanoDrop 2000c Spectrophotometer (Thermo Scientific). Synthesis of cDNA was performed with the MultiScribe-Reverse Transcriptase kit (Applied Biosystems). Target gene expression was analysed using TaqMan Universal PCR Master Mix (Applied Biosystems). Quantitative PCR was performed using the following TaqMan assays: Hsp90aa1 (Hs00743767_sH), Hspa1a (Hs00359163_s1), Hspa1b (Hs01040501_sH), Pdcd1 (Hs01550088_m1), Ifit2 (Hs01584837_s1), Ifit3 (Hs01922752_s1), Oasl (Hs00984387_m1), IL-6 (Hs00174131_m1), IL-1b (Hs01555410_m1), Ccl3 (Hs00824185_s1), Ccl4 (Hs99999148_m1), Ccl5 (Hs00982282_m1), Tnf (Hs00174128_m1), CD274 (Hs00204257_m1) and Actb (Hs01060665_g1) used as housekeeping gene. 96-well plates were run on ABI PRISM 7900HT system (Applied Biosystems). DEGs analysis and heatmap representation Differentially expressed gene (DEG) analysis was performed on each cluster using FindAllMarkers function of Seurat v4 package. The “MAST” and “wilcox” tests were used on the RNA assay, and 0.25 was set as the minimum percentage of cell to be considered in each cluster and as the LogFC threshold. ComplexHeatmap package [92] was used to visualise the results of the DEG analysis. Molecular detection of mutational status Tumor genomic DNA (gDNA) from each sample was extracted and analysed by using a capture-based target sequencing panel, including coding exons and splice site of 111 genes recurrently mutated in T-cell lymphomas. DNA was extracted from cultured cells using the MagCore Genomic DNA Whole Blood Kit following the manufacturer’s protocol. DNA was quantified using the fluorometric method Qubit dsDNA HS Assay Kit (Invitrogen) according the manufacturer’s protocol. Criteria used to design the targeted gene panel We designed a target-resequencing panel including coding exons or hotspot regions of 111 genes that are recurrently mutated in T-cell tumors. The gene panel was optimized as follows:i.Genes were chosen if recurrently mutated in > 5% of T-cell tumors;ii.For genes with a well-defined hotspot only regions that included mutations are covered;iii.For genes lacking a well-defined hotspot, all coding exons are covered. i.Genes were chosen if recurrently mutated in > 5% of T-cell tumors; Genes were chosen if recurrently mutated in > 5% of T-cell tumors; ii.For genes with a well-defined hotspot only regions that included mutations are covered; For genes with a well-defined hotspot only regions that included mutations are covered; iii.For genes lacking a well-defined hotspot, all coding exons are covered. For genes lacking a well-defined hotspot, all coding exons are covered. SeqCap EZChoice Library (Roche Diagnostics Nimblegen) probe system was designed to capture that target regions. Libraries preparation and ultra-deep sequencing Germline gDNA libraries were prepared starting from at least 200 ng of gDNA which were sheared through sonication (Covaris) before library construction to obtain 200 bp fragments. Libraries were then generated using the Kapa Library Preparation Kit (Kapa Biosystem) following the manufacturer’s instructions. Each library was labelled with a ‘barcode’, also called adapter, to allow recognition in the analysis phase. For a quality check, Bioanalyzer was used for fragment size control (200–500 bp expected) and the Qubit dsDNA HS Assay Kit to quantify libraries. The enrichment of the regions of interest were obtained using a system of SeqCap EZ Choice Library probes (Roche Diagnostics Nimblegen) properly designed. Final libraries were sequenced using the Miseq (Illumina) instrument by pair end sequencing (2 × 150 cycles). A total of 8 multiplexed libraries were simultaneously sequenced following the manufacturer's procedure, along with an internal quality control (PhiX; Illumina). The number of the libraries loaded in the sequencer were tailored to obtain at least a coverage > 1000 × in 80% of the region of interest. Bioinformatic pipeline Raw sequences were aligned to the human genome (ENSEMBL hg19) using Burrows Wheeler Aligner with default parameters. Sorted and indexed reads were then locally realigned around InDels, duplicates removed, base quality scores assessed and reads recalibrated using the genome analysis toolkit (GATK). MUtect2 was used to call SNVs and insertions/deletions (InDels) [93]. Finally, Variant Effect Predictor (VEP) version 108 was used to annotate the variants and MafTools to generate the oncoplots. In order to convert vcf files to maf files we took advantage of mskcc-vcf2maf. Cytoscan analysis The CytoScan HD was used to detect possible chromosomal aberrations and enable low-level mosaicism visualization, LOH and CN change analysis. Genomic DNA was extracted from PBMC according to the manufacturer’s instructions (Maxwell CSC Blood DNA Kit). SNPs Array was performed using CytoScan HD (Thermo Fisher) following manufacturer’s protocol. The PCR product was hybridised using CytoScan HD Array Gene Chip and processed with the Fluidic Station (Thermo Fisher) and scanned with GeneChip Scanner (ThermoFisher). The SNPs Array results were analysed using Chromosome Analysis Suite (ChAS) software. Regions of copy number alterations > 400 Kb and CN-LOH > 4 Mb identified by ChAS software were denoted as true aberrations. Identified aberrations were described according to ISCN 2018. Supplementary Information Additional file 1Additional file 2. Table S1. In this table we report the differentially expressed genes of each global clusterAdditional file 3. Table S2. In this table we report the differentially expressed genes of each CD8 T cell clusterAdditional file 4. Table S3. In this table we report the differentially expressed genes of each Monocytes cluster Additional file 1 Additional file 2. Table S1. In this table we report the differentially expressed genes of each global cluster Additional file 3. Table S2. In this table we report the differentially expressed genes of each CD8 T cell cluster Additional file 4. Table S3. In this table we report the differentially expressed genes of each Monocytes cluster