From June 2020 to August 2022, a total of 52 patients who were diagnosed with LGLL were enrolled in the trial (Supplementary Fig. 1). Among the 52 patients, 7 were diagnosed with NK-LGLL and 45 with T-LGLL, among which 37 cases were of the \(\alpha \beta\) type and 8 cases were of the \(\gamma \delta\) type. Among the 52 patients, 8 were diagnosed with severe anemia (<6 g/dL), 4 with granulocytopenia (<0.5 × 109/L), and 2 with both severe anemia and granulocytopenia. Six patients in this study were diagnosed with concurrent autoimmune disorders. Two patients presented with RA, one with unclassified connective tissue disease, one with Hashimoto’s thyroiditis, and one with autoimmune hemolytic anemia. The baseline characteristics are summarized in Table 1. The severity of cytopenia at baseline for each patient, along with their transfusion dependency status, is presented in Table 1 and Supplementary Table 2.

Among the 52 patients, 42 were newly diagnosed, whereas the remaining 10 had previously received at least one immunosuppressive therapy. The baseline characteristics of the two groups were comparable, as detailed in Supplementary Table 3. The front-line treatments for these 10 patients included CsA in six patients, CTX in one patient, and other cytotoxic drugs in three patients. The treatment outcomes varied, with three patients achieving CR, two achieving PR, and five showing stable disease (SD). PFS survival ranged from 4 to 100 months (Supplementary Table 4).

The most common reason for therapeutic intervention among the enrolled patients was anemia (46.2%), followed by a combination of anemia and neutropenia (42.3%). The primary hematological manifestation was anemia (90.4%), followed by neutropenia (50.0%) and thrombocytopenia (15.4%). Splenomegaly was observed in 18 patients (38.3%). Additionally, 23.1% of patients had pure red cell aplasia (PRCA), and 11.5% had autoimmune diseases. Among the patients, 39 underwent STAT3 mutation testing via next-generation sequencing (NGS), and the mutation rate was 56.4%.

In total, 47 patients (90.4%, 95% confidence interval (CI), 82.4% to 98.4%) achieved hematological and symptomatic responses. Among them, 39 patients achieved CR (75.0%, 95% CI, 63.0% to 87.0%), and eight patients achieved PR (15.4%, 95% CI, 5.6% to 25.2%). Among the 12 patients who were transfusion-dependent at baseline, 10 patients (83.3%) achieved transfusion independence following treatment (Supplementary Table 2).

When the 47 patients with anemia at baseline were evaluated, there was a significant increase in the median HGB level, from 6.9 g/dL (range: 3.8–10.6) to 11.8 g/dL (range: 4.9–14.6) (P < 0.001, Fig. 1a, b). Approximately 83.0% of patients achieved normal HGB levels (> 11 g/dL), with a median percentage increase from baseline of 71.0%. The median time to return to normal levels was 6.0 months (range: 3.0–36.0). Similarly, the median neutrophil count significantly increased, from 0.8 × 109/L (range: 0.3–1.5) to 1.7 × 109/L (range: 0.6–2.9) (P < 0.001, Fig. 1c, d). Approximately 73.1% (18/26) of patients achieved normal neutrophil counts ( > 1.5 × 109/L), with a median time to return to the optimal or normal level of 6.0 months (range: 3.0–33.0). The results for newly diagnosed and previously treated patients are presented separately in Supplementary Fig. 2. Spleen size decreased or returned to normal in 70.0% of patients with splenomegaly. Importantly, this regimen demonstrated a good capacity to clear LGLL clones. Twenty-seven patients underwent posttreatment peripheral blood FCM and TCR gene rearrangement analysis, with three (3/27, 11.1%) patients achieving CMR.

Changes of hemoglobin and absolute neutrophil counts before and after treatment. a, b Hemoglobin. c, d Absolute neutrophil counts; ***P < 0.001

The median follow-up time was 29.0 months (range: 4.0–42.0), and the median duration of the TPM regimen was 21.5 months (range: 3.0–42.0). Moreover, the median time to response was 3.0 months (range: 3.0–9.0), and the median time to CR was 6.0 months (range: 3.0–36.0).

No deaths were observed during the follow-up period. Eighteen patients (34.6%) experienced disease progression due to the relapse of cytopenias. The median PFS was 40.0 months (95% CI: 38.0–42.0), and the median DoR was 39.0 months (95% CI: 36.1–41.9) (Fig. 2a, b). The median OS was not reached. PFS and DoR data for newly diagnosed and previously treated patients are presented separately in Supplementary Fig. 3.

Survival curves of the enrolled patients. a Progression-free survival; b duration of response

Moreover, the coronavirus disease 2019 (COVID-19) pandemic impacted the timely follow-up of some patients. Consequently, 17 patients continued using the Thal-based regimen for more than 2 years, exceeding the planned treatment duration. In order to evaluate the impact of the prolong using this regimen to our primary endpoint, we compared the CR rate between these 17 patients and the other 35 patients and found that there was no significant difference (88.2% vs. 68.6%, P = 0.232). The treatment response of these 14 patients with severe anemia and/or granulocytopenia was comparable to that of the other 38 patients (CR rate: 64.5% vs. 78.9%, P = 0.300; ORR: 100.0% vs. 86.8%, P = 0.307).

Among the 18 patients who experienced disease progression, 11 patients progressed during the treatment period, and seven patients progressed after discontinuation of treatment. Among the seven patients who discontinued treatment, five patients were retreated with the TPM regimen, and all of them achieved a treatment response, with three patients achieving CR and two patients achieving PR. Prior to retreatment, these patients discontinued treatment for varying durations, with a median duration of 5.0 months, ranging from 4.0 to 14.0 months. The subsequent treatment regimens and responses of the patients are summarized in Supplementary Table 5.

Fifty-four patients who had received the TPM regimen for at least one month were included in the safety analysis set. During the study, a total of 35 AEs were reported for 27 patients (27/54, 50.0%), four of which (4/54, 7.4%) had serious adverse events (sAEs, grade ≥3) (Table 2). The most common AEs of any grade were peripheral neuropathy (13/54, 24.1%) and neutropenia (4/54, 7.4%). Among the 4 patients who developed neutropenia, 2 (3.7%) experienced associated infections, including tonsillitis and urinary tract infection, both of which were classified as grade 1–2 AEs. The most common grade 3–4 AE was peripheral neuropathy (2/54, 3.7%). Overall, the AEs occurred mainly during the first cycle (12/37, 32.4%). AEs led to the discontinuation of treatment in four patients (4/54, 7.4%).

Among the 13 patients with peripheral neuropathy, symptoms completely resolved in 7 individuals following symptomatic treatment, dose reduction, or discontinuation of Thal. Additionally, symptoms were mitigated in 5 patients. Only one patient presented with grade 2 numbness in the extremities without improvement.

Thrombotic events were observed in one patient (1.9%, deep vein thrombosis of grade 2), a 68-year-old female with a Caprini score of 2, three months after TPM treatment. This thrombotic event was considered related to Thal. However, the patient was not concurrently taking aspirin for thrombosis prevention. After three months of anticoagulation therapy with rivaroxaban, the patient recovered without any sequelae. Additionally, we analyzed the cumulative dose of Pred and its related AEs. The median steroid dose was 1361 mg (505–2143 mg). Pred-related AEs were observed in 7 out of 52 (13.5%) patients, with none experiencing AEs of grade 3 or higher (Supplementary Table 7).

We focused on the AEs of those 17 patients who used the TPM regimen for more than 2 years. The incidence rates of both AEs and sAEs were similar between patients who used the regimen for more than 2 years and those who used it for less than 2 years (sAE: 0% vs. 5.7%, P = 1.000; AEs: 41.2% vs. 51.4%, P = 0.488).

In this study, 42 newly diagnosed patients and ten previously treated patients were enrolled. The ORR and CR rates were comparable between the newly diagnosed and previously treated patient groups (ORR: 90.5% vs. 90.0%, P = 1.000; CR rate: 73.8% vs. 80.0%, P = 1.000). The median follow-up times of the newly diagnosed and previously treated patient groups were 27 months and 30 months, respectively. Compared with previously treated patients, newly diagnosed patients exhibited similar PFS (median, 40.0 vs. 38.0 months; P = 0.116) and DoR (median, 39.0 vs. 37.0 months; P = 0.100; Supplementary Fig. 4a, b).

The eight patients with the γδ type exhibited comparable CR rate and ORR to the 37 patients with the αβ type (CR rate: 87.5% vs. 67.6%, P = 0.405; ORR: 87.5% vs. 88.9%, P = 1.000). Meanwhile, TPM regimen efficacy in the seven patients with NK-LGLL was similar to that in the 45 patients with T-LGLL (ORR: 100.0% vs. 88.9%, P = 1.000; CR rate: 85.7% vs. 73.3%, P = 0.815). The PFS and DoR rates were also comparable between the NK-LGLL and T-LGLL groups (PFS: median, not reached vs. 40.0 months, P = 0.489; DoR: median, not reached vs. 39.0 months, P = 0.607; Supplementary Fig. 4c, d).

We then analyzed the association between STAT3 mutation status and treatment efficacy. The mutational sites and variant allele frequencies of the STAT3 gene are shown in Supplementary Table 6. Patients with STAT3 mutations had CR rates and ORRs similar to those with wild-type STAT3 (CR rates: 63.6% vs. 88.2%, P = 0.169; ORR: 90.9% vs. 88.2%, P = 1.000). Compared with the wild-type group, the STAT3-mutated group had similar PFS and DoR (PFS: median, 38.0 months vs. not reached, P = 0.376; DoR: median, 37.0 months vs. not reached, P = 0.269; Supplementary Fig. 4e, f).

Moreover, the TPM regimen was also an effective option for patients with PRCA (n = 12). The CR rate (83.3% vs. 72.5%, P = 0.704) and ORR (91.7% vs. 90.0%, P = 1.000) were comparable between patients with and without PRCA. The PFS and DoR were also comparable between patients with and without PRCA (median PFS: 38.0 vs. 40.0 months, P = 0.293; median DoR: not 37.0 months vs. 39.0 months, P = 0.113; Supplementary Fig. 4g, h).

In this study, cytokine profiling, performed with the Olink® inflammation panel, revealed significant differences in the levels of inflammation-related cytokines between LGLL patients and healthy controls (HCs). Among the 92 cytokines tested, 30 were identified as differentially expressed proteins (DEPs) in LGLL patients, with 23 upregulated and 7 downregulated cytokines compared with HCs (Fig. 3a and Supplementary Table 8). The key cytokines with significantly elevated expression in LGLL patients included IL-8, CCL3, and CXCL5, which are known to play roles in immune and inflammatory processes. Paired samples were available for 17 patients before and after treatment. The expression of six cytokines (IL-6, IL-8, CCL3, CXCL5, CXCL11, and GROA) was notably downregulated after treatment (Fig. 3b and Supplementary Table 9). Pathway enrichment analysis revealed that the cytokine‒cytokine receptor interaction pathway was prominently enriched, suggesting a potential role of these pathways in disease pathophysiology and treatment response (Fig. 3c). We performed an additional analysis of cytokine changes before and after treatment in 10 patients with STAT3 mutations. Notably, we observed significant downregulation of CXCL11 and GROA2 expression following treatment (Supplementary Fig. 5 and Supplementary Table 10).

Serum cytokine profiling via the Olink multiplex proteomic platform. a Volcano plot of differentially expressed proteins between healthy donors and LGLL patients; b volcano plot of differentially expressed proteins of 17 paired patients before and after treatment; c Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of significantly differentially expressed proteins