Increased Inflammatory Cytokines in Plasma Are Associated with Sustained Treatment-Free Remission in Chronic Myeloid Leukaemia

Background: The immune status at the time of tyrosine kinase inhibitor (TKI) cessation may play a critical role in achieving sustained treatment-free remission (TFR) in chronic myeloid leukaemia (CML). Cytokine profiles may provide a measure of host immune activation against residual leukaemia.

Aim: To investigate the expression of plasma cytokines at the time of TKI cessation (TOC) and assess their predictive significance for sustained TFR in two TFR cohorts from Australia (AU) and US.

Methods: Blood was collected from 113 CML patients (pts) at TOC from an AU discovery cohort, 72 pts in the LAST study (US) as an independent validation cohort, and 20 healthy donors. The levels of 38 cytokines, chemokines and growth factors were measured in plasma samples using a MILLIPLEX Human Cytokine & Chemokine 38-plex panel and a Luminex 200 instrument. All samples were measured in duplicate and averaged. Unsupervised analysis was performed using t-Distributed Stochastic Neighbour Embedding. Multivariable model was performed with Cox proportional hazards regression. Hazard ratios (HR) and 95% confidence interval (CI) were reported for covariates, along with P values from the Wald test. All statistical analyses were conducted using the R v.4.1.1 and Prism GraphPad v9.

Results: In the AU cohort most pts were treated with imatinib (51%), followed by nilotinib (31%) and dasatinib (17%) prior to TOC. The median duration of TKI and MR4 was 6 years (range 4.3-10) and 4 years (range 3.2-6.5), respectively. Median follow-up after TKI cessation was 24.2 months (mths; range 20.7-26.8). The probability of sustained MMR after TKI cessation at 12 and 36 mths was 54% (95% CI 46-64%) and 47.6% (95% CI 36-62%), respectively.

We used an unsupervised analysis to group samples based on the similarity of their 38 cytokine expression profiles. Two distinct clusters in the AU samples were identified. Cluster 1 (n=62) had lower global 38-cytokines expression compared to cluster 2 (n=51, median 84 vs 104 pg/mL, p<0.001). Herein, we defined cluster 1 as a “cold” cluster and cluster 2 as “hot”. This dichotomy of cytokine expression was also seen in 20 healthy controls in similar proportions. 11 pts had pre-TOC and TOC matched samples several weeks apart. Only 1 pt (9%) had a different cytokine cluster profile between timepoints suggesting that the cluster category was generally stable for each pt in the short term.

At 12 mths, the hot cluster pts had a higher probability of sustained MMR (73%) compared to the cold cluster (39%, p<0.001). This difference was more pronounced at 36 mths, with the hot cluster group maintaining a sustained MMR rate of 73%, compared to 18% in the cold cluster (p=0.000035). We did not identify statistically significant differences in sex (p=0.92), age (p=0.12), MR4 duration (p=0.45), BCR::ABL1 transcript types (p=0.46), initial BCR::ABL1 halving time (HT, p=0.26), TKI type prior to TOC (p=0.48), or TKI duration (p=0.77) between hot and cold clusters. Multivariable analysis showed that cytokine expression clusters (HR 2.7; p=0.007), HT (HR 2.5; p=0.007) and BCR::ABL1 transcript type (HR 2.1; p=0.02) were independent predictors of TFR.

The US LAST cohort (n=72; 50% TFR at 12 mths) was used as an independent validation cohort. We identified hot and cold clusters in this cohort using the same criteria. At 12 mths, the hot cluster group (n=21) had a higher probability of sustained MMR (76%) compared to the cold cluster group (39%, n=51, p=0.0095). The hot:cold cytokine cluster sample ratio was 1:2.4 for the US cohort, and 1:1.2 for the AU cohort.

We identified 6 cytokines (IL-15, TNFb, IL-13, IL-6, IL-1a and G-CSF) as the most discriminating to define hot and cold clusters. The expression levels of all six cytokines were higher in the hot cluster compared to the cold cluster. Using this restricted set of predominantly inflammatory cytokines, at 12 mths the hot cluster pts had a higher probability of sustained MMR compared to the cold cluster both in AU (69% vs 43%, p=0.008) and US (75% vs 40%, p=0.017) cohorts.

Conclusion: Pts with higher levels of inflammatory cytokine expression have higher rates of sustained TFR. We postulate that the two distinct clusters of cytokine expression at the TOC reflect distinct states of immune activation. These findings support the development of predictive cytokine-based biomarkers to guide TFR treatment strategies and further support the case for immune activation as a strategy to enhance TFR outcomes for CML pts.