Machine learning for prediction of immunotherapeutic outcome in non-small-cell lung cancer based on circulating cytokine signatures.

BACKGROUND: Immune checkpoint inhibitor (ICI) therapy has substantially improved the overall survival (OS) in patients with non-small-cell lung cancer (NSCLC); however, its response rate is still modest. In this study, we developed a machine learning-based platform, namely the Cytokine-based ICI Response Index (CIRI), to predict the ICI response of patients with NSCLC based on the peripheral blood cytokine profiles. METHODS: We enrolled 123 and 99 patients with NSCLC who received anti-PD-1/PD-L1 monotherapy or combined chemotherapy in the training and validation cohorts, respectively. The plasma concentrations of 93 cytokines were examined in the peripheral blood obtained from patients at baseline (pre) and 6 weeks after treatment (early during treatment: edt). Ensemble learning random survival forest classifiers were developed to select feature cytokines and predict the OS of patients undergoing ICI therapy. RESULTS: Fourteen and 19 cytokines at baseline and on treatment, respectively, were selected to generate CIRI models (namely preCIRI14 and edtCIRI19), both of which successfully identified patients with worse OS in two completely independent cohorts. At the population level, the prediction accuracies of preCIRI14 and edtCIRI19, as indicated by the concordance indices (C-indices), were 0.700 and 0.751 in the validation cohort, respectively. At the individual level, patients with higher CIRI scores demonstrated worse OS [hazard ratio (HR): 0.274 and 0.163, and p<0.0001 and p=0.0044 in preCIRI14 and edtCIRI19, respectively]. By including other circulating and clinical features, improved prediction efficacy was observed in advanced models (preCIRI21 and edtCIRI27). The C-indices in the validation cohort were 0.764 and 0.757, respectively, whereas the HRs of preCIRI21 and edtCIRI27 were 0.141 (p<0.0001) and 0.158 (p=0.038), respectively. CONCLUSIONS: The CIRI model is highly accurate and reproducible in determining the patients with NSCLC who would benefit from anti-PD-1/PD-L1 therapy with prolonged OS and may aid in clinical decision-making before and/or at the early stage of treatment.

Clinical significance of plasma-free amino acids and tryptophan metabolites in patients with non-small cell lung cancer receiving PD-1 inhibitor: a pilot cohort study for developing a prognostic multivariate model.

BACKGROUND: Amino acid metabolism is essential for tumor cell proliferation and regulation of immune cell function. However, the clinical significance of free amino acids (plasma-free amino acids (PFAAs)) and tryptophan-related metabolites in plasma has not been fully understood in patients with non-small cell lung cancer (NSCLC) who receive immune checkpoint inhibitors. METHODS: We conducted a single cohort observational study. Peripheral blood samples were collected from 53 patients with NSCLC before treatment with PD-1 (Programmed cell death-1) inhibitors. The plasma concentrations of 21 PFAAs, 14 metabolites, and neopterin were measured by liquid chromatography-mass spectrometry. Using Cox hazard analysis with these variables, a multivariate model was established to stratify patient overall survival (OS). Gene expression in peripheral blood mononuclear cells (PBMCs) was compared between the high-risk and low-risk patients by this multivariate model. RESULTS: On Cox proportional hazard analysis, higher concentrations of seven PFAAs (glycine, histidine, threonine, alanine, citrulline, arginine, and tryptophan) as well as lower concentrations of three metabolites (3h-kynurenine, anthranilic acid, and quinolinic acid) and neopterin in plasma were significantly correlated with better OS (p<0.05). In particular, the multivariate model, composed of a combination of serine, glycine, arginine, and quinolinic acid, could most efficiently stratify patient OS (concordance index=0.775, HR=3.23, 95% CI 2.04 to 5.26). From the transcriptome analysis in PBMCs, this multivariate model was significantly correlated with the gene signatures related to immune responses, such as CD8 T-cell activation/proliferation and proinflammatory immune responses, and 12 amino acid-related genes were differentially expressed between the high-risk and low-risk groups. CONCLUSIONS: The multivariate model with PFAAs and metabolites in plasma might be useful for stratifying patients who will benefit from PD-1 inhibitors.

Diagnostic yield of the plasma free amino acid index for pancreatic cancer in patients with diabetes mellitus.

OBJECTIVES: A multivariate index calculated using plasma free amino acids (PFAA index) was reported as a diagnostic biomarker for pancreatic cancer (PaC). Although diabetes mellitus (DM) is expected to be an early diagnostic indicator of PaC, identifying the high-risk individuals among patients with DM is warranted. We evaluated the diagnostic yield of the PFAA index for PaC in patients with DM. METHODS: We compared the diagnostic yield of the PFAA index between individuals with and those without DM. Cases and controls were recruited prospectively, and controls were matched to cases at a 1:1 ratio for age, sex, and DM status. RESULTS: A total of 180 case-control pairs were included in the analysis. The prevalence of DM was 53.3%. The sensitivity of the PFAA index was 66.7% in cases with DM and 56.0% in those without DM (P = 0.14), and the specificity was 92.7% in controls with DM and 94.0% in those without DM (P = 0.95). CONCLUSIONS: This matched case-control study revealed a comparable diagnostic yield of the PFAA index for PaC in individuals with and those without DM. The PFAA index can be used as a biomarker for further diagnostic imaging in selected patients with DM.

Development and validation of a novel IL-10 deficient cell transfer model for colitis.

A number of rodent models for inflammatory bowel disease (IBD) have been developed, but most cannot be used to develop and validate new therapies for IBD. From the models developed, the IL-10 deficient mouse model is the one that results in a disease similar to human IBD; however, in this model, colitis occurs with variable incidence taking 3-4 months to develop. These are serious problems with the model when evaluating a new therapy because of the large-scale experiments required and the difficulty in performing an accurate pharmacological analysis. In this study, the IL-10 deficient mouse model was modified by transferring whole spleen and mesenteric lymph node cells from IL-10 deficient mice to CB-17 SCID mice. In this IL-10 deficient cell transfer model, chronic intestinal inflammation developed in all recipients within 2-3 weeks, which was far earlier than in donor IL-10 deficient mice. The pathological phenotypes were similar to those of IL-10 deficient mice and CD45RBhi T cell-transfer models. In addition, we assessed several agents for inflammatory bowel disease to validate the general utility of this cell transfer model. It is worth noting that TNFR-Ig or prednisolone, which is effective for treatment of patients with severe-fulminant Crohn's disease, markedly attenuated pathological clinical indices in this colitis model, whereas the immunosuppressive agents, azathioprine, tacrolimus, and cyclosporine A produced no significant effect. These results suggest that the IL-10 deficient cell transfer model is a good experimental model to use for developing new and effective therapies for active IBD.

A role for fungal {beta}-glucans and their receptor Dectin-1 in the induction of autoimmune arthritis in genetically susceptible mice.

A combination of genetic and environmental factors can cause autoimmune disease in animals. SKG mice, which are genetically prone to develop autoimmune arthritis, fail to develop the disease under a microbially clean condition, despite active thymic production of arthritogenic autoimmune T cells and their persistence in the periphery. However, in the clean environment, a single intraperitoneal injection of zymosan, a crude fungal beta-glucan, or purified beta-glucans such as curdlan and laminarin can trigger severe chronic arthritis in SKG mice, but only transient arthritis in normal mice. Blockade of Dectin-1, a major beta-glucan receptor, can prevent SKG arthritis triggered by beta-glucans, which strongly activate dendritic cells in vitro in a Dectin-1-dependent but Toll-like receptor-independent manner. Furthermore, antibiotic treatment against fungi can prevent SKG arthritis in an arthritis-prone microbial environment. Multiple injections of polyinosinic-polycytidylic acid double-stranded RNA also elicit mild arthritis in SKG mice. Thus, specific microbes, including fungi and viruses, may evoke autoimmune arthritis such as rheumatoid arthritis by stimulating innate immunity in individuals who harbor potentially arthritogenic autoimmune T cells as a result of genetic anomalies or variations.

Altered thymic T-cell selection due to a mutation of the ZAP-70 gene causes autoimmune arthritis in mice.

Rheumatoid arthritis (RA), which afflicts about 1% of the world population, is a chronic systemic inflammatory disease of unknown aetiology that primarily affects the synovial membranes of multiple joints. Although CD4(+) T cells seem to be the prime mediators of RA, it remains unclear how arthritogenic CD4(+) T cells are generated and activated. Given that highly self-reactive T-cell clones are deleted during normal T-cell development in the thymus, abnormality in T-cell selection has been suspected as one cause of autoimmune disease. Here we show that a spontaneous point mutation of the gene encoding an SH2 domain of ZAP-70, a key signal transduction molecule in T cells, causes chronic autoimmune arthritis in mice that resembles human RA in many aspects. Altered signal transduction from T-cell antigen receptor through the aberrant ZAP-70 changes the thresholds of T cells to thymic selection, leading to the positive selection of otherwise negatively selected autoimmune T cells. Thymic production of arthritogenic T cells due to a genetically determined selection shift of the T-cell repertoire towards high self-reactivity might also be crucial to the development of disease in a subset of patients with RA.

The skewing to Th1 induced by lentinan is directed through the distinctive cytokine production by macrophages with elevated intracellular glutathione content.

In vivo lentinan (LNT)-elicited peritoneal macrophages (Mps) showed the reduced release of prostaglandins (PGs), IL-10 and IL-6, while it endowed Mps with the elevated capability to produce IL-12 and nitric oxide (NO) upon in vitro triggering, due to the elevated intracellular glutathione (GSH) content in Mps. Deprivation of intracellular GSH completely ablated the production of IL-12. Conversely, lipopolysaccharide (LPS) induced peritoneal Mps with the reduced intracellular GSH content and the reciprocal profile of mediator production. Mps with the elevated intracelluar GSH is arbitrarily termed as reductive Mp (RMp) and that with reduced amount as oxidative Mp (OMp). OMp was converted to RMp when GSH was replenished with glutathione monoethylester (GSH-OEt). The IL-2 administration in combination with LNT exerted the synergistic induction of RMp, resulting in synergistic augmentation of IL-12, NO and reduction of IL-6 production. It was also confirmed that CD4+T cells derived of LNT-administered mice showed augmented IFN-gamma and reduced IL-4 production upon in vitro anti-CD3 stimulation. Taken together it is concluded that skewing of Th1/Th2 balance to Th1 by a beta-(1-3)-glucan, LNT, is directed through the distinctive production of IL-12 versus IL-6, IL-10 and prostaglandin E2 (PGE2) by Mps, depending on intracellular GSH redox status. To the efficient tumor immunotherapy, it may be one of the critical elements to induce a reductive form of Mps in tumor stromal tissues to maintain Th1 response.