AP-1 Transcription Factor Complex Members FOSB and FOS are Linked With CNS Infiltration and Inferior Prognosis in Childhood T-ALL.

Children with acute lymphoblastic leukemia (ALL) have a favorable prognosis; however, diagnosis and treatment of central nervous system (CNS)-involvement remain challenging, particularly in T-cell ALL (T-ALL).1 Due to insufficient strategies to conclusively diagnose and treat CNS-disease, all patients receive risk-adapted potent CNS-directed chemotherapies, which can cause severe toxicity and is associated with neurocognitive deficits.1,2 Recent research has identified distinguishing features of CNS-ALL cells such as hypoxic adaptation,3 reprogramming of fatty-acid metabolism,4 and enhanced pre-B-cell receptor-5 and IL7R-signaling.6 Yet, these mostly apply to B-cell precursor (BCP)-ALL. For T-ALL in the CNS (CNS-T-ALL), only few risk factors are established, including elevated white blood cell (WBC) counts.1 Besides this, some molecules and pathways are associated with CNS-T-ALL, such as chemokine receptor 7 (CCR7) and C-X-C chemokine receptor type 4 (CXCR4), yet they are not measured in clinical routine.7,8 The aim of this study was to identify novel markers for CNS-leukemia, especially in T-ALL. Here, we show that activating protein-1 (AP-1) molecules, particularly FOSB are associated with CNS-involvement in T-ALL patient-derived xenograft (PDX) mice and can augment diagnostics of CNS-involvement in T-ALL patients. First, to identify such markers, PDX-cells from 6 T-ALL patients were transplanted into NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG)-mice and comparative RNA sequencing of leukemic cells recovered from CNS versus bone marrow (BM) was performed (Figure 1A; Suppl. Table S1; PDX1-6). Indeed, we detected the upregulation of various genes in CNS-ALL as compared with BM-ALL cells (Figure 1B; Suppl. Table S2). Notably, among the top 15 differentially upregulated genes in CNS-ALL cells, 3 members of the AP-1 family were identified, namely JUN, FOS, and FOSB. FOS and FOSB were detected as significantly upregulated in DESeq2-analysis9 (both q = 0.033; Figure 1B; Suppl. Table S2). Accordingly, we found a significant enrichment of the AP-1-pathway in gene set enrichment analysis (GSEA) in ALL cells recovered from the CNS compared with BM (normalized enrichment score = 1.46, false discovery rate q = 0.021; Figure 1C). AP-1 is a transcription factor (TF) complex involved in various cellular functions such as survival and proliferation.10 Deregulation of AP-1-signaling was shown to be associated with cancer, including ALL.10,11 Thus, we hypothesized a central role of AP-1 genes for CNS-colonization by T-ALL cells.Figure 1.: AP-1 transcription factors JUN , FOS , and FOSB and the AP-1 pathway are upregulated in CNS T-ALL blasts. (A) Schematic of the T-ALL PDX-model. Patient-derived T-ALL cells were injected into NSG-mice (n = 6). (B) Gene signature of the 15 most upregulated genes in CNS-derived blasts included AP-1 transcription factors JUN (q = 0.999), FOS (q = 0.033), and FOSB (q = 0.033) as determined by GSEA compared with blasts from the BM of T-ALL PDX-mice. Names of significant upregulated genes (q < 0.05) according to the DESeq2 method are labelled red. (C) GSEA identified enrichment of the AP-1 pathway in blasts retrieved from the CNS compared with blasts from the BM niche of 6 T-ALL PDX-mice (NES = 1.46; P = 0.019; FDR-q = 0.021). (D) qPCR analysis showed significant upregulation of mRNA levels of JUN (P = 0.002), FOS (P = 0.002), and FOSB (P = 0.004) in blasts from the CNS compared with blasts from the BM of T-ALL PDX-mice (n = 10). Wilcoxon-Test.*P < 0.05; **P < 0.01; ***P < 0.001. BM = bone marrow; CNS = central nervous system; DESeq2 = differential gene expression analysis based on the negative binomial distribution; FDR = false discovery rate; GSEA = gene set enrichment analysis; NES = normalized enrichment score; PDX = patient-derived xenograft; T-ALL = T-cell acute lymphoblastic leukemia.To validate our approach and the role of AP-1 in CNS-T-ALL, we performed GSEA applying Hallmark and Pathway Interaction Database (PID) gene sets. As expected, the AP-1 gene set was among the 10 most upregulated PID gene sets in CNS compared with BM-T-ALL cells (Suppl. Table S3). Interestingly, a strong enrichment in CNS-T-ALL cells was found for hypoxia-annotated genes, including VEGFA, which is regulated by AP-1 and is associated with CNS-BCP-ALL (Suppl. Table S3).3,12 This indicates that hypoxia adaption plays a central role in CNS-involvement also in T-ALL and may act via AP-1-signaling. Moreover, to validate the role of JUN, FOS, and FOSB in CNS-T-ALL cells, their mRNA levels were analyzed via quantitative real-time polymerase chain reaction (qPCR) in CNS- versus BM-T-ALL cells from 10 PDX-mice including 2 mice bearing PDX-cells from relapsed or refractory (r/r) T-ALL patients (Suppl. Table S1; PDX7–10). As expected, a significant upregulation of JUN (5.23-fold; P = 0.002), FOS (17.79-fold; P = 0.002), and FOSB (7.47-fold; P = 0.004) mRNA was observed in CNS-T-ALL cells (Figure 1D). We next asked whether these candidate genes may serve as prognostic biomarkers for CNS-involvement in T-ALL patients. Therefore, mRNA levels of JUN, FOS, and FOSB were measured in diagnostic BM-samples of 112 pediatric T-ALL patients treated within the ALL-BFM 2000 study, which contained 15 initially CNS-positive (CNS+) and 8 CNS relapse cases (Suppl. Table S4).8 We tested for associations of AP-1-genes with clinical parameters, including initial CNS status and CNS relapse. We detected significantly elevated levels of FOSB in CNS+ compared with CNS-negative (CNS−) patients in this cohort (21.7 versus 6.7; P = 0.038; Suppl. Table S4), confirming a role of FOSB-levels in CNS-T-ALL. No significant differences in JUN- and FOS-levels between CNS+ and CNS− patients were observed, which could be explained by experimental differences between PDX-models and patient analysis (Suppl. Table S4). However, higher JUN-levels were measured in patients with a WBC ≥100,000/µL compared with lower WBCs (13.0 versus 4.7, 8.8, and 6.4; P = 0.070; Suppl. Table S4). Furthermore, we found elevated FOS-levels in patients stratified into the ALL-BFM-high-risk group (considering minimal residual disease and prednisone response [PR]) as compared with the standard- and medium-risk groups (70.3 versus 29.3 and 33.6; P = 0.011; Suppl. Table S4). Accordingly, FOS- and FOSB-levels were significantly higher in patients with poor PR than good PR (71.8 versus 34.9; P = 0.004 and 10.1 versus 6.4; P = 0.023, respectively; Suppl. Table S4). Patients who died showed significantly elevated levels of FOS and FOSB compared with surviving patients (73.2 versus 38.6; P = 0.040 and 19.3 versus 6.7; P = 0.009; Suppl. Table S4). Multivariate logistic regression analysis (MVL), excluding the effect of the known risk parameters age and WBC, showed that FOS-expression above the median (FOShigh) was associated with an increased risk for poor PR (odds ratio [OR] = 2.543; 95% confidence interval [CI], 1.110-5.825; P = 0.027; Figure 2A). FOSB did not hold up as an independent predictor of PR in MVL, but FOSBhigh patients had an elevated risk for poor PR in univariate analysis (OR = 2.267 [95% CI, 1.052-4.886]; P = 0.037; Figure 2B). Moreover, FOShigh and FOSBhigh patients exhibited an increased risk for death in MVL (OR = 3.384 [95% CI, 1.237-9.260]; P = 0.018 and OR = 3.356 [95% CI, 1.233-9.136]; P = 0.018, respectively, Figure 2A, B). FOS-expression in the upper quartile was associated with an enhanced risk for a high-risk stratification (OR = 4.403 [95% CI, 1.341-14.464]; P = 0.015; Figure 2A). Importantly, MVL confirmed a significantly ≈4-fold increased risk for CNS-positivity of patients with FOSB-expression in the fourth quartile compared with lower quartiles (OR = 3.986 [95% CI, 1.181-13.459]; P = 0.026; Figure 2B), proposing FOSB as an independent predictor of CNS-T-ALL that can be assessed in BM biopsies.Figure 2.: FOSB and FOS have independent prognostic relevance for CNS-involvement and prognosis in T-ALL patients. FOS and FOSB mRNA levels of 112 pediatric T-ALL patients were measured in diagnostic bone marrow samples by qPCR and analyzed for associations with clinical parameters. Expression data were divided into medians or quartiles as indicated. Low = mRNA level below median; high = mRNA level is equal to or above median. (A, B) Univariate and multivariate logistic regression analysis of (A) FOS-expression groups and risk of prednisone poor response, deadly course of the disease and stratification to the HR group and of (B) FOSB expression groups and risk of initial CNS-involvement and a deadly course of the disease. Multivariate analysis was controlled for age and WBC at diagnosis.1Reference category. One patient had no available data on initial CNS status. (C–D) Kaplan-Meier survival curves showing reduced 5-year event-free survival of (C) FOS high compared with FOS low patients (pEFS = 66% vs 84%; P = 0.031) and of (D) FOSB high patients compared with FOSB low patients (pEFS = 64% vs 86%; P = 0.011). Events included incomplete remission, relapse, second cancer, or death from any cause. Black arrows indicate CNS relapses. Log-rank statistics. *P < 0.05. (E, F), Machine learning algorithms were applied to identify a CNS-high-risk marker panel. (E) Comparison of prediction performance of clinical and molecular marker sets in CNS-involvement. The accuracy of clinical markers (age, WBC, sex) (red item; sensitivity = 0%) in predicting the CNS status is compared with the group of molecular markers (including the new markers introduced in this study) (green item; sensitivity = 47%) and the combination of clinical and molecular markers (blue item; sensitivity 67%). (F) Influence of individual markers on CNS status within the identified marker panel. Coefficients of the penalized regression model for the identified markers are shown. The size of the bar indicates the predictive influence of the respective parameter on the CNS-classifier. The larger the absolute value of a coefficient, the stronger the effect of the respective parameter on the CNS status. Coefficients greater than zero indicate a positive contribution of the marker on CNS-involvement, whereas coefficients below zero indicate negative contributions. Colors indicate spearman correlations of marker expression and CNS status. - = negative; + = positive; CI = confidence interval; CNS = central nervous system; HR = high-risk; No. = number; OR = odds ratio; pEFS = probability of event-free survival; T-ALL = T-cell acute lymphoblastic leukemia; WBC = white blood cell.Subsequently, event-free survival (EFS) rates of FOShigh versus FOSlow and FOSBhigh versus FOSBlow patients were analyzed via Kaplan-Meier method. The 5-year EFS of FOShigh and FOSBhigh patients was significantly reduced compared with FOSlow and FOSBlow patients (probability of EFS [pEFS] = 0.66 versus pEFS = 0.84; P = 0.031 and pEFS = 0.64 versus pEFS = 0.86; P = 0.011; respectively; Figure 2C, D). Forty-eight of the 112 (42.9%) patients were FOShigh/FOSBhigh simultaneously and 48 of 112 (42.9%) were FOSlow/FOSBlow (Suppl. Table S5). EFS rates did not differ as compared with single marker positivity (data not shown). Eighteen of 28 (64.3%) of all events occurred in patients that were FOShigh/FOSBhigh (Suppl. Table S6). Of note, 6 of 8 CNS relapses were detected in FOShigh/FOSBhigh patients, substantiating the adverse impact of these genes (Figure 2C, D; Suppl. Table S7). To investigate if markers such as FOSB may be additionally beneficial for CNS-ALL-diagnostics when combined with clinical parameters, we used machine learning algorithms (Suppl. Figure S1). These algorithms integrated multivariate regression models considering clinical parameters (age, sex, and WBC) and 16 molecular mRNA markers, including molecules with potential involvement in CNS-ALL, such as CXCR4 and CCR7 (Suppl. Table S8).7,8 Indeed, a CNS-high-risk marker panel containing molecular and clinical markers, which were identified by the penalized regression model elastic net, predicted CNS-positivity more reliably than molecular markers and clinical risk factors alone (67% versus 47% and 0% sensitivity; Figure 2E; Suppl. Table S8). In accordance with previous reports, high WBC and CCR7 were associated with CNS-positivity. CXCR4 had a negligible predictive impact on CNS infiltration indicating that albeit its potential biological role for CNS tropism, CXCR4 might be less suitable for CNS-ALL diagnostics. Notably, FOSB showed the strongest predictive power for CNS-positivity compared with other markers including WBC and CCR7-expression, further validating FOSB as a predictor of CNS-positivity in our T-ALL cohort (Figure 2F). Identification and stratification of CNS+ versus CNS− patients is critical for tailored therapy of ALL and to spare CNS-low-risk patients toxic therapy. Furthermore, the assumption that the majority of ALL patients bear occult ALL cells in the CNS upon diagnosis promotes the view that CNS relapse may be prevented by targeting survival rather than migration pathways.13 In this respect, the AP-1-pathway may be an interesting new target. AP-1-signaling is closely interconnected with hypoxia-signaling, an important pathway for CNS-BCP-ALL cells, which we also found upregulated in CNS-T-ALL.3,12 Therefore, AP-1 may represent a central TF needed for hypoxia adaption of T-ALL cells for survival in the hostile CNS microenvironment. AP-1-signaling is also associated with resistance to glucocorticoids and chemotherapy, which may explain the adverse outcome of FOS/FOSBhigh patients, particularly regarding CNS-disease.14 Moreover, AP-1 seems to be of importance for other hematological malignancies beyond T-ALL: upregulation of the AP-1-pathway was shown to be critical in KMT2a-rearranged BCP-ALL, a prognostically unfavorable subgroup with frequent CNS-involvement.11 Moreover, own unpublished data suggest the AP-1 genes JUN, FOS, and FOSB as markers of CNS-disease in BCP-ALL. Additionally, AP-1 inhibitors such as T-5224 have proven preclinically effective in B-cell neoplasias including ALL and could represent a strategy to target CNS-T-ALL.11,15 Yet, the different AP-1 family members may fulfill separate functions depending on their dimerization partners.10 Therefore, the exact mechanism underlying the adverse prognosis associated with high JUN/FOS/FOSB-levels and their potential targetability needs to be investigated. Overall, our data reveal AP-1 molecules, particularly FOSB as novel diagnostic markers for adverse prognosis, CNS-involvement, and relapse in T-ALL warranting further mechanistic and clinical investigation (Suppl. Figure S2). ACKNOWLEDGMENTS We thank the patients and physicians who contributed samples and data for this study. We thank Gabriele Riesen, Birthe Fedders, Katrin Neumann, and Katrin Timm-Richert for the excellent technical assistance. AUTHOR CONTRIBUTIONS LS, LL, and JZ designed and performed experiments. MVG supervised statistical analysis. JZ, TB, and LB supervised machine learning analysis. AB and SB provided transcriptomic data. DW, FV, TW, and HB performed experiments. MS and GC provided ALL samples and clinical data. DMS, LL, and AA initiated and designed the study and discussed the research direction. LS and LL wrote the article. All authors discussed the article. DISCLOSURES LL and DMS received research funding from OSE Immunotherapeutics outside the submitted work. DMS was an advisory board member for Bayer, SOBI, and Jazz Pharmaceuticals. MS received research funding from Shire, and from Servier, as well as fees for Advisory Board functions from Jazz Pharmaceuticals and Servier. All the other authors have no conflicts of interest to disclose. SOURCES OF FUNDING LS and LL have been supported by the Faculty of Medicine, University of Kiel, Germany. DMS is funded by the Deutsche Krebshilfe e. V. (111963), the Wilhelm Sander Stiftung (2016.110.1 and 2019.119.1), the Deutsche José-Carreras-Leukämiestiftung (DJCLS 17 R/2017), and the Deutsche Forschungsgemeinschaft (SCHE2046/1).