• Unsupervised consensus clustering of cytokine profiles in 128 patients with TAM identified groups at higher risk for early death.

  • Measurement of levels of cytokine provides valuable information for patients with TAM that may help determine therapeutic interventions.

Abstract

Transient abnormal myelopoiesis (TAM) occurs in 10% of neonates with Down syndrome (DS). Although most patients show spontaneous resolution of TAM, early death occurs in ∼20% of cases. Therefore, new biomarkers are needed to predict early death and determine therapeutic interventions. This study aimed to determine the association between clinical characteristics and cytokine levels in patients with TAM. A total of 128 patients with DS with TAM enrolled in the TAM-10 study conducted by the Japanese Pediatric Leukemia/Lymphoma Study Group were included in this study. Five cytokine levels (interleukin-1b [IL-1b], IL-1 receptor agonist, IL-6, IL-8, and IL-13) were significantly higher in patients with early death than in those with nonearly death. Cumulative incidence rates (CIRs) of early death were significantly associated with high levels of the 5 cytokines. Based on unsupervised consensus clustering, patients were classified into 3 cytokine groups: hot-1 (n = 37), hot-2 (n = 42), and cold (n = 49). The CIR of early death was significantly different between the cytokine groups (hot-1/2, n = 79; cold, n = 49; hot-1/2 CIR, 16.5% [95% confidence interval (CI), 7.9-24.2]; cold CIR, 2.0% [95% CI, 0.0-5.9]; P = .013). Furthermore, cytokine groups (hot-1/2 vs cold) were independent poor prognostic factors in the multivariable analysis for early death (hazard ratio, 15.53; 95% CI, 1.434-168.3; P = .024). These results provide valuable information that cytokine level measurement was useful in predicting early death in patients with TAM and might help to determine the need for therapeutic interventions. This trial was registered at UMIN Clinical Trials Registry as #UMIN000005418.

Transient abnormal myelopoiesis (TAM), also known as transient leukemia or transient myeloproliferative disorder, is a unique clonal myeloproliferation characterized by immature megakaryoblasts. It occurs in 10% of neonates with Down syndrome (DS).1 Although most patients show spontaneous resolution of TAM without therapeutic interventions, ∼20% of TAM cases result in early death (death within 9 months), and ∼20% of the survivors develop acute megakaryoblastic leukemia within 4 years.2-6 Our previous reports showed that high white blood cell (WBC) count (≥100 × 109/L), systemic edema, low birth weight, preterm birth at <37 weeks of gestational age, and elevated direct bilirubin level >5 mg/dL were associated with early death.2-6 Low-dose cytarabine (LDAC) is a common therapy for TAM. It has been reported that LDAC should be considered for patients with life-threatening symptoms and risk factors associated with early death.7 Additionally, it has been reported that the LDAC intervention rate was adversely associated with the early death rate.2 However, further studies are needed to determine the criteria for consensus therapeutic intervention.

Previous reports showed that cytokine levels are associated with liver failure, which is a cause of early death in patients with TAM.8 Thus, cytokine levels can be new biomarkers to predict early death in patients with TAM. However, no large cohort data are available for cytokine analyses in patients with TAM. Thus, this study aimed to determine the association between clinical characteristics and cytokine levels in patients with TAM by analyzing 128 patients with DS with TAM enrolled in the TAM-10 prospective observational study conducted by the Japanese Pediatric Leukemia/Lymphoma Study Group (JPLSG).

Patients

A total of 167 neonates diagnosed with TAM were prospectively registered in the TAM-10 study between May 2011 and February 2014 conducted by the JPLSG of the Japan Children’s Cancer Group. The TAM-10 study was registered with the University Hospital Medical Information Network (UMIN) Clinical Trials Registry (http://www.umin.ac.jp/ctr/index.htm, number UMIN000005418). The details of the eligibility criteria and the central review system, including the GATA1 mutation analysis, are shown in the previous report.2 Clinical data and sample collections in the clinical trials were approved by the institutional review boards of each participating institution. Written informed consent was obtained from all patients’ parents/guardians. This study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the ethical review board of the JPLSG. Cytokine levels were analyzed in 128 of the 167 patients for whom serum samples were available in the early postnatal period (days 0-8). A comparison of clinical characteristics between 128 and 39 patients with or without available samples, respectively, is shown in supplemental Table 1.

Cytokine analysis

Serum concentrations of the following 27 cytokines were determined using the Bio-Plex cytokine assay (Bio-Rad, Hercules, CA), measured using a Luminex System (Austin, TX), and quantified using Bio-Plex software (Bio-Rad).9 The details of 27 cytokines are described in supplemental Table 2. Serum samples were frozen at –80°C immediately after collection and stored until analysis. Each sample was analyzed twice. The mean values of measurements were used as representative values for each subject.

Statistical analysis

Optimal cutoff values for biomarkers were determined using the Youden index of the receiver operator characteristic curve based on logistic regression analyses. The association between the covariates and early death (<9 months of age) was evaluated in univariable and multivariable Cox proportional hazard models using the stepwise Akaike information criterion method. Between-group comparisons were performed using the Mann-Whitney U test or Fisher exact test, as appropriate. Differences in cytokine levels between groups were determined using the Mann-Whitney U test. A correction for multiple testing was performed using the Benjamini-Hochberg method with the threshold P value set at < .05.

Cluster analysis was performed by 2-step unsupervised consensus clustering of 27 cytokine variables. Five cytokines (interleukin-2 [IL-2], IL-12, IL-15, IL-17, and RANTES) with missing values in ≥10 patients were excluded from subsequent analyses (supplemental Figure 1A). Details of missing values for the remaining 22 cytokines are shown in supplemental Figure 1B-C. Missing values of 22 cytokines were imputed using the random forest-based algorithm, missForest.10 The features were log-standardized for data preprocessing. For sensitivity analyses, cluster analyses were also performed on the data set complemented with different imputation methods based on the k-nearest neighbor (kNN) and principal component analysis (PCA; supplemental Figure 1D-E).11 In addition, complete data analysis was performed on 43 patients without missing data for all 27 cytokines.

All statistical analyses were performed using EZR software version 1.36 (Saitama Medical Center, Jichi Medical University, Saitama, Japan)12 and R Version 4.3.2 with “naniar,” “ConsensusClusterPlus,”13 “ComplexHeatmap,” and “ggplog2” packages. A 2-tailed P value < .05 was considered statistically significant. Details and other information on statistical analysis are described in supplemental Methods.

Patient characteristics

Table 1 shows the clinical characteristics and laboratory findings of 128 patients with TAM. The median values of gestational age, body weight at birth, WBC count, and percentage of blasts at diagnosis were 37 weeks (range, 29-40), 2588 g (range, 1438-3714), 48.3 × 109/L (range, 4.7 × 109/L to 478.7 × 109/L), and 41% (range, 1%-96%), respectively. Of the 128 patients, 87 (68%) had congenital heart disease, and 14 (11%) had other congenital abnormalities. Trisomy 21 was observed in 126 patients (98%), trisomy 21 mosaicism in 1 patient (1%), and a normal karyotype in 1 patient (1%). Systemic edema was observed in 26 patients (20%) and organ hemorrhage in 12 patients (9%). Somatic GATA1 gene mutations were confirmed in 127 patients (99%) using Sanger and/or next-generation sequencing. One patient with undetectable GATA1 mutations had flow cytometry markers (CD7+/CD117+/CD56+), consistent with a TAM phenotype. The expression type of GATA1 mutations was determined based on a previous report.2 High-expression mutations were observed in 57 patients (45%), whereas low-expression mutations were observed in 58 patients (45%). Of the 128 patients, 46 (36%) received LDAC.

Table 1.

Clinical characteristics of 128 patients with TAM

All patients,
N = 128
Patients with early death, n = 14Patients without early death, n = 114P value 
Sex (male:female) 64:64 9:5 55:59 .396 
Median gestational age (range), wk 37 (29-40) 34 (29-38) 37 (31-40) .003 
Median birth weight (range), g 2588 (1438-3714) 2249 (1438-3044) 2624 (1598-3714) .035 
Median age at diagnosis (range), d 0 (0-8) 0 (0-8) 0 (0-8) .854 
Congenital heart disease, n (%) 87 (68) 5 (36) 82 (72) .012 
Other congenital abnormally, n (%) 14 (11) 2 (14) 12 (11) .651 
Chromosomal status, n     
Trisomy 21:mosaic trisomy 21:normal karyotype 126:1:1 14:0:0 112:1:1 1.000 
Median WBC count at diagnosis (range), ×109/L 48.3 (4.7-478.7) 157.3 (14.3-238.5) 44.3 (4.7-478.7) .006 
Median blasts percentage in PB at diagnosis (range), % 41 (1-96) 60 (5-95) 37 (1-96) .057 
Direct bilirubin, median (range), mg/dL 0.8 (0-12.3) 1.0 (0.3-5.6) 0.7 (0-12.3) .134 
Hepatomegaly, median (range), cm 3 (0-8) 5 (0-8) 2 (0-7) .043 
Systemic edema, n (%) 26 (20) 11 (79) 15 (13) <.001 
Organ hemorrhage, n (%) 12 (9) 4 (29) 8 (7) .027 
Therapeutic interventions, n (%) 60 (47) 11 (79) 49 (43) .021 
LDAC, n (%) 46 (36) 6 (43) 40 (35) .586 
Exchange blood transfusion, n (%) 16 (13) 4 (29) 12 (11) .076 
Systemic steroid therapy, n (%) 24 (19) 8 (57) 16 (14) <.001 
Classification of GATA1 mutation     
High-expression type mutation, n (%) 57 (45) 6 (43) 51 (45) 1.000 
Low-expression type mutation, n (%) 58 (45) 6 (43) 52 (46) 1.000 
Unclassified mutation, n (%) 12 (9) 2 (14) 10 (9) .620 
Negative, n (%) 1 (1) 1 (1) 1.000 
Events , n (%) 42 (33) 14 (100) 28 (25) <.001 
Early deaths (age <9 mo), n (%) 14 (11) 14 (100) 0 (0) <.001 
Later phase deaths (after 9 mo), n (%) 5 (4) 0 (0) 5 (4) 1.000 
Leukemia development, n (%) 23 (18) 0 (0) 23 (20) .073 
All patients,
N = 128
Patients with early death, n = 14Patients without early death, n = 114P value 
Sex (male:female) 64:64 9:5 55:59 .396 
Median gestational age (range), wk 37 (29-40) 34 (29-38) 37 (31-40) .003 
Median birth weight (range), g 2588 (1438-3714) 2249 (1438-3044) 2624 (1598-3714) .035 
Median age at diagnosis (range), d 0 (0-8) 0 (0-8) 0 (0-8) .854 
Congenital heart disease, n (%) 87 (68) 5 (36) 82 (72) .012 
Other congenital abnormally, n (%) 14 (11) 2 (14) 12 (11) .651 
Chromosomal status, n     
Trisomy 21:mosaic trisomy 21:normal karyotype 126:1:1 14:0:0 112:1:1 1.000 
Median WBC count at diagnosis (range), ×109/L 48.3 (4.7-478.7) 157.3 (14.3-238.5) 44.3 (4.7-478.7) .006 
Median blasts percentage in PB at diagnosis (range), % 41 (1-96) 60 (5-95) 37 (1-96) .057 
Direct bilirubin, median (range), mg/dL 0.8 (0-12.3) 1.0 (0.3-5.6) 0.7 (0-12.3) .134 
Hepatomegaly, median (range), cm 3 (0-8) 5 (0-8) 2 (0-7) .043 
Systemic edema, n (%) 26 (20) 11 (79) 15 (13) <.001 
Organ hemorrhage, n (%) 12 (9) 4 (29) 8 (7) .027 
Therapeutic interventions, n (%) 60 (47) 11 (79) 49 (43) .021 
LDAC, n (%) 46 (36) 6 (43) 40 (35) .586 
Exchange blood transfusion, n (%) 16 (13) 4 (29) 12 (11) .076 
Systemic steroid therapy, n (%) 24 (19) 8 (57) 16 (14) <.001 
Classification of GATA1 mutation     
High-expression type mutation, n (%) 57 (45) 6 (43) 51 (45) 1.000 
Low-expression type mutation, n (%) 58 (45) 6 (43) 52 (46) 1.000 
Unclassified mutation, n (%) 12 (9) 2 (14) 10 (9) .620 
Negative, n (%) 1 (1) 1 (1) 1.000 
Events , n (%) 42 (33) 14 (100) 28 (25) <.001 
Early deaths (age <9 mo), n (%) 14 (11) 14 (100) 0 (0) <.001 
Later phase deaths (after 9 mo), n (%) 5 (4) 0 (0) 5 (4) 1.000 
Leukemia development, n (%) 23 (18) 0 (0) 23 (20) .073 

Bold indicates P < 0.05.

PB, peripheral blood.

P value was evaluated between patients with early death vs patients without nonearly death using Fisher exact test or Mann-Whitney U test.

Under costal margin.

Events were defined by death or leukemia development.

Of the 128 patients, 20 (16%) died, and early death (<9 months of age) occurred in 14 (11%). The causes of early death were as follows: multiple organ failure (n = 5), liver failure (n = 1), respiratory failure (n = 3), sepsis (n = 1), congenital heart disease (n = 1), and other reasons (n = 3; supplemental Table 3). Cumulative incidence rate (CIR) of early death at 9 months was 11.0% (95% confidence interval [CI], 5.3-16.2), and the leukemia development rate at 4 years was 20.9% (95% CI, 12.9-28.2; supplemental Figure 2). The early death group had a significantly lower gestational age (P = .003), lower birth weight (P = .035), higher WBC counts (P = .006), higher rate of organ hemorrhage (P = .012), and higher rate of systemic edema (P < .001) than the nonearly death group, which are poor prognostic factors associated with early death (Table 1). The median values and ranges of the 27 cytokines and the number of subjects for each cytokine are shown in supplemental Table 2. Five cytokines (IL-2, IL-12, IL-15, IL-17, and RANTES) with missing values in ≥10 patients were excluded from subsequent analyses (supplemental Figure 1A).

Relation between cytokine levels and clinical characteristics

The comparison between 29 patients with a high WBC count (≥100 × 109 cells/L, a poor prognostic factor in patients with TAM) and 99 patients without a high WBC count for 22 cytokine levels showed that the levels of 6 cytokines (IL-1b, IL-6, IL-7, IL-8, IL-13, and monocyte chemoattractant protein-1b) were significantly higher in patients with high WBC counts (Table 2). The association between expression types of GATA1 mutations and 22 cytokine levels is shown in supplemental Table 4. Six cytokines (IL-4, eotaxin, platelet-derived growth factor-BB, basic fibroblast growth factor, macrophage inflammatory protein 1β, and tumor necrosis factor α) were significantly higher in the high GATA1 expression group than in the low GATA1 expression group.

Table 2.

Serum concentrations (pg/mL) of cytokines between patients with TAM with or without a high WBC count

Patients with a high WBC count, n = 29Patients without a high WBC count, n = 99P value 
IL-1b, median (range) 3.13 (1.42-325.2) 2.49 (0.63-3662) .005 
IL-1ra 211.6 (19.71-868.1) 116.0 (8.85-9285) .064 
IL-4 4.06 (1.69-10.43) 3.79 (0.89-32.97) .698 
IL-5 2.295 (0.11-16.44) 2.47 (0.06-20.88) .625 
IL-6 117.9 (8.49-1537.9) 33.70 (1.75-6851) .003 
IL-7 31.67 (4.91-613.2) 12.10 (0.97-184.5) .007 
IL-8 93.43 (19.42-8350) 44.97 (8.81-37418) .023 
IL-9 26.71 (2.46-130.8) 21.89 (1.26-250.6) .225 
IL-10 14.69 (5.36-260.5) 11.62 (1.56-170.6) .066 
IL-13 19.84 (1.01-123.0) 8.980 (0.58-199.2) .029 
Eotaxin 176.7 (48.34-2265) 148.4 (12.50-892.0) .090 
PDGF-bb 5610 (300.0-18523) 4459 (61.75-18489) .060 
Basic FGF 60.79 (16.31-935.3) 47.42 (6.77-254.3) .079 
G-CSF 65.73 (25.02-7232) 55.71 (9.39-1770) .078 
GM-CSF 173.8 (47.04-851.1) 132.0 (6.05-1836) .081 
IFN-r 93.31 (16.52-597.9) 74.79 (10.22-6328) .606 
IP-10 2011 (296.0-18554) 1742 (70.84-18686) .602 
MCP-1(MCAF) 617.2 (86.33-4051) 195.8 (31.28-10398) .020 
MIP-1a 7.710 (0.53-759.1) 6.640 (0.71-565.6) .460 
MIP-1b 380.7 (121.7-4387) 280.1 (59.98-50908) .229 
TNF-a 50.40 (21.43-362.8) 40.81 (10.23-1029) .216 
VEGF 112.6 (12.17-1974) 76.85 (8.27-4490) .083 
Patients with a high WBC count, n = 29Patients without a high WBC count, n = 99P value 
IL-1b, median (range) 3.13 (1.42-325.2) 2.49 (0.63-3662) .005 
IL-1ra 211.6 (19.71-868.1) 116.0 (8.85-9285) .064 
IL-4 4.06 (1.69-10.43) 3.79 (0.89-32.97) .698 
IL-5 2.295 (0.11-16.44) 2.47 (0.06-20.88) .625 
IL-6 117.9 (8.49-1537.9) 33.70 (1.75-6851) .003 
IL-7 31.67 (4.91-613.2) 12.10 (0.97-184.5) .007 
IL-8 93.43 (19.42-8350) 44.97 (8.81-37418) .023 
IL-9 26.71 (2.46-130.8) 21.89 (1.26-250.6) .225 
IL-10 14.69 (5.36-260.5) 11.62 (1.56-170.6) .066 
IL-13 19.84 (1.01-123.0) 8.980 (0.58-199.2) .029 
Eotaxin 176.7 (48.34-2265) 148.4 (12.50-892.0) .090 
PDGF-bb 5610 (300.0-18523) 4459 (61.75-18489) .060 
Basic FGF 60.79 (16.31-935.3) 47.42 (6.77-254.3) .079 
G-CSF 65.73 (25.02-7232) 55.71 (9.39-1770) .078 
GM-CSF 173.8 (47.04-851.1) 132.0 (6.05-1836) .081 
IFN-r 93.31 (16.52-597.9) 74.79 (10.22-6328) .606 
IP-10 2011 (296.0-18554) 1742 (70.84-18686) .602 
MCP-1(MCAF) 617.2 (86.33-4051) 195.8 (31.28-10398) .020 
MIP-1a 7.710 (0.53-759.1) 6.640 (0.71-565.6) .460 
MIP-1b 380.7 (121.7-4387) 280.1 (59.98-50908) .229 
TNF-a 50.40 (21.43-362.8) 40.81 (10.23-1029) .216 
VEGF 112.6 (12.17-1974) 76.85 (8.27-4490) .083 

Bold indicates P < 0.05.

G-CSF, granulocyte colony-stimulating factor; GM-CSF, granulocyte-macrophage colony-stimulating factor; MCAF, monocyte chemotactic and activating factor; MCP-1, monocyte chemoattractant protein-1; MIP, macrophage inflammatory protein; PDGF, platelet-derived growth factor; TNF-a, tumor necrosis factor-alpha; VEGF, vascular endothelial growth factor.

A high WBC count was defined as >100 × 109 cells per liter.

P value was evaluated using the Mann-Whitney U test, followed by the Benjamini-Hochberg method used as a correction for multiple testing.

Furthermore, the correlation between cytokine levels and hepatomegaly and serum markers of liver fibrosis (procollagen type III peptide, type IV collagen, and hyaluronic acid) was evaluated (supplemental Table 5). The serum levels of several cytokines, especially IL-13 (r = 0.35; P < .05) and eotaxin (r = 0.35; P < .05), were correlated with hepatomegaly. Additionally, 5 cytokines (IL-6, IL-9, eotaxin, interferon-gamma-induced protein-10, and macrophage inflammatory protein 1β) were positively correlated with 2 liver fibrosis markers. Cytokine levels were compared in patients with or without leukemia development (n = 23 vs n = 105), and no cytokine showed significant differences between the 2 groups (supplemental Table 6).

Strong association between cytokine levels and early death

Cytokine levels were compared between the early death (n = 14) and nonearly death groups (n = 114). The levels of 5 cytokines (IL-1b [P = .031], IL-1ra [P = .007], IL-6 [P = .022], IL-8 [P = .004], and IL-13 [P = .037]) were significantly higher in the early death group than in the nonearly death group (Table 3). The optimal cytokine cutoff points of IL-1b, IL-1ra, IL-6, IL-8, and IL-13 were determined as 2.9 pg/mL, 256.0 pg/mL, 141.0 pg/mL, 102.0 pg/mL, and 9.2 pg/mL, respectively, to predict early death using receiver operator characteristic curves, which yielded the highest sum of sensitivity and specificity (supplemental Figure 3). The CIR of early death was significantly associated with higher levels of these 5 cytokines (Figure 1). Additionally, in a subgroup analysis restricted to 99 patients with low WBC counts (<100 × 109/L), high levels of these 5 cytokines were significantly associated with early death (supplemental Figure 4).

Table 3.

Serum concentrations (pg/mL) of cytokines between patients with TAM with or without early death

Patients with early death, n = 14Patients without early death, n = 114P value 
IL-1b, median (range) 4.435 (1.94-351.1) 2.630 (0.63-3662) .031 
IL-1ra 372.1 (101.9-1544) 122.0 (8.85-9285) .007 
IL-4 4.690 (1.69-9.35) 3.725 (0.89-32.97) .178 
IL-5 5.210 (0.26-20.88) 2.140 (0.06-17.35) .232 
IL-6 222.9 (23.55-1537) 37.43 (1.75-6851) .022 
IL-7 31.30 (4.43-613.2) 12.86 (0.97-301.1) .081 
IL-8 217.1 (28.92-8350) 45.87 (8.81-37418) .004 
IL-9 34.92 (2.46-110.2) 23.43 (1.26-250.6) .335 
IL-10 28.38 (1.85-260.5) 11.71 (1.56-170.6) .069 
IL-13 22.80 (3.52-173.5) 9.045 (0.58-199.2) .037 
Eotaxin 178.2 (46.17-2265) 150.4 (12.50-892.0) .424 
PDGF-bb 5755 (300.0-18523) 4568 (61.75-18489) .434 
Basic FGF 60.28 (19.26-935.3) 49.75 (6.77-254.3) .261 
G-CSF 58.34 (25.63-7232) 56.40 (9.39-1770) .354 
GM-CSF 141.7 (59.65-851.1) 138.1 (6.05-1836) .750 
IFN-r 106.4 (26.36-597.9) 74.79 (10.22-6328) .329 
IP-10 1360 (195.7-17268) 1857 (70.84-18686) .604 
MCP-1(MCAF) 658.6 (93.37-4051) 292.3 (31.28-10398) .222 
MIP-1a 7.79 (2.63-759.1) 6.660 (0.53-565.6) .347 
MIP-1b 454.5 (116.8-50908) 283.8 (59.98-40790) .185 
TNF-a 63.46 (21.97-362.8) 40.66 (10.23-1029) .065 
VEGF 105.7 (12.17-1974) 76.85 (8.27-4490) .275 
Patients with early death, n = 14Patients without early death, n = 114P value 
IL-1b, median (range) 4.435 (1.94-351.1) 2.630 (0.63-3662) .031 
IL-1ra 372.1 (101.9-1544) 122.0 (8.85-9285) .007 
IL-4 4.690 (1.69-9.35) 3.725 (0.89-32.97) .178 
IL-5 5.210 (0.26-20.88) 2.140 (0.06-17.35) .232 
IL-6 222.9 (23.55-1537) 37.43 (1.75-6851) .022 
IL-7 31.30 (4.43-613.2) 12.86 (0.97-301.1) .081 
IL-8 217.1 (28.92-8350) 45.87 (8.81-37418) .004 
IL-9 34.92 (2.46-110.2) 23.43 (1.26-250.6) .335 
IL-10 28.38 (1.85-260.5) 11.71 (1.56-170.6) .069 
IL-13 22.80 (3.52-173.5) 9.045 (0.58-199.2) .037 
Eotaxin 178.2 (46.17-2265) 150.4 (12.50-892.0) .424 
PDGF-bb 5755 (300.0-18523) 4568 (61.75-18489) .434 
Basic FGF 60.28 (19.26-935.3) 49.75 (6.77-254.3) .261 
G-CSF 58.34 (25.63-7232) 56.40 (9.39-1770) .354 
GM-CSF 141.7 (59.65-851.1) 138.1 (6.05-1836) .750 
IFN-r 106.4 (26.36-597.9) 74.79 (10.22-6328) .329 
IP-10 1360 (195.7-17268) 1857 (70.84-18686) .604 
MCP-1(MCAF) 658.6 (93.37-4051) 292.3 (31.28-10398) .222 
MIP-1a 7.79 (2.63-759.1) 6.660 (0.53-565.6) .347 
MIP-1b 454.5 (116.8-50908) 283.8 (59.98-40790) .185 
TNF-a 63.46 (21.97-362.8) 40.66 (10.23-1029) .065 
VEGF 105.7 (12.17-1974) 76.85 (8.27-4490) .275 

Bold indicates P < 0.05.

G-CSF, granulocyte colony-stimulating factor; GM-CSF, granulocyte-macrophage colony-stimulating factor; MCAF, monocyte chemotactic and activating factor; MCP-1, monocyte chemoattractant protein-1; MIP, macrophage inflammatory protein; PDGF, platelet-derived growth factor; TNF-a, tumor necrosis factor-alpha; VEGF, vascular endothelial growth factor.

P value was evaluated using the Mann-Whitney U test, followed by the Benjamini-Hochberg method used as a correction for multiple testing.

Figure 1.

Cytokine levels are valuable markers for predicting early death in patients with TAM. (A) CIR of early death between patients with TAM with high and low levels of IL-1b (high, n = 64; low, n = 64; CIR, 20.3% [95% CI, 9.8-29.6] vs 1.6% [95% CI, 0.0-4.6]; P < .001); (B) high and low levels of IL-1ra (high, n = 36; low, n = 90; CIR, 30.6% [95% CI, 13.8-44.1] vs 3.4% [95% CI, 0.0-7.0]; P < .001); (C) high and low levels of IL-6 (high, n = 28; low, n = 98; CIR, 32.1% [95% CI, 12.4-47.4] vs 4.1% [95% CI, 0.0-8.0]; P < .001); (D) high and low levels of IL-8 (high, n = 35; low, n = 90; CIR, 31.4% [95% CI, 14.2-45.2] vs 2.2% [95% CI, 0.0-5.3]; P < .001); and (E) high and low levels of IL-13 (high, n = 69; low, n = 59; CIR, 18.8% [95% CI, 9.1-27.6] vs 1.7% [95% CI, 0.0-4.9]; P < .001).

Figure 1.

Cytokine levels are valuable markers for predicting early death in patients with TAM. (A) CIR of early death between patients with TAM with high and low levels of IL-1b (high, n = 64; low, n = 64; CIR, 20.3% [95% CI, 9.8-29.6] vs 1.6% [95% CI, 0.0-4.6]; P < .001); (B) high and low levels of IL-1ra (high, n = 36; low, n = 90; CIR, 30.6% [95% CI, 13.8-44.1] vs 3.4% [95% CI, 0.0-7.0]; P < .001); (C) high and low levels of IL-6 (high, n = 28; low, n = 98; CIR, 32.1% [95% CI, 12.4-47.4] vs 4.1% [95% CI, 0.0-8.0]; P < .001); (D) high and low levels of IL-8 (high, n = 35; low, n = 90; CIR, 31.4% [95% CI, 14.2-45.2] vs 2.2% [95% CI, 0.0-5.3]; P < .001); and (E) high and low levels of IL-13 (high, n = 69; low, n = 59; CIR, 18.8% [95% CI, 9.1-27.6] vs 1.7% [95% CI, 0.0-4.9]; P < .001).

Close modal

An unsupervised clustering analysis was performed using the values of 22 cytokines. Missing values (1.1%; supplemental Figure 1B) were computationally imputed using the missForest method. The patients were divided into 3 groups: hot-1 (n = 37), hot-2 (n = 42), and cold (n = 49) (Figure 2A). The hot-1 group showed high inflammatory cytokine levels, including IL-8, IL-6, and IL-1β (Figure 2B). The hot-2 group was characterized by elevated IL-5 levels (Figure 2C). The cold group did not show any significant cytokine elevation (Figure 2D). The clinical characteristics of the 3 groups are described in supplemental Table 7. The CIR of early death was significantly different between the cytokine groups (hot-1/2, n = 79; cold, n = 49; hot-1/2 CIR, 16.5% [95% CI, 7.9-24.2]; cold CIR, 2.0% [95% CI, 0.0-5.9]; P = .013). The cytokine hot-1/2 groups showed significantly higher early mortality than the cytokine cold group (Figure 3; supplemental Figure 5). For sensitivity analyses, cluster analyses were conducted on data sets complemented using other imputation methods: kNN and PCA (supplemental Figure 6A-B). In addition, complete data analysis was performed for 43 patients without missing data for 27 cytokines (supplemental Figure 6C). The reproducibility of the 3 identified clusters was high while using the missForest-imputed data set as a reference; the concordance rates with the kNN, PCA, and complete data analysis were 0.94, 0.99, and 0.95, respectively (supplemental Figure 1D-E).

Figure 2.

Total of 128 patients with TAM are classified into 3 groups by an unsupervised consensus clustering analysis based on cytokine profiling. (A) Based on unsupervised clustering, patients were classified into 3 cytokine groups (hot-1 [n = 37], hot-2 [n = 42], and cold groups [n = 49]). Missing data (1.1%) in 22 cytokines were imputed using the missForest method. Black boxes indicate each clinical feature. Gray boxes indicate patients with no data. (B-D) The mean cytokine differences (x-axis) and the negative log10-transformed statistical P value (y-axis) between hot-1 group and other groups (B), hot-2 group and other groups (C), cold group and other groups (D) are shown in the volcano plot.

Figure 2.

Total of 128 patients with TAM are classified into 3 groups by an unsupervised consensus clustering analysis based on cytokine profiling. (A) Based on unsupervised clustering, patients were classified into 3 cytokine groups (hot-1 [n = 37], hot-2 [n = 42], and cold groups [n = 49]). Missing data (1.1%) in 22 cytokines were imputed using the missForest method. Black boxes indicate each clinical feature. Gray boxes indicate patients with no data. (B-D) The mean cytokine differences (x-axis) and the negative log10-transformed statistical P value (y-axis) between hot-1 group and other groups (B), hot-2 group and other groups (C), cold group and other groups (D) are shown in the volcano plot.

Close modal
Figure 3.

Cytokine group is significantly associated with the early death rate in patients with TAM. The CIR of early death in patients with TAM between cytokine hot-1/2 and cold groups (hot-1/2, n = 79; cold, n = 49; hot-1/2 CIR, 16.5% [95% CI, 7.9-24.2]; cold CIR, 2.0% [95% CI, 0.0-5.9]; P = .013).

Figure 3.

Cytokine group is significantly associated with the early death rate in patients with TAM. The CIR of early death in patients with TAM between cytokine hot-1/2 and cold groups (hot-1/2, n = 79; cold, n = 49; hot-1/2 CIR, 16.5% [95% CI, 7.9-24.2]; cold CIR, 2.0% [95% CI, 0.0-5.9]; P = .013).

Close modal

The univariable analysis showed that the following covariates were correlated with early death: cytokine group, gestational age, organ hemorrhage, systemic edema, congenital heart disease, high WBC counts in peripheral blood, systemic steroid therapy, and hepatomegaly (supplemental Table 8). The multivariable analysis was performed in 2 models using factors extracted using the stepwise Akaike information criterion method, which were identified as significantly different in univariable analysis. The multivariable analysis (model 1), without incorporating cytokine group, identified the following independent risk factors for early death: high WBC counts (hazard ratio [HR], 3.450; 95% CI, 1.127-10.56; P = .030), systemic edema (HR, 13.76; 95% CI, 3.784-50.06; P < .001), hepatomegaly (HR, 3.375; 95% CI, 1.108-10.28; P = .032), and congenital heart disease (HR, 0.294; 95% CI, 0.096-0.903; P = .033); and the multivariable analysis (model 2), incorporating cytokine group, showed that cytokine hot-1/2 groups was an independent prognostic factor (HR, 15.53; 95% CI, 1.434-168.3; P = .024; Table 4).

Table 4.

Multivariable Cox regression analyses of early death

CovariatesNumberMultivariable analysis: model 1 without incorporating cytokine groupMultivariable analysis: model 2 incorporating cytokine group
HR (95% CI)P valueHR (95% CI)P value
Cytokine group      
Cold 49 Exclusion  (1) .024 
Hot-1/2 79   15.53 (1.434-168.3)  
Systemic edema      
No 102 (1) <.001 (1) <.001 
Yes 26 13.76 (3.784-50.06)  19.24 (4.787-77.30)  
Congenital heart disease      
No 41 (1) .033 (1) .012 
Yes 87 0.294 (0.096-0.903)  0.174 (0.044-0.681)  
WBC      
<100 × 109 /L 99 (1) .030 (1) .607 
≥100 × 109 /L 29 3.450 (1.127-10.56)  1.383 (0.401-4.770)  
Hepatomegaly      
<5 cm 100 (1) .032 (1) .006 
≥5 cm 28 3.375 (1.108-10.28)  5.839 (1.639-20.80)  
Akaike information criterion  102.4 96.17 
Likelihood ratio  40.00 (P < .001) 48.21 (P < .001) 
CovariatesNumberMultivariable analysis: model 1 without incorporating cytokine groupMultivariable analysis: model 2 incorporating cytokine group
HR (95% CI)P valueHR (95% CI)P value
Cytokine group      
Cold 49 Exclusion  (1) .024 
Hot-1/2 79   15.53 (1.434-168.3)  
Systemic edema      
No 102 (1) <.001 (1) <.001 
Yes 26 13.76 (3.784-50.06)  19.24 (4.787-77.30)  
Congenital heart disease      
No 41 (1) .033 (1) .012 
Yes 87 0.294 (0.096-0.903)  0.174 (0.044-0.681)  
WBC      
<100 × 109 /L 99 (1) .030 (1) .607 
≥100 × 109 /L 29 3.450 (1.127-10.56)  1.383 (0.401-4.770)  
Hepatomegaly      
<5 cm 100 (1) .032 (1) .006 
≥5 cm 28 3.375 (1.108-10.28)  5.839 (1.639-20.80)  
Akaike information criterion  102.4 96.17 
Likelihood ratio  40.00 (P < .001) 48.21 (P < .001) 

Bold indicates P < 0.05.

A total of 22 cytokine levels were analyzed in 128 patients with DS with TAM who were enrolled in the TAM-10 prospective observational study to determine the association between cytokine levels and clinical characteristics. Five cytokines (IL-1b, IL-1ra, IL-6, IL-8, and IL-13) were significantly associated with early death. Furthermore, an unsupervised clustering analysis based on the 22 cytokine levels generated 3 groups (cytokine hot-1, hot-2, and cold). The cytokine hot-1/2 groups showed significantly higher early death rates than the cytokine cold group.

The univariable analysis showed a strong association between the cytokine hot-1/2 groups and early death (HR, 8.509; 95% CI, 1.113-65.05), and a multivariable model incorporating cytokine group (model 2) identified the cytokine hot-1/2 groups as an independent prognostic factor. These findings indicate that the cytokine group is a potent prognostic factor for TAM and may outperform the traditional clinical prognostic factor, WBC count.

The IL-1 family consists of proinflammatory cytokines such as IL-1b and anti-inflammatory cytokines such as IL-1ra.14 IL-1b is a potent proinflammatory cytokine, originally identified as an endogenous thermogenic agent, and IL-1ra is an acute phase protein secreted by the liver in response to inflammatory stimuli and can inhibit signal transduction.15 It has been reported that patients with TAM who died early had significantly elevated levels of both IL-1b and IL-1ra. However, IL-1ra is considered much less effective than agonists, requiring up to 1000-fold excess IL-1ra to inhibit IL-1 signaling.16 These findings suggest that the observed IL-1ra elevation is a secondary event, and IL-1 signaling is activated in patients with severe TAM. IL-6 promotes B and T lymphocyte differentiation and immunoglobulin G production.17,18 Furthermore, IL-6 has been reported to be involved in cancer cell proliferation via STAT3 activation19 and promote cancer cell migration and invasion.20-22 Shitara et al23 reported a case of severe TAM that showed IL-6 elevation in the pericardial fluid. Targeted therapy with cytokine antagonists, such as anakinra and canakinumab, which inhibit IL-1 signaling, and tocilizumab, which inhibits IL-6, have already been approved and demonstrated clinical efficacy for the treatment of hypercytokinemia in various diseases. These cytokine antagonists are expected to be evaluated in clinical studies as a potential future treatment for hypercytokinemia in severe TAM.

The correlation between cytokine levels and other clinical features was evaluated, except for early death. This study revealed that no cytokine levels were associated with leukemia development. Only the flow cytometric minimal residual disease positivity has been reported to be a valuable marker for predicting leukemia development.2,24 These results implied that it might be difficult to predict leukemia development from any data at the time of diagnosis. Furthermore, the association between cytokine levels and GATA1 expression type was investigated. The results showed that 7 cytokine levels were significantly associated with the GATA1 expression type. All 7 cytokine levels were higher in patients in the GATA1 high-expression group than in those in the GATA1 low-expression group. Kanezaki25 reported that the mutation types of GATA1 affected the amount of the mutant, and the GATA1 expression type significantly affected the TAM phenotype. The study findings might imply that the GATA1 high-expression type caused high levels of their cytokines.

This study has several limitations. First, this study included patients enrolled in the JPLSG TAM-10 study, and clinical samples immediately after diagnosis for cytokine measurement in 23% (39 patients) were unavailable and could not be included in the analysis. Most clinical characteristics did not show significant differences between patients with and without cytokine information; however, WBC counts at diagnosis, blast rates, and percentage of patients receiving LDAC were significantly higher in cases with cytokine information (supplemental Table 1). Second, of the 27 cytokines measured, the percentage of deficient values for 22 cytokines used in the analysis was only 1.1% (supplemental Figure 1B); however, 5 cytokines were deficient in >10% of cases and had to be excluded from subsequent analyses. Moreover, we performed a complete data analysis of 43 cases for which we had data for all 27 cytokines and found consistent results (supplemental Figure 1D-E). Third, the dosage and intervention criteria of LDAC were not standardized, although a relatively high percentage (36%) of patients were treated with LDAC as per the policy of the participating centers. Fourth, the clinical significance of cytokine profiling analysis has not been validated due to the absence of a validation cohort. This limitation is largely unavoidable given the rarity of TAM and the scarcity of international prospective studies in this field. However, we plan to re-evaluate the cytokine profiling analysis in the future using clinical samples from patients enrolled in our ongoing prospective clinical trial (jRCTs041190063).

In conclusion, this study showed that cytokine profiling provides supportive information along with previous clinical prognostic factors such as WBC count as a biomarker for predicting early death and may contribute to precision medicine for patients with TAM.

The authors thank Enago for the English language review.

This work was supported by Health and Labor Sciences Research grants 201128038B and 201324122A and the Japan Agency for Medical Research and Development Innovative Cancer Medical Practice Research Project (15Aek0109055h0002 and 18ck0106435h0001).

Contribution: G.Y. conducted the study, analyzed the data, and wrote the manuscript; Y.H. designed and conducted the study, led the project, and wrote the manuscript; Y.T. and H.M. wrote the manuscript and analyzed the data; A.S., N.S., T. Kaburagi, T.D., T. Kawai, and Y.Y. analyzed the data; T.I. performed statistical analyses; H.T. performed the research and bioinformatics analysis; Y.T. wrote the manuscript; K.T. and E.I. performed the GATA1 mutation analysis; K.W. collected clinical samples and data; and all authors critically reviewed and revised the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Yasuhide Hayashi, Department of Hematology/Oncology, Gunma Children's Medical Center, 779, Shimohakoda, Hokkitsu, Shibukawa, 377-8577 Japan; email: [email protected].

1.
Pine
SR
,
Guo
Q
,
Yin
C
,
Jayabose
S
,
Druschel
CM
,
Sandoval
C
.
Incidence and clinical implications of GATA1 mutations in newborns with down syndrome
.
Blood
.
2007
;
110
(
6
):
2128
-
2131
.
2.
Yamato
G
,
Deguchi
T
,
Terui
K
, et al
.
Predictive factors for the development of leukemia in patients with transient abnormal myelopoiesis and Down syndrome
.
Leukemia
.
2021
;
35
(
5
):
1480
-
1484
.
3.
Gamis
AS
,
Alonzo
TA
,
Gerbing
RB
, et al
.
Natural history of transient myeloproliferative disorder clinically diagnosed in Down syndrome neonates: a report from the Children’s Oncology Group Study A2971
.
Blood
.
2011
;
118
(
26
):
6752
-
6996
.
4.
Klusmann
JH
,
Creutzig
U
,
Zimmermann
M
, et al
.
Treatment and prognostic impact of transient leukemia in neonates with Down syndrome
.
Blood
.
2008
;
111
(
6
):
2991
-
2998
.
5.
Massey
GV
,
Zipursky
A
,
Chang
MN
, et al
.
Children’s Oncology Group (COG). a prospective study of the natural history of transient leukemia (TL) in neonates with Down syndrome (DS): Children’s Oncology Group (COG) study POG-9481
.
Blood
.
2006
;
107
(
12
):
4606
-
4613
.
6.
Muramatsu
H
,
Kato
K
,
Watanabe
N
, et al
.
Risk factors for early death in neonates with Down syndrome and transient leukaemia
.
Br J Haematol
.
2008
;
142
(
4
):
610
-
615
.
7.
Tunstall
O
,
Bhatnagar
N
,
James
B
, et al;
British Society for Haematology
.
Guidelines for the investigation and management of transient leukaemia of Down syndrome
.
Br J Haematol
.
2018
;
182
(
2
):
200
-
211
.
8.
Kinjo
T
,
Inoue
H
,
Kusuda
T
, et al
.
Chemokine levels predict progressive liver disease in Down syndrome patients with transient abnormal myelopoiesis
.
Pediatr Neonatol
.
2019
;
60
(
4
):
382
-
388
.
9.
Kato
M
,
Tsukagoshi
H
,
Yoshizumi
M
, et al
.
Different cytokine profile and eosinophil activation are involved in rhinovirus- and RS virus-induced acute exacerbation of childhood wheezing
.
Pediatr Allergy Immunol
.
2011
;
22
(
1 Pt 2
):
e87
-
e94
.
10.
Stekhoven
DJ
,
Bühlmann
P
.
MissForest -- non--parametric missing value imputation for mixed-type data
.
Bioinformatics
.
2012
;
28
(
1
):
112
-
118
.
11.
Josse
J
,
Husson
F
.
missMDA: a package for handling missing values in multivariate data analysis
.
J Stat Softw
.
2016
;
70
(
1
):
1
-
31
.
12.
Kanda
Y
.
Investigation of the freely available easy-to-use software 'EZR' for medical statistics
.
Bone Marrow Transplant
.
2013
;
48
(
3
):
452
-
458
.
13.
Wilkerson
MD
,
Hayes
DN
.
ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking
.
Bioinformatics
.
2010
;
26
(
12
):
1572
-
1573
.
14.
Turner
MD
,
Nedjai
B
,
Hurst
T
,
Pennington
DJ
.
Cytokines and chemokines: at the crossroads of cell signalling and inflammatory disease
.
Biochim Biophys Acta
.
2014
;
1843
(
11
):
2563
-
2582
.
15.
Boraschi
D
,
Lucchesi
D
,
Hainzl
S
, et al
.
IL-37: a new anti-inflammatory cytokine of the IL-1 family
.
Eur Cytokine Netw
.
2011
;
22
(
3
):
127
-
147
.
16.
Gabay
C
,
Lamacchia
C
,
Palmer
G
.
Palmer. Il-1 pathways in inflammation and human diseases
.
Nat Rev Rheumatol
.
2010
;
6
(
4
):
232
-
241
.
17.
Muraguchi
A
,
Kishimoto
T
,
Miki
Y
, et al
.
T cell-replacing factor-(TRF) induced IgG secretion in a human B blastoid cell line and demonstration of acceptors for TRF
.
J Immunol
.
1981
;
127
(
2
):
412
-
416
.
18.
Lotz
M
,
Jirik
F
,
Kabouridis
P
, et al
.
B cell stimulating factor 2/interleukin 6 is a costimulant for human thymocytes and T lymphocytes
.
J Exp Med
.
1988
;
167
(
3
):
1253
-
1258
.
19.
Ara
T
,
Declerck
YA
.
Interleukin-6 in bone metastasis and cancer progression
.
Eur J Cancer
.
2010
;
46
(
7
):
1223
-
1231
.
20.
Walter
M
,
Liang
S
,
Ghosh
S
,
Hornsby
PJ
,
Li
R
.
Interleukin 6 secreted fromadipose stromal cells promotes migration and invasion of breast cancer cells
.
Oncogene
.
2009
;
28
(
30
):
2745
-
2755
.
21.
Foran
E
,
Garrity-Park
MM
,
Mureau
C
, et al
.
Upregulation of DNA methyltransferase-mediated gene silencing, anchorage-independent growth, and migration of colon cancer cells by interleukin-6
.
Mol Cancer Res
.
2010
;
8
(
4
):
471
-
481
.
22.
Mano
Y
,
Aishima
S
,
Fujita
N
, et al
.
Tumor-associated macrophage promotes tumor progression via STAT3 signaling inhepatocellular carcinoma
.
Pathobiology
.
2013
;
80
(
3
):
146
-
154
.
23.
Shitara
Y
,
Takahashi
N
,
Aoki
Y
, et al
.
Cytokine profiles in pericardial effusion in a Down syndrome infant with transient abnormal myelopoiesis
.
Tohoku J Exp Med
.
2017
;
241
(
2
):
149
-
153
.
24.
Flasinski
M
,
Scheibke
K
,
Zimmermann
M
, et al
.
Low-dose cytarabine to prevent myeloid leukemia in children with Down syndrome: TMD prevention 2007 study
.
Blood Adv
.
2018
;
2
(
13
):
1532
-
1540
.
25.
Kanezaki
R
,
Toki
T
,
Terui
K
, et al
.
Down syndrome and GATA1 mutations in transient abnormal myeloproliferative disorder: mutation classes correlate with progression to myeloid leukemia
.
Blood
.
2010
;
116
(
22
):
4631
-
4638
.

Author notes

Data are available on request from the corresponding author, Yasuhide Hayashi ([email protected]).

The full-text version of this article contains a data supplement.