DNAM-1/CD226 is functionally expressed on acute myeloid leukemia (AML) cells and is associated with favorable prognosis

DNAM-1 is reportedly expressed on cytotoxic T and NK cells and, upon interaction with its ligands CD112 and CD155, plays an important role in tumor immunosurveillance. It has also been reported to be functionally expressed by myeloid cells, but expression and function on malignant cells of the myeloid lineage have not been studied so far. Here we analyzed expression of DNAM-1 in leukemic cells of acute myeloid leukemia (AML) patients. We found substantial levels of DNAM-1 to be expressed on leukemic blasts in 48 of 62 (> 75%) patients. Interaction of DNAM-1 with its ligands CD112 and CD155 induced release of the immunomodulatory cytokines IL-6, IL-8 IL-10 and TNF-α by AML cells and DNAM-1 expression correlated with a more differentiated phenotype. Multivariate analysis did not show any association of DNAM-1 positivity with established risk factors, but expression was significantly associated with clinical disease course: patients with high DNAM-1 surface levels had significantly longer progression-free and overall survival compared to DNAM-1low patients, independently whether patients had undergone allogenic stem cell transplantation or not. Together, our findings unravel a functional role of DNAM-1 in AML pathophysiology and identify DNAM-1 as a potential novel prognostic maker in AML.

Flow cytometry. DNAM-1 surface expression was assessed by staining with a specific antibody or isotype control followed by incubation with goat-anti-mouse IgG-PE. Samples were analyzed using FACSCanto II instrument (BD Biosciences).
Leukemic cells in patient samples were first selected by FSC/SSC and then gates were applied according to the surface markers based on the individual immunophenotype defined upon routine diagnosis (CD33, CD34 or CD117). Dead cells were excluded based on 7-aminoactinomycin staining (7-AAD, BD Biosciences). Specific fluorescence indices (SFI) were calculated by dividing median fluorescence values measured with specific antibody by median fluorescence values obtained with isotype control.

Measurements of cellular viability and cytokine production.
For stimulation with DNAM-1 ligands, 96-well cell culture plates (Greiner Bio-One) were coated with 10 µg/ml solution of rhCD112, rhCD155 or hIgG1 in PBS overnight. Afterwards wells were washed with PBS and freshly thawed AML cells were plated in concentration 2 × 10 6 cells/ml. Cytokines in supernatants from cultured AML blasts were quantified by ELISA according to the manufacturer's instructions using OptEIA assays from BD Pharmingen. Viability was measured using CellTiter-Glo® Luminescent Cell Viability Assay (Promega) according to the manufacturer's instructions.
Statistical analysis. Statistical  www.nature.com/scientificreports/ Log-rank test was performed for estimating survival differences between the groups. Cutoff value for separation of individuals into DNAM-1 high and DNAM-1 low was determined using receiver operating characteristic (ROC) analysis in JMP® Pro as previously described 13 . Value of highest Youden's index was used as cutoff. P values of < 0.05 were considered statistically significant. Patients with missing data were excluded from statistical analysis.

Results
Patient characteristics. The clinical characteristics of the patient cohort are summarized in Table 1 DNAM-1 is expressed on malignant myeloid cell lines and leukemic blasts from AML patients. As a first step, we used flow cytometry to determine DNAM-1 surface expression on cancer cell lines representing a wide range of myeloid malignancies: K562 (chronic myeloid leukemia), U937 (promyelocytic leukemia), KG-1 (AML M6), EOL-1 (AML M4), Kasumi-1 (AML M2), HL-60 (AML M2), TF-1 (AML M6) (Fig. 1A). We detected no relevant expression in KG-1 and K-562 (SFI > 1.5 as threshold defined for positivity), whereas modest expression was observed with EOL-1, Kasumi and HL-60 cells (SFI = 1.8, 1.7 and 1.7, respectively). TF-1 and U937 cells showed strong DNAM-1 positivity (SFI = 12 and 7.5, respectively). We further confirmed DNAM-1 expression by quantifying mRNA expression in leukemic cells using qRT-PCR ( Fig. 1B,C). Notably, DNAM-1 transcript expression showed no correlation with surface levels, which indicates that DNAM-1 surface expression may be regulated posttranscriptionally. Next, we analyzed DNAM-1 expression on leukemic blasts from AML patients. The gaiting strategy and examples of surface staining are shown in Fig. 1D,E. Again, considering AML specimen with SFI ≥ 1.5 to be DNAM-1 positive, 77% (48 of 62) of all AML cases were found to express substantial DNAM-1 surface levels (Fig. 1F). Individual SFI values of all patients are given in Table 1. The number of positive cells within the leukemic population of individual patients varied considerably, ranging from 0% to almost 100% (Fig. 1G). When DNAM-1 expression in the AML samples was quantified using RT-PCR (Fig. 1H,I), alike with the cell lines, DNAM-1 transcripts were detectable in some samples without relevant surface expression on leukemic cells. In line, no correlation between DNAM-1 mRNA levels quantified by qRT-PCR and surface expression was observed (Fig. 1I). These results further hint at posttranscriptional regulation of surface DNAM-1 levels. Together, our data reveal widespread expression of DNAM-1 in a substantial number of AML cases.

DNAM-1 stimulates cytokine release by AML cells.
As interaction of DNAM-1 with its ligands promotes activation and cytokine secretion in various cell types [14][15][16] , we next set out to determine whether DNAM-1 was functional in AML cells. To this end we analyzed cytokine production in cultures of primary AML specimens upon stimulation with the two different DNAM-1 ligands CD112 and CD155. We selected thirteen AML samples with high DNAM-1 expression on leukemic cells (SFI ≥ 5, 10-90% positive cells, median positivity 70%) and incubated them on immobilized recombinant human DNAM-1 ligands (rhCD112-Fc or rhCD155-Fc) or isotype control (human IgG1) for 24 h. Afterwards secretion of the immunomodulatory cytokines IL-6, IL-8, TNF-α and IL-10 that are implicated in AML pathophysiology 17,18 were analyzed by ELISA. Exemplary cytokine profiles obtained with 2 DNAM-1-positive (UPN34 and UPN20) and 1 negative specimen (UPN18) as control are shown in Fig. 2A. Treatment with DNAM-1 ligands potently stimulated cytokine release in the positive samples, whereas no effect was observed with DNAM-1-negative leukemia. Stimulation with LPS, a non-specific inducer of cytokine secretion in leukemia cells 19 , induced release of cytokines in DNAM-1 negative samples, confirming that effects observed with positive leukemic cells were specifically mediated by DNAM-1 signaling ( Fig. 2A). Substantial interindividual differences with regard to cytokine release were observed (Fig. 2B,C). Out of 10 samples that responded to stimulation, none responded with release of all four cytokines. In one sample, production of IL-6, IL-8 and IL-10 was observed, whereas 5 and 4 samples secreted at least 2 and only one of the cytokines, respectively (Fig. 2C). Of note, no correlation of DNAM-1 expression with release of the cytokines was observed (Fig. 2D).
Together, these findings demonstrate that DNAM-1 is functional in AML cells with substantial interindividual differences with regard to the consequences of signaling that may differentially contribute to the cytokine milieu.
Analysis of the correlation of DNAM-1 expression with factors affecting prognosis did not reveal an association for NCCN risk class (Fig. 3E), primary versus secondary disease (Fig. 3F) and age (< 60 vs. ≥ 60 years) (Fig. 3G). Moreover, no relationship between DNAM-1 positivity and white blood count was observed (Fig. 3H). Analysis of correlation of DNAM-1 expression with AML subtypes according to WHO criteria revealed significantly lower DNAM-1 expression among AML cases classified as "AML with myelodysplasia-related changes" compared to "AML with recurrent genetic aberrations" and "AML, not otherwise specified" (p = 0.004 and p = 0.04 respectively) (Fig. 3I). Expression of DNAM-1 did not show any correlation with NPM1 and CEBPA mutations (Fig. 3J,K).

DNAM-1 expression correlates with survival of AML patients.
To determine whether DNAM-1 expression has prognostic value in AML, we correlated DNAM-1 surface positivity with survival of 59 patients; for 3 individuals survival data were not available. For initial analysis, individuals were divided into quartiles according to DNAM-1 expression, and overall survival (OS) in each quartile was assessed. Kaplan-Meier analysis did not reveal any difference in survival between the four groups (p = 0.49) (Fig. 4A). Next, we employed receiver operating characteristic (ROC) analysis to determine a predictive cut-off for separation of patients into groups. A cut-off value of 31.3% was selected to subdivide AML samples into DNAM-1 high (n = 20) and DNAM-1 low (n = 39) groups (Fig. 4B). Upon separation of patients according to this threshold, DNAM-1 high patients displayed significantly longer OS (hazard ratio 0.43, p = 0.04, Fig. 4C). Progression free survival (PFS) analysis did not reveal statistically significant differences between the groups (p = 0.05), but demonstrated a similar trend (Fig. 4D). When only patients who received hematopoietic stem cell transplantation (HSCT) were included in the analysis, an even more pronounced difference between the groups was revealed: individuals in the DNAM-1 high group showed significantly longer OS and PFS (p = 0.0048 and 0.0033 respectively) (Fig. 4E,F). These results identify DNAM-1 as potential prognostic marker in AML. 7,8,20,21 . Whereas DNAM-1 is mostly known for its cytotoxicity-stimulating properties in NK cells and T cells, in myeloid cells it was reported to influence migration 8 . Beyond healthy cells, DNAM-1 expression was also reported for chronic lymphocytic leukemia (CLL) 22 . Whether DNAM-1 is expressed in malignant myeloid cells and whether it contributes to disease pathophysiology is so far unknown.

DNAM-1 is an important mediator of effector functions in NK cells and T cells. Besides cytotoxic lymphocytes, functionally relevant DNAM-1 expression was described on B cells, but also on healthy cells of the myeloid lineage like monocytes and platelets
Here we report that DNAM-1 is expressed in cell lines of myeloid lineage and primary leukemic cells of AML patients. When we investigated the functionality of the DNAM-1 in AML cells, we found that interaction with its ligands resulted in production of the cytokines IL-6, IL-8, TNF-α and IL-10. This is in line with reports that interaction of DNAM-1 on cytotoxic cells with CD155 or CD112 triggers signaling via immunoreceptor tail tyrosine (ITT)-like motifs and stimulates cytokine secretion 14,23 . No correlation between the extent of cytokine release and the surface levels of DNAM-1 was observed, suggesting that the signaling capacity of the receptor is influenced by other factors, for example by co-expression of inhibitory molecules such as CD96 24,25 .
IL-6 and IL-8 are implicated to affect cellular survival in various cancer entities including AML [26][27][28][29][30][31] and associate with disease outcome [32][33][34] . IL-10 reportedly influences proliferation and cytokine production by AML cells 26,35,36 . Elevated plasma levels of IL-10 correlate with better survival of patients. For TNF-α controversial effects in AML pathogenesis have been reported 37 . Notably, variable patterns of cytokine release upon DNAM-1 signaling were observed. Whereas IL-6 or IL-8 were released by more than a half of the samples tested, secretion of IL-10 www.nature.com/scientificreports/ and TNF-α was observed less frequently. Likely the ultimate effect of DNAM-1 signaling may be is determined by combinatorial action of several secreted factors. Regarding the association of DNAM-1 expression and clinical parameters, we found that more mature AML cells displayed higher DNAM-1 positivity compared to immature FAB subclasses. Thus, DNAM-1 may serve as additional maturation marker upon immunophenotyping. Further in line with the association of DNAM-1 positivity and a more differentiated phenotype, we observed substantially higher expression on mature CD34-negative subpopulations of leukemic cells.
When we employed ROC analysis and separated AML patients in two groups according to a defined cut-off value of 31.3% of DNAM-1 positive cells, a significant correlation of high DNAM-1 surface levels with better OS was observed. Notably, high DNAM-1 expression was also associated with improved OS and PFS in patients after HSCT.
The role of NK cell-expressed DNAM-1 in recognition and killing of AML cells is well established. Interaction of DNAM-1 with its ligands CD112 and CD155 on AML cells is required for NK cell activation and cytotoxicity 47 . Attenuated expression of DNAM-1 on NK cells or DNAM-1 ligands on AML cells was suggested as an immune escape mechanism 48 . In agreement with this, higher expression of DNAM-1 on NK cells was associated with increased survival in AML patients 49 . The data presented here show that elevated DNAM-1 expression on AML cells themselves was also associated with improved survival of AML patients. So far, it remains unclear to which extent improved survival of individuals in the DNAM-1 high group can in fact be attributed to engagement of the signaling axis or is rather an epiphenomenon of the association with more differentiated status of AML cells. This is even more since, as stated above, DNAM-1 expression was found to be associated with a CD34-negative blast phenotype and survival of patients negatively correlates with the amount of CD34-positive leukemic cells 50 .
To conclude, we here provide first evidence that DNAM-1 is functionally expressed on AML cells and expression correlates with survival of the patients. Based on the data reported, confirmatory studies in larger cohorts are warranted and hold promise to establish DNAM-1 expression as a prognostic marker for risk stratification and ultimately to improve outcome for AML patients.  www.nature.com/scientificreports/ www.nature.com/scientificreports/

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.