Dipeptidylpeptidase 4 promotes survival and stemness of acute myeloid leukemia stem cells

SUMMARY

Leukemic-stem-cell-specific targeting may improve the survival of patients with acute myeloid leukemia (AML) by avoiding the ablative effects of standard regimens on normal hematopoiesis. Herein, we perform an unbiased screening of compounds targeting cell surface proteins and identify clinically used DPP4 inhibitors as strong suppressors of AML development in both murine AML models and primary human AML cells xenograft model. We find in retrovirus-induced AML mouse models that DPP4-deficient AML cell-transplanted mice exhibit delay and reversal of AML development, whereas deletion of DPP4 has no significant effect on normal hematopoiesis. DPP4 activates and sustains survival of AML stem cells that are critical for AML development in both human and animal models via binding with Src kinase and activation of nuclear factor κB (NF-κB) signaling. Thus, inhibition of DPP4 is a potential therapeutic strategy against AML development through suppression of survival and stemness of AML cells.

Graphical Abstract

In brief

Chen et al. show that leukemic cells, when compared with normal hematopoietic stem cells, overexpress a cell surface protein called DPP4, which is essential for their survival and confers stemness. Currently available inhibitors of this cell surface protein exhibit potential as effective leukemia therapeutic agents.

INTRODUCTION

Acute myeloid leukemia (AML) is the most common acute leukemia in adults and is responsible for the highest number of annual deaths among leukemias in the United States. Recent advances have resulted in targeted therapies and low-intensity regimens tailored toward mutation-positive AML and subjects >65 years old, respectively.¹ Although the novel treatments significantly improve clinical outcome for some AML subsets, their benefit to a large proportion of patients has been limited. The overall prognosis remains poor with 5 year survival of only 29%. A major challenge in the field of AML therapy is that a limited number of specific targets have been identified in AML cells. Moreover, most current treatments affect normal hematopoietic cells, resulting in significant adverse effects.² In this regard, standard chemotherapy with “cytarabine + anthra-cycline” or allogeneic hematopoietic stem cell transplant (alloSCT) is associated with significant morbidity and mortality.³ Therefore, an effective yet tolerable therapy for AML is an urgent unmet need.

Cell surface proteins are important therapeutic targets due to their role as cell markers and their extracellular accessibility.⁴ Importantly, the expression and function of these cell surface proteins and downstream signaling may differ in leukemia cells from those of normal hematopoietic cells.^5,6 We screened a group of compounds, which are thought to target cell surface proteins expressed in both human and mouse AML cells, and identified CD26/dipeptidylpeptidase 4 (DPP4), as a potential target for AML treatment. DPP4 exists primarily as a membrane-bound extracellular peptidase that cleaves Xaa-Pro or Xaa-Ala dipeptides from the N terminus of polypeptides.⁷ DPP4 is expressed by many types of cells and in different organ systems, including the hematopoietic system.⁸ DPP4 substrates G-CSF and C-X-C motif chemokine 12 (CXCL12) have been shown to play a role in the regulation of hematopoietic stem cell (HSC) trafficking.^9,10 DPP4 has been identified as a biomarker and target of leukemia stem cells for both chronic myeloid leukemia (CML) and acute lymphoblastic leukemia (ALL).^11-13 In contrast, DPP4-truncated granulocyte-macrophage colony-stimulating factor (GM-CSF) and interleukin-3 (IL-3) have been shown to inhibit normal and malignant hematopoiesis.¹⁴ Recently, inhibition of DPP4 activity in AML exosomes was shown to reverse exosome-mediated myelosuppression.¹⁵ Nevertheless, the function of DPP4 in AML cells and AML development is unknown. Herein, we demonstrate that inhibition or deletion of DPP4 suppressed AML development.

RESULTS

Growth of AML cells is dependent on DPP4

To identify potential cell surface targets that are effective for AML treatment, we explored compound libraries (MedChemExpress, #HY-L022, #HY-L026) for more than 1,200 compounds with clinical utility that target cell surface proteins and identified 57 candidate compounds to perform further phenotype-based in vitro screening (Figures 1A-1D and S1A). We tested the efficacy on cell growth in 3 human leukemic cell lines and inhibition of colony-forming unit ability (CFU) of primary mouse AML cells. We observed that linagliptin, a highly DPP4-specific inhibitor and a medication widely used in clinics for type two diabetes mellitus (T2DM), displayed the most significant and consistent suppression of growth of both human and mouse AML cells (Figures 1A-1D and S1B). To further assess the dependence of human AML (hAML) development on DPP4, we selected THP-1 cells, which consistently manifest the highest cell surface expression and enzyme activity of DPP4 among the human leukemic cell lines examined (Figures 1E and 1F). We found that both linagliptin and vildagliptin (another DPP4 inhibitor) significantly suppressed hAML cell growth by increasing apoptosis in vitro in a dose-dependent manner, with IC50s of 223.1 and 1,084.8 nM, respectively (Figures 1G, 1H, S1C, and S1D). In addition, two different short hairpin RNAs (shRNAs) independently and efficiently decreased total expression of DPP4 and essentially blocked the in vitro growth of AML cells, likely via significantly increasing apoptosis of leukemia cells, but did not alter the cell-cycle status, which is consistent with the effects of DPP4 inhibitors (Figures 1I-1L and S1E).

Figure 1. Growth of AML cells is dependent on DPP4.

(A–C) Cell numbers of THP1 (A), MV4-11 (B), and U937 (C) on day 6 of treatment with 57 select compounds (details in Figure S1A), each at a concentration of 10 μM (n = 3 wells). The dotted red line indicates the cell number with DMSO treatment (control), and the red bar indicates treatment with compound 37 (linagliptin, DPP4 inhibitor).

(D) Proportion of apoptotic THP1 cells on day 6 of treatment with 57 select compounds (n = 3 wells).

(E) The DPP4 enzyme activity of 1 × 10⁶ human leukemic cells was measured by a fluorescence assay. Results are reported as relative light units (RLUs) (n = 3 wells).

(F) Human leukemic cells’ medium fluorescent intensity (MFI) of DPP4 relative to isotype as determined by flow cytometry (n = 3 wells).

(G) DPP4 inhibitors (linagliptin 200 nM, vildagliptin 300 nM) suppress the growth of THP1 cells (n = 3 wells).

(H) Apoptosis in linagliptin-treated THP1 cells at indicated concentrations (n = 3 wells).

(I–K) Treatment with shRNA-targeting DPP4 inhibits the growth of THP1 cells. Two different DPP4-targeting shRNAs inhibit the growth of THP1 cells (I, n = 3 wells). Knockdown of DPP4 in THP1 cells was evaluated by western blot (WB) (J). Proportion of apoptotic cells was measured on the indicated days (K, n = 3 samples). (L) DPP4 knockdown inhibited the growth of MV4-11, MOLM, Kasumi1, and MM6 cells as measured on day 6 (n = 3 wells).

Error bars, SEM. *p < 0.05, ***p < 0.001. Data were analyzed using two-tailed Student’s t test; when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses.

DPP4 inhibitors suppress AML development, not the activity of HSCs and progenitors (HSPCs)

DPP4 is expressed on the majority of leukemic stem cell (LSC)-enriched populations, identified as YFP⁺Kit⁺Mac1⁺ cells in the bone marrow (BM) of MLL-AF9 AML mice.¹⁶ Furthermore, the more restricted AML stem cell (AML-SC) population in this model, GMP-like leukemic (L-GMP) cells (YFP⁺Lin⁻Sca-1⁻Kit⁺CD34⁺FcγRII/III⁺),¹⁷ had increased levels of DPP4 expression when compared with normal GMP cells (YFP⁻ GMP) after MLL-AF9 transformation (Figure S2A), which is consistent with DPP4 being a SC biomarker similar to CML and ALL^11-13 (Figure 2A). Interestingly, DPP4 inhibitors significantly decrease CFU ability of primary mouse AML cell in a dose-dependent manner (Figures 2B and 2C) while not affecting that of normal BM cells (Figures 2D and 2E), in part due to the differences in inhibition of DPP4 enzyme activity (Figures 2C, 2E, and S2B). Next, we compared the efficacy of linagliptin with cytarabine (Ara-C), the mainstay induction therapy for most patients with AML, by treating mouse AML cells at various concentrations^18,19 and observed weak inhibition of cell growth at low concentrations of linagliptin (0.1–1 ng/mL) but similar treatment efficacy to Ara-C at higher concentrations (100–1,000 ng/mL) (Figure 2F).^16,17 In support of in vitro efficacy, 20 days of continuous treatment with linagliptin (1 mg/kg/day) led to significantly increased survival of AML cell-transplanted mice due to slower AML development, similar to Ara-C treatment (100 mg/kg/day over 5 days) (Figures 2G-2I). Furthermore, there were no common adverse drug effects (ADEs), such as hypoglycemia (Figure 2J) and BM suppression/myelosuppression indicated by SC activity of YFP⁻ BM cells (non-AML cells) (Figure 2K).

Figure 2. DPP4 inhibitors suppress AML development, not the activity of HSPCs.

(A) Dot plots showing DPP4 expression on DPP4^+/+YFP⁺Mac-1⁺Kit⁺AML cells (4 weeks after transplantation) as determined by flow cytometry.

(B and C) Comparative efficacies of two different DPP4 inhibitors at indicated concentrations in suppression of mouse MLL-AF9 cell CFU numbers (B, 2,000 cells/well, n = 3 wells) and DPP4 enzyme activity of 1 × 10⁶ cells (C, n = 3 repeats). L, linagliptin; V, vildagliptin.

(D) The effect of linagliptin treatment (200 nM) on normal mouse BM cell CFU-forming ability (10,000 cells/well n = 3 wells).

(E) DPP4 enzyme activity inhibition by linagliptin treatment (200 nM) on 1 × 10⁶ mouse BM HSPCs (n = 3 wells).

(F) Comparison of the efficacy of linagliptin (Lgp; at 0.21, 2.1, 21, 210, and 2,100 nM, respectively) and Ara-C (at 0.41, 4.1, 41, 410, and 4,100 nM, respectively) chemotherapy in vitro, corresponding to 0.1, 1.0, 10, 100, and 1,000 ng/mL of each (n = 3, AML cells were obtained from 3 MLL-AF9 AML cell-transplanted mice).

(G) Schematic representation of in vivo treatment protocol. The regimen consisted of daily intraperitoneal injections of cytarabine (Ara-C, 100 mg/kg/day over 5 days) or vehicle (V)/Lgp (1 mg/kg/day over 20 days). D, day.

(H) The percentage of YFP⁺ AML cells in PB of V−, Lgp−, or Ara-C-treated mice at 30 days post-transplantation (n = 5 mice).

(I) Survival curve of MLL-AF9 AML mice treated with V (n = 20 mice), Lgp (n = 12 mice), or Ara-C (n = 12 mice). p values were derived from a log rank test.

(J) Fasting glucose levels in V−, Lgp−, or Ara-C-treated mice at 30 days post-transplantation (n = 5 mice).

(K) Comparison of CFU-forming ability of non-AML cells (YFP⁻ BM cells, 10,000 cells per well, n = 3 wells) obtained from V−, Lg−, or Ara-C-treated mice at 30 days post-transplantation.

Error bars, SEM. *p < 0.05, ***p < 0.001. Data were analyzed using two-tailed Student’s t test; when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses.

hAML growth is dependent on increased expression and activity of DPP4

Query of 3 human databases revealed that high expression of DPP4 was associated with reduced survival of patients with AML (Figures S3A-S3C). To examine whether high expression of DPP4 is essential for hAML development, we measured cell surface expression and enzyme activity of DPP4 and observed significantly higher levels in primary AML cells from human patients compared with peripheral blood (PB) myeloid cells from healthy controls (Figures 3A-3C and S3D). Furthermore, CD34⁺ hAML blast cells express DPP4 at a significantly higher level than CD34⁺ healthy controls or CD34⁻ non-blast AML cells, indicating that DPP4 is specifically expressed higher in LSC-enriched AML cells (Figures 3D and 3E). Intriguingly, linagliptin treatment of high DPP4-expressing AML cells from 3 human patients significantly inhibited LSC activity and increased AML cell apoptosis (Figures 3F and 3G). To test whether DPP4 inhibitor is effective for AML treatment in vivo, we transplanted primary hAML cells to xenograft NOD/SCID-IL2RG (NSG) mice and treated the mice with linagliptin (Figure 3H). We observed that linagliptin treatment significantly reduced AML cell infiltration in PB after 30 days treatment (35.1% vs. 2.4%) and increased overall survival of the mice when compared with controls (vehicle treatment) (Figures 3I-3K). Together, these data support the premise that DPP4 inhibition significantly suppresses hAML development. Next, we investigated whether suppression of DPP4 expression of hAML cells alters in vitro LSC activity and in vivo AML cell growth in transplanted NSG mice (Figure 3L). We observed greatly reduced CFU ability and increased apoptosis in DPP4-deficient hAML samples (Figures 3M and 3N). Transplantation of DPP4-deficient hAML cells into NSG mice suppressed leukemia development, as demonstrated by significantly prolonged overall survival and reduced engraftment of GFP⁺ cells (marking the lentivirus-infected hAML cells) into BM, spleen, liver, and PB (Figures 3O-3R). These results demonstrate that DPP4 is essential for hAML development, which further supports the notion that high expression of DPP4 negatively correlates with the overall survival of patients with AML (Figures S3A-S3C).

Figure 3. Human AML growth is dependent on increased expression and activity of DPP4.

(A) Histograms depicting DPP4 expression in the PB of human healthy controls and patients with AML: red indicates a positive staining reaction, blue indicates a negative staining reaction, and the isotype-matched control antibody is shown in gray.

(B and C) Proportion of DPP4⁺ myeloid cells (B) or DPP4 enzyme activity (C) in PB of patients with AML (n = 17) and healthy controls (n = 8).

(D) Representative flow cytometry plots to measure cell surface expression of DPP4; the numbers indicate the MFI value of gated cells. mlin⁻ indicates mobilized PB lineage⁻ cells.

(E) Summary of DPP4 expression in CD34⁺ and CD34⁻ populations of human healthy mobilized PB lineage⁻ cells (n = 5 samples) and patient with AML PB cells (n = 8 samples).

(F) Lgp suppressed DPP4⁺ AML cell CFU-forming ability in serial plating of AML patients #6, #13, and #14 (10,000 cells per well, n = 3 wells). In addition, 10,000 live cells from the first CFUs were used to perform the second CFU assay to test the stemness of AML cells.

(G) Lgp treatment increased the proportion of apoptotic AML cells after primary CFU assay (n = 3 wells), Annexin V was used (100,000 cells) to measure apoptosis on the cells from the first CFU assay.

(H) Schematic representation of the treatment protocol in hAML cell (sample #14)-transplanted NSG mice, which consisted of daily intraperitoneal injections of V or Lgp (1 mg/kg/day over 30 days). D, day.

(I) Survival curve of hAML cell-transplanted NSG mice treated with V or Lgp (n = 9 mice). p values were derived from a log rank test.

(J and K) Flow cytometry analysis (J) and histograms (K) of PB cells on day 65 showed Lgp treatment greatly reduced AML cell engraftment in NSG mice (n = 9 mice in each group). The number indicated the percentage of gated human CD45⁺ AML cells with 7AAD staining to exclude the dead cells.

(L) Schematic representation of the experiment design used to knock down dpp4 in hAML cells (sample #14) and then perform CFU and xenograft assays.

(M) Comparison of CFU-forming ability in serial plating after DPP4 knockdown in hAML cells (10,000 cells per well, n = 3 wells).

(N) DPP4 knockdown increased the proportion of apoptotic AML cells in primary CFU assay (n = 3 wells).

(O) Survival curve of hAML cell-transplanted NSG mice infected with scramble or shdpp4-2 virus (n = 6 mice). p values were derived from a log rank test.

(P) Summary of the proportions of GFP⁺ cells in BM, spleen, liver, and PB at day 150 after transplantation or when they were sacrificed (n = 6 mice).

(Q and R) Comparison of the weight of spleens (Q) and livers (R) of the mice transplanted with hAML cells expressing shRNA-targeting dpp4 or scramble control (n = 6 mice).

Error bars, SEM. *p < 0.05, **p < 0.01, ***p < 0.001. Data were analyzed using two-tailed Student’s t test; when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses.

Hematopoietic-cell-specific deletion of DPP4 does not affect normal hematopoiesis

We generated a hematopoietic-system-specific deletion of Dpp4 in mice to study the role of DPP4 in both normal hematopoiesis and AML development (Figure S4A).²⁰ The Dpp4 ^fl/fl: Vav-Cre⁺ (Tm1d or DPP4^−/−) mice displayed an apparently normal phenotype including a normal lifespan when compared with the control Dpp4 ^fl/fl: Vav-Cre⁻ (DPP4^+/+) mice. The complete blood count (CBC) of DPP4^−/− mice is comparable to that of DPP4^+/+ mice (Figure S4B). Furthermore, DPP4^−/− mice have normal frequency and numbers of HSPC (total of longterm HSCs, short-term HSCs, and multiple potential progenitors) and downstream progenitor cells (including common myeloid progenitors [CMPs], granulocyte-macrophage progenitors [GMPs], and common lymphoid progenitors [CLPs]) in the BM and regular cell-cycle activity of HSCs (Figures S4C-S4F). Consistent with DPP4 inhibition in normal BM cells (Figure 2D), DPP4-deficient BM cells have equal CFU-forming ability as wild-type (WT) cells (Figure S4G). DPP4^−/− HSCs have higher engraftment ability with no biased lineage reconstitution (Figures S4H and S4I), which is consistent with findings in the DPP4 global knockout mouse model.^21,22

DPP4 supports AML development in two mouse AML models

Utilizing the DPP4^−/− and DPP4^+/+ mice generated above with normal hematopoiesis, we retrovirally transduced BM HSPCs derived from these mice with two oncogenes, namely MLL-AF9^16,17 and AML1-ETO9a²³ separately, to generate two AML models to study the role of DPP4 in leukemia development. The efficiency of retrovirus transduction in DPP4^+/+, DPP4^+/−, and DPP4^−/− HSPCs is comparable as shown in Figure 4A (23% vs. 21% vs. 22%). Intriguingly, MLL-AF9 transformation can greatly raise the level of DPP4 in DPP4^+/+ cells (Figure 4A). The recipient mice transplanted with MLL-AF9-transduced DPP4^−/− AML cells (henceforth called DPP4^−/− AML mice) develop leukemia much more slowly than mice transplanted with DPP4^+/− (heterozygous) and DPP4^+/+ AML cells (on average, DPP4^−/− AML mice survive >300 days vs. DPP4^+/− AML mice survive 69.4 days vs. DPP4^+/+ AML mice survive only 40.7 days) in primary transplantation (Figure 4B). Dose dependence of AML development on DPP4 is also evident from the counts of AML cells in PB and DPP4 enzyme activity at 4 weeks post-transplantation (Figures 4C and 4D). Serial tracking of the proportion of YFP⁺ AML cells in the PB of DPP4^−/− AML mice revealed the significantly slow time course of AML development compared with the DPP4^+/+ AML mice at 3 weeks post-transplantation (2.1% vs. 7.6%) that reached relative maximum numbers at 6 weeks post-transplantation (29.4% vs. 88.3%), which then gradually decreased and remained at very low levels (1%–6%, in 8 to 24 week period) (Figure 4E). No comparison was possible after 6 weeks since DPP4^+/+ AML mice were moribund and euthanized. Similar to PB, the counts and infiltration of DPP4^−/− AML cells are substantially lower in the various hematopoietic organs compared with those of DPP4^+/+ AML cells at 6 weeks post-transplantation (Figure 4F), which is likely responsible for the significantly smaller size of spleen and liver (Figures 4G and 4H) and less infiltration of AML cells in the hematopoietic organs in recipient mice (Figure 4I). Similar results for survival and engraftment were obtained in another widely used mouse AML model utilizing AML1-ETO9a-transformed AML cells (Figures 4J, 4K, and S5).^5,23 Together, our data indicate that DPP4 plays an essential role in the development of AML.

Figure 4. DPP4 supports AML development in two mouse AML models.

(A) Contour plots showing the infection efficiency (numbers in the parenthesis indicate the proportions of YFP⁺ cells) of DPP4^+/+, DPP4^+/−, or DPP4^−/− HSPCs transduced with MLL-AF9-IRES-YFP retrovirus 40 h after infection. DPP4 MFIs of YFP⁺ and YFP⁻ populations have been indicated.

(B) Survival curve of mice receiving MLL-AF9-infected DPP4^+/+, DPP4^+/−, or DPP4^−/− HSPCs (n = 10 mice).

(C) The proportions of YFP⁺ AML cells in PB of mice transplanted with DPP4^+/+, DPP4^+/−, or DPP4^−/− AML cells at 4 weeks post-transplantation (n = 10 mice).

(D) The DPP4 enzyme activity of 1 × 10⁶ AML cells were measured by fluorescence assay (n = 3 mice).

(E) Summary of the proportions of YFP⁺ DPP4^+/+ and DPP4^−/− MLL-AF9 leukemia cells in PB of primary transplanted mice over time (n = 10 mice). Red dots indicate the moribund stage of the mice, which were sacrificed at that time point.

(F–H) Comparison of the proportions of YFP⁺ AML cells in BM, PB, spleen, and liver (F), the spleen weight (G), and liver weight (H) of mice transplanted with DPP4^+/+ MLL-AF9 cells and those with DPP4^−/− MLL-AF9 cells at 6 weeks post-transplantation (n = 3 mice).

(I) Histological analysis of AML infiltration in the spleen, liver, and PB of mice at 6 weeks posttransplantation (hematoxylin and eosin staining for spleen and liver, HEMA 3 staining for PB samples). Shown are representative images from at least three similar images. Scale bar is 100 μM.

(J) Survival curves of mice receiving AML1-ETO9a-infected DPP4^+/+ or DPP4^−/− HSPCs (n = 12 mice; p < 0.0001, log rank test).

(K) Proportions of AML1-ETO9a GFP⁺ leukemia cells in BM at 20 weeks after transplantation (n = 6 mice).

Error bars, SEM. *p < 0.05, **p < 0.01, ***p < 0.001. Data were analyzed using two-tailed Student’s t test; when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses, except that the p values of (B) and (J) were derived from log rank tests.

DPP4 is required for the stemness/activity and survival of AML cells

To further understand the mechanisms underlying dependence of AML development on DPP4, we first measured LSC markers in the BM of the MLL-AF9 mouse model. Surprisingly, at 6 weeks post-transplantation, there were considerably more L-GMP cells in the BM of DPP4^−/− AML mice compared with DPP4^+/+ AML mice (2.51-fold increase; Figures 5A and 5B),¹⁷ despite the fact that this difference is not immediately apparent at the transformation stage (Figures S6A and S6B). Furthermore, the LSC-harboring population (YFP⁺Mac1⁺Kit⁺) also increased in the BM of DPP4^−/− AML mice compared with DPP4^+/+ AML mice (2.1-fold; Figures 5C, S6C, and S6D).^5,16 Next, to understand why DPP4 deletion enhanced the number of phenotypic LSCs while preventing AML development, we relied on functional measures to evaluate the activity of LSCs using serial CFU replating and secondary transplantation assays. We observed dramatic reduction in self-renewal of AML-SCs as indicated by stepwise reduction in the activity of DPP4^−/− AML-SCs in CFU-forming ability of ~50%, ~60%, and ~65% upon serial plating (Figure 5D). Moreover, secondary AML transplantation resulted in much more delayed AML development in DPP4^−/− AML mice in vivo (Figures 5E-5G vs. Figure 4C). Thus, AML-SC activity in DPP4^−/− AML mice was greatly lowered compared with DPP4^+/+ AML mice. To quantify how DPP4 deficiency affects the AML-SC frequency, we performed transplantations with limiting dilutions of sorted YFP⁺ DPP4^+/+ and DPP4^−/− MLL-AF9 BM cells that were collected from primary recipients. Figure 5H depicts the progression of leukemia as evaluated by the survival ratio and latency days. Strikingly, the frequency of functional AML-SCs in the DPP4^−/− primary MLL-AF9 AML model was only 1/675 (= 161/108,740) of that in DPP4^+/+ primary AML mice. Together, these results indicate that DPP4 is required for the stemness of AML cells.

Figure 5. DPP4 deletion leads to an AML remission-like state via increase in apoptosis of AML cells and reduction in the activity of AML-SCs.

(A) Immunophenotypic analysis of myeloid progenitors from the BM of DPP4^+/+ and DPP4^−/− MLL-AF9 AML mice at 6 weeks post-transplantation to identify the L-GMP population.

(B) Summary of absolute numbers of L-GMP in the BM of DPP4^+/+ and DPP4^−/− MLL-AF9 AML mice at 6 weeks post-transplantation (n = 5 mice).

(C) Summary of absolute numbers of YFP⁺ Mac1⁺ Kit⁺ AML cells in the BM of DPP4^+/+ and DPP4^−/− MLL-AF9 AML mice at 6 weeks post-transplantation (n = 5 mice).

(D) Comparison of CFU capability of DPP4^+/+ and DPP4^−/− MLL-AF9 AML cells during serial re-plating (2,000 cells/well, n = 3 wells).

(E) Survival curve of secondary transplanted mice receiving 2,000 DPP4^+/+ or DPP4^−/− AML cells (n = 10 mice; p < 0.001, log rank test).

(F) The proportion of YFP⁺ AML cells in PB of mice secondary transplanted with DPP4^+/+ or DPP4^−/− AML cells at 3 weeks post-transplantation (n = 10 mice).

(G) Comparison of the spleen weight of mice secondary transplanted with DPP4^+/+ MLL-AF9 cells and those with DPP4^−/− AML cells at 3 weeks post-transplantation (n = 3 mice).

(H and I) Limiting dilution assays comparing the frequencies of AML stem cells in DPP4^+/+ and DPP4^−/− MLL-AF9⁺ AML. The indicated YFP⁺ DPP4^+/+ and DPP4^−/− MLL-AF9⁺ BM cells that were collected from primary recipients were co-transplanted with 2 × 105 BM competitor cells into lethally irradiated recipients. Mice were followed for 6 months, and at the end of the follow-up period, flow cytometry analysis confirmed the absence of YFP⁺ donor leukemia cells in BM and subclinical disease in surviving mice. The competitive repopulating units (CRUs) were calculated by L-Calc software (v.1.1; StemCell Technologies).

(J and K) Representative image and plots of flow cytometry analysis of apoptosis in mouse YFP⁺ AML cells from PB at 6 weeks post-transplantation; apoptosis was detected as Annexin V⁺/PE⁺/7-AAD⁻ staining (n = 5 mice).

(L) Comparison of the proportions of YFP⁺ AML cells in BM, PB, spleen, and liver of mice transplanted with DPP4^−/− AML cells at 13, 24, and 50 weeks posttransplantation (n = 5 mice).

(M) Flow cytometry analysis of apoptosis in mouse MLL-AF9 YFP⁺ (AML) cells and YFP⁻ (normal) cells of the BM of DPP4^−/− MLL-AF9 at the indicated time points. Apoptosis was detected as Annexin V⁺ 7-AAD⁻ staining (n = 5 mice).

(N) Histological analysis of AML infiltration in the BM of mice transplanted with DPP4^+/+ AML cells (4 weeks post-transplantation) or DPP4^−/− AML cells (50 weeks post-transplantation). Hematoxylin and eosin (H&E) stain, original magnification, ×600. Scale bar is 50 μM. The dashed lines outline a sampling of immature blasts (blue arrow), and trilineage cells, erythroid (yellow arow), megakaryocyte (green arrow), and mature myeloid (red arrow).

Error bars, SEM. *p < 0.05, ***p < 0.001. Data were analyzed using two-tailed Student’s t test; when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses, except that the p value of (E) was derived from log rank tests.

In addition to the role of DPP4 in supporting LSC activity, we examined the mechanisms for substantial reduction of leukemia cells in the PB of DPP4^−/− AML mice at 6 weeks post-transplantation (Figure 4E) and found that DPP4 deficiency significantly increased apoptosis in the PB of DPP4^−/− AML mice, (Figures 5J and 5K), which is consistent with our observation in hAML cells treated with DPP4 inhibitors and DPP4 targeting shRNA (Figures 3G and 3N). Serial tracking of AML cell engraftment in DPP4^−/− AML mice revealed a dramatic reduction in DPP4^−/− AML cells in all hematopoietic organs from 13 through 50 weeks post-transplantation (Figures 5L, S6E, and S6F). Specifically, BM tracking showed that DPP4^−/− AML cells were greatly reduced due to apoptosis from 49.3% (apoptotic ratio: 17.3%) at 13 weeks to 12.3% (apoptotic ratio: 3.1%) at 24 weeks and 2.2% (apoptotic ratio: 2.6%) at 50 weeks post-transplantation (Figures 5L and 5M), which is likely responsible for the substantial reduction of leukemia cells in DPP4^−/− AML mice during this AML-suppression stage. Consistent with these findings, normal trilineage hematopoiesis was present in the BM of DPP4^−/− AML mice at 50 weeks post-transplantation, while DPP4^+/+ AML mice at 4 weeks post-transplantation had the marrow largely replaced by immature blasts with large irregular nuclei (Figures 5N and S6G). These data support the notion that DPP4 deletion may suppress/reverse AML development, resulting in AML remission.

NF-κB activation mediates DPP4-initiated intracellular effects on AML-SCs

To determine the mechanisms by which DPP4 regulates the activity and survival of AML cells, we analyzed the genome-wide transcriptome change in DPP4^+/+ vs. DPP4^−/− MLL-AF9 AML cells. We found that DPP4^+/+ AML cells had much greater expression of genes associated with AML-SC self-renewal and maintenance (Figures 6A and 6B). Quantitative PCR analysis demonstrated significant downregulation of several self-renewal associated genes, such as Hox genes, in MLL-AF9 AML cells, which corroborated with the deep sequencing results (Figure 6C). Reintroduction of DPP4 into DPP4^−/− AML cells enhanced the level of the self-renewal associated genes (Figure S7A). Importantly, expression of the same self-renewal-associated genes was not different between DPP4^+/+ vs. DPP4^−/− normal HSPCs (Figure S7B). Gene set enrichment analysis (GSEA) on RNA sequencing (RNA-seq) data revealed that nuclear factor κB (NF-κB) is the most suppressed transcriptional signature in DPP4^−/− AML cells (normalized enrichment score [NES] = 2.2) compared with DPP4^+/+ AML cells (Figure 6D). To assess the signature of NF-κB inactivation in DPP4^−/− AML cells, we determined the activation of the signaling intermediates of NF-κB, IκB, and p65 and observed that the phosphorylation of IkB and p65 is dramatically lower in DPP4^−/− AML cells compared with DPP4^+/+ AML cells. Interestingly, DPP4^−/− AML cells had higher caspase 3 activity (cleaved caspase 3), which is consistent with our data in hAML, demonstrating that either DPP4 inhibition or deletion can lead to apoptosis in AML cells. In further support of the dependence of DPP4 signaling on NF-κB, other signaling pathways such as pP38MAPK or AKT were not affected (Figure 6E). Importantly, we validated these findings in THP1 AML cells and hAML samples either with shRNA knockdown or linagliptin treatment and observed downregulation of NF-κB intermediates (Figures 6F-6H).

Figure 6. NF-κB activation mediates DPP4-initiated intracellular effects on AML-SCs.

(A) Heatmap of 2,172 differentially expressed genes from RNA-seq (n = 2 DPP4^+/+ or DPP4^−/− BM AML cells, red is higher and blue is lower expression).

(B) Gene set enrichment analysis (GSEA) plot evaluating changes in leukemia initiation/maintenance gene signatures upon DPP4 signaling depletion in DPP4^+/+ and DPP4^−/− MLL-AF9 AML cells. NES, normalized enrichment score; NOM p, normal p value; FDR q, false discovery rate q value.

(C) Bar graphs compare the change in expression levels of the indicated transcripts in DPP4^+/+ and DPP4^−/− MLL-AF9 AML cells. All transcripts levels were normalized to levels of actin expression (n = 3 wells).

(D) GSEA plot shows that DPP4 deletion negatively coordinates NF-κB signaling in AML cells.

(E) Comparison of p-IκB, p-p65, p-AKT, p-P38, and cleaved caspase3 protein levels in DPP4^−/− (k, KO) and DPP4^+/+ (w, WT) MLL-AF9⁺ BM cells from each of 3 primary transplanted mice.

(F–H) Immunoblotting of p-IκB and p-p65 in scramble (control [Ctrl])- and DPP4 shRNA (KD)-treated THP1 cells (F) and hAML cells (G, samples #14, #13, and #6) and in DMSO- or Lgp (200 nM)-treated hAML cells (H, sample #14).

Error bars, SEM. *p < 0.05, Data were analyzed using two-tailed Student’s t test; when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses.

Src is essential to the regulation by DPP4-NF-κB axis of AML development

To understand the molecular mechanism by which DPP4 regulates NF-κB activation, we searched for putative binding partners of DPP4. Among them, the cellular tyrosine kinase Src (c-Src) is an intracellular protein that is a known regulator of AML development.^24,25 Indeed, immunoprecipitation of DPP4 confirmed DPP4-Src binding in vitro (Figure 7A). Deficiency of DPP4 in mouse and human AML cells and hAML samples had a marked reduction in Src activation (Figures 7B-7D). In contrast, the activation of Src, IκB, and p65 in normal BM cells of both DPP4^+/+ and DPP4^−/− mice was low but equal (Figure 7E), suggesting that deletion of DPP4 does not affect Src-NF-κB signaling in normal hematopoiesis. To further test whether DPP4-dependent stemness of AML cells is mediated by Src, we performed CFU assays with DPP4^+/+ and DPP4^−/− AML cells in the presence of Src family kinase (SFK)-specific inhibitor saracatinib (5 μM). We observed that saracatinib significantly reduces the CFU ability of DPP4^+/+ AML cells but not DPP4^−/− AML cells (Figure 7F). To further determine whether Src is essential for DPP4-mediated effects in AML cells, we introduced retroviruses encoding src into DPP4^−/− AML cells. The ectopic expression of Src was capable of reversing the inhibitory phenotype of DPP4^−/− AML cells to the same level as DPP4^+/+ AML cells in transplantation and CFU assays (Figures 7F-7I). Moreover, the rescue assay samples showed that ectopically expressed Src could activate NF-κB signaling and inhibit caspase signaling, as well as improve the transcriptional level of LSC self-renewal associated genes in DPP4^−/− AML cells (Figures 7J and 7K). These results demonstrate that DPP4-mediated promotion of AML cell stemness, growth, and NF-κB signaling activation are Src dependent.

Figure 7. Src is essential to the regulation by DPP4-NF-κB axis of AML development.

(A) DPP4 binds to Src in transfected 293T cells. FLAG-tagged DPP4 and/or hemagglutinin (HA)-tagged Src was overexpressed in 293T cells. FLAG antibody was used to precipitate DPP4 protein, and the FLAG or HA antibodies were used in WB.

(B–D) Comparison of p-Src levels in DPP4^−/− (k) and DPP4^+/+ (w) MLL-AF9⁺ BM cells from each of 3 primary transplanted mice (B) and scramble- (Ctrl) and DPP4 shRNA (KD)-treated THP1 cells (C) or hAML cells (samples #14, #13, and #6) (D) determined by WB.

(E) p-Src, p-IκB, and p-p65 in non-AML DPP4^+/+ and DPP4^−/− BM cells were determined by WB.

(F) Comparison of colony-forming activity of DPP4^+/+ and DPP4^−/− MLL-AF9 AML cells upon saracatinib (5 μM) treatment (2,000 cells/well, n = 3 wells).

(G) Retrovirally expressed src reversed the DPP4 knockout phenotype in transplantation assay. MLL-AF9⁺, DPP4^+/+, or DPP4^−/− BM cells in primarily transplanted mice were collected at 40 days and were infected with Src-encoding or control virus. Survival curves of mice transplanted with 3,000 of these ectopically src-expressing or control cells (n = 10 mice; p < 0.0001, log rank test).

(H) Proportions of retrovirus infected (GFP⁺) AML cells in PB of secondary recipient mice after 28 days of transplantation (n = 10 mice).

(I) Retrovirally expressed src increased CFU numbers of DPP4^−/− AML cells compared with cells that were infected with control virus in colonyforming assays (2,000 cells/well, n = 3 wells).

(J) WB showing the activity of IκB and p65 in DPP4^+/+ and DPP4^−/− AML cells upon Src overexpression.

(K) Bar graphs compare the change in expression levels of the indicated transcripts in DPP4^−/− MLL-AF9 AML cells ectopically expressing either control or src gene (n = 3 wells).

Error bars, SEM. *p < 0.05. Data were analyzed using two-tailed Student’s t test; when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses, except that the p value of (G) was derived from log rank test.

DISCUSSION

Currently, the mainstay of AML treatment involves induction chemotherapy followed by consolidation. For AML with high-risk characteristics such as FLT3-ITD mutations or a monosomal karyotype,²⁶ alloSCT is often the only curative approach. Chemotherapy and alloSCT are highly toxic due to indiscriminate inhibition of a plethora of cellular processes, the so-called off-target effects, which manifest as severe myelosuppression; gastrointestinal symptoms such as vomiting and diarrhea due to effects on dividing cells; infections; bleeding; and graft versus host disease.^1,27 This study identified cell surface protein DPP4 inhibition as a potential treatment for AML. DPP4 inhibitors have been successfully used in the treatment of T2DM for ~15 years and are well tolerated with minimal ADEs such as upper respiratory tract infections, headache, and bloating. Importantly, DPP4 inhibition does not cause severe ADEs such as hypoglycemia, gastrointestinal (GI) side effects (due to inhibition of cell division), and BM suppression.^28-33 This study further validated that DPP4 inhibitors do not affect normal hematopoeisis, which is probably one of the most vexing problems associated with chemotherapy and other novel therapies.

The specificity of DPP4 as a target in AML may be conferred due to the high expression and/or activity of DPP4 and its associated proteins in leukemic cells compared with normal hematopoietic cells. Our studies in mouse AML models showed that DPP4 expression is greatly increased during the HSPC transforming stage (Figures 4A, S2A, and S5), which is consistent with the data that DPP4 expression is significantly higher in CD34⁺ hAML blast cells than CD34⁻ non-blast cells. It was noted that DPP4 expression was significantly higher in a subset of patient-derived hAML samples, implying that DPP4 expression may be associated with specific AML genetic subtypes. FLT3-ITD mutations constitute the largest subgroup of patients with AML (25%) and are associated with a poor prognosis.³⁴ Significantly, DPP4 is highly expressed in LSCs of patients with AML with FLT3-ITD mutations.^11,13,35 Although FLT3-ITD inhibitors (midostaurin) are approved, they may cause severe ADEs such as fever, respiratory tract infections, and bleeding, all related to BM suppression. Therefore, future studies are needed on the role of DPP4 in the diagnosis and therapy of patients with FLT3-ITD mutant AML.

For decades, the notion of removing the block in LSC differentiation has been explored both to avoid toxicity of chemotherapy to the BM and to guide differentiated cells toward apoptosis in the PB.^36-41 Apart from the success of all-trans retinoic acid (ATRA) in acute promyelocytic leukemia (APL), differentiation therapy has not lived up to its promise in other AML or leukemias, either due to variable responses or due to the development of differentiation syndrome.^40,41 It was intriguing to explore the notion that DPP4 inhibitors may be involved in differentiating leukemia cells since DPP4 has been shown to be involved in cell-cycle arrest and senescence.^42,43 However, when we looked at the cell-cycle stages in our shRNA-mediated knockdown experiments, we did not see evidence of cell-cycle arrest. Moreover, we observed no changes in differentiated populations in the BM of DPP4^−/− AML mice compared with DPP4^+/+ AML mice, suggesting that DPP4 inhibition does not contribute to AML cell differentiation. Apart from differentiation therapy, other strategies to improve AML treatment include overcoming resistance to chemotherapy by addressing new mutations and/or tumor heterogeneity. Our studies using primary hAML samples further supported the dependence on DPP4 for AML development. This suggests DPP4 deficiency may overcome resistance to therapy in AML. However, true resistance attributable to tumor heterogeneity will require further studies in long-term AML development models.

The role of DPP4 in cancer SCs is not well understood. For example, DPP4 has been identified as a SC marker for both CML and ovarian cancer. While DPP4 is considered as an oncogene in CML, it has been shown to act as a tumor suppressor in ovarian cancer. The differences in downstream mechanisms may be attributed to variations in DPP4 expression levels and the substrates of DPP4 associated with the specific cancer cells. In addition, DPP4 gene expression is regulated via methylation at CpG islands in the 5′ region, which may be altered differently in individual cancers, suggesting that epigenetic regulation may affect the gene function. Furthermore, DPP4 protein expression could be subject to multiple post-translational modifications including glycosylation and acetylation, even sumoylation. In our study, although we observed a paradoxical increase in the phenotypic LSC or LSC-harboring population in DPP4^−/− AML cells, our data support an increasing dependence of mouse AML-SCs on DPP4 through functional experiments, which manifests as a reduction in engraftment ability after second transplantation and continuous reduction of colony numbers in serial CFU assays. Moreover, the differences in leukemic potential were confirmed by limiting dilution transplantation studies, which revealed that the functional AML-SCs were reduced by 675-fold in the MLL-AF9 model after DPP4 deletion. Taken together, these data suggest that DPP4 confers stemness to the AML cell.

Another interesting finding from our study is that DPP4 inhibition/deletion slowed down growth of AML cells but not normal BM cells. While cell-cycle progression was unchanged in both normal BM and AML cells, apoptosis was prominently affected in the AML cells, and this effect was distinct from Ara-C, which affected both normal BM and AML cells. The increase in apoptosis was observed in both mouse and hAML models and under both DPP4 inhibition and deficient conditions. Increased apoptosis of AML cells and AML-SCs likely resulted in decreased AML cell growth and suppressed AML development that ultimately led to an AML remission-like phenotype on our periodic follow-up studies. DPP4 is a transmembrane protein with a short 6 amino acid sequence with a Src binding motif. We confirmed binding of DPP4 to Src and observed a reduction in phosphorylation and activation of Src-NF-κB signaling with DPP4 deletion in both mouse and hAML, which resulted in improved survival of DPP4^−/− AML cells transplanted mice. The dependence on Src for AML growth and survival is consistent with recent findings that the inhibitors of SFK suppress AML cell growth and AML development.^44,45 Transcriptome analysis and cell signaling studies identified that DPP4-Src-NF-κB signaling works together to promote the survival and activity of AML cells. Inactivation of NF-κB signaling downstream of Src inactivation is consistent both with increased apoptosis in DPP4^−/− AML cells and with reports that NF-κB signaling pathway is involved in AML development.^46-48

In our study, there was a survival difference in AML mice that received DPP4 inhibitor (linagliptin) vs. those that received DPP4-deleted mouse and primary hAML cells. The difference can be attributed to the different levels of DPP4 enzyme inhibition from these two approaches or the contribution of non-enzymatic activity of DPP4 to AML suppression through signaling. Moreover, as DPP4 is expressed on many cell types including AML cells, each with exposure to different substrates, the effect of DPP4 inhibitors is more complex as it involves the net effect of multiple cells and their functions based on different DPP4 substrates.

In summary, the higher rates of apoptosis and reduced stemness of DPP4-deficient and DPP4-inhibited AML cells led to less AML development. Additionally, both of these factors led to reduced tumor burden in the PB and other organs, which might translate to amelioration of leukostasis, clotting problems, shortness of breath, and stroke in patients with leukemia.^49,50 Therefore, DPP4 inhibition may be a viable strategy for treatment of AML, especially in those patients who are unwilling to take or cannot tolerate conventional therapy.^51-53

Limitations of the study

Our study has identified DPP4 inhibitors as a novel treatment for AML, which specifically targets AML-SCs and not normal HSCs. Although hematopoietic-cell-specific deletion of DPP4 does not affect normal hematopoiesis, and linagliptin does not affect the CFU ability of HSPCs, we cannot completely exclude the impact of linagliptin on LT-HSC function. Price et al. observed no effects on the normal differentiation into the different leukocyte subsets with sitagliptin treatment for 28 days, which further supports the notion that DPP4 inhibition has minimal effects on hematopoiesis of healthy individuals.⁵⁴ We and others³⁵ have observed that DPP4 expression is high in CD34⁺ hAML, specifically the FLT3-ITD subset of patients with AML. However, we have also observed differential expression of DPP4 on CD34⁺ AML-SCs vs. CD34⁻ AML cells, i.e., some hAML samples that are not CD34⁺ have moderate DPP4 expression; whether DPP4 plays role in CD34⁻ AML cells is unknown. These observations align with the MLL-AF9 oncogene-transformed AML cells, which, clinically, are mostly CD34⁻ AML cells, having significantly higher DPP4 expression, partially indicating that DPP4 plays role in CD34⁻ AML cells as well.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xunlei Kang(Kangxu@health.missouri.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

The TCGA datasets analyzed are available in the UCSC Xena Browser (https://xena.ucsc.edu). The RNA-seq datasets generated in the current study have been deposited in NCBI “SRA database: SRP323430.”

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

C57 BL/6, CD45.2, CD45.1 mice were purchased from Charles River, Inc. DPP4 flox/flox mice were obtained by breeding targeted C57Bl/6NTac-DPP4tm1a Wtsi/Ics mice (European Mouse Mutant Cell Repository, EUCOMM) with 129S4/Bl6-Gt(ROSA) 26Sortm2(FLP*)Sor/J (stock #012930, The Jackson Laboratory, Bar Harbor, ME). The offspring were then crossed with Vav-iCre mice (stock #018968, The Jackson Laboratory, Bar Harbor, ME) (de Boer et al., 2003),⁵⁶ to obtain DPP4 ^fl/fl;Vav-Cre mice. All mice were cared for in accordance with National Institutes of Health guidelines. No statistical method was used to predetermine the sample size. Mice were randomly assigned to the experiments after genotyping with the same sex, similar age group (6–8 weeks old) and approximately same weight. The investigators were not blinded to the allocation of animals during the experiments and outcome assessment. All procedures were approved in advance by the Institutional Animal Care and Use Committee of the University of Missouri.

Human samples

Primary human AML samples were obtained from the tissue banks at University of Missouri, School of Medicine. Informed consent was obtained under protocols reviewed and approved by the Institutional Review Board at University of Missouri (IRB #2018932) to obtain fresh human peripheral blood samples and primary human AML samples at the Ellis Fischel Cancer Center. Diagnosis of AML was confirmed by a pathologist utilizing BM aspiration and biopsy and Cytochemical and Immunohistochemical tests. Cyotgenetics analysis and Next Generation Sequence were performed on the bone marrow by Mayo Clinic Laboratories-Rochester Main Campus. The blood samples were processed using SepMate^™-50 (IVD) (STEMCELL Technology GmbH, Germany) according to the manufacturer’s instructions and frozen in FBS with 10% DMSO and stored in liquid nitrogen. Information of patient samples including the age and gender have been listed in Figure S3D.

METHOD DETAILS

Cell culture

293T cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% FBS at 37°C in 5% CO2. Primary mouse AML cells were cultured in serum-free StemSpan^™ (STEMCELL Technologies, Cambridge MA) supplemented with 50 ng/mL SCF, 10 ng/mL IL-3, 10 ng/mL IL-6 (STEMCELL Technologies, Cambridge MA). Human monocytic AML cells THP-1 (ATCC, TIB-202), macrophage B myelomonocytic leukemia MV4-11 (ATCC, CRL-9591), and histiocytic lymphoma U937 (ATCC, CRL-1593.2)) were cultured in Roswell Park Memorial Institute (RPMI) 1640 supplemented with 10% FBS at 37°C in 5% CO₂ and normal O₂. All cell lines were routinely tested using a mycoplasma contamination kit (R&D Systems, Minneapolis MN).

Compound screen for cell surface protein targets

We selected two compound libraries #HYL026 (Clinical Compound Library, MedChemExpress, Monmouth Junction, NJ) and #HYL022 (FDA-Approved Drug Library, MedChemExpress, Monmouth Junction, NJ), based on a broad panel of cell surface target proteins. #HYL022 and #HYL026 have been used to identify novel targets (Perez-Gomez et al., Nat Commun. 2018 Dec 10; 9(1):5272, Kunder et al. Cell Chem Biol. 2021 Sep 8; S2451-9456(21)00400-1). We did phenotype-based screening, and this approach allows several drugs to be tested without limiting oneself to any single mechanism of action, thereby resulting in first-in-class drugs. We identified 57 candidate compounds based on three pre-specified criteria 1) Low expression of targets of these compounds improved overall survival of AML patients as documented in The Cancer Genome Atlas (TCGA) (Supplementary 1a). 2) in silico analysis of gene expression in AML cells. 3) The availability of genetically modified mice for targets of interest. We also ensured that there was no bias in the form of a specific hypothesis in determining which compound to select. We performed the in vitro screenings at a concentration of 10mM, including for linagliptin, a DPP4 inhibitor.

Virus construction and infection

pLentiLoxp3.7 (Addgene, Watertown MA) was used to construct DPP4 shRNA expressing lentiviral vectors. The target sequences for shDPP4-1 and shDPP4-2 were 5′ AATGCCAGGAGGAAGGAATCTTT -3′ and 5′ AACAAGTTGAGTACCTCCTTATT -3′, respectively. For virus packaging, retroviral constructs MSCV-MLL-AF9-IRES- YFP/GFP, Mig-AML1-ETO9a or MSCV-SRC-IRES-GFP, were mixed with PCL-ECO (2:1), and the lentivirus constructs were mixed with pSPAX2 and pMD2.G (4:3:1), followed by transfection into 293T cells using PolyJet (SignaGen Laboratories, Frederick MD). Virus-containing supernatant was collected 48–72 h posttransfection and used for infection as described previously.⁵ Briefly, the virus supernatant was collected and filtered through a 0.45-mm filter. Lin⁻ cells were isolated from the bone marrow of littermate DPP4^−/− or DPP4^+/+ mice pretreated with 5-FU (150 mg/kg). For infection of Lin⁻ cells from bone marrow (BM), Lin⁻ cells were suspended in virus supernatant supplemented with 4 μg/mL polybrene at 10⁵/mL and placed in six-well plates for spin infection (2000 rpm, 90 min at 32°C).

AML transplantation

For primary transplantation, Infected Lin⁻ cells (2 × 10⁵) were transplanted into lethally irradiated (1,000 cGy) C57BL/6 mice (6–8 weeks old) by retro-orbital injection. For MLL-AF9 AML model, the engraftment was assessed at 4, 6, 13, 24 and 50 weeks post-transplantation, and for AML1-ETO9a AML model, the engraftment was assessed at 20 weeks post-transplantation. For secondary transplantation, we used FACS to isolate YFP⁺ or GFP⁺ BM cells from primary recipient mice and transplanted 2,000 cells (MLL-AF9 model) together with 2 × 10⁵ normal BM cells into lethally irradiated recipients and engraftment assessed at 3 weeks post-transplantation.

Human AML xenograft

Xenografts were performed essentially as we described previously.⁵ Briefly, adult NSG mice (6-8 weeks old) were sublethally irradiated with 250 cGy total body irradiation before transplantation. shRNA-infected cells were resuspended in 200mL PBS containing 2% FBS at a final concentration of 5 × 10⁵ GFP⁺ primary human AML viable cells per mouse for retro-orbital injection.

DPP4 activity assay

Plasma-EDTA, bone marrow extracellular fluid (BMEF), or cell lysates were used for DPP4 activity assay. 20 μL each of plasma-EDTA or BMEF or 100 μg protein was diluted in DPP4 assay buffer (Tris-HCl [pH 8.0], 150 mM NaCl, and protease inhibitor cocktail [Roche Diagnostics]) in a black 96-well plate to a final volume of 50 μL, and 50 μL of the substrate, 200 mM H-Ala-Pro-AFC (I-1680; Bachem Americas, Torrance CA), was added to each well and the plate incubated for 10 min at room temperature in the dark. Fluorescence was measured with a Synergy Microplate Reader at the excitation wavelength of 405 nm and an emission wavelength of 535 nm. Results are reported as relative light units (RLUs).

Leukemia characterization

We performed the procedure as previously described.⁵⁷ After transplantation, we monitored the survival, examined the size and histological properties of BM, spleen, and liver, and analyzed the numbers and infiltration of leukemia cells in PB, BM, spleen, and liver. Complete blood counts in PB were measured using a Hemavet 950FS. We also characterized different populations of leukemia cells using flow cytometry. Leukemia characterization was performed by investigators blinded to the experimental groups. Moribund leukemic mice were euthanized and the time was recorded as the time of death. Mice that died for reasons unrelated to leukemia within 10 d after transplantation were excluded from the data analysis.

Flow cytometry

For flow cytometry analyses of mouse PB, BM, Spleen, and Liver, hematopoietic cells were labeled with the following antibodies from Biolegend (San Diego CA) unless specifically indicated: Anti-CD3e-PE/Cyanine5 (mouse, #100310, 1:200 dilution), Anti-Ly-6G/Ly-6C (Gr-1)-PE/Cyanine5 (mouse, #108410, 1:200 dilution), Anti-CD11b-PE/Cyanine5 (mouse, #101210, 1:200 dilution), Anti-CD45R-PE/Cyanine5 (mouse, #103210, 1:200 dilution), Anti-Ter-119-PE/Cyanine5 (mouse, #116210, 1:200 dilution), Anti-CD117 (c-Kit)-APC (mouse, #105812, 1:200 dilution), Anti-Sca-1-PE-cy7 (mouse, #108114, 1:200 dilution), Anti-CD150-PE (mouse, #115904, 1:200 dilution), Anti-CD48-APC/Cyanine7 (mouse, #103432, 1:200 dilution), Anti-Ki67-FITC (mouse, #652410, 1:200 dilution), Hoechst34580 (#565877, BD Pharmingen), Anti-CD16/32-PE (mouse, #101308, 1:200 dilution), Anti-CD34-FITC (mouse, #11-0341-82, 1:200 dilution eBioscience), Anti-CD127-APC/Cyanine7 (mouse, #135040, 1:200 dilution), Anti-CD135-Brilliant Violet421 (mouse, #135314, 1:200 dilution), Annexin V (#640941, 1:20 dilution), PI (#421301, 1:50). Intracellular staining was performed using the Foxp3/Transcription factor staining set (eBioscience, Grand Island NY) and Fixation/Methanol protocol (eBioscience, Grand Island NY) provided by the manufacturers.

Colony assays

Mouse AML cells were diluted to the indicated concentration in IMDM with 2% FBS and were then seeded into methylcellulose medium M3534 or M3434 or H4436 (STEMCELL Technologies, Cambridge MA) for myeloid colony formation analysis, as previously described.⁵⁷

RNA-seq analysis

RNA obtained from sorted BM AML cells of two DPP4^+/+ and two DPP4^−/− AML cells transplanted mice, was purified with QIAGEN miRNeasy Mini Kit. The sample concentration was determined by Qubit fluorometer (Invitrogen, Waltham MA) using the Qubit HS RNA assay kit, and RNA integrity was assessed using the Fragment Analyzer automated electrophoresis system. Next, poly-A containing mRNA was purified from total RNA (1mg), RNA was fragmented, double-stranded cDNA was generated from fragmented RNA, and the index-containing adapters were ligated to the ends. Libraries were constructed following the manufacturer’s protocol with reagents supplied in Illumina’s TruSeq mRNA stranded sample preparation kit. The amplified cDNA constructs were purified by addition of Axyprep Mag PCR Clean-up beads. The final construct of each purified library was evaluated using the Fragment Analyzer automated electrophoresis system, quantified with the Qubit fluorometer using the Qubit HS dsDNA assay kit, and diluted according to Illumina’s standard sequencing protocol for sequencing on the NextSeq 500. High-throughput sequencing was performed at the University of Missouri DNA Core Facility.

GSEA

GSEA was used to identify gene sets and pathways associated with gene expression data. The input for GSEA was a list of significantly differentially expressed genes (log2 (fold change) > 2, p < 0.05, and RPKM >0.5) between two biological states as determined by edgeR. http://software.broadinstitute.org/gsea/index.jsp. GSEA (version 2.2.0) is an analytic tool for relating gene expression data to gene sets to identify unifying biological themes.^55,58 Leukemia stem cell gene sets were obtained from the indicated publication. Gene heat maps were clustered by hierarchical clustering (https://software.broadinstitute.org/morpheus/).

Western blotting

Cells were lysed in Laemmli sample buffer (Sigma-Aldrich, St. Louis MO) supplemented with protease inhibitor cocktail (Roche Diagnostics, Indianapolis IN). Samples were separated on SDS–PAGE gels (Bio-Rad, Hercules CA) and transferred to nitrocellulose membranes (Bio-Rad, Hercules CA) for protein detection as described. Primary antibodies were obtained from Cell Signaling Technology (Danvers MA): p-IKB (2859s, 1:1000 dilution, p-p65 (3039s, 1:1000 dilution), p-Akt (4046s, 1:2000 dilution), p-38MAPK (4092s, 1:1000 dilution), p-SRC (59548s, 1:1000 dilution), Dpp4 (67138s, 1:1000 dilution). Other antibodies were obtained thus: c-src/src (sc-8056, 1:1000 dilution, Santa Cruz Biotechnologies, Dallas TX), β-actin (sc-47778, 1:2000 dilution, Santa Cruz Biotechnologies, Dallas TX), Flag M2 (F1804, 1:10,000 dilution, Sigma-Aldrich, St. Louis MO), and anti HA (16B12, 1:10,000, Covance, Princeton NJ) as well as horseradish peroxidase (HRP) conjugated secondary antibodies (HAF007, 1:1,000, and HAF005, 1:1,000, R&D system, Minneapolis MN) and chemi-luminescent substrate (Invitrogen, Waltham MA).

Blood glucose measurements

Blood samples were analyzed for glucose (AlphaTRACK, Abbott IL) at time 0 and 240 min following Linagliptin (Boehringer Ingleheim, Ingelheim am Rhein, Germany) administration by oral gavage at a concentration of 1 mg/kg.

Quantitative RT-PCR

RT-PCR was performed on 5 ng total RNA with gene-specific primers (as shown below) and a QIAGEN One Step RT-PCR kit (210210; Qiagen, Germantown MD) according to the manufacturer’s protocol. β-actin was used as the reference gene to normalize the relative expression for quantitative RT-PCR analysis. RT-qPCR primers used in this study were as follows: dpp4: forward (5′- TAC AAA AGT GAC ATG CCT CAG TT-3′) and reverse (5′- TGT GTA GAG TAT AGA GGG GCA GA-3′); β-actin: forward (5′-GCT CTT TTC CAG CCT TCC TT-3′) and reverse (5′-CTT CTG CAT CCT GTC AGC AA-3′); hoxa5: forward (5′- CTC ATT TTG CGG TCG CTA TCC-3′) and reverse (5′- ATC CAT GCC ATT GTA GCC GTA-3′); hoxa3: forward (5′- TCA GCG ATC TAC GGT GGC TA′) and reverse (5′- GAG GCA AAG GTG GTT CAC CC-3′); hoxa7: forward (5′- GCG CTT TTT AGC AAA TAT ACG GC-3′) and reverse (5′- GGG ATG TTT TGG TCG TAG GAG-3′); hoxa9: forward (5′- CCC CGA CTT CAG TCC TTG C-3′) and reverse (5′- CCA GGA GCG CAT ATA CCT GC-3′); hoxa11: forward (5′- TCT TCG CGC CCA ATG ACA TAC-3′) and reverse (5′- GGC TCA ATG GCG TAC TCT CT-3′); Meis1: forward (5′- TCA GCA AAT CTA ACT GAC CAG C-3′) and reverse (5′- AGC TAC ACT GTT GTC CAA GCC-3′); Myb: forward (5′- AAC TCC CAC ACC ATT CAA AC-3′) and reverse (5′- GCC ACA ATT CCA GAT TCA TCC-3′); Mef2c: forward (5′- ATC CCG ATG CAG ACG ATT CAG-3′) and reverse (5′- AAC AGC ACA CAA TCT TTG CCT-3′).

Bone marrow histology

A bone marrow biopsy was performed at the indicated time points. The biopsy was decalcified in EDTA, embedded in frozen section compound 22 blue (Leica), sectioned at 4 mm, placed onto coated Superfrost Plus Microscope Slides (Fisher Scientific), and fixed in 4% phosphate-buffered formalin. Fixed slides were then stained with hematoxylin/eosin (Thermo Scientific^™ Shandon^™ Rapid-Chrome H & E Frozen Section Staining Kit). Images were acquired by BZ-8000 fluorescence microscope with a 20x and 60x objective (resulting in 200X and 600x magnification.)

In silico analyses and survival analysis

Data were obtained from the TCGA AML database (https://tcga-data.nci.nih.gov/tcga/, n = 187), the GSE6891 dataset (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE6891, n = 520), and the GSE10358 dataset (n = 91). Expression was normalized to total mRNA. Patients were separated into three groups (low, medium, and high) based on the expression levels of DPP4 to perform Kaplan-Meier survival analysis (GraphPad Prism, version 7.0, GraphPad, San Diego CA).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical details for each experiment can be found in the respective figure legends. Statistical analysis was performed using Graphpad Prism version 9.0 (GraphPad, San Diego CA). For comparison of two groups, we used Student’s t-test and when comparing 3 or more groups, one-way or two-way ANOVA was used with Fisher’s test for post-hoc analyses. The survival rates of the two groups were analyzed using a log–rank test and were considered statistically significant if p < 0.05. Analysis of RNAseq was analyzed as described in respective methods. Data with error bars represent means ± standard error of the means (SEM). All the experiments had at least 2–3 independent repeats. Different levels of statistical significance were denoted by p values (*p < 0.05, **p < 0.01, ***p < 0.001).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-mCD3e-PE/Cyanine5	Biolegend	Cat#100310; RRID:AB_312675
Anti-mLy-6G/Ly-6C (Gr-1)-PE/Cyanine5	Biolegend	Cat#108410; RRID:AB_313375
Anti-mCD11b-PE/Cyanine5	Biolegend	Cat#101210; RRID: AB_312793
Anti-mCD45R-PE/Cyanine5	Biolegend	Cat#103210; RRID: AB_312995
Anti-mTer-119-PE/Cyanine5	Biolegend	Cat#116210; RRID: AB_313711
Anti-mCD117 (c-Kit)-APC	Biolegend	Cat#105812; RRID: AB_313221
Anti-mSca-1-PE-cy7	Biolegend	Cat#108114; RRID: AB_493596
Anti-mCD150-PE	Biolegend	Cat#115904; RRID: AB_313683
Anti-mCD48-APC/Cyanine7	Biolegend	Cat#103432; RRID: AB_2561463
Anti-mKi67-FITC	Biolegend	Cat#652410; RRID: AB_2562141
Hoechst34580	BD Pharmingen	Cat#565877; RRID:AB_2869723
Anti-mCD16/32-PE	Biolegend	Cat#101308; RRID: AB_312807
Anti-mCD34-FITC	eBioscience	Cat#11-0341-82; RRID: AB_465021
Anti-mCD127-APC/Cyanine7	Biolegend	Cat#135040; RRID: AB_2566161
Anti-mCD135-Brilliant Violet421	Biolegend	Cat#135314; RRID: AB_2562339
Anti-hCD45-PE	Biolegend	Cat#304039; RRID: AB_2562057
Annexin V	Biolegend	Cat#640941; RRID: AB_2616657
7-AAD	Biolegend	Cat#420404
PI	Biolegend	Cat#421301
Bacterial and virus strains
pLentiLoxp3.7	Addgene, Watertown MA	# 11795
MSCV-MLL-AF9-IRES- YFP/GFP	Addgene	#71443
Mig-AML1-ETO9a	Addgene	#12433
MSCV-SRC-IRES-GFP		This manuscript
Flag-DPP4		This manuscript
MSCV-IP N-HAonly SRC	Addgene	#35017
Biological samples
Primary human AML samples	Ellis Fischel Cancer Center, University of Missouri	N/A
Human peripheral blood samples	Ellis Fischel Cancer Center, University of Missouri	N/A
Chemicals, peptides, and recombinant proteins
5-Fluorouracil	Sigma	F6627
Protease inhibitor cocktail	Roche Diagnostics
H-Ala-Pro-AFC	Bachem Americas	I-1680.0050BA
Ara-C	Sigma-Aldrich	147-94-4
Linagliptin	MedChemExpress	HY-10284
Vildagliptin	MedChemExpress	HY-14291
EDTA	Sigma-Aldrich	10378-23-1
4% phosphate-buffered formalin	Sigma-Aldrich	1.00496
Critical commercial assays
mycoplasma contamination kit	R&D Systems	CUL001B
Foxp3/Transcription factor staining set	eBioscience	00-5523-00
miRNeasy Mini Kit	QIAGEN	217084
Qubit HS RNA assay kit	Invitrogen	Q32852
Qubit HS dsDNA assay kit	Invitrogen	Q32851
One Step RT-PCR kit	QIAGEN	210212
Shandon™ Rapid-Chrome H & E Frozen Section Staining Kit	Thermo Scientific	Epredia™ 9990001
Deposited data
RNA-seq datasets	This study	“SRA database: SRP323430”
Experimental models: Cell lines
THP-1	ATCC	TIB-202
MV4-11	ATCC	CRL-9591
K562	ATCC	CCL-243
697	DSMZ	ACC 42
Kasumi2	DSMZ	ACC 526
Sup15B	ATCC	CRL-1929
KOPN8	DSMZ	ACC 552
RCH-ACV	DSMZ	ACC 548
REH	ATCC	CRL-8286
MHH-Call-2	DSMZ	ACC 341
NALM-6	ATCC	CRL-3273
MOLM	DSMZ	ACC 554
MM6	DSMZ	ACC 124
Kasumi1	ATCC	CRL-2724
U937	ATCC	CRL-1593.2
Experimental models: Organisms/strains
C57 BL/6	Charles River	N/A
CD45.2	Charles River	N/A
CD45.1	Charles River	N/A
C57Bl/6NTac-DPP4tm1a Wtsi/Ics	European Mouse Mutant Cell Repository	N/A
129S4/Bl6-Gt(ROSA)26Sortm2 (FLP*)Sor/J	Jackson	JAX: 012930
Vav-iCre	Jackson	JAX: 018968
Oligonucleotides
Mouse primers for qPCR, see Method details	Sigma	N/A
Human primers for qPCR, see Method details	Sigma	N/A
Software and algorithms
GraphPad Prism version 7.0	GraphPad software
Flowjo v10	BD	https://www.flowjo.com/solutions/flowjo
Quant StudioTM Real-Time PCR software v1.3	Applied Biosystems	https://www.thermofisher.com/de/de/home/global/forms/life-science/quantstudio-6-7-flex-software.html
edgeR		http://software.broadinstitute.org/gsea/index.jsp
hierarchical clustering		https://software.broadinstitute.org/morpheus
GSEA version 2.2.0	Mootha et al., 200355

Highlights.

• AML cell survival depends on DPP4-mediated upregulation of Src-NF-κB axis

• DPP4 deficiency renders AML cells susceptible to apoptosis while sparing normal HSCs

• DPP4 confers stemness to AML cells

• Currently available DPP4 inhibitors could be added to therapeutic regimen in AML