A novel HLA-A*0201 restricted peptide derived from cathepsin G is an effective immunotherapeutic target in acute myeloid leukemia

Abstract

Purpose: Immunotherapy targeting aberrantly expressed leukemia-associated antigens has shown promise in the management of acute myeloid leukemia (AML). However, because of the heterogeneity and clonal evolution that is a feature of myeloid leukemia, targeting single peptide epitopes has had limited success, highlighting the need for novel antigen discovery. In this study, we characterize the role of the myeloid azurophil granule protease cathepsin G (CG) as a novel target for AML immunotherapy.

Experimental design: We used Immune Epitope Database and in vitro binding assays to identify immunogenic epitopes derived from CG. Flow cytometry, immunoblotting, and confocal microscopy were used to characterize the expression and processing of CG in AML patient samples, leukemia stem cells, and normal neutrophils. Cytotoxicity assays determined the susceptibility of AML to CG-specific cytotoxic T lymphocytes (CTL). Dextramer staining and cytokine flow cytometry were conducted to characterize the immune response to CG in patients.

Results: CG was highly expressed and ubiquitinated in AML blasts, and was localized outside granules in compartments that facilitate antigen presentation. We identified five HLA-A*0201 binding nonameric peptides (CG1-CG5) derived from CG, and showed immunogenicity of the highest HLA-A*0201 binding peptide, CG1. We showed killing of primary AML by CG1-CTL, but not normal bone marrow. Blocking HLA-A*0201 abrogated CG1-CTL-mediated cytotoxicity, further confirming HLA-A*0201-dependent killing. Finally, we showed functional CG1-CTLs in peripheral blood from AML patients following allogeneic stem cell transplantation.

Conclusion: CG is aberrantly expressed and processed in AML and is a novel immunotherapeutic target that warrants further development.

Figure 1

Cathepsin G is overexpressed in myeloid leukemia. (A) Western blot (WB) showing CG expression in whole cell lysates (WCL) from 8 acute myeloid leukemia (AML) samples. Purified CG (15 ng), P3 (0.5 μg) and U-937 WCL (30 μg) were used as positive controls for CG, P3 and NE, respectively. (B) WB showing high CG expression in AML samples, in comparison with granulocytes (Gran) from four different normal individulas. WCL (35 μg) from each sample were loaded into each well; purified CG (15 ng) was loaded as positive control.

Figure 2

CG is located outside azurophil granules in AML. (A) Confocal microscopy staining of normal granulocytes and two patient leukemia samples show CG outside azurophil granules in AML blasts, in contrast with granulocytes where CG is located within granules. (B, C) Graphical representation of the granularity in leukemia blasts from two AML patients, AML #2 and AML#5 (n= 1376 cells) vs. normal granulocytes from a healthy individual (n=171 cells). (B) The ratio of the dimmest to brightest 10% of pixels in each cell was determined and the distribution of cells was graphed. Low dim/bright pixel ratio indicates granularity. (C) Percentage of granulocytes and AML blasts with a dim:bright pixel ratio <0.6 from 2 different AML patient samples (AML #2 and AML #5) and 1 normal granulocyte sample. Dim:bright ratio <0.6 indicates granularity.

Figure 3

CG1 binds HLA-A*0201 with a high affinity. (A) Schematic of CG showing the signal peptide, which is cleaved in the endoplasmic reticulum and is the source of CG1-CG3. Dashed lines demonstrate the activation dipeptide that is cleaved to generate active protease. (B) T2 cells were co-cultured for 90 minutes with CG derived peptides at increasing concentration and then stained with FITC conjugated anti-HLA-A*0201 antibody. Results presented are the mean ± SD mean fluorescence intensity (MFI) from triplicate staining groups and are representative of 3 independent experiments. (C) Peptides were incubated in HLA-coated wells at concentrations ranging from 10⁻⁴ to 10⁻⁸ molar (M) at 21°C overnight. Results represent duplicate data points and were graphed relative to the binding of the positive (POS) control peptide (FLPSDFFPSV) at 10⁻⁴ M. The ED₅₀ was determined using GraphPad Prism’s nonlinear regression ‘log (agonist) versus response –variable slope (four parameter)’ curve. (D) CG1 was incubated in HLA-coated wells at a concentration of 11 mM at 21°C overnight, then washed and incubated at 37°C and read over the course of 8 hours. Results represent duplicate data points and were graphed relative to the positive (POS) control peptide as 100% binding. CG1 ED₅₀=1.1 μM; t_1/2≈14 hours. The t_1/2 was calculated using GraphPad Prism’s nonlinear regression, ‘dissociation – one phase exponential decay’ curve.

Figure 4

Killing of HLA-A*0201 leukemia by CG1-specific CTLs. (A) Calcein AM cytotoxicity assay using CG1-CTL as effector cells shows specific killing of CG-expressing HLA-A*0201 (HLA-A2⁺) primary AML, with minimal to no killing of HLA-A*0201 negative (HLA-A2⁻) or normal HLA-A2⁺ bone marrow. CG1-pulsed and PR1-pulsed T2 cells were used as positive and negative controls, respectively. (B) CG1-CTL cytotoxicity is HLA-A*0201 dependent. HLA-A*0201 primary AML targets were cultured with CG1-CTLs +/− the anti-HLA-A*0201 antibody BB7.2. Calcein AM release was used to measure cytotoxicity. (C) Cytotoxicity assays showing that CG1-CTL specifically lyse CG-expressing HLA-A*0201 AML. The median fluorescence intensity (MFI) of intracellular CG staining of each of the AML samples was determined using flow cytometry and is shown in parenthesis (Supplementary Fig. S3). Pt. 6 is HLA-A2 negative. ND indicates not determined by flow cytometry.

Figure 5

Higher expression of CG is detected in leukemia stem cells (LSC). (A) Sorted Lin⁻ CD34⁺ CD38⁻ stem cells were permeabilized and stained with alexa-647 conjugated anti-CG antibody (red) and DAPI (blue) nuclear stain. Confocal microscopy imaging of stem cells demonstrates higher expression of CG in LSC in contrast with normal stem cells (NSC). (B) Graphical representation showing the mean±SD intensity of CG staining in 2 LSC samples, LSC 1 (n=234 cells) and LSC 2 (n=570 cells), in contrast with normal stem cells pooled from four different individuals (n=357 cells). Asterisks indicate statistically significant higher expression of CG in LSC: P<0.0001. ANOVA followed by Tukey test was performed using Prism 5.0 software