Article Text
PDF
Abstract
Background Immune checkpoint inhibitors have revolutionized the treatment of solid tumors, enhancing clinical outcomes by releasing T cells from inhibitory effects of receptors like programmed cell death protein 1 (PD-1). Despite these advancements, achieving durable antitumor responses remains challenging, often due to additional immunosuppressive mechanisms within the tumor microenvironment (TME). Tumor-associated macrophages (TAMs) contribute significantly to the immunosuppressive TME and play a pivotal role in shaping T cell-mediated antitumor responses. Leukocyte immunoglobulin-like receptor subfamily B member 2 (LILRB2), expressed on myeloid cells, including TAMs, is an inhibitory receptor, which contributes to macrophage-mediated immunosuppression. In this study, we present AZD2796, a high-affinity anti-LILRB2 antibody designed to repolarize TAMs from an immunosuppressive to a proinflammatory phenotype.
Methods Anti-LILRB2 antibodies were identified using single-B-cell encapsulation Immune Replica technology. The ability of AZD2796 to enhance proinflammatory responses from macrophages treated with CD40 ligand or lipopolysaccharide was assessed using a macrophage stimulation assay. A tumor cell/macrophage/T cell co-culture assay was developed to evaluate the effect of AZD2796, as a single agent and in combination with an anti-PD-1 antibody, on the cytolytic activity of antigen-specific T cells. In vivo assessments were then carried out to determine the ability of AZD2796 to alter tumor growth rate in mice humanized with CD34 hematopoietic stem cells.
Results In preclinical assessments, AZD2796 skewed macrophage differentiation away from an immunosuppressive phenotype and enhanced the proinflammatory function of macrophages. AZD2796 significantly increased the anti-tumor response of T cells following PD-1 checkpoint blockade, while AZD2796 monotherapy reduced tumor growth in humanized mouse models.
Conclusions These findings support the potential of AZD2796 as an anti-cancer therapy, with the ability to synergize with T-cell-based therapeutics.
- Immunotherapy
- Immune Checkpoint Inhibitor
- Macrophage
- Tumor infiltrating lymphocyte - TIL
Data availability statement
Data are available on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Immune checkpoint inhibitors have shown improved clinical outcomes across various cancer types and, by releasing T cells from the inhibitory effects of programmed cell death protein (PD)-1 or its ligand (PD-L1), have heightened their ability to detect and eliminate malignant cells. However, increasing both the frequency of responding patients and the depth of response requires targeting of additional pathways that facilitate immunosuppression in the tumor microenvironment. Leukocyte immunoglobulin-like receptor subfamily B member 2 (LILRB2) is an inhibitory receptor, which contributes to macrophage-mediated immunosuppression.
WHAT THIS STUDY ADDS
The anti-LILRB2 monoclonal antibody, AZD2796, binds to LILRB2 with high affinity and has been shown to enhance the proinflammatory function of immune-stimulated macrophages as indicated by increased production of tumor necrosis factor alpha (TNFα) and granulocyte macrophage colony-stimulating factor. AZD2796 skews macrophage differentiation away from an immunosuppressive phenotype while increasing metabolic activity and promoting pathways involved with cell migration and antigen presentation. Furthermore, in vitro and in vivo antitumor efficacy was demonstrated, both alone and in combination with an anti-PD-1 antibody.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Together, these preclinical data demonstrate the role of LILRB2 in limiting the impact of T-cell-mediated antitumor immunity. AZD2796 can repolarize tumor-associated macrophages from an immunosuppressive to a proinflammatory phenotype and acts synergistically with other T-cell-based therapies.
Introduction
Immune checkpoint inhibitors (ICIs) have transformed the treatment landscape for solid tumors, yielding improved clinical outcomes across various cancer types. By releasing T cells from the inhibitory effects of receptors such as programmed cell death protein 1 (PD-1) and its ligand programmed cell death ligand 1 (PD-L1), ICIs heighten their ability to detect and eliminate malignant cells. However, the extent and durability of antitumor responses remain a challenge, with resistance factors often stemming from additional immunosuppressive mechanisms within the tumor microenvironment (TME).
Tumor-associated macrophages (TAMs) are prevalent within tumors and contribute to an immunosuppressive TME through a variety of mechanisms.1 2 Macrophages are functionally heterogeneous and display considerable plasticity, and while TAMs are mostly considered to be immunosuppressive (termed “M2”-like), some possess a proinflammatory phenotype (termed “M1”-like TAMs).3 High “M2”-like macrophage infiltration is linked to tumor progression and poor prognosis, whereas “M1”-like TAMs correlate with improved clinical outcomes.4 These associations have been observed following ICI immunotherapies,5–8 emphasizing the impact of macrophage phenotype on shaping T cell-mediated antitumor responses. Consequently, novel antitumor therapies targeting TAMs as well as other myeloid cells are under development9 to potentially act synergistically with ICIs and overcome resistance, to increase the proportion of responders and enhance the durability of response.
Leukocyte immunoglobulin-like receptor subfamily B member 2 (LILRB2; also known as immunoglobulin-like transcript 4 (ILT4), CD85D, monocyte/macrophage immunoglobulin-like receptor 10 (MIR10) and leukocyte immunoglobulin-like receptor 2 (LIR2)) is an inhibitory receptor expressed on most myeloid cell subsets10 11 including TAMs. LILRB2 signaling occurs through immunoreceptor tyrosine-based inhibitory motifs located within its cytoplasmic region which, on ligand binding, recruit SHP1 and SHP2 phosphatases that inhibit proinflammatory signaling of activating receptors,11–14 thereby contributing to the immunosuppression of macrophages.
Multiple ligands have been described for LILRB2, including both the classical and non-classical human leukocyte antigen (HLA) Class I molecules11 13 15 16 and several members of the angiopoietin-like protein (ANGPTL) family.17 18 Of the HLA class I ligands, LILRB2 displays the strongest binding to the non-classical HLA-G, with an affinity that is 3–4 fold higher than to classical HLA class I.19 The primary role of HLA-G is to mediate maternal–fetal immune tolerance. However, aberrant expression of HLA-G has been described in multiple cancer types and is associated with poor clinical prognosis,20 where its interaction with LILRB2 has been postulated to contribute to tumor immune escape.21 Other ligands present in the TME may also drive an M2-like phenotype in TAMs following engagement with LILRB2.
In this study, we explore the impact of LILRB2 activity on macrophage function and suppression of anti-cancer activity of T cells and explore the ability of AZD2796, a high-affinity anti-LILRB2 therapeutic antagonistic monoclonal antibody (mAb), to reverse these immune-suppressive effects and enhance antitumor immunity.
AZD2796 is under development as a potential anti-cancer therapy for patients with solid tumors and offers opportunities for combination with T-cell-based therapeutics.
Methods
Generation of anti-LILRB2 antibodies using immune replica technology
The precursor of AZD2796 was isolated using immune replica technology,22 which creates natively paired immune libraries for phage display selections. Further details are provided in the online supplemental methods.
Supplemental material
Binding affinity assessments for AZD2796
The binding affinity (equilibrium dissociation constant, KD) for AZD2796 antigen-binding fragment (Fab) interacting with human LILRB2 allelic variant 1 was measured by performing multiple kinetic exclusion assays (KinExA, Sapidyne Instruments, Boise, Idaho, USA). Association rate estimate was also performed using KinExA. A comparison of the kinetics and affinity of chip-immobilized biotinylated-AZD2796 IgG interacting with all four human LILRB2 allelic variants was performed using surface plasmon resonance on an 8K Biacore. Further details of all techniques are provided in the online supplemental methods.
Assessment of AZD2796 binding to recombinant LILR proteins by homogeneous time-resolved fluorescence
Homogeneous time-resolved fluorescence (HTRF) assay was used to assess the binding of AZD2796 to allelic variants of LILRB2 and to assess off-target binding to its closest related members of the LILR family. Further details are provided in the online supplemental methods.
Assessment of AZD2796 binding to recombinant LILR proteins by biolayer interferometry
Biolayer interferometry was used to assess off-target binding of AZD2796 to six members of the LILR family and the related molecule leukocyte-associated immunoglobulin-like receptor 1 (LAIR-1). Further details are provided in the online supplemental methods.
Cellular binding of AZD2796
The binding of AZD2796 to cell surface expressed LILRB2 on transfected Jurkat cells was assessed by flow cytometry. Further details are provided in the online supplemental methods.
Blockade of LILRB2 binding to HLA-B
The ability of AZD2796 to block the binding of a human leukocyte antigen-B (HLA-B) tetramer to cell surface expressed LILRB2 on transfected Jurkat cells was assessed by flow cytometry. Further details are provided in the online supplemental methods.
Reporter cell assay to assess blockade of LILRB2/HLA-G binding
A reporter assay was developed to assess the ability of AZD2796 to inhibit LILRB2 interaction with HLA-G. Further details are provided in the online supplemental methods.
Macrophage stimulation assay
A macrophage stimulation assay was used to assess the ability of AZD2796 to enhance proinflammatory responses from macrophages treated with CD40L or lipopolysaccharide (LPS). Statistical analyses were performed using GraphPad Prism, V.9.4.0. For dose response assessments, data across each donor were normalized by transforming tumor necrosis factor alpha (TNFα) levels into a linear scale, where the maximum value achieved was defined as 100%, while 0 pg/mL TNFα was defined as 0%, to give percentage of maximum values. Half maximal effective concentration (EC50) values were calculated using a nonlinear regression model (log agonist vs response–variable slope (four parameters)). The difference in cytokine levels achieved by AZD2796 versus isotype control antibody was determined using a two-way paired t-test. Further details are provided in the online supplemental methods.
Proteomic assessment of AZD2796 treatment throughout differentiation of tumor conditioned macrophages
The effect of AZD2796 treatment throughout macrophage differentiation was performed in the presence of tumor-conditioned media derived from MDA-MB-231 cell culture and assessed by mass spectrometry-based proteomics. Further details are provided in the online supplemental methods.
Three cell co-culture assay
A tumor cell/macrophage/T cell co-culture assay was developed to assess the effect of AZD2796, both as a single agent and in combination with an anti-PD-1 antibody, on the cytolytic activity of antigen-specific T cells. Statistical analysis was performed using one-way analysis of variance (ANOVA) with Holm-Šídák’s multiple comparisons test (GraphPad Prism). Further details are provided in the online supplemental methods.
In vivo assessments of AZD2796
The ability of AZD2796 to alter tumor growth rate in vivo was assessed in NSG-SGM3 mice humanized with CD34 hematopoietic stem cells (HSC). To assess for statistical difference in tumor growth following treatment, a two-way ANOVA, assuming sphericity (equal variability of differences) with a Holm-Šídák’s test for multiple comparisons was performed (GraphPad Prism). Further details are provided in the online supplemental methods.
Results
AZD2796 binds selectively and with high affinity to LILRB2
The affinity of AZD2796 binding to recombinant LILRB2 protein was determined using a kinetic exclusion assay (KinExA).23 AZD2796 was found to bind with an affinity of 3.2 pM, with a 95% CI 1.6 to 5.2 pM (online supplemental figure S1).
Nine non-synonymous polymorphisms occur within the extracellular region of LILRB2 in alleles with a worldwide frequency of greater than 10% (online supplemental table 1). To assess the ability of AZD2796 to bind these allelic variants of LILRB2, a panel of recombinant polymorphic LILRB2 proteins was generated which carried all four extracellular Ig-like domains (online supplemental table 1).
Binding assessments were conducted using HTRF, a biochemical assay based on fluorescence resonance energy transfer that can assess the binding of labeled proteins in solution. The results for these assessments are shown in figure 1A where AZD2796 was found to bind all variants tested, indicating that commonly occurring polymorphisms within LILRB2 do not disrupt AZD2796 binding. Surface plasmon resonance comparisons of the AZD2796 IgG interacting with these four allelic variants confirmed that the interactions were kinetically and thermodynamically undistinguishable (online supplemental figure S2; online supplemental table 2). A KinExA-based association rate estimate is presented in online supplemental figure S3.
Binding assessments of AZD2796. (A) AZD2796 binding to commonly-occurring polymorphic variants of LILRB2 were performed by HTRF using a panel of recombinant LILRB2 proteins (labeled H1–H4 on figure). A negative isotype control antibody and a positive control LILRB2-specific control antibody (1E1) were also assessed. Technical replicates, mean, and SD shown. (B) The binding of AZD2796 to members of the LILR receptor family most closely related to LILRB2 was assessed by HTRF. Binding was compared with a negative isotype control antibody, while for each individual LILR protein, an anti-LILR positive control antibody was used (the details of which are provided in the methods section). Technical replicates, mean, and SD shown. (C) Binding assessment of AZD2796 to the remaining members of the LILR family was performed by biolayer interferometry. Binding was compared with a negative isotype control antibody, while for each individual LILR protein (except LILRB2), a positive control antibody was used. (D) Determination of the binding domain within LILRB2 recognized by AZD2796 using chimeric LILRB1/B2 recombinant proteins. A panel of chimeric recombinant proteins was generated that consisted of differing combinations of LILRB2 Ig domain pairs incorporated into a LILRB1 backbone, as depicted in the diagrams. Binding of AZD2796 to those proteins was assessed using HTRF, alongside an isotype control antibody and an anti-LILRB2 positive antibody (clone 1E1). A second anti-LILRB2 (clone 287219, indicated as anti-LILRB2 (2) on figure) was used as a positive control for the LILRB2 Ig D3&D4 construct. (E) The binding of AZD2796 to Jurkat cells transfected with either LILRB2 or LILRB1 was assessed by flow cytometry alongside an isotype control antibody. Representative results from one out of three experiments are shown. The staining of Jurkat-LILRB1 cells was performed using antibodies at a single concentration of 50 nM. An anti-LILRB1/LILRB2 cross-reactive antibody was used as a positive control for LILRB1 binding. Technical replicates, mean, and SD shown. HTRF, homogeneous time-resolved fluorescence; LAIR1, leukocyte-associated immunoglobulin-like receptor 1; LILRA, leukocyte immunoglobulin-like receptor subfamily A; LILRB, leukocyte immunoglobulin-like receptor subfamily B.
We explored the specificity of AZD2796 by assessing the binding to the other members of the LILR-family of receptors in humans. HTRF was used to assess the binding of AZD2796 to LILRB1, LILRA1, LILRA2, and LILRA3 protein. These LILR are the most closely related to LILRB2 and share a high degree of sequence homology within their extracellular Ig domains. AZD2796 did not show off-target binding to these LILR proteins (figure 1B).
In addition, AZD2796 did not bind to the remaining members of the LILR family (LILRB3, LILRB4, LILRB5, LILRA4, LILRA5, LILRA6 and the related LAIR-1 molecule, figure 1C), as assessed using Octet Biolayer Interferometry, an optical biosensing technology that measures protein interactions.
Taken together, AZD2796 was demonstrated to be LILRB2-specific with no cross-reactivity with other LILR molecules.
To identify the region of LILRB2 bound by AZD2796, a panel of four recombinant proteins was generated that consisted of differing combinations of LILRB2 Ig domain pairs incorporated into a LILRB1 backbone (figure 1D). Binding assessments were performed by HTRF, and AZD2796 was found to bind two of the four constructs (figure 1D), both of which carried the D1 Ig domain which was absent in the remaining proteins. This indicates that the epitope of AZD2796 is located within the D1 Ig domain. The positive control LILRB2 antibody 1E1 also showed the same pattern of binding.
Flow cytometry-based cell binding assays were performed to evaluate binding of AZD2796 to human LILRB2. Jurkat T lymphocyte cells were engineered to stably express full-length human LILRB2 protein, while Jurkat cells stably transduced with full-length human LILRB1 were used as a negative control.
AZD2796 bound to the LILRB2 expressing Jurkat cells in a concentration-dependent manner (figure 1E). The EC50 values obtained by AZD2796 were in the pico-molar range, achieving a mean average EC50 of 54.5 pM (SD=19.5). AZD2796 did not bind to Jurkat cells expressing LILRB1 (figure 1E), which is consistent with the lack of binding observed to recombinant soluble LILRB1 protein (figure 1B). In addition, binding to monocytes from peripheral blood and monocyte-derived macrophages was performed (online supplemental figure S4).
AZD2796 blocks ligand binding of LILRB2
To assess the ability of AZD2796 to block LILRB2/HLA-G interactions, a two-cell co-culture bioactivity assay was developed (figure 2A). Jurkat NFAT-luciferase reporter T cells were genetically engineered to express an activating version of human LILRB2, which consisted of the extracellular region of LILRB2 fused to a CD28 transmembrane portion, followed by CD137 and CD3ζ cytoplasmic activating domains, to generate a chimeric construct that activates the luciferase reporter cassette following ligand binding. The LILRB2 Jurkat reporter cells were then co-cultured with Raji B cells that were genetically engineered to overexpress the non-classical major histocompatibility complex (MHC) class I molecule HLA-G, a high-affinity ligand of LILRB2. AZD2796 showed a concentration-dependent activity (figure 2B) with an IC50 of 220 pM and was able to fully inhibit the interaction of LILRB2 with HLA-G.
Assessment of the antagonistic properties of AZD2796. (A) Graphical representation of the LILRB2-Jurkat NFAT-luciferase reporter cell assay used to measure the bioactivity of AZD2796. The reporter cell line was co-cultured with Raji cells overexpressing HLA-G, a high affinity ligand of LILRB2. (B) Results from the assessment of AZD2796 in the LILRB2-Jurkat NFAT-luciferase reporter cell assay. AZD2796 and an isotype control antibody were assessed for their ability to block LILRB2 interaction with its HLA-G ligand. Data are reported as mean±SD. (C) AZD2796 blocking of an HLA-B tetramer binding to cell surface LILRB2 expressed on a Jurkat LILRB2 transfectant. Data reported as mean±SD. (D) HLA-B tetramer binding to Jurkat parental cells (labeled “Jurkat” on figure), Jurkat cells transfected with either LILRB2 or LILRB1 (labeled “LILRB2” or “LILRB1”, respectively) in the presence or absence of AZD2796. Data are reported as mean±SD. CD, cluster of differentiation; HLA-G, human leukocyte antigen G; LILRB, leukocyte immunoglobulin-like receptor subfamily B; Luc, luciferase.
We assessed the ability of AZD2796 to block the binding of a classical MHC class I molecule to cell surface LILRB2, using a fluorescently labeled recombinant HLA-B tetramer (allele HLA-B*0801). AZD2796 blocked the binding of the HLA-B*0801 tetramer to a Jurkat LILRB2 transfectant in a dose-dependent manner, with an IC50 of 274 pM (figure 2C). As expected, AZD2796 was unable to block the binding of the tetramer to Jurkat cells transfected with LILRB1 (figure 2D).
LILRB2 binds to a region within MHC class I that is conserved across both classical and non-classical molecules, including allelic variants of HLA-A, HLA-B and HLA-C.24 The blocking of LILRB2 binding to both HLA-G and HLA-B*0801 demonstrated that AZD2796 can inhibit LILRB2 interactions with all MHC class I molecules.
AZD2796 enhances proinflammatory macrophage responses to immune stimuli
We evaluated the ability of AZD2796 to enhance proinflammatory responses of macrophages to immune stimuli. Initially, the activity of AZD2796 was assessed on human monocyte-derived macrophages stimulated with soluble CD40L (sCD40L) to mimic activation by T cells via CD40. In the first assessment, the level of secretion of the proinflammatory cytokine TNFα into culture supernatants was evaluated following treatment with AZD2796 over a concentration range from 0.0032 pM to 1 nM. AZD2796 enhanced TNFα release from macrophages in a dose-dependent manner across all 12 donors tested (figure 3A), with an average EC50 of 5 pM±2 pM (SD) (figure 3B).
Cytokine production from CD40L-stimulated macrophages in response to AZD2796. Human monocyte-derived macrophages were incubated with AZD2796, isotype control antibody, or in the absence of antibody (no mAb) for 24 hours in the presence of 500 ng/mL soluble CD40L. (A) Level of secreted TNFα from macrophages from 12 separate donors. For each donor, maximal TNFα level was normalized to 100%. Mean values (±SD) achieved for each individual donor (from technical triplicates) following AZD2796 treatment are plotted alongside the combined values achieved across all donors to the isotype control antibody and no antibody (no mAb) treatments. (B) The EC50 values for AZD2796 from the assessment of each individual donor are shown. Mean±SD plotted. (C) Macrophages were incubated with 50 nM AZD2796. Results show levels in secreted TNFα, GM-CSF, and VEGF-A (all pg/mL) generated from 12 separate donors. Bars represent mean values. Statistical significance between AZD2796 and that of the controls was calculated using a one-way ANOVA incorporating Holm-Šidák’s test for multiple comparisons. P values are presented in the figures as *p<0.05; ***p<0.001. Means of technical triplicates plotted for each donor. Lines show means for the 12 donors. ANOVA, analysis of variance; EC50, half maximal effective concentration; GM-CSF, granulocyte macrophage colony-stimulating factor; mAb, monoclonal antibody; TNFα, tumor necrosis factor alpha; VEGF-A, vascular endothelial growth factor-A.
In a follow-up experiment, we explored the effect of AZD2796 on the secretion of TNFα, granulocyte macrophage colony-stimulating factor (GM-CSF) and vascular endothelial growth factor-A (VEGF-A) from monocyte-derived macrophages (from a cohort of 12 donors) stimulated with sCD40L (figure 3C). Response to AZD2796 (used at 50 nM) was compared with two negative controls: an isotype control antibody and macrophages not treated with antibody. AZD2796 treatment increased levels of TNFα across all donors (p<0.001, one-way ANOVA for AZD2796 comparisons with both negative controls). The proinflammatory cytokine GM-CSF was also increased across all donors following treatment with AZD2796 (p<0.05, one-way ANOVA for comparisons with both negative controls), while the proangiogenic cytokine VEGF-A was significantly decreased across all donors (figure 3C) (p<0.001, one-way ANOVA).
We next explored the impact of LILRB2 blockade on macrophages stimulated with a pathogen-associated molecular pattern to mimic response to bacterial infection. To achieve this, AZD2796 treatment was assessed on macrophages stimulated with LPS using cells derived from six donors (online supplemental figure S5). AZD2796 significantly increased the production of TNFα from LPS-stimulated macrophages (p<0.05, one-way ANOVA for AZD2796 comparison with both negative controls), while VEGF-A was significantly decreased following AZD2796 treatment in conjunction with LPS stimulation (p<0.05 for AZD2796 comparison with both negative controls). An increase in GM-CSF was observed across all donors following AZD2796 treatment of LPS-stimulated macrophages, although this did not reach significance (p=0.16, online supplemental figure S5A). In contrast, AZD2796 had minimal effect on GM-CSF and VEGF-A production from unstimulated macrophages (online supplemental figure S5B), and while TNFα was raised significantly following AZD2796 treatment (achieving a mean average of 134.3±73.11 pg/mL vs 53.0±38.22 pg/mL for the no treatment control (NTC) and 55.2±38.4 for the isotype control (p<0.05 for both comparisons)), this level was considerably lower than those achieved by AZD2796 treatment of stimulated macrophages (3516±1888 pg/mL and 11134±648 pg/mL for CD40L and LPS stimulated, respectively).
AZD2796 can influence macrophage differentiation, polarization and function
To explore the impact of LILRB2 blockade during macrophage differentiation, monocytes were treated with AZD2796 throughout differentiation, and a comprehensive proteomic assessment of the resulting macrophages was performed using mass spectrometry. To increase the relevance to cancer, cell cultures were supplemented with tumor-conditioned media derived from MDA-MB-231 breast cancer cells to restore a TAM-like phenotype in the macrophages.25
In total, over 6000 proteins were identified per treatment condition in the macrophages, of which >400 proteins were significantly upregulated by AZD2796 treatment when compared with the two negative control reactions (isotype control antibody treatment and in the absence of a test antibody), and >250 proteins significantly downregulated (figure 4A). The downregulated cluster of proteins included key immunosuppressive macrophage markers CD14, MRC1 (CD206) and CD163 (figure 4B–D). These were among the proteins displaying the greatest reduction in expression along with the “M2” marker CD209 (figure 4E and F), and SIGLEC-1 (CD169) and C1QC, both of which are enriched in specific subsets of TAMs.26 27 Within the upregulated cluster, the most significantly altered proteins included CCRL2 and TNFSF14. These have been demonstrated to induce immunostimulatory effects in macrophages and anti-tumor immune response.28 29 As expected, minimal differences were observed when comparing responses of macrophages to the two negative control reactions (figure 4G). Hierarchical clustering of 817 differentially regulated proteins further demonstrated the robustness of the data and revealed markedly different protein expression patterns between the AZD2796-treated macrophages and the two negative controls (figure 4H).
Proteomic assessment of tumor-conditioned macrophages treated with AZD2796 during differentiation. Mass spectrometry-based proteomics was used to evaluate biological changes in TAMs on AZD2796 treatment during differentiation from monocytes. (A) Overall number of upegulated and downregulated proteins was represented on the bar chart. Expression patterns of immunosuppressive markers CD14 (B), MRC1 (CD206) (C) and CD163 (D) were assessed. Results were generated from six separate donors. Bars represent mean values from two MS-technical replicates. Statistical significance between AZD2796 and controls was calculated using one-way ANOVA incorporating Tukey’s pairwise comparisons test (ns p>0.05, *p≤0.05, **p≤0.002, ***p≤0.0002). Error bars indicate SD. Differentially expressed proteins were derived using a two-sided Student’s t-test (false discovery rate (FDR) <0.05, fold change ≥2) and represented on volcano plots for AZD2796 versus No mAb (E), AZD2796 versus Isotype (F) and Isotype versus No mAb (G). (H) The heatmap illustrated the overall trends in protein expression and interdonor variability in the statistically significant differentially expressed proteins in AZD2796 versus control samples. (I) Ingenuity pathway analysis revealed significantly enriched pathways in the upregulated and downregulated datasets for the AZD2796 versus isotype control comparison (p value cutoff≤0.05). ANOVA, analysis of variance; CD, cluster of differentiation; IL, interleukin; LFQ, label-free quantification; LXR/RXR, liver X receptor-retinoid X receptor; L1CAM, L1 cell adhesion molecule; mAb, monoclonal antibody; MHC, major histocompatibility complex; MRC1, mannose receptor C-type 1; ns, no significance; PD-L1, programmed cell death ligand 1; PD-1, programmed cell death protein 1; PI3K, phosphoinositide 3 kinase; TAM, tumor-associated macrophages.
Pathway analysis revealed that AZD2796 treatment altered a broad range of cellular functions. Pathways associated with the uptake of metabolites, endocytosis, and lipid metabolism were upregulated on AZD2796 treatment (figure 4I), consistent with an increase in metabolic activity. AZD2796 also upregulated proteins involved with cell adhesion and migration, antigen presentation, and immune cell interactions (figure 4I), while downregulating proteins involved in apoptotic and pyroptotic cell death.
Taken together, the results from the proteomic assessment indicate that AZD2796 skewed macrophage differentiation away from an immunosuppressive phenotype, while increasing metabolic activity and pathways involved with cell migration and antigen presentation.
AZD2796 enhances T-cell-mediated cytotoxicity in the presence of human macrophages
To assess the ability of AZD2796 to modulate human T-cell cytolytic activity, as a single treatment as well as in combination with an anti-PD-1 antibody, an assay was established to evaluate T-cell-mediated lysis of a human tumor cell line in the presence of macrophages. For this purpose, an HLA-A*02:01 positive human epithelial breast cancer cell line, MDA-MB-231, was engineered to express a human cytomegalovirus (HCMV)-encoded peptide derived from pp65. This tumor cell line was used as a target for HCMV-specific T cells previously expanded in vitro from two HLA-A*02:01 positive donors with known reactivity to HCMV. These HCMV-specific T cells were co-cultured with non-autologous macrophages from three donors, culminating in an assessment of six T cell/macrophage donor pairings in the presence of tumor cells.
The presence of macrophages in the co-culture was found to substantially reduce the level of tumor killing by T cells, demonstrating the suppressive effect of macrophages on T cell effector function (figure 5A). When macrophages were present in the T cell/tumor cell co-culture, AZD2796 significantly increased tumor cell death when compared with either NTCs or with anti-PD-1 mAb treatment.
AZD2796 increases tumor cell killing by antigen-specific T cells in the presence of anti-PD-1 mAb, and promotes proinflammatory phenotype on macrophages in T cell/macrophage/tumor co-culture assay. MDA-MB-231 breast cancer cell line engineered to present a CMV-encoded peptide in the context of HLA*A2 was co-cultured in the presence of CMV-specific T cells, macrophages, and drugs for 3 days. Additional cell co-culture combinations in the absence of drug/antibody (no mAb) were also assessed as indicated in the figure. Tumor killing (A) and level of expression of CD86 (B) and CD163 (C) on CD14+ macrophages were assessed by flow cytometry. The levels of the proinflammatory cytokine TNFα in supernatants were assessed by ELISA, and values for fold change compared with no antibody control treated cells were plotted (D). Results were generated from six donor pairs. The arithmetic mean and SD values are plotted. Statistical analysis was performed by one-way ANOVA incorporating Holm-Šidák’s test for multiple comparisons (ns=>0.05, *p≤0.05, **p≤0.01). ANOVA, analysis of variance; CD, cluster of differentiation; CMV, cytomegalovirus; mAb, monoclonal antibody; MFI, mean fluorescence intensity; ns, no significance; PD-1, programmed cell death protein 1; TNFα, tumor necrosis factor alpha.
The combination of AZD2796 and the anti-PD-1 mAb significantly increased the proportion of dead tumor cells compared with that achieved by both monotherapies (figure 5A).
In the same assay, AZD2796 was found to repolarize macrophages from a suppressive phenotype to a proinflammatory phenotype, as demonstrated by a significant increase in the expression of the “M1” marker CD86 on macrophages compared with the NTC (figure 5B) and a significant decrease in the expression of the suppressive “M2” marker CD163 (figure 5C). Levels of TNFα, a key proinflammatory cytokine produced by macrophages, were also significantly increased in culture supernatants following AZD2796 treatment and are consistent with the polarization of macrophages to a proinflammatory phenotype (figure 5D).
Combination of AZD2796 with an anti-PD-1 mAb further increased macrophage polarization (as determined by increased CD86 (figure 5B) and TNFα (figure 5D), and decreased CD163 expression (figure 5C)) when compared with monotherapy.
Taken together, these results demonstrate that AZD2796 can overcome macrophage-mediated suppression of T cells and can improve responses to PD-1 blockade.
AZD2796 is efficacious in vivo
The antitumor activity of AZD2796 was investigated in three independent studies in NSG-SGM3 mice engrafted with CD34+human stem cells (CD34+HSC) derived from cord blood. Each study explored a different human tumor xenograft cell line (SK-Mel-5, a malignant melanoma line; NCI-H358, a lung bronchioalveolar carcinoma line or PC-9, a lung adenocarcinoma line) and used mice reconstituted with a different CD34+HSC cord blood donor. Intraperitoneal administration of AZD2796 monotherapy to animals significantly inhibited the growth of tumors, compared with the activity of its isotype control antibody, across all three studies as shown in figure 6. AZD2796 treatment inhibited tumor growth as early as 5 days post first dose (figure 6).
Effect of AZ2796 on tumor growth in three xenograft models in humanized mice CD34-HSC engrafted NSG-SGM3 mice implanted with (A) NCI-H358 lung cancer tumor cells (isotype n=7, AZD2796 n=8 mice), or (B) SK-MEL-5 melanoma tumor cells (isotype n=10, AZD2796 n=10 mice) or (C) PC-9 lung cancer cells (isotype n=12, AZD2796 n=12 mice) were administered intraperitoneal (IP) with 10 mg/kg of AZD2796 or isotype control antibody twice weekly for 3–5 weeks. Data shown for all mice, no data points were excluded. Time of dosing is indicated on each graph with dashed vertical lines. Intergroup differences were analyzed for statistical significance by a two-way ANOVA, assuming sphericity (equal variability of differences) incorporating a Šídák’s multiple comparison test. The main effect of treatment was significant in each of these models with p<0.0001 for (A, B) and p=0.0001 for (C). Significant differences at each time point are indicated within each graph: *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. Error bars represent SD. Spider plots for each individual mouse per group, are shown in the bottom row. ANOVA, analysis of variance; HSC, hematopoietic stem cell.
Discussion
In this study, we describe the characterization of AZD2796, an antibody designed to repolarize TAMs from an immunosuppressive to a proinflammatory phenotype by blocking the inhibitory receptor LILRB2. This approach aims to enhance T cell antitumor function.
TAMs possess multiple mechanisms that suppress T cell antitumor activity. PD-L1 expressed on TAMs inhibits T cells via the binding of the immune checkpoint molecule PD-1 on activated T cells.30 Furthermore, TAMs release anti-inflammatory cytokines such as transforming growth factor-beta (TGF-β) and interleukin-10 (IL-10),31 creating an immunosuppressive environment that hinders cytotoxic T cells and promotes the development of regulatory T cells (Treg), while also recruiting Tregs through the production of chemokines like CCL22.32 Moreover, TAMs express the enzymes arginase and indoleamine 2,3-dioxygenase that deplete arginine and tryptophan within the TME, impairing T-cell proliferation and anti-tumor function.33 In addition, suppressive TAMs display attenuated antigen presentation, along with decreased expression of costimulatory molecules such as CD86 and reduced production of proinflammatory cytokines, factors that are likely to lessen their capacity to support T-cell activation and recruit CD8+effector T cells.34–36 Multiple strategies are being explored both preclinically and clinically to alleviate the pro-tumor effect of TAMs, including depletion of TAMs, metabolic reprogramming of TAMs, and repolarization of TAMs from a suppressive phenotype to a proinflammatory phenotype.9 Blockade of LILRB2 falls into the latter category.
The ability of LILRB2 to negatively modulate myeloid cell immune activation has been known since LILRB2’s discovery in the late 1990s. More recently, LILRB2 blockade has been shown to repolarize macrophages from an “M2”-like suppressive phenotype to a proinflammatory “M1”-like state and promote anti-tumor activity in preclinical models.12 This potential led to the development of AZD2796, AstraZeneca’s first myeloid checkpoint inhibitor for cancer treatment.
AZD2796 is a fully human antibody specific to LILRB2 that does not bind to any other member of the LILR family. AZD2796 possesses an IgG1 Fc domain that carries three mutations (TM) that reduce Fc-mediated immune effector functions.37 AZD2796 binds to all commonly occurring alleles of LILRB2 and blocks ligand binding. AZD2796 binds LILRB2 with high affinity (KD=3.2 pM), greater than that of other published clinical anti-LILRB2 antibodies (eg, MK-4830, KD=17 nM38 ; JTX-8064, KD=3.18 nM39 ; IO-108, KD=2.1 nM40).
AZD2796 enhances the proinflammatory function of immune-stimulated macrophages, as indicated by increased production of TNFα and GM-CSF. Notably, the impact of AZD2796 on the production of proinflammatory cytokines from resting macrophages is modest. This aligns with the recognized functional role of LILRB2 and the capacity of LILRB2 blockade to enhance existing activation stimuli, rather than triggering spontaneous immune activation.11 13 The latter observation may explain the relatively benign clinical safety profile associated with anti-LILRB2 antibody monotherapies39 41 compared with some therapeutic strategies that can directly activate myeloid cells.42–45
Macrophages play a critical role in the wound healing process, including the promotion of vasculature remodeling through the production of cytokines that drive angiogenesis.46 47 In cancer, macrophages can contribute to tumor growth through the same mechanisms,48 including the production of VEGF-A, a potent angiogenic factor associated with tumor progression and poor prognosis.49 Therefore, the reduction of VEGF-A production by macrophages following AZD2796 treatment may further enhance its antitumor effects in addition to the proinflammatory impact of LILRB2 blockade.
Key proof-of-concept in vitro data using a three-cell co-culture assay demonstrated that AZD2796 can enhance antitumor activity. This assay involved co-culturing antigen-specific T cells with tumor cells in the presence of macrophages. The results revealed the ability of macrophages to inhibit T-cell-mediated tumor killing, with AZD2796 mono-treatment significantly increasing tumor cell lysis. This effect may be attributed to the upregulation of CD86 expression on macrophages and TNFα production, providing additional activation stimuli to the T cells. Notably, the combination of AZD2796 with an anti-PD-1 antibody further restored tumor cell lysis, providing evidence in support of the combination of AZD2796 with T-cell checkpoint inhibitors.
AZD2796 treatment during monocyte differentiation significantly altered the phenotype and metabolic profile of the resulting macrophages. These macrophages exhibit an overall enrichment of pathways involved with metabolic activity, cell migration, immune-related activity including antigen presentation, and a clear downregulation of markers associated with suppressive macrophages (namely CD163, CD206 (MRC1), CD14 and CD209). This suggests that AZD2796 promotes the differentiation of monocytes into macrophages with a greater propensity for proinflammatory function. This finding, together with the effect of AZD2796 on macrophages treated post-differentiation, supports the hypothesis that AZD2796 can repolarize macrophages and influence the differentiation of newly infiltrated monocytes.
Mice lack a true ortholog to LILRB2. Consequently, in vivo efficacy assessments of AZD2796 were performed using humanized mice. These mice (NSG/SGM3) lack a murine adaptive immune system and have reduced and dysfunctional murine innate immune cells; instead, they carry a functioning human immune system (including LILRB2+ myelomonocytic cells) following reconstitution with human CD34+ stem cells derived from cord blood. AZD2796 monotherapy resulted in reduced tumor growth compared with an isotype control antibody in three xenograft models of human cancer. These results demonstrate that AZD2796 heightens antitumor activity in vivo, aligning with our in vitro observations.
Together, these preclinical data demonstrate the importance of the LILRB2 axis in limiting the impact of T cell-mediated antitumor immunity.
Data availability statement
Data are available on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
The animals were humanely treated and housed according to Institutional Animal Care and Use Committee under the approved protocol no. AUP-22-17 in the Laboratory Animal Resources facility at AstraZeneca, an Association for Animal Accreditation of Laboratory Animal Care and US Department of Agriculture-licenced facility. Human PBMCs were obtained from leucocyte cones, which are by-products of the apheresis process of blood donated by healthy individuals to the National Health Service Blood and Transplant (NHSBT). Donors give generic consent for research use and are not identifiable.
Acknowledgments
We thank Leon Venegas for his contributions to this manuscript and the AstraZeneca flow cytometry core team for performing the flow-cytometry-based cell binding assays. Medical editing support, under the direction of the authors, was provided by Asad Mustafa, MSc, of Ashfield MedComms (London, UK), an Inizio company, and was funded by AstraZeneca.
References
Footnotes
DCJ and LI contributed equally.
SJD, RWW, FA and MC contributed equally.
Contributors DCJ serves as guarantor and accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish. Conceptualization/study design: DCJ, LI, KM, RWW and FA. Methodology: DCJ, LI, MW, SP, GB, SA, AMS and SR. Performed experiments: RD, EIC, SB, MW, SP, DGR, GB, SS, KB, SA, AK, RV, MK and SR. Data analysis and interpretation: DCJ, LI, RD, EIC, SB, MW, SA, SP, DGR, AMS, GB, KB, RV, MK, AK and SR. Data collection or curation: DCJ, LI, RD, MW, AMS, SA, EIC, SP, DGR, GB, RV, MK, AK and SR. Writing–original draft: DCJ and LI. Writing–reviewing and editing: DCJ, LI, GB and KB. Approval of the final version: all authors. Supervision: DCJ, LI, KM, RWW, FA, MC, SJD, MSG and SH. DCJ and RWW are currently employed by Immunocore, Abingdon, UK.
Funding This study was sponsored by AstraZeneca.
Competing interests All authors report employment and stocks with AstraZeneca at the time of content development.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.