Targeting COX-2 abrogates mammary tumorigenesis

Epidemiological studies and clinical trials show therapeutic protection against cancers of the colon, prostate, lung, breast, and other tissues afforded by long-term use of cyclooxygenase-2 (COX-2) selective inhibitors, the ‘coxibs.’¹ These findings have been recapitulated across COX-2 deficient cancer mouse models, with suppressed COX-2-derived prostaglandin E₂, a dominant pro-inflammatory mediator that also promotes proliferation and angiogenesis while suppressing apoptosis and immune function, identified as the principal anticancer event.² However, emergence of significant cardiovascular risks, arising from collateral suppression of the cardioprotective endothelial COX-2 derived mediator prostacyclin, have raised doubts about the benefits of systemic COX-2 inhibition.³ We engineered mice lacking COX-2 selectively in distinct cellular compartments—mammary epithelium (COX-2 KO^epi) or myeloid cells (COX-2 KO^myeloid)—to delineate the role of COX-2 in mammary tumorigenesis and to determine the protective effects of cell-specific COX-2 inhibition, thus avoiding systemic cardiovascular side-effects.^4-6

In COX-2 KO^epi mice mammary tumor onset, induced either by the carcinogen DMBA or transgenic expression of an activated HER2/neu oncogene, was delayed.^4,5 Although qualitatively and quantitatively different, in both models significant changes were evident in the tumor microenvironment involving tumor-associated macrophages (TAM) and tumor-infiltrating lymphocytes (TIL), which are essential to immune destruction of tumor cells.

In the DMBA model, the dominant effect of epithelial COX-2 deletion was augmented TAM,⁴ a counterintuitive finding since TAM are typically of the M2 phenotype that promotes tumor growth by mechanisms that include suppression of innate and adaptive anti-tumor immunity.⁷ However, phenotypic marker and cytokine analysis revealed that COX-2 KO^epi mammary TAM were skewed toward the M1 phenotype that supports type 1 anti-tumorigenic CD4⁺ T helper type 1 lymphocytes (T_h1). Concordantly, increased CD4^+, and correlation between M1-TAM and T_h1 markers, was detected in DMBA-induced COX-2 KO^epi mammary tumors. In addition, natural killer (NK) cell markers were increased in COX-2 KO^epi tumors, further supporting a paracrine influence of epithelial cell COX-2 to suppress anti-tumor immune function.

In the HER2/neu model, which more closely resembles human disease, the phenotype was more profound – COX-2 KO^epi tumor onset was delayed, and multiplicity reduced, co-incident with decreased cell proliferation, suppressed angiogenesis, and a shift toward increased infiltrating CD3⁺CD4⁺ (likely T_h1), CD3⁺CD8⁺ (cytotoxic T-lymphocytes or CTL), and CD3^- CD8⁺ (likely NK) immune cells.⁵ Mechanistically, this shift toward anti-tumor immune function appeared related not to TAM function but rather to modified COX-2 KO tumor cell function to produce more of the lymphocyte chemo-attractant CXCL-9 and less of the lymphocyte co-inhibitory molecule PD-L1. This novel link between COX-2 and PD-L1 function was confirmed when lower PD-L1 expression was observed in HER2/neu transformed mammary tumor cells transduced with shRNA to knock down (KD) COX-2. Moreover, COX-2 KD cells failed to grow orthotopic tumors in wild type mouse fat pads. The critical role of CD8⁺ cells was apparent when CD8⁺ cell depletion in wild type mice restored COX-2 KD tumor growth to the level seen with normal tumor cells.

Together these studies indicated a significant role for tumor cell COX-2 in tumor immune suppression. Given the central role of TAM in the tumor immune microenvironment, and the central function of COX-2 in inflammatory macrophages, we developed HER2/neu transgenic mice lacking COX-2 in macrophages, as well as other myeloid cells (COX-2 KO^myeloid). Significant suppression of mammary tumor onset, multiplicity and growth was evident.⁶ Importantly, orthotopic tumor growth of normal mammary tumor cells was dramatically reduced in COX-2 KO^myeloid hosts compared with wild type, underscoring the significant pro-tumor role of myeloid/macrophage COX-2. Similar to the COX-2 KO^epi model, increased CD4⁺ and CD8⁺ TIL were evident in COX-2 KO^myeloid tumors. This appeared secondary to reduced TAM-mediated T-cell suppression that was driven by impaired colony stimulating factor-1-dependent migration of COX-2 KO macrophages to tumors, as well as restrained differentiation to the immune-suppressive M2 TAM phenotype. Host depletion of CD8⁺, but not CD4⁺, cells restored orthotopic tumor growth in COX-2 KO^myeloid mice, again underscoring the contribution of COX-2 to suppressing the crucial anti-tumor role of CD8⁺ cells.

Immunosurveillance by CD8⁺ CTL and NK cells has emerged as critical for controlling nascent in situ tumor growth. This has led to a wave of promising new therapeutic strategies, including blockade of immune checkpoints such as CTLA-4, PD-L1, and its receptor PD-1, or adoptive transfer of anti-tumor T lymphocytes, with significant and durable clinical responses reported in trials.^8,9 While the efficiency of COX-2 inhibition in promoting cytotoxic immune function should be verified in human breast and other cancers, our mouse studies implicate COX-2 in the ‘broken’ immune-surveillance that allows for mammary tumor progression (Fig. 1), and demonstrate that, rather than the hazardous systemic inhibition, targeted cell-specific COX-2 inhibition is sufficient to achieve the benefit. These studies open possibilities for combination of COX-2 inhibitors to potentiate the clinical effect of checkpoint blockade or adoptive T-cell transfer. Remarkably, our studies show that targeted inhibition of COX-2 in macrophages, a component of the complex tumor microenvironment, produced substantial antitumor immune function and suppressed tumor growth, revealing opportunities to combine targeted delivery of COX-2 selective inhibitors directly to macrophages. One approach is to incorporate coxibs into HDL nanoparticles, which are readily engulfed by macrophages,¹⁰ to achieve the tumor suppressive effect of systemic COX-2 inhibition but without thromboembolic risk, because COX-2 function in healthy vascular endothelium is spared. With such targeted strategies, the promise of COX-2 inhibitors in cancer prevention and therapy, which was all but lost to the cardiovascular risk of systemic inhibition, may be renewed and realized.