Here we show that human NK cells express FAP, which is a key regulator of NK cell migration, invasion, extravasation and tumor infiltration. This novel finding significantly broadens the existing understanding of NK cell migration and tissue infiltration, and illustrates a mechanism for NK cell extravasation from blood vessels. Our findings reveal that both knockout and inhibition of FAP restrict NK cell tumor infiltration, and attenuate NK cell-mediated tumor cell lysis, underscoring the critical role of FAP-mediated migratory mechanisms in the anti-cancer activity of NK cells. Importantly, FAP overexpression enhances NK cell invasion through matrix, promoting tumor infiltration both in vitro and in vivo. Therefore, this work reveals novel insights into FAP biology and NK cell biology, with important implications for emerging NK cell-focused therapeutic strategies.
For extravasation or tissue invasion, cells must penetrate the basement membrane and interstitial tissue. During this process they are confronted by a 3D ECM that provides a substrate for adhesion and traction, as well as biomechanical resistance. For cells to traffic effectively through the ECM, which can offer narrow or non-existent pores for passage, leukocytes must adopt contracted shapes. Excessive cellular deformation can result in nuclear rupture that causes genomic damage, long-term genomic alterations and limited cellular survival. To circumvent nuclear damage, some cells employ proteolytic digestion to widen pores in the ECM20. Although proteolytic migration is considered less common in leukocytes versus other cell types, it has been documented. Zebrafish neutrophils and macrophages use proteolytic digestion for basement membrane transmigration34. Human neutrophils secrete elastase, a serine protease, to facilitate their endothelial transmigration35.
Unlike other immune cell types, there are few studies investigating the physical mechanisms driving NK cell migration. Decades-old research demonstrated that mouse and rat NK cell migration through Matrigel was dependent on MMPs27,36,37. More recent studies have used physiologic models. Putz et al. showed that heparinase regulated mouse NK cell infiltration into murine tumors38. Prakash et al. showed that granzyme B released from murine cytotoxic lymphocytes, including NK cells, enhanced lymphocyte extravasation via ECM remodeling, although it did not affect interstitial migration. They confirmed that a granzyme B inhibitor reduced human donor T cell transmigration through a Matrigel coated semi-permeable membrane (i.e., Boyden chamber assay)39. Although these authors did not assess changes in human donor NK cell migration in response to a granzyme B inhibitor, it is reasonable to assume it would be similar to that of T cell migration, since both cell types express and release granzyme B. However, our finding that FAP is expressed in human NK cells, but not in all murine NK cells or other human immune cell types (Fig. 1), suggests that some migratory mechanisms can be cell-type and species-specific. Unlike these previous studies that investigated either extravasation or tumor infiltration, we investigated both and found that NK cells use the same proteolytic migration strategy for basement membrane degradation/extravasation as well as tumor tissue infiltration. We further demonstrate that defects in proteolytic migration directly impair the ability of NK cells to infiltrate and lyse tumor cells.
FAP is a well-studied protein. Although once thought to be restricted to activated fibroblasts, FAP expression has been found in additional cell types such as epithelial tumor cells40–42, melanocytes43 and macrophages44,45. In non-immune cells, FAP enhances cellular invasion43,46–49. The role of FAP in macrophages is less clear. Arnold et al. showed that in murine tumors there is a FAP+ minor sub-population of immunosuppressive F4/80hi/CCR2+/CD206+ M2 macrophages. While this study highlighted how FAP+ macrophages affect tumor growth, FAP’s function in these macrophages was not described44. Tchou et al. identified FAP+CD45+ cells in human breast tumors by immunofluorescence. They then used flow cytometry to demonstrate that a portion of these FAP+CD45+ cells were CD11b+CD14+MHC−II + tumor associated macrophages. Since the flow cytometry panel used to categorize these FAP+CD45+ cells consisted of only macrophage markers, those data do not exclude the possibility that some of the FAP+CD45+ tumor cells were NK cells. In contrast to that study, we did not identify FAP expression in human macrophages (CD14+ cells) (Fig. 1F). However, we examined circulating cells, as opposed to cells in the tumor microenvironment. Future studies are needed to further categorize FAP expression in tumor immune cell populations, potentially using multicolor immunofluorescent staining. Additionally, more studies are needed to determine if the function of FAP in FAP+ tumor macrophages is the same as we have described here in NK cells.
The finding that human NK cells express FAP (Fig. 1D) has several clinical implications for existing FAP-targeted therapies. For example, an anti-FAP/IL-2 fusion protein has been utilized in clinical trials though the results are not yet published (NCT02627274). The proposed mechanism of action of this drug is that it targets IL-2 to FAP expressing tumor stroma, thereby limiting on-target, off-site toxicities associated with IL-2 cytokine therapy. Our findings that FAP is expressed on the NK cell surface suggests that anti-FAP/IL-2 fusion protein may also target IL-2 directly to NK cells, enhancing NK cell activation and potentially tumor clearance.
Anti-FAP CAR therapies are also in development to treat conditions such as cardiac fibrosis50,22, malignant pleural mesothelioma51, lung adenocarcinoma52 and other cancers53,54. Our data suggest that anti-FAP CAR cells may also be useful in NK cell malignancies such as aggressive NK-cell leukemia. There are potential caveats to the clinical use of anti-FAP CAR T cells. It is plausible that an anti-FAP CAR T cell could induce NK cell lysis, resulting in NK cell leukopenia in humans, this toxicity might be missed in preclinical murine models. For cancer immunotherapy, an ideal anti-FAP CAR would be engineered to target FAP expression by fibroblasts while sparing NK cells. It should be noted that Gulati et al. performed the first-in-human trial of an anti-FAP CAR T cell therapy and demonstrated that a FAP CAR T cell therapy induced stable disease for 1 year in a patient with malignant pleural mesothelioma without any treatment-terminating toxicities51.
Our finding that FAP regulates NK cell tissue infiltration (Figs. 6 and7) has clinical implications as well. These results imply the potential value of NK cells engineered to overexpress FAP in enhancing tumor infiltration and cell lysis.
Existing strategies aimed at enhancing NK cell infiltration into tumors rely on manipulating chemokine/receptor pathways. For example, Wennerberg et al. demonstrated that ex vivo expanded NK cells express higher levels of chemokine receptor CXCR3 than unexpanded NK cells which then demonstrated increased migration towards CXCL10 expressing melanomas18. Another approach that has been utilized is engineering NK cells to overexpress CXCR2, a chemokine receptor. This study showed that CXCR2 overexpressing NK cells had enhanced trafficking towards and lysis of renal cell carcinoma cells in vitro19. These findings suggest that these strategies to enhance NK cell migration are feasible, however, chemokine pathway-altering strategies require not only elevated expression of the chemokine receptor on NK cells, but also secretion and maintenance of chemoattractants by the tumor. Additionally, many chemoattractants recruit multiple immune cell types, including immunosuppressive cells. For example, CXCL10 is a chemoattractant for cytotoxic T lymphocytes and NK cells, but also for regulatory T cells56. We postulate that the ideal migration-altering therapeutic approach would increase cytotoxic immune cell infiltration in tumor masses, without influencing or even reducing immunosuppressive immune cell content in the TME. Since overexpressing FAP enhances NK92 cell tumor infiltration and lysis in vitro and in vivo (Figs. 6 and7), we speculate that engineering NK cells to overexpress FAP, either in autologous NK cell or NK CAR-NK therapies, could increase NK cell tumor infiltration and lysis. This approach is independent of tumor-associated factors, such as chemoattractant secretion, and would not be expected to induce the infiltration or expansion of immunosuppressive cell populations into the tumor microenvironment. Since proteolytic migration is required for NK cell killing of malignant cells (Fig. 6), the ability to alter protease expression or activity to enhance NK cell tumor infiltration represents a potentially promising approach to altering NK cell anti-tumor activity. Future studies are needed to explore the benefit of FAP-overexpressing NK cells in preclinical models and in clinical studies, and to determine what, if any, toxicities they induce.
In this study we have demonstrated that human NK cells express FAP and that FAP directly affects NK cell migration, extravasation and tumor infiltration. These findings further the understanding of both FAP and NK cell biology. Importantly, FAP overexpression promotes the infiltration of NK92 cells into human tumor xenografts, suggesting a role for manipulating FAP expression to promote NK cell therapeutics. Future studies will determine if these novel findings have meaningful implications for NK cell-based therapy strategies currently in development.