Granzymes Regulate Proinflammatory Cytokine Responses

Our immune defense against tumor cells and virally infected cells is mediated by cytotoxic lymphocytes: NK cells, NKT cells, γδ TCR cells, and cytotoxic T lymphocytes. These cytotoxic cells can induce apoptosis in aberrant cells via the death receptor pathway and the granule exocytosis pathway (1). The death receptor pathway depends on the engagement of death receptors on the target cell by death receptor ligands on the effector cell, whereas the granule exocytosis pathway involves the release of a set of cytotoxic proteins. These are stored in granules in cytotoxic lymphocytes and include the pore-forming protein perforin and a family of structurally related serine proteases called granzymes (Grs) (2, 3). Upon recognition, Grs are released into the immunological synapse between the cytotoxic lymphocyte and the target cell. The Grs then enter the target cell with the aid of perforin and activate various proapoptotic pathways by cleavage of intracellular substrates (1, 4).

In humans, five different Grs exist: GrA, GrB, GrH, GrK, and GrM. This nonsequential nomenclature is explained by the existence of a more extensive Gr family in rodents. Mice express Grs A–G, K, M, and N, but not H, whereas rats express Grs I and J, in addition to Grs A–C, F, K, and M (5). Human Grs (hGrs), named after their rodent homologs, are encoded on three distinct chromosomal regions: GrA and GrK on chromosome 5, GrB and GrH on chromosome 14, and GrM on chromosome 19 (6). Grs A, B, K, and H are expressed in higher quantities in cytotoxic T lymphocytes compared with NK cells, whereas GrM is more abundant in NK cells (1, 2, 4).

Grs are serine proteases that consist of two six-stranded β-barrels that regulate substrate specificity, in the middle of which lies a catalytic triad containing the amino acids serine, histidine, and aspartic acid. Mutation of one of the amino acids in the catalytic triad renders the Gr catalytically inactive. Although human Grs are ∼40% homologous in amino acid sequence (7), they each cleave their own specific set of substrates, a phenomenon partly caused by differences in the primary substrate specificity or P1 of each Gr. GrA and GrK display tryptase-like activity, cleaving their substrates after an arginine or a lysine. Similar to caspases, GrB cleaves substrates after an aspartic acid or glutamic acid. GrM cleaves after a leucine or a methionine, and GrH cleaves after a tyrosine or a phenylalanine (2, 4). Some Grs share substrates, but none have exactly the same degradome. Thus, Grs have substrate specificities that only partially overlap (8). The likeliness that in vitro hydrolysis of a given substrate is physiologically relevant can be estimated by determining the concentration of substrate required for effective hydrolysis (indicated by the Michaelis constant K_m, the concentration of enzyme at which the enzyme reaction is at half of its maximum rate) and the time required for the Gr to hydrolyze one substrate molecule (indicated by the apparent unimolecular rate constant or turnover number k_cat). When K_m is low and k_cat is high, the catalysis has a greater probability to be of physiological relevance.

During the past decades, the dogma has been held that all Grs induce cell death. However, this dogma has recently been debated, in particular for GrA and GrK (6, 9–11). Additionally, increased Gr levels in serum, plasma, synovial fluid, and/or bronchoalveolar lavage fluid have been described in patients suffering from inflammatory diseases, including rheumatoid arthritis (GrA, B), viral infections (GrA, B, K), Plasmodium falciparum infections (GrA, GrB), experimental endotoxemia or sepsis (GrA, GrB, GrK, GrM), hypersensitivity pneumonitis (GrA, GrB), and acute airway inflammation (GrK) (12–22). These observations prompted researchers to investigate alternative Gr functions in inflammation. The first evidence for an extracellular role for Grs was provided by Sayers et al. (23), who noted that treatment with extracellular rat Gr causes growth inhibition in tumor cell lines (23). More recently, novel extracellular and perforin-independent functions have been identified (24, 25). It has been demonstrated that Gr-mediated extracellular matrix (ECM) degradation may contribute to inflammation (26–30). GrB degrades several ECM components, such as vitronectin, laminin, fibronectin, decorin, and the proteoglycans biglycan and betaglycan (31–35). In mouse models, GrB degrades the ECM proteins fibronectin, decorin, fibrillin-1, and vitronectin, and it contributes to several disease processes, including delayed wound closure, abdominal aortic aneurysm, and skin aging (36–39). These findings point to a pivotal role of GrB in mediating ECM remodeling in disease. Furthermore, mouse and rat GrA degrade human ECM components (fibronectin and collagen type IV) (40–43). However, it is unclear to what extent these findings have physiological relevance. Also, it is unknown whether human GrA, GrH, GrK, and GrM degrade human ECM substrates.

Apart from their role in ECM degradation, a growing body of evidence now implicates GrA, GrB, GrK, and GrM in the production, release, and/or processing of proinflammatory cytokines. This immunomodulatory potential of Grs is discussed in the present review.

Studies in mice show a role for GrA in the inflammatory response to the Gram-negative bacterial cell wall component LPS. GrA^−/− mice better withstand a lethal LPS challenge in that they survive longer than do wild-type (WT) mice (11, 44). Furthermore, hGrA exerts direct effects on several cell types to induce cytokine release in vitro. Extracellular GrA stimulates release of proinflammatory cytokines IL-6 from fibroblast cell lines and IL-8 from fibroblast and epithelial cell lines (45). Additionally, extracellular GrA activates primary human monocytes to release IL-1β, TNF-α, IL-6, and IL-8 (11, 46). These effects of GrA are dependent on its catalytic activity, implying that at least one downstream signaling protein needs to be cleaved. The effects of GrA are augmented upon intracellular delivery (11), which suggests that GrA substrates necessary for cytokine induction localize in the cell. Until now, however, these substrates have not been identified. hGrA cleaves and activates pro–IL-1β in vitro and was therefore designated an IL-1β–converting enzyme (47), but the physiological relevance of this finding has never been shown. Moreover, some have failed to demonstrate such an effect of GrA (11). GrA-induced cytokine release from monocytes is blocked in the presence of caspase-1 inhibitors (11), further indicating that IL-1β maturation upon GrA treatment is not mediated by GrA directly, although the inflammasome is involved (Fig. 1). Finally, we have recently found that human recombinant GrA does not induce cytokine release from human monocytes, but synergistically potentiates the LPS-induced cytokine response by these cells (48). This is in line with the observation that mouse macrophages primed with LPS respond to mouse Gr (mGr)A and mGrK (10, 11). We are currently investigating the molecular mechanism underlying this GrA effect. Interestingly, hGrA blocks growth of intracellular mycobacteria in human macrophages (49). Human γδ T cells efficiently inhibit the growth of these bacteria inside macrophages. Recombinant GrA, as well as GrA produced by γδ T cells, induces TNF-α production in macrophages, which in turn inhibits the growth of the pathogen (49). The mechanism by which TNF-α exerts this effect remains unknown. In this regard, it would be interesting to investigate whether GrA, in addition to potentiating TLR4-mediated cytokine responses to LPS (48), also synergizes with other TLR ligands such as mycobacterial products to more efficiently activate monocytes and other innate immune cells.

Data on GrB knockout mice survival upon LPS challenge are conflicting. Anthony et al. (44) show that deletion of GrB has no effect on survival upon LPS challenge, whereas Metkar et al. (11) showed that GrB knockout results in a marked increase in survival that exceeds the effect of GrA deletion. Thus, the effect of GrB on the immune response to LPS remains to be elucidated. The reported differences in survival of GrB^−/− mice upon LPS challenge may relate to the age of the mice, the injected amount of LPS per gram of body weight, the source of LPS, or other differences in experimental setup.

Interestingly, Metkar et al. (11) reported a worse survival of GrAB double knockout mice compared with GrA or GrB single knockout mice upon LPS challenge (11). Anthony et al. (44), however, found no difference in survival of GrA^−/− mice and GrAB^−/− mice after LPS injection, consistent with their finding that GrB does not influence survival. GrB^−/− and GrAB^−/− mice were not directly compared in their experiments (44). The interplay between GrA and GrB in the immune response to LPS is unresolved and deserves further study. Possibly, different Grs are preferentially involved in response to different types of LPS.

In vitro evidence that GrB is involved in cytokine release during inflammatory responses is scarce. hGrB itself does not directly induce HeLa and HUVEC cells to produce IL-6 and IL-8 (50). However, we recently demonstrated that hGrB synergistically enhances LPS-induced TNF-α release from human monocytes in vitro, but it does not induce cytokine release from these cells on its own (48). This function of hGrB is shared with at least hGrK and hGrA (48), a redundancy consistent with a limited effect of mGrB on survival in response to LPS (44). A role for human GrB in the processing of proinflammatory cytokines also has been described. GrB cleaves pro–IL-18 in vitro and ex vivo, at the same position as caspase-1, although with slower kinetics (51, 52). Furthermore, hGrB cleaves the 31-kDa precursor of IL-1α, which enhances the biological activity of the cytokine several fold, an in vitro result that was confirmed in mice (50). Thus, it appears that extracellular GrB is able to cleave and activate several important proinflammatory cytokines. In conclusion, GrB may have more effect on cytokine cleavage and activation than on initial cytokine release.

Mice deficient in GrK have not been described yet. Hence, it is currently unknown whether GrK deficiency in mice results in altered responses to LPS challenges. The effects of GrK on different cell types in vitro, however, have been investigated. Similar to GrA (45), extracellular hGrK releases IL-6 and IL-8 from human lung fibroblasts, dependent on GrK proteolytic activity (48, 53). This effect of GrK involves cleavage of protease-activated receptor 1 (53). Additionally, recombinant mGrK, but not its catalytically inactive proform, induces release of IL-1β from peritoneal mouse macrophages (10), an effect also observed for mGrA (11). This effect is seen after priming the cells with LPS, which may indicate an additive or synergistic effect of Grs and bacterial compounds. Intracellular GrK delivery is required, although extracellular effects are observed at concentrations >600 nM (53). Cytotoxic T cells isolated from lymphocytic choriomeningitis virus–infected GrA/B-deficient mice, which mainly expressed GrK, also induce IL-1β release from mouse macrophages (10). We have recently shown that extracellular human GrK synergistically potentiates LPS-induced release of the proinflammatory cytokines TNF-α, IL-6, and IL-8 in human monocytes in vitro (48). Additionally, a combination of GrK and LPS enhances release of TNF-α, IL-1β and IL-6 in vivo, compared with LPS alone. This effect is independent of GrK proteolytic activity. GrK liberates individual LPS molecules from micelles and promotes complex formation between LPS and CD14 (48). Taken together, these data indicate that the proinflammatory effects of GrK are diverse and cell type specific. The proinflammatory effects of GrA, GrB, and GrK on different cell types are summarized in Fig. 1.

Upon LPS challenge, GrM^−/− mice survive longer and produce significantly lower serum levels of the proinflammatory cytokines IL-1α, IL-1β, TNF, and IFN-γ, compared with WT mice (44). Perforin knockout partially shows the same phenotype as GrM knockout (44), indicating that Grs may function intracellularly in the LPS response. However, recombinant mGrM does not cleave pro–IL-1β in vitro, and the molecular role of GrM in promoting responsiveness to an endotoxin challenge remains elusive (44). Consistent with results obtained with single knockout mice, GrM/GrA double knockout mice are even more resistant to LPS than are single GrM or GrA knockout mice (44). This additional effect of a double Gr knockout probably indicates that GrM and GrA function in the immune response to LPS via different mechanisms.

GrM colocalizes with the chemotactic protein MIP-1α in cytotoxic vesicles of human NK cells (54). GrM knockout leads to impaired MIP-1α secretion from NK cells and macrophages isolated from the liver of mice challenged with LPS or the Gram-positive bacterium Listeria monocytogenes (54). Because MIP-1α is important for NK cell recruitment to the liver during infection, the authors propose that GrM regulates this process (54). Similar to the response upon LPS challenge, GrM^−/− mice survive longer and exhibit reduced serum levels of cytokines following L. monocytogenes infection (54). Interestingly, however, MIP-1α serum levels were not different between WT and GrM^−/− serum levels after L. monocytogenes infection (54). Apparently, the local effect of GrM knockout on MIP-1α secretion is not reflected systemically. Taken together, GrM appears to be an important regulator of cytokine release in response to infection with Gram-negative as well as Gram-positive bacteria, at least in mice. However, in vitro data explaining the mechanism(s) behind these observations are lacking.

It is not known how Grs are released during inflammation. Leakage of Grs out of the immunological synapse during ongoing cytotoxic lymphocyte responses could explain their extracellular presence. Alternatively, Grs could be actively secreted during inflammation. GrA and GrB are released upon LPS injection into healthy human volunteers and upon incubation of whole-blood cultures with LPS or bacteria (19), suggesting that purposeful degranulation may occur in the absence of cytotoxicity. Alternatively, secretion of Grs may take place independently from degranulation, because significant amounts of GrA and GrB continue to be secreted from cytotoxic lymphocytes in the absence of continued stimulation by antigenic cells (55).

The source of extracellular Grs is also unknown. Cytotoxic lymphocytes may evidently produce these extracellular Grs. Anthony et al. (44) observed that NK cell depletion, similar to GrM deletion in WT mice, confers LPS resistance to RAG-1^−/− mice. Thus, NK cell–derived mGrM may play an important role in the inflammatory response to LPS (44). However, alternative sources of Grs also have to be considered. There is ample evidence that both human myeloid dendritic cells (mDCs) and human plasmacytoid DCs (pDCs) express GrB (56–60). Constitutive mRNA and protein expression has been found in pDCs (57, 60), whereas IL-3, IL-10, and IL-21 upregulate GrB mRNA and protein expression in these cells (57, 58, 60). In mDCs, no constitutive GrB protein expression is reported, but GrB and perforin protein expression can be induced by treatment with TLR7 and TLR8 agonists (59). In contrast, TLR7 and TLR9 agonists inhibit GrB protein expression in pDCs (57–60).

GrB expression by DCs may contribute to their cytotoxic potential in perforin-dependent as well as perforin-independent ways (61). Human mDCs upregulate perforin protein expression upon stimulation with TLR7 and/or TLR8 ligands and release perforin together with GrB (59). However, no such upregulation is found in pDCs stimulated with TLR7/8 ligands (59) or IL-3/IL-10 (58). GrB-expressing mDCs (59) and pDCs (56, 57, 59) eliminate endothelial cells and tumor cell lines. This cytotoxic activity may provide DCs with extra possibilities to regulate inflammatory processes. Interestingly, human immature DCs eliminate CD8⁺ T cells via a perforin- and GrA-dependent mechanism (62). Gr expression could therefore depend on the developmental stage of the DC, because mature pDCs do not express GrA (57). Human B cells also express GrB (63–65), and as for DCs, GrB expression is induced by IL-21, albeit in the context of other stimuli, including viral Ags (65). GrB from B cells was shown to inhibit T cell proliferation via the cleavage of the TCR ζ-chain (63), an extracellular and perforin-independent mechanism also shown for GrB derived from cytotoxic lymphocytes (66). Furthermore, GrB-secreting B cells possess cytotoxic potential in the absence of perforin expression (64), which is thought to contribute to their immunoregulatory potential (66). In contrast, Hagn et al. (67) could not detect GrB expression in resting or activated murine B cells.

GrB protein expression has also been found in human macrophages (68), basophils (69), mast cells (70), and several other cell types (66). GrB expression in mast cells and basophils was not accompanied by perforin expression (69, 70), indicating perforin-independent and possibly extracellular functions for GrB released by these cells. Further studies are required regarding the cellular origin of Gr in individual diseases, the regulation of Gr release, and the molecular immunoregulatory mechanisms Grs may employ.

Another issue to be clarified is whether Grs stimulate the release of previously synthesized cytokines stored inside the cell, or whether they exert a stimulatory effect on cytokine mRNA transcription and protein synthesis, or both. Not much is known about the pathways that are activated after extracellular Gr treatment. GrA can be internalized by monocytes in the absence of perforin, and it induces cytokine release (11), but the link between these events is unclear. Upon cleavage of protease-activated receptor 1 by GrK, ERK1/2 and p38 MAPK are activated (53), which then leads to cytokine release from fibroblasts. More research is needed to clarify which signaling pathways are activated by Grs to trigger cytokine responses. When Gr catalytic activity is required, identification of intra- or extracellular Gr substrates is essential. In cases where Grs potentiate pathways induced by TLR ligands independent of the Gr catalytic activity (48), the focus should lie on clarifying how Grs modulate these pathways via interaction with signaling molecules. A common characteristic in LPS-enhancing proteins such as LPS-binding protein (71, 72), Grs (48), the neutrophil granule protein azurocidin (73, 74), high mobility group box-1 protein (75, 76), protamines (77), and apolipoprotein C1 (78, 79) is the presence of cationic patches of arginines and/or lysines that drive the interaction with LPS. Endocytosis may also play a role, because internalization is required for the stimulating effect of azurocidin on LPS-induced cytokine responses (80). Interestingly, GrA binds to and is internalized by monocytes (11).

Another question is whether extracellular circulating Grs are active, or whether their activity is reduced due to binding by protease inhibitors. In vivo, the catalytic activity of Grs and other serine proteases is inhibited by serine protease inhibitors or serpins (81). Inhibition occurs through proteolytic attack of the serpin by the Grs, after which a covalent bond is formed between the serpin and the active site of Gr (81). Enzyme/serpin complexes are subsequently removed from the circulation by the liver. Removal is rapid and often occurs within minutes after complex formation (82, 83).

Several endogenous serpins for human Grs have been identified. Furthermore, several general protease inhibitors with the potential to inhibit the catalytic activity of Grs are present in the human circulation. GrA is inhibited by the serpin Kazal (84) and extracellular GrA is inhibited by anti-thrombin III and α₂-macroglobulin (85). The Gr is protected from inhibition when it is complexed to proteoglycans (85). GrB is inhibited by serpin B9, also known as proteinase inhibitor 9 (PI-9) (86). PI-9 is thought to operate intracellularly, but is also present in the circulation where it is complexed to GrB (87). Still, GrB activity is retained in the presence of 80% plasma (88), and therefore it is unclear whether PI-9 inhibits GrB activity in blood. GrB catalytic activity is not reduced by α₁-antitrypsin (88). The three major extracellular protease inhibitors present in normal lung fluid (α₁-antitrypsin, elafin, and secretory leukocyte protease inhibitor) do not inhibit GrA or GrB, and GrA is active in bronchoalveolar lavage fluid (20). GrH is inhibited by serpin B1 (89), and GrM is inhibited by serpin B4 (90). Furthermore, GrM is inhibited by α1-antichymotrypsin and α₁-proteinase inhibitor, although weakly (91). GrK catalytic activity is inhibited by inter–α-trypsin inhibitor (92). In sepsis patients, circulating amounts of inter–α-trypsin inhibitor are lower than in healthy controls (15, 93) Also, the molecular mass of GrK in the serum of sepsis patients is lower than in healthy controls. GrK mainly circulates with a molecular mass of ∼26 kDa in sepsis patients (15). This finding implies that GrK is freed from protease inhibitors under inflammatory conditions, although it was not determined whether this GrK is proteolytically active.

In summary, inhibition of catalytic activity of Grs may be reduced under inflammatory conditions, although this subject requires further study. Apart from intracellular inhibition of GrB by PI-9, the physiological relevance of serpins or other protease inhibitors to inhibit Gr activity remains unknown. Furthermore, it has to be considered that proteolytic activity may not be essential for all extracellular Gr functions (48). In this regard, steric hindrance by binding of inhibitors or proteoglycans could be of superior importance.

Increasing evidence now demonstrates that Grs are involved in ECM remodeling and modulation of inflammatory pathways. Grs may induce cytokine release from cells and/or cleave cytokines after cytokine release. Involvement of GrA, GrB, GrK, and GrM in these processes has been demonstrated in humans and/or mice (11, 44, 48–50, 53, 54). Most of these novel Gr functions are perforin-independent and extracellular, although effects may be enhanced upon intracellular delivery. Furthermore, Grs have immunomodulatory functions independent of their catalytic activity (48). Taken together, these data indicate that Grs have multiple regulatory functions in (innate) proinflammatory immune responses via many at least partly overlapping pathways (Fig. 1). This functional redundancy underlines the significance of Grs in immunity and ensures proper Gr function when one of the Grs or Gr-induced mechanisms is inhibited. The notion that multiple Grs influence inflammatory processes is reminiscent of the cytotoxic potential shared by several Grs, which is regarded as a safeguard against inhibition of one or more Grs (4).

Several important questions remain to be solved. First, not all Grs have been studied in the context of inflammation. It is unknown whether GrH and GrM possess direct cytokine-inducing and/or -processing capacity. Second, in many cases the intracellular mechanisms behind the effects of Grs on cytokine secretion have not been elucidated. Third, it is not understood how Gr secretion is regulated under inflammatory conditions and from which cell type(s) Grs originate. Finally, the regulation of Gr activity in the circulation also deserves further study. Elucidation of all these different aspects of Gr biology will help us to better understand the implications of Gr activity in disease. Once clear molecular mechanisms have been defined for Grs in inflammatory disease, and their physiological relevance has been demonstrated, Gr inhibition might in the future contribute to treatment of infectious diseases and autoimmune diseases.

The authors have no financial conflicts of interest.