Granzyme B Promotes Cytotoxic Lymphocyte Transmigration via Basement Membrane Remodeling

Immunity

Article Granzyme B Promotes Cytotoxic Lymphocyte Transmigration via Basement Membrane Remodeling Monica D. Prakash,1,7 Marcia A. Munoz,2,3,7 Rohit Jain,2,3 Philip L. Tong,2,3 Aulikki Koskinen,4 Matthias Regner,4 Oded Kleifeld,1 Bosco Ho,1 Matthew Olson,5 Stephen J. Turner,5 Paulus Mrass,2 Wolfgang Weninger,2,3,6,8 and Phillip I. Bird1,8,* 1Department

of Biochemistry and Molecular Biology, Monash University, Melbourne, VIC 3800, Australia Institute of Cancer Medicine and Cell Biology, Newtown, NSW 2042, Australia 3Discipline of Dermatology, Sydney Medical School, University of Sydney, NSW 2006, Australia 4Department of Emerging Pathogens and Vaccines, John Curtin School of Medical Research, College of Medicine, Biology, and Environment, Australian National University, Canberra, ACT 2600, Australia 5Department of Microbiology and Immunology, University of Melbourne, Melbourne, VIC 3010, Australia 6Department of Dermatology, Royal Prince Alfred Hospital, Camperdown, NSW 2050, Australia 7These authors contributed equally to the work 8These authors contributed equally to the work *Correspondence: [email protected] http://dx.doi.org/10.1016/j.immuni.2014.11.012 2Centenary

SUMMARY

Granzyme B (GzmB) is a protease with a well-characterized intracellular role in targeted destruction of compromised cells by cytotoxic lymphocytes. However, GzmB also cleaves extracellular matrix components, suggesting that it influences the interplay between cytotoxic lymphocytes and their environment. Here, we show that GzmB-null effector T cells and natural killer (NK) cells exhibited a cell-autonomous homing deficit in mouse models of inflammation and Ectromelia virus infection. Intravital imaging of effector T cells in inflamed cremaster muscle venules revealed that GzmB-null cells adhered normally to the vessel wall and could extend lamellipodia through it but did not cross it efficiently. In vitro migration assays showed that active GzmB was released from migrating cytotoxic lymphocytes and enabled chemokine-driven movement through basement membranes. Finally, proteomic analysis demonstrated that GzmB cleaved basement membrane constituents. Our results highlight an important role for GzmB in expediting cytotoxic lymphocyte diapedesis via basement membrane remodeling.

INTRODUCTION Cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells, collectively known as cytotoxic lymphocytes, eliminate foreign, infected, or neoplastic cells. A killing mechanism used by cytotoxic lymphocytes involves perforin-mediated delivery of proteases (granzymes) from the degranulating cytotoxic lymphocyte into the cytoplasm of the target cell, resulting in target cell death. Granzyme B (GzmB) is the best characterized granzyme: it is a widely used activation marker of both CTLs and NK cells and is a potent cytotoxin mediating target cell apoptosis, primarily 960 Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc.

via caspase activation. It has also been implicated in immune regulation via regulatory-T-cell-mediated suppression of activated cells and elimination of T helper (Th) cells by activationinduced cell death (Cao et al., 2007; Devadas et al., 2006). Recent evidence suggests that granzymes have broader roles (Joeckel and Bird, 2014). For example, GzmB is present in many noncytotoxic immune cell types (Table S1, available online), its relative importance in direct target cell killing in vivo has been questioned (Regner et al., 2009), and a role in proinflammatory signaling has been suggested (Afonina et al., 2011). In addition, several lines of evidence suggest that GzmB functions independently of perforin outside the cell (Hiebert and Granville, 2012). Notably, extracellular GzmB occurs in the circulation and increases during infection, inflammation, allergy, or autoimmune disease. This soluble GzmB is most likely the result of constitutive release by NK cells and CTLs in the absence of target cell engagement (Prakash et al., 2009) or stimulated release on binding to extracellular matrix (ECM) proteins (Takahashi et al., 1991). There are at least ten documented GzmB substrates from the ECM (Hendel et al., 2010). Given that proteolysis of ECM components is important in leukocyte trafficking (Leppert et al., 1995), we and others have proposed that GzmB potentially plays a role in cytotoxic lymphocyte migration (Buzza and Bird, 2006; Buzza et al., 2005; Kramer and Simon, 1987). The studies we report here showed that GzmB contributed to cytotoxic lymphocyte extravasation and homing in vivo without affecting interstitial motility. In vitro studies showed that it played a role in basement membrane remodeling by cytotoxic lymphocytes via ECM cleavage. A role for extracellular GzmB in cell trafficking might partly explain its expression by other leukocyte types and nonimmune cells, as well as its association with several immune pathologies linked to ECM degradation. RESULTS Impaired Homing of GzmB-Deficient Cytotoxic Lymphocytes during Virus Infection We infected mice with the attenuated natural pathogen Ectromelia virus Hampstead egg (ECTV-HE) to examine CTL and NK cell

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

recruitment in Gzmb / mice. We focused on the peritoneum, given that immune cells must enter the cavity from the vasculature via diapedesis and crossing the squamous epithelium lining the peritoneal wall. Wild-type (WT) or Gzmb / mice were infected intraperitoneally (i.p.), and spleens and peritoneal exudates were collected and analyzed postinfection. Numbers and proportions of NK cells and CTLs in peritoneal exudate cells (PECs) were compared to splenic populations, which acted as controls. As indicated by analysis of total CD8+ T cells or a subset of virus-specific CTLs marked by EV-Kb tetramer (Regner et al., 2009), peritoneal exudates from WT mice showed substantial accumulation of CTLs by 6 days after infection, but this was completely abrogated in Gzmb / mice (Figure 1A). No difference was observed in the proportion or number of splenic CTLs, suggesting that the decrease in the peritoneum was a result of defective recruitment. A recruitment deficit was also observed for NK cells (Figure 1B). Spleens from WT and Gzmb / mice contained similar numbers of splenocytes, as well as similar total and proportional numbers of NK cells at all time points. However, although there was a substantial increase in peritoneal NK cells in WT mice at days 4 and 6, only a slight increase was observed in Gzmb / mice. Thus, both CTLs and NK cells lacking GzmB appear unable to efficiently extravasate in the context of viral infection. To determine whether this recruitment deficit is cell autonomous, we adoptively transferred NK cells from WT or Gzmb / mice into WT Ptprca congenic recipient mice (Figure 1C). In this normal environment, Gzmb / NK cells still displayed defective recruitment to the peritoneal cavity during infection, showing that the defect is intrinsic to the migrating cell and does not result from altered external stimuli. GzmB-Deficient CTLs Exhibit Decreased Homing in Models of Inflammation The Gzmb / cytotoxic lymphocyte homing defect observed during ECTV-HE infection was also observed in models of nonantigen-specific inflammation (Figure 2). In the mechanical tape-strip assay (Shulman et al., 2011), activated WT ovalbumin-tcr-I (OT-I) or Gzmb / OT-I donor CTLs were adoptively transferred into Ptprca recipient mice after local inflammation had been induced in one ear. We used spleen as the control organ to account for the number of systemic OT-I cells and compared the noninflamed ear to the irritated ear to ensure that an inflammatory response had occurred. The ratio of donor cells in the ear to donor cells in the spleen showed significantly less Gzmb / CTL homing than WT cell homing to the inflamed ear (Figure 2A). We also adapted the contact-hypersensitivity model (Gaspari and Katz, 1991) to examine the recruitment of WT and Gzmb / CTLs in vivo. Lymphocytes from either WT or Gzmb / mice were labeled and adoptively transferred into sensitized WT recipient mice. A local inflammatory response was induced chemically, and recruitment of labeled lymphocytes to the inflamed site (ear) was compared to that of control tissue (untreated ear). As shown in Figure 2B, compared to recruitment of control cells, recruitment of Gzmb / lymphocytes to the inflamed ear was significantly impaired. To examine NK cell recruitment during inflammation, we differentially labeled naive WT and Gzmb / NK cells, coadoptively transferred them into WT recipients, and peritoneally injected

them with poly(I:C). Using accumulation of donor cells within the spleen as a comparison, we examined peritoneal exudate for labeled donor NK cells. Significantly fewer Gzmb / NK cells than WT cells were recovered in the peritoneal cavity (Figure 2C). Similar results were obtained with cells from animals lacking GzmA, GzmB, and GzmM (Gzma / Gzmb / Gzmm / ; Figure 2C), suggesting that GzmA and GzmM do not contribute to cytotoxic lymphocyte recruitment (see also Figure 4Aii). GzmB Contributes to CTL Transmigration in Postcapillary Venules To directly visualize the behavior of Gzmb / CTLs within inflamed venules, we performed intravital imaging in a sterile inflammation model (Shulman et al., 2011). Adoptively transferred WT and Gzmb / CTLs were covisualized in dermal vessels of the ear. WT and Gzmb / CTLs showed similar rolling fractions and mean rolling velocities (Figures 3Ai and 3Aii; Figure S1 and Movie S1, available online), suggesting similar engagement with E- and P-selectins on the endothelium. Furthermore, there was no significant difference in firm adhesion between the two populations (Figure 3Aiii), indicating identical binding of integrins on CTLs to their respective endothelial ligands. We did not observe transmigration of CTLs during 30 min live imaging sessions. Nevertheless, whole-mount stains from ear skin samples obtained 6 hr posttransfer revealed fewer Gzmb / than WT CTLs in the dermis, suggesting that GzmB expression facilitates lymphocyte exit into the interstitium (data not shown). To directly investigate the role of GzmB CTLs in transmigration, we imaged differentially labeled Gzmb / and WT CTLs in inflamed cremaster muscle by multiphoton microscopy (Hyun et al., 2012). This model allows precise application of proinflammatory mediators and covisualization of differentially labeled CTLs in the same postcapillary venules. We found that extravasation of CTLs was rare but invariably followed prolonged adherence of the cells to the endothelium. To capture sufficient extravasation events for statistical comparison, we therefore recorded long 3D sequences of single fields for up to 6 hr (Figure 3B; Movies S2 and S3). Our imaging revealed that WT CTLs extravasated in significantly higher numbers than Gzmb / CTLs (extravasation fractions: 0.48 for WT versus 0.11 for Gzmb / , p = 0.0064; Figure 3B). We also observed that once penetration started, most WT CTLs smoothly and rapidly translocated through the vessel wall (Figure 3Bi; Figure S2A; Movie S2). By contrast, Gzmb / CTLs usually remained firmly attached to the vasculature wall; they sometimes projected lamellipodia into the extravascular space but ultimately retreated into the lumen (Figure 3Bii; Figure S2A; Movie S3). Furthermore, transmigration appeared to take longer for Gzmb / than for WT cells (average 77 min for Gzmb / versus 28 min for WT), although the small number of extravasation events recorded for Gzmb / CTLs did not allow statistical comparison (Figure S2C). Together, these data demonstrate that GzmB expression by CTLs is dispensable for rolling and firm adhesion but plays an important role in crossing the vessel wall, the final step in diapedesis. GzmB Is Not Required for Interstitial Migration To investigate the contribution of GzmB to interstitial movement of cytotoxic lymphocytes, we compared the motility of WT and Gzmb / CTLs in a tumor microenvironment by using multiphoton Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc. 961

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

(legend on next page)

962 Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc.

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

Figure 2. Deficient Recruitment of Gzmb–/– Cytotoxic Lymphocytes during Inflammation (A) Homing of activated WT or Gzmb / OT-I CTLs in a tape-strip inflammation model. Shown is the ratio of donor cells in the tape-stripped ear to donor cells in spleen 24 hr after stimulation. The graph shows mean ± SD of data from two separate experiments. (B) Homing of naive WT or Gzmb / lymphocytes in a contact-hypersensitivity inflammation model. Shown is the ratio of donor cells in the 2,4,6-trinitrochlorobenzene (TNCB)-treated ear to donor cells in the untreated ear 24 hr after elicitation. The graph shows mean ± SE of data pooled from eight experiments. (C) Homing of WT or Gzma / Gzmb / Gzmm / NK cells in response to poly(I:C). Shown is the ratio of donor cells recovered from the peritoneal cavity to donor cells in spleen 48 hr after poly(I:C) administration. The graph shows mean ± SE of data pooled from three experiments.

time-lapse microscopy. As shown in Figure 3Ci, there was no difference in tumor homing between WT and Gzmb / CTLs, which is consistent with the compromised, and hence more permissive, nature of barriers between vascular and tumor tissue (Hashizume et al., 2000). We analyzed movies of motile CTLs to measure cellular mean speed and displacement (Figures 3Cii and 3Ciii; Movies S4 and S5). No differences were observed between WT and Gzmb / CTLs in any of these aspects of interstitial migration, suggesting that GzmB is not required to remodel the extravascular ECM to promote CTL motility. GzmB Contributes to Migration of Human CTLs through Basement Membrane In Vitro To gain insight into the molecular and cellular aspects of GzmBmediated transmigration, we performed in vitro migration assays. Primary human peripheral-blood CTLs placed in the upper compartment of a Boyden chamber (Transwell) transmigrated freely through the basal filter toward the chemokine SDF-1a, which was added to the lower chamber (Zhang et al., 2006). Progress was impeded if Matrigel, a widely used model basement membrane (Benton et al., 2011; Kleinman and Martin, 2005), was used to coat the filter. Progress was further decreased if cells were pretreated to inhibit either granule release (degranulation) or GzmB activity (Figure 4Ai). The degranulation inhibitor concanamycin A (CMA) reduced chemokine-directed CTL transmigration through Matrigel by 43%. Compound-20 (C-20), the specific cell-permeable inhibitor of human GzmB (Willoughby et al., 2002), caused a 57% decrease in transmigration, which correlated well with the 60% inhibition of cellular GzmB enzymatic activity caused by C-20 pretreatment (data not shown). Thus, GzmB contributes to CTL transmigration in vitro. Migration of Primary Murine CTLs through Basement Membrane Is Facilitated by Both Granular and Extragranular GzmB Using cells from three different granzyme-deficient mouse strains (Gzma / , Gzmb / , and Gzma / Gzmb / ), we showed

that CTLs had a pronounced defect in transmigration through Matrigel in the absence of GzmB, but not in the absence of the closely related protease GzmA (Figure 4Aii). Use of Matrigel stripped of growth factors and other soluble components did not alter the result, suggesting that cytotoxic lymphocyte migration does not involve GzmB-mediated activation or release of ECM-embedded signaling molecules (data not shown). Importantly, in the absence of basement membrane, there was no defect in chemokine-driven motility of Gzmb / CTLs (Figure 4Aiii). Jinx mutant mice possess a truncated, nonfunctional version of UNC13D, which is an important component of the granule exocytosis machinery. Thus, degranulation in Jinx lymphocytes is severely compromised (10% of WT cells; Crozat et al., 2007). We showed that analogous to transmigration of CMAtreated human CTLs (Figure 4Ai), transmigration of Jinx CTLs through basement membrane decreased by half in comparison to WT CTLs (Figure 4Aii). Together, these data support the idea that GzmB facilitates CTL migration via basement membrane degradation in vitro. CTLs Moving through Basement Membrane Release Active GzmB Granzymes stored in granules are quiescent (inactive) as a result of the acidic lumenal environment. Upon receptor stimulation, they are released from granules and gain activity in a process involving granule mobilization and polarization followed by fusion with the plasma membrane (de Saint Basile et al., 2010). Granzyme release from static CTLs in response to purified ECM components has been reported (Takahashi et al., 1991), but we sought to examine whether this also occurs in cells moving on or through basement membrane via degranulation. Mouse CTLs were placed on Matrigel, and the distribution of GzmB-containing granules was assessed (Figure 4B). Granules were marked by a GzmB-mCherry fusion protein (Bird et al., 2010) introduced by retroviral transduction. GzmB-containing granules polarized toward the cell-matrix interface,

Figure 1. Gzmb–/– Mice Exhibit a Cell-Autonomous Cytotoxic Lymphocyte Recruitment Defect in Response to Ectromelia Virus WT or Gzmb / mice (n = 3–4) were infected i.p. with ECTV-HE, and spleens and PECs were collected postinfection (p.i.). (A and B) CTL (A) and NK cell (B) recruitment to the peritoneum was measured as CD8+EV-Kb+ and NK1.1+ cells present in peritoneal exudate as compared to in the spleen. (C) Donor WT and Gzmb / NK cells were adoptively transferred to congenic recipients (n = 9–10) prior to ECTV-HE infection. Donor cells in PECs and spleen were assessed 4 days postinfection. The graph shows mean ± SE of data from three experiments.

Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc. 963

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

Figure 3. GzmB Facilitates CTL Extravasation, but Not Interstitial Migration (A) Interaction of adoptively transferred WT OT-I and Gzmb / OT-I CTLs with vessel walls was measured by intravital microscopy of inflamed ear venules (see also Figure S1 and Movie S1). (i) Mean T cell rolling velocity. Each dot represents a single event; 50 events were analyzed per group. (ii and iii) Rolling fraction (ii) and sticking fraction (iii). Each dot represents data from a single mouse (n = 3 mice). Statistical analysis was performed with a Mann Whitney two-tailed t test (n = 3 mice). n.s. indicates ‘‘not significant.’’ (B) CTL extravasation was visualized by intravital microscopy of inflamed cremaster muscle. The numbers of events analyzed and the corresponding extravasation fractions for each mouse are in Figure S2A. (i) 3D time-lapse images of differentially labeled WT (green) and Gzmb / (red) CTLs moving in a postcapillary venule (gray) highlighted by Evans Blue staining (endothelial cells are indicated with dotted lines). The scale bar represents 10 mm. The upper row shows an extravasating WT CTL (see also Movie S2), and the lower row shows a Gzmb / cell that extends its front into the extravascular space but then retracts into the vascular lumen (see also Movie S3). (ii) Proportions of WT and Gzmb / cells extravasating within individual animals (populations originating from a particular animal are joined by a line). (iii) Average extravasation indexes. Statistical analysis was performed with a Mann Whitney two-tailed t test (n = 7 mice). See also Figure S2. (C) Interstitial migration of CTLs. (i) WT or Gzmb / OT-I CTLs were transferred into tumor-bearing mice (n = 5). Tumors were explanted 2–3 days posttransfer, and CTLs were observed via two-photon microscopy. Shown are CTL numbers in tumors 3–4 days posttransfer in relation to splenic CTLs. (ii and iii) Mean speed (ii) and displacement (iii) between WT and Gzmb / CTLs were measured (n = 700–800 cells were examined over four separate experiments, each with at least three tumors per group). See also Movies S4 and S5.

964 Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc.

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

mainly in the uropod, suggesting directed release of granule contents toward the cell-matrix interface (Figure 4B, gray arrowheads). CTLs embedded in Matrigel spontaneously migrate (Leppert et al., 1995). By incorporating a diffusible quenched fluorescent GzmB peptide substrate into the matrix, we demonstrated that foci of active GzmB were mainly in the uropod and midbody of embedded migrating cells (Figure 4Ci). Up to five foci per WT cell, far fewer than the total number of granules, were apparent at any one time, indicating an ordered process involving sequential granule fusion and ongoing release of GzmB (Figure 4Cii). By contrast, GzmB-deficient cells exhibited very low, nonpunctate background substrate cleavage always associated with membrane blebs from the uropod (Figure 4C). This most likely resulted from caspase-8, which has a similar specificity to GzmB and can also cleave this peptide substrate. To confirm that granzymes are released from CTLs in response to ECM contact, we cultured cells on either uncoated or Matrigel-coated surfaces and assessed cell lysates and culture supernatants for GzmA (benzyloxycarbonyl-L-lysine thiobenzyl [BLT] esterase) activity. We assessed total stored and releasable BLT esterase by treating controls with the degranulating agents phorbol myristate acetate (PMA) and calcimycin. There was a small but significant increase in extracellular BLT esterase activity from CTLs on Matrigel (Figure 4D), indicating that granzymes are released in response to the ECM. Importantly, Jinx CTLs, which cannot efficiently degranulate (Crozat et al., 2007), showed no stimulation of BLT esterase release by Matrigel (Figure 4D). GzmB-Dependent Basement Membrane Remodeling by Migrating Cytotoxic Lymphocytes We used live cell confocal imaging and scanning electron microscopy (SEM) to further study the interaction of migrating CTLs with basement membrane. Cells moving across Matrigel altered the surface and left a visible trail, suggesting that the matrix had been modified (Figure 5A; Movie S6). If embedded in the matrix and induced to move along a chemokine gradient, the cells burrowed to form two distinct types of spatial structures, which we termed ‘‘tunnels’’ and ‘‘nests’’ (Figure 5Bi; Movie S7). Gzmb / CTLs formed fewer tunnels and appeared in nests more often than WT CTLs (Figure 5Bii). The burrows created by Gzmb / CTLs were also significantly smaller in volume than those created by WT CTLs (Figure 5Biii). Viewed by SEM, CTLs placed on matrix over a source of chemokine exhibited morphological characteristics of migrating cells (e.g., polarization; Figure 5Ci) and appeared to clear the surrounding matrix and/or burrow under it (Figure 5Cii). Fewer Gzmb / CTLs exhibited clearing, although cell polarization and burrowing were unaffected (Figure 5Cii). Taken together, these results indicate that migrating cytotoxic lymphocytes use GzmB to alter (remodel) basement membrane structure. GzmB Cleaves Basement Membrane Components Matrigel is the ‘‘gold-standard’’ basement membrane model (Benton et al., 2011; Kleinman and Martin, 2005). Viewed by SEM, Matrigel is a dense, fibrillar structure (Abrams et al., 2000; Van Goethem et al., 2010) that covers the pores in the Transwell membrane and must be cleared or remodeled to

allow cells into the lower chamber (Figure 6Ai). However, a formal possibility remains that cytotoxic lymphocytes might not use GzmB to traverse a basement membrane synthesized by cells and assembled in situ. We therefore followed established procedures (Ferrell et al., 2011; Tammi et al., 2000) to allow monolayers of epithelial Madin-Darby canine kidney (MDCK) cells to produce basement membrane in Transwells. As shown in Figure 6Ai, MDCK-cell-derived basement membrane is structured differently from Matrigel in that it forms a dense continuous sheet. Nevertheless, SEM analysis of MDCK cell and Matrigel basement membrane in Transwells before and after exposure to CTLs showed that the integrity of both was significantly disrupted by CTLs, which exposed the underlying filter (Figure 6Ai). Importantly, Gzmb / CTLs also exhibited a transmigration deficit on MDCK cell basement membrane (Figure 6Aii), further supporting the idea that GzmB facilitates cytotoxic lymphocyte migration through basement membranes in vivo. Given that GzmB probably contributes to cytotoxic lymphocyte migration via basement membrane modification, we investigated whether ECM components are cleaved by GzmB. We found that exposure to purified recombinant GzmB disrupted Matrigel or MDCK cell basement membrane on Transwell filters (Figure 6Bi). Membranes were less affected by exposure to GzmB inactivated by an alteration in its active site (mGzmBS203A), consistent with its low residual activity (data not shown). Pretreatment of Matrigel in Transwells with recombinant GzmB also enhanced transmigration of Gzmb / CTLs at a rate comparable to that seen in the absence of matrix (Figure 6Bii). GzmB-mediated Matrigel degradation was also demonstrated by a proteomics approach. Matrigel was set and acetylated to block free N termini and treated with GzmB; then, newly exposed N termini were detected. Numerous cleavage products were evident upon GzmB treatment (Figure 6C). Although it is known that GzmB can cleave isolated ECM proteins (Hendel et al., 2010), this experiment demonstrates cleavage of proteins in the context of a complex 3D matrix. We then used the Protein Topography and Migration Analysis Platform (PROTOMAP) to identify specific matrix proteins that are directly or indirectly cleaved by GzmB (Dix et al., 2008). PROTOMAP identified 300 Matrigel proteins, many of which are cleaved by GzmB (Table S2). In addition to including the major ECM constituents, Matrigel also contains a large number of contaminating intracellular proteins (Hansen et al., 2009; Hughes et al., 2010). Consistent with the documented ability of GzmB to cleave a wide range of cytosolic substrates (Van Damme et al., 2009), intracellular proteins composed the majority of GzmB targets identified. Nevertheless, ten of the proteins cleaved by GzmB are known ECM components, and in four a putative GzmB cleavage site could be readily identified. These substrates are shown in Table 1 and are ordered by decreasing similarity of the putative mouse GzmB cleavage site to the consensus sequence motif (Kaiserman et al., 2006; Van Damme et al., 2009). Fibrinogen-b has previously been reported as a substrate of GzmB (Buzza et al., 2008), and the indicated cleavage site has been demonstrated by N-terminal sequencing. These results suggest that GzmB remodels basement membrane during transmigration by cleaving structural components. Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc. 965

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

Figure 4. GzmB Expedites Cytotoxic Lymphocyte Migration through Model Basement Membrane In Vitro and Is Released from Granules of Mobile Cells (A) (i) Transmigration of human CTLs treated with CMA or C-20 to inhibit GzmB release or activity, respectively. Shown is mean ± SE of data from five experiments. (ii) Transmigration of Gzma / , Gzmb / , Gzma / Gzmb / , or Jinx CTLs. Shown is mean ± SE of data from more than three experiments. (iii) Transmigration of WT and Gzmb / CTLs in the absence of Matrigel. Shown is mean ± SE of data from two experiments. (B) Representative images of GFP (green)- and mCherry-GzmB (magenta)-expressing CTLs migrating on CellTrace Violet-labeled Matrigel (blue). Localization of GzmB-containing granules is viewed from the top (i and ii) and from the side (iii). White arrowheads indicate lamellipodia (lamell.; leading end of the cell), and gray arrowheads indicate the uropod (lagging end). The scale bar represents 7 mm. (legend continued on next page)

966 Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc.

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

DISCUSSION It is emerging that granzymes do more than destroy cells (Joeckel and Bird, 2014). We have shown here that the key cytotoxin GzmB has an additional function as an important contributor to cytotoxic lymphocyte extravasation and hence homing. This is consistent with the accepted role of proteolysis in leukocyte extravasation, which can also involve neutrophil elastase, matrix metalloproteases (e.g., MMP-1, MMP-2, MMP-8, MMP9, and MMP-13), and cathepsins B, G, K, and L. Taken together, our data suggest that extracellular-GzmBmediated cleavage and/or release of basement membrane components (remodeling) assists cytotoxic lymphocytes to cross the endothelium during transmigration. It is unlikely that GzmB is required for initial penetration of the vessel wall, given that GzmB-null CTLs can extend lamellipodia through it and retain the ability to extravasate, albeit very poorly. Rather, it is likely that remodeling by GzmB facilitates passage of the cytotoxic lymphocyte body—containing the relatively rigid nucleus— through the breach. (By contrast, neutrophils with their multilobular nuclei can presumably squeeze through smaller holes and require little or no basement membrane remodeling after penetration.) That GzmB is not required for CTL homing to tumors or for interstitial motility can be explained by (1) the abnormal structure of tumor vasculature, which is more permissive to CTL extravasation (Boissonnas et al., 2007; Hashizume et al., 2000), and (2) the distinct, amoeboid-like lymphocyte movement within tissues, which does not rely on ECM proteolysis (Weninger et al., 2014). Previous studies showing ECM cleavage by GzmB have generally used purified individual protein substrates in soluble or matrix conformations (for example, see Buzza et al., 2005). We have extended these observations by demonstrating GzmB-mediated cleavage of complex, intact basement membrane and by identifying GzmB substrates within it. The presence of substrates lacking strong GzmB cleavage motifs can be explained by (1) GzmB-mediated activation of other proteases within Matrigel via zymogen conversion or inhibitor degradation or (2) exosite GzmB-substrate interactions, which allow cleavage at an ostensibly suboptimal site, as previously documented for GzmB-mediated cleavage of vitronectin (Buzza et al., 2005) and supported by subsequent proteomics studies (Plasman et al., 2011; Van Damme et al., 2009). Cytotoxic lymphocytes unable to degranulate are impaired in migration but to a lesser extent than GzmB-null cells. This discrepancy suggests that some active GzmB is released via an alternative pathway or that zymogen GzmB released by the default secretory pathway is activated extracellularly. Previously, we showed that human and mouse CTLs constitutively release both active and zymogen GzmB in the absence of targets via degranulation and the default secretory pathway (Prakash et al., 2009). We have also shown that in addition to cathepsin

C, which normally activates GzmB, cathepsin H also possesses proGzmB convertase activity, illustrating that other activators of GzmB exist (D’Angelo et al., 2010). This, together with the data presented here, supports the existence of an extragranular pathway to activation of zymogen GzmB. Recent studies have revealed proinflammatory roles for GzmA, GzmK, and GzmM, but not GzmB, via cytokine modulation following lipopolysaccharide challenge (reviewed in Joeckel and Bird, 2014). It has also been reported that GzmB (along with calpain-1, elastase, and mast cell chymase) cleaves the alpha subunit of interleukin-1 (IL-1) to enhance proinflammatory effects (Afonina et al., 2011). This can occur either intracellularly during target cell killing by cytotoxic lymphocytes or as a consequence of IL-1 released by necrotic cells encountering GzmB in the extracellular space. Therefore, there is a formal possibility that GzmB might also enhance CTL migration via cytokine modulation. We consider this unlikely given that our infection, inflammation, and tumor models involved adoptive transfer of WT and Gzmb / cytotoxic lymphocytes into WT animals. Because we noted no difference in efficiency of recruitment of WT or Gzmb / CTLs to tumors, we conclude that GzmB produced by CTLs does not modulate cytokines originating from tumor or other cells to amplify CTL recruitment. In closing, our finding that extracellular GzmB was released by infiltrating cytotoxic lymphocytes to expedite transmigration through basement membrane provides a simple explanation for several long-standing and puzzling observations: (1) the increase in circulating GzmB in several inflammatory and infectious disease states (Buzza and Bird, 2006), (2) the links between GzmB and atherosclerotic lesion severity (Hendel et al., 2010) and between GzmB and impaired wound healing (Hiebert et al., 2013), and (3) the expression of GzmB by other noncytotoxic immune cell types, nonimmune cells, and neoplastic cells. It could be that GzmB is used by a variety of cell types to assist basement membrane remodeling. EXPERIMENTAL PROCEDURES Mice All mice were on the C57BL6/J background. Gzma / , Gzmb / , Gzma / Gzmb / , and Gzma / Gzmb / Gzmm / lines were from the Peter MacCallum Cancer Centre or the Animal Services Division of Australian National University. OT-I strains were bred in house at the Centenary Institute. Jinx mice were from the Mutant Mouse Regional Resource Center (University of California, Davis), and Ptprca mice were from the Animal Resources Centre (Canning Vale) or the Walter and Eliza Hall Institute (Melbourne). Albino C57BL/6 (C57BL/6-Tyrc-2J) mice were from the Jackson Laboratory. All experiments were conducted under the appropriate animal ethics regime. Cell Culture Human CTLs were isolated from peripheral-blood mononuclear cells from informed human volunteers (Monash University Standing Committee on Ethics in Research Involving Humans project no. CF07/0849). CTLs were isolated via magnetic-activated cell separation (MACS, Miltenyi Biotec) based on CD8

(C) GzmB activity is associated with mobile cytotoxic lymphocytes. (i) Representative images of mouse WT and Gzmb / CTLs migrating through Matrigel containing the diffusible quenched fluorescent GzmB substrate z-IETD-R110. The scale bar represents 10 mm. Single-cell images were taken as 3D stacks and are displayed as consecutive z planes 1 mm apart. (ii) Scoring of GzmB release by migrating CTLs, as indicated by cell-associated fluorescent puncta arising from pericellular substrate cleavage (an arrow points to an example in i). Fluorescent signal (low versus high) was assigned in relation to background intensity. (D) Degranulation of WT or Jinx CTLs in response to Matrigel. Data are expressed as a percentage of total BLT esterase cleavage in cell lysate and supernatant. Shown is mean ± SE of data from three experiments.

Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc. 967

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

Figure 5. GzmB Facilitates Cytotoxic Lymphocyte Migration via Basement Membrane Remodeling (A) Snapshots from a representative live cell movie (Movie S6) of a GFP-expressing CTL moving across fluorescently labeled Matrigel show remodeling of the ECM surface. (B) (i) Representative 3D reconstructions from two experiments of characteristic spaces, termed ‘‘tunnels’’ and ‘‘nests,’’ created by cytotoxic lymphocytes moving through Matrigel in response to CXCL10. (ii) Proportion of WT and Gzmb / CTLs showing tunnels or nests (see also Movie S7). (iii) Quantitation of burrows generated by individual cells. (C) (i) SEM of cytotoxic lymphocytes cultured on Matrigel in the presence of a CXCL12 gradient and quantitation of cytotoxic lymphocyte polarization. (ii) Scoring of matrix clearing and burrowing by cytotoxic lymphocytes on Matrigel-CXCL12 via SEM. Scale bars represent 15 mm. expression after CD56+ cell depletion. CTLs were cultured in complete RPMI1640 medium (10% heat-inactivated fetal calf serum [FCS], 2 mM glutamine, 50 U/ml penicillin, 50 mg/ml streptomycin, 90 mM mercaptoethanol, 0.1 mM MEM nonessential amino acids, and 1 mM sodium pyruvate; Gibco) with

968 Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc.

5 mg/ml concanavalin A and 100 U/ml IL-2 for 72–120 hr. CTLs were 70.11% ± 13.90% CD3+ and 74.14% ± 8.01% CD8+ (data not shown). GzmB secretion and activity were inhibited by 4 hr pretreatment with 100 ng/ml CMA and 100 mM C-20, respectively, at 37 C.

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

Figure 6. GzmB Facilitates Migration through Natural Basement Membrane and Cleaves Basement Membrane Components (A) (i) SEM of Matrigel and MDCK-cell-derived basement membrane before and after WT CTL passage. (ii) Transmigration of WT and Gzmb / CTLs through MDCK cell basement membrane. The graph shows mean ± SE of data from six separate experiments. The scale bar represents 3 mm (high magnification) or 15 mm. (B) (i) SEM of Matrigel and MDCK-cell-derived basement membrane treated with 100 nM recombinant mouse GzmB (mGzmB) or active-site mutant GzmB (mGzmBS203A). White arrows indicate pre-existing pores no longer obscured by matrix in the Transwell membrane. (ii) Migration of Gzmb / CTLs through Matrigel treated with the given concentrations of exogenous recombinant GzmB. Graph shows mean ± SE of data from five separate experiments. The scale bar represents 3 mm. (C) Matrigel was N-terminally dimethylated and exposed to 100 nM recombinant mouse GzmB. Neo-N-termini were biotinylated, separated by SDS-PAGE, and visualized by immunoblotting. Murine CTLs were derived from splenocytes via MACS selection of CD8+ cells and were cultured as described for human CTLs for 48–72 hr. Alternatively, WT OT-I and Gzmb / OT-I CTLs were stimulated in vitro with OVA257-264 (SIINFEKL) peptide for 2 hr and cultured for 5–6 days prior to use. MDCK cells were cultured in complete Dulbecco’s modified Eagle’s medium (10% FCS, 2 mM glutamine, 50 U/ml penicillin, and 50 mg/ml streptomycin; Gibco). For generation of basement membranes, confluent MDCK cell monolayers were maintained in Transwells for 18–22 days and then removed while the basement membrane was left intact, as previously described (Tammi et al., 2000). Ectromelia Infection Attenuated ECTV-HE was prepared as previously described (Wallich et al., 2001). Four to five mice per group were infected i.p. with 1 3 103 plaque-forming units of ECTV-HE in 100 ml PBS (Xu et al., 2004). At given days postinfection, mice were killed, and spleens and peritoneal exudates were isolated. CD8+ and EV-Kb tetramer+ populations were analyzed by flow cytometry. For NK cell adoptive-transfer experiments, NK cells were isolated from splenocytes via MACS based on NK1.1 expression. Twenty-four hours prior to infection, 1 3 106 to 3 3 106 freshly isolated donor NK cells were adoptively trans-

ferred into Ptprca congenic recipient mice. Recipient mice were subsequently infected as described above. Tape-Strip Assay Ptprca congenic recipient mice were anesthetized with ketamine-xylazine (80/ 10 mg/kg), hair was removed from the ears by depilation, and the dorsal side of the ear was tape stripped (Shulman et al., 2011). Mice were left to recover from anesthesia for 24 hr before adoptive cotransfer of 2 3 107 WT OT-I and 2 3 107 Gzmb / OT-I CTLs (for discriminating between populations, one population was labeled with CMTMR, and the labeled population was alternated in successive experiments). Mouse spleens and ears were harvested 24 hr posttransfer, and cells were isolated from the ear skin as previously described (Sumaria et al., 2011). Cell suspensions were stained with fluorochrome-conjugated antibodies against B220, CD8, CD11b, CD45, CD45.2, and CD90.2 (BD Biosciences) for the identification of donor cells and were analyzed by flow cytometry. Contact-Hypersensitivity Assay This assay was adapted from Gaspari and Katz, 1991. Naive WT or Gzmb / lymphocytes were labeled with either PKH-26 or CellTracker-Orange CMRA

Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc. 969

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

Table 1. GzmB-Directed Cleavage of Basement Membrane Proteins in Matrigel Accession Number

Protein

Description

Putative mGzmB Cleavage Site

Q08879-2

fibulin-1C

matrix-associated glycoprotein

393 FYFD

Q3UQ28

peroxidasin

basement membrane component

1273 ILCD

Q61292

laminin beta-2

basement membrane component

264 NLLD

Q8KOE8

fibrinogen beta

blood coagulation factor

227 IQPD

P07214

SPARC, osteonectin, BM-40

ECM scaffold

no putative mGzmB cleavage site found at event; position 220

P08122

collagen IV alpha-2

basement membrane component

no putative mGzmB cleavage site found at event; position 1,500

O88322

nidogen-2

basement membrane scaffold

cleaved at multiple sites

E9Q557

desmoplakin

desmosome-associated protein

cleaved at multiple sites

P02463

collagen IV alpha-1

basement membrane component

ND

Q99JR5

tubulointerstitial nephritis antigen-like

basement membrane component

ND

The following abbreviations are used: ECM, extracellular matrix; and ND, no data (insufficient data for prediction of cleavage site).

(Invitrogen). A total of 2 3 107 labeled cells in 100 ml PBS were transferred into WT recipients. Seven percent 2,4,6-trinitrochlorobenzene (TNCB) was applied to a shaved flank (sensitization dose). Six days later, an elicitation dose of 1% TNCB was applied to both sides of the left ear. After 24 hr, ears and spleen were analyzed for the presence of donor cells by flow cytometry. Poly(I:C)-Induced NK Cell Recruitment to the Peritoneal Cavity Naive WT and Gzmb / NK cells were labeled with either CellTracker-Orange CMRA or CellTrace Violet (Invitrogen). A total of 1 3 107 donor NK cells of each genotype were cotransferred into WT recipients. After 24 hr, recipient mice were injected i.p. with 100 mg poly(I:C), and after 48 hr spleens and peritoneal exudates were analyzed by flow cytometry for the presence of donor NK cells. Intravital Imaging of CTL Diapedesis WT and Gzmb / OT-I CTLs (5–6 days postactivation) were labeled in serumfree RPMI at 37 C with CFSE, CellTrace violet (6.7 mM, 1 3 107 cells/ml for 20 min), or CMTMR (5 mM, 1 3 107 cells/ml for 15 min) (Life Technologies). In the ear inflammation model, ears of albino C57BL/6 mice were tape stripped 18–24 hr prior to imaging. CTLs were mixed at a 1:1 ratio (Figure S1), injected, and imaged as previously described (Shulman et al., 2011). CTL rolling was defined as the movement of cells along the vessel wall at a lower velocity than free-flowing cells. Sticking was defined as CTL arrest for more than 30 s without further movement greater than one cell diameter. The rolling fraction was the number of rolling cells out of the total number of cells observed; the rolling velocity was the distance traveled by the cells in blood vessels (mm) as a function of time (s); the sticking fraction was the number of cells sticking in blood vessels out of the total number of cells observed; and the extravascular fraction was the number of cells in the interstitium out of the total number of cells observed. In the cremaster muscle model, C57BL/6 mice were anesthetized by intraperitoneal injection of ketamine-xylazine (80/10 mg/kg) with regular halfdose administrations. Mice were injected intrascrotally with 0.5 mg TNF-a (0.25 mg in 200 ml per testicle) 1–2 hr before surgery. Anesthetized mice were cannulated (right carotid artery) and placed on a customized stage for maintaining body temperature, and cremaster muscles were surgically exposed. The tissue was immersed in buffer containing 10 nM CXCL10 at 37 C, covered with a glass coverslip, and sealed with vacuum grease. Immediately before imaging, 1 3 108 labeled cells (WT and Gzmb / OT-I CTLs at a 1:1 ratio and previously checked for GzmB expression; Figure S2B) were injected via the cannula together with Evans Blue (50–100 ml at 33.3 mg/ml) to highlight blood vessels. Imaging was performed with a multiphoton LaVisionBioTec microscope with a motorized intravital stage (163 water-dipping objective [Nikon CFI75 LWD 163W], 0.80 numerical aperture, and 3.0 mm working distance). MaiTai HP and MaiTai Insight laser systems were tuned to 790 nm (0.81 W infrared [IR] source power) and 1,080 nm (1.35 W IR source power), respectively. Chosen fields of view encompassed postcapillary veins 20–60 mm in

970 Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc.

diameter (Hyun et al., 2012), and 3D stacks were acquired at 1 min intervals for up to 6 hr. As a control for potential labeling artifacts affecting CTL behavior, dyes (CMTMR and CellTrace Violet) were alternated between the two CTL populations in consecutive experiments. For calculating extravasation fractions, the proportion of CTLs exiting the vasculature was counted in relation to the total number of cells firmly engaged with the vascular wall. Firm engagement was defined as a cell displaying little or no displacement over 20 min or longer. Most extravasating cells initiate transmigration 20 min after arrest. Successful extravasation involved complete translocation of the cell body into the extravascular space. For minimizing bias during image acquisition and analysis, the identity of labeled CTLs was unmasked only after analysis was complete. 3D images were processed with Imaris, Volocity, and ImageJ. Tumor Homing OT-I T cells were expanded, and noncognate tumors were established and explanted for ex vivo observation of infiltrating OT-I CTLs via two-photon microscopy as previously described (Manjunath et al., 2001; Mrass et al., 2006). Mean speed and displacement were used for assessing migration of individual CTLs. Mean speed measures the mean velocity of vectors in a defined track. Displacement is the distance from the origin (time 0: point A) to the final point in the track (point X) and is normalized for the duration (seconds) of each track. Transwell Migration Assay Transwell (6.5 mm, 5 mm pore size; Corning) membranes were coated with either 1.5 mg/ml Matrigel (BD Biosciences) in a total volume of 50 ml or MDCK-cell-derived basement membrane as described above. Transwells were placed in a 24-well tray containing 600 ml complete medium per well with or without 100 ng/ml SDF-1a (Sigma-Aldrich). Cells were resuspended in 200 ml complete medium and added to the top chamber. After incubation for 16 hr at 37 C, cell transmigration to the bottom chamber was quantitated by CellTiter-Glo Luminescent Cell Viability Assay (Promega). Where indicated, Matrigel was treated with recombinant mouse GzmB (mGzmB) for 24 hr prior to assay. Data are expressed as a migration index, which is the ratio of the number of cells that migrate in response to SDF-1a over the number of cells that migrate in the absence of chemokine (random migration). Fluorescence Imaging of CTL Response to Basement Membrane OT-I CTLs retrovirally transduced with GFP and/or GzmB-mCherry (Bird et al., 2010) were cultured on Matrigel labeled with CellTracer Violet and were imaged with confocal microscopy. For visualization of GzmB release in a 3D environment, OT-I CTLs expressing and lacking GzmB were embedded in Matrigel containing 400 mM GzmB quenched fluorescent substrate z-IETD-R110 (Molecular Probes) and imaged by confocal microscopy. The appearance of fluorescence, representative of substrate cleavage, was assessed for each snapshot.

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

Degranulation Assay Splenocyte-derived CTLs were incubated in the presence or absence of Matrigel for 4 hr at 37 C. Cells and medium were separated, cells were lysed in 1% NP-40, and lysates and supernatants were assayed for BLT esterase activity (Takahashi et al., 1991; Takayama and Sitkovsky, 1987). Cells were treated with 2.5 mg/ml PMA and 0.5 mg/ml calcimycin for the culture period as a positive control. 3D Matrigel Migration and Remodeling Assay OT-I CTLs expressing or lacking GzmB were embedded in Matrigel labeled with DDAO-se and were imaged by confocal microscopy with CXCL10 as they moved along a chemokine gradient. 3D representations were constructed from consecutive z stacks of individual CTLs and their burrows and were categorized according to shape. Z stack volumes and image sequences were analyzed with FIJI and Volocity. SEM Samples were fixed in 2.5% gluteraldehyde in 0.1 M sodium cacodylate pH 7.4 for R2 hr, dehydrated in a series of increasing ethanol concentrations, and critical-point dried by hexamethyldisilazane (ProSciTech). After being coated with gold, samples were imaged with a Hitachi S570 scanning electron microscope (Monash Micro Imaging). Cells were scored from six random fields for polarization and interaction with the underlying matrix. ECM Cleavage Assay ECM protein was N-terminally dimethylated with 40 mM formaldehyde and 20 mM sodium cyanoborohydride. Blocking reagent was quenched with 100 mM glycine, and samples were treated with 100 nM protease overnight at 37 C. Neo-N-termini were biotinylated with 1 mg/ml sulfo-NHS-SS-biotin for 2 hr at ambient temperature. Protein was collected into an appropriate buffer and analyzed by SDS-PAGE and immunoblotting. Biotinylation was detected with streptavidin-HRP (Silenus). PROTOMAP Matrigel was treated with either 100 nM recombinant mouse GzmB or buffer alone overnight at 37 C. Samples were then submitted to PROTOMAP analysis (Dix et al., 2008). GzmB cleavage-site motifs were identified with consensus sequences generated from a published phage display (Kaiserman et al., 2006) and proteomic (Van Damme et al., 2009) data sets. Statistical Analyses Data were assessed for normal distribution, and when only two data sets were compared, significance was tested with a two-tailed Student’s t test or twotailed Mann-Whitney’s t test. Error bars indicate mean ± either SD or SE as indicated. SUPPLEMENTAL INFORMATION Supplemental Information includes two figures, two tables, and seven movies and can be found with this article online at http://dx.doi.org/10.1016/j.immuni. 2014.11.012. AUTHOR CONTRIBUTIONS P.I.B. conceived the study. M.D.P., M.M., P.I.B., M.R., P.M., and W.W. designed the experiments. M.D.P., O.K., A.K., M.M., R.J., and P.L.T. performed the experiments. M.D.P. and P.I.B. wrote the manuscript. M.O. and S.J.T. provided preliminary data and advice. ACKNOWLEDGMENTS M.D.P. and P.I.B. were supported by a grant from Australian National Health and Medical Research Council (NHMRC) project no. 1007830. P.L.T. was supported by the Dean’s Fellows Scholarship (University of Sydney) and an NHMRC Postgraduate Scholarship. W.W., P.M., M.M., and R.J. were supported by grants from the NHMRC and the Australian Research Council and a fellowship from the New South Wales Cancer Institute. W.W. and P.M. are fellows of the Cancer Institute New South Wales. We thank A. Matthews

(Monash University) for recombinant GzmB, J. Clark (Monash University) for assistance with ultrastructural analysis, and M. Biro, B. Roediger, and M. Kuligowski (Centenary Institute) for help with imaging experiments. Received: December 11, 2013 Accepted: November 25, 2014 Published: December 18, 2014 REFERENCES Abrams, G.A., Goodman, S.L., Nealey, P.F., Franco, M., and Murphy, C.J. (2000). Nanoscale topography of the basement membrane underlying the corneal epithelium of the rhesus macaque. Cell Tissue Res. 299, 39–46. Afonina, I.S., Tynan, G.A., Logue, S.E., Cullen, S.P., Bots, M., Lu¨thi, A.U., Reeves, E.P., McElvaney, N.G., Medema, J.P., Lavelle, E.C., and Martin, S.J. (2011). Granzyme B-dependent proteolysis acts as a switch to enhance the proinflammatory activity of IL-1a. Mol. Cell 44, 265–278. Benton, G., Kleinman, H.K., George, J., and Arnaoutova, I. (2011). Multiple uses of basement membrane-like matrix (BME/Matrigel) in vitro and in vivo with cancer cells. Int. J. Cancer 128, 1751–1757. Bird, C.H., Rizzitelli, A., Harper, I., Prescott, M., and Bird, P.I. (2010). Use of granzyme B-based fluorescent protein reporters to monitor granzyme distribution and granule integrity in live cells. Biol. Chem. 391, 999–1004. Boissonnas, A., Fetler, L., Zeelenberg, I.S., Hugues, S., and Amigorena, S. (2007). In vivo imaging of cytotoxic T cell infiltration and elimination of a solid tumor. J. Exp. Med. 204, 345–356. Buzza, M.S., and Bird, P.I. (2006). Extracellular granzymes: current perspectives. Biol. Chem. 387, 827–837. Buzza, M.S., Zamurs, L., Sun, J., Bird, C.H., Smith, A.I., Trapani, J.A., Froelich, C.J., Nice, E.C., and Bird, P.I. (2005). Extracellular matrix remodeling by human granzyme B via cleavage of vitronectin, fibronectin, and laminin. J. Biol. Chem. 280, 23549–23558. Buzza, M.S., Dyson, J.M., Choi, H., Gardiner, E.E., Andrews, R.K., Kaiserman, D., Mitchell, C.A., Berndt, M.C., Dong, J.-F., and Bird, P.I. (2008). Antihemostatic activity of human granzyme B mediated by cleavage of von Willebrand factor. J. Biol. Chem. 283, 22498–22504. Cao, X., Cai, S.F., Fehniger, T.A., Song, J., Collins, L.I., Piwnica-Worms, D.R., and Ley, T.J. (2007). Granzyme B and perforin are important for regulatory T cell-mediated suppression of tumor clearance. Immunity 27, 635–646. Crozat, K., Hoebe, K., Ugolini, S., Hong, N.A., Janssen, E., Rutschmann, S., Mudd, S., Sovath, S., Vivier, E., and Beutler, B. (2007). Jinx, an MCMV susceptibility phenotype caused by disruption of Unc13d: a mouse model of type 3 familial hemophagocytic lymphohistiocytosis. J. Exp. Med. 204, 853–863. D’Angelo, M.E., Bird, P.I., Peters, C., Reinheckel, T., Trapani, J.A., and Sutton, V.R. (2010). Cathepsin H is an additional convertase of pro-granzyme B. J. Biol. Chem. 285, 20514–20519. de Saint Basile, G., Me´nasche´, G., and Fischer, A. (2010). Molecular mechanisms of biogenesis and exocytosis of cytotoxic granules. Nat. Rev. Immunol. 10, 568–579. Devadas, S., Das, J., Liu, C., Zhang, L., Roberts, A.I., Pan, Z., Moore, P.A., Das, G., and Shi, Y. (2006). Granzyme B is critical for T cell receptor-induced cell death of type 2 helper T cells. Immunity 25, 237–247. Dix, M.M., Simon, G.M., and Cravatt, B.F. (2008). Global mapping of the topography and magnitude of proteolytic events in apoptosis. Cell 134, 679–691. Ferrell, N., Groszek, J., Li, L., Smith, R., Butler, R.S., Zorman, C.A., Roy, S., and Fissell, W.H. (2011). Basal lamina secreted by MDCK cells has size- and charge-selective properties. Am. J. Physiol. Renal Physiol. 300, F86–F90. Gaspari, A.A., and Katz, S.I. (1991). Induction of in vivo hyporesponsiveness to contact allergens by hapten-modified Ia+ keratinocytes. J. Immunol. 147, 4155–4161. Hansen, K.C., Kiemele, L., Maller, O., O’Brien, J., Shankar, A., Fornetti, J., and Schedin, P. (2009). An in-solution ultrasonication-assisted digestion method

Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc. 971

Immunity GzmB Expedites Cytotoxic Lymphocyte Extravasation

for improved extracellular matrix proteome coverage. Mol. Cell. Proteomics 8, 1648–1657. Hashizume, H., Baluk, P., Morikawa, S., McLean, J.W., Thurston, G., Roberge, S., Jain, R.K., and McDonald, D.M. (2000). Openings between defective endothelial cells explain tumor vessel leakiness. Am. J. Pathol. 156, 1363–1380. Hendel, A., Hiebert, P.R., Boivin, W.A., Williams, S.J., and Granville, D.J. (2010). Granzymes in age-related cardiovascular and pulmonary diseases. Cell Death Differ. 17, 596–606. Hiebert, P.R., and Granville, D.J. (2012). Granzyme B in injury, inflammation, and repair. Trends Mol. Med. 18, 732–741.

Regner, M., Pavlinovic, L., Koskinen, A., Young, N., Trapani, J.A., and Mu¨llbacher, A. (2009). Cutting edge: rapid and efficient in vivo cytotoxicity by cytotoxic T cells is independent of granzymes A and B. J. Immunol. 183, 37–40. Shulman, Z., Cohen, S.J., Roediger, B., Kalchenko, V., Jain, R., Grabovsky, V., Klein, E., Shinder, V., Stoler-Barak, L., Feigelson, S.W., et al. (2011). Transendothelial migration of lymphocytes mediated by intraendothelial vesicle stores rather than by extracellular chemokine depots. Nat. Immunol. 13, 67–76.

Hiebert, P.R., Wu, D., and Granville, D.J. (2013). Granzyme B degrades extracellular matrix and contributes to delayed wound closure in apolipoprotein E knockout mice. Cell Death Differ. 20, 1404–1414.

Sumaria, N., Roediger, B., Ng, L.G., Qin, J., Pinto, R., Cavanagh, L.L., Shklovskaya, E., Fazekas de St Groth, B., Triccas, J.A., and Weninger, W. (2011). Cutaneous immunosurveillance by self-renewing dermal gammadelta T cells. J. Exp. Med. 208, 505–518.

Hughes, C.S., Postovit, L.M., and Lajoie, G.A. (2010). Matrigel: a complex protein mixture required for optimal growth of cell culture. Proteomics 10, 1886– 1890.

Takahashi, K., Nakamura, T., Adachi, H., Yagita, H., and Okumura, K. (1991). Antigen-independent T cell activation mediated by a very late activation antigen-like extracellular matrix receptor. Eur. J. Immunol. 21, 1559–1562.

Hyun, Y.M., Sumagin, R., Sarangi, P.P., Lomakina, E., Overstreet, M.G., Baker, C.M., Fowell, D.J., Waugh, R.E., Sarelius, I.H., and Kim, M. (2012). Uropod elongation is a common final step in leukocyte extravasation through inflamed vessels. J. Exp. Med. 209, 1349–1362.

Takayama, H., and Sitkovsky, M.V. (1987). Antigen receptor-regulated exocytosis in cytotoxic T lymphocytes. J. Exp. Med. 166, 725–743.

Joeckel, L.T., and Bird, P.I. (2014). Are all granzymes cytotoxic in vivo? Biol. Chem. 395, 181–202. Kaiserman, D., Bird, C.H., Sun, J., Matthews, A., Ung, K., Whisstock, J.C., Thompson, P.E., Trapani, J.A., and Bird, P.I. (2006). The major human and mouse granzymes are structurally and functionally divergent. J. Cell Biol. 175, 619–630. Kleinman, H.K., and Martin, G.R. (2005). Matrigel: basement membrane matrix with biological activity. Semin. Cancer Biol. 15, 378–386. Kramer, M.D., and Simon, M.M. (1987). Are proteinases functional molecules of T lymphocytes? Immunol. Today 8, 140–142. Leppert, D., Waubant, E., Galardy, R., Bunnett, N.W., and Hauser, S.L. (1995). T cell gelatinases mediate basement membrane transmigration in vitro. J. Immunol. 154, 4379–4389. Manjunath, N., Shankar, P., Wan, J., Weninger, W., Crowley, M.A., Hieshima, K., Springer, T.A., Fan, X., Shen, H., Lieberman, J., and von Andrian, U.H. (2001). Effector differentiation is not prerequisite for generation of memory cytotoxic T lymphocytes. J. Clin. Invest. 108, 871–878. Mrass, P., Takano, H., Ng, L.G., Daxini, S., Lasaro, M.O., Iparraguirre, A., Cavanagh, L.L., von Andrian, U.H., Ertl, H.C., Haydon, P.G., and Weninger, W. (2006). Random migration precedes stable target cell interactions of tumor-infiltrating T cells. J. Exp. Med. 203, 2749–2761. Plasman, K., Van Damme, P., Kaiserman, D., Impens, F., Demeyer, K., Helsens, K., Goethals, M., Bird, P.I., Vandekerckhove, J., and Gevaert, K. (2011). Probing the efficiency of proteolytic events by positional proteomics. Mol. Cell. Proteomics 10, 003301. Prakash, M.D., Bird, C.H., and Bird, P.I. (2009). Active and zymogen forms of granzyme B are constitutively released from cytotoxic lymphocytes in the absence of target cell engagement. Immunol. Cell Biol. 87, 249–254.

972 Immunity 41, 960–972, December 18, 2014 ª2014 Elsevier Inc.

Tammi, R.H., Tammi, M.I., Hascall, V.C., Hogg, M., Pasonen, S., and MacCallum, D.K. (2000). A preformed basal lamina alters the metabolism and distribution of hyaluronan in epidermal keratinocyte ‘‘organotypic’’ cultures grown on collagen matrices. Histochem. Cell Biol. 113, 265–277. Van Damme, P., Maurer-Stroh, S., Plasman, K., Van Durme, J., Colaert, N., Timmerman, E., De Bock, P.-J., Goethals, M., Rousseau, F., Schymkowitz, J., et al. (2009). Analysis of protein processing by N-terminal proteomics reveals novel species-specific substrate determinants of granzyme B orthologs. Mol. Cell. Proteomics 8, 258–272. Van Goethem, E., Poincloux, R., Gauffre, F., Maridonneau-Parini, I., and Le Cabec, V. (2010). Matrix architecture dictates three-dimensional migration modes of human macrophages: differential involvement of proteases and podosome-like structures. J. Immunol. 184, 1049–1061. Wallich, R., Simon, M.M., and Mu¨llbacher, A. (2001). Virulence of mousepox virus is independent of serpin-mediated control of cellular cytotoxicity. Viral Immunol. 14, 71–81. Weninger, W., Biro, M., and Jain, R. (2014). Leukocyte migration in the interstitial space of non-lymphoid organs. Nat. Rev. Immunol. 14, 232–246. Willoughby, C.A., Bull, H.G., Garcia-Calvo, M., Jiang, J., Chapman, K.T., and Thornberry, N.A. (2002). Discovery of potent, selective human granzyme B inhibitors that inhibit CTL mediated apoptosis. Bioorg. Med. Chem. Lett. 12, 2197–2200. Xu, D., Regner, M., Smith, D., Ruby, J., Johnstone, R., and Mu¨llbacher, A. (2004). The multidrug resistance gene mdr1a influences resistance to ectromelia virus infection by mechanisms other than conventional immunity. Immunol. Cell Biol. 82, 462–470. Zhang, T., Somasundaram, R., Berencsi, K., Caputo, L., Gimotty, P., Rani, P., Guerry, D., Swoboda, R., and Herlyn, D. (2006). Migration of cytotoxic T lymphocytes toward melanoma cells in three-dimensional organotypic culture is dependent on CCL2 and CCR4. Eur. J. Immunol. 36, 457–467.