The Molecular Basis for Perforin Oligomerization and Transmembrane Pore Assembly

Katherine Baran,1 Michelle Dunstone,2,3 Jenny Chia,1 Annette Ciccone,1 Kylie A. Browne,1 Christopher J.P. Clarke,1,5 Natalya Lukoyanova,8 Helen Saibil,8 James C. Whisstock,2,4 Ilia Voskoboinik,1,6,9,* and Joseph A. Trapani1,2,7,9,* 1Cancer Immunology Program, Peter MacCallum Cancer Centre, St Andrew’s Place, East Melbourne, Victoria 3002, Australia 2Department of Biochemistry and Molecular Biology 3Department of Microbiology 4ARC Centre of Excellence in Structural and Functional Microbial Genomics Monash University, Clayton, Victoria 3800, Australia 5Department of Pathology 6Department of Genetics 7Department of Microbiology and Immunology The University of Melbourne, Parkville, Victoria 3010, Australia 8Department of Crystallography, Institute of Structural and Molecular Biology, Birkbeck College, Malet Street, London WC1E 7HX, UK 9These authors contributed equally to this work *Correspondence: [email protected] (I.V.), [email protected] (J.A.T.) DOI 10.1016/j.immuni.2009.03.016

SUMMARY effector molecule of CTL and cooperates with proapoptotic serine proteases (granzymes) to induce cell death. Granzymes Perforin, a pore-forming protein secreted by cyto- may enter the target cell independently of PRF through fluid- toxic lymphocytes, is indispensable for destroying phase endocytosis after initially binding to the target cell virus-infected cells and for maintaining immune membrane electrostatically (Bird et al., 2005; Trapani et al., homeostasis. Perforin polymerizes into transmem- 2003). However, the pore-forming activity of PRF is critical for brane channels that inflict osmotic stress and facili- the regulated access of granzymes to the cytosol, where they tate target cell uptake of proapoptotic granzymes. cleave key substrates to initiate apoptotic death (Froelich et al., 1996; Nakajima et al., 1995; Shi et al., 1997). PRF is essential Despite this, the mechanism through which perforin for all granule-mediated cytotoxicity, as has been extensively monomers self-associate remains unknown. Our demonstrated in gene-targeted PRF-deficient mice (Kagi et al., current study establishes the molecular basis for per- 1994) and in humans, where congenital PRF deficiency results forin oligomerization and pore assembly. We show in the potentially fatal immunoregulatory disorder, type 2 familial that after calcium-dependent membrane binding, haemophagocytic lymphohistiocytosis (FHL) (Stepp et al., 1999; direct ionic attraction between the opposite faces of Voskoboinik et al., 2006). Incomplete loss of PRF function has adjacent perforin monomers was necessary for pore also been linked to hematological cancer (Clementi et al., 2005; formation. By using mutagenesis, we identified the Mehta et al., 2006; Voskoboinik et al., 2007), although the findings opposing charges on residues Arg213 (positive) and remain controversial and a feasible mechanism to explain the Glu343 (negative) to be critical for intermolecular association is still lacking. In other experimental scenarios, PRF interaction. Specifically, disrupting this interaction appears to be a critical mediator of tissue damage, for instance the auto-immune beta cell destruction that results in juvenile- had no effect on perforin synthesis, folding, or traf- onset (type 1) diabetes in nonobese diabetic mice (Kagi et al., ficking in the killer cell, but caused a marked kinetic 1997). defect of oligomerization at the target cell membrane, Despite its function in regulating immune homeostasis and its severely disrupting lysis and granzyme B-induced critical role in viral immunity, PRF’s molecular function remains apoptosis. Our study provides important insights poorly understood, and predictions on the functional significance into perforin’s mechanism of action. of certain of its subdomains have never been experimentally tested. For more than two decades, PRF has been known to share sequence and antigenic similarity with lytic proteins of the INTRODUCTION complement membrane attack complex (MAC) (Tschopp et al., 1986; Young et al., 1986a). The electron microscopic appear- The granule exocytosis pathway is used by cytotoxic lympho- ances of PRF and MAC pores were also shown to be quite similar cytes (CTL or NK cells) to bring about the death of virus-infected (Blumenthal et al., 1984; Dennert and Podack, 1983; Masson or transformed cells (Bolitho et al., 2007). After the formation of an and Tschopp, 1985; Podack and Dennert, 1983; Young et al., immune synapse, CTL secretory granules polarize to the site of 1986a). The major area of sequence similarity between PRF contact with the target cell, fuse with the CTL cell membrane, and MAC proteins was aptly named the membrane attack and release their luminal contents into the synapse (Stinchcombe complex-perforin (MACPF) domain, reflecting the fact that no and Griffiths, 2007; Trapani, 1998). Perforin (PRF) is a crucial other known proteins possessed such a domain (Kwon et al.,

684 Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. Immunity Pore Formation by Lymphocyte Perforin

1989; Lichtenheld et al., 1988; Shinkai et al., 1988). Soon after of individual FHL-associated mutations and polymorphisms this, the members of this protein family were proposed to insert (Voskoboinik et al., 2005b) and to map the key residues in the into the target cell membrane via two closely adjacent amphi- PRF C2 domain responsible for calcium-dependent membrane pathic a-helical motifs in the MACPF region, and this view of binding (Voskoboinik et al., 2005a), a mandatory step that pore formation has remained dogma ever since (Peitsch et al., precedes membrane insertion and oligomerization. 1990; Shinkai et al., 1988). Our previous studies have shown that only 20% of missense Despite the suggested importance of the proposed amphi- PRF mutations identified in FHL patients specifically affect PRF’s pathic a helices, a lack of suitable methodologies for expressing function at the level of the target cell (and are therefore termed and purifying PRF and indeed other MACPF family members has ‘‘postsynaptic’’ mutations), whereas the majority map presynap- meant that direct testing of their function during pore formation tically; that is, they map to the CTL and produce deleterious has not been feasible. However, we have now developed robust effects on protein folding or trafficking to CTL secretory granules, and validated expression systems for purifying recombinant PRF resulting in protein degradation (Voskoboinik et al., 2005b). The and also for examining its function in the context of an intact CTL current study aimed to understand the basis for PRF’s interac- (Voskoboinik et al., 2004, 2005a, 2005b), enabling us to revisit this tion with the target cell membrane, so our initial step was to issue. It has also become apparent of late that the MACPF identify those mutations that resulted in postsynaptic defects domain is far more widely distributed in nature than previously of function. We performed an initial screen of 37 missense suspected, with evolutionary conservation of the MACPF fold mutations at any of 17 charged or polar residues in this region, extending back as far as the pore-forming toxins of bacteria by expressing PRF in RBL-2H3 cells (Shiver and Henkart, 1991). (Rosado et al., 2007). Publication of the first two crystal structures These residues have previously been thought to form the internal of MACPF proteins human C8a (Hadders et al., 2007; Slade et al., ‘‘aqueous’’ surface of the PRF pore. 2008) and bacterial Plu-MACPF (a putative toxin synthesized Three distinct types of outcome were noted with respect to by Photorhabdus luminescens)(Rosado et al., 2007) provided protein expression and PRF-dependent cytotoxicity (summa- a considerable advance in our ability to predict structure-function rized in Table S1 available online and Figure 1A). The smallest relationships for PRF (Rosado et al., 2008). The structural data but most interesting group of mutations (n = 6) produced post- also revealed that the MACPF domain was much larger than orig- synaptic dysfunctions, but these all occurred at one of only inally suspected (350 amino acids) and predicted that folding two of the 17 amino acids tested, residues R213 and D191. occurred almost identically to another family of bacterial pore- These mutations underwent further characterization and, as forming toxins, the cholesterol-dependent cytolysins (CDCs). described below, turned out to provide major insights on the These recent structural findings are also revolutionary because mechanism of PRF polymerization. A second, more common they reveal that many human pathogens (for instance, Clos- outcome (n = 14) was presynaptic dysfunction, which we have tridium, Streptococcus, Chlamydia) secrete pore-forming toxins previously defined as reduced expression and/or PRF degrada- that fold through a similar molecular mechanism as the very tion or truncation in the RBL cells (Voskoboinik et al., 2004). As proteins elaborated by the immune system to eliminate them. was observed previously with FHL-associated mutations (Ishii In the present study, we performed a comprehensive functional et al., 2005; Voskoboinik et al., 2005b), this group of substitutions analysis to investigate the molecular mechanism of PRF pore invariably led to a marked loss of function (Table S1). Most formation. We definitively demonstrated that the highly conserved commonly, presynaptic dysfunction occurred at residues that amphipathic a-helical region, formerly thought to constitute the showed extremely high conservation between species as diver- membrane-spanning domain, is involved in a distinct and critical gent as humans and fish (for instance, residues 189, 220, 221, function, namely the assembly of PRF monomers into a pore 231, or 239), presumably because they are critical for protein complex. Apart from the more generic calcium-binding C2 folding during its biosynthesis. The third and largest category domain (Uellner et al., 1997; Voskoboinik et al., 2005a), this study of mutations (n = 17) resulted in normal PRF expression and lytic is, to our knowledge, the first instance in which a molecular activity that was equivalent to wild-type (WT)-PRF. function of PRF has been unequivocally assigned to a specific Taken together, our extensive mutational analysis identified subregion of the molecule. R213 and D191 as crucial residues where mutation significantly compromised cytotoxic function but had no apparent effect on RESULTS protein folding.

Mutational Analysis of PRF Residues 187–239 A Positive Charge at Residue 213 Is Critical Recent advances in our understanding of the structures of for Postsynaptic PRF Function MACPF proteins Plu-MACPF (Rosado et al., 2007), human C8a Point mutations affecting residue R213 had variable effects on (Hadders et al., 2007), and C8a complexed to C8g (Slade et al., PRF function, depending on the nature of the amino acid substi- 2008) have clearly indicated that the highly conserved region of tution (Table S1). Retention of a positive charge with the conser- PRF hitherto thought to be an amphipathic a-helical structure vative substitution R213K had minimal effects on PRF expression, responsible for target membrane insertion must perform a processing, and function. Charge reversal (R213E) resulted in WT different function (Peitsch et al., 1990; Persechini et al., 1992). protein expression, but dramatically decreased cytotoxicity To resolve this issue, we undertook detailed site-directed muta- (Figures 1B and 1C). By contrast, introduction of an uncharged, genesis of residues 187–239 of mouse PRF. We made use of hydrophobic side chain at this position (R213L) resulted in both mammalian and baculovirus-based expression modalities that decreased cytotoxicity and reduced protein expression (Figures we had previously established to dissect the functional impact 1B and 1C).

Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. 685 Immunity Pore Formation by Lymphocyte Perforin

Figure 1. Two Views of the Structure of C8a (A) The position of residues mutated in this study (see Table S1) are indicated by spheres and are numbered with PRF numbering (see Figure S4 for an alignment between PRF and C8a). D191, R213, and E343 are highlighted in blue (R) and red (D and E). Mutation of these residues results in a postsynaptic defect. The G and H helices are colored pink and are labeled. (B) Substitution of residue R213 results in a postsynaptic defect in PRF lysis. Mouse PRF residue R213 was substituted with a hydrophobic (L) or charged (K or E) amino acid of similar size, and the cytotoxicity of PRF-transfected RBL-2H3 cells against Jurkat target cells was analyzed in a 4 hr 51Cr-release assay at the E:T ratios indicated. The data points represent the mean ± SD of the triplicate assay and are representative of at least three independent experiments. (C) Immunoblots showing PRF expression in the RBL-2H3 transfectant cells tested in (B). Actin staining was used to demonstrate equivalent protein loading. (D) The lytic activity of WT- and R213E-PRF. SRBC (top) and Jurkat (bottom) cell lysis assays were each performed with equal quantities of WT- and R213E-PRF at 37C for 20 min in buffer containing 2 mM calcium. Data points represent data pooled from at least three experiments performed on different days ± SD.

in solution exclusively as monomers (Figure S1B). This was in sharp contrast to recombinant A91V-PRF, a presynaptic mutation for which we have previously used the same methodology to demonstrate spontaneous aggregation (Vosko- boinik et al., 2007). To further confirm that R213E-PRF was correctly folded, we compared its proteolytic degradation These results strongly suggested that the presence of to that of WT-PRF when both proteins were digested with a charged side chain at position 213 is important for PRF folding varying concentrations of trypsin or chymotrypsin. The degrada- within the effector cell and, additionally, that a positive charge at tion patterns of the two proteins were indistinguishable on this position is critical for its lytic function after exocytosis. We SDS-PAGE, confirming that folding of R213E-PRF was equiva- therefore elected to further investigate R213E-PRF’s reduced lent to WT (Figure S1C). ability to bring about target cell lysis. A Role for R213 in PRF Polymerization, but Not Plasma The R213E Mutant Is Expressed and Purified as a Stable Membrane Binding Monomeric Protein We next compared the ability of recombinant WT- and R213E- To directly demonstrate the postsynaptic nature of the R213E PRF to lyse target cells. To insure that we used closely compa- substitution, WT- and R213E-PRF were purified from cultures rable amounts of the two proteins, each protein batch was of baculovirus-infected Sf21 cells (Figure S1A). R213E-PRF carefully titrated and quantitated by immunoblotting and the was expressed in similar quantity to WT and migrated at the resulting images analyzed by densitometry (Figure S2 and Experi- expected size (67 kDa) on SDS-PAGE. A key prediction for mental Procedures). Consistent with the results obtained when substitutions such as R213E that cause purely a postsynaptic R213E-PRF was delivered by RBL cells (Figures 1B and 1C), dysfunction is that accurate folding of the protein should approximately five times as much R213E-PRF as WT was required preclude the formation of nonspecific higher order complexes to achieve equivalent lysis of nucleated (Jurkat) or non-nucleated in solution. By using gel filtration chromatography under nonde- target cells (sheep erythrocytes), indicating 80% loss of function naturing, nonreducing conditions followed by immunoblotting of (Figure 1D). The difference in lysis was greatly amplified when the the eluted fractions, we confirmed that both WT- and R213E- same assay was performed at 22C, in that the mutated PRF was PRF (as well as all other mutants studied below) were present now unable to achieve 100% lysis (discussed further below).

686 Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. Immunity Pore Formation by Lymphocyte Perforin

Figure 2. Calcium-Dependent Plasma Membrane Binding and Kinetics of WT- and R213E-PRF-Mediated Lysis (A) Immunoblot analysis showing membrane-bound PRF after equal amounts (150 ng) of WT- and R213E-PRF were incubated with SRBC and various concentrations of calcium at 4C. The input lane contains puriﬁed PRF alone without the addition of any membranes. (B) A SRBC lysis assay measuring Hb release in response to equal amounts of WT and R213E-PRF was performed at 37C for 20 min in the presence of 2 mM calcium or 0.0625 mM calcium. The data are representative of at least three experiments performed on different days. (C) The kinetics of WT- and R213E-PRF lysis. SRBC lysis assays performed at 37C with equal quantities (400 ng) of WT- and R213E-PRF. (D) SRBC lysis assays performed at 22C for 20, 40, 60, or 120 min with equal quantities of WT- and R213E-PRF. Data are representative of at least three experiments ± SD of the triplicate assay.

R213E-PRF’s defective lysis could not be put down to a reduction To establish the molecular basis for these findings, we used in target cell membrane binding, because this was shown to be silver-stained SDS-PAGE to compare the rate of incorporation equivalent across a wide range of calcium concentrations down of the 67 kDa PRF membrane-bound monomers into SDS-resis- to approximately 0.0625 mM. As expected for these authentically tant higher-order membrane-inserted PRF oligomers. It has folded proteins, no membrane binding whatsoever was observed previously been shown that MAC and ‘‘polyperforin’’ typically when calcium was used at even lower concentrations or deleted migrate at >1000 kDa under such conditions, suggesting the altogether (Figure 2A). When SRBC lysis was examined by incorporation of at least 16 monomers into the complexes measuring hemoglobin (Hb) release at the suboptimal concentra- (Podack and Tschopp, 1982; Young et al., 1986b). Lysis of tion of Ca2+ (0.0625 mM), the loss of function associated with the SRBC at 37C resulted in the progressive appearance of the R213E substitution was not augmented, further confirming our polyperforin signal, which was clearly detectable by 2 min. This finding that defective calcium-dependent lipid binding was not was accompanied by a corresponding loss of the signal for the responsible for loss of function (Figure 2B). PRF monomer (Figure 3A). By contrast, the rate of incorporation To further investigate R213E-PRF dysfunction, we analyzed of monomeric R213E-PRF into polyperforin was slowed (for the kinetics of cell lysis. We found that whereas WT-PRF was example, compare the relative signals at 2 min), and some able to induce 100% Hb release within 2 min at 37C, complete residual monomer was still present at the later time points lysis by R213E-PRF took 7 min (Figure 2C). The kinetic defect (10 and 20 min; Figure 3A). Immunoblot analysis confirmed that could be compensated by adding 5 times as much R213E- after 20 min at 37C, more of the R213E-PRF remained in its PRF to the assay (data not shown). When the assay was repeated membrane-bound monomeric form than WT-PRF, consistent at 22C with various amounts of PRF, maximal Hb release was with less having been incorporated into polyperforin (Figure 3B). achieved with WT-PRF by 20 min. R213E-PRF released far As PRF monomers associate into pore-like structures through less Hb over 20 min; however, increasing the incubation time to noncovalent bonding, we reasoned that it should be possible to 120 min resulted in a dose-related but still partial rescue of the recover the signal for monomeric PRF by applying suitable dena- defect (Figure 2D). These results indicated that target cell pore turing conditions to membrane-inserted polymerized PRF. When formation was considerably less efficient for R213E-PRF. the lysed erythrocyte ‘‘ghosts’’ were dissolved and boiled in SDS

Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. 687 Immunity Pore Formation by Lymphocyte Perforin

the reduction in oligomer formation by R213E-PRF over a 20 min incubation with SRBC at 37C was approximately 50% (Figure 3C, p < 0.01). Overall, these experiments established that the impaired cytotoxic activity of R213E-PRF was not due to reduced membrane binding, but rather to its inefﬁcient oligomerization and pore formation.

R213E-PRF Forms Pores of Similar Size to WT We then went on to investigate whether R213E-formed pores were distinct from those formed by WT-PRF. PRF pores formed in vitro and viewed by electron microscopy typically have an internal diameter of 16 nm and are thought to consist of 16–18 monomers (Young et al., 1986b). Although the lytic defect of R213E-PRF could be rescued by using a higher concentration of the mutated protein at 37C, we wished to determine whether the pores formed by R213E-PRF were of similar or smaller size to those formed by WT-PRF. One possibility was that the defect was entirely kinetic, so that R213E-PRF pores were of similar size to WT but formed at a slower rate, resulting in fewer pores at early time points. An equally feasible explanation was that the mutated PRF formed smaller pores than did WT-PRF, but still of sufficient size to permit the release of Hb or 51Cr from erythrocytes or nucleated cells, respectively. To distinguish these possibilities, three types of experiment were undertaken. First, the influx of Ca2+ into PRF-treated cells was examined to determine whether the slowed polymerization seen with R213E-PRF would result in reduced calcium flux. Jurkat cells labeled with the ratiometric intracellular calcium ionophore Indo-1AM were treated with WT- or R213E-PRF at 37C and Ca2+ flux was followed over 15 min via flow cytometry (Figure 4A; Figure S3). When equal quantities of WT- and R213E-PRF were added, calcium flux was greater and pro- ceeded much faster with WT-PRF. When cells were exposed to a 4-fold molar excess of R213E-PRF, similar calcium influx into the cells resulted (Figure 4A), again consistent with a lag in poly- Figure 3. Reduced Incorporation of Monomeric R213E-PRF into Pol- merization for R213E-PRF relative to WT. The second set of yperforin, Compared with WT-PRF experiments used negative stain electron microscopy to directly (A) Equal amounts of WT- and R213E-PRF (1 mg) were incubated with SRBC at compare the size of pores formed in bilamellar liposomes by WT- 4Cor37C for up to 20 min and both the membrane-inserted polymeric PRF and R213E-PRF. This highly sensitive technology also enabled us (1000 kDa) and monomeric membrane-bound PRF (65 kDa) were visualized to confirm visually that both WT- and R213E-PRF remained by silver-stained SDS-PAGE. The input lane contains purified PRF alone exclusively in the monomeric state unless both lipid vesicles without the addition of SRBC. The data are representative of experiments performed a minimum of three times on different days. and calcium were also present (data not shown). We confirmed (B) Immunoblot analysis comparing membrane-bound monomeric PRF that there was no difference in the size or shape of WT- or (65 kDa) after equal amounts of WT- and R213E-PRF (150 ng) were incubated R213E-PRF pores when amounts of PRF resulting in equivalent with SRBC at 4Cor37C for up to 20 min. The data are representative of lysis were used (Figure 4B). Both formed cylindrical pores of western blot experiments performed seven times on different days. 14–22 nm diameter, although pore formation at 37C required (C) Immunoblots from seven similar experiments were scanned and densitom- 3–5 times longer incubation times for R312E-PRF compared etry was used to compare the levels of membrane-bound R213E- and WT-PRF remaining unincorporated into polyperforin after incubation at 37C. The signal with WT-PRF. A pore diameter of 14–22 nm is consistent with for membrane-bound PRF at 37C was plotted relative to membrane-bound the pore sizes determined for native PRF during the 1980s PRF at 4C. Data represent the mean ± SD of densitometry measurements (Podack et al., 1985; Young et al., 1986b). Finally, the ability of pooled from seven experiments performed on different days. R213E-PRF to facilitate granzyme-B (GrzB)-dependent target cell apoptosis was determined in comparison with WT-PRF by sample buffer that additionally contained 8 M urea, the signals for measuring the appearance of Annexin V/PI double-positive monomeric WT- and R213E-PRF were equalized (Figure 3B). Jurkat cells (Figure 4C; Figure S4A) or 51Cr release (Figure S4B) Once again, this confirmed that a greater proportion of WT- than upon exposure to PRF and GrzB. When equimolar quantities of R213E-PRF had been incorporated into oligomers and thus R213E- and WT-PRF were used, the capacity of GrzB to induce participated in pore formation (Figure 3B). By using densitometry apoptosis was severely compromised in cells treated with to estimate PRF oligomerization over seven similar experiments, R213E-PRF (Figure 4C; Figure S4B). However, when the weaker

688 Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. Immunity Pore Formation by Lymphocyte Perforin

lytic activity of R213E-PRF was compensated by using 5-fold more mutant protein, there was no difference in the number of cells killed through GrzB-dependent apoptosis (Figure 4C; Figure S4B). This result was further reﬂected by the percentage of viable cells remaining 24 hr after PRF/GrzB treatment (Figure S4A). Thus, our results across three very different types of assay consistently indicated that R213E-PRF produces pores of similar size to WT, conﬁrming that the defect in R213E-PRF function was principally kinetic in nature.

PRF Oligomerization Involves Protein-Protein Interactions between the Flat Faces of the Molecule The crystal structure of the MACPF domains of C8a and Plu- MACPF revealed that the MACPF domain adopts an unusual flat- tened cuboid-like shape (Hadders et al., 2007; Rosado et al., 2007). By analogy with CDCs, it is proposed that PRF oligmerizes through protein-protein interactions made between the two largest flat faces of the MACPF domain (Figure S5). Consistent with a role in oligomerization, our data reveal that R213 maps to one of two a helices (the H helix) (Rosado et al., 2007) that we predict would be buried in the interface of a PRF oligomer (Figure 1A; Figure S5). We therefore postulated that residues in the other helix (the G helix) may perform a similar function. Exam- ination of the structures of both C8a and Plu-MACPF suggested that PRF residue D191 on the G helix would be surface exposed (Figure 1A; Figure S5). We therefore made mutations at this position to test the idea that residues on both the G and H helices are important in oligomerization. Consistent with our prediction, and also with the finding that mutations introduced at residue D191 produced postsynaptic defects (Table S1), we found that substitutions at D191 (D191V, D191K, and D191S) all exhibited WT expression but markedly decreased cytotoxicity (Figures 5A and 5B). Charge reversal (D191K) or substitution with a hydrophobic residue (D191V) resulted in a more dramatic phenotype than did substitution with a polar noncharged side chain (D191S). The D191K substitution resulted in a complete loss of cytotoxicity and some residual lysis could be demonstrated for D191V, so baculovirus-expressed D191V-PRF was purified to examine its capacity for lysis and polymerization. Similar to the R213E substitution, D191V-PRF exhibited a reduction in lytic activity of 75%, which became more pronounced when the assay was performed at 22C(Figure 5C). Once again, a defect in polymerization was noted, in that less of the D191V monomer became incorporated into polyperforin (Figure 5D). Together, these data indicated the importance of an acidic residue at position 191, functioning cooperatively in PRF oligomerization with a basic side chain at position 213.

drawn as an approximation of calcium flux. The data are representative of Figure 4. Analyzing the Size and Function of WT- and R213E-PRF experiments performed a minimum of three times on different days. Pores (B) Representative negative stain electron microscopy images of pores formed (A) Indo-1AM-labeled Jurkat cells were exposed to equal concentrations of by recombinant WT- or R213E-PRF in bilamellar liposomes. Both PRFs formed WT- and R213E-PRF, or concentrations resulting in equal cell lysis (35%) for pores of similar shape and size, with diameters ranging from 14 to 22 nm. 20 min at 37C in buffer containing 1 mM calcium. Calcium flux was deter- (C) Apoptosis induced by purified gzmB and PRF. The proapoptotic effect of mined by the ratio of violet:blue wavelengths over time, via flow cytometry. human gzmB (25–50 nM) was assessed in synergy with equal concentrations Only viable (PI-negative) cells were analyzed. A software program designed of WT- or R213E-PRF, or with PRF concentrations adjusted to produce equal by V. Gromov (IT Department, Peter MacCallum Cancer Centre) allowed an levels of cell lysis when used alone. Apoptotic Jurkat cells were enumerated by average readout of 10 data points to be plotted per unit time. This represented flow cytometry measuring Annexin+PI double-positive cells. Data are repre- an average of 60–70,000 cells at a flow rate of 80–120 cells/s. A trendline was sentative of at least three experiments performed on different days.

Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. 689 Immunity Pore Formation by Lymphocyte Perforin

Figure 5. The Effect of Amino Acid Substitutions at Residue D191 on Mouse PRF Function (A) D191 was substituted with hydrophobic (V), charged (K), or polar (S) amino acids of similar size, and the cytotoxicity of PRF-transfected RBL-2H3 cells against Jurkat target cells was analyzed in a 4 hr 51Cr-release assay at the E:T ratios indicated. The data shown are the mean ± SD of the triplicate assay and are representative of at least three independent experiments. (B) Western blotting showing the levels of PRF expression in transfected RBL-2H3 cells. Actin staining was used to demonstrate equivalent protein loading. (C) Measured Hb release after equal quantities of WT- and D191V-PRF were incubated with SRBC at 37C (top) and 22C (bottom) for 20 min in buffer containing 2 mM calcium. Data are representative of at least three experiments performed on different days ± SD of the triplicate assay. (D) Western blot analysis showing a marked increase of membrane-bound D191V-PRF remaining in the monomeric form (65 kDa) compared to WT-PRF, after equal amounts were incubated with SRBC at 4Cor37C for up to 20 min. The data are representative of experiments performed a minimum of four times on different days.

R213 and E343 on Adjacent Monomers Form a Salt E343R-PRF exhibited WT expression, but had markedly Bridge during PRF Polymerization decreased cytotoxicity (Figures S6A and S6B). Recombinant Our data revealed that R213 and D191 are both important for E343R-PRF was next purified from baculovirus-infected cul- PRF oligomerization and pore formation and that disturbances tures. Like R213E, purified E343R-PRF remained a monomer in of charge at these positions resulted in a marked loss of function. solution (see Figure S1B). Consistent with the idea that E343 is Structural analyses suggested that both residues map to one of important for oligomerization and function, we noted a substan- the large flat faces of the PRF MACPF domain (Figure 1A; tial loss of lytic function for E343R-PRF, which became far more Figure S5). To build on this discovery, we decided to investigate pronounced at 22C(Figures 7A and 7B). Furthermore, a defect whether residues on the opposite flat face of the PRF MACPF in polymerization similar to that of R213E-PRF was also seen, in domain perform an analogous role. Sequence mapping revealed that incorporation of E343E-PRF monomer into polyperforin was that the acidic residue E343 was located in the center of the slowed, as shown by immunoblot (Figure 7C). These data sug- opposite flat face and thus may contribute to oligomerization gested that residue E343 plays an analogous role to D191 and (Figure 1A; Figure S5). Because E343 is located in a conserved R213 in PRF oligomerization. Given the clear phenocopying region with no insertions or deletions and is adjacent to a highly observed with R213E- and E343R-PRFs, it was not surprising conserved proline residue (Figure S5), we were able to map the that R213E-PRF showed a similar substantial reduction in func- position of this residue with confidence. Further, our molecular tion when the lysis reaction was carried out at 22C(Figures 7A modeling of a PRF oligomer suggested that this residue might and 7B). This result could not be put down to unfolding of the be appropriately placed to interact with R213 by means of mutant PRF at the lower temperature, as shown by the fact a salt bridge (Figure 6; Figure S5). that shifting the temperature back to 37C after the prior incuba- To test these ideas, we expressed and tested the function tion at 22C caused similar lysis to incubation at 37C alone of E343R-PRF in RBL-2H3 cells. We found that like R213E-, (Figure 7D).

690 Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. Immunity Pore Formation by Lymphocyte Perforin

et al., 1988). In the absence of functional or structural data to the contrary, this model has remained unchallenged ever since. Typically, helix 1 was suggested to comprise residues 187–203 and helix 2 residues 211–239, connected by a short cytoplasmic loop (residues 204–210) (Peitsch et al., 1990; Shinkai et al., 1988). In support of a functional role, this region also contains the highly

conserved MACPF ‘‘signature’’ motif Y/W-G-T/S-H-F/Y-X6-GG. About half (>30) of all the missense mutations identified in FHL patients have also been mapped to the conserved signature motif and the surrounding regions, further emphasizing its functional importance (Molleran Lee et al., 2004; Voskoboinik et al., 2005b, 2006). The current study has capitalized on newly developed expression modalities that enabled us to directly test the long-held Figure 6. Molecular Model of the Interactions between PRF Mono- prediction that sequences highly conserved between MACPF mers proteins (particularly residues 187–239) are critical for insertion Three PRF monomers are shown. Our data predict that R213 and E343 form and spanning of the target cell membrane. After a detailed a salt bridge (dashed lines) in the PRF oligomer. A schematic is also shown mutational and structure-function analysis, we demonstrated in Figure S5. unequivocally that the role of certain residues in this region is to mediate PRF oligomerization, and we found no evidence of To directly test the prediction that E343 and R213 form an ionic this region’s involvement in membrane spanning. Our data are interaction, we incorporated both substitutions into the same consistent with very recent structure-based predictions that the monomer to form the R213E+E343R-PRF double mutant, thereby membrane-spanning domains of PRF actually comprise the introducing a salt bridge ‘‘swap.’’ Remarkably, the double muta- regions analogous to transmembrane helices 1 and 2 (residues tion completely rescued both the lytic and polymerization defects 125–140 and 255–270) in MACPF proteins and CDCs (Rosado seen at 37C. Even more notably, the double mutation also et al., 2007; Rossjohn et al., 1997), which unfold into b strands completely reversed the more profound loss of function observed and ultimately form a b-barrel pore. in individual mutants R213E and E343R at 22C(Figures 7A Of 17 polar or charged residues that were mutated, two key and 7B). The effect on R213E-PRF function was particularly residues (D191 and R213) were found to be directly involved profound at reduced temperature (>99% loss of function), as in polymerization, in that several amino acid substitutions at indicated by the fact that complete lysis was not achieved at those positions had no substantial effect on PRF processing or even the highest concentration of PRF. trafficking within the effector cell, but nonetheless resulted in E343 and R213 map to opposite flat faces of the PRF mono- a marked reduction in target cell lysis, thus representing post- mer and thus are unlikely to form intramolecular interactions, synaptic defects. Overall, the data provide incontrovertible func- so the data clearly pointed to these two residues forming an tional evidence against the long-held view that most conserved intermolecular salt bridge that is essential for oligomerization residues of PRF and MACPF proteins in general play a critical (Figure 6). role in membrane insertion. Our major observations were centered on a substitution (R213E) that introduced a charge DISCUSSION reversal at this position. This change in charge was mapped to one of the side chains protruding from the H helix in the PRF For more than 20 years, the functional domains of PRF involved in MACPF domain. Consistent with the reduced function seen pore formation have remained unclear, and despite its critical role when R213E-PRF was tested in a cell-based assay, purified re- in viral immunity and immune homeostasis, PRF’s molecular combinant R213E-PRF also had decreased lytic function function remains the least understood of all the CTL effector compared with WT-PRF, resulting from slowed polymerization molecules. PRF was first purified in the early 1980s when its and pore formation. Importantly, multiple assays including the antigenic cross-reactivity with complement component 9, a direct visualization of the pore by negative stain electron micros- membrane-spanning constituent of the MAC, was first appreci- copy clearly demonstrated that R213E-PRF formed pores of ated (Tschopp et al., 1986; Young et al., 1986a), and ultrastruc- similar size to those formed by WT-PRF or described in EM tural studies showed that PRF forms barrel-shaped pores with studies that used PRF purified from primary CTLs (Blumenthal striking similarity to complement lesions (Blumenthal et al., 1984; et al., 1984; Masson and Tschopp, 1985; Young et al., 1986a). Masson and Tschopp, 1985; Young et al., 1986a). Sequence Our results for residue R213 were validated by investigations alignments revealed significant similarity between PRF and C9 on a second amino acid D191, which maps to the adjacent G in two adjacent domains, the MACPF domain and an adjacent helix. Our data revealed that D191V-PRF almost exactly phe- EGF-like domain (Kwon et al., 1989; Lichtenheld et al., 1988; nocopied R213E-PRF. Insight into the exquisite specificity of Shinkai et al., 1988). Subsequently, several independent molec- intermolecular PRF interactions can also be gleaned by analysis ular modeling studies predicted this region to contain two of mutations that have a modest or no effect on lytic activity. adjacent amphipathic a helices that each inserted into the lipid Analysis of other charged residues such as K194 and R197 via bilayer by virtue of its opposing hydrophobic and hydrophilic the structure of C8a suggest that, like D191, both of these resi- surfaces (Peitsch et al., 1990; Persechini et al., 1992; Shinkai dues are located on helix G. However, as a result of the helical

Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. 691 Immunity Pore Formation by Lymphocyte Perforin

Figure 7. The Lytic Activity of WT-, R213E-, E343R-, or R213E+E343R-PRF (A and B) Hb released from SRBC after incubation with equal quantities of WT-, R213E-, E343R-, and R213E/E343R-PRF at 37C (A) or 22C (B) for 20 min in buffer containing 2 mM calcium. The data are representative of at least three experiments performed on different days ± SD of the triplicate assay. (C) Immunoblot comparing the recovery of membrane-bound PRF monomers (65 kDa) after equal amounts of WT-, E343R-, and R213E/E343R-PRF were incubated with SRBC at 4Cor37C for up to 20 min. Data are representative of experiments performed a minimum of four times on different days. (D) The activity of R213E-PRF was restored to its maximal level relative to WT-PRF when a 22C reaction similar to that shown in (B) was followed by a further 20 min incubation at 37C. rotation, we predict that the side chains of both K194 and R197 predicted that these residues might be ‘‘docking’’ partners are positioned such that they point away from the plane of the through a salt bridge. In strong support of this hypothesis, when proposed oligomerization interface. we ‘‘swapped’’ the amino acid charges and generated the Sequence mapping and molecular modeling with the recently E343R+R213E double mutation, the lytic and polymerization published crystal structures of human C8a and the bacterial phenotypes were restored completely to WT. protein Plu-MACPF reveal that both R213 and D191 map to Our data provide detailed mechanistic understanding of PRF one of the two ‘‘flat faces’’ of the MACPF domain (Hadders pore assembly. Taken together with the structural and mecha- et al., 2007; Rosado et al., 2007). Cryo-EM studies (Tilley et al., nistic information available for CDCs, our data also provide 2005) of the homologous CDC family reveal that the correspond- a broader insight into pore formation by lytic MACPF proteins. ing region in CDCs is buried in the oligomeric pore form of these This model suggests that these molecules assemble into pores proteins. Together, these data suggest a consistent structural via a common mechanism; namely oligomerization via the large explanation for the observation that R213 and D191 are important flat faces of the MACPF domain. However, it is interesting to note for oligomerization. Based upon these data, we postulated that that none of the positions identified in this study as being impor- ionic interactions on the opposite face of the PRF MACPF domain tant for oligomerization (i.e., D191, R213, and E343) correspond would also be important for oligomerization. Accordingly, we to highly conserved positions in either the MACPF superfamily or hypothesized that the residue E343 might be involved in PRF olig- a subset of lytic membrane-inserting MACPF proteins such as omerization. Indeed, the E343R mutation resulted in substantially C9 or C8a (Figure S5; Rosado et al., 2007). Thus we suggest reduced lytic activity, likewise attributable to a polymerization that residues on the flat ‘‘oligomeric’’ faces of different subfam- defect. Given the opposite charge of R213 and E343, we further ilies of lytic MACPF proteins may have evolved for the particular

692 Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. Immunity Pore Formation by Lymphocyte Perforin

requirements of pore formation by that subfamily. For example, it WT and mutant PRFs were used in functional assays, various volumes of the is essential that oligomerization and pore formation of PRF is protein stocks were sampled and probed by immunoblotting with P1-8. Densi- prevented prior to its exocytic release from CTL granules. Our tometry on the resultant images was used to construct a standard curve for each batch of protein (Figure S2). Spontaneous protein aggregation in the mutational analysis of PRF reveals that both acidic and basic preparations of purified PRF was excluded by performing gel filtration, fol- residues are important for pore assembly. Because the pH of lowed by western blotting of the eluted fractions. Gel filtration was carried the lysosome-like granules is low (4.9), key acidic residues out under nondenaturing, nonreducing conditions, as follows. A Superdex such as D191 and E343 are predicted to be protonated where 200 HPLC column (GE Healthcare) was equilibrated with PRF storage buffer PRF is granule bound, so that a durable ionic interaction with (aqueous buffered salt solution) and successive fractions of equal volume other monomers is unlikely. We have previously shown that were eluted with the same buffer (Figure S1B). The column was calibrated by running Mr standards of known mass with the same buffers and under a similar mechanism is responsible for preventing premature the same conditions (Voskoboinik et al., 2007). activation of PRF within the CTL granule, because the proton- ation of aspartate residues within the C2 domain prevents Trypsin- and Chymotrypsin-Mediated Cleavage of PRF calcium binding within the acid granule milieu (Voskoboinik Trypsin and chymotrypsin were serially diluted (25–6.25 mg/ml) in PRF elution et al., 2005a). Therefore, a single mechanism that regulates buffer (10 ml). Purified PRF (3 mg) was added, the reaction volume was adjusted PRF’s intra- and intermolecular ionic interactions appears to to 20 ml, and the samples were incubated at 37C for 20 min. Protein sample be responsible for at least two key controls on premature PRF buffer (23) was added, and samples were analyzed by SDS-PAGE stained activation. with 0.25% w/v Coomassie blue. In summary, the current study has assigned a unique function SRBC Lysis (oligomerization) to a region of PRF long thought to perform an Sheep red blood cells (SRBC, Centre for Animal Biotechnology, University of alterative critical function, membrane insertion. Rather, the Melbourne) were extensively washed in (20 mM [pH 7.4]) buffered saline solu- recently available structural data clearly suggest that the trans- tion (HE) to remove plasma proteins prior to use. SRBC lysis was determined by membrane region instead consists of helices 1 and 2 (TMH1 estimating haemoglobin release into the supernatant (absorbance at 405 nm). and TMH2). The crystal structures of C8a and Plu-MACPF have Controls included SRBC resuspended in water (total release) or HE (sponta- also revealed their protein fold to be similar to those of the neous release). b-pore-forming toxin family, the CDCs. The CDCs do not possess Lysis and Apoptosis of Jurkat Cells alpha helices capable of membrane spanning, but instead Jurkat cell lysis and apoptosis was analyzed by measuring annexin+PI-posi- are thought to form a ‘‘pre-pore’’ prior to insertion, to enable tive cells via flow cytometry and specific 51Cr release as described elsewhere membrane-interacting domains to be revealed (Dang et al., (Sutton et al., 2008). 2005; Tilley et al., 2005). This mechanism of pore formation sees polymerization occurring after membrane binding and Binding of Recombinant PRF to Plasma Membranes prior to membrane insertion. Such a finding challenges current This assay was modified from a method described by us (Voskoboinik et al., perceptions regarding PRF function, in that it has always been 2005a). All stages of the assay were at 4C, and cells were washed with ice- 7 assumed that oligomerization would come after the insertion of cold buffers. Recombinant PRF was incubated with SRBC (2 3 10 ) and CaCl in a final volume of 200 ml HE for 5 min. SRBC were pelleted by centrifu- PRF monomers. Based on the high degree of structural similarity 2 gation, resuspended in the same buffer, and repelleted. The cells were then predicted between PRF and these b-PFT, further studies are lysed with DDW, subjected to SDS-PAGE, and transferred to nylon membranes warranted to determine whether PRF inserts into membranes for immunoblotting. via such a CDC-like mechanism. PRF Polymerization 7 EXPERIMENTAL PROCEDURES PRF (150 ng) was incubated with SRBC (2 3 10 ) for 5 min in HE containing 2 mM CaCl2 at 4 C, pelleted by centrifugation at 4 C, washed in the same Cell Culture, Transfection, and Immunoblotting ice-cold buffer, and resuspended in 50 ml of the same buffer at 37 C for varying Rat basophil leukemia (RBL-2H3, referred to as RBL) cells were cultured, times. The cells were immediately chilled on ice, pelleted once again at 4 C, transfected, and used in cytotoxicity assays as described (Voskoboinik resuspended in SDS sample buffer or SDS sample buffer containing 8 M et al., 2005b). To eliminate potential differences in transfection efficiency, tran- urea, subjected to SDS-PAGE, and transferred to nylon membranes for immu- sient transfections of the effector RBL cells were performed with a pIRES-GFP noblotting. Silver-stained gels were run on step SDS-PAGE with the stack vector, and GFP-positive transfectants were sorted 20–24 hr later on the basis section at 4% (w/v), the top separating section at 6% (w/v), and the bottom of identical GFP fluorescence (Sutton et al., 2008; Voskoboinik et al., 2005b). separating section at 10% acrylamide. Although IRES-driven GFP fluorescence was used as a surrogate marker for PRF expression in this initial cell sort, all cell populations used in cytotoxicity Calcium Flux experiments were also tested directly for PRF expression by western blot, Jurkat cells were labeled with 2 mM Indo-1AM (Invitrogen), an emission-ratio- with the rat anti-mouse PRF monoclonal antibody P1-8 (for example, as shown metric fluorescence indicator for quantifying intracellular Ca2+, in HE buffer in Figures 1B and 1C). Mouse anti-human actin monoclonal antibody was used for 20 min at 37C and washed three times. 1 3 105 labeled Jurkat cells as an independent control for protein loading (ICN Biomedicals). To generate were resuspended in HE/1 mM CaCl2 containing 0.5 mg/ml 7AAD. The ratio missense mutations, site-directed mutagenesis of the mouse PRF cDNA was of absorbance at 420:510 nm was estimated to obtain a baseline reading prior performed with the QuickChange methodology (Stratagene). to PRF addition.

Purification and Characterization of Recombinant PRF Structural Analysis Recombinant PRF was purified from the supernatants of baculovirus-infected Structural analysis of PRF was performed with the structures of C8a and Sf21 cells, as described (Sutton et al., 2008). All batches of mutant PRF Plu-MACPF. Superpositions of C8a and Plu-MACPF were performed with (R213E, D191V, E343R, and R213E/E343R) were expressed, purified, and MUSTANG (Konagurthu et al., 2006). Alignments were produced with clustalW stored in parallel with simultaneously expressed WT-PRF, and functional (Larkin et al., 2007) and manually edited to take secondary structure into assays were carried out within 2–3 days. To ensure that similar amounts of account. All figures were produced with PYMOL (DeLano Scientific LLC).

Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. 693 Immunity Pore Formation by Lymphocyte Perforin

A template homology model of murine PRF was built with the X-ray crystal Dennert, G., and Podack, E.R. (1983). Cytolysis by H-2-specific T killer cells. structure of human C8a (2QQH, 2RD7) as a template in the program MODELER Assembly of tubular complexes on target membranes. J. Exp. Med. 157, (Eswar et al., 2008) as deployed in Quanta (Accelrys inc) and via modeling/ 1483–1495. refinement techniques as described (Myers et al., 2000; Whisstock et al., Eswar, N., Eramian, D., Webb, B., Shen, M.Y., and Sali, A. (2008). Protein 1996). The alignment between murine PRF and human C8a was used as a basis structure modeling with MODELLER. Methods Mol. Biol. 426, 145–159. for building the model (Figure S3). Residues that were disordered in electron Froelich, C.J., Orth, K., Turbov, J., Seth, P., Gottlieb, R., Babior, B., Shah, density in the template C8a or that represent significant insertions C8a with G.M., Bleackley, R.C., Dixit, V.M., and Hanna, W. (1996). New paradigm for respect to PRF were not modeled (residues 81–83, 131–139). A schematic of lymphocyte granule-mediated cytotoxicity. Target cells bind and internalize a PRF trimer (Figure 6) was generated by manually positioning PRF monomers granzyme B, but an endosomolytic agent is necessary for cytosolic delivery with respect to one another so that no steric clashes were apparent. and subsequent apoptosis. J. Biol. Chem. 271, 29073–29079.

Negative Stain Electron Microscopy of PRF Pores in Liposomes Hadders, M.A., Beringer, D.X., and Gros, P. (2007). Structure of C8alpha- Liposomes were prepared by extrusion (via Mini-Extruder, Avanti Polar Lipids, MACPF reveals mechanism of membrane attack in complement immune Alabaster, USA) from a 1:1 molar ratio of cholesterol and phosphatidylcholine defense. Science 317, 1552–1554. (MacDonald et al., 1991). Monomeric PRF was added to fresh liposomes (final Ishii, E., Ueda, I., Shirakawa, R., Yamamoto, K., Horiuchi, H., Ohga, S., Furuno, lipid concentration of 2 mM) at a molar ratio of 1:4000À7000 total lipids in K., Morimoto, A., Imayoshi, M., Ogata, Y., et al. (2005). Genetic subtypes of the following buffer: 0.15 M NaCl, 1 mM CaCl2, 20 mM HEPES (pH 8.0). The familial hemophagocytic lymphohistiocytosis: correlations with clinical mixture was placed on ice for 1–2 min and then incubated at 37C for 1–3 min features and cytotoxic T lymphocyte/natural killer cell functions. Blood 105, for WT-PRF and 5–10 min for R213E-PRF. Immediately after this, the sample 3442–3448. was applied to a freshly glow-discharged carbon-coated copper grid and nega- Kagi, D., Ledermann, B., Burki, K., Seiler, P., Odermatt, B., Olsen, K.J., tively stained with a few drops of 1% w/v uranyl acetate. Micrographs were Podack, E.R., Zinkernagel, R.M., and Hengartner, H. (1994). Cytotoxicity collected under low-dose conditions on a Tecnai T12 electron microscope (FEI, mediated by T cells and natural killer cells is greatly impaired in perforin-defi- Eindhoven, The Netherlands) operating at an accelerating voltage of 120 keV. cient mice. Nature 369, 31–37. m Images were taken at 52,0003 magnification (defocus range 0.7–0.9 m). Kagi, D., Odermatt, B., Seiler, P., Zinkernagel, R.M., Mak, T.W., and Hengart- ner, H. (1997). Reduced incidence and delayed onset of diabetes in perforin- SUPPLEMENTAL DATA deficient nonobese diabetic mice. J. Exp. Med. 186, 989–997. Konagurthu, A.S., Whisstock, J.C., Stuckey, P.J., and Lesk, A.M. (2006). Supplemental Data include six figures and one table and can be found with this MUSTANG: A multiple structural alignment algorithm. Proteins 64, 559–574. article online at http://www.cell.com/immunity/supplemental/S1074-7613(09) Kwon, B.S., Wakulchik, M., Liu, C.C., Persechini, P.M., Trapani, J.A., Haq, 00189-7. A.K., Kim, Y., and Young, J.D. (1989). The structure of the mouse lymphocyte pore-forming protein perforin. Biochem. Biophys. Res. Commun. 158, 1–10. ACKNOWLEDGMENTS Larkin, M.A., Blackshields, G., Brown, N.P., Chenna, R., McGettigan, P.A., J.A.T. and I.V. are supported by Fellowships, a project grant, and a Program McWilliam, H., Valentin, F., Wallace, I.M., Wilm, A., Lopez, R., et al. (2007). Grant from the National Health and Medical Research Council (NHMRC) of Clustal W and Clustal X version 2.0. Bioinformatics 23, 2947–2948. Australia. J.C.W. is an Australian Research Council Federation Fellow and is Lichtenheld, M.G., Olsen, K.J., Lu, P., Lowrey, D.M., Hameed, A., Hengartner, supported by a program and project grant from the NHMRC as well as a Centre H., and Podack, E.R. (1988). Structure and function of human perforin. Nature of Excellence/Discovery Project from the Australian Research Council. M.D. is 335, 448–451. an NHMRC Peter Doherty Training Fellow and recipient of a Monash University MacDonald, R.C., MacDonald, R.I., Menco, B.P., Takeshita, K., Subbarao, Faculty of Medicine Strategic Grant Scheme Early Career Researcher award. N.K., and Hu, L.R. (1991). Small-volume extrusion apparatus for preparation K.B. is supported by a postdoctoral fellowship from the Cancer Council of large, unilamellar vesicles. Biochim. Biophys. Acta 1061, 297–303. Victoria, Australia. We thank C. House for expert assistance with gel filtration Masson, D., and Tschopp, J. (1985). Isolation of a lytic, pore-forming protein experiments. (perforin) from cytolytic T-lymphocytes. J. Biol. Chem. 260, 9069–9072.

Received: August 26, 2008 Mehta, P.A., Davies, S.M., Kumar, A., Devidas, M., Lee, S., Zamzow, T., Elliott, Revised: November 24, 2008 J., Villanueva, J., Pullen, J., Zewge, Y., and Filipovich, A. (2006). Perforin Accepted: March 6, 2009 polymorphism A91V and susceptibility to B-precursor childhood acute Published online: May 14, 2009 lymphoblastic leukemia: A report from the Children’s Oncology Group. Leukemia 20, 1539–1541. REFERENCES Molleran Lee, S., Villanueva, J., Sumegi, J., Zhang, K., Kogawa, K., Davis, J., and Filipovich, A.H. (2004). Characterisation of diverse PRF1 mutations Bird, C.H., Sun, J., Ung, K., Karambalis, D., Whisstock, J.C., Trapani, J.A., and leading to decreased natural killer cell activity in North American families Bird, P.I. (2005). Cationic sites on granzyme B contribute to cytotoxicity by with haemophagocytic lymphohistiocytosis. J. Med. Genet. 41, 137–144. promoting its uptake into target cells. Mol. Cell. Biol. 25, 7854–7867. Myers, M.A., Davies, J.M., Tong, J.C., Whisstock, J., Scealy, M., Mackay, I.R., Blumenthal, R., Millard, P.J., Henkart, M.P., Reynolds, C.W., and Henkart, P.A. and Rowley, M.J. (2000). Conformational epitopes on the diabetes autoantigen (1984). Liposomes as targets for granule cytolysin from cytotoxic large gran- GAD65 identiﬁed by peptide phage display and molecular modeling. J. Immu- ular lymphocyte tumors. Proc. Natl. Acad. Sci. USA 81, 5551–5555. nol. 165, 3830–3838. Bolitho, P., Voskoboinik, I., Trapani, J.A., and Smyth, M.J. (2007). Apoptosis Nakajima, H., Park, H.L., and Henkart, P.A. (1995). Synergistic roles of gran- induced by the lymphocyte effector molecule perforin. Curr. Opin. Immunol. zymes A and B in mediating target cell death by rat basophilic leukemia 19, 339–347. mast cell tumors also expressing cytolysin/perforin. J. Exp. Med. 181, 1037– Clementi, R., Locatelli, F., Dupre, L., Garaventa, A., Emmi, L., Bregni, M., 1046. Cefalo, G., Moretta, A., Danesino, C., Comis, M., et al. (2005). A proportion Peitsch, M.C., Amiguet, P., Guy, R., Brunner, J., Maizel, J.V., Jr., and Tschopp, of patients with lymphoma may harbor mutations of the perforin gene. Blood J. (1990). Localization and molecular modelling of the membrane-inserted 105, 4424–4428. domain of the ninth component of human complement and perforin. Mol. Dang, T.X., Hotze, E.M., Rouiller, I., Tweten, R.K., and Wilson-Kubalek, E.M. Immunol. 27, 589–602. (2005). Prepore to pore transition of a cholesterol-dependent cytolysin visual- Persechini, P.M., Ojcius, D.M., Adeodato, S.C., Notaroberto, P.C., Daniel, ized by electron microscopy. J. Struct. Biol. 150, 100–108. C.B., and Young, J.D. (1992). Channel-forming activity of the perforin

694 Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. Immunity Pore Formation by Lymphocyte Perforin

N-terminus and a putative a-helical region homologous with complement C9. Tilley, S.J., Orlova, E.V., Gilbert, R.J., Andrew, P.W., and Saibil, H.R. (2005). Biochemistry 31, 5017–5021. Structural basis of pore formation by the bacterial toxin pneumolysin. Cell Podack, E.R., and Dennert, G. (1983). Assembly of two types of tubules with 121, 247–256. putative cytolytic function by cloned natural killer cells. Nature 302, 442–445. Trapani, J.A. (1998). Dual mechanisms of apoptosis induction by cytotoxic Podack, E.R., and Tschopp, J. (1982). Circular polymerization of the ninth lymphocytes. Int. Rev. Cytol. 182, 111–192. component of complement. Ring closure of the tubular complex confers resis- Trapani, J.A., Sutton, V.R., Thia, K.Y., Li, Y.Q., Froelich, C.J., Jans, D.A., tance to detergent dissociation and to proteolytic degradation. J. Biol. Chem. Sandrin, M.S., and Browne, K.A. (2003). A clathrin/dynamin- and mannose- 257, 15204–15212. 6-phosphate receptor-independent pathway for granzyme B-induced cell Podack, E.R., Young, J.D., and Cohn, Z.A. (1985). Isolation and biochemical death. J. Cell Biol. 160, 223–233. and functional characterization of perforin 1 from cytolytic T-cell granules. Tschopp, J., Masson, D., and Stanley, K.K. (1986). Structural/functional simi- Proc. Natl. Acad. Sci. USA 82, 8629–8633. larity between proteins involved in complement- and cytotoxic T-lymphocyte- Rosado, C.J., Buckle, A.M., Law, R.H., Butcher, R.E., Kan, W.T., Bird, C.H., mediated cytolysis. Nature 322, 831–834. Ung, K., Browne, K.A., Baran, K., Bashtannyk-Puhalovich, T.A., et al. (2007). Uellner, R., Zvelebil, M.J., Hopkins, J., Jones, J., MacDougall, L.K., Morgan, A common fold mediates vertebrate defense and bacterial attack. Science B.P., Podack, E., Waterfield, M.D., and Griffiths, G.M. (1997). Perforin is acti- 317, 1548–1551. vated by a proteolytic cleavage during biosynthesis which reveals a phospho- Rosado, C.J., Kondos, S., Bull, T.E., Kuiper, M.J., Law, R.H.P., Buckle, A.M., lipid-binding C2 domain. EMBO J. 16, 7287–7296. Voskoboinik, I., Bird, P.I., Trapani, J.A., Whisstock, J.C., and Dunstone, M.A. Voskoboinik, I., Thia, M.C., De Bono, A., Browne, K., Cretney, E., Jackson, J.T., (2008). The MACPF/CDC family of pore forming toxins. Cell. Microbiol. 10, Darcy, P.K., Jane, S.M., Smyth, M.J., and Trapani, J.A. (2004). The functional 1765–1774. basis for hemophagocytic lymphohistiocytosis in a patient with co-inherited Rossjohn, J., Feil, S.C., McKinstry, W.J., Tweten, R.K., and Parker, M.W. missense mutations in the perforin (PFN1) gene. J. Exp. Med. 200, 811–816. (1997). Structure of a cholesterol-binding, thiol-activated cytolysin and a model Voskoboinik, I., Thia, M.C., Fletcher, J., Ciccone, A., Browne, K., Smyth, M.J., of its membrane form. Cell 89, 685–692. and Trapani, J.A. (2005a). Calcium-dependent plasma membrane binding and Shi, L., Mai, S., Israels, S., Browne, K., Trapani, J.A., and Greenberg, A.H. cell lysis by perforin are mediated through its C2 domain: A critical role for aspar- (1997). Granzyme B (GraB) autonomously crosses the cell membrane and tate residues 429, 435, 483, and 485 but not 491. J. Biol. Chem. 280, 8426–8434. perforin initiates apoptosis and GraB nuclear localization. J. Exp. Med. 185, Voskoboinik, I., Thia, M.C., and Trapani, J.A. (2005b). A functional analysis of 855–866. the putative polymorphisms A91V and N252S and 22 missense perforin muta- Shinkai, Y., Takio, K., and Okumura, K. (1988). Homology of perforin to the tions associated with familial hemophagocytic lymphohistiocytosis. Blood ninth component of complement (C9). Nature 334, 525–527. 105, 4700–4706. Shiver, J.W., and Henkart, P.A. (1991). A noncytotoxic mast cell tumor Voskoboinik, I., Smyth, M.J., and Trapani, J.A. (2006). Perforin-mediated line exhibits potent IgE-dependent cytotoxicity after transfection with the target-cell death and immune homeostasis. Nat. Rev. Immunol. 6, 940–952. cytolysin/perforin gene. Cell 64, 1175–1181. Voskoboinik, I., Sutton, V.R., Ciccone, A., House, C.M., Chia, J., Darcy, P.K., Slade, D.J., Lovelace, L.L., Chruszcz, M., Minor, W., Lebioda, L., and Sodetz, Yagita, H., and Trapani, J.A. (2007). Perforin activity and immune homeostasis: J.M. (2008). Crystal structure of the MACPF domain of human complement The common A91V polymorphism in perforin results in both presynaptic and protein C8 alpha in complex with the C8 gamma subunit. J. Mol. Biol. 379, postsynaptic defects in function. Blood 110, 1184–1190. 331–342. Whisstock, J., Lesk, A.M., and Carrell, R. (1996). Modeling of serpin-protease Stepp, S.E., Dufourcq-Lagelouse, R., Le Deist, F., Bhawan, S., Certain, S., complexes: antithrombin-thrombin, alpha 1-antitrypsin (358Met/Arg)- Mathew, P.A., Henter, J.I., Bennett, M., Fischer, A., de Saint Basile, G., and thrombin, alpha 1-antitrypsin (358Met/Arg)-trypsin, and antitrypsin-elastase. Kumar, V. (1999). Perforin gene defects in familial hemophagocytic lymphohis- Proteins 26, 288–303. tiocytosis. Science 286, 1957–1959. Young, J.D., Cohn, Z.A., and Podack, E.R. (1986a). The ninth component of Stinchcombe, J.C., and Griffiths, G.M. (2007). Secretory mechanisms in complement and the pore-forming protein (perforin 1) from cytotoxic T cells: cell-mediated cytotoxicity. Annu. Rev. Cell Dev. Biol. 23, 495–517. structural, immunological, and functional similarities. Science 233, 184–190. Sutton, V.R., Waterhouse, N.J., Baran, K., Browne, K., Voskoboinik, I., and Young, J.D., Podack, E.R., and Cohn, Z.A. (1986b). Properties of a purified Trapani, J.A. (2008). Measuring cell death mediated by cytotoxic lymphocytes pore-forming protein (perforin 1) isolated from H-2-restricted cytotoxic T cell or their granule effector molecules. Methods 44, 241–249. granules. J. Exp. Med. 164, 144–155.

Immunity 30, 684–695, May 22, 2009 ª2009 Elsevier Inc. 695