You are on page 1of 13

REVIEWS

Perforin and granzymes: function,


dysfunction and human pathology
Ilia Voskoboinik1,2, James C. Whisstock3,4 and Joseph A. Trapani1,2
Abstract | A defining property of cytotoxic lymphocytes is their expression and regulated
secretion of potent toxins, including the pore-forming protein perforin and serine protease
granzymes. Until recently, mechanisms of pore formation and granzyme transfer into the
target cell were poorly understood, but advances in structural and cellular biology have
now begun to unravel how synergy between perforin and granzymes brings about
target cell death. These and other advances are demonstrating the surprisingly broad
pathophysiological roles of the perforin–granzyme pathway, and this has important
implications for understanding immune homeostasis and for developing immunotherapies
for cancer and other diseases. In particular, we are beginning to define and understand a
range of human diseases that are associated with a failure to deliver active perforin to
target cells. In this Review, we discuss the current understanding of the structural, cellular
and clinical aspects of perforin and granzyme biology.

Central to both the innate and adaptive arms of the granzymes (FIG. 1) and granulysin1 — into the synaptic
Perforinopathies
A range of human
immune system is the ability of cytotoxic lymphocytes cleft. Defects of this cytotoxic pathway — particularly
immune-mediated disorders to recognize virus-infected or transformed cells and a failure to secrete functional perforin — are associated
that are caused by impaired kill them by apoptosis. The role of cytotoxic lympho- with various human disorders (which we have termed
perforin delivery or function. cytes in defending against pathogens has long been ‘perforinopathies’ (REF. 2)), including familial haemo­
appreciated, but the more controversial notion of phagocytic lymphohistiocytosis (FHL), protracted
immune surveillance of cancer — which was proposed viral infections and susceptibility to haemato­logical
1
Cancer Immunology
Program, Peter MacCallum
by Macfarlane Burnet in the 1950s — only recently malignancies3–5. Although the mutations that disrupt
Cancer Centre, East gained broad acceptance. Subsequently, the develop­ granule trafficking or exocytosis are mostly rare, recent
Melbourne, Victoria 3002, ment of immune-based therapies that exploit the attention has turned to a common polymorphism in
Australia. activity of cytotoxic lymphocytes has begun to revo- the perforin (PRF1) gene that encodes moderate-to-
2
Sir Peter MacCallum
lutionize cancer therapy. Thus, new cancer therapies severe dysfunction of the perforin protein2. In this
Department of Oncology,
University of Melbourne, that promote antitumour cytotoxicity, such as gene- Review, we discuss the mechanisms by which perforin
Parkville, Victoria 3010, engineered chimeric antigen receptor-expressing and granzymes kill target cells, and the consequences
Australia. T cells and recombinant antibodies, take advantage of of defects in this essential facet of immune function.
3
Department of Biochemistry the perforin–granzyme effector pathway. Advancing
and Molecular Biology,
Monash University, Clayton,
our understanding of cytotoxic lymphocyte biology The cytotoxic cargo of cytotoxic lymphocytes
Victoria 3800, Australia. should support the further development of immune- The discovery of perforin and granzymes. The first
4
Australian Research Council based therapies for cancer, as well as for infectious report that lymphocytes kill specific target cells was
Centre of Excellence in diseases and autoimmunity. more than 50 years ago6. Subsequently, the Golgi and
Advanced Molecular Imaging,
Cytotoxic lymphocytes include cytotoxic T lympho­ electron-dense secretory granules were shown to re‑­
Monash University, Clayton,
Victoria 3800, Australia. cytes (CTLs) and natural killer (NK) cells. Despite orient (or become ‘polarized’) towards the target cell7,8,
Correspondence to I.V. and substantial differences in how these two cell types are and the dynamics of CTL secretory granule movement
J.A.T. activated and how they recognize their targets, the key towards the immunological synapse were determined9.
e‑mails: pathways that mediate target cell death are conserved. In 1980, pore-like membrane lesions were observed on
ilia.voskoboinik@petermac.org;
joe.trapani@petermac.org
Once conjugated to a target cell, the cytotoxic secre- cells that were targeted by antibody-dependent degranu-
doi:10.1038/nri3839 tory granules traffic to the immunological synapse and lation of NK cells10. The groups of Henkart, Okumura
Published online 22 May 2015 release a cargo of deadly proteins — including perforin, and Podack11–16 then isolated a monomeric 67 kDa

388 | JUNE 2015 | VOLUME 15 www.nature.com/reviews/immunol

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

a b
MTOC Secretory granule
Granzyme
Perforin Microtubule
Secretory granule
trafficking

Cytotoxic
lymphocyte Target cell

c d
Release of perforin
and granzymes
Fusion of granules
with the presynaptic
membrane
Perforin
pore

↑ Ca2+

Ca2+

Figure 1 | Interaction of a cytotoxic lymphocyte with a


target cell. The cytotoxic lymphocyte recognizes its target
cell and forms an immunological synapse (part a). The
e microtubule-organizing centre (MTOC) of the cytotoxic
lymphocyte polarizes and secretory granules traffic towards
the presynaptic membrane (part b). The secretory granules
then fuse with the presynaptic membrane and release perforin
and granzymes into the synaptic cleft (parts c and d). At the
postsynaptic membrane, perforin forms large transmembrane
pores that enable the diffusion of granzymes into the target
cell cytosol. Granzymes then initiate apoptosis of the
target cell, and the cytotoxic lymphocyte detaches from
the dying cell (indicated by the arrow; part e) and can interact
Dying cell with another target cell to carry out serial killing (not shown).

granule protein that oligomerized (at neutral pH and Knockout studies in mice showed that loss of per-
in the presence of Ca2+) into membrane-spanning pores forin abolished granule-dependent target cell death;
Nature Reviews | Immunology
and named it perforin. Crucially, these authors also these data supported a role for perforin as the key
noted that perforin pores and the complement mem- enabler of granzyme-induced apoptosis (reviewed in
brane attack complex had marked similarity by electron REFS 31,32). Conversely, disrupting single granzyme
microscopy. Around the same time, several granzymes genes produced far subtler phenotypes, which high-
were identified as abundant granule constituents17–20, and lighted some redundancy in the functions of gran-
they were later shown to induce target cell apoptosis in zymes33–35. Great effort has been made to understand
synergy with otherwise non-lethal amounts of perforin how granzymes induce specific death signalling path-
in vitro 21–28. The lipid-disrupting toxin granulysin — ways and to understand their role in infectious dis-
which is secreted by human T cells — was discovered eases, inflammation and cancer. More recently, the
soon after 29, but its orthologues exist only in ruminants discovery that congenital perforin deficiency causes
(BOX 1). Detailed biochemical studies characterized these the fatal human hyperinflammatory disease FHL36
perforin-enriched and granzyme-enriched granules as has reignited interest in perforin and demonstrated its
a subset of ‘secretory lysosomes’ with an electron-dense involvement in a far broader range of pathologies than
core surrounded by small vesicles17–19,30. previously suspected.

NATURE REVIEWS | IMMUNOLOGY VOLUME 15 | JUNE 2015 | 389

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

Box 1 | Granulysin
As discussed below, this and other lines of evidence sug-
gest that even mild changes in perforin activity, as occurs
Granulysin is a small, saposin-like, antibacterial toxin that is expressed by natural in carriers of hypomorphic PRF1 mutations, might
killer (NK) cells and CD8+ T cells of primates and ruminants, but not rodents166. contribute to immune‑based pathology over time.
Unlike co‑secreted perforin and granzymes, it kills intracellular bacteria — such as
Listeria monocytogenes and Mycobacterium tuberculosis — by rupturing their
Safe trafficking and storage of perforin in granules. How
membranes167. Similar to granzymes, the delivery of granulysin into infected cells
is dependent on perforin; granulysin diffuses through perforin pores and kills
do killer cells safely express and store large amounts of
pathogenic bacteria that are escaping from the endolysosomal compartment of highly toxic proteins in their granules? Granzyme traf-
dying target cells168. To investigate granulysin delivery into a target cell and the ficking and activation are well understood45–47. By con-
possible role of perforin and granzymes in its antibacterial activity, a recent study trast, the trafficking of perforin and management of its
used mice that were transgenic for granulysin. The results of this study confirmed inherent toxicity are less well characterized. The activity
earlier reports that intracellular antibacterial activity of granulysin in vivo depends of perforin is highly dependent on pH and Ca2+; perforin
on perforin; however, wild-type mice that were transgenic for granulysin cleared is active at the neutral pH of the immunological synapse,
L. monocytogenes as efficiently as granzyme-deficient animals that were also but it is inactive under acidic conditions (that is, less than
transgenic for granulysin169. In vitro experiments have shown that granulysin can pH 6)48–50. Therefore, the acidic secretory granule (which
deliver lethal levels of granzymes into bacteria, suggesting that perhaps under
is at pH 5) is a safe storage compartment for perforin,
certain conditions these effector proteins may synergize in vivo. Despite showing
little cytotoxicity towards mammalian cells under controlled conditions, high
which only becomes active upon secretion. However,
levels of extracellular granulysin might contribute to Stevens–Johnson syndrome until recently, it was unclear how perforin is safely pro-
and toxic epidermal necrolysis, both of which result from immune responses to cessed and trafficked from the Ca2+-rich and pH-neutral
certain drugs and allergens170. In these conditions, the recruitment and activation environment of the endoplasmic reticulum (ER) to the
of NK cells and CD8+ T cells within epidermal blisters leads to the secretion of granules. One hypothesis is that perforin is synthesized
granulysin, which causes potentially life-threatening skin necrosis170. as an inactive precursor, with an inhibitory carboxy‑
terminal dodecapeptide that blocks Ca2+ binding to the
C2 domain51. However, this is inconsistent with the find-
Regulation of perforin expression in cytotoxic lympho­ ing that recombinant perforin with an intact C terminus
cytes. Stimulation with cytokines and/or specific has the same activity and stability as perforin that lacks
antigen causes naive CD8+ CTLs and NK cells to dif- the C terminus52. Furthermore, the X‑ray crystal struc-
ferentiate into efficient killers, which require the rapid ture of the perforin monomer showed that the C termi-
synthesis, safe trafficking and storage of large amounts nus is located far from the Ca2+-binding region of the
of perforin and granzymes. A major advance in under- C2 domain, making the proposed mechanism unlikely53,54.
standing PRF1 transcription in T cells came from the dis- Instead, it has been suggested that the distal end of the
covery of ten DNase I hypersensitivity sites in its 150‑kb C‑terminal region encodes a highly efficient ER export
long locus control region, which signifies an open chro- motif that is required for protein delivery from the ER to
matin configuration that enables gene transcription37,38. the granules52. Deletion of the last 12 amino acids or even
Subsequently, some of the sites were shown to bind to a mutation of the last C-terminal residue, tryptophan,
runt-related transcription factor 3 (RUNX3) — which results in the retention of functional and stable perforin
is necessary for PRF1 expression — and to an addi- in the ER and the death of cytotoxic lymphocytes owing
tional PRF1 transcriptional regulator, eomesodermin to perforin pore-forming activity52. This is further sup-
(which is encoded by EOMES)38–40. More recent stud- ported by the observation that cytotoxic cells that express
ies have shown that PRF1 is also post-transcriptionally an inactive perforin variant — which cannot properly
regulated. MicroRNAs (mi­RNAs) bound to the 3ʹ untrans- bind to Ca2+ — are protected from the toxic effects of
lated region of Prf1 and PRF1 mRNAs directly down- mutations of the C‑terminal region52. Accordingly, an
regulated their translation in mouse and human CTLs explanation for cell survival during perforin synthesis is
and NK cells, whereas other mi­RNAs indirectly down­ that it is rapidly exported from the ER, thus minimizing
regulated perforin expression by reducing the expression the concentration of active perforin in the ER at any one
of eomesodermin41–44. Depending on the mode of acti- time. Furthermore, it is notable that N‑linked glycans are
vation, several mi­RNAs can suppress PRF1 translation essential to enable the delivery of perforin from the Golgi
including: miR‑27a* (REF. 42), miR‑30e and miR‑378 in to secretory granules. The exact carrier (or carriers) of
human NK cells44; miR‑150 in mouse NK cells41; and nascent perforin are unknown52; however, this process
miR‑139 and miR‑342 in human and mouse CTLs43. may involve constitutive protein trafficking machinery
Although modulation of miRNA expression and activ- (such as lysosome-­associated membrane glycoprotein 1
ity in cytotoxic lymphocytes in vitro markedly reduced (LAMP1)55 or the ‘leaky’ mannose 6‑phosphate receptor).
the production of perforin, the overall reduction in Finally, it is important to note that the perforin C‑terminal
cytotoxicity was relatively mild (approximately two- peptide is removed; however, the precise purpose of this
Toxic epidermal necrolysis
A rare life-threatening skin fold)41,42, implying that residual perforin levels were processing (if any) remains unclear.
disease in which the dermis sufficient to support granzyme delivery. Despite this,
detaches from the epidermis. adoptively transferred miR‑150‑deficient NK cells Granule exocytosis and the immunological synapse.
This disease often leads to with enhanced perforin expression rejected trans- CTLs respond to T cell receptor-mediated recognition
sepsis and death. It is also
known as Lyell syndrome;
planted tumours better than wild-type NK cells 41, of peptide–MHC complexes, and NK cells respond to
one of its forms is called and miR‑27a*-­overexpressing NK cells (with reduced the balance of stimulatory versus inhibitory signals
Stevens–Johnson syndrome. perforin levels) were less effective than control cells42. that are expressed by target cells. In both cell types,

390 | JUNE 2015 | VOLUME 15 www.nature.com/reviews/immunol

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

recognition of the target gives rise to signalling events of the granzyme B inhibitor serine protease inhibitor B9
that result in a pronounced increase in cytosolic Ca2+ (serpin B9)68. Knockout studies show that mice lack-
concentration. The microtubule-organizing centre (MTOC) ing serpin B9 have much more CTL death during the
rapidly moves towards the target cell, and then cyto- response to infection than wild-type mice69. Together,
toxic granules migrate along the MTOC, fuse with the these data suggest that although the reservoir of ser-
presynaptic membrane and release their cargo into the pin B9 in CTLs may be exhausted during a prolonged
synaptic cleft 1,4,56–58 (FIG. 1). It was recently proposed immune attack, the presence of this inhibitor may
that granule clustering at the presynaptic membrane provide cytotoxic cells with limited protection from
might precede MTOC polarization and granule dock- perforin and granzyme toxicity during killing.
ing, which suggests some plasticity or redundancy in
mechanisms of cytotoxic immunological synapse forma- Death of the target cell
tion59. Consistent with this notion, human disorders that Structural basis of perforin oligomerization and pore
are caused by cytoskeletal abnormalities show varying formation. Until recently, the molecular basis for
cytotoxic impairment — for example, individuals with perforin pore formation remained unclear. The X‑ray
Wiskott–Aldrich syndrome (which is caused by disordered crystal structures of several perforin-related proteins70–72
F‑actin remodelling) show only mildly defective NK cell showed that these proteins are homologous in their
cytotoxicity189. However, disorders that are associated pore-forming domain (known as the membrane attack
with mutations in adhesion molecules — such as leuko­ complex and perforin (MACPF) domain) to the bac-
cyte adhesion deficiency 1 (LAD1), LAD3, hyper-IgE terial cholesterol-dependent cytolysins (CDCs). This
recurrent infection syndrome (which is due to a defi- unexpected relationship provided several mechanistic
ciency in the gene encoding dedicator of cytokinesis 8 clues about perforin function. CDCs initially bind to and
(DOCK8)) and a deficiency in the gene encoding the oligomerize on the membrane surface into a pre-pore of
guanine nucleotide exchange factor VAV1 — all cause 20–50 monomers (FIG. 2a–c). Following pre-pore assem-
an incomplete loss of lymphocyte cytotoxicity (for exam- bly, two small clusters of α‑helices (transmembrane
ple, see REFS 60,61). By contrast, failure to deliver per- helical region 1 (TMH1) and TMH2) in each mono-
forin to the immunological synapse owing to biallelic mer unwind and insert into the membrane as a pair
mutations in UNC13D (which encodes MUNC13‑4), of amphipathic β‑hairpins. Therefore, a full pore com-
STX11 (which encodes syntaxin 11) and STXBP2 (which prises a large β‑barrel of 80–200 strands73–75. Through an
encodes syntaxin-binding protein 2; also known as alternative but similar mechanism, several studies have
MUNC18‑2) results in the development of FHL type 3, shown that CDCs, and most likely perforin itself, form
FHL type 4 and FHL type 5, respectively 4,5. These pro- incomplete pores (arc-like structures) that can penetrate
teins regulate cytotoxic granule docking, priming or and damage the membrane76–78.
fusion to the presynaptic membrane; consistent with On the basis of these structural data, it was suggested
these loss‑of‑function consequences, data indicate that that perforin and other MACPF domain-containing
their functions are non-redundant (discussed below). proteins form pores in a CDC-like manner 70,71. This pre-
diction supports the finding that mutations in TMH1 or
Protection of cytotoxic cells following perforin secre­ TMH2 — which are essential for membrane piercing —
tion. The notion that killer cells are resistant to the toxic frequently cause perforin dysfunction and disease53,79.
effects of perforin and granzymes following granule Furthermore, the crystal structure of perforin itself,
exocytosis is remarkable, and how they do so remains together with the 28 Å resolution cryo-electron micro­
unclear 62–65. Without this resistance, it would not be pos- scopy structure of the pore, provided broad support for
sible for a single cytotoxic lymphocyte to kill multiple the idea that perforin forms pores through a CDC-like
Microtubule-organizing target cells, and an infected host would need to gener- mechanism53 (FIG. 2d–f).
centre ate at least as many cytotoxic cells as infected cells54,66.
(MTOC). An intracellular
However, cytotoxic lymphocytes are not intrinsically Access of granzymes to the target cell cytosol. Early elec-
structure from which
microtubules originate. resistant to granule exocytosis: CTLs were readily killed tron microscopy 10 and recent X‑ray and cryo-electron
In cytotoxic lymphocytes, when they presented cognate peptide–MHC class I to microscopy studies have shown that perforin pores (with
the MTOC moves towards the identical, non-peptide-pulsed clonogenic CTLs, which an internal diameter of 16–22 nm) should easily enable
immunological synapse. went unscathed90. Recently, it was proposed that, upon granzymes (with a diameter of approximately 4 nm) to
Wiskott–Aldrich syndrome
degranulation, the NK cell surface becomes ‘coated’ diffuse across the plasma membrane into the cytosol of
A life-threatening X‑linked with LAMP1, which protects it against newly released the target cell53,80. However, several studies using puri-
immunodeficiency that is perforin67. However, LAMP1‑deficient and wild-type fied perforin could demonstrate only small solutes such
caused by mutations in the NK cells were equally cytotoxic in long-term in vitro as Ca2+ or Na+ passing through the pores, whereas the
gene encoding the Wiskott–
assays, which suggests that serial cytotoxicity is not larger fluorescent molecule propidium iodide seemed to
Aldrich syndrome protein. The
condition is characterized by reduced in the absence of LAMP1 (REF. 67). It is still be excluded from cells that were simultaneously under-
thrombocytopaenia with small possible that several such factors may contribute to going perforin-dependent apoptosis81,82. This paradox
platelets, eczema, recurrent cytotoxic lymphocyte protection; a role has been pro- provided support to the ‘endosomolysis’ hypothesis,
infections caused by posed for cathepsin B63, but the survival of cathepsin which proposed that small perforin pores or rents in the
immunodeficiency and an
increased incidence of
B‑deficient cytotoxic cells following encounters with tar- membrane are internalized together with granzyme mol-
autoimmune manifestations get cells is no different from wild-type cytotoxic cells64. ecules into clathrin-dependent and dynamin-dependent
and malignancies. Finally, it has been noted that CTLs contain high levels endosomes by the target cell. Approximately 15 minutes

NATURE REVIEWS | IMMUNOLOGY VOLUME 15 | JUNE 2015 | 391

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

later, perforin pores are thought to assemble within the related toxins can also form curved membrane disrup-
endosome (or a larger endosomal structure known as tions (‘arcs’) re‑enlivened the debate77,78. This situation
the ‘gigantosome’) to form channels that are large raised problems for both major hypotheses. On the one
enough for granzymes to exit into the cytosol81–86. hand, how did granzymes enter an intact target cell when
Structural studies with perforin in liposomes found entry of propidium iodide could not be demonstrated
no evidence of pores with a ‘small’ lumen, and full pore through transmembrane pores? On the other hand, how
structures comprising fewer than 16 monomers were could endosomolysis occur when the endosomal envi-
extremely rare. However, the finding by independ- ronment (with an acidic pH and extremely low Ca2+
ent groups that recombinant perforin and structurally concentration87) seems unsuitable for perforin action49,65?

a CDC monomer b CDC pre-pore c CDC pore

Lipid
bilayer

d Perforin domain structure e Perforin monomer f Perforin pore


MACPF domain (pore forming) EGF-like domain C-terminal domain
21 112–165 240–296 371 377 397 397 412 526 542

TMH1 TMH2 413 525


C2 domain
(Ca2+ and
membrane
binding)

Figure 2 | Structural transitions of cholesterol-dependent cytolysins attack complex and perforin (MACPF) domain contains two membrane-
and perforin. a | The structure of a cholesterol-dependent cytolysin (CDC) penetrating regions (TMH1 and TMH2) and is followed
Nature by |an
Reviews epidermal
Immunology
in its monomeric state, with the pore-forming ‘warhead’ (blue), the central growth factor (EGF)-like domain. The C‑terminal domain (comprising a
sheet (red) and the transmembrane helical (TMH) regions (orange). b | CDCs short disulfide-bonded region) contains the Ca 2+ -binding and
initially interact with the lipid bilayer of the cell membrane through the membrane-binding C2 domain. e | The structure of mouse perforin. The
carboxy-terminal immunoglobulin-like domain (yellow; lower panel). Here, pore-forming warhead is homologous to that of CDCs; the central, highly
monomers oligomerize to form a distinct pre-pore intermediate. A surface- twisted sheet is shown in red, the TMH regions are shown in orange and the
contoured view (upper panel) shows the proposed structure of the pre-pore. remainder of the MACPF domain is shown in blue. It is not known whether
c | CDC pre-pores undergo a marked collapse, which results in insertion of perforin forms a pre-pore before membrane insertion. f | A cryo-electron
the two pore-forming regions (orange) into the membrane as a large microscopy reconstruction of a perforin pore. The diameter of the pore is
β‑barrel (lower panel). Modelling studies suggest that the linker domain calculated to be approximately 150 Å and therefore would easily enable the
(green) rotates during pre-pore collapse175. The β‑strands in the final pore passage of granzyme B, which is approximately 50 Å in diameter (not
are thought to adopt a slightly tilted (off-vertical) architecture (upper shown). All of the images in this figure were prepared specifically for this
panel)75. d | The domain structure of perforin. The pore-forming membrane Review based on data from REFS 53,74,75,175.

392 | JUNE 2015 | VOLUME 15 www.nature.com/reviews/immunol

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

To address some of these issues, we recently re‑­ presence of recombinant perforin65. In the context of the
investigated whether propidium iodide could pass immunological synapse, clear morphological changes
through perforin pores. In previous studies, propidium (such as cell rounding and membrane blebbing) within
iodide had arbitrarily been used at 0.5–1.5 μM82,85,86 and 10 minutes of the propidium iodide influx indicated
occasionally at 15 μM78,88. As the diffusion of a solute granzyme-induced caspase activation; this occurred
through a pore is proportional to its concentration but much faster than predicted by endosomolysis. Human
inversely proportional to its radius89, these conditions and mouse cytotoxic cells all produced similar results
favoured the diffusion of Ca2+ (which is typically used at a against diverse target cells90, indicating that the function
concentration of 0.5–2 mM) into cells over that of propid- of perforin is conserved across species.
ium iodide by several orders of magnitude. We found that Taken together, we argue that these data do not sup-
simply increasing the concentration of propidium iodide port the endosomolysis hypothesis. However, these
to 100 μM was sufficient to demonstrate obvious free dif- findings do not preclude a role for arcs in permitting
fusion of propidium iodide across the plasma membrane rapid granzyme delivery in the context of the immuno-
in the presence of recombinant perforin65. Consistent with logical synapse. Indeed, it has been proposed that arcs
these findings, we repeated these experiments in the con- may transition into pore-like structures92, and recent
text of physiologically relevant immunological synapses studies on the CDC suilysin indicate that arcs that form
formed by human or mouse cytotoxic cells; these con- in close proximity may develop even larger membrane
ditions led to a pronounced entry of propidium iodide ruptures77. Therefore, it is possible that perforin — the
into the target cell near to the immunological synapse, secretion of which is concentrated onto a confined area
which indicates the formation of large perforin pores of the target cell membrane — may form a mixture of
(FIG. 3). This influx of propidium iodide occurred within arcs, pores and coalesced arcs that increase the diffusion
60–80 seconds of Ca2+ flux into the cytotoxic cell cytosol, of granzymes into the target cell. At the same time, exo-
which is the stimulus that signals degranulation65,90. The cytic membrane repair of the target cell at the immuno­
potentially lytic effect of perforin on the target cell was logical synapse65 seems to be sufficient to prevent lysis
rapidly (within approximately 30 seconds) countered by while enabling enough time for granzymes to enter the
efficient exocytic membrane repair mechanisms91,93 that target cell and initiate apoptosis65,90.
prevented osmotic lysis, but still enabled the movement of Finally, it is important to note that many of the
granzymes into the target cell cytosol65. studies that address perforin pore formation in the con-
The authenticity of these kinetics was evident from text of the target cell rely (by technical necessity) on
the direct cleavage of BH3‑interacting domain death applying recombinant perforin to cells78, liposomes, lipid
agonist (BID) — which is the cytosolic substrate of gran- monolayers or lipid-coated chips53. Therefore, although
zyme B — within 2 minutes of adding it to cells in the these in vitro studies are informative, it is important

00:00 10 μm 03:25 04:55

05:25 06:15 06:45


Figure 3 | Perforin pore formation in action. The cytotoxic T lymphocyte (CTL) was labelled with a Ca2+ fluorophore
and added to adherent target cells in medium that was supplemented with 100 μM of propidium Nature Reviews
iodide . After
65,90 | Immunology

3–5 minutes, CTL degranulation was observed, as measured by an increase in green fluorescence (which reflected Ca2+
influx). Approximately 1 minute and 30 seconds later, the target cell membrane was punctured, as shown by a focal influx
of propidium iodide into the cytosol, where it fluoresced red following binding to RNA. The propidium iodide then spread
throughout the target cell, which started to acquire a round morphology (a sign of caspase activation). Finally, the CTL
detached from the rounded target. The whole process from degranulation to detachment was remarkably rapid and
took less than 5 minutes. These images have not been previously published but are analogous to those in REFS 65,90.

NATURE REVIEWS | IMMUNOLOGY VOLUME 15 | JUNE 2015 | 393

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

Table 1 | Hierarchical granzyme-mediated cell death pathways


Cell type Cytotoxicity* Morphology Phosphatidylserine Propidium Initiation Refs
(functional exposure‡ iodide phase§
granzymes) uptake‡ (minutes)
Wild-type +++++ Classical apoptosis Early Late <7 108
NK cells (secondary
(predominantly necrosis)
granzyme B)
GzmA−/− +++++ Classical apoptosis Early Late <7 108
NK cells (secondary
(predominantly necrosis)
granzyme B)
GzmB−/− NK cells ++ Athetosis (worm-like Simultaneous Simultaneous ~15 108
(predominantly morphology, (secondary necrosis) (secondary
granzyme A) elongated cells and no necrosis)
membrane blebbing)
GzmA−/−GzmB−/− +/− Similar to apoptosis Late Late Typically 108,
CTLs (other but slower (secondary >20 176
granzymes) necrosis)
CTL, cytotoxic T lymphocyte; Gzm, granzyme; NK, natural killer. *The relative amount of cell death in a typical 4‑hour assay, in
which ‘+’ indicates the death of approximately 20% of target cells. +/– indicates the death of <20% of target cells. ‡Propidium
iodide used at 1 μM, as used in conventional cell death assays that measure annexin V binding and propidium iodide uptake.
Using propidium iodide at 100 μM demonstrates a much earlier breach of the target cell plasma membrane at the time of perforin
pore formation. §Defined as the time elapsed from conjugation to target cell rounding.

to note that they may not accurately reflect the in vivo Granzymes and cell death. Granzymes induce target
situation. For example, the concentration of perforin cell death through diverse, non-redundant pathways
and granzymes at the immunological synapse and the that function in a complementary and hierarchical
structures that these molecules form within the synapse manner (as recently reviewed in REF. 98; TABLE 1). In an
remain to be understood. Recent advances in high- evolutionary context, this reflects a balance between
resolution electron tomography may provide key three mechanisms: first, the immune mechanisms that
insights into these issues in the future. induce suicide of infected cells; second, the opposing
pathways that are activated by intracellular patho-
Lethal and non-lethal granzyme function gens to provide a safe haven (for example, protection
Since the mid‑1990s, the increased interest in how gran- from neutralizing antibodies); and third, the cellular
zymes induce cell death led to diminished research on machinery of replication. Granzyme B is the most pow-
the non-cytotoxic functions of granzymes. Granzymes erful pro-apoptotic granzyme, as its ability to cleave
are characterized by their individual substrate speci- target cell proteins at sites after selected aspartate
ficity and the gene cluster by which they are encoded. residues mimics the caspases. This ‘imposed’ death
Despite being broadly labelled as trypsin-like or is rapid and effective; target cells that are exposed to
chymotrypsin-like, granzymes induce tightly regulated nanomolar amounts of recombinant granzyme B and
signalling (but not indiscriminate protein digestion), an otherwise innocuous dose of recombinant perforin
as exemplified by their highly orchestrated activation (which provides access for granzyme B to the target
of caspase-driven cell death pathways21,94. Granzymes cell cytosol) die within 5–8 minutes by apop­tosis65,99–101.
also regulate inflammation95 and can directly subvert Granzyme A cleaves proteins at sites after basic amino
intracellular infections by interfering with viral bio- acids and activates a slower form of cell death through
synthetic pathways96. However, as much of the work different substrates (see below). Although most other
on granzymes has been done in vitro, there are gaps in granzymes also activate cell death in vitro, many of
our understanding of the potential pathophysiological the assays do not include granzyme delivery across a
relevance of interesting laboratory findings. Whereas bona fide immunological synapse, and thus it remains
congenital human perforin deficiency is well character- unclear whether granzymes other than granzyme A and
ized, no known human disease arises from granzyme granzyme B have generic cytotoxic activity (BOX 2).
deficiency. Papillon–Lefèvre syndrome (which is associ- There is still controversy about what the physio­logical
ated with cathepsin C deficiency) was proposed to cause substrates of granzymes are, but recent findings have
granzyme inactivity, as all pro-granzymes were thought resolved several contentious issues. The most impor-
to require cathepsin C for amino-terminal process- tant finding was reported by several groups showing
ing 46,97. Although this is the case for granzyme A and that there are species-specific differences in the fine sub-
the myeloid proteases cathepsin G and elastase, gran- strate specificity of orthologous granzymes102–104. It took
zyme B is also activated by cathepsin H45, which explains a decade to reach a consensus on whether granzyme B
why neutrophils of patients with Papillon–Lefèvre syn- preferentially processes pro-caspases directly or activates
drome are poorly bactericidal, but NK cell cytotoxicity the mitochondrial cell death pathway by cleaving BID.
remains intact. In fact, human granzyme B very efficiently activates the

394 | JUNE 2015 | VOLUME 15 www.nature.com/reviews/immunol

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

Box 2 | Which granzymes are authentic inducers of cell death?


a form of cell death that is dependent on an intact actin
cytoskeleton and is marked by a writhing (‘twisting
Many granzymes have been reported to cause cell death in vitro, but there are many and turning’) morphology of the dying cell108. Mouse
inconsistencies among the findings. For example, three groups (including ours) have
CTLs and NK cells deficient in granzyme B phenocopy
reported that granzyme M kills target cells in vitro but through very different and
athetosis that is induced by recombinant granzyme A.
incompatible mechanisms133,171–174. These studies used different methods to deliver
recombinant granzyme M to target cells (perforin, synthetic detergent or adenovirus). Athetosis does not occur when both granzyme A and
The reports differed on whether activated caspases were required for cell death, and granzyme B are absent, confirming granzyme A as the
they identified various substrates or ligands, including inhibitor of caspase-activated inducer of athetosis. However, human granzyme A has
DNase (ICAD; also known as DFFA), the cytoskeletal proteins ezrin and α‑tubulin, minimal cytotoxic activity even at high concentrations
FAS-associated death domain protein (FADD) and topoisomerase IIα. Granzyme (in the micromolar range), which also suggests that
M‑deficient mice have no defect in generic cell death pathways114, even when different species have been shaped by their microbial
possible masking effects of granzyme A and granzyme B are absent108. Overall, the environment and have evolved their own ‘customized’
notion that granzyme M has authentic cytotoxic activity remains unproven. range of granzyme-dependent cell death pathways109.
In addition to the possibility that some granzymes might have redundant functions,
In the absence of both granzyme A and granzyme B,
there are many reasons why in vitro findings may not translate into the in vivo setting.
mouse CTLs and NK cells induce an even slower form of
Granzyme substrate specificity can vary between species, but many studies combine
cells and reagents from disparate species. Moreover, in examinations of whether cell death than that activated by granzyme A alone. The
a substrate is genuinely processed, robust kinetic analyses are rarely performed. morphology of the double-deficient dying cells resem-
Commonly used commercially available reagents, such as monoclonal antibodies, bles that of cells undergoing apoptosis, but phosphatidyl­
can give misleading results owing to a lack of specificity175. Granzyme concentrations serine exposure on the dying cell is greatly delayed110.
used in vitro vary from low nanomolar to high micromolar, but it is not known how Loss of this signal for phagocytosis reduces uptake of
these levels relate to cell death induced by intact cytotoxic T lymphocytes (CTLs) the dying cells by professional antigen-presenting cells
or natural killer (NK) cells. Finally, contaminants other than granzymes, such as and reduces their immunogenicity 110. It is likely that this
lipopolysaccharide or proteases, can markedly influence results. Indeed, we have ‘residual’ cell death is activated by mouse granzyme M
found that trypsin (which is commonly used to harvest adherent cells) caused
and by human granzyme M and granzyme H111–113, and
dose-dependent death of HeLa cells when delivered with recombinant perforin
disrupting these minor pathways in otherwise normal
(J.A.T. and K. A. Brown, unpublished observations).
Thus, caution is advised when extrapolating in vitro results to intact animals or CTLs has little effect on cell death in vitro or in intact
humans without corroborating genetic evidence. Perforin-deficient CTLs and NK cells mice114. Genetic evidence supporting a pro-apoptotic
from both mice and humans have severely reduced cytotoxicity, but genetic evidence role for granzyme M is still lacking 114, but several cell
for cytotoxic functions has only been provided for mouse granzyme A and granzyme B. death pathways can potentially be activated in vitro
(BOX 2). This lack of phenotype or specific function for
granzymes other than granzyme A and granzyme B has
BID-dependent, B cell lymphoma 2 (BCL‑2)‑inhibitable led to them being referred to as ‘orphan’ granzymes,
pathway, but mouse granzyme B preferentially activates although granzyme H is known to be crucial for defence
caspases directly 102. Both mouse and human gran- against human adenovirus V, as it disrupts capsid
zyme B are polymorphic, with variability of residues synthesis and viral assembly 115.
in mouse granzyme B having an effect on substrate Studies have catalogued many potential granzyme
choice102. Owing to strong founder effects, there is only substrates to decipher the ‘proteome’ by mixing cell
one granzyme B allele in all inbred laboratory mouse lysates with an arbitrary concentration of recombinant
strains (designated granzyme BP)105. However, C57BL/6 granzyme116. One database currently lists 1,984 granzyme
mice that are congenic for the most divergent ‘wild’ B substrates, which is more than for all but three other
allele (designated granzyme BW) showed a profound proteases117. Although unbiased, this approach does not
susceptibility to mouse cytomegalovirus (MCMV) due explain how granzyme B could ever encounter its sub-
to unrestrained viral replication in the liver 106. Failure to strates physiologically, thus drawing some criticism116
control MCMV was due to BCL‑2‑like pro-survival viral (BOX 2). By contrast, functional genomics approaches
proteins that blocked killing by granzyme BW‑expressing that are designed to identify whole pathways, and there-
CTLs, but not by granzyme BP‑expressing CTLs. By con- fore some ‘indirect’ granzyme targets, are now providing
trast, granzyme genes that are closely linked to the gene insight. This approach showed that granzyme B‑mediated
encoding granzyme B have minimal poly­morphism, processing of BID is controlled epigenetically; loss of
indicating that they are less crucial for defence against the transcriptional cofactors histone acetyltransferase
specific viruses than granzyme B, and that they have KAT2B (also known as PCAF) or transcriptional adap-
other roles105. The pro-apoptotic role of granzyme B tor 3 markedly reduced target cell death due to aberrant
was also confirmed in granzyme B‑deficient mice that trafficking of uncleaved BID118. This powerful approach
have similar susceptibility to MCMV as granzyme can identify key substrates in granzyme B signalling that
BW‑congenic mice35,106,107. are not cleaved by granzyme B or any other protease118.
In granzyme B‑deficient mice, which lack the rapidly
acting granzyme B-mediated cell death pathway, a novel Granzymes in inflammation and infection. A long hiatus
granzyme A‑induced cell death pathway was revealed108. separated the first two studies reporting that granzymes
Mouse granzyme A acts more slowly than granzyme directly influence inflammation109,119 by releasing pro-
B through a distinct, caspase-independent mechanism; inflammatory cytokines from monocytes, which focused
recombinant mouse granzyme A that is delivered to tar- initially on granzyme A and granzyme B, but more
get cells by purified perforin causes athetosis, which is recently also on granzyme K and granzyme M120–124.

NATURE REVIEWS | IMMUNOLOGY VOLUME 15 | JUNE 2015 | 395

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

Granzyme M‑deficient mice resist lethal Toll-like recep- Perforinopathies: defects of granule exocytosis
tor 4 (TLR4)-mediated endotoxicosis owing to reduced The term perforinopathy denotes the range of autosomal-­
circulating levels of interleukin‑1α (IL‑1α), IL‑1β and recessive, immune-mediated diseases that are caused by
tumour necrosis factor 124. This study did not examine insufficient perforin delivery to the immunological syn-
whether granzyme M directly processes these cytokines, apse, due either to PRF1 mutations or to impaired granule
but granzyme B can directly cleave pro‑IL‑1α close to exocytosis2. In approximately half of the cases in which
sites that are recognized by calpains and elastin125, and PRF1 is not mutated139, UNC13D, STX11 or STXBP2 are
direct processing of IL‑18 has also been proposed126,127. the genes that are most commonly mutated139,141–143. The
Granzyme K might also function by promoting the bind- corresponding proteins (MUNC13‑4, syntaxin 11 and
ing of lipopolysaccharide to CD14; surprisingly, this STXBP2, respectively) all take part in secretory granule
effect persisted with enzymatically inactive granzyme K, transport to the immunological synapse and/or mem-
raising questions about the mechanism underlying this brane fusion; therefore, deleterious mutations in these
process128. Given their potential role in inflammation, genes impair perforin secretion and the killing activity
therapeutic targeting of granzymes has been considered. of cytotoxic cells139,141–142. Perforinopathies are classified
For example, the concept of treating septic shock with into three groups: acute (also known as early-onset ‘clas-
granzyme inhibitors was supported by a recent study sic’ FHL, which invariably occurs before 2 years of age),
showing that granzyme A‑deficient mice survived lethal subacute (such as atypical FHL or another pathology,
infection and did not show compromised CTL-mediated which usually occurs after 5 years of age) and chronic (in
killing of infected cells129. which heterozygosity can predispose to disease in adults);
In addition to mediating cell death and depriving the chronic disease grouping has not yet been well char-
viruses of the cellular processes that are necessary for acterized and requires further study. Complete loss of
their survival and replication, granzymes can directly perforin function or secretion almost invariably results in
limit the release of viable virus following the demise of FHL by 12 months of age (with a median of 3–9 months),
the cell96. Thus, mouse CTLs that are specific for influ- whereas hypomorphic PRF1 alleles result in delayed
enza virus upregulate the expression of granzyme K130, FHL, or other inflammatory or neoplastic disorders32,144.
which interferes with viral replication by cleaving As classic FHL has been previously reviewed5, we
host importin‑α1 and importin‑β, and by preventing highlight one recent advance that causally links failed
the nuclear transport of viral nucleoprotein121. In an target cell death with the fatal ‘cytokine storm’, which
analogous manner, heterogeneous nuclear ribonucleo­ is the hallmark of FHL. It was shown that detachment
protein K (hnRNPK) is cleaved by granzyme M to pre- of cytotoxic lymphocytes from their targets requires
vent the replication of human CMV123, and eukaryotic a caspase-dependent signal from dying cells140. In the
initiation factor 4γ3 (eIF4γ3) is cleaved by granzyme B absence of perforin, or if caspases are inhibited in the
to stop viral protein translation in dying cells131. target cell, the duration of engagement of cytotoxic
cells with the target cell is increased by approximately
Granzymes in tumour immunity. As described below, fivefold. A failure to detach results in many successive
perforin-deficient mice and humans are susceptible to rounds of Ca2+ flux into the cytotoxic cell, which trig-
cancer. As perforin-dependent cell death is mediated by gers substantial secretion of pro-inflammatory cytokines
granzymes, some granzyme substrates are likely to be and chemokines. In turn, secreted interferon‑γ causes
important for clearing transformed cells. Thus, cleav- hyperactivation of the myeloid cell compartment and
age of hnRNPK by several granzymes (including gran- stimulates macrophages to secrete IL‑6, which is a major
zyme H, granzyme K and granzyme M) reduces tumour instigator of the fatal systemic inflammation of FHL140.
cell viability 123,132. Granzyme M also targets topoiso­ Below, we focus on subacute and chronic perforin­
merase IIα113 and processes human (but not mouse) opathies caused by PRF1 mutations.
FAS-associated death domain protein (FADD), caus-
ing FADD dimerization and increasing pro-caspase 8 Partial perforin deficiency increases cancer risk. The role
processing 133,134. Some granzyme-mediated cell death of perforin in surveillance of carcinogen-induced cancer
pathways are subverted by tumour cells. For example, or transplanted tumours in mice was demonstrated many
transformed cells may target granzyme B for degrada- years ago145,146, but of greater physiological relevance was
tion when undergoing autophagy 135. Granzyme B can the finding that approximately 60% of Prf1‑deficient
also contribute to the demise of CTLs that are internal- mice spontaneously developed B cell lymphomas with
ized by tumour cells through a process of engulfment age147. Broadly speaking, congenital or acquired immune
known as emperitosis136. Cancer cells often overexpress deficiency has long been associated with increased
BCL‑2 or related pro-survival molecules that protect cancer risk in humans, particularly those that are caused
against myriad metabolic insults and block the mito- by viruses148, but a direct link with impaired cytotoxic
chondrial death pathway that is activated by human lymphocyte function has been difficult to prove. We pre-
granzyme B137. Pertinent to tumour cell survival, pro- viously reported that carriers of biallelic PRF1 mutations
Emperitosis cessing of BID by human granzyme B is long-lived (with have a bimodal age distribution32, and patients presenting
A process by which a a half-life greater than 16 hours), which potentially offers in the later peak (that is, at 5–20 years of age) typically
granzyme B‑expressing
cytotoxic lymphocyte is
sufficient opportunity to administer pharmaco­logical have subacute perforinopathy 144. Approximately 40%
taken up by the target cell BCL‑2 antagonists, such as ABT‑737, to restore the of patients present with various haematological cancers
and killed by apoptosis. therapeutic benefit of granzyme B‑expressing CTLs138. (but generally not those with a known viral cause),

396 | JUNE 2015 | VOLUME 15 www.nature.com/reviews/immunol

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

Table 2 | Association of the Ala91Val polymorphism of perforin with human immunopathologies*


Perforin mutations Total FHL cases Haematological Other pathology Healthy
(late onset)‡ malignancy
Ala91Val + null§ 11 (7) 3 1 1
Ala91Val + Ala91Val|| 7 (6) 6 – 5
FHL, familial haemophagocytic lymphohistiocytosis. *Data from REFS 163,165,177–188. The number of cases of late-onset FHL

of the total number of FHL cases is shown in brackets. §Indicates co‑inheritance of Ala91Val with perforin-inactivating frameshift,
nonsense or missense mutations. ||Indicates homozygous Ala91Val polymorphisms.

approximately 20% of patients present with a life- population) perforin polymorphism C>T272 (which
threatening viral infection (commonly Epstein–Barr causes the Ala91Val substitution) provides an oppor-
virus149) and the remainder of patients suffer with vari- tunity to test this proposition. Although initially con-
able manifestations of FHL or a relapsing, corticosteroid- sidered functionally neutral, Ala91Val impairs perforin
refractory, systemic inflammatory disorder 144 . activity in several ways: it reduces protein stability and
Surprisingly, most of the PRF1 missense mutations that slows intracellular trafficking, which reduces cyto­
have been identified in these patients had no function toxicity by 50–90%, depending on the assay 161. Notably,
in vitro and the protein was retained in the ER, raising primary NK cells from healthy carriers of monoallelic
the issue of why FHL did not develop during child- Ala91Val were less cytotoxic than wild-type cells162. So
hood144. Culturing the CTLs from these patients in con- far, of the 34 reported cases of Ala91Val homozygosity
ditions that optimized protein folding also improved or compound heterozygosity (that is, Ala91Val inher-
perforin trafficking and partly rescued CTL function, ited with a second PRF1 mutation), almost 80% pre-
and thus it is likely that these mutations are ‘leaky’ in sented either with atypical FHL or with haematological
intact organisms144, as previously proposed for a range cancer (TABLE 2).
of protein-misfolding diseases, including cystic fibrosis Can monoallelic inheritance of Ala91Val be patho-
and α1‑antitrypsin deficiency. Consistent with this, PRF1 genic? Apart from its own reduced function, Ala91Val
mutations with the greatest recoverable function pre- perforin also has dominant-negative function, presum-
sented later in life, and presented more commonly with ably by disrupting pore formation when co‑packaged
leukaemia or lymphoma than with inflammation144. We and released with wild-type perforin161. Genotype analy-
concluded that subacute perforinopathy caused by PRF1 sis of large cohorts of patients with cancer showed that
mutations represent a new protein-misfolding disease Ala91Val was not over-represented in more than 2,000
that is similar to other misfolding disorders150,151. cases of paediatric leukaemia163; however, Ala91Val was
These observations directly linked partial perforin over-represented threefold in the subset of patients with
deficiency with predisposition to haematological cancer, acute lymphocytic leukaemia with BCR−ABL trans­
providing that sufficient residual perforin function locations (that is, fusion of ABL1 on chromosome 9
remained to avoid early-onset FHL (designated acute with BCR on chromosome 22, which is found in approx-
perforinopathy)3. Although it has still not been demon­ imately 10% of cases)163. In another study, 4 of 15 patients
strated, we hypothesize that hypomorphic defects of with a lifetime history of both malignant melanoma
MUNC13‑4, STX11 or STXBP2 might have similar and lymphoma carried either the monoallelic Ala91Val
outcomes owing to inadequate perforin delivery 152–154. polymorphism or the Arg28Cys mutation, which closely
Although broad immune suppression states (particularly phenocopies Ala91Val164, suggesting that silencing one
in HIV infection, AIDS and the post-transplant setting) PRF1 allele may predispose to dual cancers in humans.
can greatly increase cancer risk148,155–157, the association of Although significance was achieved in these studies,
hypomorphic PRF1 mutations with cancer links cancer the number of patients was small and the findings need
susceptibility with a protein for which the only known further testing in other cohorts. Intriguingly, individu-
function is to mediate target cell death. Impaired clear- als with a single mutation in UNC13D, STX11, STXBP2
ance of transformed cells by cytotoxic lymphocytes or RAB27A, together with the Ala91Val perforin allele,
seems to be the most feasible way to link PRF1 mutations can present with late-onset FHL165, raising the possibil-
with cancer predisposition, and it would be consistent ity that mutations affecting different components of the
with the remarkable recent success of immune modula- cytotoxic pathway may be complementary.
tors that are directed at programmed cell death protein 1
(PD1) and CTL antigen 4 (CTLA4) as cancer therapeu- Concluding remarks
tics: these agents de‑repress cytotoxic lymphocyte killing Recent advances in molecular, structural and cellular
by inhibiting signalling by these proteins158–160. biology have enhanced our understanding of how cyto-
toxic lymphocytes kill their targets, but many fascinat-
The common human perforin polymorphism Ala91Val ing questions remain. The perforin–granzyme pathway
and disease. The prediction that monoallelic mutations is increasingly being studied in contexts that are patho-
in PRF1 can lead to pathology (mainly cancer) in later physiologically relevant, such as in viral immunity, and
life — tentatively designated chronic perforinopathy — for understanding susceptibility to various inflamma-
requires far more study, particularly at the population tory and neoplastic diseases. The importance of per-
level. The highly prevalent (8–9% among the Caucasian forin and granzyme poly­morphisms is being explored,

NATURE REVIEWS | IMMUNOLOGY VOLUME 15 | JUNE 2015 | 397

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

as is the mechanistic link between failure to release inflammation, will provide opportunities to improve the
active perforin and the often catastrophic inflammatory current therapies. Future work will test the intriguing
consequences of failed target cell death. Informative possi­bility that the common hypomorphic polymorphism
mouse models will guide this work, but the focus must in PRF1 (that is, Ala91Val) is pathogenic and will study
shift to human subjects. At last, rational immune-based its health impact at a population level. In other settings,
therapies are showing promise in patients with cancer, the role of perforin in promoting autoimmune patho­
and more detailed knowledge of how tumour cells logies and allograft rejection warrant an investigation of
are killed by the immune system, and of intratumoral perforin as a target for therapeutic blockade.

1. Ritter, A. T., Angus, K. L. & Griffiths, G. M. The role 20. Young, J. D. et al. Isolation and characterization of a 39. Cruz-Guilloty, F. et al. Runx3 and T‑box proteins
of the cytoskeleton at the immunological synapse. serine esterase from cytolytic T cell granules. Cell 47, cooperate to establish the transcriptional
Immunol. Rev. 256, 107–117 (2013). 183–194 (1986). program of effector CTLs. J. Exp. Med. 206, 51–59
2. Voskoboinik, I. & Trapani, J. A. Perforinopathy: 21. Shi, L., Kam, C. M., Powers, J. C., Aebersold, R. & (2009).
a spectrum of human immune disease caused by Greenberg, A. H. Purification of three cytotoxic 40. Pipkin, M. E. et al. Interleukin‑2 and inflammation
defective perforin delivery or function. Front. Immunol. lymphocyte granule serine proteases that induce induce distinct transcriptional programs that promote
4, 441 (2013). apoptosis through distinct substrate and target cell the differentiation of effector cytolytic T cells.
3. Brennan, A. J., Chia, J., Trapani, J. A. & Voskoboinik, I. interactions. J. Exp. Med. 176, 1521–1529 (1992). Immunity 32, 79–90 (2010).
Perforin deficiency and susceptibility to cancer. 22. Nakajima, H. & Henkart, P. A. Cytotoxic lymphocyte 41. Kim, N. et al. MicroRNA‑150 regulates the
Cell Death Differ. 17, 607–615 (2010). granzymes trigger a target cell internal disintegration cytotoxicity of natural killers by targeting perforin‑1.
4. de Saint Basile, G., Ménasché, G. & Fischer, A. pathway leading to cytolysis and DNA breakdown. J. Allergy Clin. Immunol. 134, 195–203 (2014).
Molecular mechanisms of biogenesis and exocytosis J. Immunol. 152, 1057–1063 (1994). 42. Kim, T. D. et al. Human microRNA‑27a* targets Prf1
of cytotoxic granules. Nature Rev. Immunol. 10, 23. Shi, L. et al. Granzyme B (GraB) autonomously crosses and GzmB expression to regulate NK‑cell cytotoxicity.
568–579 (2010). the cell membrane and perforin initiates apoptosis and Blood 118, 5476–5486 (2011).
5. Janka, G. E. Familial and acquired hemophagocytic GraB nuclear localization. J. Exp. Med. 185, 855–866 43. Trifari, S. et al. MicroRNA-directed program of
lymphohistiocytosis. Annu. Rev. Med. 63, 233–246 (1997). cytotoxic CD8+ T‑cell differentiation. Proc. Natl Acad.
(2012). 24. Nakajima, H., Park, H. L. & Henkart, P. A. Synergistic Sci. USA 110, 18608–18613 (2013).
6. Rosenau, W. & Moon, H. D. Lysis of homologous cells roles of granzymes A and B in mediating target cell 44. Wang, P. et al. Identification of resting and type I IFN-
by sensitized lymphocytes in tissue culture. death by rat basophilic leukemia mast cell tumors also activated human NK cell miRNomes reveals
J. Natl Cancer Inst. 27, 471–483 (1961). expressing cytolysin/perforin. J. Exp. Med. 181, microRNA‑378 and microRNA‑30e as negative
7. Bykovskaja, S. N., Rytenko, A. N., 1037–1046 (1995). regulators of NK cell cytotoxicity. J. Immunol. 189,
Rauschenbach, M. O. & Bykovsky, A. F. Ultrastructural 25. Shiver, J. W., Su, L. & Henkart, P. A. Cytotoxicity with 211–221 (2012).
alteration of cytolytic T lymphocytes following their target DNA breakdown by rat basophilic leukemia 45. D’Angelo, M. E. et al. Cathepsin H is an additional
interaction with target cells. II. Morphogenesis of cells expressing both cytolysin and granzyme A. convertase of pro-granzyme B. J. Biol. Chem. 285,
secretory granules and intracellular vacuoles. Cell 71, 315–322 (1992). 20514–20519 (2010).
Cell. Immunol. 40, 175–185 (1978). 26. Heusel, J. W., Wesselschmidt, R. L., Shresta, S., 46. Pham, C. T. & Ley, T. J. Dipeptidyl peptidase I is
8. Bykovskaja, S. N., Rytenko, A. N., Russell, J. H. & Ley, T. J. Cytotoxic lymphocytes required for the processing and activation of
Rauschenbach, M. O. & Bykovsky, A. F. Ultrastructural require granzyme B for the rapid induction of DNA granzymes A and B in vivo. Proc. Natl Acad. Sci. USA
alteration of cytolytic T lymphocytes following their fragmentation and apoptosis in allogeneic target cells. 96, 8627–8632 (1999).
interaction with target cells. I. Hypertrophy and change Cell 76, 977–987 (1994). 47. Griffiths, G. M. & Isaaz, S. Granzymes A and B are
of orientation of the Golgi apparatus. Cell. Immunol. 27. Su, B., Bochan, M. R., Hanna, W. L., Froelich, C. J. & targeted to the lytic granules of lymphocytes by the
40, 164–174 (1978). Brahmi, Z. Human granzyme B is essential for mannose‑6‑phosphate receptor. J. Cell Biol. 120,
9. Yannelli, J. R., Sullivan, J. A., Mandell, G. L. & DNA fragmentation of susceptible target cells. 885–896 (1993).
Engelhard, V. H. Reorientation and fusion of cytotoxic T Eur. J. Immunol. 24, 2073–2080 (1994). 48. Young, J. D., Damiano, A., DiNome, M. A., Leong, L. G.
lymphocyte granules after interaction with target cells 28. Trapani, J. A. & Smyth, M. J. Killing by cytotoxic T cells & Cohn, Z. A. Dissociation of membrane binding
as determined by high resolution cinemicrography. and natural killer cells: multiple granule serine and lytic activities of the lymphocyte pore-forming
J. Immunol. 136, 377–382 (1986). proteases as initiators of DNA fragmentation. protein (perforin). J. Exp. Med. 165, 1371–1382
10. Dourmashkin, R. R., Deteix, P., Simone, C. B. & Immunol. Cell Biol. 71, 201–208 (1993). (1987).
Henkart, P. Electron microscopic demonstration 29. Jongstra, J. et al. The isolation and sequence of 49. Voskoboinik, I. et al. Calcium-dependent plasma
of lesions in target cell membranes associated a novel gene from a human functional T cell line. membrane binding and cell lysis by perforin are
with antibody-dependent cellular cytotoxicity. J. Exp. Med. 165, 601–614 (1987). mediated through its C2 domain: a critical role for
Clin. Exp. Immunol. 42, 554–560 (1980). 30. Peters, P. J. et al. Cytotoxic T lymphocyte granules aspartate residues 429, 435, 483, and 485 but not
11. Henkart, P. A., Millard, P. J., Reynolds, C. W. & are secretory lysosomes, containing both perforin 491. J. Biol. Chem. 280, 8426–8434 (2005).
Henkart, M. P. Cytolytic activity of purified cytoplasmic and granzymes. J. Exp. Med. 173, 1099–1109 50. Praper, T. et al. Human perforin permeabilizing activity,
granules from cytotoxic rat large granular lymphocyte (1991). but not binding to lipid membranes, is affected by pH.
tumors. J. Exp. Med. 160, 75–93 (1984). 31. Bolitho, P., Voskoboinik, I., Trapani, J. A. & Mol. Immunol. 47, 2492–2504 (2010).
12. Millard, P. J., Henkart, M. P., Reynolds, C. W. & Smyth, M. J. Apoptosis induced by the lymphocyte 51. Uellner, R. et al. Perforin is activated by a proteolytic
Henkart, P. A. Purification and properties of effector molecule perforin. Curr. Opin. Immunol. 19, cleavage during biosynthesis which reveals a
cytoplasmic granules from cytotoxic rat LGL tumors. 339–347 (2007). phospholipid-binding C2 domain. EMBO J. 16,
J. Immunol. 132, 3197–3204 (1984). 32. Voskoboinik, I., Smyth, M. J. & Trapani, J. A. 7287–7296 (1997).
13. Shinkai, Y., Takio, K. & Okumura, K. Homology of Perforin-mediated target-cell death and immune 52. Brennan, A. J. et al. Protection from endogenous
perforin to the ninth component of complement (C9). homeostasis. Nature Rev. Immunol. 6, 940–952 perforin: glycans and the C terminus regulate exocytic
Nature 334, 525–527 (1988). (2006). trafficking in cytotoxic lymphocytes. Immunity 34,
14. Podack, E. R., Young, J. D. & Cohn, Z. A. Isolation 33. Mullbacher, A., Hla, R. T., Museteanu, C. & 879–892 (2011).
and biochemical and functional characterization of Simon, M. M. Perforin is essential for control of 53. Law, R. H. et al. The structural basis for membrane
perforin 1 from cytolytic T‑cell granules. Proc. Natl ectromelia virus but not related poxviruses in mice. binding and pore formation by lymphocyte perforin.
Acad. Sci. USA 82, 8629–8633 (1985). J. Virol. 73, 1665–1667 (1999). Nature 468, 447–451 (2010).
15. Young, J. D., Cohn, Z. A. & Podack, E. R. 34. Pardo, J., Balkow, S., Anel, A. & Simon, M. M. This study reports the X‑ray crystal structure of
The ninth component of complement and the pore- The differential contribution of granzyme A and perforin and the three-dimensional reconstruction
forming protein (perforin 1) from cytotoxic T cells: granzyme B in cytotoxic T lymphocyte-mediated of the entire pore.
structural, immunological, and functional similarities. apoptosis is determined by the quality of target cells. 54. Lopez, J. A., Brennan, A. J., Whisstock, J. C.,
Science 233, 184–190 (1986). Eur. J. Immunol. 32, 1980–1985 (2002). Voskoboinik, I. & Trapani, J. A. Protecting a serial
16. Young, J. D., Hengartner, H., Podack, E. R. & 35. Simon, M. M. et al. In vitro- and ex vivo-derived killer: pathways for perforin trafficking and self-defence
Cohn, Z. A. Purification and characterization of a cytolytic leukocytes from granzyme A x B double ensure sequential target cell death. Trends Immunol.
cytolytic pore-forming protein from granules of knockout mice are defective in granule-mediated 33, 406–412 (2012).
cloned lymphocytes with natural killer activity. apoptosis but not lysis of target cells. J. Exp. Med. 55. Krzewski, K., Gil-Krzewska, A., Nguyen, V., Peruzzi, G.
Cell 44, 849–859 (1986). 186, 1781–1786 (1997). & Coligan, J. E. LAMP1/CD107a is required for
17. Jenne, D. et al. Identification and sequencing of 36. Stepp, S. E. et al. Perforin gene defects in familial efficient perforin delivery to lytic granules and NK‑cell
cDNA clones encoding the granule-associated serine hemophagocytic lymphohistiocytosis. Science 286, cytotoxicity. Blood 121, 4672–4683 (2013).
proteases granzymes D, E and F of cytolytic 1957–1959 (1999). 56. Dustin, M. L. & Long, E. O. Cytotoxic immunological
T lymphocytes. Proc. Natl Acad. Sci. USA 85, This is the first study reporting biallelic perforin synapses. Immunol. Rev. 235, 24–34 (2010).
4814–4818 (1988). mutations in patients with FHL. 57. Lagrue, K. et al. The central role of the cytoskeleton
18. Masson, D. & Tschopp, J. A family of serine 37. Pipkin, M. E. et al. Chromosome transfer activates in mechanisms and functions of the NK cell immune
esterases in lytic granules of cytolytic T lymphocytes. and delineates a locus control region for perforin. synapse. Immunol. Rev. 256, 203–221 (2013).
Cell 49, 679–685 (1987). Immunity 26, 29–41 (2007). 58. Stinchcombe, J. C., Majorovits, E., Bossi, G., Fuller, S.
19. Krahenbuhl, O. et al. Characterization of granzymes A 38. Pipkin, M. E., Rao, A. & Lichtenheld, M. G. & Griffiths, G. M. Centrosome polarization delivers
and B isolated from granules of cloned human cytotoxic The transcriptional control of the perforin locus. secretory granules to the immunological synapse.
T lymphocytes. J. Immunol. 141, 3471–3477 (1988). Immunol. Rev. 235, 55–72 (2010). Nature 443, 462–465 (2006).

398 | JUNE 2015 | VOLUME 15 www.nature.com/reviews/immunol

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

59. Bertrand, F. et al. An initial and rapid step of lytic 80. Baran, K. et al. The molecular basis for perforin 103. Casciola-Rosen, L. et al. Mouse and human granzyme
granule secretion precedes microtubule organizing oligomerization and transmembrane pore assembly. B have distinct tetrapeptide specificities and abilities
center polarization at the cytotoxic T lymphocyte/ Immunity 30, 684–695 (2009). to recruit the bid pathway. J. Biol. Chem. 282,
target cell synapse. Proc. Natl Acad. Sci. USA 110, 81. Browne, K. A. et al. Cytosolic delivery of granzyme B by 4545–4552 (2007).
6073–6078 (2013). bacterial toxins: evidence that endosomal disruption, 104. Cullen, S. P., Adrain, C., Luthi, A. U., Duriez, P. J. &
60. Ruiz-Garcia, R. et al. A case of partial dedicator of in addition to transmembrane pore formation, is an Martin, S. J. Human and murine granzyme B exhibit
cytokinesis 8 deficiency with altered effector phenotype important function of perforin. Mol. Cell. Biol. 19, divergent substrate preferences. J. Cell Biol. 176,
and impaired CD8+ and natural killer cell cytotoxicity. 8604–8615 (1999). 435–444 (2007).
J. Allergy Clin. Immunol. 134, 218–221 (2014). 82. Keefe, D. et al. Perforin triggers a plasma membrane- 105. Thia, K. Y. & Trapani, J. A. The granzyme B gene is
61. Randall, K. L. et al. DOCK8 deficiency impairs CD8 repair response that facilitates CTL induction of highly polymorphic in wild mice but essentially
T cell survival and function in humans and mice. apoptosis. Immunity 23, 249–262 (2005). invariant in common inbred laboratory strains.
J. Exp. Med. 208, 2305–2320 (2011). 83. Froelich, C. J. et al. New paradigm for lymphocyte Tissue Antigens 70, 198–204 (2007).
62. Kupfer, A., Singer, S. J. & Dennert, G. On the granule-mediated cytotoxicity. Target cells bind and 106. Andoniou, C. E. et al. A natural genetic variant of
mechanism of unidirectional killing in mixtures of two internalize granzyme B, but an endosomolytic agent granzyme B confers lethality to a common viral
cytotoxic T lymphocytes. Unidirectional polarization of is necessary for cytosolic delivery and subsequent infection. PLoS Pathog. 10, e1004526 (2014).
cytoplasmic organelles and the membrane-associated apoptosis. J. Biol. Chem. 271, 29073–29079 References 105 and 106 identify and characterize
cytoskeleton in the effector cell. J. Exp. Med. 163, (1996). natural variants of mouse granzyme B and show
489–498 (1986). 84. Metkar, S. S. et al. Cytotoxic cell granule-mediated that their inheritance has a crucial role in the
63. Balaji, K. N., Schaschke, N., Machleidt, W., apoptosis: perforin delivers granzyme B–serglycin clearance of viral pathogens.
Catalfamo, M. & Henkart, P. A. Surface cathepsin B complexes into target cells without plasma membrane 107. Mullbacher, A. et al. Granzymes are the essential
protects cytotoxic lymphocytes from self-destruction pore formation. Immunity 16, 417–428 (2002). downstream effector molecules for the control of
after degranulation. J. Exp. Med. 196, 493–503 85. Thiery, J. et al. Perforin activates clathrin- and primary virus infections by cytolytic leukocytes.
(2002). dynamin-dependent endocytosis, which is required for Proc. Natl Acad. Sci. USA 96, 13950–13955 (1999).
64. Baran, K. et al. Cytotoxic T lymphocytes from plasma membrane repair and delivery of granzyme B 108. Susanto, O. et al. Mouse granzyme A induces a novel
cathepsin B‑deficient mice survive normally in vitro for granzyme-mediated apoptosis. Blood 115, death with writhing morphology that is mechanistically
and in vivo after encountering and killing target cells. 1582–1593 (2010). distinct from granzyme B‑induced apoptosis.
J. Biol. Chem. 281, 30485–30491 (2006). 86. Thiery, J. et al. Perforin pores in the endosomal Cell Death Differ. 20, 1183–1193 (2013).
65. Lopez, J. A. et al. Perforin forms transient pores on the membrane trigger the release of endocytosed 109. Metkar, S. S. et al. Human and mouse granzyme A
target cell plasma membrane to facilitate rapid access granzyme B into the cytosol of target cells. induce a proinflammatory cytokine response.
of granzymes during killer cell attack. Blood 121, Nature Immunol. 12, 770–777 (2011). Immunity 29, 720–733 (2008).
2659–2668 (2013). 87. Gerasimenko, J. V., Tepikin, A. V., Petersen, O. H. & 110. Hoves, S. et al. A critical role for granzymes in antigen
This paper demonstrates the precise moment of Gerasimenko, O. V. Calcium uptake via endocytosis cross-presentation through regulating phagocytosis of
target cell membrane disruption by perforin in a with rapid release from acidifying endosomes. killed tumor cells. J. Immunol. 187, 1166–1175 (2011).
live immunological synapse. Curr. Biol. 8, 1335–1338 (1998). 111. Ewen, C. L., Kane, K. P. & Bleackley, R. C. Granzyme H
66. Isaaz, S., Baetz, K., Olsen, K., Podack, E. & 88. Metkar, S. S. et al. Perforin rapidly induces plasma induces cell death primarily via a Bcl‑2‑sensitive
Griffiths, G. M. Serial killing by cytotoxic T membrane phospholipid flip-flop. PLoS ONE 6, mitochondrial cell death pathway that does not require
lymphocytes: T cell receptor triggers degranulation, e24286 (2011). direct Bid activation. Mol. Immunol. 54, 309–318
re‑filling of the lytic granules and secretion of lytic 89. Renkin, E. M. Filtration, diffusion, and molecular (2013).
proteins via a non-granule pathway. Eur. J. Immunol. sieving through porous cellulose membranes. 112. de Poot, S. A. & Bovenschen, N. Granzyme M: behind
25, 1071–1079 (1995). J. Gen. Physiol. 38, 225–243 (1954). enemy lines. Cell Death Differ. 21, 359–368 (2014).
67. Cohnen, A. et al. Surface CD107a/LAMP‑1 protects 90. Lopez, J. A. et al. Rapid and unidirectional perforin 113. de Poot, S. A. et al. Granzyme M targets topoisomerase
natural killer cells from degranulation-associated pore delivery at the cytotoxic immune synapse. IIα to trigger cell cycle arrest and caspase-dependent
damage. Blood 122, 1411–1418 (2013). J. Immunol. 191, 2328–2334 (2013). apoptosis. Cell Death Differ. 21, 416–426 (2014).
68. Bird, C. H. et al. Selective regulation of apoptosis: 91. Reddy, A., Caler, E. V. & Andrews, N. W. Plasma 114. Pao, L. I. et al. Functional analysis of granzyme M and
the cytotoxic lymphocyte serpin proteinase inhibitor 9 membrane repair is mediated by Ca2+-regulated its role in immunity to infection. J. Immunol. 175,
protects against granzyme B‑mediated apoptosis exocytosis of lysosomes. Cell 106, 157–169 (2001). 3235–3243 (2005).
without perturbing the Fas cell death pathway. 92. Gilbert, R. J. Inactivation and activity of cholesterol- 115. Andrade, F. Non-cytotoxic antiviral activities of
Mol. Cell. Biol. 18, 6387–6398 (1998). dependent cytolysins: what structural studies tell us. granzymes in the context of the immune antiviral state.
69. Zhang, M. et al. Serine protease inhibitor 6 protects Structure 13, 1097–1106 (2005). Immunol. Rev. 235, 128–146 (2010).
cytotoxic T cells from self-inflicted injury by ensuring 93. McNeil, P. L. & Steinhardt, R. A. Plasma membrane 116. Joeckel, L. T. & Bird, P. I. Blessing or curse?
the integrity of cytotoxic granules. Immunity 24, disruption: repair, prevention, adaptation. Proteomics in granzyme research. Proteomics Clin.
451–461 (2006). Annu. Rev. Cell Dev. Biol. 19, 697–731 (2003). Appl. 8, 351–381 (2014).
70. Hadders, M. A., Beringer, D. X. & Gros, P. Structure of 94. Shi, L., Kraut, R. P., Aebersold, R. & Greenberg, A. H. 117. Rawlings, N. D., Waller, M., Barrett, A. J. &
C8α–MACPF reveals mechanism of membrane attack A natural killer cell granule protein that induces DNA Bateman, A. MEROPS: the database of proteolytic
in complement immune defense. Science 317, fragmentation and apoptosis. J. Exp. Med. 175, enzymes, their substrates and inhibitors.
1552–1554 (2007). 553–566 (1992). Nucleic Acids Res. 42, D503–D509 (2014).
71. Rosado, C. J. et al. A common fold mediates 95. Hiebert, P. R. & Granville, D. J. Granzyme B in injury, 118. Brasacchio, D. et al. A functional genomics screen
vertebrate defense and bacterial attack. Science 317, inflammation, and repair. Trends Mol. Med. 18, identifies PCAF and ADA3 as regulators of human
1548–1551 (2007). 732–741 (2012). granzyme B‑mediated apoptosis and Bid cleavage.
References 70 and 71 report the first X‑ray crystal 96. Andrade, F., Fellows, E., Jenne, D. E., Rosen, A. & Cell Death Differ. 21, 748–760 (2014).
structures of MACPF domain-containing proteins Young, C. S. Granzyme H destroys the function of 119. Irmler, M. et al. Granzyme A is an
and demonstrate the similarity of these proteins to critical adenoviral proteins required for viral DNA interleukin 1 β-­converting enzyme. J. Exp. Med. 181,
bacterial CDCs. replication and granzyme B inhibition. EMBO J. 26, 1917–1922 (1995).
72. Slade, D. J. et al. Crystal structure of the MACPF domain 2148–2157 (2007). 120. Hiebert, P. R., Wu, D. & Granville, D. J. Granzyme B
of human complement protein C8α in complex with the This paper shows that granzymes can synergize by degrades extracellular matrix and contributes to
C8γ subunit. J. Mol. Biol. 379, 331–342 (2008). cleaving viral substrates that hijack antiviral host delayed wound closure in apolipoprotein E knockout
73. Shatursky, O. et al. The mechanism of membrane defence. mice. Cell Death Differ. 20, 1404–1414 (2013).
insertion for a cholesterol-dependent cytolysin: a novel 97. Pham, C. T., Ivanovich, J. L., Raptis, S. Z., 121. Zhong, C. et al. Granzyme K inhibits replication of
paradigm for pore-forming toxins. Cell 99, 293–299 Zehnbauer, B. & Ley, T. J. Papillon–Lefèvre syndrome: influenza virus through cleaving the nuclear transport
(1999). correlating the molecular, cellular, and clinical complex importin α1/β dimer of infected host cells.
74. Tilley, S. J., Orlova, E. V., Gilbert, R. J., Andrew, P. W. & consequences of cathepsin C/dipeptidyl peptidase I Cell Death Differ. 19, 882–890 (2012).
Saibil, H. R. Structural basis of pore formation by the deficiency in humans. J. Immunol. 173, 7277–7281 122. Bovenschen, N. et al. Elevated granzyme M‑expressing
bacterial toxin pneumolysin. Cell 121, 247–256 (2005). (2004). lymphocytes during cytomegalovirus latency and
75. Reboul, C. F., Mahmood, K., Whisstock, J. C. & 98. Sutton, V. R. & Trapani, J. A. Proteases in lymphocyte reactivation after allogeneic stem cell transplantation.
Dunstone, M. A. Predicting giant transmembrane β-barrel killer function: redundancy, polymorphism and Clin. Immunol. 150, 1–11 (2014).
architecture. Bioinformatics 28, 1299–1302 (2012). questions remaining. Biol. Chem. 391, 873–879 123. van Domselaar, R. et al. Granzyme M targets host cell
76. Gilbert, R. J., Dalla Serra, M., Froelich, C. J., (2010). hnRNP K that is essential for human cytomegalovirus
Wallace, M. I. & Anderluh, G. Membrane pore 99. Heibein, J. A. et al. Granzyme B‑mediated replication. Cell Death Differ. 20, 419–429 (2013).
formation at protein-lipid interfaces. Trends Biochem. cytochrome c release is regulated by the Bcl‑2 family This paper is an example of how an orphan
Sci. 39, 510–516 (2014). members bid and Bax. J. Exp. Med. 192, 1391–1402 granzyme may have evolved to interfere with a
77. Leung, C. et al. Stepwise visualization of membrane (2000). specific viral pathogen.
pore formation by suilysin, a bacterial cholesterol- 100. Sutton, V. R. et al. Initiation of apoptosis by granzyme 124. Anthony, D. A. et al. A role for granzyme M in
dependent cytolysin. eLife 3, e04247 (2014). B requires direct cleavage of bid, but not direct TLR4‑driven inflammation and endotoxicosis.
78. Metkar, S. S. et al. Perforin oligomers form arcs in granzyme B‑mediated caspase activation. J. Immunol. 185, 1794–1803 (2010).
cellular membranes: a locus for intracellular delivery J. Exp. Med. 192, 1403–1414 (2000). 125. Afonina, I. S. et al. Granzyme B‑dependent proteolysis
of granzymes. Cell Death Differ. 22, 74–85 (2015). 101. Sutton, V. R., Wowk, M. E., Cancilla, M. & Trapani, J. A. acts as a switch to enhance the proinflammatory
References 77 and 78 demonstrate the formation Caspase activation by granzyme B is indirect, and activity of IL‑1α. Mol. Cell 44, 265–278 (2011).
of incomplete pores (‘arcs’) by perforin and a caspase autoprocessing requires the release of 126. Akeda, T. et al. CD8+ T cell granzyme B activates
bacterial CDC. proapoptotic mitochondrial factors. Immunity 18, keratinocyte endogenous IL‑18. Arch. Dermatol. Res.
79. Voskoboinik, I., Dunstone, M. A., Baran, K., 319–329 (2003). 306, 125–130 (2014).
Whisstock, J. C. & Trapani, J. A. Perforin: structure, 102. Kaiserman, D. et al. The major human and mouse 127. Omoto, Y. et al. Granzyme B is a novel interleukin‑18
function, and role in human immunopathology. granzymes are structurally and functionally divergent. converting enzyme. J. Dermatol. Sci. 59, 129–135
Immunol. Rev. 235, 35–54 (2010). J. Cell Biol. 175, 619–630 (2006). (2010).

NATURE REVIEWS | IMMUNOLOGY VOLUME 15 | JUNE 2015 | 399

© 2015 Macmillan Publishers Limited. All rights reserved


REVIEWS

128. Wensink, A. C. et al. Granzyme K synergistically 148. Gatti, R. A. & Good, R. A. Occurrence of malignancy 170. Chung, W. H. et al. Granulysin is a key mediator for
potentiates LPS-induced cytokine responses in in immunodeficiency diseases. A literature review. disseminated keratinocyte death in Stevens–Johnson
human monocytes. Proc. Natl Acad. Sci. USA 111, Cancer 28, 89–98 (1971). syndrome and toxic epidermal necrolysis.
5974–5979 (2014). 149. Katano, H. et al. Chronic active Epstein–Barr virus Nature Med. 14, 1343–1350 (2008).
129. Arias, M. A. et al. Elucidating sources and roles of infection associated with mutations in perforin that 171. Bovenschen, N. et al. NK cell protease granzyme M
granzymes A and B during bacterial infection and impair its maturation. Blood 103, 1244–1252 targets α-tubulin and disorganizes the microtubule
sepsis. Cell Rep. 8, 420–429 (2014). (2004). network. J. Immunol. 180, 8184–8191 (2008).
130. Jenkins, M. R., Trapani, J. A., Doherty, P. C. & 150. Sekijima, Y. et al. The biological and chemical basis 172. Hu, D. et al. Cleavage of survivin by granzyme M
Turner, S. J. Granzyme K expressing cytotoxic for tissue-selective amyloid disease. Cell 121, 73–85 triggers degradation of the survivin-X‑linked inhibitor
T lymphocytes protects against influenza virus in (2005). of apoptosis protein (XIAP) complex to free caspase
granzyme AB−/− mice. Viral Immunol. 21, 341–346 151. Wiseman, R. L., Powers, E. T., Buxbaum, J. N., activity leading to cytolysis of target tumor cells.
(2008). Kelly, J. W. & Balch, W. E. An adaptable standard J. Biol. Chem. 285, 18326–18335 (2010).
131. Marcet-Palacios, M. et al. Granzyme B inhibits for protein export from the endoplasmic reticulum. 173. Kelly, J. M. et al. Granzyme M mediates a novel form
vaccinia virus production through proteolytic Cell 131, 809–821 (2007). of perforin-dependent cell death. J. Biol. Chem. 279,
cleavage of eukaryotic initiation factor 4 gamma 3. 152. Machaczka, M. et al. Development of classical 22236–22242 (2004).
PLoS Pathog. 7, e1002447 (2011). Hodgkin’s lymphoma in an adult with biallelic STXBP2 174. Lu, H. et al. Granzyme M directly cleaves inhibitor
132. van Domselaar, R., Quadir, R., van der Made, A. M., mutations. Haematologica 98, 760–764 (2013). of caspase-activated DNase (CAD) to unleash CAD
Broekhuizen, R. & Bovenschen, N. All human 153. Meeths, M. et al. Spectrum of clinical presentations in leading to DNA fragmentation. J. Immunol. 177,
granzymes target hnRNP K that is essential for tumor familial hemophagocytic lymphohistiocytosis type 5 1171–1178 (2006).
cell viability. J. Biol. Chem. 287, 22854–22864 patients with mutations in STXBP2. Blood 116, 175. Reboul, C. F., Whisstock, J. C. & Dunstone, M. A.
(2012). 2635–2643 (2010). A new model for pore formation by cholesterol-
133. Wang, S., Xia, P., Shi, L. & Fan, Z. FADD cleavage by This study uncovers a range of previously dependent cytolysins. PLoS Comput. Biol. 10,
NK cell granzyme M enhances its self-association to unsuspected clinical symptoms in patients e1003791 (2014).
facilitate procaspase‑8 recruitment for auto-processing with FHL type 5. 176. Sedelies, K. A. et al. Discordant regulation of
leading to caspase cascade. Cell Death Differ. 19, 154. Pagel, J. et al. Distinct mutations in STXBP2 are granzyme H and granzyme B expression in human
605–615 (2012). associated with variable clinical presentations in lymphocytes. J. Biol. Chem. 279, 26581–26587
134. de Poot, S. A., Lai, K. W., Hovingh, E. S. & patients with familial hemophagocytic (2004).
Bovenschen, N. Granzyme M cannot induce cell death lymphohistiocytosis type 5 (FHL5). Blood 119, 177. Busiello, R. et al. Atypical features of familial
via cleavage of mouse FADD. Apoptosis 18, 533–534 6016–6024 (2012). hemophagocytic lymphohistiocytosis. Blood 103,
(2013). 155. Adami, J. et al. Cancer risk following organ 4610–4612 (2004).
135. Baginska, J. et al. Granzyme B degradation by transplantation: a nationwide cohort study in Sweden. 178. Cannella, S. et al. Germline mutations of the perforin
autophagy decreases tumor cell susceptibility to Br. J. Cancer 89, 1221–1227 (2003). gene are a frequent occurrence in childhood anaplastic
natural killer-mediated lysis under hypoxia. 156. Frisch, M., Biggar, R. J., Engels, E. A., Goedert, J. J. & large cell lymphoma. Cancer 109, 2566–2571 (2007).
Proc. Natl Acad. Sci. USA 110, 17450–17455 AIDS–Cancer Match Registry Study Group. 179. Clementi, R. et al. Adult onset and atypical
(2013). Association of cancer with AIDS-related presentation of hemophagocytic lymphohistiocytosis
136. Wang, S. et al. Rapid reuptake of granzyme B leads to immunosuppression in adults. JAMA 285, in siblings carrying PRF1 mutations. Blood 100,
emperitosis: an apoptotic cell‑in‑cell death of immune 1736–1745 (2001). 2266–2267 (2002).
killer cells inside tumor cells. Cell Death Dis. 4, e856 157. Grulich, A. E., van Leeuwen, M. T., Falster, M. O. & 180. Clementi, R. et al. A proportion of patients with
(2013). Vajdic, C. M. Incidence of cancers in people with lymphoma may harbor mutations of the perforin gene.
137. Sutton, V. R., Vaux, D. L. & Trapani, J. A. Bcl‑2 HIV/AIDS compared with immunosuppressed Blood 105, 4424–4428 (2005).
prevents apoptosis induced by perforin and transplant recipients: a meta-analysis. Lancet 370, 181. Feldmann, J. et al. Functional consequences of
granzyme B, but not that mediated by whole cytotoxic 59–67 (2007). perforin gene mutations in 22 patients with familial
lymphocytes. J. Immunol. 158, 5783–5790 (1997). 158. Leach, D. R., Krummel, M. F. & Allison, J. P. haemophagocytic lymphohistiocytosis. Br. J. Haematol.
138. Sutton, V. R. et al. Granzyme B triggers a prolonged Enhancement of antitumor immunity by CTLA‑4 117, 965–972 (2002).
pressure to die in Bcl‑2 overexpressing cells, defining blockade. Science 271, 1734–1736 (1996). 182. Mancebo, E. et al. Familial hemophagocytic
a window of opportunity for effective treatment with 159. Naidoo, J., Page, D. B. & Wolchok, J. D. Immune lymphohistiocytosis in an adult patient homozygous for
ABT‑737. Cell Death Dis. 3, e344 (2012). checkpoint blockade. Hematol. Oncol. Clin. North Am. A91V in the perforin gene, with tuberculosis infection.
139. Cote, M. et al. Munc18‑2 deficiency causes familial 28, 585–600 (2014). Haematologica 91, 1257–1260 (2006).
hemophagocytic lymphohistiocytosis type 5 and 160. Zamarin, D. et al. Localized oncolytic virotherapy 183. Marcenaro, S. et al. Analysis of natural killer-cell function
impairs cytotoxic granule exocytosis in patient overcomes systemic tumor resistance to immune in familial hemophagocytic lymphohistiocytosis (FHL):
NK cells. J. Clin. Invest. 119, 3765–3773 (2009). checkpoint blockade immunotherapy. Sci. Transl Med. defective CD107a surface expression heralds Munc13‑4
140. Jenkins, M. R. et al. Failed CTL/NK cell killing and 6, 226ra32 (2014). defect and discriminates between genetic subtypes of
cytokine hyper-secretion are directly linked through 161. Voskoboinik, I. et al. Perforin activity and immune the disease. Blood 108, 2316–2323 (2006).
prolonged synapse time. J. Exp. Med. 212, 307–317 homeostasis: the common A91V polymorphism in 184. Okur, H. et al. Clinical and molecular aspects of Turkish
(2015). perforin results in both presynaptic and postsynaptic familial hemophagocytic lymphohistiocytosis patients
This study demonstrates the link between perforin defects in function. Blood 110, 1184–1190 (2007). with perforin mutations. Leuk. Res. 32, 972–975
deficiency, the duration of the immunological 162. House, I. G. et al. Heterozygosity for the common (2008).
synapse and hypersecretion of pro-inflammatory perforin mutation, p.A91V, impairs the cytotoxicity of 185. Risma, K. A., Frayer, R. W., Filipovich, A. H. & Sumegi, J.
cytokines. It also provides the cellular mechanism primary natural killer cells from healthy individuals. Aberrant maturation of mutant perforin underlies the
of FHL. Immunol. Cell Biol. http://dx.doi.org/10.1038/ clinical diversity of hemophagocytic lymphohistiocytosis.
141. Feldmann, J. et al. Munc13‑4 is essential for cytolytic icb.2015.1 (2015). J. Clin. Invest. 116, 182–192 (2006).
granules fusion and is mutated in a form of familial 163. Mehta, P. A. et al. Perforin polymorphism A91V 186. Santoro, A. et al. A single amino acid change A91V
hemophagocytic lymphohistiocytosis (FHL3). and susceptibility to B‑precursor childhood in perforin: a novel, frequent predisposing factor
Cell 115, 461–473 (2003). acute lymphoblastic leukemia: a report from to childhood acute lymphoblastic leukemia?
142. zur Stadt, U. et al. Familial hemophagocytic the Children’s Oncology Group. Leukemia 20, Haematologica 90, 697–698 (2005).
lymphohistiocytosis type 5 (FHL‑5) is caused by 1539–1541 (2006). 187. Trambas, C. et al. A single amino acid change, A91V,
mutations in Munc18‑2 and impaired binding to 164. Trapani, J. A. et al. Human perforin mutations and leads to conformational changes that can impair
syntaxin 11. Am. J. Hum. Genet. 85, 482–492 (2009). susceptibility to multiple primary cancers. processing to the active form of perforin.
143. zur Stadt, U. et al. Linkage of familial hemophagocytic Oncoimmunology 2, e24185 (2013). Blood 106, 932–937 (2005).
lymphohistiocytosis (FHL) type‑4 to chromosome 6q24 165. Zhang, K. et al. Synergistic defects of different 188. Trizzino, A. et al. Genotype–phenotype study of familial
and identification of mutations in syntaxin 11. molecules in the cytotoxic pathway lead to clinical haemophagocytic lymphohistiocytosis due to perforin
Hum. Mol. Genet. 14, 827–834 (2005). familial hemophagocytic lymphohistiocytosis. mutations. J. Med. Genet. 45, 15–21 (2008).
References 139 and 141–143 determine three Blood 124, 1331–1334 (2014). 189. Orange, J. S. et al. Wiskott–Aldrich syndrome protein
genes that are responsible for perforin and This study demonstrates that the co‑inheritance is required for NK cell cytotoxicity and colocalizes
granzyme secretion into the immunological of monoallelic mutations in two genes associated with actin to NK cell-activating immunologic synapses.
synapse; defects in these genes lead to FHL. with FHL may lead to the disease; the Ala91Val Proc. Natl Acad. Sci. USA 99, 11351–11356 (2002).
144. Chia, J. et al. Temperature sensitivity of human polymorphism in PFR1 has been responsible for
perforin mutants unmasks subtotal loss of cytotoxicity, almost 50% of cases. Acknowledgements
delayed FHL, and a predisposition to cancer. 166. Pena, S. V., Hanson, D. A., Carr, B. A., Goralski, T. J. & I.V. and J.C.W. are supported by fellowships from the National
Proc. Natl Acad. Sci. USA 106, 9809–9814 (2009). Krensky, A. M. Processing, subcellular localization, Health and Medical Research Council of Australia (NHMRC).
This study defines partial perforin deficiency as a and function of 519 (granulysin), a human late T cell I.V., J.C.W. and J.A.T. are supported by project grants from the
protein misfolding disease and determines its link activation molecule with homology to small, lytic, NHMRC. J.C.W. is supported by the Australian Research
with cancer predisposition. granule proteins. J. Immunol. 158, 2680–2688 Council (ARC) and he acknowledges the support of an ARC
145. van den Broek, M. E. et al. Decreased tumor (1997). Federation fellowship. J.A.T. is supported by a programme
surveillance in perforin-deficient mice. J. Exp. Med. 167. Anderson, D. H. et al. Granulysin crystal structure and grant from the NHMRC. The authors thank the many members
184, 1781–1790 (1996). a structure-derived lytic mechanism. J. Mol. Biol. 325, of their laboratories, especially J. A. Lopez and A. J. Brennan,
146. van den Broek, M. F., Kagi, D., Zinkernagel, R. M. & 355–365 (2003). and their collaborators, especially H. Saibil, M. A. Dunstone,
Hengartner, H. Perforin dependence of natural killer 168. Stenger, S. et al. An antimicrobial activity of cytolytic
R. H. P. Law and N. Lukoyanova, for their contributions to
cell-mediated tumor control in vivo. Eur. J. Immunol. T cells mediated by granulysin. Science 282, 121–125
many of the findings discussed in this Review.
25, 3514–3516 (1995). (1998).
147. Smyth, M. J. et al. Perforin-mediated cytotoxicity is 169. Walch, M. et al. Cytotoxic cells kill intracellular
critical for surveillance of spontaneous lymphoma. bacteria through granulysin-mediated delivery of Competing interests statement
J. Exp. Med. 192, 755–760 (2000). granzymes. Cell 157, 1309–1323 (2014). The authors declare no competing interests.

400 | JUNE 2015 | VOLUME 15 www.nature.com/reviews/immunol

© 2015 Macmillan Publishers Limited. All rights reserved

You might also like