REVIEW ARTICLE
Programmed cell death
Apoptosis and alternative deathstyles
Cinthya Assunc¸a
˜o Guimara
˜es* and Rafael Linden
Instituto de Biofı´sica da UFRJ, Rio de Janeiro, Brazil
Programmed cell death is a major component of both
normal development and disease. The roles of cell death
during either embryogenesis or pathogenesis, the signals
that modulate this event, and the mechanisms of cell
demise are the major subjects that drive research in this
field. Increasing evidence obtained both in vitro and
in vivo supports the hypothesis that a variety of cell death
programs may be triggered in distinct circumstances.
Contrary to the view that caspase-mediated apoptosis
represents the standard programmed cell death, recent
studies indicate that an apoptotic morphology can be
produced independent of caspases, that autophagic exe-
cution pathways of cell death may be engaged without
either the involvement of caspases or morphological signs
of apoptosis, and that even the necrotic morphology of
cell death may be consistently produced in some cases,
including certain plants. Alternative cell death programs
may imply novel therapeutic targets, with important
consequences for attempts to treat diseases associated with
disregulated programmed cell death.
Keywords: programmed cell death; apoptosis; autophagy;
necrosis; neurodegenerative diseases.
Introduction
Programmed cell death is a major component of both
normal development and disease [1–11]. The roles of cell
death during either embryogenesis or pathogenesis, the
signals that induce or regulate this event, and the mecha-
nisms of cell demise are common subjects that drive research
in this field [12–17]. The purpose of this article is to review
major morphological, biochemical and molecular hallmarks
of distinct forms of programmed cell death, and to examine
the limits of some prevailing views of cell death modes and
mechanisms.
The classical ultrastructural studies of Kerr and coworkers
[18] provided evidence that cells may undergo at least two
distinct types of cell death: The first type is known as necrosis,
a violent and quick form of degeneration affecting extensive
cell populations, characterized by cytoplasm swelling,
destruction of organelles and disruption of the plasma
membrane, leading to the release of intracellular contents
and inflammation. A remarkably distinct type of cell death
was called apoptosis, identified in single cells usually
surrounded by healthy-looking neighbors, and characterized
by cell shrinkage, blebbing of the plasma membrane,
maintenance of organelle integrity, and condensation and
fragmentation of DNA, followed by ordered removal
through phagocytosis [18,19]. During the last 30 years, cell
death has usually been classified within this dichotomy.
The work of Kerr and collaborators stirred interest in
programmed cell death, both because it provided a visible
object (the apoptotic profile) to be consistently approached
in experimental studies prior to the disappearance of the
dead cells, as well as due to the evidence provided for
controlled events that justify the operational definition of
ÔprogrammedÕ[20]. At a first approximation, necrosis was
attributed to accidental, uncontrolled degeneration, whereas
apoptosis presented the defining characteristics of a cell
death program. Indeed, many research groups began to
consider apoptosis and programmed cell death as a single
entity, despite the knowledgeable criticism of pioneers in
the field [21]. Nonetheless, this simplified and generalized
scheme neglects the exceptions; for example, morphologies
of cell death that do not fit in the original classification
(reviewed in [22]). On the other hand, evidence is now
available for multiple alternative cell death pathways, as
well as for cross-talk of intracellular mechanisms involved in
distinct aspects of cell degeneration. This review will focus
on the growing evidence that, besides apoptosis, autophagic
and necrotic forms of cell degeneration may be pro-
grammed, and underlie cell death either in isolation or
combined with mechanisms of apoptosis.
Correspondence to R. Linden, Instituto de Biofı
´sica da UFRJ,
Centro de Cieˆncias da Sau´ de, bloco G, Cidade Universita
´ria,
21949–900, Rio de Janeiro, Brazil. Tel.: + 55 21 25626553,
Fax: + 55 21 22808193, E-mail: rlinden@biof.ufrj.br
Abbreviations: AIF, apoptosis inducing factor; DAPk, death-associ-
ated protein kinase; DRP, DAPk-related protein kinase; FADD,
Fas-associated protein with death domain; IAP, inhibitor of
apoptosis; LEI, leucocyte elastase inhibitor; L-DNase II,
LEI-DNase II; MPT, mitochondrial permeability transition;
PCD, programmed cell death; PI3K, phosphatidylinositol-3-kinase;
TNFa, tumor necrosis factor-alpha; TUNEL, TdT-mediated
biotin-dUDP nick-end labeling.
Note: a website is available at http://www.biof.ufrj.br/neurogen
*Present address: The Hebrew University of Jerusalem, Department of
Biological Chemistry, Institute of Life Sciences, The Edmond J. Safra
Campus, Givat Ram, Jerusalem 91904 Israel.
(Received 12 January 2004, revised 17 February 2004,
accepted 10 March 2004)
Eur. J. Biochem. 271, 1638–1650 (2004) FEBS 2004 doi:10.1111/j.1432-1033.2004.04084.x
Defining features of programmed cell death
Despite the tremendous impact of research in apoptosis
upon the understanding both of cellular and molecular
mechanisms of cell demise, as well as of mechanisms of
degenerative diseases, the confusion between apoptosis and
programmed cell death has somewhat obscured the field.
Regardless of whether this paradox is attributable to either
disconnection of modern science from its philosophical
foundations [23] or to a more trivial neglect of classical
papers (reviewed in [20]), it is likely that progress in the
identification and understanding of nonapoptotic forms
of programmed cell death may have been unnecessarily
delayed.
Indeed, well before the upsurge in the understanding of
mechanisms of apoptosis, a clear warning had been issued
to avoid confusion between the form of cell death called
apoptosis, and the concept of programmed cell death as a
sequence of events, but not necessarily those that led to the
morphology of apoptosis [21].
Although the original work that led to the concept of
programmed cell death was carried out in developing
organisms [24–29], there is nothing intrinsically develop-
mental in the concept. The apoptotic form has been long
identified in adult tissues [18], and there is no evidence that
any particular form of cell death, much less the operational
concept of programmed cell death, can be attributed
exclusively to either developing or mature cells. Conversely,
it has been argued that apoptotic forms of cell death induced
by cytotoxic drugs or physical stimuli could not be taken as
programmed cell death because the latter represents normal
degeneration that is part of the life of an organism [30].
However, those instances of induced degeneration reflected
no less an orderly sequence of cellular events than naturally
occurring cell death in the form of apoptosis found in
developing organisms.
Thus, a simple and noncommittal definition of pro-
grammed cell death as Ôa sequence of events based on
cellular metabolism that lead to cell destructionÕis likely
both to preserve the concept as originally defined, as well as
to discard decorative qualifications based on particular
experimental findings. A disturbing example of the latter is
the requirement for protein synthesis [31,32], that had a
great impact in the acceptance of cell death as controlled by
gene expression (and thus Ôgenetically programmedÕ). Not
only do many cells die under inhibition of either transcrip-
tion or translation in a controlled way indistinguishable
from that underlying cell death dependent on protein
synthesis [33–35], but the rapid progress in the understand-
ing of post-translational mechanisms of cell death has
largely overshadowed transcriptional control and even the
classical requirement for protein synthesis [36,37]. The
caveat that the sequence of events in programmed cell death
must be based on cell metabolism allows for the irony that
even the time between hitting a cell with a hammer and the
death of the former is finite, notwithstanding that the
intervening events are not resolvable with current tech-
niques.
Acceptance of the minimalist concept should help
attribute appropriate weight to alternative forms of pro-
grammed cell death, despite the overwhelming dominance
of apoptosis in the literature.
Multiple mechanisms of apoptosis
Cell death with apoptotic morphology can be triggered
by several stimuli, including intracellular stress and
receptor-mediated signaling. These signals feed into an
evolutionarily conserved intracellular machinery of execu-
tion [36,38], the mechanisms of which have mainly been
traced to the activity of the caspase family of cysteine-
proteases [39–41].
Caspase-mediated apoptotic cell death has been
extensively reviewed, e.g. [16,36,38,42–44]. Briefly, the
caspases are synthesized as zymogens and upstream signals
convert these precursors into mature proteases. Initiator
caspases (caspase-1, -2, -4, -5, -8, -9, -10 and -14] are
activated via oligomerization-induced autoprocessing
[45–50], while effector caspases [caspase-3, -6 and -7] are
activated by other proteases, including initiator caspases
and granzyme B. Proteolytic cleavage of cellular substrates
by effector caspases largely determines the features of
apoptotic cell death ([51–53]; reviewed in [54–56]).
Three major pathways have been identified according to
their initiator caspase: the death receptor pathway involving
caspase-8 [57], the endoplasmic reticulum stress pathway
attributed to activation of caspase-12 [58], and the mito-
chondrial pathway, in which various signals can trigger
the release of harmful proteins by mitochondria into the
cytoplasm, leading to activation of caspase-9 and down-
stream cleavage of caspase-3, -7 or -6 [46,59–62].
Although caspase-3 is widely involved in the execution of
apoptosis [63], its effector functions may be dispensable for
apoptotic-like cell death [64,65]. The use of either pharma-
cological inhibitors or knockout animals further showed
that cells can trigger alternative mechanisms of cell demise.
For example, sympathetic and dorsal root ganglion neurons
deprived of nerve growth factor (NGF) die in a caspase-2-
dependent manner, but the same neurons derived from
caspase-2 knockout mice still die following nerve growth
factor deprivation, this time depending on activation of
caspase-9, which does not occur in wild-type mice [66].
Thus, rather than a single linear mechanism, alternative
caspase-mediated pathways may be activated for apoptotic
cell death, depending on whether a preferential caspase is
blocked. It is likely that the network of intrinsic regulatory
pathways that impinge upon the activity of caspases, such as
the inhibitors of apoptosis (IAPs) and IAP-binding proteins
[67], may regulate the choice between alternative pathways
in normal cells, depending on metabolic state, stage of
differentiation and other conditions.
In addition, caspase inhibition fails to block programmed
cell death with apoptotic morphology in several experimental
models [68–72]. For example, the ultrastructural features of
apoptosis inducing factor (AIF)-induced cell death represent
an example of a slight variation from the standard pattern of
apoptotic morphology, which appears to be independent of
caspase activation ([73]; see also [74]). Cell death pathways
independent of caspase activation have been described, for
example, even in some forms of cell death induced either
by the Bcl-family protein Bax [75], as well as in cell death
involving the activation of other proteases, such as calpain
[76], proteasome [77] and serine proteases.
The latter enzymes have an important role in early
chromatin cleavage [78], and are activated in the classical
FEBS 2004 Alternative pathways of programmed cell death (Eur. J. Biochem. 271) 1639
model of apoptosis of thymocytes induced by glucocortic-
oids [79]. Serine proteases participate in a cell death pathway
that involves the activation of the endonuclease leucocyte
elastase inhibitor (LEI)-DNase II (L-DNase II), and is not
inhibited in HeLa cells by pancaspase inhibitors [80].
Activation of L-DNase II was first described in lens cell
differentiation, which is related to apoptosis [81]. The
activation of this enzyme also occurs under other physio-
logical conditions, such as the death of retinal cells during
development [82].
The key molecule of this pathway is LEI, which is a
member of the superfamily of protease inhibitors called
serpins (serine protease inhibitors). In its active form, LEI
inhibits elastase, cathepsin G and probably other proteases
[83]. LEI can undergo post-translational modifications
either under acidic pH or by the action of proteases,
including elastase. Once LEI is exposed to these conditions,
a decrease in the molecular mass is observed simultaneously
with the appearance of endonuclease activity [84]. The
DNase generated by the action of the serine protease
elastase was named L-DNase II, as it shows dependence on
the same ions and pH required by DNase II.
Recent reports shows that the serine protease Omi/HtrA2
is a mitochondrial direct X-chromosome-linked inhibitor of
apoptosis protein (XIAP)-binding protein, which is released
from mitochondria upon induction of apoptosis together
with cytochrome cand Smac/Diablo ([85,86]; reviewed in
[87]), and its release can be inhibited by Bcl-2 [88]. These
data suggest that in some cases there may be a cooperative
action between serine proteases and caspases in the execu-
tion of cell death.
The data show that the classically defined apoptotic
morphology can be achieved either by activation of
caspases, or through the mediation of other families of
proteases [Fig. 1], although the exact cytological features
of cell demise may vary slightly among these various forms
of apoptosis.
Autophagy and autophagic cell death
As part of normal development, cells depend on a strictly
regulated balance of protein synthesis and degradation, as
well as organelle biogenesis and dismantlement. While
proteasome-mediated degradation is responsible for most of
the protein recycling, the turnover of organelles is mainly
attributed to autophagy [89].
Autophagy occurs in many eukaryotic cell types, where
organelles and other cell components are sequestered into
lysosomes and degraded. The lysosome is a cellular
compartment enriched in hydrolases able to cleave proteins,
lipids, nucleic acids and carbohydrates that may lead
to organelle degradation through macroautophagy [90].
Autophagy has been described both as a means to resist
starvation, and as part of cellular remodeling during
differentiation, metamorphosis, aging, cell transformation,
physiological whole-organ changes such as growth of the
uterus during pregnancy and its atrophy after childbirth, as
well as in the removal of anomalous cellular components
that accumulate following toxic insults or during cell death
[91]. In the nervous system, for example, morphological
signs of autophagy are observed in physiological processes,
such as the removal of outer segments of retinal photo-
receptors by the pigment epithelium [92], which is not
associated with cell death.
Notwithstanding, autophagic profiles identified by ultra-
structural features have been associated with cell death in
certain circumstances [22]. Cells in the early stages of
autophagy contain several autophagic vacuoli, and both the
nucleoplasm and the cytoplasm appear slightly darkened,
although nuclear structure still appears normal. Mitochon-
dria and the endoplasmic reticulum are sometimes dilated,
and the Golgi apparatus is often enlarged. The plasma
membrane loses specializations such as microvilli and
junctional complexes, and blebbing can occur. In several
cases, an intense endocytosis is observed, and this probably
Fig. 1. Multiple pathways to apoptosis, both dependent and independent of the activation of caspases. The diagram summarizes the major components
of the pathways reported to underlie cell death in various types of cells and tissues.
1640 C. Assunc¸ a
˜o Guimara
˜es and R. Linden (Eur. J. Biochem. 271)FEBS 2004
leads to a reduction in the area of the plasma membrane.
During late stages, both the number and size of vacuoli
increase, and many of them contain myelin figures or are
filled with lipids, which appear as pale gray inclusions in the
cytoplasm [22].
The nucleus of a cell undergoing autophagic cell death
can become pyknotic and identifiable as such by light
microscopy, either in early or in late stages of the
degenerative process. Nevertheless, this nuclear condensa-
tion is neither as common nor as remarkable as that of
apoptosis. The late autophagic cell debris is frequently
removed by heterophagy, but this tends to occur in very late
stages, and seems to be less conspicuous than the clearance
of apoptotic bodies [22].
Autophagic cell death is not an exclusive feature of
multicellular organisms. In the protozoan pathogen Leish-
mania donovani, treatment with antimicrobial peptides
induced cytoplasmic vacuolization and dismantling of the
cellular organization without disruption of the plasma
membrane, with no nuclear fragmentation or DNA
laddering, and independent of caspase-like activity. Instead,
monodansylcadaverine, a biochemical marker of auto-
phagy, specifically labeled the vacuoles induced by anti-
microbial peptides [93].
Endostatin, an inhibitor of angiogenesis, was shown to
induce the formation of autophagic vacuoles in endothelial
cells. Cell death was not prevented by antioxidants or
caspase inhibitors, but was reduced by 3-methyladenine, a
specific inhibitor of autophagy, and serine and cysteine
lysosomal protease inhibitors [94]. Neuregulin (NRG; a
ligand of ErbB), also activates ErbB-2/ErbB-3 heterodimers
and induces cell death of prostate cancer LNCaP cells.
Neuregulin-induced cell death was not inhibited by broad-
spectrum caspases inhibitors, but was blocked by 3-methyl-
adenine [95].
Ionizing radiation induced a dose-dependent suppression
of cell proliferation and autophagic cell changes in several
glioblastoma multiform cell lines [96]. Arsenic trioxide, an
agent that causes remission in patients with acute promyelo-
cytic leukemia and multiple myeloma without severe side-
effects, was shown to inhibit proliferation of glioma cell
lines. The G2/M arrest was accompanied by ultrastructural
features of autophagy, and was inhibited by the autophagy
inhibitor bafilomycin A1, whereas general caspase inhibitors
did not block As
2
O
3
-induced cell death [97]. In neurobla-
stoma cells, dopamine leads to autophagic changes charac-
terized by the presence of numerous cytoplasmic vacuoles
with inclusions, and accompanied by mitochondrial aggre-
gation, activation of the stress-response kinases SAPK/JNK
and p38, and increased a-synuclein expression. Both cell
viability and the increase in a-synuclein expression were
prevented by antioxidants, by the specific inhibitors of
p38 and SAPK/JNK, and by 3-methyladenine [98]. Thus,
various agents can lead to autophagic cell death in tumor
cell lines.
Indeed, Bursch and collaborators [99] had long since
shown that the MCF-7 breast carcinoma cell line, which
does not express caspase-3 [100], undergoes autophagic cell
death upon treatment with tamoxifen. More recent work
showed that in apoptotic cell death induced by tyrphostin
A25 in the human colon cancer cell line HT29/HI1, early
stages of the death process are associated with depolymer-
ization of actin and degradation of intermediate filaments.
In contrast, during tamoxifen-induced autophagic cell death
of MCF-7 cells, intermediate and microfilaments are
redistributed, but largely preserved even beyond the stage
of nuclear collapse [101]. These data support the concept
that autophagic cell death is a separate form of programmed
cell death that is distinctly different from apoptosis.
In keeping with this interpretation, intensive irradiation
led to up to 30% cell death in MCF-7 cells without any signs
of apoptosis. In this case, cell death was accompanied by the
formation of acidic vesicular organelles and lamellar
structures, which was prevented by 3-methyladenine. How-
ever, following low-dose irradiation, the presence of acidic
vesicular organelles correlated with an increased chance of
survival, suggesting that moderate signs of autophagy may
be associated with a defensive reaction of nonlethally
damaged cells [102]. The data are consistent with the view
that nonlethal injury can trigger an autophagic defensive
reaction, whereas harsh treatment of certain cells can lead to
cell death largely dependent on autophagy itself.
The first step of autophagy is the formation of an
autophagosome, which occurs when a portion of the
cytoplasm is engulfed by a double membrane vacuole that
does not contain either acid phosphatases or aryl-sulphatase
activity. The double membrane is derived from ribosome-
free areas of rough endoplasmic reticulum [91]. After a
maturation period that includes the acidification of the
vacuole, hydrolases are inserted into the autophagosome by
fusion with pre-existing lysosomes or elements deriving
from the Golgi complex. This process appears to involve
mannose-6-phosphate receptors located in the autophago-
some membrane, resulting in the formation of a degradative
vacuole limited by a single membrane named an autolyso-
some [103]. Vacuole formation can be regulated by amino
acids and hormones and by stress [104,105].
Similarly to yeast [90], autophagy in mammalian cells is
highly dependent on phosphorylation events. In mammalian
hepatocytes, the phosphorylation of the ribosomal protein
S6 correlates strongly with inhibition of macroautophagy
[106]. The activity of the p70S6-kinase is regulated by the
mTor kinase, and inhibition of S6 phosphorylation caused
by inactivation of mTor with rapamycin induces autophagy
even under nutrient-rich conditions. In yeast, inhibition of
the Tor2 kinase results in activation of protein phosphatase
2A and induction of autophagy [107]. Also, in hepatocytes,
the effect of the phosphatase inhibitor okadaic acid upon
protein phosphatase 2A inhibits the autophagic process
[108]. Furthermore, various classes of phosphatidylinositol-
3-kinase (PI3K) control the autophagic pathway in distinct
ways: class IA PI3K inhibits cytoplasm sequestration and
degradation, while class III stimulates the sequestration of
cytoplasm, implicating the PI3K family as key regulators of
the autophagic pathway [109].
In a recent study, Inbal and coworkers [110] showed that
the expression of death-associated protein kinase (DAPk)
and DAPk-related protein kinase (DRP)-1, members of a
family of Ca
2+
/calmodulin-regulated Ser/Thr death kinases,
triggered two major caspase-independent cytoplasmic
events. These were membrane blebbing, a feature common
to various forms of cell death, and extensive autophagy,
which is typical of autophagic cell death. Furthermore,
either the expression of the dominant negative mutant of
FEBS 2004 Alternative pathways of programmed cell death (Eur. J. Biochem. 271) 1641
DRP-1 or DAPk antisense mRNA reduced autophagy
induced by antiestrogens, amino acid starvation, or admin-
istration of interferon-c. The finding of DRP-1 inside the
autophagic vesicles suggests a direct involvement of this
kinase in the process of autophagy.
Liang and collaborators [111] showed that beclin-1, a
protein that interacts with bcl-2, promotes autophagy in
both an autophagy-defective yeast cell line and in the
MCF-7 cell line, which normally does not express detectable
levels of beclin-1. It was also shown that beclin-1 expression
is frequently low in epithelial breast carcinomas, but is
widely expressed in normal tissue. The authors suggested
that beclin-1 may be a mammalian gene for autophagy, and
that it inhibits tumorigenesis through activation of this cell
death pathway [111]. Kihara and coworkers [112] suggested
that beclin-1 is a component of the PI3K complex, which is
also required for autophagy, and that beclin-1 and PI3K
control autophagy as a complex at the Trans Golgi network
[112]. However, in contrast with the poor autophagic
response of MCF-7 cells to amino acid deprivation [111],
strong autophagic responses were elicited in this cell line
bothbytamoxifen[101]andbyirradiation[102].Thus,
either beclin-1 can be replaced by other autophagy-inducing
proteins, or alternative pathways of autophagy may operate
under various stimuli.
The death of cerebellar Purkinje cells in lurcher animals is
due to a mutation in GluRd2 that results in its constitutive
activation (GluRd2-Lc). Yue and collaborators [113]
showed that GluRd2, nPIST and beclin-1 interact, and that
autophagy can be induced by nPIST and beclin-1 synergy
and by the mutated GluRd2-Lc, but not by the wildtype
GluRd2. Dying lurcher Purkinje cells displayed morpholo-
gical features of autophagy in vivo, providing evidence both
for a direct link between GluRd2-Lc receptor and the
autophagic pathway in these cells [113], and that beclin-1
can exert its autophagic functions through interaction with
multiple proteins in the cells [111–113].
Current evidence therefore indicates that that in at least
some instances, autophagy may lead to a caspase-inde-
pendent program of cell demise that fits the concept of
programmed cell death, subject to complex, multivariate
control [Fig. 2]. The next section will examine its relation-
ship with apoptosis.
Apoptosis vs. autophagy
Notwithstanding the abundant evidence for a role for both
autophagy and apoptosis in various diseases, their interplay
in those pathologies is not yet fully understood. In
Parkinson’s disease, the ultrastructural study of neurons
of the substantia nigra of affected patients showed signs of
autophagy, as well as apoptosis [114]. However, it has been
shown that expression of a-synuclein mutants, a condition
frequently found among certain families with Parkinson’s
disease, induces autophagic cell death with no caspase
activation, due to alterations of the ubiquitin-dependent
protein degradation system [115]. This suggests that there
may be no obligatory interdependence between apoptosis
and autophagy in the pathology of this disease.
In Alzheimer’s disease, the endosomal–lysosomal system
was found to be enlarged and highly activated, showing that
endocytosis and/or autophagy are accelerated in the neu-
rons of affected brains, even at early stages. The early
activation of this system could be related to major
Fig. 2. A summary diagram of the major
components identified in pathways leading to
autophagy. These may lead to either cell death
or recovery from an insult such as aminoacid
deprivation.
1642 C. Assunc¸ a
˜o Guimara
˜es and R. Linden (Eur. J. Biochem. 271)FEBS 2004
