PERGAMON
Micron 31 (2000) 97–111
www.elsevier.com/locate/micron
Review
Recent structural insight into mitochondria gained by microscopy
G.A. Perkins*, T.G. Frey
Department of Biology, San Diego State University, San Diego, CA 92182-4614, USA
Received 18 May 1999; received in revised form 26 July 1999; accepted 27 July 1999
Abstract
Novel applications of microscopy have recently provided new insights into mitochondrial structures. Diverse techniques such as high
resolution scanning electron microscopy, transmission electron microscopy, electron microscope tomography and light microscopy have
contributed a better understanding of mitochondrial compartmentalization, dynamic networks of mitochondria, intermembrane bridges,
segregation of mitochondrial DNA and contacts with the endoplasmic reticulum among other aspects. This review focuses on advances
reported in the last five years concerning aspects of mitochondrial substructure or dynamics gained through new techniques, whether they be
novel microscope methods or new ways to prepare or label specimens. Sometimes these advances have produced surprising results and more
often than not, they have challenged current conceptions of how mitochondria work. q 1999 Published by Elsevier Science Ltd. All rights
reserved.
Keywords: Electron microscopy; Electron microscope tomography; High-resolution scanning electron microscopy; Light microscopy; Mitochondria; Mitochondrial structure
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1. Mitochondria: more than just the cell’s power plant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2. Mitochondria and cell death . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3. Thrust of mitochondrial structural research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2. Historical perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1. Baffles or septa? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2. Which model prevailed? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3. What have we learned lately? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1. HRSEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2. Electron tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3. Electron microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4. Light microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4. Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
97
98
98
99
99
100
101
101
102
106
107
109
110
110
1. Introduction
1.1. Mitochondria: more than just the cell’s power plant
* Corresponding author. Tel.: 1 1-619-534-7968; fax: 1 1-619-5347497. NCMIR, University of California, San Diego, La Jolla, CA 920930608, USA.
E-mail address: [email protected] (G.A. Perkins)
One of the most significant differences between prokaryotes and eukaryotes is the possession by eukaryotes of
mitochondria. Indeed, mitochondria are believed to be
derived from bacteria engulfed by primitive eukaryotes
hundreds of millions of years ago. The major function of
0968-4328/99/$ - see front matter q 1999 Published by Elsevier Science Ltd. All rights reserved.
PII: S0968-432 8(99)00065-7
98
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
these organelles (although by no means the only one) is the
synthesis of ATP (adenosine triphosphate) from ADP
(adenosine diphosphate) and inorganic phosphate by the
process of oxidative phosphorylation. In eukaryotic cells,
oxidative phosphorylation occurs only in mitochondria.
Every cell in humans contains between 5 and 2000 mitochondria. Their overall shape is variable. In many instances,
they have a fairly regular shape, either spherical or cylindrical. Spherical mitochondria, for example found in liver,
generally vary from 0.5–5 mm in diameter. Simple cylindrical mitochondria tend to be at least 0.2 mm in diameter and
up to 20 mm long. However, highly branched mitochondria
and mitochondria exhibiting specialized shapes, such as
annuli, discs or cup-shapes have been documented (Munn,
1974). Such shapes increase the surface area to volume
ratio, and may enhance the exchange of metabolites
between mitochondria and the cytosol. Mitochondria
consume over 80% of the oxygen we breathe and make
over 90% of the energy our cells use. They use oxygen to
obtain energy by oxidizing hydrogen rich molecules in food.
This process converts food into ATP, water and carbon
dioxide. ATP is the energy currency used universally by
life on earth. Without ATP, muscles could not contract
and neurons could not fire. Mitochondria literally make it
possible to move and to think. Clearly, an understanding of
the structure and operation of mitochondria and their interaction with other cellular components is essential to
appreciate the success of eukaryotes, from one-celled organisms to the complexity that defines humans.
In order to appreciate how mitochondria are more than
just the cell’s power plant, it is important to grasp the diversity of mitochondrial function, and how defects in function
can lead to disease. Mitochondria in different organisms and
tissues differ dramatically in their consumption of oxygen
and production of ATP. For example, heart mitochondria
consume 50 times the oxygen that liver mitochondria
consume because of the high-energy demand in the continuous pumping of blood. However, liver mitochondria have
adopted a different task and are specialized to detoxify
ammonia, a waste product of protein metabolism, by enzymatic action. Mitochondria also differ in the fuels they can
utilize. For example, heart mitochondria depend primarily
on fatty acids to meet their energy needs. In contrast, liver
mitochondria use both sugars and fats. It is this diversity of
tasks, which reflects the observed diversity of mitochondrial
structure that makes mitochondria integral to cell activity.
The motivation to study how mitochondrial function
relates to structure has recently intensified because defects
in function have now been linked to many of the most
common diseases of aging. Among these are Alzheimer
dementia, Parkinson disease, type II diabetes mellitus,
stroke, atherosclerotic heart disease and cancer. Even such
autoimmune diseases as rheumatoid arthritis, multiple
sclerosis and systemic lupus erythmatosus manifest mitochondrial components. While it cannot yet be stated that
mitochondrial dysfunction causes these diseases, it is clear
that mitochondria are intimately involved because their
function is measurably disturbed, and sometimes gross
morphological changes are observed. Even though, the
causative link of mitochondria in these diseases has not
been substantiated, the role of mitochondria as an orchestrator of apoptosis (programmed cell death) has now been
firmly established (Green and Reed, 1998).
1.2. Mitochondria and cell death
Not only are mitochondria pivotal in fueling cell life, but
they are also pivotal in controlling cell death. The role of
mitochondria in apoptosis has generated a flurry of recent
activity utilizing microscopy to endeavor to detect structural
alterations that may provide a glimpse at the mechanisms of
apoptosis triggered by mitochondria. How do these organelles initiate programmed cell death? At least three general
mechanisms have been proposed, and their effects may be
interrelated (Green and Reed, 1998). They are: (1) release of
proteins, such as cytochrome c, that trigger activation of
caspase family proteases; (2) disruption of electron transport, oxidative phosphorylation and ATP production; and
(3) change in cellular redox potential. Participation of proand antiapoptotic Bcl-2 family proteins in these mechanisms remains a fertile ground for discovery, including the
use of immunomicroscopy to complement the biochemical
assays. In certain apoptosis scenarios, the potential across
the mitochondrial inner membrane (DC m) collapses; this
indicates that the permeability transition (PT) pore, a
nonspecific high conductance channel, has opened.
However, cytochrome c release and activation of caspases
can occur before any detectable loss of DC m, implying that
PT pore opening may occur downstream of caspase activation. Alterations in mitochondrial ultrastructure may accompany the release of cytochrome c, and perhaps other
activators such as apoptosis-inducing factor and intramitochondrial caspases. Two mechanisms have been proposed.
One involves osmotic disequilibrium, which would sequentially expand the matrix space, rupture the outer membrane
and release these caspase activators. The other envisions
opening of channels or transient holes in the outer
membrane that would release cytochrome c without concomitant swelling. Electron microscopy may provide clues as
to how general these two mechanisms are, and whether only
a few mitochondria need to release their caspase activators
to cause cell death, or whether the great majority is
involved.
1.3. Thrust of mitochondrial structural research
In this decade, the thrust of mitochondrial research involving microscopy has progressed along two lines. One, motivated by observations described above, is to attempt to
understand the different signals that converge on mitochondria, whether in normal processes such as apoptosis, or in
diseases. Tried-and-true microscopic techniques are
employed in these endeavors where the novelty is in probing
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
99
Fig. 1. Models of mitochondrial membrane structures. (a) Infolding or “baffle” model, which is the representation (reproduced by permission from Lodish et
al., 1995, Fig. 5-43) most commonly depicted in textbooks. This model originated with Palade in the 1950s and has been prominent until recently. (b) Crista
junction model, which supplants the baffle model for all mitochondria examined to date from higher animals. Electron tomography has been instrumental in
providing the improved 3D visualizations of mitochondria in situ that have generated a new model for membrane architecture. Instead of the large openings
connecting the intracristal space to the membrane space present in the baffle model, narrow tubular openings (crista junctions) connect these spaces in this
model. Most cristae have more than one crista junction and these can be arranged on the same side of the mitochondrial periphery, or on opposite sides if the
crista extends completely across the matrix. (Model courtesy of M. Bobik and M. Martone, University of California, San Diego.)
structural changes based on biochemical stimuli. The other
thrust is the use of novel microscopy on “normal” mitochondria to try to tease out structural features that were either not
visualized or not fully appreciated with the techniques available to early investigators. These novel techniques, which
include three-dimensional (3D) deconvolution light microscopy, high-resolution scanning electron microscopy
(HRSEM) and electron microscope tomography, are providing new and sometimes surprising answers to the questions
of mitochondrial compartmentalization and interactions
with their nearest neighbors. This review focuses on the
latter thrust, which was aptly described by one contributor
to the field as “new views of an old organelle” (Mannella et
al., 1997).
2. Historical perspective
2.1. Baffles or septa?
To understand what constitutes a “new view”, a brief
recapitulation of the controversy concerning mitochondrial
structure that occurred during the formative years of cell
biology as a field may provide perspective as to how the
currently displayed models of mitochondria found their way
into modern textbooks (Rasmussen, 1997). Models of the
structure of mitochondria came principally from the transmission electron microscope (TEM). Views of the internal
structure originated in the 1950s with Palade (1952) and
Sjostrand (1953) and their colleagues, who interpreted the
3D architecture of the multiple, membrane-bound compartments from electron micrographs of sectioned tissue. They
found that mitochondria have two distinct membrane
systems, the outer membrane and the inner membrane.
According to Palade, the inner membrane protrudes into
the mitochondrion interior in a “baffle-like” manner, but
does not go all the way across and connect to the opposite
periphery; i.e. it does not form septa. The leaves or lamellae
that form the baffle were dubbed cristae, and were aligned
more or less orthogonally to the long axis. The baffle
arrangement meant that there was a large opening on the
side of the lamellae facing the periphery. Fig. 1 shows the
salient features. This model of cristae fit well with the opinion
of contemporary biochemists that many of the mitochondrial enzymes involved in oxidative reactions must be
supported in some definite spatial relationship. Palade
thought that the inner membrane baffles might bear these
enzymes in the proper order in linear chains analogous to an
assembly line. This view is consistent with observations that
mitochondria from highly active tissues, e.g. bird flight
muscle, typically have large numbers of tightly packed cristae, whereas mitochondria in organisms living in an almost
anaerobic environment, e.g. tapeworms, contain few cristae.
In contrast to Palade’s model was Sjostrand’s view, that the
cristae are actually a stack of independent membranous
lamellae, which were referred to as “septa”. This view
held that with few exceptions, there was no continuity
between the cristae and peripheral membranes.
How could two such disparate models be proposed from
similar electron micrographs? Part of the difference arose
100
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
because of the necessity of inferring 3D organization from
TEM images of thin sections, which provide projection
images that are inherently 2D. The early debate centered
on who had superior-quality micrographs and on the issue
of specimen preservation. Both groups acknowledged that
Sjostrand’s pictures were better, but a debate ensued over
the best preservation techniques. Sjostrand believed that it
was safest to maintain tissue in a physiological state until
the moment of fixation. Thus, he injected live animals with
fixatives, excised tissue rapidly and immersed it in fixative
prepared in phosphate-buffered saline that was adjusted to
physiological pH and salt content. Palade made no attempt
to employ a physiologically compatible vehicle for preserving tissue because he did not think it was important. It is an
interesting aside that Sjostrand’s and not Palade’s views on
the best preservation techniques have been universally
adopted today.
and 1970s, but was generally ignored in textbooks. Pediculi
seemed to have been forgotten until 1994, when two seminal
papers were published in a special issue of Microscopy
Research and Technique (Lea et al., 1994; Mannella et al.,
1994). These papers focused on new views of mitochondrial structures afforded by recent technical advances in
HRSEM and electron tomography, respectively. The
HRSEM work showed that tubular cristae, and not
lamellar cristae, predominate in many tissue types in
higher animals, while the electron tomography work
confirmed the pediculi concept, and extended the topological mapping of two mitochondrial membrane
systems. Key to these new views was the improved
3D resolution provided by techniques not available to
earlier investigators. The remainder of this review highlights new insights on mitochondrial structure obtained
by novel applications of microscopy.
2.2. Which model prevailed?
3. What have we learned lately?
To understand why Palade’s model gained acceptance,
instead of Sjostrand’s, one need look no further than the
manner used to interpret micrographs. Both agreed almost
completely about what constitutes a good micrograph.
However, for Palade, interpretation flowed from biochemistry to morphology; for Sjostrand, interpretation flowed the
other way. Naturally, Palade’s approach appealed to the
biochemists of the time. Also, Sjostrand (1956) conceded
a little ground in 1956 by admitting that in “isolated cases“
cristae are continuous with the inner bounding membrane.
By the late 1950s, Palade had moved on to other interests,
but Sjostrand directed a meticulous set of 3D reconstructions made by stacking projection images of serial thinsections through mitochondria. From these, it was noted
that narrow stalks might connect cristae to the bounding
membrane. However, Sjostrand qualified this feature by
stating that these stalks rarely appeared in sections of
good preparations, and that there was no fluid in them.
For biochemists of the era, whether Palade’s folds or Sjostrand’s septa accurately described cristae was less important
than the conceptualization of a membranous framework to
support ordered chains of enzymes. Both models support
this framework concept and it was not until the acceptance
of Mitchell’s chemiosmotic theory of oxidative phosphorylation, proposed in 1961 (Mitchell, 1961), that Palade’s
model prevailed. It was easier to accommodate Palade’s
model with chemiosmosis, although the wide-held notion
that Sjostrand’s model is incompatible with Mitchell’s
theory was never substantiated.
Daems and Wisse (1966) reported that cristae attach to
the inner boundary membrane by small tubes, which were
called pediculi. This study involved careful serial sectioning
of well-preserved and stained mitochondria, and the resulting model could be thought of as an improvement on Sjostrand’s model. The pediculi model received attention in
books focusing exclusively on mitochondria in the 1960s
New insights into mitochondrial structure have been
realized with such diverse techniques as HRSEM, electron
tomography, wide-field light microscopy and confocal
microscopy (see Table 1). The excitement of discovering
new facets of mitochondrial structure has initiated a renaissance in imaging mitochondria, leading a new group of
structural biologists to re-examine a question which was
thought to have been answered. Only new results from
“normal” mitochondria will be treated here. The structural
alterations found in diseased mitochondria or through the
action of drugs are beyond the scope of this review.
3.1. HRSEM
When describing mitochondria, one usually emphasizes
the structural features that are common among all or most
mitochondria along with a brief mention of important variations. However, recent HRSEM work emphasized the
importance of structural variety. Lea and Hollenberg
(1989) and Lea et al. (1994) challenged the prevailing
view that cristae in higher animals are invariably lamellar.
By using 3D stereo HRSEM, they showed the existence of
tubular cristae about 30 nm in diameter in every mammalian
tissue type examined, with the exception of brown fat. They
were one of the pioneers to point out the inadequacy of
serial-section reconstructions in delimiting cristae morphology, because the limited z-resolution does not allow for a
reliable mapping of the convoluted membrane topology.
The discovery that tubular cristae are common in higher
animals was a significant link to the bacterial origin of
mitochondria. The main argument against this link had
been the observation that tubular cristae predominate in
lower eukaryotes, which were thought to correspond to
similar tubes found in purple bacterium, believed to be the
organism from which mitochondria evolved. Lamellar cristae were believed to be the only form in higher animals, and
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
101
Table 1
Recent advances in understanding mitochondrial structure using microscopy techniques
Technique
Advance
Reference
HRSEM
Tubular cristae in higher animals, except for lamellar cristae in brown fat
Stalks connecting bands of mitochondria in human muscle fibers
High-resolution examination with metal-less coating
Three kinds of intermembrane bridges
Lea et al. (1994)
Ogata and Yamasaki (1997)
Ogata and Yamasaki (1997)
Senda and Yoshinaga-Hirabayashi (1998)
Electron tomography
Identification of crista junctions
Spatial distribution and dimensions of contact sites
Mapping of the inner membrane topography
Entire mitochondrial structure at 5–10 nm resolution
Measurements of sub-volumes and surface areas from segmented volumes
Mitochondrial morphology in a near-physiological state
Punctate association of mitochondria with the endoplasmic reticulum
Crystal-like cristae membranes in mitochondria from an amoeba
Mannella et al. (1994), Perkins et al. (1997a)
Perkins et al. (1997a, 1998)
Mannella et al. (1997), Perkins et al. (1997a)
Mannella et al. (1994), Perkins et al. (1997a)
Mannella et al. (1997), Perkins et al. (1997b)
Mannella et al. (1998)
Mannella et al. (1996b), Perkins et al. (1997a)
Deng et al. (1998)
Electron microscopy
Visualization of a mitochondrion-dividing apparatus
Observation of plasmid-like DNA in animal mitochondria
Resistance of brain mitochondria to the induction of the permeability
transition
Fusion of mitochondria
Nonfissional means of mitochondrial propagation
Respiration is not required for normal mitochondrial morphology
Kuroiwa et al. (1998)
Endoh et al. (1994)
Hastings et al. (1998)
Light microscopy
Discovery of linear mtDNA molecules in plants and fungi
Single dynamic network of mitochondria in yeast, fibroblasts, and HeLa
cells
Mitochondrial reticulum maintained by a balance of fusion and fission
Growth, retraction, fission and fusion of the reticulum occurring within one
minute
Mitochondrial fusion requires a transmembrane GTPase
Active mtDNA segregation
Microheterogeneity of calcium sensed by the mitochondrial calcium uptake
system
Link of mitochondrial morphology and inheritance
hence, likely did not develop from the tubular form.
However, the existence of tubular cristae in mitochondria
of higher animals (see Fig. 2) removed this argument against
the link between mitochondria and purple bacteria. Furthermore, the observation that striated muscle mitochondria
contain cristae tubes and lamellae in varying proportions,
and that brown fat mitochondria contain only lamellar mitochondria led to the hypothesis that cristae conformation is a
direct consequence of the specialized function of the respective tissue. For example, the lamellar cristae may be a
consequence of the heat-producing aspect of brown fat.
A recent technical advance in HRSEM allowed new
insight into the diversity of the structure of human muscle
fiber mitochondria that was previously unimagined. The
novel technique was to couple osmium-hydrazine impregnation with a well-known aldehyde-osmium-DMSOosmium (A-ODO) procedure (Ogata and Yamasaki, 1997).
The A-ODO exposed large areas of mitochondria, and the
impregnation permitted examination without metal coating
with the potential for increased resolution. Using ultraHRSEM, it was found that mitochondrial morphology
varied significantly between white, red, and intermediate
muscle fibers. Mitochondrial shape and configuration were
Bereiter-Hahn and Voth (1994)
Dai et al. (1998)
Church and Poyton (1998)
Bendich (1996)
Handran et al. (1997), Nunnari et al. (1997),
Rizzuto et al. (1998)
Nunnari et al. (1997)
Rizzuto et al. (1998)
Hermann et al. (1998)
Nunnari et al. (1997)
Rizzuto et al. (1998)
Berger and Yaffe (1996)
simplest in white fibers; paired long thin mitochondria
encircled myofibrils at the level of the I-band. In contrast,
mitochondria in red fibers were stubby, and often were
connected by a slender stalk across the A-band to the next
row of mitochondria. Mitochondria in intermediate fibers
resembled those in red fibers, but were thinner and longer
and possessed a slenderer stalk. This architecture of mitochondria in human muscle fiber types was markedly different from that found in rat muscle types, which was
considered the architypical mammalian muscle. The takehome-lesson from these recent HRSEM studies is that the
diversity of mitochondrial shape and placement may permit
these organelles to respond to different cellular environments. We might take this suggestion one step further and
ask, “can mitochondria in different regions of the same cell
show diversity of structure that is physiologically relevant,
such as might occur when a calcium wave propagates
through a cell?” One can imagine that a calcium wave
might induce a change of inner membrane conformation,
since calcium can affect the energized state of this
membrane. A transition from condensed (shrunken or
condensed matrix volume) to orthodox (expanded matrix
volume) conformation, as defined by Hackenbrock (1968),
102
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
Fig. 2. Ultra fine-structure of mitochondria revealed by HRSEM. Rat pancreas exocrine cells were fixed with 1% osmium tetroxide, postfixed with 0.5%
Tannin, dehydrated through an alcohol series and critical-point dried. Cryofracturing was done in liquid nitrogen and maceration was performed using 0.1%
osmium tetroxide in a microwave oven. No metal coating was used and images were recorded at 3 kV with a HRSEM equipped with a low-voltage, fieldemission in-lens gun. (1) Fracture plane revealing the surface (left) and interior structure (right) of mitochondria (M). The right-hand side mitochondrion
appears to have both tubular and lamellar cristae. Labeled features are outer mitochondrial membrane (OM), inner mitochondrial membrane (IM), cristae (C),
tubular cristae (tC), rough endoplasmic reticulum (rER), ribosome (r). (2) View emphasizing the intermembrane space separating the outer and inner
membranes (arrowheads) and “elemental particles” (arrows), which may be the ATP synthase. (Courtesy of Naoko Yamada, University of California, San
Diego.)
and used subsequently to delimit inner membrane conformation in relation to energy states of mitochondria
described in literature, might be instigated by such a wave.
Using HRSEM, Senda and Yoshinaga-Hirabayashi
(1998) showed that three kinds of intermembrane bridges
exist within the mitochondria that may play a role in the
structural integrity of the two-membrane systems. One
bridge connects the outer and inner boundary membranes.
The other two bridges link adjacent cristae membranes,
either across the intracristal space or across the matrix.
They used a quick-freeze, deep-etch technique with a
wide variety of tissue types and with both chemically
fixed and fresh specimens to demonstrate that these bridges
are a consistent feature that are likely to occur in vivo. It was
proposed that the bridges across the outer and inner boundary membranes might keep these membranes apart and thus
maintain the intermembrane space, or perhaps to prevent its
dilation. The intercristae bridges might help support the
orientation of the cristae.
3.2. Electron tomography
Electron tomography is a powerful technique that is
closely related to computerized axial tomography used by
CAT-scanners in radiological imaging, in that computa-
tional methods are used to calculate a 3D structure from
many 2D images or projections recorded over a wide
range of tilt angles. The first use of electron tomography
with thick-sections of mitochondria described improvements in the 3D visualization of the interior of rat liver
mitochondria (Mannella et al., 1994) over serial-section
reconstructions and HRSEM. It was emphasized that
because HRSEM can only image-exposed surfaces, full
3D information about individual mitochondria could not
be obtained. The limitation with serial-section reconstructions from thin sections is the low z-resolution, which corresponds to the section thickness, typically 30–100 nm, and
with this resolution certain important membranous structures, such as contact sites and crista junctions, cannot be
resolved along this dimension. Crista junctions are the
narrow, tubular openings that connect the cristae
membranes to the inner boundary membrane, and contact
sites are defined as regions where the outer and inner boundary membranes come in such close apposition that they
cannot be separately resolved (Perkins et al., 1997a).
Mannella and coworkers found that to accurately map the
convoluted inner membrane topography, the z-resolution
should be at least 5–10 nm. This resolution is routinely
attained with electron tomography.
Electron tomography presently is the imaging technique
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
Fig. 3. Electron tomography of mitochondria. (a) Two views of a 3D
reconstruction of a Purkinje cell dendritic mitochondrion imaged in situ.
Electron tomography was performed on ,0.5 mm slices of rat brain tissue
using a 400 kV microscope. The outer membrane is shown in dark blue, the
inner boundary membrane in light blue, and the cristal membranes in
yellow. The inner boundary and cristal membranes are continuous surfaces
but were segmented independently to highlight their separate topographies.
(b) Four cristae were segmented independently to demonstrate crista
morphology and connectivity and are displayed together with the inner
boundary membrane. The yellow crista is an example of a lamellar
compartment with multiple crista junctions connecting it to the inner
boundary membrane. The gray crista has seven tubes that connect to the
inner boundary membrane on different sides. Three tubes connect to the
boundary in the foreground and four tubes connect in the background,
which is obscured by the lamellar portion of this crista and the green crista.
A tubular crista with only one junction is displayed in blue near the top of
the mitochondrion just right of center. (c) Three edge views of the yellow
(lamellar) crista (from b) showing the four junctions (black openings) on
this side of the crista. Scale bar ˆ 120 nm:
that provides the highest 3D resolution of the internal structure of mitochondria in thick sections. In contrast to serialsection techniques, where the specimen is sliced as thin as
possible, electron tomography employs sections cut thick
enough, typically 0.25–1.0 mm, to contain a large portion
of the mitochondrial volume. By tilting the section in the
microscope and recording images from the same mitochondrion, a full 3D density distribution can be reconstructed by
computational means (see Fig. 3); hence one is not obliged
to draw inferences from partial surface views or thin slices.
The tilting is performed around one or more axes over an
angular range of typically ^ 608 in tilt increments of 1–28.
Because thick sections are used, it is necessary to use
higher-voltage electron microscopes, ranging from 400 to
3000 kV, in order to obtain suitable images without high
background from inelastically scattered electrons.
Mannella et al. (1994, 1997) showed that the commonly
accepted concept that liver mitochondria have cristae
composed of extended sheet-like, parallel involutions of
103
the inner membrane was incorrect. Instead, they showed
that intracristal spaces in mitochondria with a condensed
matrix (Hackenbrock, 1968) connected both to the mitochondrial periphery and to each other by narrow tubular
extensions, and that these interconnected cristae possess
multiple crista junctions (openings) at the inner boundary
membrane. In contrast, their orthodox mitochondria
(expanded matrix; Hackenbrock, 1968) had cristae with
few interconnections and each had a single, narrow junction,
also tubular, to the inner boundary membrane. Perkins et al.
(1997a,b) obtained similar results from neuronal mitochondrial imaged in situ, but with significant differences. The
neuronal mitochondria were in the orthodox conformation
and contained both tubular and lamellar cristae. Some cristae remained tubular throughout while others merged to
form large or small lamellar cristae. Thus, the lamellar cristae joined the inner boundary membrane only via multiple
crista junctions; the larger the crista the more crista junctions they possessed (see Fig. 3). Perkins and coworkers
found that crista junctions are relatively narrow and uniform
in size (28 ^ 6 nm diameter), and they never observed
connections between lamellar cristae. On the basis of
insights gained from electron tomographic reconstructions
of liver (Mannella et al., 1994, 1997) and neuronal (Perkins
et al., 1997a,b) mitochondria, it was suggested that crista
junctions might be a uniform structural principle. In order to
test the hypothesis that this biomembrane junction is the
governing architecture among all higher animals, Perkins
et al. (1998) examined the structure of brown fat mitochondria by electron tomography, because HRSEM indicated
that brown fat mitochondria contain only lamellar cristae.
They found that both chemically fixed and cryofixed brown
fat mitochondria exhibited no tubular cristae, but instead
had only large lamellar cristae (Perkins et al., 1998).
However, all connections between the inner boundary and
cristae membranes were through tubular crista junctions
virtually identical to those found in neuronal mitochondria.
Furthermore, it was found that the basic cristae architecture
of cryofixed mitochondria was the same as that found in
chemically fixed mitochondria suggesting that while no
fixation procedure is completely free of artifacts, this architectural feature is likely found in vivo.
Mannella et al. (1994, 1997) hypothesized that crista
junctions might represent a barrier to the free diffusion of
ions between the intracristal space and the intermembrane
space. If diffusion is limited, then microcompartmentation
of protons may result in greater pH gradients locally across
the cristae membranes. This hypothesis is of great interest to
bioenergeticists because of Mitchell’s chemiosmotic theory
of ATP generation. The argument is presented as follows. If
instead of crista junctions there were large openings into the
intracristal space, a salient feature of the baffle model, then
protons ejected from the matrix should rapidly diffuse into
the cytosol through the numerous pores in the outer
membrane. So, the pH component of the chemiosmotic
potential would be simply proportional to pHmatrix 2 pHcytosol.
104
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
Fig. 4. Comparison of a double-tilt tomogram with a single-tilt tomogram
of a liver mitochondrion. (a) X, Y and Z slices through a double-tilt tomogram showing improved visualization of the inner and outer membranes
when compared to the same slices through a single-tilt tomogram shown in
(b). The arrowheads point to crista junctions. The two perpendicular lines
indicate where the X slice (horizontal line) and Y slice (vertical line) intersect in comparison with the Z slice. (Reproduced by permission from
Penczek et al., 1995. Double-tilt electron tomography, Ultramicroscopy,
60, 393–410.) (Reproduced by permission from Kuroiwa et al., 1998. The
division apparatus of plastids and mitochondria, Int. Rev. Cytol. 181, 1–
40.)
In contrast, if a proton diffusion barrier existed at crista
junctions, then a greater chemiosmotic potential could be
established locally. The diameters of crista junctions, ca.
30 nm, are very large compared to protons, however, and
if they are fully open, they would not present a barrier to
diffusion of protons. On the other hand, crista junctions
might be a barrier to diffusion of membrane proteins. This
would allow compartmentalization of inner membrane function by segregating different inner membrane proteins
between the inner boundary membrane and the crista
membrane.
As electron tomography has become a more routine tool
in mitochondrial research, a number of technical improvements have been implemented to improve the quality of
reconstructions. One such advance is double-tilt electron
tomography (Penczek et al., 1995). Double-tilt electron
tomography simply consists of collecting a second singletilt series after rotating the specimen by 908 around an axis
perpendicular to the section plane, i.e. the z-axis. Both series
are aligned with respect to each other and the 3D information is merged to more completely fill the sampled Fourier
space. Since the addition of a second tilt series increases
radiation damage to the specimen, a number of tilt selection
schemes have been investigated in which fewer projections
at low tilt angles and more projections at higher tilt angles
were taken, leaving the total number of projections constant.
It was found that a more faithful reconstruction could be
obtained with the double-tilt technique without using a
greater number of tilt images than were used with the
conventional single-tilt technique. Fig. 4 is a side-by-side
comparison of a double-tilt tomogram with a single-tilt
tomogram showing improved visualization of the inner
and outer membranes in the vicinity of the crista junctions.
Technical improvements in computational tools have
recently been made to further the capability to extract quantitative information from tomograms, such as membrane
surface areas, compartment volumes and the distribution
of contact sites (Perkins et al., 1997a, 1998). The interpretation and measurement of the substructures of mitochondria
depend critically on the use of tools for segmenting and
classifying the features of interest, software for making
3D measurements, and programs for interactively visualizing components of the structure. For example, the capability
to segment and classify features of the tomographic volume
was essential for the interpretation of mitochondrial architecture, and led to the full visualization of crista junctions
(Mannella et al., 1994, 1997; Perkins et al., 1997a). Li et al.
(1997) developed another tool for interactive segmentation
that has been successfully applied to mitochondria. The
advantage of their tool is the 3D cursor, which is used to
define and isolate features of interest displayed, and maybe
hidden, inside the volume. Once isolated, these features can
be viewed from any angle, from which measurements can be
made and relationships analyzed.
Computational tools were also recently developed in
order to perform automated tomography. Computer control
of the relevant microscope functions and on-line auto-focus,
cross-correlation and alignment have been implemented to
automate the collection of a tilt series and can achieve low
total electron exposure (Koster et al., 1992, 1997). Automated tomography is especially useful when dealing with
cryo-specimens (Frank, 1995) that are embedded in vitreous
ice and are extremely sensitive to exposure to the electron
beam. The attractiveness of cryo-techniques is the ability to
maintain the sample in physiological states and to monitor
dynamic events through very rapid freezing. By using a
CCD-based imaging system for automated tomography,
Mannella et al. (1998) showed that unfixed, unstained mitochondria quick-frozen in vitreous ice had the same crista
junction morphology found in chemically fixed and stained
mitochondria embedded in plastic.
Because of the relatively high 3D resolution afforded by
electron tomography, new insight into contact sites were
gained (Mannella et al., 1996a; Perkins et al., 1997a,
1998). The outer and inner boundary membranes come
close together with no discernible space between them at
numerous locations called contact sites (Hackenbrock,
1966). These sites are believed to be the main locations at
which proteins imported from the cytosol are transported
across both membranes to the matrix through protein translocation pores (reviewed by Schatz and Dobberstein, 1996).
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
The number of sites of peptide translocation is highly
dependent on the energy state of the mitochondrion, because
an energized inner membrane is required to complete the
passage of the precursor peptide through the outer
membrane (Pfanner and Neupert, 1990; Baker and Schatz,
1991). It is not clear which components are responsible for
the formation of contact sites and whether there exist two
types of contact sites: stable and labile (van der Klei et al.,
1994). Contact sites may be held together by creatine kinase
octomers (Schnyder et al., 1988) that reside in the intermembrane space and/or by the protein translocation
complexes, MOM and MIM (now called TOM and TIM;
Berthold et al., 1995) that would act as protein “grommets”
anchored in the outer and inner membranes.
Perkins et al. (1997a,b)characterized the sizes and distributions of contact sites in orthodox mitochondria observed
in chemically fixed and embedded neuronal tissue by electron tomography. The width of contact sites …,14 ^ 2 nm†
implied a very close association of the two membranes, but
did not support the “semi-fusion” model (Venetie (1982)).
Since the contact site width was not less than twice the
width of a single membrane (,7 nm), there was no basis
to invoke partial membrane fusion to account for them. The
extent of contact sites parallel to the membranes was similar
to their width, 14 ^ 7 nm; suggesting that they may be
formed by a single complex of proteins. By examining
projection images, it was previously hypothesized that
contact sites might be proximal to crista junctions in order
to facilitate transport of proteins destined for the cristae.
However, measurements on tomographic volumes indicate
that contact sites are randomly located with respect to these
junctions. The implications of this result must be reexamined in light of results from cryofixed mitochondria. The
only significant structural difference observed between
cryofixed and chemically fixed brown fat mitochondria is
that in the cryofixed mitochondria, almost all the outer
membranes were close to the inner boundary membrane
with no visible space observed between them (Perkins et
al., 1998). The average width from the outer surface of
the outer membrane to the inner surface of the inner boundary membrane in cryofixed mitochondria was 17 nm, 5 nm
narrower than the comparable measurement in chemically
fixed mitochondria and close to the 14 nm width measured
for contact sites. The close apposition of the outer
membrane and inner boundary membrane essentially over
all the periphery in cryofixed mitochondria indicates that
this entire area is dimensionally competent for protein
import, and that there may be different classes of contact
sites, e.g. stable and labile, with different functions.
In comparing the structures of chemically fixed liver,
neuronal and BAT mitochondria, a number of similarities
stand out, which may be significant for mitochondrial
biogenesis. The outer and inner boundary membranes
were far enough apart to reveal discrete stable contact
sites holding the two together. In all three types of mitochondria, the inner boundary membrane invaginated at
105
numerous sites, the crista junctions, which were tubular
and relatively uniform in diameter. In brown fat mitochondria, the crista junctions were tubular for only a short extent
before broadening into lamellar compartments. One important point to note is that this increase in cristae surface area
did not cause an increase in the diameter of the crista junctions. In liver and neuronal mitochondria, however, the
tubular structure of the crista junctions often persisted to
the extent that some crista were entirely tubular while
other tubular cristae merged to form lamellar cristae of
various sizes. The uniform size and shape of crista junctions
between mitochondria types with different cristae morphologies suggested that there might be some components that
cause their formation and stabilize their structure. This led
to the hypothesis that cristae begin by a budding of the inner
boundary membrane at crista junctions and grow by the
addition of membrane components, perhaps at the junction
site, causing tubular cristae to grow longer and eventually
merge with other cristae to form large lamellar compartments. In contrast to the network of merging tubular cristae
observed in neuronal mitochondria, no such network was
observed in brown fat mitochondria.
Another intriguing feature observed in tomographic
reconstructions is an association between the outer mitochondrial membrane and the endoplasmic reticulum (ER)
membrane (Mannella et al., 1996b, 1998; Perkins et al.,
1997a). These contacts are punctate and about 15 nm in
diameter. Vance and Shiao (1996) showed that the newly
synthesized phosphatidylserine was imported into the mitochondria through contacts with an “endoplasmic reticulumlike mitochondria-associated membrane.” The views of
punctate contacts between ER and mitochondrial
membranes could be a “snapshot” of lipid transfer to the
mitochondrion.
A novel technique, at least as applied to mitochondria, is
to embed unfixed and unstained isolated mitochondria in
vitreous ice (Mannella et al., 1998). These “naked” organelles were rapidly frozen in solution (frozen-hydrated) and
imaged with a cryo-electron microscope at low temperature.
The frozen-hydrated specimens are very sensitive to the
electron beam, and so fewer images in a tilt series could
be collected. Hence, the 3D resolution was necessarily
lower. An automated CCD-based tomography system facilitated data collection. The advantage of this technique is that
direct visualization of mitochondrial substructure close to
the native state can be achieved. This method confirmed the
presence of crista junctions found previously in chemically
fixed and cryofixed mitochondria. The inner boundary
membrane and the outer membrane were observed to be
close together as observed in cryofixed and freeze-substituted specimens, but there was a space between them that
was not observed in the freeze-substituted specimens. The
latter were stained, however, which might have obscured
any space between the two membranes.
Double-tilt electron tomography (Deng et al., 1998) as
well as mathematical modeling (Deng and Mieczkowski,
106
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
Fig. 5. Electron micrographs of sections through C. merolae cut in the
plane parallel to the longitudinal axis of mitochondria (M) and chloroplasts
(C). Nucleus denoted by N. (a) Cell that is about to divide as indicated by
the microbody (arrowhead) in close proximity to the mitochondrion. (b–e)
Formation of the (mitochondrion-dividing) MD ring. When the microbody
contacts the mitochondrion, the MD ring is formed and appears as a long
bar in a tangential section (arrows in b and c) or deposits in cross-section
(large arrows in d and e). (f–i) Contraction of the MD ring. As the MD ring
(large arrows in f–i) contract, they increase in thickness. (j) During cytokinesis, two daughter mitochondria, two daughter chloroplasts, a dividing
cell nucleus and two microbodies are visible.
1998) have also been useful for a structural analysis of the
crystal-like cubic structure found in the cristae membranes
of mitochondria from the unfed amoeba Chaos carolinensis.
By comparing projections along defined lattice lines (e.g.
111) in tomographic reconstructions and a mathematical
model, it was found that the highly curved cristae
membranes could be accurately modeled by periodic
cubic surfaces. It was also found that the matrix was ultracondensed, i.e. confined to a very small continuous space
between enlarged cristae. Even with enlarged cristae, the
crista junction architecture persisted. Because the cristae
membranes can be modeled mathematically, studies of
geometrical distortions in the process of tomographic data
collection and studies of cristae organization in relation to
mitochondrial bioenergetics are possible.
3.3. Electron microscopy
Even though a tremendous amount of information on
mitochondrial structure has been obtained over the past
almost 50 years by the conventional TEM, new aspects
continue to be discovered. New insight into mitochondrial
biogenesis (sometimes called mitochondriokinesis) involving a mitochondrion-dividing apparatus (MD ring) has
recently been reported (Kuroiwa et al., 1998). It has long
been known that mitochondria multiply in the cell cytoplasm by binary division of preexisting organelles.
However, until recently, no rigid mitochondrial contractile
ring had been found. The existence of an MD ring was
hypothesized by the extrapolation of the FtsZ ring found
in dividing prokaryotes and the contractile ring characterized in eukaryotes. The first structural evidence for the MD
ring was found in mitochondria of the red alga Cyanidioschyzon merolae (Fig. 5). This ring was not observed
sooner because of the small size of the dividing apparatus,
the lack of specific staining for it, and the finding that mitochondrial division is generally not synchronized in eukaryotic cells. The characterization of division in C. merolae,
from the formation of the MD ring during the early stage to
its disappearance during the late stage, was made possible
because the cell cycle in this eukaryote could be made
synchronous in a population of cells. As an aside, to visualize the MD ring in other eukaryotes, it may be necessary to
label and locate by immunoelectron microscopy one of its
components, as was done with the bacterial contractile ring.
A part of the mitochondrial division involves the partitioning of DNA. Mitochondria in some organisms have a
nucleoid, which can be visualized by DAPI staining, where
the DNA is organized by proteins. During mitochondriokinesis, the DNA molecules, which are ,1=100 the size of
bacterial DNA, were separated equally between daughter
mitochondria. The mechanism of separation appeared to
be under different control than the MD ring constriction.
The MD ring actually consists of two rings, an outer ring
and an inner ring. The outer ring is in the cytoplasm and the
inner ring is in the matrix adjacent to the inner boundary
membrane at the constricted isthmus of dividing mitochondria. During constriction, the outer ring thickened while the
inner ring remained unchanged. It was suggested that the
outer ring generates the motive force during division and the
inner ring may simply be a remnant of the bacterial contractile ring (FtsZ ring). The MD ring was observed to be a
bundle of fine filaments. Interestingly, this ring was not
labeled by anti-actin antibody or by phalloidin. Furthermore, division was not inhibited by cytochalasin B.
Hence, it does not appear that actin is a component of the
MD ring and its composition remains unsure. However, it
was determined that the MD ring proteins are encoded by
genes in the nuclear genome, and that the growth of this ring
is independent from nuclear DNA synthesis. In summary,
the identification of the MD ring by conventional electron
microscopy has added insight into the division mechanism
of mitochondria that is similar in principle to the division of
bacteria, plastids and eukaryotic cells.
On a smaller scale, electron microscopy has provided
views of plasmid-like DNA in mitochondria in the animal
kingdom (Endoh et al., 1994). Numerous kinds of mitochondrial plasmids have been reported in fungi and plants.
However, until recently such plasmids were not known
among mitochondria of animals. Mitochondrial DNA from
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
Paramecium caudatum was isolated by electrophoresis into
bands of three distinct types. Mitochondrial genomic DNA
was found in a high molecular weight band and was designated as mtDNA-type. The plasmid-like DNAs were designated either type I or type II DNAs depending on the
molecular weight of the bands. Remarkable features of
type II DNA strands were visualized in electron micrographs. One observed feature was a hairpin structure that
had a loop at one terminus and a protruding segment of
single-stranded DNA at the other terminus. Another feature
was the presence of stable dimer structures. The monomers
paired with base pairing in opposite directions suggest the
presence of inverted repeats. It was hypothesized that these
unusual dimers may aid DNA replication by having the
constituent monomers serve as a primer for each other.
Another viewpoint is that, type II DNAs may not be plasmids, but rather parasitic DNAs. In either case, the observation of plasmid-like DNA in mitochondria is intriguing
because of how they might have been inserted, the implication for replication mechanism, and whether they can replicate autonomously or are under the control of host
mitochondria.
While tissue-dependent differences in mitochondrial
morphology have been documented ever since thin sections
of tissue were first examined in electron microscope, it has
been generally assumed that mitochondrial function and
response to stimuli are similar across tissues. The notable
exception to functional homogeneity is brown fat mitochondria, which produce heat by uncoupling ATP production
from the electron transport chain. A new study of brain
and liver mitochondria provided evidence that mitochondrial response to stimuli is not as homogeneous as assumed
previously. Interest in the PT pore has been increasing
because of its putative role in the response of brain cells
to ischemia and excitotoxic agents. The PT pore has been
studied mostly in isolated liver mitochondria, and it has
been assumed that the properties discovered in this model
system would be the same in brain mitochondria. However,
through the use of electron microscopy and biochemical
tools, it was found that those chemicals which open the
PT pore and cause swelling and loss of glutathione in
liver mitochondria, did not produce the same effect in
isolated brain mitochondria (Hastings et al., 1998). The
implication is that brain mitochondria are more resistant
to the induction of permeability transition. It can be argued
from these results that properties found in mitochondria of
one tissue type may not translate easily to mitochondria of
other tissue types.
While fusion of adjacent mitochondria has been observed
for many years at the level of light microscopy, only
recently have fusion events been visualized at the level of
electron microscopy. Mitochondria do not form de novo, but
assemble from pre-existing mitochondria by fission and
fusion events. By using phase-contrast light microscopy,
Bereiter-Hahn and Voth (1994) were able to record video
sequences of fusing mitochondria. Fixation was used to
107
arrest these mitochondria in various stages of fusion, and
thin-section electron microscopy provided high-resolution
images of this process. By employing correlative light and
electron microscopies, it was possible to differentiate
between fusing mitochondria and mitochondria that were
transiently close to one another. Fusion does not appear to
depend on physiological condition, but rather on mitochondrial motility. It was seen that the peripheral mitochondrial
membranes in the contact zone, where early fusion events
take place, exhibited a high electron density. This high
density might reflect clustering of membrane proteins
involved with fusing the outer and inner boundary
membranes of the apposed mitochondria. It was found
that the fusing regions were commonly devoid of cristae,
which also argues for the localized presence of specialized
fusion machinery in the regions of the merging peripheral
membranes. While fusion of the peripheral membranes
occurred on a relatively fast time-scale …,1–3 s†; rearrangement of cristae progressed on a longer time scale (.20 s).
Rearrangement in the transition region between two mitochondria occurs via the elongation of cristae. Everywhere
else, the cristae maintain the same arrangement that existed
in the separate mitochondria before merging. De novo cristae, hypothesized to bud from crista junctions, have yet to be
documented in the transition region or for that matter
anywhere else.
Electron microscopic evidence that mitochondria may be
capable of propagating by means other than fission was
recently published (Dai et al., 1998). A slow-sedimenting
(ss-) mitochondrion was identified in mitochondrial fractions and observed in situ within rapidly growing mung
bean seedlings. Besides possessing cristae (in the larger,
but undetected in smaller ss-mitochondria), these mitochondria had cytochrome oxidase subunit III (detected in situ
with immunoelectron microscopy) and mitochondrial
DNA (mtDNA). Smaller than their neighbors (70–300 nm
in size), ss-mitochondria lacked demonstrable respiratory
capability, and were almost always in contact with a filament-aligned membrane-like structure. The correlation of
size with the appearance of sparse cristae suggests that
these mitochondria may not be static, but rather intermediates in a developmental pathway leading to respiratorycompetent mitochondria. Further, despite several unique
mitochondrial characteristics, ss-mitochondria appeared to
be immature structurally as well as functionally. Binary
fission of mitochondria with more classical structure was
also detected in mung bean seedlings. These observations
imply that small ss-mitochondria may be “nascent” mitochondria from a yet unidentified mitochondria-propagation
scheme, and that mitochondria may be capable of propagating by more than one pathway.
3.4. Light microscopy
A novel technique for imaging mtDNA has shown that
heterogeneity of mitochondrial ultrastructure extends to
108
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
Fig. 6. High-speed 3D imaging of mitochondria and the endoplasmic
reticulum (ER) using wide-field deconvolution microscopy. (a) HeLa cell
transiently expressing mtGFP. Each image was recorded 30 s apart. The
mitochondrial network (reticulum) is constantly rearranging. (b) 3D image
of mitochondria taken with a 100 × objective lens and computationally
deblurred. Note the mitochondrial reticulum. (c) Recovery of mtGFP fluorescence after photobleaching. The first and second images were recorded
immediately before and after photobleaching. The following three images
were recorded at 2 min intervals after and the final image 30 min after
photobleaching. The rapid recovery indicates an interconnected mitochondrial network with luminal continuity. (d) Combined 3D imaging of mitochondria and ER in a HeLa cell transiently expressing mtGFP and erGFP.
The two 3D images are superimposed. Mitochondria are in red, ER in green
and the overlaps of mitochondria and ER result in a white color. The
mitochondrial reticulum in close proximity to the ER was estimated to be
,5–20% of the total. (Reproduced by permission from Rizzuto et al., 1998.
Close contacts with the endoplasmic reticulum as determinants of mitochondrial calcium responses. Science, 280, 1763–1766.)
DNA molecules as well (Bendich, 1996). Molecules of
mtDNA in the 50–150 kb size region from various fungi
and plants were obtained from gel bands and after soaking
in ethidium bromide to make fluorescent were analyzed as
moving pictures during electrophoresis with video fluorescence microscopy. The lengths of mtDNA in fields of
between 10 and 100 molecules were measured and their
structures were noted (circular or linear) by examining
frames of videotape. The longest molecules were measured
and compared to a size standard (lambda DNA, 48.5 kb)
after encounters with agarose fibers had extended the
strands to their maximum length. It was found that most
of the fractionated mtDNA were linear and ranged in size
from 50 to 150 kb. Very little of the mtDNA was in the
conformation of large circular molecules. This finding is
important because nearly all mtDNA extracted from
metazoan animal cells has been in the form of genomesized circular molecules. It was expected that circular mitochondrial genomes would also be the norm in other
organisms. Circular molecules are universally depicted in
reviews and textbooks describing mtDNA of fungi and
plants. Moreover, recent models of recombination,
rearrangement and loss of plant mtDNA sequences were
based on the assumption of circular genomes. Since the
existence of circular mtDNA molecules has been established in only a subset of eukaryotes, the conformation
from a diverse sampling of organisms needs to be demonstrated, and more realistic models need to be constructed for
those genomes with known linear conformation.
A recent application of wide-field fluorescence microscopy has revealed that the structure of mitochondria in
yeast has direct consequences for cytoplasmic inheritance
(Nunnari et al., 1997). In haploid cells of opposite mating
type, mtDNA and mitochondrial proteins were labeled with
different fluorescent dyes. After mating, it was found that
parental mitochondrial protein markers were redistributed
rapidly throughout the zygotes, providing evidence that
parental mitochondria fuse and their constituent proteins
intermix. In fact, it appears that yeast mitochondria form a
single dynamic network of compartments, and its interconnectedness is maintained by a balance of fusion and fission
events. The generation of such a mitochondrial reticulum is
thought to facilitate signaling, and passage of metabolites
and energy throughout cells (Ichas et al., 1997). One of the
fluorescent markers was a matrix-targeted GFP. Its
complete mixing in the zygote indicated that fusion of
both outer and inner membranes occurred. Of necessity,
fusion of separate membranes requires distinct fusion
steps. It is likely that both steps involve a unique fusion
apparatus operating on the respective membranes. In
contrast to the mixing of proteins after fusion, the mtDNA
remained separately localized to each half of the zygote.
The lack of diffusion through the mitochondrial network
implied that daughter cells could have three different mitochondrial genotypes. Those that bud from either end are
likely to contain mtDNA from only one of the two parents,
but those that bud from the middle might contain mtDNA
from both parents. The precise mechanism for mtDNA
transmission during mating is unknown; however, lack of
diffusion might be an important vehicle for controlling the
accurate inheritance of the mitochondrial genome. This
further suggests that active mechanisms for maintaining
separate mtDNA are involved, which may include anchoring of mtDNA molecules to specific locations within the
mitochondrion. Active mtDNA segregation offers a structural explanation for the nonrandom transmission patterns
that have been observed genetically for years now.
A dynamic network of largely interconnected mitochondria was also observed in normal cultured fibroblasts
(Handran et al., 1997) and in an immortalized human cancer
cell line (HeLa cells; Rizzuto et al., 1998) (Fig. 6). The
vitality of fibroblast mitochondria was monitored with the
vital dyes JC-1 and TMRE. Confocal light microscopy was
used in this study to collect 0.4 mm optical sections. Each
optical section was the average of 64 frames recorded at
a rate of 30 frames/s. The stack of sections was deconvolved and rendered into an image that simulates 3D
space. These images displayed a reticulum of mitochondrial
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
compartments. By using two differently colored green fluorescent proteins specifically targeted to either mitochondria
or ER in HeLa cells, the spatial relation between these
organelles was analyzed by Rizzuto and coworkers with a
high-speed wide-field fluorescence imaging system (Fig.
6c,d). 3D reconstructions were made by optical sectioning
and were computationally deblurred to reveal that the mitochondrial network was mostly tubular and underwent
continuous rearrangement (Fig. 6b). The imaging system
used laser illumination, which allowed 128 × 128 pixel
images to be collected in 30 ms using a low-noise CCD
camera. With this high-speed system, 41 optical sections
with a z-step of 0.25 mm were collected in ,1 s. Stacks of
optical sections were taken 30 s apart. 3D reconstructions of
mitochondria from these stacks demonstrated that growth,
retraction, fission and fusion had occurred within 1 min of
observation indicating a high structural plasticity in agreement with cinematic recordings using phase contrast
microscopy (Bereiter-Hahn and Voth, 1994). The interconnectedness of mitochondrial networks was confirmed by the
recovery of fluorescence within 30 min after photobleaching
a portion of the network.
The binding of calcium to aequorin produces luminescence. Rizzuto et al. (1998) used aequorin chimeras specifically targeted to the cytosol, mitochondrial intermembrane
space, or matrix to monitor the calcium concentrations
sensed by the mitochondrial calcium uptake systems.
Through the opening of IP3-gated channels, calcium is
released from the ER, the intracellular calcium storehouse.
It was shown that the areas of close apposition between the
ER and mitochondria were also the areas where an increase
in calcium concentration saturated the binding of calcium to
aequorin in the mitochondrial intermembrane space. It thus
appeared that at contacts between the ER and mitochondria,
microdomains of high calcium concentration were generated, which could allow for a rapid uptake of a relatively
large amount of calcium by mitochondria. After uptake,
there appeared to be a rapid diffusion of calcium within
the mitochondrial matrix, because a large portion of the
matrix aequorin discharged its luminescence. This rapid
diffusion might be a mechanism for quickly tuning mitochondrial metabolism to the cell’s needs for ATP. The
microheterogeneity of the calcium signal emphasizes the
importance of the cellular distribution of mitochondria in
relation to the ER.
Mitochondrial morphology in yeast may be linked to
inheritance, i.e. to the mechanism of mitochondrial fission.
Coupling light and electron microscopy with genetic
analyses, it has been shown that mutations in genes required
for normal inheritance and distribution alter mitochondrial
morphology (Berger and Yaffe, 1996). These genes encode
proteins found in either the outer or inner boundary
membranes. Instead of the normal mitoplasmic reticulum,
these mutant yeast exhibited either small round mitochondria (Fisk and Yaffe, 1997), or giant round mitochondria
(Berger et al., 1997) depending on the mutation.
109
Using light microscopy, Visser et al. (1995) showed that
the number of mitochondria in yeast can vary from one to
more than 40 depending on the growth phase, the oxygen
availability and the carbon source. When yeast are grown on
a carbon source that represses genes for respiratory proteins,
2–3 long, branched mitochondria are observed that lie near
the cell periphery. However, when a carbon source is used
that does not repress these genes, many small mitochondria
are seen dispersed throughout the cell. More recently,
Church and Poyton (1998) used light and electron microscopy to show that neither assembled cytochrome oxidase,
nor cellular respiration is required for normal mitochondrial
morphology or volume in yeast. This work suggests that the
morphological changes documented by Visser and coworkers are not a consequence of respiratory capacity, but rather
to another, as yet undetermined factor. It was pointed out
that although respiration is not involved in the maintenance
of morphology, oxygen availability and glucose repression
might influence other metabolic processes, which in turn
may affect mitochondrial morphology. An interesting question is whether mitochondrial morphology can be governed
by the principle of symmorphosis (Weibel et al., 1991),
which has as its tenet that “no more structure is formed
and maintained than is required to satisfy functional
needs”. If symmorphosis can be applied to yeast mitochondria, then the observation that mitochondrial morphology is
normal in the absence of respiration implies that mitochondrial structure is maintained for purposes other than respiration. It is important to note that in addition to being the site
for oxidative phosphorylation, mitochondria also synthesize
certain amino acids and precursors for the synthesis of pyrimidines and heme. This capability makes mitochondria
essential to the normal operation of eukaryotic cells,
whether they are grown aerobically or anaerobically.
4. Perspectives
What has allowed the advances in our understanding of
mitochondrial structures reviewed here? First and foremost,
the capabilities of microscope technology have been
improving. This is most evident in electron tomography
and SEM, where improvement in technique in the latter
has added the superlatives “ultra high-resolution” to SEM.
It should be emphasized again that electron tomography
increased the resolution attained in 3D reconstructions of
whole mitochondria by about an order of magnitude over
that attained by the technique previously used, serial-section
reconstruction. It is no wonder then that with this increased
resolution new facets of membrane topography were
revealed. As with most scientific endeavors that utilize
computers, microscopy has benefited from the revolution
in computer technology experienced this decade. Nowhere
is this more apparent than in the resurgence of light microscopy, where the capability to monitor dynamic events by
fast recording media is made possible by fast computers and
110
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
CCD cameras. Computers that can handle the analysis and
storage of large volumes of data have also been instrumental
in the growth of electron tomography. Moreover, the availability of new software tools for viewing and analyzing
segmented volumes have been invaluable for extracting
hidden features and understanding spatial relationships.
Finally, the advent of a greater number and more sensitive
biochemical probes that can be used in conjunction with
various microscopies has broadened how we can dissect
working mitochondria.
Acknowledgements
TGF acknowledges support from Grant-in-Aid #98-256A
from the Western States Affiliate of the American Heart
Association. Some of the work included here was conducted
at the National Center for Microscopy and Imaging
Research at San Diego, which is supported by NIH Grant
RR04050 to Mark H. Ellisman.
References
Baker, K.P., Schatz, G., 1991. Mitochondrial proteins essential for viability
mediate protein import into yeast mitochondria. Nature 349, 205–208.
Bendich, A.J., 1996. Structural analysis of mitochondrial DNA molecules
from fungi and plants using moving pictures and pulsed-field gel electrophoresis. J. Mol. Biol. 255, 564–588.
Bereiter-Hahn, J., Voth, M., 1994. Dynamics of mitochondria in living
cells: shape, changes, dislocations, fusion, and fission of mitochondria.
Microsc. Res. Tech. 27, 198–219.
Berger, K.H., Yaffe, M.P., 1996. Mitochondrial distribution and inheritance. Experientia 52, 1111–1116.
Berger, K.H., Sogo, L.F., Yaffe, M.P., 1997. Mdm12p, a component
required for mitochondrial inheritance that is conserved between
budding and fission yeast. J. Cell Biol. 136, 545–553.
Berthold, J., Bauer, M.F., Schneider, H.-C., Klaus, C., Dietmeier, K.,
Neupert, W., Brunner, M., 1995. The MIM complex mediates preprotein translocation across the mitochondrial inner membrane and couples
it to the mt-Hsp70/ATP driving system. Cell 81, 1085–1093.
Church, C., Poyton, R.O., 1998. Neither respiration nor cytochrome c
oxidase affects mitochondrial morphology in Saccharomyces cerevisiae. J. Exp. Biol. 201, 1729–1737.
Daems, W.T., Wisse, E., 1966. Shape and attachment of the cristae mitochondriales in mouse hepatic cell mitochondria. J. Ultrastruct. Res. 16,
123–140.
Dai, H., Lo, Y.-S., Jane, W.-N., Lee, L.-W., Chiang, K.-S., 1998. Population heterogeneity of higher-plant mitochondria in structure and function. Eur. J. Cell Biol. 75, 198–209.
Deng, Y.R., Mieczkowski, M., 1998. Three-dimensional periodic cubic
membrane structure in the mitochondria of amoebae Chaos carolinensis. Protoplasma 203, 16–25.
Deng, Y., Marko, M., Buttle, K.F., Leith, A., Mannella, C.A., 1998. Mitochondria in Chaos. Albany Conference Abstracts.
Endoh, H., Yazaki, K., Takahashi, M., Tsukii, Y., 1994. Hairpin and dimer
structures of linear plasmid-like DNAs in mitochondria of Paramecium
caudatum. Curr. Genet. 27, 90–94.
Fisk, H.A., Yaffe, M.P., 1997. Mutational analysis of Mdm1p function in
nuclear and mitochondrial inheritance. J. Cell Biol. 138, 484–494.
Frank, J., 1995. Approaches to large-scale structures. Curr. Op. Struc. Biol.
5, 194–201.
Green, D.R., Reed, J.C., 1998. Mitochondria and apoptosis. Science 281,
1309–1312.
Hackenbrock, C.R., 1966. Ultrastructural bases for metabolically linked
mechanical activity in mitochondria. I. Reversible ultrastructural
changes with change in metabolic steady state in isolated liver mitochondria. J. Cell Biol. 30, 269–297.
Hackenbrock, C.R., 1968. Chemical and physical fixation of isolated mitochondria in low-energy and high-energy states. Proc. Natl. Acad. Sci.
USA 61, 598–605.
Handran, S.D., Werth, J.L., DeVivo, D.C., Rothman, S.M., 1997. Mitochondrial morphology and intracellular calcium homeostasis in cytochrome oxidase-deficient human fibroblasts. Neurobiol. Dis. 3, 287–
298.
Hastings, T.G., Watkins, S.W., Berman, S.B., 1998. Evidence that brain
mitochondria do not readily undergo permeability transition. Albany
Conference Abstracts.
Hermann, G.J., Thatcher, J.W., Mills, J.P., Hales, K.G., Fuller, M.T.,
Nunnari, J., Shaw, J.M., 1998. Mitochondrial fusion in yeast requires
the transmembrane GTPase Fzo1p, J. Cell Biol. 143, 359–373.
Ichas, F., Jouaville, L.S., Mazat, J.P., 1997. Mitochondria are excitable
organelles capable of generating and conveying electrical and calcium
signals. Cell 89, 1145–1153.
Koster, A.J., Chen, H., Sedat, J.W., Agard, D.A., 1992. Automated microscopy for electron tomography. Ultramicroscopy 46, 207–227.
Koster, A.J., Grimm, R., Typke, D., Hegerl, R., Stoschek, A., Walz, J.,
Baumeister, W., 1997. Perspectives of molecular and cellular electron
tomography. J. Struct. Biol. 120, 276–308.
Kuroiwa, T., Kuroiwa, H., Sakai, A., Takahashi, H., Toda, K., Itoh, R.,
1998. The division apparatus of plastids and mitochondria. Int. Rev.
Cytol. 181, 1–40.
Lea, P.J., Hollenberg, M.J., 1989. Mitochondrial structure revealed by high
resolution scanning electron microscopy. Am. J. Anat. 184, 245–257.
Lea, P.J., Temkin, R.J., Freeman, K.B., Mitchell, G.A., Robinson, B.H.,
1994. Variations in mitochondrial ultrastructure and dynamics observed
by high resolution scanning electron microscopy (HRSEM). Micros.
Res. Tech. 27, 269–277.
Li, Y., Leith, A., Frank, J., 1997. Tinkerbell—a tool for interactive segmentation of 3D data. J. Struct. Biol. 120, 266–275.
Lodish, H., Berk, A., Matsudaira, P., Baltimore, D., Zipursky, L., Darnell,
J., 1995. Molecular Cell Biology, 3rd. Scientific American Books, New
York.
Mannella, C.A., Marko, M., Penczek, P., Barnard, D., Frank, J., 1994. The
internal compartmentation of rat-liver mitochondria: tomographic study
using the high-voltage transmission electron microscope. Micros. Res.
Tech. 27, 278–283.
Mannella, C.A., Buttle, K.F., O’Farrell, K.A., Leith, A., Marko, M., 1996.
Structure of contact sites between the outer and inner mitochondrial
membranes investigated by HVEM tomography. Proc. Microsc. Microanal., 966–967.
Mannella, C.A., Marko, M., Buttle, K.F., O’Farrell, K., Tessitore, K., 1996.
Three-dimensional structure of rat liver mitochondria: interactions of
the outer membrane with the inner membrane and endoplasmic reticulum. Albany Conference Abstracts.
Mannella, C.A., Marko, M., Buttle, K.F., 1997. Reconsidering mitochondrial structure: new views of an old organelle. TIBS 22, 37–38.
Mannella, C.A., Buttle, K.F., Pfeiffer, D.R., Rath, B.K., Marko, M., 1998.
Mitochondrial design: implications for the structural basis of the permeability transition. Albany Conference Abstracts.
Mitchell, P., 1961. Coupling of phosphorylation to electron and hydrogen
transfer by a chemi-osmotic type of mechanism. Nature 191, 144–148.
Munn, E.A., 1974. The Structure of Mitochondria, Academic Press, New
York.
Nunnari, J., Marshall, W.F., Straight, A., Murray, A., Sedat, J.W., Walter,
P., 1997. Mitochondrial transmission during mating in Saccaromyces
cerevisiae is determined by mitochondrial fusion and fission and the
intramitochondrial segregation of mitochondrial DNA. Mol. Biol. Cell
8, 1233–1242.
G.A. Perkins, T.G. Frey / Micron 31 (2000) 97–111
Ogata, T., Yamasaki, Y., 1997. Ultra-high-resolution scanning electron
microscopy of mitochondria and sarcoplasmic reticulum arrangement
in human red, white, and intermediate muscle fibers. Anat. Rec. 248,
214–223.
Palade, G., 1952. The fine structure of mitochondria. Anat. Rec. 114, 427–
451.
Penczek, P., Marko, M., Buttle, K., Frank, J., 1995. Double-tilt electron
tomography. Ultramicroscopy 60, 393–410.
Perkins, G., Renken, C., Martone, M.E., Young, S.J., Ellisman, M., Frey,
T., 1997. Electron tomography of neuronal mitochondria: three-dimensional structure and organization of cristae and membrane contacts. J.
Struct. Biol. 119, 160–172.
Perkins, G., Young, S., Renken, C., Song, J.Y., Lamont, S., Martone, M.,
Lindsey, S., Frey, T., Ellisman, M., 1997. Electron tomography of large,
multicomponent biological structures. J. Struct. Biol. 120, 219–227.
Perkins, G.A., Song, J.Y., Tarsa, L., Deerinck, T.J., Ellisman, M.H., Frey,
T.G., 1998. Electron tomography of mitochondria from brown adipocytes
reveals crista junctions. J. Bioenergetics Biomembranes 30, 431–442.
Pfanner, N., Neupert, W., 1990. The mitochondrial protein import apparatus. Annu. Rev. Biochem. 59, 331–353.
Rasmussen, N., 1997. Picture Control: the Electron Microscope and the
Transformation of Biology in America, 1940–1960, Stanford University Press, Stanford, CA.
Rizzuto, R., Pinton, P., Carrington, W., Fay, F.S., Fogarty, K.E., Lifshitz,
L.M., Tuft, R.A., Pozzan, T., 1998. Close contacts with the endoplasmic
reticulum as determinants of mitochondrial calcium responses. Science
280, 1763–1766.
Schatz, G., Dobberstein, B., 1996. Common principles of protein translocation across membranes. Science 271, 1519–1526.
111
Schnyder, T., Engel, A., Lustig, A., Wallimann, T., 1988. Native mitochondrial creatine kinase forms octameric structures. II. Characterization of
dimers and octamers by ultracentrifugation, direct mass measurements
by scanning transmission electron microscopy, and image analysis of
single mitochondrial creatine kinase octamers. J. Biol. Chem. 263,
16954–16962.
Senda, T., Yoshinaga-Hirabayashi, T., 1998. Intermembrane bridges within
membrane organelles revealed by quick-freeze deep-etch electron
microscopy. Anat. Rec. 251, 339–345.
Sjostrand, F.S., 1953. Electron microscopy of mitochondria and cytoplasmic double membranes. Nature 171, 30–31.
Sjostrand, F.S., 1956. The ultrastructure of cells as revealed by the electron
microscope. Int. Rev. Cytol. 5, 455–533.
van der Klei, I.J., Veenhuis, M., Neupert, W., 1994. A morphological
view on mitochondrial protein targeting. Micros. Res. Tech. 27,
284–293.
Vance, J.E., Shiao, Y.J., 1996. Intracellular trafficking of phospholipids:
import of phosphatidylserine into mitochondria. Anticancer Res. 16,
1333–1339.
van Venetie, R., Verkleij, A.J., 1982. Possible role of non-bilayer lipids in
the structure of mitochondria: a freeze-fracture electron microscopy
study. Biochim. Biophys. Acta 692, 397–405.
Visser, W., van Spronsen, E.A., Nanninga, N., Pronk, J.T., Gijs Kuenen, J.,
van Dijken, J.P., 1995. Effects of growth conditions on mitochondrial
morphology in Saccaromyces cerevisiae. Antonie Van Leeuwenhoek
67, 243–253.
Weibel, E.R., Taylor, C.R., Hoppeler, H., 1991. The concept of symmorphosis: a testable hypothesis of structure-function relationship. Proc.
Nat. Acad. Sci. USA 88, 10357–10361.