TiBS 22 - FEBRUARY 1997
Reconsidering mitochondrial
structure: new views of
an old organelle
difficulties for interpreting complicated
three-dimensional structures like cristae.
Examining consecutive sections through
cells helps somewhat, but cannot fully
resolve the internal structure of the
organelles3'1, in part owing to practical
limitations in obtaining series of suffi
ciently thin (5-10 nm) sections.
Tomography of rat-liver mitochondria
In a recent Reflections article, Rasmussen1
recounted the controversy about mito
chondrial structure that developed dur
ing the formative years of the field of cell
biology. A key issue regarded the organiz
ation of the infoldings of the mitochon
drial inner membrane, the cristae, and
the connectivity of the compartments
they enclose with the space between the
outer and inner membranes. One group
held that cristae are shaped like baffles
and that the intracristal spaces communi
cate freely with the peripheral compart
ment. The opposing camp argued that
cristae define discrete compartments
that have no (or only minimal) openings
into the intermembrane space.
The differences between the two mito
chondrial models are of more than pass
ing interest to bioenergeticists. Energytransducing chemiosmotic gradients are
established by respiration-driven trans
port of protons across the mitochondrial
inner membrane. According to the gen
erally accepted 'baffle' model, protons
ejected from the matrix (the compart
ment bounded by the inner membrane)
should rapidly diffuse from the cristae,
and into the cytosol through the numer
ous pores in the outer membrane2. Thus,
it is usually assumed that the pH com
ponent of the chemiosmotic potential is
simply proportional to pHmatnx-pHcytosol.
If, instead, diffusion between the intra
cristal and peripheral spaces were re
stricted, microcompartmentation of pro
tons might result in locally greater pH
gradients across the cristal membranes.
The recently developed technique of
electron microscopic tomography5 over
remains to be learned about the interior comes many of the limitations associated
design of mitochondria. Novel techniques with TEM of thin sections. Instead of
not available to early investigators have cutting the specimen as thin as possible,
begun to provide new and sometimes sections are cut that are thick enough
surprising answers to the question of to contain a representative portion of the
mitochondrial compartmentation.
object of interest, 0.25-1.0 jim in the case
Much of the early debate about mito of mitochondria. Numerous images are
chondrial structure centered around recorded from the same field, each micro
images obtained by transmission elec graph representing a different projection
tron microscopy (TEM) of sections of direction. This is achieved by tilting the
tissue typically 50-80 nm thick. Because specimen around one or more axes over
such projection images are inherently a wide angular range, typically **/- 60°
two-dimensional, they present obvious in increments of 1-2°. Imaging of these
Conventional transmission electron
microscopy
In describing the process by which
the baffle model gained acceptance,
Rasmussen stressed that reaching
'proper conclusions' hinges on deter
mining 'proper methodology'. However,
as practicing biologists are all too aware,
proper techniques might not be avail
able at a particular point in time. Thus,
while the baffle model finds its way into
textbooks based on a consensus dating
Figure 1
Three-dimensional reconstruction of an isolated, condensed rat-liver mitochondrion, obtained
by double-tilt electron tomography7, (a) Stacked contours showing the outer membrane
(red), the inner membrane (periphery, yellow; cristae, blue), and matrix granules (yellow).
The mitochondrion has an outer diameter of 1.5 jxm. Arrows point to some of the narrow
tubular connections of cristae to the periphery of the inner membrane. Contours were
drawn using the Sterecon system11 from 2.6nm-thick slices parallel to the plane of the tilt
axes (i.e. the plane of the page), (b-d) Surface renderings of the model in (a) showing
selected cristae with one tubular connection to the inner peripheral membrane (blue), two
TIBS 22 - FEBRUARY 1997
Figure 2
ii'tochondrion in Fi
might represent a single crista with several interconnected compartments, or two cristae
in close apposition at the arrow in (b). The arrow in (a) points to a region that is over ljim
from the nearest opening into the inner peripheral membrane.
discrepancies in specimen preparation
between various investigators. Nonethe
less, Rasmussen's caveat that good con
clusions require good methodology still
applies. There has been considerable pro
gress in optimizing the structural preser
vation of specimens for electron micros
copy, particularly involving cryo-fixation,
low-denaturation embedding and the use
of frozen-hydrated specimens. In fact,
several of these developments have their
roots in the early period of the mitochon
dria controversy. Tomography, coupled
with application of new specimen prep
arative and immunolabelling procedures,
has the potential for providing fresh in
sights into a wide range of questions
about cell structure and organization.
Acknowledgements
'thick' specimens is done on microscopes
operating at 400-1000 kV (vs 100 kV for
conventional instruments). After digitiz
ation and alignment, the tilt-series pro
jection images are used to reconstruct
(by modified back-projection or algebraic
algorithms) the full three-dimensional
density distribution in the object.
The first applications of tomography to
mitochondria6-8 have employed specimens
prepared by conventional fixation and
plastic-embedding procedures. Various
views of a tomographic reconstruction
(tomogram) of a 0.45 u.m slice through
an isolated rat-liver mitochondrion are
presented in Figs 1, 2. The organelle is in
the so-called 'condensed' configuration,
characterized by a contracted matrix
space and expanded intracristal spaces.
The cristal membranes generally have a
dual character, defining large pleiomorphic compartments that are connected to
the periphery of the inner membrane via
one or more tubular regions 30-40 nm in
diameter. Of the 17 cristae contained
entirely in this section, six have one con
nection to the inner membrane periphery,
while the rest have two or more. In addi
tion, several of the internal compartments
are connected to each other by similar
narrow tubes (Figs 1, 2). While some of
the peripheral or internal tubular connect
ors are short, others are hundreds of
nanometers in length. The same basic
organization has been observed for 'ortho
dox' (matrix-expanded) mitochondria,
both isolated6 and in situ (C. A. Mannella
et al., unpublished). The intracristal com
partments are flattened, but are still
connected to the outside and occasion
ally to each other by tubular regions. In
some orthodox mitochondria, the frac-
The three-dimensional images of mito
chondria provided by tomography argue
against the commonly portrayed baffle
model for cristae, at least for conven
tionally prepared rat-liver mitochondria.
The narrow tubular connections that
link the cristae to each other and to the
periphery of the inner membrane suggest
that diffusion between the intracristal
and intermembrane spaces might be re
stricted, and that microcompartmentation within the organelle might be more
complex than in the conventional model.
In fact, there have been earlier indi
cations from transmission3-9 and scan
ning10 electron microscopy that tubular
regions occur in the cristae of various
types of mammalian mitochondria. The
results provided by tomography differ in
a fundamental respect from the earlier
studies. Tomograms provide complete
three-dimensional information about the
density distribution within an object, so it
is not necessary to draw inferences from
partial surface views or thin slices. It is a
relatively straightforward matter to ex
tract from tomograms quantitative infor
mation about compartment volumes and
membrane surface areas within individ
ual mitochondria, as well as the distri
bution and shape of features like matrix
granules (Fig. la) and the contacts be
tween the inner and outer membranes8.
Conclusions and outlook
The results reported here employed
specimens that were fixed and embedded
using basically the same techniques
available 40 years ago. Thus, acceptance
of the baffle model for the mitochondrion
can be attributed to difficulties in ex
trapolating three-dimensional structures
This paper is dedicated to the
memory of VV. D. Bonner, Jr. The study is
supported by NSF grant MCB-9606l'l3,
using the facilities of the Wadsworth
Center's Biological Microscopy and
Image Reconstruction Resource, which
is supported by N1H/NCRR grant RR01219
(Biomedical Resource Technology
Program) and NSF grant BIR-9219043
(Computational Biology Program). We
gratefully acknowledge the contributions
of several members of the resource to
this ongoing project, particularly J. Frank,
M. Radermacher, P. Penczek and A. Leith.
The mitochondrial specimen used in this
study was prepared by S. Konstantinova.
References
1 Rasmussen. N. (1996) Trends Biochem. Sci.
21.319-321
2 Mannella. C. A. (1992) Trends Biochem. Sci.
17. 315-320
3 Daems. W. T. and Wisse. E. (1966)
J. Ultrastruct. Res. 16. 123-140
4 Winslow. J. L. Hollenberg. M. J. and Lea. P. J.
(1991) J. Electron Microsc. Tech. 18. 241-248
5 Frank. J., ed. (1992) Electron Tomography.
Plenum
6 Mannella. C. A. er al. (1994) Microsc. Res.
Tech. 27. 278-283
7 Penczek. P.. Marko, M.. Buttle. K. and Frank. J.
(1995) Ultramicroscopy 60. 393-410
8 Mannella. C. A. er al. (1996) in Proceedings of
Microscopy and Microanalysis. 1996 (Bailey, G. W.
etal.. eds). pp. 966-967, San Francisco Press
9 Brdiczka. D. and Reith. A. (1987) in The
Organization of Cell Metabolism (Welch. G. R.
and Clegg. J. S.. eds). pp. 277-287. Plenum
10 Lea. P. J. and Hollenberg. M. J. (1989) Am. J.
Anat. 184. 245-257
11 Marko. M. and Leith, A. (1996) J. Struct. Biol.
116.93-98
CARMEN A. MANNELLA. MICHAEL
MARKO AND KAROLYN BUTTLE
Biological Microscopy and Image
Reconstruction Resource. Wadsworth Center.
New York State Department of Health. Albany.
NY 12201-0509. USA.