Proc. Natl. Acad. Sci. USAVol. 92, pp. 5251-5257, June 1995

Review

Rethinking cell structure(electron microscopy/cytoskeleton/nuclear matrix/mitosis/cell membranes)

Sheldon PenmanDepartment of Biology, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139

Contributed by Sheldon Penman, January 6, 1995

ABSTRACT Cell structure, emergingfrom behind the veil of conventional elec-tron microscopy, appears far more com-plex than formerly realized. The standardplastic-embedded, ultrathin section canimage only what is on the section surfaceand masks the elaborate networks of thecytoplasm and nucleus. Embedment-freeelectron microscopy gives clear, high-contrast micrographs of cell structurewhen combined with removal ofobscuringmaterial such as soluble proteins. Theresinless ultrathin section is the tech-nique of choice; it is simple and inexpen-sive, and it uses ordinary electron micro-scopes. The resulting pictures reveal aworld of complex cell structure and func-tion. These images necessarily change ourconception of the cytoskeleton, nuclearmatrix, mitosis, and the relation of mem-branes to cytostructure.

The observational instruments of an ex-perimental science often delineate itsmost fundamental ideas. In astronomy,every pivotal innovation in teloscopy lit-erally recasts the cosmos: the simple Gal-ilean telescope's discovery of Jupiter'smoons was an unparalleled liberationfrom static, heliocentric heavens of crys-talline spheres. Today, radio and y-raytelescopes give us an unsettled universe ofneutron stars and black holes. In likemanner, biology owes many of its basicconcepts to a single class of instrumentsthat reveal the invisible. The microscopesof Leuwenhoeck and Hooke first showedcells, and our notions of cell structure haveevolved with improvements in light andelectron optical instruments. Given theprofound importance of microscopy tocell biology thought, it is quite surprisingthat the most powerful magnifying instru-ment of all, the transmission EM, affordsa very truncated view of reality.

Cell biology has long labored under aserious misperception of cell structuresince conventional electron micrographsdo not show most of the cell's architectural

components. This is not an intrinsic tech-nical weakness of electron imaging, how-ever, and recent, very simple EM tech-niques give a far more accurate and de-tailed picture of cell structure. This reviewbriefly explains the problem with conven-

tional EM and how it is solved. Examplesof the resulting EM images show that theydiffer profoundly from the customary pic-ture of internal cell architecture.

Introduction of a more accurate meansof visualizing cell structure is timely sincethe perceived roles of cell architecture are

evolving rapidly. I believe the resultingconceptual transformation will prove ap-posite and, in Kuhn's wonderfully apt, ifnow lamentably cliched, phrase, a "para-digm shift." An emerging school of re-search looks beyond current interests,largely grounded in solution biochemistry,to focus on cell structure, its design andorganization, and, most important, its cru-cial roles in biochemical and developmen-tal functions. As described here, EM can

fully and accurately image cell structure,establish new concepts, and serve as a

powerful adjunct to structure research.Unfortunately, at times the study of cell

structure has been demeaned as of merelyobservational interest and unfavorablycompared with a supposedly more rigor-ous and reductionist biochemistry. Even ifsuch denigration were ever justified, it isincreasingly inappropriate and irrelevant;contemporary studies of cell structure areas disciplined and reductionist as any.Moreover, cell architecture is proving crit-ical to understanding many differentiatedcell functions (1-3), cellular form (4),cellular growth controls, neoplastic trans-formation, and a wide spectrum of generegulation events (5, 6). Such thoroughlyrigorous studies add cellular mechanicsand plumbing to the more usual emphasison chemistry. Only such research can dis-cover the mechanisms of cell collabora-tion in constructing tissue patterns andorganism form, subjects recalcitrant to themore conventional chemical approach.Ultimately, structure research will illumi-nate that most vexing and refractory ofpuzzles-the nature and location of thegenomic instructions dictating the form ofcells, tissue, and, ultimately, organisms.

5251

A major impediment to the study of cellstructure has been the understandableperception that there was simply not verymuch of it. The conventional electronmicrograph of cells shows cells with littlemore of a framework than a marshmallow.The largely featureless interior stronglydiscouraged searching for organizing prin-ciples that, instructed by the genome, de-termine cell, tissue, and organism form.Despite its universality, the conventionalEM image was realized as seriously in-complete. Paradoxically, optical micros-copy could display elaborate, often dra-matic cytoskeletal architecture (7-9) butat relatively low optical resolution. The farmore powerful EM could show the cellinterior only as formless "plasm" withorganelles seemingly scattered about.There appeared no cell framework fordetermining cell shape, cell polarity, andthe obviously nonrandom morphologyand distribution of organelles. The nu-

cleus showed no scaffold or matrix fororganizing enormous lengths of chroma-tin. Also, the chromatin appeared only asdots rather than as the expected thickchromatin solenoids. Mitosis seemed alargely inexplicable, almost magical pro-cess. Nevertheless, despite concerns aboutaccuracy, the embedded thin-section elec-tron micrograph continued to be thenorm. They do show some things ex-tremely well-e.g., membrane profilesand cross-sections of organelle interiors.

Failure to image internal cell structureis not a fault of the EM. The "marshmal-low" image is entirely due to the plasticembedding used in traditional EM sam-ples to allow cutting ultrathin sections.The plastic-embedded section originatedin the earliest days of EM when its short-comings were not obvious and cells wereconsidered extremely fragile. Today, em-

bedding is often an obstacle to under-standing.EM of embedment-free samples reveals

the cell interior comprising an astonishingarray of complex structures. These elabo-rate and dynamic structural networks arepresent throughout the cytoplasm and nu-clei in all cells and, I believe, should be

Abbreviation: DGD, diethylene glycol distea-rate.

The publication costs of this article were defrayedin part by page charge payment. This article musttherefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicatethis fact.

Proc. Natl. Acad. Sci. USA 92 (1995)

considered the true "cytoskeleton." Cy-toskeleton, as a term, is often consideredroughly synonymous with the three fila-ments most often imaged by fluorescencemicroscopy (microfilaments, intermediatefilaments, and microtubules). These fila-ments are but a small part of the newlyrevealed cytostructure (perhaps 10-15%by biochemical measurement). Removingobscuring plastic and soluble proteins al-lows strikingly clear EM of the intricatefilament armatures in the cytoplasm andnucleus. Research into cell structure needno longer struggle with the featurelessmarshmallow cell.

Why the Conventional Embedded EMSection Conceals Most of the Cell'sInterior Structure

A brief reflection on elementary opticalprinciples shows why the usual embeddedsection shows so little of reality. All mi-croscopy requires that a sample differoptically (or electron optically) from itssurroundings. Both light microscopes andEMs form a magnified image by focusingperturbations in the illuminating rays. Theimages in a bright-field microscope resultfrom differential absorption affected byselective stains. Similarly, phase-sensitiveoptical instruments use interference toimage the differential retardation of lightby the sample compared with the sur-rounding milieu. Such interference instru-ments can be quite sensitive to subtleoptical properties of the sample. Althoughbiology textbooks often incorrectly stateotherwise, the EM is primarily a phase-contrast instrument. Although not oftenthought so, it is an extremely sensitiveinstrument that can easily image relativelyelectron lucent objects, such as small cal-iber protein fibers, and needs no heavymetal staining.The extreme sensitivity and resolution of

the EM are completely subverted by em-bedding samples in a plastic resin thatmatches their electron optical properties.The atomic composition of plastics, whichdetermines their electron index of refrac-tion, is almost identical to that of biologicalsamples. To an electron, the sample volumeis now essentially homogeneous and it ex-periences little phase difference betweenthe sample and its surroundings that couldserve to form an EM image. Anyone whohas, by oversight, tried to view an EM sec-tion without poststaining knows how blankthe viewing screen will be.

Producing an image with an embeddedsample requires "staining" the section sothat heavy metal atoms bind to bits ofsample protruding through the sectionsurface. Only these surface atoms formthe familiar EM image; nothing in thesection interior is portrayed. Most or-ganelles and fibers lie below the actualsection surface where they remain invisi-ble. Filaments in particular are difficult to

image since they must lie in the plane ofthe section surface or else appear as amere dot. Apart from the image of stainbound to those dense organelles andmembranes penetrating the section sur-face, the cell interior appears largely form-less. This technique of dense plastic em-bedding might strike us today as curious,but 40 years ago it served to make somesense from images otherwise hopelesslyconfused by the unexpected appearance ofsoluble proteins. The sections showedmembranes very well, which were then ofprimary interest. The price exacted for thisseeming clarity was the masking of whatwe now consider critical internal cellstructures.

Embedment-free EM

About 15 years ago, Keith Porter (10)reintroduced a much older EM technique,

one that viewed whole mounts of cellswithout any embedding material. The re-sulting images showed a dense, complex,polymorphic mesh that he termed the"microtrabecular lattice." The imagesseemed mysterious and could not be re-lated to anything in orthodox micro-graphs; "artifact," used in its pejorativesense, was heard. Repeating Porter's stud-ies was initially difficult since few hadaccess to the necessary 1-MeV EM. Thehigh voltage served to penetrate the densemicrotrabecular lattice. However, a sim-ple technical development has made ahigh-voltage microscope unnecessary; theultrathin, resinless section allows viewingcells with a garden variety 80-kV micro-scope and achieves results equivalent tothose of Porter.

Resinless sections are made by embed-ding the sample in a temporary medium and

FIG. 1. Comparison of transmission electron micrographs of whole HeLa cell sectionsprepared through the conventional Epon technique (A) and the DGD resinless method (B). (A)Mitochondria (arrowheads) are seen "floating" in the cytoplasm (Cy) of a HeLa cell that wasembedded in Epon, sectioned, and conventionally poststained. N, nucleus. (B) Resinless sectionof a fixed HeLa cell that was initially embedded in DGD shows mitochondria (arrowheads)enmeshed in the cytoplasmic microtrabecular lattice that is revealed through this technique. (Bar= 100 nm.)

5252 Review: Penman

Proc. Natl. Acad. Sci. USA 92 (1995) 5253

cutting sections on the ultramicrotome. Theresulting ultrathin sections are placed on agrid, extracted with solvent to remove thetemporary embedding medium, and thendried through the CO2 critical point. Glu-taraldehyde-fixed samples, even when ex-tensively extracted, are remarkably toughand appear to maintain their form evenwhen undergoing considerable manipula-tion. The resinless section technique wasfirst developed by Wolosewick (11) usingpolyethylene glycol (PEG) as the provi-sional embedment. We found using PEGtechnically difficult and developed an alter-native technology based on using diethyleneglycol distearate (DGD) for interim embed-ding (12). Most find the DGD cuts far moreeasily than PEG and, since it is not watersoluble, it can be floated in the usual water-filled microtome trough. Recently, we haveadopted a modified form ofDGD (Antibed;the kind gift of the EMCorp, Chestnut Hill,MA) that avoids some problems with brit-tleness.

Coping with the Newly Visible SolubleProteins

Creating embedment-free sections is butthe first step in unveiling cytoarchitecture.Now the soluble proteins, convenientlyhidden within embedded sections for ahalf century, are visible. The contrast withconventional micrographs is striking butnot very informative. Fig. 1A shows HeLacell cytoplasm in a familiar, Epon-embed-ded, poststained section and Fig. 1B shows

an unstained, resinless section of the samecell material. The ghostly trabecular net-work in Fig. 1B, present but invisible inFig. 1A, consists of the soluble proteinscross-linked by the fixative into a stablestructure. The soluble proteins are most ofthe cell mass and obscure organelles, suchas mitochondria, which can be only dimlyglimpsed. Their presence surely discour-aged the early microscopists since, oncevisible, they effectively hid everythingelse. The soluble protein problem verylikely led to adoption of the embeddedsection, because it at least showed mem-brane outlines.With our increased knowledge of cell

structure, we can now do much better thanenshroud the soluble proteins in embed-ding plastic, thus covering up most of thecell's interior. Instead, we can selectivelyand completely remove the soluble pro-teins by extraction with nonionic deter-gent. The mechanism of the detergentextraction is often misunderstood; non-ionic (in contrast to ionic) detergents donot affect proteins and so they do not"extract" in the usual sense of that term.Instead, the detergent simply dissolvesmembrane phospholipids, destroying theplasma membrane barrier and allowingsoluble proteins to passively diffuse away.If the extraction buffer salts and pH areclose to the intact cell's original interiormilieu, the procedure leaves cytoskeletalstructures intact and with nearly nativemorphology.

The detergent-extracted cytoskeleton,free of soluble proteins, is potentially idealfor both EM and biochemical studies.However, EM using conventional embed-ded sections is particularly poor for view-ing the detergent-extracted cytoskeleton.The three-dimensional filament networksof the cytoskeleton are precisely the typeof material that images most poorly inconventional sections. Fig. 2A shows thefeeble picture of detergent-prepared cy-toskeletal structures afforded by an Epon-embedded section. Portions of both thecytoplasmic region (upper right) and ofthe nucleus (lower left) are visible. Ribo-somes are the most visible component inthe cytoplasm due to the avidity of theirRNA for poststaining uranyl ions. Theconspicuous, nonrandom clusters of thepolyribosomes reflect their binding to thecytoskeleton that, of course, is not visiblehere. Some ribosomes are associated withwhat may be remnants of the rough en-doplasmic reticulum. The image of thenucleus is also uninformative, showingchromatin, mostly clumped, and little else.

Fig. 2B shows the striking, high-con-trast cytoskeleton image afforded by anunstained, resinless section. The microtra-becular lattice of soluble proteins is, ofcourse, gone, leaving an unobstructedview of the cytoskeleton. The profusion ofcytoskeletal structures includes micro-and intermediate filaments, polyribo-somes attached to the cytoskeleton, andmyriad small structures not yet identified.The nuclear space shows thick, knobby

FIG. 2. HeLa cell cytoskeletal preparations as seen by conventional (A) and unembedded (B) EM. The resinless section shows, in contrast to

the conventional one, the interconnected cytoskeletal filaments that span throughout the cytoplasm (Cy) and are anchored on the nuclear lamina.(Bar = 100 nm.)

Review: Penman

Proc. Natl. Acad. Sci. USA 92 (1995)

agglomerates of chromatin and thin fibers,previously unseen, which are nuclear ma-trix filaments.The simple and effective nonionic deter-

gent extraction reveals a highly intricatecytostructure that, unlike the structurelesscell, appears suitable for carrying out com-plex architectural instructions. The resinlesssections are three-dimensional objects andyield true stereoscopic images when viewedwith a tilting stage. In three dimensions, thecell networks resemble "tensegrity" struc-tures-i.e., built of elements that experiencetension and compression but no bending.However, the detergent-extraction proce-dure does exact a serious price: solubiliza-tion of all cell membranes. We will see laterthat saponin extraction that selectively re-moves cholesterol from membranes, leavingmuch of the phospholipid, preserves a sig-nificant portion of the internal cell mem-branes.

The Core Filaments of the NuclearMatrix

Selective extraction can do more thansimply remove soluble proteins. With suit-able modification, extraction can be usedto peel away cell structures, enabling us topeer more deeply into the cell interior.This has been particularly useful in resolv-ing the controversies surrounding the ex-istence of the nuclear matrix. Few biolog-ical subjects in recent times have arousedsuch passionate dispute. Exactly why theidea of nonchromatin nuclear structureengendered so much contention will prob-ably remain a puzzle. The existence of anorganizing scaffold would surely seemcompelled by the complexities of nuclearorganization and behavior that hardly ap-

pear due to soluble components. Verylikely the complete invisibility of the pur-ported nuclear matrix in contemporarymicrographs, even after chromatin deple-tion, helped incite apprehensions as to itsreality (13).The nuclear matrix is hidden under

overwhelming amounts of electron-densechromatin that must be removed beforemicroscopy or biochemical analysis. Be-rezney and Coffey (14, 15) first detectedthe nuclear matrix biochemically as thenonchromatin nuclear material remainingafter vigorous extraction of chromatinwith DNase and high salt. This and similarpreparations served in much importantresearch and established the matrix as aparticipant in important nuclear events(16-23). The preparations were particu-larly useful for the early specific hormonereceptor studies (18, 25, 26). However, thenuclear matrix remained invisible in theconventional electron micrographs, pro-voking further skepticism concerningwhether the matrix really existed. Ofcourse, invisibility was also a serious ob-stacle to further research.The embedded EM section cannot im-

age the nuclear matrix for the same reasonthat it cannot show the cytoskeleton; res-inless sections are far better for this pur-pose. The need to remove the overwhelm-ing amounts of electron-opaque chroma-tin is an added complication. Chromatin isbound tenaciously to the nuclear matrixand removing it requires relatively harshprocedures. Effecting an essentially com-plete elimination of chromatin while min-imizing distortion of the matrix has been along-standing technical goal. Fig. 3 showsresinless section images of our most recent

preparation that we believe is the mostnative produced to date.

Fig. 3A shows the nuclear matrix filamentnetwork while Fig. 3B is a higher-magnifi-cation view. The matrix consists of a dense,anastomosing network of 10- to 13-nm fila-ments. We have suggested these be called"core filaments" since they are the inner-most structures of the cell. Although theysuperficially resemble intermediate fila-ments, they share no epitopes with eitherlamins or intermediate filaments and soappear unrelated. Enmeshed in the fila-ments are many dense bodies, some ofwhichare rich in RNA splicing components.The identification, localization, and

characterization of nuclear matrix compo-nents are just beginning. An importantimpetus comes from finding nuclear ma-trix proteins that are cell-type specific(27-29) and others that are markers ofmalignancy (30-32). Direct measure-ments of transcription show nuclear ma-trix participation in its regulation (5).These results suggest a role for matrixproteins in determining gene expression.Many antibodies have been created that

specifically immunostain nuclear matrixproteins. Many, although not all, of theseantibodies give a speckled immunofluo-rescence staining pattern as shown in Fig.4A. Fig. 4B shows that immunogold stain-ing with the same antibody decorates thedense bodies enmeshed in the core fila-ment network. The "speckles" clearly cor-respond to some but not all dense bodies.Several anti-nuclear matrix antibodies,such as B1C8 shown here, bind to proteinsin spliceosomes, suggesting that the densebodies are related to RNA splicing.

FIG. 3. Embedment-free ultrastructure of nuclear matrix core filaments. Micrographs show at low (A) and high (B) magnification theinterrelations between the nuclear matrix components. The anastomosing network of the core filaments (CF) enmeshes electron-dense aggregatesthat form the dense bodies (DB) and connects with the nuclear lamina (L). (A, bar = 500 nm; B, bar = 200 nm.)

5254 Review: Penman

Proc. Natl. Acad. Sci. USA 92 (1995) 5255

I

4.

FIG. 4. Immunostaining of core filament preparations. (A) Immunofluorescent image of Caski cell nuclei stained with B4A11 monoclonalantibody. Speckled staining pattern is further revealed by immunogold resinless EM of the core filament nuclear matrix (B). Dense bodies (DB)are decorated with 10-nm immunogold and are enmeshed in core filament (CF) matrix. (Bar = 100 nm.)

Cell Structure at Mitosis

The rearrangements during cell divisionare surely the most profound in a cell'srepertoire. While the biochemistry uniqueto mitosis has been an active and fruitfularea of research, little has been adducedconcerning cell architecture at mitosis.Apart from studies of microtubule func-tioning, most important questions aboutmitotic cytostructure remain. These puz-zles include the mysterious agents effect-ing chromosome movement before thespindle forms, the machinery directingspindle fibers precisely to the kineto-chores, the mechanisms for nuclear disso-lution and reformation, etc.

Conventional embedded-section EMcan only hint at the profound rearrange-ments of the cytoskeleton and nuclearmatrix at mitosis. When the nuclear lam-ina disappears at late prophase, the cy-toskeleton and nuclear matrix meld.into anew, cell-wide structure until now mostlyunseen. At best, the embedded sectionsuggests only a few large components ofthe mitotic apparatus such as the chromo-somes, short stretches of spindle fibers,and the occasional centriole.The resinless section displays a new world

of mitotic cell cytostructure. Only the solu-ble proteins need be extracted; there is nolonger a need to extract chromatin since ithas been compacted into chromosomes.Space limitations prevent a comprehensivepresentation here but the metaphase platein Fig. 5 shows the unique information thetechnique affords. Compared with thevague, shadowy suggestions of conventional

micrographs, the filament and spindle fibernetworks emerge here as sharp, well de-fined, and far more complex than previouslysuspected.A complex filament network enmeshes

the metaphase chromosomes in Fig. 5, fila-ments that are mostly invisible in conven-tional electron micrographs. The dense fil-ament bundles extending horizontally fromthe chromosomes are the spindle fibers that,

when lying in a thick-section plane as here,appear in their entirety. Especially arrestingare the many thin filaments interconnectingchromosomes and coupling spindle fibers.These may originate in the core filaments ofthe interphase cell but firm identificationawaits the development of core filament-specific antibodies. The major import ofthese pictures is that, rather than floating inan amorphous sea, the major elements of

-t. - |-J.- .'

FIG. 5. Resinless electron micrograph of a mitotic HeLa cell. Chromosomes (Chr) are shown"suspended" on spindle fibers (arrowheads), which are interconnected through numerous thinfilaments (double arrowheads). (Bar = 500 nm.)

Review: Penman

Proc. Natl. Acad. Sci. USA 92 (1995)

the mitotic cell are part of a complex skel-etal network that underlies the mechanicalevents of mitosis.

Membranes in Resinless Sections:Extraction with Saponin

The absence of cell membranes is a short-coming of the preparations for resinlesssections shown so far. This is not an in-trinsic weakness of the technique but re-sults from the detergent extraction of allphospholipids to remove soluble proteins.Saponin, substituted for Triton X-100,punctures the plasma membrane by pref-erentially extracting cholesterol (24). Thisweak detergent leaves much of the plasmamembrane phospholipid bilayer intactwhile allowing the soluble proteins to dif-fuse away from the cell.

Fig. 6 shows mitochondria and othermembrane-bounded cellular organellesenmeshed in the cytoskeleton in saponin-extracted cells. The mitochondrial cristaeappear as dark zebra stripes inside theorganelle. Cytoskeletal filaments anasto-mose with the membrane surfaces of theorganelles. This picture suggests that theusual view of membrane-enveloped or-ganelles as structurally autonomous enti-ties is incomplete. The resinless sectionmicrographs intimate that cytoskeletonfilaments may serve as organelle frame-works enveloped by the lipid bilayers ascanvas covers the framework of a tent. Inthis view, the cytoskeleton plays a more

central role in organelle structure than hasbeen appreciated. Such a role could ex-plain the often complex morphologies ofmembranes and organelles.

Rethinking Cell Structure

Resinless microscopy affords images thatlead to some radical conclusions. First,and probably most important, the resinlesssection can be enormously powerful butonly if combined with removing obscuringmaterial. No single method is completelysatisfactory for all purposes. The solubleproteins, which confounded early embed-ment-free microscopy, are most easily ex-tracted with nonionic detergent but at thecost of losing all internal membranes. Ex-traction with saponin effects the sameremoval while leaving much but not all ofthe internal membrane structures. Thesaponin preparation offers unique possi-bilities for deeper study of membranecytoskeleton interactions. Removing te-naciously bound chromatin while preserv-ing a nearly native nuclear matrix has beenfar more challenging. The core filamentnuclear matrix preparations have been ourmost successful effort to date.A second important point is that cell

structures are remarkably tough. They canretain their detailed morphology throughextraction, embedding and deembedding,sectioning, and critical point drying. Isuspect one reason for originally adoptingembedded sections was a belief in the

ephemerality of cells once removed fromtissue. The concern was doubtless justifiedin the early days before autolysis wasappreciated and controlled.

Finally, the resinless section techniquesare especially simple; most microscopistsfind them far less difficult than preparingthe conventional ultrathin Epon sections.They should not be a technical obstacle tothose interested in adopting these proce-dures.The images revealed by the techniques

described here compel serious reexamina-tion ofwhat we believe a cell to be. The oldview, which might be oversimplified as a"fluid" model, is of moderately viscous"plasm" bounded and contained by aplasma membrane composed of a lipidbilayer into which proteins are inserted.The new idea is that the cytoskeletonserves as a solid framework or armaturewith an outer boundary of a protein sheetwith lipids inserted. This framework isanchored on a second protein sheet orlamina that is the surface of the nucleusand bounds a separate framework, thenuclear matrix. Both nuclear and cytoplas-mic frameworks probably conform to thedefinition of tensegrity (33) since mostbiological fibers can be strong in tensionand even in compression but very weak inbending.Many specific questions of cytoskeleton

and nuclear matrix function can now beexamined in light of our deeper insightinto cell structure. In addition, more pro-

FIG. 6. Ultrastructure of a smooth muscle cell from rat aorta after saponin extraction. Whole-mount electron micrograph shows mitochondria(M) intimately connected with the cytoskeletal framework. Arrowheads point to mitochondrial cristae. (Bar = 200 nm.)

5256 Review: Penman

Proc. Natl. Acad. Sci. USA 92 (1995) 5257

found intellectual questions are posed bythe intricate structural networks now seenin cells-e.g., where are the complex in-structions organizing the elaborate net-works that underlie architecture? Likeastronomical dark matter, the coding forarchitectural information seems invisibleand largely ignored although its effects areeverywhere evident. The subject could bedisregarded when cells appeared lackingmuch structure. Once we add a wholly newspectrum of architectural complexity togenome encoding, we can no longer ig-nore the >90% of the vertebrate genomethat does not directly instruct protein syn-thesis. This "extra" DNA has long beenvexatious and seemingly illogical in prop-erties. Despite its often intense transcrip-tional activity, it is frequently dismissed as"junk," a designation that may be intel-lectually soothing but hardly helpful sci-entifically. Molecular cell biology canscarcely aspire to conceptual complete-ness while ignoring the vast majority oforganismic DNA.Form and structure are not natural sub-

jects for biochemistry that, in the macro-scopic world, deals with scalar quanti-ties-i.e., amounts, rates, etc. Building thecomplex designs glimpsed in any anatomyor physiology text requires, at the veryleast, instructions that are vectorial-i.e.,that specify direction and place. Theseinstructions are encoded somewhere-itseems very likely that they reside in theheavily transcribed but "non-protein cod-ing" DNA. Building staggeringly complexorgans-e.g., brains or kidneys-by sim-ply specifying the constituent proteincomponents (as suggested by the moreextreme formulations of molecular biol-ogy that genes are simply proteins) isunlikely. Such a strategy would be tanta-mount to trying to specify a bridge or anedifice by merely giving a list of parts.Indeed, Gray's Anatomy, seen with anengineer's eye, suggests that the complex-ity of the instruction sets for mammalianmorphology require large regions of the

genome: very likely much or most of thecurrently ignored, non-protein coding,90% (or more) of the genome. I suspectthat future cell scientists will marvel at thedensity and ingenuity of genome instruc-tions for structure while wondering howwe could overlook them for so long.

EM was made possible by the remarkabletechnical skills of Gabrielle Krockmalnic. Kather-ine Wan conducted much of the difficult bio-chemical measurements. Jeffery Nickerson con-ducted several of the important studies but, mostimportantly, has been a constant source of stim-ulation and wise reflection. It is not possible tocredit all who have contributed to these thoughtsbut among them are Alice Fulton, Gary Stein,Laura Manelides, Don Coffee, and many others.This work was supported by grants from theNational Institutes of Health (CA45480-08,CA67628-28, and AR42262-01).

1. Aggeler, J., Ward, J., Blackie, L. M., Bar-cellos-Hoff, M. H., Streuli, C. H. & Bis-sell, M. J. (1991) J. Cell Sci. 99, 407-417.

2. Folkman, J. & Moscona, A. (1978) Nature(London) 273, 345-349.

3. Glowacki, J., Trepman, E. & Folkman, J.(1983) Proc. Soc. Exp. Bio. Med. 172,93-98.

4. Ingber, D. E., Dike, L., Hansen, L., Karp,S., Liley, H., Maniotis, A., McNamee, H.,Mooney, D., Plopper, G., Sims, J. &Wang, N. (1994) Int. Rev. Cytol. 150, 173-224.

5. Bidwell, J. P., Van-Wijnen, A. J., Fey,E. G., Dworetzky, S., Penman, S., Stein,J. L., Lian, J. B. & Stein, G. S. (1993) Proc.Natl. Acad. Sci. USA 90, 3162-3166.

6. Dworetzky, S. I., Wright, K. L., Fey, E. G.,Penman, S., Lian, J. B., Stein, J. L. &Stein, G. S. (1992) Proc. Natl. Acad. Sci.USA 89, 4178-4182.

7. Lazarides, E. & Weber, K. (1974) Proc.Natl. Acad. Sci. USA 71, 2268-2272.

8. Lazarides, E. & Burridge, K. (1975) Cell 6,89-98.

9. Goldman, R. D., Yerna, M. J. & Schloss,J. A. (1976) J. Supramol. Struct. 5, 155-183.

10. Porter, K. R. (1984)J. Cell Biol. 99, Suppl.,3s-12s.

11. Wolosewick, J. (1980) J. Cell Biol. 86,675-681.

12. Capco, D. G., Krockmalnic, G. & Pen-man, S. (1984) J. Cell Biol. 98, 1878-1885.

13. Herman, R., Weymouth, L. & Penman, S.(1978) J. Cell Biol. 78, 663-674.

14. Berezney, R. & Coffey, D. S. (1974) Bio-chem. Biophys. Res. Commun. 60, 1410-1417.

15. Berezney, R. & Coffey, D. S. (1977) J. CellBiol. 73, 616-637.

16. Berezney, R. & Coffey, D. S. (1975) Sci-ence 189, 291-292.

17. Ben-Ze'ev, A. & Aloni, Y. (1983) Virology125, 475-479.

18. Barrack, E. R. & Coffey, D. S. (1980) J.Biol. Chem. 255, 7265-7275.

19. Carter, K. C., Bowman, D., Carrington,W., Fogarty, K., McNeil, J. A., Fay, F. S. &Lawrence, J. B. (1993) Science 259, 1330-1335.

20. Ciejek, E. M., Tsai, M.-J. & O'Malley, B. W.(1983) Nature (London) 306, 607-609.

21. Jackson, D. A. & Cook, P. R. (1985)EMBO J. 4, 919-925.

22. Jarman, A. P. & Higgs, D. R. (1988)EMBO J. 7, 3337-3344.

23. Pardoll, D. M., Vogelstein, B. & Coffey,D. S. (1980) Cell 19, 527-536.

24. Lin, A., Krockmalnic, G. & Penman, S.(1990) Proc. Natl. Acad. Sci. USA 87,8565-8569.

25. Buttyan, R., Olsson, C. A., Sheard, B. &Kallos, J. (1983) J. Biol. Chem. 258,14366-14370.

26. Kirsch, T. M., Miller-Diener, A. &Litwack, G. (1986) Biochem. Biophys. Res.Commun. 137, 640-648.

27. Fey, E. G. & Penman, S. (1988) Proc. Natl.Acad. Sci. USA 85, 121-125.

28. Getzenberg, R. H. & Coffey, D. S. (1990)Mol. Endocrinol. 4, 1336-1342.

29. Stuurman, N., Van-Driel, R., De-Jong, L.,Meijne, A. M. & Van-Renswoude, J.(1989) Exp. Cell Res. 180, 460-466.

30. Getzenberg, R. H. & Coffey, D. S. (1991)Cancer Res. 51, 6514-6520.

31. Partin, A. W., Getzenberg, R. H., Car-Michael, M. J., Vindivich, D., Yoo, J.,Epstein, J. I. & Coffey, D. S. (1993) Can-cer Res. 53, 744-746.

32. Khanuja, P. S., Lehr, J. E., Soule, H. D.,Gehani, S. K., Noto, A. C., Choudhury, S.,Chen, R. & Pienta, K. J. (1993) CancerRes. 53, 3394-3398.

33. Ingber, D. E. (1993) J. Cell Sci. 104, 613-627.

Review: Penman