The conduction system in the human heart at midgestation-immunohistochemical demonstration of the intermediate filament protein skeletin

Summary The intermediate filaments have during recent years increasingly attracted attention and new information on the distribution of the subunits of the filaments has become available by the use of specific antibodies. In the present study human fetal hearts at midgestation were studied with immunofluorescence microscopy for a demonstration of the intermediate filament subunit skeletin. In the ordinary ventricular and atrial myocytes the fluorescence was moderate, mainly concentrated to the Z disk levels. The proximal parts of the conduction system also showed a moderate fluorescence, while the fluorescence was intense in the cells in the peripheral parts. The shift from moderate to intense fluorescence occurred in the very proximal part of the left bundle branch and in the distal part of the intramural right bundle branch. We conclude that the conduction cells attain Purkinje fibre-like characteristics at these levels, as evidenced by the content of skeletin. Furthermore, the human fetal Purkinje fibres are more easily distinguished with the technique used herein than with conventional histological techniques. The observed differences in content of skeletin are discussed in terms of the cytoskeletal function of intermediate filaments and the embryology of the conduction system.     

Immunological relationship between different types of bovine intermediate filaments

In order to examine the relationship between the intermediate filaments from Purkinje fibres of the cow heart conduction system and five proposed subclasses of mammalian intermediate filaments, the gel electrophoresis-derived enzyme-linked immunosorbent assay (GEDELISA) has been used to examine the specificity and crossreactivity of our antibodies against the Purkinje fibre intermediate filament protein, skeletin. Bovine tissues known to contain intermediate filaments of the five main subclasses were examined with antiskeletin and with preimmune serum and the specific antiserum absorbed with pure skeletin as controls. The antibodies raised against Purkinje fibre skeletin reacted with all three polypeptides of the "neurofilament triplet", with glial fibrillary acidic protein (GFAP), with smooth muscle desmin and also slightly with some prekeratin subunits and with endothelial vimentin. From studies with monoclonal antibodies and amino acid sequencing, certain regions of all intermediate filaments are suggested to be structurally related. Here we show that Purkinje fibre skeletin seems to share antigenic determinants with the proposed five main classes of intermediate filaments. Our antibody is the first carefully controlled experimentally induced antibody having such properties. This might be due to the special attributes of the intermediate filament system in Purkinje fibres, which themselves have unique properties.     

Intermediate (skeletin) filaments in heart Purkinje fibers. A correlative morphological and biochemical identification with evidence of a cytoskeletal function

Cow Purkinje fibers contain a population of free cytoplasmic filaments which consistently differ in ultrastructural appearance from actin and myosin filaments, irrespective of preparation technique. The fixation and staining techniques, however, influenced the filament diameter, which was found to be 7.4--9.5 nm for filaments in plastic-embedded material, and 7.0 nm in cryo-sectioned material, thus intermediate as compared to actin and myosin filaments. Cross-sectional profiles suggested that the intermediate-sized filaments are composed of four subfilaments. To provide a basis for further biochemical investigations on the filaments, extraction procedures were carried out to remove other cell organelles. Electron microscopy showed that undulating bundles of intermediate filaments converging towards desmosomes still remained, after the extractions, together with Z-disk material. In spite of the extensive extraction, the shape of the individual cells and the assemblies of cell bundles remained intact. This confirms that the intermediate filaments of cow Purkinje fibers together with desmosomes do in fact have a cytoskeletal function. On account of (a) the cytoskeletal function of the filaments, (b) the similarities to the smooth muscle "100-A filament" protein subunit skeletin, and (c) the inadequate and confusing existing terminology, we suggest that the filaments be named "skeletin filaments."     

Immunoelectron microscopic studies of desmin (skeletin) localization and intermediate filament organization in chicken skeletal muscle

We studied the localization of desmin (skeletin), the major subunit of muscle-type intermediate filaments, by high resolution immunoelectron microscopy in adult chicken skeletal muscle. Immunoferritin labeling of ultrathin frozen sections of intact fixed sartorius muscle showed the presence of desmin between adjacent Z-bands and as strands peripheral to Z-bands, forming apparent connections between the Z-bands with adjacent sarcolemma, mitochondria, and nuclei. We observed no desmin labeling, however, in the vicinity of the T-tubules. In addition, intermediate filaments were morphologically discernible at the level of the Z-bands in plastic sections of glycerol-extracted muscle that had been infused with unlabeled antidesmin antibodies. Our results indicate that the desmin present in adult skeletal muscle, that had previously been detected by immunofluorescence light microscopy, is largely if not entirely in the form of intermediate filaments. The results provide evidence that these filaments serve to interconnect myofibrils at the level of their Z-bands, and to connect Z-bands with other specific structures and organelles in the myotube, but not with the T-tubule system.     

Specific attachment of desmin filaments to desmosomal plaques in cardiac myocytes

Intercellular junctions which are similar in ultrastructure and protein composition to typical desmosomes have so far only been found in epithelial cells and in heart tissue, specifically in the intercalated disks of cardiac myocytes and at cell boundaries between Purkinje fiber cells. In epithelial cells the cytoplasmic side of desmosomes, the 'desmosomal plaque', represents a specific attachment structure for the anchorage of intermediate filaments (IF) of the cytokeratin type. Cardiac myocytes do not contain cytokeratin filaments. In primary cultures of rat cardiac myocytes, we have examined by immunofluorescence and electron microscopy, using single and double label techniques, whether other types of IF are attached to the desmosomal plaques of the heart. Antibodies to desmoplakin, the major protein of the desmosomal plaque, have been used to label specifically the desmosomal plaques. It is shown that the desmoplakin-containing structures are often associated with IF stained by antibodies to desmin, i.e., the characteristic type of IF present in these cells. Like cytokeratin filaments in epithelial cells, desmin filaments attach laterally to the desmosomal plaque. They also remain attached to these plaques after endocytotic internalization of desmosomal domains by treatment of the cells with EGTA. These desmin filaments do not appear to attach to junctions of the fascia adherens type and to nexuses (gap junctions). These observations show that anchorage at desmosomal plaques is not restricted to IF of the cytokeratin type and that IF composed of either cytokeratin or desmin, specifically attach, in a lateral fashion, to desmoplakin-containing regions of the plasma membrane. We conclude that special domains exist in these two IF proteins that are involved in binding to the desmosomal plaque.     

Associations between beta-tubulin and mitochondria in adult isolated heart myocytes as shown by immunofluorescence and immunoelectron microscopy

We have investigated the associations between beta-tubulin and mitochondria in freshly isolated cardiac myocytes from the rat. Beta-tubulin was identified by using monoclonal antibodies for immunofluorescence and high resolution immunogold electron microscopy. In addition, conventional transmission and scanning electron microscopic studies were performed. After chemical stabilization in a formaldehyde solution, the myocytes were shock-frozen at -150 degrees C, cryosectioned at -70 degrees C and subsequently processed for immunohistochemical and immunocytochemical microscopy. A characteristic of the rod shaped myocytes is the presence of a dense network of microtubules in the cytoplasm displaying a pattern of strong anti-beta-tubulin reaction. The complexity of this network however varies considerably among the myocytes reflecting microtubule dynamic instability. Further, our findings demonstrate that the beta-tubulin label in rod cells is confined to the perinuclear and interfibrillar spaces and, therefore, is largely colocalized with the cytoplasmic organelles. In myocytes undergoing severe contracture the distribution of beta-tubulin is entirely restricted to the outer mitochondrial-containing domain. This implies that, in a cell model with marked segregation of the contractile filaments and organelles, mitochondria are codistributed with microtubules in the total absence of desmin intermediate filaments. Moreover, our immunogold preparations demonstrate anti-beta-tubulin labelling in the outer mitochondrial membrane as well as of fibres in close apposition to this membrane. These results indicate the presence of a specific beta-tubulin binding to the outer mitochondrial membrane that probably also involves microtubule based translocators and/or MAPs.     

Specificity of desmin to avian and mammalian muscle cells.

The extent of invariance and heterogeneity in desmin, the major component of the muscle form of 100 Å filaments, has been investigated in avian and mammalian muscle and nonmuscle cells with two-dimensional gel electrophoresis and indirect immunofluorescence. Desmin from chick, duck and quail, smooth, skeletal and cardiac muscle cells is resolved into two isoelectric variants, α and β, with each possessing the same charge and electrophoretic mobility in all three avian species irrespective of muscle type. Guinea pig and rat muscle desmin resolves into only one variant; it also possesses the same charge and electrophoretic mobility in the two mammalian species, but it is more acidic and slower in electrophoretic mobility than the two avian variants.In immunofluorescence, desmin is localized together with α-actinin along myofibril Z lines. Antibodies to chick smooth muscle desmin, prepared against the protein purified by preparative SDS gel electrophoresis prior to immunization, cross-react with myofibril Z lines in all three avian species. These antibodies do not cross-react with either rat or guinea pig myofibril Z lines. Similarly, they do not cross-react with avian or mammalian nonmuscle cells grown in tissue culture and known to contain cytoplasmic 100 Å filaments.These results demonstrate that desmin is highly conserved within avian muscle cells and within mammalian muscle cells. It is, however, both biochemically and immunologically distinguishable between avian and mammalian muscle cells, and between muscle and nonmuscle cells. We conclude that there are biochemically and immunologically specific forms of desmin for avian and mammalian muscle cells. Furthermore, within a particular vertebrate species, there are at least two separate classes of 100 Å filaments: the muscle class whose major component is desmin, and the nonmuscle class whose major component is distinct from desmin. Taking into consideration the immunological specificity reported by other laboratories for the 100 Å filaments in glial cells, for neurofilaments and for the epidermal 80 Å keratin filaments, we propose that a given vertebrate species contains at least four major distinguishable classes of 100 Å filaments: muscle 100 Å filaments (desmin filaments), glial filaments, neurofilaments and epidermal keratin filaments.     

Isolation and partial characterization of intermediate filament protein (skeletin) from cow heart Purkinje fibres.

The intermediate filament protein skeletin from cow heart Purkinje fibres was purified to homogeneity by a selective extraction procedure and gel chromatography in the presence of sodium dodecyl sulphate. Monospecific antibodies were obtained by immunisation of rabbits with the sodium dodecyl sulphate-skeletin complex, and rocket electrophoresis made it possible to quantify the concentration of protein. The skeletin monomer has a molecular weight of 55 000. Amino acid analysis revealed that skeletin has a high content of glutamic acid, aspartic acid, alanine and leucine, together constituting more than 50% of the molecule. The isoelectric point is determined as 6.35. Skeletin is insoluble at pH 4--6 in the absence of detergent and shows increasing solubility at higher and lower pH. The biochemical characteristics are discussed in relation to the cytoskeletal function of the filaments. Comparison with intermediate-sized filament protein of other tissues show certain important similarities suggesting that the filaments may share a common evolutionary ancestry.     

The interaction of tropomodulin with tropomyosin stabilizes thin filaments in cardiac myocytes

Actin (thin) filament length regulation and stability are essential for striated muscle function. To determine the role of the actin filament pointed end capping protein, tropomodulin1 (Tmod1), with tropomyosin, we generated monoclonal antibodies (mAb17 and mAb8) against Tmod1 that specifically disrupted its interaction with tropomyosin in vitro. Microinjection of mAb17 or mAb8 into chick cardiac myocytes caused a dramatic loss of the thin filaments, as revealed by immunofluorescence deconvolution microscopy. Real-time imaging of live myocytes expressing green fluorescent protein-alpha-tropomyosin and microinjected with mAb17 revealed that the thin filaments depolymerized from their pointed ends. In a thin filament reconstitution assay, stabilization of the filaments before the addition of mAb17 prevented the loss of thin filaments. These studies indicate that the interaction of Tmod1 with tropomyosin is critical for thin filament stability. These data, together with previous studies, indicate that Tmod1 is a multifunctional protein: its actin filament capping activity prevents thin filament elongation, whereas its interaction with tropomyosin prevents thin filament depolymerization.     

Spatial Distribution of Proteins Specific for Desmosomes and Adhaerens Junctions in Epithelial Cells Demonstrated by Double Immunofluorescence Microscopy

Abstract. The spatial relationships between the protein constituents of two junctional structures, adhaerens junctions and desmosomes, were determined by double immunofluorescence microscopy using marker proteins specific for these structures. Adhaerens junctions were visualized by immunofluorescent labeling for the membrane-associated protein vinculin and by their association with actin filaments. Desmosomal components were identified by labeling with anti-bodies to a group of minor desmosomal plaque proteins (DP1 antigens) and their association with filaments stained by cytokeratin antibodies. Double immunofluorescence microscopy of these components was performed in several tissues and cultured cells, including intact intestine, dissociated intestinal cells, and two morphologically different types of epithelial cells, cultured bovine kidney (MDBK), and mammary gland (BMGE) epithelial cells. This allowed the direct demonstration that each filament system is associated exclusively with its specific membrane-bound junctional protein. Vinculin and DP1-protein were found in distinct sites in the subapical intercellular junctional complex of intestinal epithelium and MDBK cells. Cell-substrate focal contacts contained vinculin and actin and showed no apparent relationships to the tonofilament system whereas intercellular contacts of BMGE cells were characterized by positive staining for DP1-protein and associated cytokeratin filaments. Immunolabeling of the cultured cells at different intervals after plating for the cytoskeletal elements and their membrane anchorage proteins was used to determine the temporal sequence of their organization. We propose that this approach may be used for the molecular definition and identification of cellular contacts and junctions as well as for studies of junction topology, dynamics of junction-cytoskeleton interactions, and junction biogenesis.     

Disruption of the keratin filament network during epithelial cell division

The behaviour of keratin filaments during cell division was examined in a wide range of epithelial lines from several species. Almost half of them show keratin disruption as described previously: by immunofluorescence, filaments are replaced during mitosis by a 'speckled' pattern of discrete cytoplasmic dots. In the electron microscope these ' speckles ' are seen as granules around the cell periphery, just below the actin cortical mesh, with no detectable 10 nm filament structure inside them and no keratin filament bundles in the rest of the cytoplasm. A time course of the filament reorganization was constructed from double immunofluorescence data; filaments are disrupted in prophase, and the filament network is intact again by cytokinesis. The phenomenon is restricted to cells rich in keratin filaments, such as keratinocytes; it is unrelated to the co-existence of vimentin in many of these cells, and vimentin is generally maintained as filaments while the keratin is restructured. Some resistance to the effect may be conferred by an extended cycle time. Filament reorganization takes place within minutes, so that a reversible mechanism seems more likely than one involving de novo protein synthesis, at this metabolically quiet stage of the cell cycle.     

Morphological integrity of single adult cardiac myocytes isolated by collagenase treatment: immunolocalization of tubulin, microtubule-associated proteins 1 and 2, plectin, vimentin, and vinculin

Single cardiac myocytes were isolated from hearts of 9 to 12-week-old rats by means of collagenase (100 U/ml). After assessment of their functional integrity they were processed for immunofluorescence microscopy of the cytoskeletal proteins tubulin, microtubule-associated proteins 1 and 2 (MAP-1 and MAP-2), plectin, vimentin, and vinculin. Antibodies to tubulin decorated a delicate filamentous network that apparently was unrelated to any sarcomeric organization. The distribution of MAP-1 and MAP-2 was strikingly different from that of tubulin, as both antigens were confined to Z-line structures. These structures were also prominently stained by affinity-purified antibodies to plectin and a monoclonal antibody to vimentin. Co-distribution of plectin and vimentin was also observed at the former intercalated disk region of the heart cell. Anti-vinculin antibodies decorated an intricate meshwork consisting of delicate filaments with predominantly irregular orientation and occasional assembly into whorls. These immunolocalization data indicate that the cell shape and cytoskeletal architecture characteristic of cardiac myocytes in tissues is maintained in single isolated cells. Furthermore, intermediate filaments rather than microtubules seem to be instrumental in the preservation of cell morphology.     

Widespread occurrence of intermediate-sized filaments of the vimentin-type in cultured cells from diverse vertebrates.

Specific antibodies against vimentin, the major constitutive protein of intermediate-sized filaments present in cytoskeletons of mesenchymal cells of vertebrates, have been raised in guinea pigs. Antibodies to murine and human vimentin are of three types. The first two types produced against murine vimentin show an exclusive or preferential reaction with vimentin filaments of rodents. The third type raised against murine or human vimentin reacts with intermediate-sized filaments in species as diverse as mammals, birds and amphibia. This latter type is used here to show, both by immunoreplica techniques and by immunofluorescence microscopy, that almost all vertebrate cells growing in culture contain filaments of the vimentin type which are usually present in extended arrays. These immunological findings also suggest that the vimentin molecule contains both sequences conserved during evolution and regions different in different vertebrate species. The cells studied include not only cells of mesenchymal origin, but also cells derived from epithelia, in which it is now possible to demonstrate extensive arrays of vimentin filaments in interphase cells as well as intermediate-sized filaments of the prekeratin type. The data are consistent with the idea that most cells grown in culture contain intermediate-sized filaments of the vimentin type, irrespective of the state of differentiation of the cells from which they are derived.     

Myoseverin disrupts sarcomeric organization in myocytes: an effect independent of microtubule assembly inhibition

Although disruption of the microtubule (MT) array inhibits myogenesis in myocytes, the relationship between the assembly of microtubules (MT) and the organization of the contractile filaments is not clearly defined. We now report that the assembly of mature myofibrils in hypertrophic cardiac myocytes is disrupted by myoseverin, a compound previously shown to perturb the MT array in skeletal muscle cells. Myoseverin treated cardiac myocytes showed disruptions of the striated Z-bands containing alpha-actinin and desmin and the localization of tropomyosin, titin and myosin on mature sarcomeric filaments. In contrast, MT depolymerization by nocodazole did not perturb sarcomeric filaments. Similarly, expression of constitutively active stathmin as a non-chemical molecular method of MT depolymerization did not prevent sarcomere assembly. The extent of MT destabilization by myoseverin and nocodazole were comparable. Thus, the effect of myoseverin on sarcomere assembly was independent of its capacity for MT inhibition. Furthermore, we found that upon removal of myoseverin, sarcomeres reformed in the absence of an intact MT network. Sarcomere formation in cardiac myocytes therefore, does not appear to require an intact MT network and thus we conclude that a functional MT array appears to be dispensable for myofibrillogenesis.     

Expression of keratin and vimentin intermediate filaments in rabbit bladder epithelial cells at different stages of benzo[a]pyrene-induced neoplastic progression

Rabbit bladder epithelium, grown on collagen gels and exposed to the chemical carcinogen benzo[a]pyrene, produced nontumorigenic altered foci as well as tumorigenic epithelial cell lines during 120-180 d in culture. Immunofluorescence studies revealed extensive keratin filaments in both primary epithelial cells and benzo[a]pyrene-induced altered epithelial foci but showed no detectable vimentin filaments. The absence of vimentin expression in these cells was confirmed by two- dimensional gel electrophoresis. In contrast, immunofluorescence staining of the cloned benzo[a]pyrene-transformed rabbit bladder epithelial cell line, RBC-1, revealed a reduction in filamentous keratin concomitant with the expression of vimentin filaments. The epithelial nature of this cell line was established by the observation that cells injected into nude mice formed well-differentiated adenocarcinomas. Frozen sections of such tumors showed strong staining with antikeratins antibodies, but no detectable staining with antivimentin antibodies. These results demonstrated a differential expression of intermediate filament type in cells at different stages of neoplastic progression and in cells maintained in different growth environments. It is apparent that the expression of intermediate filaments throughout neoplastic progression is best studied by use of an in vivo model system in parallel with culture studies.     

Gangliosides of the bovine heart impulse conducting system

The ganglioside analysis of the heart impulse conducting system was carried out, comparing it with that of ordinary myocardium. The heart impulse conducting system contained about 3-times more gangliosides than ordinary myocardium and showed a distinctly different ganglioside composition. These observations seemed to indicate that the differentiation between myoblasts and each type of cardiac muscle cell, impulse conducting system and ordinary myocardium cells, resulted in their characteristic ganglioside compositions.     

Organization of cytokeratin bundles by desmosomes in rat mammary cells

In a rat mammary epithelial cell line, LA-7, cytokeratin bundles recognized in immunofluorescence by a monoclonal antibody (24B42) disappear after trypsinization of cultures and are gradually reformed after replating. We have followed the time course of cytokeratin filament reappearance by growing cells in low calcium medium (0.1 mM) which prevents desmosome formation, and then shifting to high calcium (1.8 mM) to start the process. By fixing the cells at various intervals and staining them in immunofluorescence for 24B42 cytokeratin and for desmosomal proteins, we found that cell to cell contact and desmosome formation are prerequisites for keratin filament formation in these cells. EGTA treatment, by disassembling desmosomes, causes the cytokeratin filaments to disappear and the 24B42 protein to pass into a soluble form in this cell line, as ascertained by 100,000 g fractionation and immunoenzymatic assay. Cycloheximide treatment also causes cytokeratin filaments to disappear, indicating that protein synthesis is needed for normal filament maintenance. In another related cell line (106A-10a) and in HeLa cells, trypsinization and EGTA exposure do not cause a complete loss of 24B42 immunofluorescence, although distinct filaments disappear, indicating the presence in these cells of different organizing centers, besides desmosomes, for cytokeratin bundle formation. LA7 cells therefore seem to have a cytokeratin system strictly dependent on the presence of desmosomes, which act as an organizing center for filament assembly. 106A-10a cells (also rich in desmosomes) and HeLa cells (showing instead a reduced number of desmosomes) have a cytokeratin system partially or totally independent from that of desmosomes, with different organizing centers.     

TRANSLOCATION OF PHOTOSYNTHATE IN THE BROWN ALGA NEREOCYSTIS

Translocation of photosynthates was found to occur when blades of Nereocystis were illuminated in the presence of C¹⁴ bicarbonates. Rates of translocation averaging 37 cm/hr in the laboratory were observed. Samples from the epidermis, cortex, and medulla of the stipes of plants with blades exposed to tracer showed that the radioactivity in organic compounds was confined to medulla where sieve filaments occurred. Girdle preparations of blades, interrupting the mucilage ducts and leaving the blade medulla intact, allowed translocation to take place. These data indicate that conduction of photosynthate takes place in the medulla. Similarities between the anatomy of algal sieve filaments and angiosperm sieve tubes, coupled with the continuity of the sieve filaments from blade medulla to stipe medulla suggested indirectly that the sieve filaments were conducting elements. Further support of this hypothesis was provided from collections of radioactive exudate from cells in the medulla of the lower stipe that were continuous with the sieve filaments. Tracer applied to the blades was partially recovered as organic material in a clear fluid that collected inside the pneumatocyst. Continued accumulation of radioactivity in this fluid was dependent on living blades; fluid with low specific activity that did not increase during the experiment accumulated slowly if blades were killed with ethanol after an exposure to tracer. It is possible that the system that produced the stipe fluid was part of (or a side effect of) the system responsible for maintaining volume in the conducting system. It may also provide an alternate route (other than the sieve filaments) for delivery of photosynthates to the base of the plant. Carbon-14 applied to blade tips as bicarbonate was recovered in part as radioactive mannitol in the translocation stream.     

Immunofluorescence evidence for cytoskeletal rearrangement accompanying pigment redistribution in goldfish xanthophores

Immunofluorescence and phase-contrast microscopic studies of goldfish xanthophores with aggregated or dispersed pigment show two unusual features. First, immunofluorescence studies with anti-actin show punctate structures instead of filaments. These punctate structures are unique for the xanthophores and are absent from both goldfish dermal non-pigment cells and a dedifferentiated cell line (GEM-81) derived from a goldfish xanthophore tumor. Comparison of immunofluorescence and phase-contrast microscopic images with electron microscopic images of thin sections and of Triton-insoluble cytoskeletons show that these punctate structures represent pterinosomes with radiating F-actin. The high local concentration of actin around the pterinosomes results in strong localized fluorescence such that, when the images have proper brightness for these structures, individual actin filaments elsewhere in the cell are too weak in their fluorescence to be visible in the micrographs. Second, whereas immunofluorescence images with anti-tubulin show typical patterns in xanthophores with either aggregated or dispersed pigment, namely, filaments radiating out from the microtubule organizing center, immunofluorescence images with anti-actin or with anti-intermediate filament proteins show different patterns in xanthophores with aggregated versus dispersed pigment. In cells with dispersed pigment, the punctate structures seen with anti-actin are relatively evenly distributed in the cytoplasm, and intermediate filaments appear usually as a dense perinuclear band and long filaments elsewhere in the cytoplasm. In cells with aggregated pigment, both intermediate filaments and pterinosomes with associated actin are largely excluded from the space occupied by the pigment aggregate, and the band of intermediate filaments surrounds not only the nucleus but also the pigment aggregate. The patterns of distribution of the different cytoskeleton components, together with previous results from this laboratory, indicate that formation of the pigment aggregate depends at least in part on the interaction between pigment organelles and microtubules. The possibility that intermediate filaments may play a role in the formation/stabilization of the pigment aggregate is discussed.