SDS GEL ELECTROPHORESIS OF RAPIDLY TRANSPORTED PROTEINS IN GARFISH OLFACTORY NERVE

Abstract— Proteins undergoing rapid axonal transport in the garfish olfactory nerve were examined by sodium dodecyl sulphate gel electrophoresis. The distribution of polypeptides and the extent of their labeling by transported molecules was determined in several nerve subfractions including: total particulate, total membrane, mitochondrial and two membrane subfractions rich in axolemma. The polypeptide composition of the various fractions was found to be relatively similar, with each showing a major protein with an estimated MW of 58,000. Specific differences in the concentrations of certain proteins were noted between fractions, including differences between the lower and higher density axolemma rich subfractions. Axonally transported radioactivity was predominantly localized among high molecular weight proteins, with all fractions, except mitochondrial pellet, displaying a major peak of radioactivity centered at 126,000-MW. Several major proteins including the 58,000-MW band were labeled by rapid transport to a much smaller extent. Certain labeled peaks were found to be concentrated in individual fractions, particularly a polypeptide (MW 35,000) more predominantly found in the lower density axolemma rich fraction.
Systemic labeling of the nerve is found to give a general distribution of radioactivity on gels, which is clearly different from the pattern obtained after axonal transport labeling.     

Subaxolemmal cytoskeleton in squid giant axon. II. Morphological identification of microtubule- and microfilament-associated domains of axolemma

In the preceding paper (Kobayashi, T., S. Tsukita, S. Tsukita, Y. Yamamoto, and G. Matsumoto, 1986, J. Cell Biol., 102:1710-1725), we demonstrated biochemically that the subaxolemmal cytoskeleton of the squid giant axon was highly specialized and mainly composed of tubulin, actin, axolinin, and a 255-kD protein. In this paper, we analyzed morphologically the molecular organization of the subaxolemmal cytoskeleton in situ. For thin section electron microscopy, the subaxolemmal cytoskeleton was chemically fixed by the intraaxonal perfusion of the fixative containing tannic acid. With this fixation method, the ultrastructural integrity was well preserved. For freeze-etch replica electron microscopy, the intraaxonally perfused axon was opened and rapidly frozen by touching its inner surface against a cooled copper block (4 degrees K), thus permitting the direct stereoscopic observation of the cytoplasmic surface of the axolemma. Using these techniques, it became clear that the major constituents of the subaxolemmal cytoskeleton were microfilaments and microtubules. The microfilaments were observed to be associated with the axolemma through a specialized meshwork of thin strands, forming spot-like clusters just beneath the axolemma. These filaments were decorated with heavy meromyosin showing a characteristic arrowhead appearance. The microtubules were seen to run parallel to the axolemma and embedded in the fine three-dimensional meshwork of thin strands. In vitro observations of the aggregates of axolinin and immunoelectron microscopic analysis showed that this fine meshwork around microtubules mainly consisted of axolinin. Some microtubules grazed along the axolemma and associated laterally with it through slender strands. Therefore, we were led to conclude that the axolemma of the squid giant axon was specialized into two domains (microtubule- and microfilament-associated domains) by its underlying cytoskeletons.     

Subaxolemmal cytoskeleton in squid giant axon. I. Biochemical analysis of microtubules, microfilaments, and their associated high-molecular- weight proteins

Using the squid giant axon, we analyzed biochemically the molecular organization of the axonal cytoskeleton underlying the axolemma (subaxolemmal cytoskeleton). The preparation enriched in the subaxolemmal cytoskeleton was obtained by squeezing out the central part of the axoplasm using a roller. The electrophoretic banding pattern of the subaxolemmal cytoskeleton was characterized by large amounts of two high-molecular-weight (HMW) proteins (260 and 255 kD). The alpha, beta-tubulin, actin, and some other proteins were also its major constituents. The 260-kD protein is known to play an important role in maintaining the excitability of the axolemma (Matsumoto, G., M. Ichikawa, A. Tasaki, H. Murofushi, and H. Sakai, 1983, J. Membr. Biol., 77:77-91) and was recently designated "axolinin" (Sakai, H., G. Matsumoto, and H. Murofushi, 1985, Adv. Biophys., 19:43-89). We purified axolinin and the 255-kD protein in their native forms and further characterized their biochemical properties. The purified axolinin was soluble in 0.6 M NaCl solution but insoluble in 0.1 M NaCl solution. It co-sedimented with microtubules but not with actin filaments. In low-angle rotary-shadowing electron microscopy, the axolinin molecule in 0.6 M NaCl solution looked like a straight rod approximately 105 nm in length with a globular head at one end. On the other hand, the purified 255-kD protein was soluble in both 0.1 and 0.6 M NaCl solution and co-sedimented with actin filaments but not with microtubules. The 255-kD protein molecule appeared as a characteristic horseshoe-shaped structure approximately 35 nm in diameter. Furthermore, the 255-kD protein showed no cross-reactivity to the anti-axolinin antibody. Taken together, these characteristics lead us to conclude that the subaxolemmal cytoskeleton in the squid giant axon is highly specialized, and is mainly composed of microtubules and a microtubule-associated HMW protein (axolinin), and actin filaments and an actin filament-associated HMW protein (255-kD protein).     

Exocytosis of neurosecretory granules from the crustacean sinus gland in freeze-fracture

Exocytosis is clearly shown in freeze-fracture preparations to be the mechanism for neurosecretion granule release from axon endings in the crayfish sinus gland. The cytoplasmic leaflet (A-face) of axon ending membrane is characterized by randomly situated depressions representing invaginations of the axolemma, which are in contact with limiting membranes of neurohormone granules in the subjacent cytoplasm. The extracellular leaflet (B-face) of the axolemma at release sites exhibits complementary volcano-shaped protrusions which are cross-fractures through necks of channels formed by invaginating plasma membrane in contact with underlying neurosecretion granules. Structural variation in B-face protrusions is consistent with a spectrum of exocytotic profiles in various stages of formation, and with granules at different stages of passage out of the endings. Evidence in this study suggests that formation of exocytotic structures may begin by alteration of axon membrane structure at the neurosecretory ending-hemolymph interface prior to contact of the neurohormone granules with the axolemma. Limiting membranes of neurosecretory granules exhibit protrusions which appear to interconnect granules adjacent to release sites and to attach granules to the axolemma. Freeze-fracture is clearly shown to be an invaluable tool for monitoring the degree of exocytosis exhibited by sinus glands under normal conditions and under experimental acceleration of hormone release. This technique is capable therefore, of detecting slight increases in numbers of exocytotic profiles much more quickly and accurately than the examination of random thin sections.     

Synaptophysin and synapsin I as tools for the study of the exo-endocytotic cycle

Synaptophysin, an integral protein of the synaptic vesicle membrane, and synapsin I, a phosphoprotein associated with the cytoplasmic side of synaptic vesicles, represent useful markers that allow to follow the movements of the vesicle membrane during recycling. The use of antibodies against these proteins to label nerve terminals during experimental treatments which stimulate secretion has provided evidence that during the exo-endocytotic cycle synaptic vesicles transiently fuse with the axolemma, from which they are specifically recovered. When recycling is blocked, exocytosis leads to the permanent incorporation of the synaptic vesicle membrane into the axolemma and to diffusion of the vesicle components in the plane of the membrane.     

Axonal microtubules necessary for generation of sodium current in squid giant axons: I. Pharmacological study on sodium current and restoration of sodium current by microtubule proteins and 260K protein

Effects of the reagents suppressing or supporting axoplasmic microtubule assembly were studied on the Na ionic current of squid giant axons by perfusing the axon internally with the solution containing the reagent. Among the reagents suppressing the assembly, colchicine, vinblastine, podophyllotoxin, sulfhydryl reagents such as DTNB and NEM, and chaotropic anions such as iodide and bromide, were examined. These reagents reduced maximum Na conductance and shifted the voltage dependence of steady-state Na activation in a depolarizing direction along the voltage axis. They also made the voltage dependence less steep, but did not affect sodium inactivation appreciably. Effects on Na ionic current of reagents which support microtubule assembly (Taxol, DMSO, D2O and temperature) were opposite the effects of those agents suppressing assembly. At the same time, we demonstrated that after Na currents were partially reduced, they could be restored by internally perfusing the axon with a solution containing microtubule proteins, 260K proteins and cAMP under conditions favorable for microtubule assembly. For full restoration, it was found that the following conditions were necessary: (1) The microenvironment within the axon is suitable for microtubule assembly. (2) Tubulins incorporated into microtubules are fully tyrosinated at their C-termini. (3) A peripheral protein having a molecular weight of 260,000 daltons (260K protein) is indispensable. These results suggest that axoplasmic microtubules and 260K proteins in the structure underlying the axolemma play a role in generating Na currents in squid giant axons.     

Axonal microtubules necessary for generation of sodium current in squid giant axons: I. pharmacological study on sodium current and restoration of sodium current by microtubule proteins and 260K protein

Summary Effects of the reagents suppressing or supporting axoplasmic microtubule assembly were studied on the Na ionic current of squid giant axons by perfusing the axon internally with the solution containing the reagent. Among the reagents suppressing the assembly, colchicine, vinblastine, podophyllotoxin, sulfhydryl reagents such as DTNB and NEM, and chaotropic anions such as iodide and bromide, were examined. These reagents reduced maximum Na conductance and shifted the voltage dependence of steady-state Na activation in a depolarizing direction along the voltage axis. They also made the voltage dependence less steep, but did not affect sodium inactivation appreciably. Effects on Na ionic current of reagents which support microtubule assembly (Taxol, DMSO, D₂O and temperature) were opposite the effects of those agents suppressing assembly. At the same time, we demonstrated that after Na currents were partially reduced, they could be restored by internally perfusing the axon with a solution containing microtubule proteins, 260K proteins and cAMP under conditions favorable for microtubule assembly. For full restoration, it was found that the following conditions were necessary: (1) The microenvironment within the axon is suitable for microtubule assembly. (2) Tubulins incorporated into microtubules are fully tyrosinated at their C-termini. (3) A peripheral protein having a molecular weight of 260,000 daltons (260K protein) is indispensable. These results suggest that axoplasmic microtubules and 260K proteins in the structure underlying the axolemma play a role in generating Na currents in squid giant axons.     

The cytoskeletal adapter protein 4.1G organizes the internodes in peripheral myelinated nerves

Myelinating Schwann cells regulate the localization of ion channels on the surface of the axons they ensheath. This function depends on adhesion complexes that are positioned at specific membrane domains along the myelin unit. Here we show that the precise localization of internodal proteins depends on the expression of the cytoskeletal adapter protein 4.1G in Schwann cells. Deletion of 4.1G in mice resulted in aberrant distribution of both glial adhesion molecules and axonal proteins that were present along the internodes. In wild-type nerves, juxtaparanodal proteins (i.e., Kv1 channels, Caspr2, and TAG-1) were concentrated throughout the internodes in a double strand that flanked paranodal junction components (i.e., Caspr, contactin, and NF155), and apposes the inner mesaxon of the myelin sheath. In contrast, in 4.1G(-/-) mice, these proteins "piled up" at the juxtaparanodal region or aggregated along the internodes. These findings suggest that protein 4.1G contributes to the organization of the internodal axolemma by targeting and/or maintaining glial transmembrane proteins along the axoglial interface.     

Mitogenic Effect of Axolemma-Enriched Fraction on Cultured Oligodendrocytes

An in vitro system has been devised to study the mitogenic effect of axolemma on cultured oligodendrocytes. Addition of axolemma-enriched fraction to cultured oligodendrocytes results in a dose-dependent mitotic response with an 11-fold stimulation at a membrane concentration of 200 micrograms/ml. The interaction between oligodendrocytes and axolemma is specific, as myelin-enriched fraction, astrocyte membrane, and red blood cell membrane showed little or no effect on the oligodendroglial proliferation under similar conditions. In addition, cultured astrocytes were tested with the same axolemma membrane, and no mitotic stimulation was observed. The mitogenicity of AEF membrane on cultured oligodendrocytes is sensitive to heat and trypsin treatment, suggesting that the axolemma mitogen may be a protein.     

Subaxolemmal filamentous network in the giant nerve fiber of the squid (Loligo pealei L.) and its possible role in excitability

A new technique utilizing the squid giant nerve fiber has been developed which permits direct examination of the inner face of the axolemma by scanning electron microscopy. The axoplasm was removed sequentially in a 15-mm long segment of the fiber by intracellular perfusion with a solution of KF, KCl, Ca++-containing seawater, or with pronase. The action potential of the fibers was monitored during these treatments. After brief prefixation in 1% paraformaldehyde and 1% glutaraldehyde, the perfused segment was opened by a lne could be related to information on the detailed morphology of the cytoplasmic face of the axolemma and the ectoplasm. The results obtained by scanning electron microscopy were further substantiated by transmission electron microscopy of thin sections. In addition, living axons were studied with polarized light during axoplasm removal, and the identification of actin by heavy meromyosin labeling and sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis was accomplished. These observations demonstrate that a three-dimensional network of interwoven filaments, consisting partly of an actinlike protein, is firmly attached to the axolemma. The axoplasmic face of fibers in which the filaments have been removed partially after perfusion with pronase displays smooth membranous blebs and large profiles which sppose the axolemma. In fibers where the excitability has been suppressed by pronase perfusion, approximately one-third of the inner face of the axolemma in the perfusion zone is free of filaments. It is hypothesized that the attachment of axoplasm filaments to the axolemma may have a role in the maintenance of the normal morphology of the axolemma, and, thus, in some aspect of excitability.     

Isolation and Characterization of Rapid Transport Vesicle Subtypes from Rabbit Optic Nerve

Subcellular fractionation of rabbit optic nerve resolves three populations of membranes that are rapidly labelled in the axon. The lightest membranes are greater than 200 nm and are relatively immobile. The intermediate density membranes consist of 84 nm vesicles which disappear from the nerve with kinetics identical to those of the rapid component. A third population of membranes, displaying a distinct protein profile, is present in the most dense region of the gradient. Immunological characterization of these membranes suggests the following. (1) The lightest peak contains rapidly transported glucose transporter and most of the total glucose transporters present in the nerve; this peak is therefore enriched in axolemma. (2) The intermediate peak contains rapidly transported glucose transporters and synaptophysin, an integral synaptic vesicle protein, and about half of the total synaptophysin; this peak therefore contains transport vesicles bound for both the axolemma and the nerve terminal, and these subpopulations can be separated by immunoadsorption with specific antibodies against the aforementioned proteins. (3) The heaviest peak contains rapidly transported synaptophysin and tachykinin neuromodulators and about half of the total synaptophysin, and 80% of the total tachykinins present in the nerve; this peak appears to represent a class of synaptic vesicle precursor bound for the nerve terminal exclusively. (4) Synaptophysin is present in the membranes of vesicles carrying tachykinins. (5) Both the intermediate and the heaviest peaks are enriched in kinesin heavy chain, suggesting that both vesicle classes may be transported by the same mechanism.     

Protein prenylation in glucose-induced insulin secretion from the pancreatic islet beta cell: a perspective.

Insulin secretion from the pancreatic beta-cell is regulated principally by the ambient concentration of glucose. However, the molecular and cellular mechanisms underlying the stimulus-secretion coupling of glucose-stimulated insulin secretion (GSIS) remain only partially understood. Emerging evidence from multiple laboratories suggests key regulatory roles for GTP-binding proteins (G-proteins) in the cascade of events leading to GSIS. This class of signaling proteins undergo a series of requisite post-translational modifications (e.g., prenylation) at their C-terminal cysteines, which appear to be necessary for their targeting to respective membranous sites for optimal interaction with their respective effector proteins. This communication represents a perspective on potential regulatory roles for protein prenylation steps (i.e., protein farnesylation and protein geranylgeranylation) in GSIS from the islet beta cell. Possible consequences of protein prenylation and potential mechanisms underlying glucose-induced regulation of prenylation, specifically in the context of GSIS are also discussed.     

Snap-25 is polarized to axons and abundant along the axolemma: an immunogold study of intact neurons

SNAP-25, synaptosomal associated protein of 25 kDa, is reported to be a t-SNARE (target receptor associated with the presynaptic plasma membrane) involved in the docking and fusion of synaptic vesicles. We present here the first ultrastructural localization of SNAP-25 in intact neurons by pre-embedding EM immunocytochemistry in rat brains, hippocampal slice cultures, and PC12 cells. In differentiated neurons, SNAP-25 labeling was clearly membrane-associated. The labeling was most prominent in the plasma membrane of axons and excluded from the plasma membranes of soma and dendrites. Furthermore, SNAP-25 did not appear to be restricted to the synaptic junctions. SNAP-25 labeling was seen in the cytoplasm of the soma and large dendrites, mostly associated with the Golgi complexes. There were also some SNAP-25 labeled tubulo-vesicular structures in the cytoplasm of the soma and the axons, but rarely in the smaller dendrites. In PC12 cells, after 5–10 minutes of high potassium (75 mM) stimulation in the presence of HRP, SNAP-25 labeling appeared, additionally, on HRP-filled early endosomes. After a longer (20–30 minutes) HRP incubation, most of the later stage endosomes and lysosomes were loaded with HRP but they were negative for SNAP-25. These results suggest that SNAP-25 is sorted out of these late endosomal compartments, and that the bulk of the SNAP-25 protein is probably recycled back to the axolemma from the early endosomes. In contrast, in those samples which were incubated with HRP for longer periods, there were still some SNAP-25–positive vesicular structures which were HRP-negative. These structures most likely represent anterograde vesicles that carry newly synthesized SNAP-25 from the soma to the axolemma by axonal transport. SNAP-25 appears to be sorted at the Golgi complex to reach the axolemma specifically. Its widespread distribution all along the axolemma does not support the view of SNAP-25 as a t-SNARE limited for synaptic exocytosis.     

p103 and A60: novel proteins of the neuronal membrane-associated cytoskeleton.

Associated with the neuronal plasma membrane are cytoskeletal proteins which probably control the specialization of the membrane into axonal and dendritic domains. Specialized isoforms of the proteins spectrin and ankyrin are located in each region and provide molecular mechanisms for locating specific transmembrane proteins at required points. However, spectrin and ankyrin were defined by extensions of the model for the erythrocyte membrane, an analogy unlikely to provide a complete account of the neuronal membrane skeleton. We have defined two new proteins of the neuronal membrane skeleton, designated p103 and A60. p103 is enriched in post-synaptic densities and binds with high affinity to integral membrane proteins--we suggest that it may have a role in linking the cytoskeleton to synaptic glycoproteins. A60 is a 60 kDa axonal protein, which appears to form a lining to the axolemma. It is almost exclusively axonal, although some neurons (such as Purkinje cells) appear to contain it in the cell body and initial dendrite segment. A60 binds both ankyrin and neurofilaments, and may have a role in transmitting information critical to axonal morphology to the membrane.     

Synapsin I partially dissociates from synaptic vesicles during exocytosis induced by electrical stimulation.

The distribution of the synaptic vesicle-associated phosphoprotein synapsin I after electrical stimulation of the frog neuromuscular junction was investigated by immunogold labeling and compared with the distribution of the integral synaptic vesicle protein synaptophysin. In resting terminals both proteins were localized exclusively on synaptic vesicles. In stimulated terminals they appeared also in the axolemma and its infoldings, which however exhibited a lower synapsin I/synaptophysin ratio with respect to synaptic vesicles at rest. The value of this ratio was intermediate in synaptic vesicles of stimulated terminals, and an increased synapsin I labeling of the cytomatrix was observed. These results indicate that synapsin I undergoes partial dissociation from and reassociation with synaptic vesicles, following physiological stimulation, and are consistent with the proposed modulatory role of the protein in neurotransmitter release.     

Effects of nitric oxide on protein-lipid interactions in the membranes of the myelinated nerve fiber

The influence of nitric oxide (NO) on the myelinated nerve fiber and the impact of modification of SHgroups of axon and myelin membrane proteins on the amplitude and propagation velocity of action potential (AP), amount of the membrane-bound calcium (Ca _mb ²⁺ , viscosity of the axon membrane, and saturation factor of phospholipid fatty acids (Sf) of myelin have been investigated. We established that the decrease in the number of extracellular SH-groups in membrane proteins induced by p-chloromercuribenzoate (pCMB, 10^?4 M), led to a decrease in the AP amplitude and a reversible desorption of Ca _mb ²⁺ but did not affect the axolemma viscosity and Sf. Nitric oxide (NO) caused a decrease in the AP amplitude and propagation velocity, an increase in the axolemma viscosity and a decrease in Sf of myelin; it also induced a reversible desorption of Ca _mb ²⁺ . Pretreatment of the nerve fiber with pCMB weakened the NO-induced desorption of Pretreatment of the nerve fiber with K⁺-channel blocker tetraethylammonium (10^?2 M) completely abolished the NO-induced change in the amount of Ca _mb ²⁺ . We suppose that NO-mediated changes in axolemma viscosity, Sf of myelin and desorption of Ca _mb ²⁺ affect protein-lipid interactions in axolemma and myelin, which in their turn influence the propagation of AP.     

Membrane-Associated Cytoskeletal Proteins in Squid Giant Axons

Abstract: Cytoskeletal proteins (e.g., tubulin, actin, and neurofilament proteins) in the squid giant axon are separable into KF-soluble and -insoluble forms. The KF-insoluble cytoskeletal components appear to constitute the major proteins in the subaxolemmal fibrous network on the inner surface of the axon. These cytoskeletal proteins and the subaxolemmal network are both highly soluble in KI solutions. Whereas giant axons tolerate prolonged perfusions in KF solutions with no loss of excitable properties, a relatively short perfusion with KI solution completely eliminates the excitability of the axon. The loss of this excitability correlates with the simultaneous dissolution of the subaxolemmal network of cytoskeletal proteins and the release of its proteins into the perfusate. These data support the hypothesis that cytoskeletal proteins associated with the inner surface of the axolemma are involved in the regulation of axonal excitability.     

GLIAL VERSUS NEURONAL ORIGIN OF MYELIN PROTEINS AND GLYCOPROTEINS STUDIED BY COMBINED INTRAOCULAR AND INTRACRANIAL LABELLING

Abstract— —The contribution of axonal transport to the production of myelin proteins and glycoproteins was investigated using the double labelling technique of combined intraocular and intracerebral injections in the same animal. Myelin and an axolemma-enriched fraction were isolated from pooled optic nerves, chiasma and optic tracts. Separation by gel electrophoresis showed that typical myelin proteins and glycoproteins were only significantly labelled by intracerebral injection. Intraocular injection labelled high molecular weight proteins other than the major Wolfgram protein and the major myelin glycoprotein. Fifteen days after intraocular injection the label was concentrated in a high molecular weight protein which migrated slightly more slowly than the major Wolfgram protein. The pattern of proteins and glycoproteins in myelin labelled by intraocular injection was very similar to that obtained in the axolemma-enriched fraction by the same route. These results indicate that neuronal metabolism and axonal transport do not contribute significantly to the synthesis of specific myelin proteins and glycoproteins, but suggest that the components of myelin fractions which are labelled by intraocular injection are contaminants of axolemmal origin. One of these glycoproteins may prove a useful marker of axolemma membranes.     

Characterization of the Resting Axolemma in the Giant Axon of the Squid

Previous electron microscope studies have shown that the Schwann cell layer is traversed by long and tortuous slit-like channels ~60Å wide, which provide the major route of access to the axolemma surface. In the present work the restriction offered by the resting axolemma to the passage of six small non-electrolyte molecules has been determined. The radii of the probing molecules were estimated from constructed molecular models. The ability of the axolemma to discriminate between the solvent (water) and each probing molecule was expressed in terms of the reflection coefficient σ. σ was then used to calculate an effective pore size for the resting axolemma. The value of 4.25 Å found for the pore radius is in excellent agreement with the 1.5 to 8.5 Å limiting values previously calculated from our measurements of water fluxes. The presence of pores with 4.25 Å radius in the resting axolemma is compatible with restricted diffusion of Na. The present paper leads to the conclusion that the axolemma is the only continuous barrier across which the ionic gradient responsible for the normal functioning of the nerve can be maintained. The combined findings of electron microscopy, water permeability, and molecular restricted filtration indicate that in all probability the axolemma is the "excitable membrane" of the physiologists.     

Markovian domain fingerprinting: statistical segmentation of protein sequences.

MOTIVATION: Characterization of a protein family by its distinct sequence domains is crucial for functional annotation and correct classification of newly discovered proteins. Conventional Multiple Sequence Alignment (MSA) based methods find difficulties when faced with heterogeneous groups of proteins. However, even many families of proteins that do share a common domain contain instances of several other domains, without any common underlying linear ordering. Ignoring this modularity may lead to poor or even false classification results. An automated method that can analyze a group of proteins into the sequence domains it contains is therefore highly desirable. RESULTS: We apply a novel method to the problem of protein domain detection. The method takes as input an unaligned group of protein sequences. It segments them and clusters the segments into groups sharing the same underlying statistics. A Variable Memory Markov (VMM) model is built using a Prediction Suffix Tree (PST) data structure for each group of segments. Refinement is achieved by letting the PSTs compete over the segments, and a deterministic annealing framework infers the number of underlying PST models while avoiding many inferior solutions. We show that regions of similar statistics correlate well with protein sequence domains, by matching a unique signature to each domain. This is done in a fully automated manner, and does not require or attempt an MSA. Several representative cases are analyzed. We identify a protein fusion event, refine an HMM superfamily classification into the underlying families the HMM cannot separate, and detect all 12 instances of a short domain in a group of 396 sequences. CONTACT: jill@cs.huji.ac.il; tishby@cs.huji.ac.il.