带有隐含神经元的单层神经网络模型

提出了一种带有隐含神经元的单层神经网络模型。把网络的记忆容量区分为信息记忆容量和物理记忆容量。新模型能记忆相关图样,其信息容量α_i(最大记忆图样数/表达神经元数)首次超过了1。所作出的计算机模拟结果,表明了理论分析的正确性,证实了由5×5个显神经元组成的点阵能记住包含26个英文字母和4个任选图样的30个图样,因此该模型为神经网络的广泛应用提供了一条重要途径。     

Quantitation of the Myelin-Associated Glycoprotein in Human Nervous Tissue from Controls and Multiple Sclerosis Patients

Myelin-associated glycoprotein (MAG) was measured by radioimmunoassay in the human CNS and peripheral nervous system (PNS). The level of MAG, expressed as ng/microgram of total protein, was approximately 20-fold higher in whole homogenates of cerebral white matter (4.7 +/- 0.60) than of peripheral nerve (0.12-0.28). MAG concentrations were only slightly higher in the isolated myelin fractions from these tissues: CNS myelin, 5.6 ng/microgram; PNS myelin, 0.37 ng/microgram. The levels of MAG were measured in nine plaques, periplaque regions, and areas of macroscopically normal-appearing white matter (NAWM) from six separate multiple sclerosis brains and compared with the levels of other myelin proteins in the same samples. MAG and other myelin proteins were reduced to very low levels in plaques. The levels of MAG and basic protein (BP) and the activity of 2',3'-cyclic nucleotide 3'-phosphodiesterase (CNP) in periplaque areas were significantly lower than those in control white matter, and MAG and BP levels were also significantly reduced in NAWM. In a periplaque region and NAWM from the most rapidly progressing case of multiple sclerosis examined, the MAG content was between 30 and 35% of the control level, whereas BP and PLP levels and CNP activity were between 50 and 85% of control values. The reduction of MAG content in periplaque regions from all nine multiple sclerosis plaques examined was significantly greater than the reductions of BP level and CNP activity. In NAWM samples, the mean reduction of MAG content was also greater than the reductions of BP level and CNP activity, but the difference was only statistically significant in comparison to CNP.(ABSTRACT TRUNCATED AT 250 WORDS)     

Evidence for Specific Polypeptide Chain Folding in Myelin Basic Protein from Reactions Between Fragments of the Protein and Monoclonal Antibodies

The specificities of two monoclonal IgM antibodies (18.25 and 21.14.2) evoked in mice with guinea pig myelin basic protein were examined and interpreted in terms of a specific folding of the protein's polypeptide chain. Studies with guinea pig and rabbit myelin basic protein fragments showed that a region encompassing the central Phe-Phe (87-88) sequence is obligatory, but not sufficient, for reactivity with antibody 18.25. Appreciable reactivity was observed for rabbit peptides 22-95 and 45-151, and lower, but significant, reactivity was shown by peptide 32-95. Only very weak reactivity was seen with peptide 44-95. No reactivity was observed with peptide 1-95 after its lysine residues were acetylated, acetamidinated, or guanidinated. These results have been interpreted in terms of a polypeptide chain folding that creates an epitope within sequence Val-Val-His-Phe-Phe-Lys-Asn-Ile-Val (84-92). The specific conformation of this epitope, which includes probably the Lys-89 and possibly the Asn-90 and Val-92 side chains, could be formed by the association of sequence 84-92 with either sequence Ile-Leu-Asp-Ser-Ile-Gly-Arg-Phe-Phe (37-45) or with sequence Val-Leu-Ser-Arg-Phe (108-112) to form beta-sheet structures essentially identical with those that appear to be present in the intact BP [Martenson R.E.J. Neurochem. 46, 1612-1622 (1986)]. The second monoclonal antibody, no. 21.14.2, reacts only with guinea pig myelin basic protein and fragments containing the species-restricted sequence Arg-Ala-Asp-Tyr-Lys-Ser-Lys (129-135).(ABSTRACT TRUNCATED AT 250 WORDS)     

Evolutionary Divergence in the Structure of Myelin Basic Protein: Comparison of Chondrichthye Basic Proteins with Those from Higher Vertebrates

Abstract: A basic protein has been purified from the CNS myelin of the gummy shark (Mustelus antarticus). Electroblotting was used to examine the capacity of rabbit antisera raised against this electrophoretically pure protein to recognize myelin basic protein from higher vertebrates. The antisera bound to two shark proteins including the original polypeptide antigen and to chicken, bovine, and human myelin basic proteins. Thus, the shark protein appeared to possess antigenic determinants that have been retained through evolutionary divergence of these proteins. Whereas bovine basic protein caused experimental allergic encephalomyelitis in guinea pigs, animals that received injections of the shark protein showed neither clinical nor histological signs of this disease. However, tests for delayed-type hypersensitivity and for Arthus reaction following injection with the shark protein revealed a T-cell-mediated response to this antigen and substantial cross-reactivity with higher vertebrate basic proteins. Analysis of the amino acid composition of the shark protein, and comparison of its tryptic peptide map with that of the bovine protein, revealed substantial changes in the amino acid sequence. Although the shark protein has some antigenic determinants in common with the proteins from higher vertebrates, it appears that much of the structure differs.     

Bony Fish Myelin: Evidence for Common Major Structural Glycoproteins in Central and Peripheral Myelin of Trout

Peripheral nervous system (PNS) myelin from the rainbow trout (Salmo gairdneri) banded at a density of 0.38 M sucrose. The main myelin proteins consisted of (1) two basic proteins, BPa and BPb (11,500 and 13,000 MW, similar to those of trout central nervous system (CNS) myelin proteins BP1 and BP2), and (2) two glycosylated components, IPb (24,400 MW) and IPc (26,200 MW). IPc comigrated with trout CNS myelin protein IP2 in sodium dodecyl sulfate-polyacrylamide gel electrophoresis, whereas trout CNS myelin protein IP1 had a lower molecular weight (23,000). Following two-dimensional separation, however, both IPb and IPc from PNS showed two components; the more acidic component of IPc comigrated with IP2 from CNS. PNS tissue autolysis led to the formation of IPa (20,000 MW), consisting of two components in isoelectric focusing of which again the more acidic one comigrated with the CNS autolysis product IP0. Limited enzymatic digestion of isolated IP proteins from PNS and CNS led to closely similar degradation patterns, being most pronounced in the case of IP2 and IPc. Immunoblotting revealed that all IP components from trout PNS and CNS myelins reacted with antibodies to trout IP1 (CNS) and bovine P0 protein (PNS) whereas antibodies to rat PLP (CNS) were entirely unreactive. All BP components from trout PNS and CNS myelins bound to antibodies against human myelin basic protein. On the basis of these studies trout PNS and CNS myelins contain at least one common IP glycoprotein, whereas other members of the IP myelin protein family appear closely related. In the CNS myelin of trout the IP components appear to replace PLP.(ABSTRACT TRUNCATED AT 250 WORDS)     

Cleavage of Rabbit Myelin Basic Protein by Plasmin: Isolation and Identification of the Major Products

Rabbit myelin basic protein (BP) was subjected to partial cleavage with plasmin, and 15 cleavage products were isolated by a combination of gel filtration and ion-exchange chromatography. Their identification was achieved by amino acid analysis and tryptic peptide mapping, supplemented in some instances by carboxy-terminal analyses with carboxypeptidases A, B, and Y and amino-terminal analyses with dipeptidyl aminopeptidase I. The results showed that major plasmic cleavage sites included the Lys89-Asn90, Lys133-Ser134, and Lys153-Leu154 bonds. Cleavages also occurred at the Arg31-His32, Lys53-Arg54, and Arg25-His26 bonds, but these appeared to be less extensive. A large number of additional peptides were produced in relatively low yield. The smaller of these were isolated from heterogeneous fractions by high-voltage electrophoresis-TLC. Amino acid analysis of these peptides showed that minor cleavage sites included the Arg9-His10, Lys13-Tyr14, Lys103-Gly104, Lys137-Gly138, Lys140-Gly141, and Arg160-Ser161 bonds. In spite of a lower selectivity toward peptide bonds in BP as compared with pepsin, cathepsin D, and thrombin, plasmin has the advantage over the former proteinases in that it does not cleave at or near the Phe44-Phe45 bond. Instead it cleaves at the Arg31-His32 and Lys53-Arg54 bonds, thus preserving the entire hydrophobic sequence Ile-Leu-Asp-Ser-Ile-Gly-Arg-Phe-Phe as well as short sequences to either side.     

Calcium-Stimulated Proteolysis in Myelin: Evidence for a Ca2+-Activated Neutral Proteinase Associated with Purified Myelin of Rat CNS

Incubation of myelin purified from rat spinal cord with CaCl2 (1-5 mM) in 10-50 mM Tris-HCl buffer at pH 7.6 containing 2 mM dithiothreitol resulted in the loss of both the large and small myelin basic proteins (MBPs), whereas incubation of myelin with Triton X-100 (0.25-0.5%) and 5 mM EGTA in the absence of calcium produced preferential extensive loss of proteolipid protein (PLP) relative to MBP. Inclusion of CaCl2 but not EGTA in the medium containing Triton X-100 enhanced degradation of both PLP and MBPs. The Ca2+-activated neutral proteinase (CANP) activity is inhibited by EGTA (5 mM) and partially inhibited by leupeptin and/or E-64c. CANP is active at pH 5.5-9.0, with the optimum at 7-8. The threshold of Ca2+ activation is approximately 100 microM. The 150K neurofilament protein (NFP) was progressively degraded when incubated with purified myelin in the presence of Ca2+. These results indicate that purified myelin is associated with and/or contains a CANP whose substrates include MBP, PLP, and 150K NFP. The degradation of PLP (trypsin-resistant) in the presence of detergent suggests either release of enzyme from membrane and/or structural alteration in the protein molecule rendering it accessible to proteolysis. The myelin-associated CANP may be important not only in the turnover of myelin proteins but also in myelin breakdown in brain diseases.     

Transport of basic amino acids in Riccia fluitans: Evidence for a second binding site

The transport of several amino acids with different side-chain characteristics has been investigated in the aquatic liverwort Riccia fluitans. i) The saturation of system I (neutral amino acids) by addition of excess -aminoisobutyric acid to the external medium completely eliminated the electrical effects which are usually set off by neutral amino acids. Under these conditions arginine and lysine significantly depolarized the plasmalemma. ii) L- and D-lysine/arginine were discriminated against in favour of the L-isomers. iii) Increasing the external proton concentration in the interval pH 9 to 4.5 stimulated plasmalemma depolarization, electrical net current, and uptake of [¹⁴C]-basic amino acids. iv) Uptake of [¹⁴C]-glutamic acid took place only at acidic pHs. v) [¹⁴C]-histidine uptake had an optimum between pH 6 and 5.5. vi) Overlapping of the transport of basic, neutral, and acidic amino acids was common. It is suggested that besides system I, a second system (II), specific for basic amino acids, exists in the plasmalemma of Riccia fluitans. It is concluded that the amino-acid molecule with an uncharged side chain is the substrate for system I, which also binds and transports the neutral species of acidic amino acids, whereas system II is specific for amino acids with a positively charged side chain. The possibility of system II being a proton cotransport is discussed.Abbreviation AiB -aminoisobutyric acid