Search

In order to generalize from a training set to a test set, it is desirable that small changes in the input space of a pattern do not change the output components. This can be done by forcing this behavior as part of the training algorithm. This is done in double backpropagation by forming an energy function that is the sum of the normal energy term found in backpropagation and an additional term that is a function of the Jacobian. Significant improvement is shown with different architectures and different test sets, especially with architectures that had previously been shown to have very good performance when trained using backpropagation. It is shown that double backpropagation, as compared to backpropagation, creates weights that are smaller, thereby causing the output of the neurons to spend more time in the linear region.

A simple purification procedure has been developed for the extracellular alkaline protease from Neurospora crassa. Key steps in the purification were: 1) the choice of gelatin as the protein inducer, which induces optimally at a much lower concentration than other commonly employed protein inducers; 2) heat treatment, during which the inducer is digested by the protease; and 3) a concentration step that eliminates the usual precipitation procedures and removes much of the digested protein inducer. These procedures were followed by routine ion exchange chromatography and gel filtration. The preparation was homogeneous, as determined by gel electrophoresis and ultracentrifugal analyses. A molecular weight of approximately 30,500 was determined by amino acid analysis, gel electrophoresis, and sedimentation equilibrium. The protease has 100% activity from pH 6.0 to 10.0, is heat labile above 45 degrees C, and susceptible to autodigestion. Hydrolysis of the beta chain from insulin indicates a preferential cleavage on the carboxyl group side of neutral and aromatic amino acids.

A new extracellular RNase, designated N4, was detected in culture filtrates from Neurospora crassa and its regulation was studied. Limitation of a nutrient obtainable from RNA alone was not sufficient to cause enzyme derepression. The addition of RNA to the medium had no inductive effect, but the addition of exogenous protein caused enzyme production. With protein in the medium, N4 was derepressible for all three elemental nutrients obtainable from RNA: carbon, nitrogen, and phosphorus. Successful carbon derepression required the addition of a small amount of proteolytic activity to the cultures, as has been reported for the carbon-derepressible proteases of N. crassa. Exogenous protein affected RNase production before translation. Effects of the exogenous protein appeared similar to those previously reported for N. crassa protease induction. N4 was under the control of the nit-2 and nuc-1 gene products. nit-2 and nuc-1 mutants were unable to derepress enzyme synthesis for nitrogen and phosphorus limitation, respectively; however, these mutants responded like wild types to the other two states of derepression. Enzyme synthesis was constitutive in the preg mutant. Results indicate that the transcription of the N4 structural gene responds to multiple regulatory gene products from different regulatory circuits and that external protein affects the synthesis of classes of hydrolases other than proteases.

An extracellular acid protease was purified 1420-fold from sulfur-starved protein-induced cultures of Neurospora crassa. The enzyme was homogeneous as determined by polyacrylamide electrophoresis. The purification procedure consisted of an ultrafiltration step, cation-exchange chromatography, and affinity chromatography on Sepharose-linked pepstatin. The enzyme is homologous to aspartyl proteases that are characterized by pepstatin inhibition and trypsinogen activation. It is extremely autolytic, especially under denaturing conditions. The protease is stable between pH 3 and 7, showing optimal activity near pH 4.0 for both trypsinogen activation and hydrolysis of bovine serum albumin. The molecular weight of the enzyme was 34,500 by gel electrophoresis and gel filtration, and 34,975 by amino acid analysis.

Extracellular RNase N4 from Neurospora crassa is derepressible by limitation of any of the three nutrient elements obtainable from RNA. We have purified and characterized the enzyme from cultures grown under each of the three states of derepression. The purification procedure consisted of an ultrafiltration step, cation-exchange chromatography, and gel filtration. We found only one enzyme (N4) that hydrolyzed RNA at pH 7.5 in the presence of EDTA in culture filtrates from nitrogen-, phosphorus-, or carbon-limited cells. In all three cases, the enzymes were identical by polyacrylamide gel electrophoresis (Mr approximately 9,500) and by gel filtration (Mr approximately 10,000). There were no differences in thermal stability or pH optimum; all three cross-reacted with antibody to the nitrogen-depressed enzyme in interfacial ring and in Ouchterlony tests. Digestion of homopolyribonucleotides indicated that N4 preferentially cleaved phosphodiester bonds adjacent to guanine residues. Results indicate that the enzymes are very similar or identical and are probably products of the same gene. N4 appears to be homologous to guanine-specific RNases from other fungal sources.

Three electrophoretically distinct acid proteases appear in culture filtrates of Neurospora crassa. Like the previously investigated alkaline and neutral proteases, these enzymes require induction by an exogenous protein. But in contrast to alkaline and neutral proteases, which are synthesized and secreted in response to limitation of any one of three nutrilites (carbon, nitrogen or sulfur), extracellular elaboration of the acidic proteases is more specifically a function of the missing nutrilite. AcP, a pepstatin-inhibitable enzyme similar to other fungal carboxyl proteases, was secreted in large amounts when protein was the sole source of sulfur. Only trace amounts were secreted when nitrogen was the limiting nutrilite, and it was undetectable under carbon limitation. M-1, a chelator-sensitive protease, was secreted when nitrogen or carbon was limiting. M-2, also chelator sensitive, was present only when nitrogen or sulfur was limiting. The evidence presented suggests that the differential regulation of the acidic proteases with respect to nutrilite deprivation may not occur at the level of transcription. AcP and M-2 were partially purified from nitrogen-derepressed cultures by ultrafiltration, cation-exchange chromatography, and gel filtration. AcP has a molecular weight of 66,000, is stable from pH 3.0 to 6.0, and is optimally active toward bovine serum albumin at pH 4.0. M-2 has a molecular weight of 18,000, is stable from pH 1.6 to 5.5, and has optimal activity at pH 4.5.

A chloride-dependent transport process for glutamate has been identified in partially purified rat brain synaptosomes. This process shares many characteristics with the chloride-dependent sequestration process for glutamate in brain sonicates, which was previously thought to represent a quisqualate receptor, such as sensitivity to specific inhibitors and regulation by anions. Increasing the concentrations of chloride led to an increase in the apparent Vmax without affecting the KT. Synaptosomes preincubated with [3H]-L-glutamate exhibit an efflux of the radiolabel, which was stimulated by a substrate for the carrier in the incubating medium, indicating the bidirectional nature of the transport. The chloride-dependent transfer process is restricted to the brain, and regional and developmental profiles clearly distinguish it from the sodium-dependent high-affinity uptake process for glutamate. Nevertheless, the effects of excitotoxic lesions strongly suggest a neuronal localization of the chloride-dependent transport.

A method is described for the dissociation of rat lungs into individual viable cells. Thermolysin was perfused through the vasculature and trachea. The lung was then minced and further dissociated by washing with sequential additions of thermolysin. The results indicate this procedure to be an effective means of dispersing lung tissue into its cellular components. Standard trypsinizing procedures generally yield approximately 5 per cent of the total available lung cells whereas the thermolysin treatment increases cell yield 10-fold without significantly affecting cell viability.

Although membrane sites from brain, labelled with [3H]glutamate (Glu) under sodium-free conditions, are thought to represent excitatory receptors, certain anomalous characteristics of the kinetics of apparent binding raised the question of whether transport might contribute to this process, prompting a closer examination of it. Hyperosmolar media and low incubation temperatures (4 degrees C) both led to decreases in the apparent specific binding of [3H]glutamate to membranes from the brain of the rat in the presence of chloride. Furthermore, only 15% of the [3H]glutamate, bound at 37 degrees C, was dissociable when the membranes were then cooled to 4 degrees C. The binding of [3H]glutamate was increased in the presence of certain dipeptides such as L-phenylalanyl-L-glutamate (Phe-Glu); and the binding augmented by the presence of Phe-Glu, was also sensitive to temperature and osmolarity of the incubation buffer. Sonication of membranes in 5 mM glutamate increased the apparent binding of [3H]glutamate and abolished the stimulatory effect of Phe-Glu. These findings are consistent with the hypothesis that chloride-dependent association of [3H]glutamate with membranes from brain reflects, in part, a sequestration process, which may be driven by glutamate exchange.

Valentine, R. C. (University of Illinois, Urbana), H. Drucker, and R. S. Wolfe. Glyoxylate fermentation by Streptococcus allantoicus. J. Bacteriol. 87:241-246. 1964.-Extracts of Streptococcus allantoicus were found to degrade glyoxylate, yielding tartronic semialdehyde and CO(2). Tartronic semialdehyde was prepared chemically, and its properties were compared with the enzymatic product: reduction by sodium borohydride yielded glycerate; heating at 100 C yielded glycolaldehyde and CO(2); autoxidation yielded mesoxalic semialdehyde; periodate oxidation yielded glyoxylate and a compound presumed to be formate. Tartronic semialdehyde reductase was present in extracts of S. allantoicus and in a species of Pseudomonas grown on allantoin. A scheme for the synthesis of acetate from glyoxylate by S. allantoicus is discussed.

Cells of Neurospora crassa strain 74A, grown on sucrose for 12 h and transferred to a medium containing protein as sole carbon source, would not produce exocellular protease in significant amounts. When a filtrate from a culture induced to make protease by normal growth on a medium containing protein as principal carbon source was added to an exponential-phase culture in protein medium, exocellular protease was made in amounts similar to those made during normal induction. The material in the culture filtrate that participated in the induction process was identified as protease by its heat lability, molecular weight, and the dependence of induction rate on units of proteolytic activity added to the exponential-phase culture. Induction of the formation of exocellular protease by exponential-phase cells appears to require a protein substrate, added proteolytic activity, and protein synthesis. The protease produced by induced exponential-phase cells was as efficient in promoting induction as normally induced enzyme, whereas constitutive intracellular enzyme was only 50% as efficient. The bacterial protease thermolysin was able to induce exocellular protease at 90.7% of the rate observed with added N. crassa exocellular protease.

The cytochrome c(3) of Desulfovibrio desulfuricans and that of D. vulgaris were purified to homogeneity as judged by disc gel electrophoresis and by ultracentrifugation. Both cytochromes had an oxidation-reduction potential of -205 +/- 5 mv at pH 7.0 and showed characteristic absorption bands at 525 and 553 nm in the reduced state. The molecular weights of the two cytochromes (calculated from sedimentation and diffusion data) were similar, with values of 13,500 to 14,300 for D. desulfuricans and 13,800 to 14,700 for D. vulgaris. The two cytochromes differed in their electrophoretic properties on Geon and polyacrylamide gel electrophoresis and did not share a common precipitating antigenic determinant as judged by immunodiffusion data.