Purification and characterization of plantaricin A, a Lactobacillus plantarum bacteriocin whose activity depends on the action of two peptides

A Lactobacillus plantarum bacteriocin, plantaricin A, has been purified to homogeneity by ammonium sulphate precipitation, binding to cation exchanger and Octyl-Sepharose, and reverse-phase chromatography. The bacteriocin activity was associated with two peptides, termed alpha and beta, which were separated upon reverse-phase chromatography. Bacteriocin activity required the complementary action of both the alpha and beta peptides. From the N-terminal end, 21 and 22 amino acid residues of alpha and beta, respectively, were sequenced. Further attempts at sequencing revealed no additional amino acid residues, suggesting that either the C terminus had been reached or that modifications in the next amino acid residue blocked the sequencing reaction. Judging from their amino acid sequence, alpha and beta may be encoded by the same gene, since alpha appeared to be a truncated form of beta. Alanine, the first amino acid residue at the N-terminal end of beta was not present at this position in alpha. Otherwise the sequences of alpha and beta appeared to be identical. The calculated molecular masses of the sequenced part of alpha and beta were 2426 and 2497 Da, respectively. The molecular masses of alpha and beta as determined by mass spectroscopy were 2687 +/- 30 and 2758 +/- 30 Da, respectively, indicating that (i) the only difference between alpha and beta was the presence of the N-terminal alanine residue in beta, and that (ii) in addition to the sequenced residues, two to three unidentified amino acid residues are present at the C-terminal ends of the alpha and beta peptides.(ABSTRACT TRUNCATED AT 250 WORDS)     

Transporters for cationic amino acids in animal cells: discovery, structure, and function

The structure and function of the four cationic amino acid transporters identified in animal cells are discussed. The systems differ in specificity, cation dependence, and physiological role. One of them, system y+, is selective for cationic amino acids, whereas the others (B[0,+], b[0,+], and y+ L) also accept neutral amino acids. In recent years, cDNA clones related to these activities have been isolated. Thus two families of proteins have been identified: 1) CAT or cationic amino acid transporters and 2) BAT or broad-scope transport proteins. In the CAT family, three genes encode for four different isoforms [CAT-1, CAT-2A, CAT-2(B) and CAT-3]; these are approximately 70-kDa proteins with multiple transmembrane segments (12-14), and despite their structural similarity, they differ in tissue distribution, kinetics, and regulatory properties. System y+ is the expression of the activity of CAT transporters. The BAT family includes two isoforms (rBAT and 4F2hc); these are 59- to 78-kDa proteins with one to four membrane-spanning segments, and it has been proposed that these proteins act as transport regulators. The expression of rBAT and 4F2hc induces system b[0,+] and system y+ L activity in Xenopus laevis oocytes, respectively. The roles of these transporters in nutrition, endocrinology, nitric oxide biology, and immunology, as well as in the genetic diseases cystinuria and lysinuric protein intolerance, are reviewed. Experimental strategies, which can be used in the kinetic characterization of coexpressed transporters, are also discussed.     

Lepidopteran peptides of the allatostatin superfamily

Peptides of the allatostatin superfamily with the C-terminal amino acid sequence -YXFGL-NH2 have been isolated and identified from the lepidopterans, the codling moth, Cydia pomonella (Tortricidae) and the bollworm, Helicoverpa armigera (Noctuidae). The peptides, designated cydiastatins and helicostatins respectively, were monitored during purification with radioimmunoassays based on the callatostatins of the blowfly Calliphora vomitoria. The eight peptides from each of the two species appear to form an homologous series with four identical and three that differ by a single amino acid. This study demonstrates the ubiquitous nature of this family of peptides in insects.     

Necrosis versus apoptosis as the mechanism of target cell death induced by Entamoeba histolytica

The human pathogen Entamoeba histolytica is known to kill a variety of host cells, including leukocytes. Using human myeloid cells as targets, we studied whether cytotoxicity of amoebic trophozoites in vitro is equivalent to the induction of apoptosis or whether these target cells die via necrosis. Based upon morphological criteria, incubation of target cells with amoebae resulted in necrosis, with cell swelling, rupture of plasma membrane, and release of cell contents including nucleic acids being detected by light and transmission electron microscopy. On the other hand, the characteristic features of apoptosis such as cell shrinking, surface blebbing, and chromatin condensation were not observed. Moreover, internucleosomal fragmentation of genomic DNA within target cells as a characteristic feature of apoptotic cell death did not occur as judged by the terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling technique in combination with flow cytometry. Consistently, cleavage of DNA was detectable upon agarose gel electrophoresis only after a substantial part of the target cell population had already been lysed. We also analyzed the mechanism of cell death induced by amoebapores, pore-forming peptides and primary candidate molecules for mediating the cytolytic activity of E. histolytica. At a time point at which the majority of target cells showed membrane injury upon incubation with purified amoebapores, no DNA degradation was detectable in the victim cells. The data suggest that the target cells used in our study undergo necrosis rather than apoptosis when they are killed by viable trophozoites as well as by isolated amoebapores.     

History and heritage: development of radiation oncology in China

The amount of polygalacturonase-inhibiting protein (PGIP) was 14 times higher in bean pods than in etiolated hypocotyls. The PGIP was extracted from bean pods and partially purified by chromatography on columns of S-Sepharose. DEAE-Sephadex A-50, and Sephadex G-75. Further purification by ion-exchange chromatography on a Mono Q column separated two isoforms of the inhibitor. The two PGIPs were similar in most properties but differed slightly in pI values. They also differed in one residue of the N-terminal amino acid sequences. Both bean pod PGIPs differed in two and possibly three residues of the deduced N-terminal amino acid sequence for hypocotyl PGIP. Small alterations in the structure of PGIP may represent a strategy in bean plants for resistance to a variety of pathogens.     

Expression of CYP71B7, a cytochrome P450 expressed sequence Tag from Arabidopsis thaliana

c-Yes was purified 322-fold from a rat liver plasma membrane fraction to a single 60-kDa band on SDS-PAGE. The purified protein contained essentially no phosphotyrosine residues and was autophosphorylated with Mg2+. ATP exclusively at tyrosine residues with a concomitant increase in the protein-tyrosine kinase activity. The autophosphorylated c-Yes was extensively digested by trypsin and the resultant two major phosphopeptides, peptides I and II, were purified by HPLC on a reversed-phase C-18 column. The amino acid sequence of peptide I was determined to be LIEDNEYTAR, which is identical with the sequence from Leu-418 through Arg-427 of mouse c-Yes, indicating that one of the autophosphorylation sites corresponds to Tyr-424 of the mouse c-Yes. After partial determination of the N-terminal sequence of 10 amino acid residues of peptide II, the 230 bp sequence of rat cDNA that encodes the N-terminal 76 amino acid residues of c-Yes covering peptide II, was determined. From the predicted amino acid sequence, the sequence of peptide II was assumed to be from Tyr-16 through Lys-46, YTPENPTEPVNTSAGHYGVEHATAATTSSTK. The purified c-Yes phosphorylated the tyrosine residue of synthetic peptides covering Tyr-32 and its surrounding sequence but did not phosphorylate peptides covering Tyr-16 and its surrounding sequence, suggesting that the other autophosphorylation site is Tyr-32.     

A novel isoform of the mitochondrial outer membrane protein VDAC3 via alternative splicing of a 3-base exon. Functional characteristics and subcellular localization

Voltage-dependent anion channels (VDACs) are pore-forming proteins found in the outer mitochondrial membrane of all eucaryotes. VDACs are the major pathway for metabolites through the outer mitochondrial membrane and, in mammals, bind several cytosolic carbohydrate kinases. Whereas yeast contain a single VDAC (YVDAC), to date three isoforms have been described in the mouse that constitute a gene family. We have observed an additional isoform of VDAC3 that appears to be generated via the tissue-specific alternative splicing of a 3-base exon (ATG). The exon is predicted to introduce a methionine 39 amino acids downstream of the amino terminus of the polypeptide. Between exons 3 and 4 is an intronic sequence that potentially encodes the exon, with flanking splice enhancer elements. Expression of this alternative form in the mouse is limited to brain, heart, and skeletal muscle. Complementation of YVDAC-deficient yeast by the two isoforms and with other sequence variants of VDAC3 suggests this residue is an important modulator of VDAC3 function. In transfected mammalian cells both isoforms localize to mitochondria. A similar variant is present in humans.     

Psychological aspects of systemic lupus erythematosus: cognitive function, mood, and self-report

Previous studies have shown that Tetrahymena citrate synthase and the Tetrahymena 14-nm filament protein are encoded by a single gene and translated from one species of mRNA, and that they are identical in terms of molecular weight, antigenicity, and some enzymatic properties. In this study, using two-dimensional gel electrophoresis, we demonstrated that the citrate synthase comprised pI 7.7 and 8.0 isoforms, while the 14-nm filament protein comprised three isoforms with isoelectric points of 7.7, 8.0, and 8.4. The amino acid sequences of the NH2-terminal portions of all isoforms were identical and the peptide maps with V8 protease were almost the same. In addition, when the citrate synthase activity of each isoform was measured after separation by non-urea isoelectric focusing without denaturing treatment, the pI 7.7 and/or pI 8.0 isoforms exhibited the citrate synthase activity, but the pI 8.4 isoform only found for the 14-nm filament protein did not possess this activity. These results suggest that the polymorphism of these isoforms is caused by some posttranslational modifications, and that it may have resulted in the different compartmentalization and functions of Tetrahymena citrate synthase and the 14-nm filament protein.     

Multiple isoforms of Arabidopsis casein kinase I combine conserved catalytic domains with variable carboxyl-terminal extensions

Three cDNA clones encoding isoforms of casein kinase I (CKI) were isolated from Arabidopsis thaliana. One full-length clone, designated CKI1, contained an open reading frame of 1371 bp encoding a protein of 51,949 D with an isoelectric point of 9.7. In addition to the highly conserved catalytic domain (of about 300 amino acids), the Arabidopsis CKI isoforms contain 150 to 180 amino acid carboxyl-terminal extensions, which show among themselves a lower level of sequence conservation. These extensions do not show any sequence similarity to nonplant CKI isoforms, such as rat testis CKI delta, which is their closest isolated homolog, or to yeast CKI isoforms. Three additional isoforms of Arabidopsis CKI were found in the data bases of expressed sequence tags and/or were isolated serendipitously in nonspecific screening procedures by others. One of them also shows a carboxyl-terminal extension, but of only 80 amino acids. Casein kinase activity was detected in the soluble fraction of Escherichia coli strains expressing the CKI1 protein. This activity showed the crucial properties of CKI, including the ability to phosphorylate the D4 peptide, a specific substrate of CKI, and inhibition by N-(2-aminoethyl)-5-chloroisoquinoline-8-sulfonamide, a specific CKI inhibitor. Like several recombinant CKI isoforms from yeast, CKI1 was able to phosphorylate tyrosine-containing acidic polymers.     

Purification and characterization of a kinin-releasing and fibrinogen-clotting serine proteinase (KN-BJ) from the venom of Bothrops jararaca, and molecular cloning and sequence analysis of its cDNA

Two forms of a proteinase, KN-BJ 1 and 2, were purified to homogeneity from the venom of Bothrops jararaca. In SDS/PAGE reduced KN-BJ 1 and 2 migrated as single bands with molecular masses of 38 kDa and 39 kDa. The two enzymes have similar N-terminal amino acid sequences and specific activities on synthetic chromogenic substrates, and both release bradykinin from bovine low-molecular-mass kininogen. KN-BJ 1 and KN-BJ 2 clot fibrinogen with specific activities of 245 NIH U/mg and 219 NIH U/mg, releasing only fibrinopeptide A. The amidolytic, kinin-releasing and coagulant activities are inhibited by phenylmethylsulfonyl fluoride, demonstrating that KN-BJ is a serine proteinase. Benzamidine derivatives, which are competitive inhibitors of trypsin-like proteinases, also inhibited the amidolytic activity of KN-BJ. A cDNA clone (HS104, 2.2 kb) has been isolated from a cDNA library of B. jararaca venom glands with an ORF of 771 bp. The deduced amino acid sequence contains segments that are identical to the sequences of the N-terminus and three tryptic peptides of KN-BJ 2. Therefore, the cDNA is believed to represent the gene of KN-BJ 2. The deduced amino acid sequence indicates that KN-BJ 2 is synthesized as a prezymogen of 257 amino acids with a putative signal peptide of 18 amino acids and an activating peptide of six amino acid residues. The sequence of 233 amino acids representing the mature enzyme exhibits high similarity to sequences of serine proteinases isolated from crotalid venoms.     

The amino acid sequence of topi pancreatic ribonuclease

Pancreatic tissue from topi (Damaliscus korrigum) contains three ribonuclease components in a ratio of 8:22:70. Two components are glycosidated, whereas the third one does not contain carbohydrate. The amino acid sequence of topi ribonuclease A was deduced from a tryptic digest of the performic acid-oxidized protein. Peptides were positioned by homology with other bovid ribonucleases. Only peptides that differed in amino acid composition from the corresponding peptides of bovine ribonuclease were sequenced. The evidence obtained for the sequence of residues 67-73 is incomplete. Among the bovid ribonucleases (cow, bison, eland, sheep, goat and gnu), topi ribonuclease shows the closest resemblance with sheep and goat ribonucleases; except that the glutamic acid residue at position 103 in the ribonucleases from sheep and goat is substituted by a lysine residue in topi. Topi ribonucleases A and B differ only in the presence of carbohydrate attached to asparagine 34.     

Block of brain sodium channels by peptide mimetics of the isoleucine, phenylalanine, and methionine (IFM) motif from the inactivation gate

Inactivation of sodium channels is thought to be mediated by an inactivation gate formed by the intracellular loop connecting domains III and IV. A hydrophobic motif containing the amino acid sequence isoleucine, phenylalanine, and methionine (IFM) is required for the inactivation process. Peptides containing the IFM motif, when applied to the cytoplasmic side of these channels, produce two types of block: fast block, which resembles the inactivation process, and slow, use-dependent block stimulated by strong depolarizing pulses. Fast block by the peptide ac-KIFMK-NH2, measured on sodium channels whose inactivation was slowed by the alpha-scorpion toxin from Leiurus quinquestriatus (LqTx), was reversed with a time constant of 0.9 ms upon repolarization. In contrast, control and LqTx-modified sodium channels were slower to recover from use-dependent block. For fast block, linear peptides of three to six amino acid residues containing the IFM motif and two positive charges were more effective than peptides with one positive charge, whereas uncharged IFM peptides were ineffective. Substitution of the IFM residues in the peptide ac-KIFMK-NH2 with smaller, less hydrophobic residues prevented fast block. The positively charged tripeptide IFM-NH2 did not cause appreciable fast block, but the divalent cation IFM-NH(CH2)2NH2 was as effective as the pentapeptide ac-KIFMK-NH2. The constrained peptide cyclic KIFMK containing two positive charges did not cause fast block. These results indicate that the position of the positive charges is unimportant, but flexibility or conformation of the IFM-containing peptide is important to allow fast block. Slow, use-dependent block was observed with IFM-containing peptides of three to six residues having one or two positive charges, but not with dipeptides or phenylalanine-amide. In contrast to its lack of fast block, cyclic KIFMK was an effective use-dependent blocker. Substitutions of amino acid residues in the tripeptide IFM-NH2 showed that large hydrophobic residues are preferred in all three positions for slow, use-dependent block. However, substitution of the large hydrophobic residue diphenylalanine or the constrained residues phenylglycine or tetrahydroisoquinoline for phe decreased potency, suggesting that this phe residue must be able to enter a restricted hydrophobic pocket during the binding of IFM peptides. Together, the results on fast block and slow, use-dependent block indicate that IFM peptides form two distinct complexes of different stability and structural specificity with receptor site(s) on the sodium channel. It is proposed that fast block represents binding of these peptides to the inactivation gate receptor, while slow, use-dependent block represents deeper binding of the IFM peptides in the pore.     

Horsetail (Equisetum telmateia) ferredoxins I and II. Amino acid sequences

Two ferredoxins were isolated from horsetail (Equisetum telmateia) and their amino acid sequences were determined by use of a sequence analyzer in combination with carboxypeptidase digestion and manual Edman degradation of tryptic peptides of carboxymethyl-ferredoxins. Ferredoxins I and II each had only four cysteine residues in a total of 95 and 93 residues, respectively. The amino-terminal residues of both ferredoxins were heterogeneous, but alanine was concluded to be their genuine terminal residue. The comparison of these isozymelike molecules showed 29 differences in amino acid residues with three inverted replacements. One gap was inserted in ferredoxin II at position 32 to align the ferredoxins with greatest homology. Despite the many differences in amino acid residues there was no difference in net charges of the two ferredoxins.     

Small peptides as potent mimetics of the protein hormone erythropoietin

Random phage display peptide libraries and affinity selective methods were used to isolate small peptides that bind to and activate the receptor for the cytokine erythropoietin (EPO). In a panel of in vitro biological assays, the peptides act as full agonists and they can also stimulate erythropoiesis in mice. These agonists are represented by a 14- amino acid disulfide-bonded, cyclic peptide with the minimum consensus sequence YXCXXGPXTWXCXP, where X represents positions allowing occupation by several amino acids. The amino acid sequences of these peptides are not found in the primary sequence of EPO. The signaling pathways activated by these peptides appear to be identical to those induced by the natural ligand. This discovery may form the basis for the design of small molecule mimetics of EPO.     

Preparation of peptides which mimic glycosphingolipids by using phage peptide library and their modulation on beta-galactosidase activity

We describe the use of a phage-displayed random pentadecamer peptide library for searching glycosphingolipid mimicking peptides. Two phage clones (AD-1 and AD-2) were selected by biopanning using monoclonal antibody AD117m, directed to lactotetraosylceramide (Lc4Cer). The amino acid sequences of the selected clones showed high homology (VPPXFXXXY) in 9-mer. Three phage clones were selected by using monoclonal antibody H11, directed to neolactotetraosylceramide (nLc4Cer), the linkage isomer of Lc4Cer, and the displayed amino acid sequences were compared. One of these peptides showed the same amino acid sequence as that of AD-2 except for one amino acid substitution. Pentadecamer, 9-mer and point mutated 9-mer peptides were synthesized on the basis of the displayed amino acid sequences. Binding activity of the peptides to the monoclonal antibodies or Ricinus communis lectin showed that 9-mer peptides are enough to mimic the epitope carbohydrate structure. Furthermore, six of the synthesized peptides inhibited Jack bean beta-galactosidase activity towards nLc4Cer at a high concentration of the enzyme, whereas at lower enzyme concentrations some peptides showed potent activation of the enzyme activity. This is the first report of carbohydrate mimicking peptides which modulate glycosidase activity.     

Identification and characterization of heparin binding regions of the Fim2 subunit of Bordetella pertussis

Bordetella pertussis fimbriae bind to sulfated sugars such as heparin through the major subunit Fim2. The Fim2 subunit contains two regions, designated H1 and H2, which show sequence similarity with heparin binding regions of fibronectin, and the role of these regions in heparin binding was investigated with maltose binding protein (MBP)-Fim2 fusion proteins. Deletion derivatives of MBP-Fim2 showed that both regions are important for binding to heparin. The role of H2 in heparin binding was confirmed by site-directed mutagenesis in which basic amino acids were replaced by alanine. These studies revealed that Lys-186 and Lys-187 are important for heparin binding of MBP-Fim2, whereas Arg-179 is not required. Peptides derived from H1 and H2 (pepH1 and pepH2) also showed heparin binding activity. Using a series of peptides, in each of which a different basic amino acid was substituted for alanine, we demonstrated that the structural requirements for heparin binding differ significantly among pepH1 and pepH2 peptides. A Pepscan analysis of Fim2 revealed regions outside H1 and H2 which bind heparin and showed that not only basic amino acids but also tyrosines may be important for binding to sulfated sugars. A comparison of the heparin binding regions of Fim2 with homologous regions of Fim3 and FimX, two closely related but antigenically distinct fimbrial subunits, showed that basic amino acids and tyrosines are generally conserved. The major heparin binding regions identified in Fim2 are part of epitopes recognized by human antibodies, suggesting that the heparin binding regions are exposed at the fimbrial surface and are immunodominant. Since B. pertussis fimbriae show weak serological cross-reactivity, the differences in primary structure in the heparin binding regions of Fim2, Fim3, and FimX may affect antibody binding but not heparin binding, allowing the bacteria to evade antibody-mediated immunity by switching the fimbrial gene expressed.     

Solution structure of P01, a natural scorpion peptide structurally analogous to scorpion toxins specific for apamin-sensitive potassium channel

The venom of the North African scorpion Androctonus mauretanicus mauretanicus possesses numerous highly active neurotoxins that specifically bind to various ion channels. One of these, P05, has been found to bind specifically to calcium-activated potassium channels and also to compete with apamin, a toxin extracted from bee venom. Besides the highly potent ones, several of these peptides (including that of P01) have been purified and been found to possess only a very weak, although significant, activity in competition with apamin. The amino acid sequence of P01 shows that it is shorter than P05 by two residues. This deletion occurs within an alpha-helix stretch (residues 5-12). This alpha-helix has been shown to be involved in the interaction of P05 with its receptor via two arginine residues. These two arginines are absent in the P01 sequence. Furthermore, a proline residue in position 7 of the P01 sequence may act as an alpha-helix breaker. We have determined the solution structure of P01 by conventional two-dimensional 1H nuclear magnetic resonance and show that 1) the proline residue does not disturb the alpha-helix running from residues 5 to 12; 2) the two arginines are topologically replaced by two acidic residues, which explains the drop in activity; 3) the residual binding activity may be due to the histidine residue in position 9; and 4) the overall secondary structure is conserved, i.e., an alpha-helix running from residues 5 to 12, two antiparallel stretches of beta-sheet (residues 15-20 and 23-27) connected by a type I' beta-turn, and three disulfide bridges connecting the alpha-helix to the beta-sheet.     

PhLPs and PhLOPs in the phosducin family of G beta gamma binding proteins

In this study, we identify new isoforms of the retinal phosducin and investigate the expression of the phosducin family, showing that an isoform, PhLP1, has sequence homology with Phd and Gbeta gamma binding capability, whereas two isoforms (phosducin-like orphan proteins, PhLOPs) share sequence homology with Phd but fail to bind Gbeta gamma. Original identification of PhLP1 and the PhLOPs was from a human retina cDNA library, using a PCR product for library hybridization screening that contained a predicted functional epitope domain. The screen identified Phd and three related, but distinct, recombinants (PhLP1, PhLOP1, and PhLOP2). By RT-PCR, all isoforms are expressed in either retina or forskolin-stimulated Y79 retinoblastoma cells; however, the new isoforms are below the level of detection on Northern blot analysis. The predicted amino acid translation of each homologue revealed major differences, arising from either splice variants or gene duplication of Phd. To test the functional interaction of all phosducin isoforms with Gbeta gamma in vitro, a glutathione S-transferase (GST) fusion protein was developed for each member. Biochemical interaction with purified retinal transducin Gbeta gamma was verified for GST-Phd and demonstrated for GST-PhLP1; however, neither GST-PhLOP1 nor GST-PhLOP2 bound Gbeta gamma. Comparable results were observed when the GST-phosducin fusion proteins selectively sequestered Gbeta gammas from retinal extracts or when functional Gbeta gamma interactions were assessed using surface plasmon resonance technology. Phosducin and its isoforms are widely distributed in body tissues where they may participate in signal transduction pathways. Phd and PhLP1 possess an 11-amino acid conserved epitope domain (TGPKGVINDWR) that controls the high-affinity binding of Gbeta gamma; these isoforms are implicated in the G-protein signaling pathway. The phosducin-like orphan proteins (PhLOPs) fail to bind Gbeta gamma, suggesting that the PhLOP isoforms may participate in still unidentified signaling pathways.     

Demonstration of a cell-specific isomerization of invertebrate neuropeptides

Neurohemal organs of the lobster Homarus americanus contain isoforms of the crustacean hyperglycemic hormone, which differ by the third amino acid (phenylalanyl) residue that is either in the L- or in the D-configuration. Polyclonal antisera have been raised in rabbit against synthetic octapeptides with the sequence corresponding to the N-terminal part of the L- or D-phenylalanine-containing isoforms. Their specificity was shown by immunoassays, indicating that they discriminate the isoforms of the lobster hyperglycemic neuropeptides. It was demonstrated that the two major forms of the crayfish Orconectes limosus hyperglycemic hormone also correspond to peptide isomers containing the L- or D-phenylalanyl residue. The cellular distribution of the isoforms among the neurosecreting cells of the major neuroendocrine complex in lobster and crayfish has been studied by immunohistochemistry. Every hyperglycemic hormone-containing cell was labelled with the anti-L antisera while only some of them were visualized with the anti-D antisera. These results constitute the first observation of peptide isomerization at the cellular level and suggest that the isomerization process occurs in specialized neuroendocrine cells.