Sites of phosphorylation and mutation in regulatory subunit of cyclic AMP-dependent protein kinase from S49 mouse lymphoma cells: mapping to structural domains

A novel peptide mapping approach has been used to map sites of charge modification to major structural domains of regulatory subunit (R) of type I cAMP-dependent protein kinase from S49 mouse lymphoma cells. Proteolytic fragments of crude, radiolabeled R were purified by cAMP affinity chromatography and displayed by two-dimensional polyacrylamide gel electrophoresis. [35S]methionine-labeled peptides containing sites of mutation or phosphorylation exhibited charge heterogeneity attributable to the modification. Phosphate-containing fragments were also labeled with [32P]orthophosphate to confirm their phosphorylation. Major fragments from [35S]methionine-labeled S49 cell R corresponded in size to carboxyterminal cAMP-binding fragments reported from proteolysis of purified type I Rs from various mammalian species; additional fragments were also visualized. End-specific markers in Rs from some mutant S49 sublines confirmed that cAMP-binding fragments extended to the carboxyterminus of R. Aminoterminal endpoints of fragments could be deduced, therefore, from peptide molecular weights. Clustering of proteolytic cleavage sites within the "hinge-region" separating aminoterminal and carboxyterminal domains of R permitted high resolution mapping in this region: the endogenous phosphate and a "phenotypically-silent" electrophoretic marker mutation fell within a 2.5-kdalton interval at its aminoterminal end. On the other hand, Ka mutations that increase the apparent constant for activation of kinase by cAMP mapped within the large cAMP-binding region of R. A map of charge density distribution within the hinge-region of R was constructed to facilitate structural comparisons between Rs from S49 cells and from other mammalian sources.     

Altered regulation of cyclic AMP-dependent protein kinase in a mouse lymphoma cell line.

The ability of cyclic AMP to inhibit growth, cause cytolysis and induce synthesis of cyclic AMP-phosphodiesterase in S49.1 mouse lymphoma cells is deficient in cells selected on the basis of their resistance to killing by 2 mM dibutyryl cyclic AMP. The properties of the cyclic AMP-dependent protein kinase (ATP:protein phosphotransferase, EC 2.7.1.37) in the cyclic AMP-sensitive (S) and cyclic AMP-resistant (R) lymphoma cells were comparatively studied. The cyclic AMP-dependent protein kinase activity or R cells cytosol exhibits an apparent Ka for activation by cyclic AMP 100-fold greater than that of the enzyme from the parental S cells. The free regulatory and catalytic subunits from both S and R kinase are thermolabile, when associated in the holoenzyme the two subunits are more stable to heat inactivation in R kinase than in S kinase. The increased heat stability of R kinase is observed however only for the enzyme in which the catalytic and cyclic AMP-binding activities are expressed at high cyclic AMP concentrations (10(-5)--10(-4) M), the activities expressed at low cyclic AMP concentrations (10(-9)--10(-6) M) being thermolabile. The regulatory subunit of S kinase can be stabilized against heat inactivation by cyclic AMP binding both at 2-10(-7) and 10(-5) M cyclic AMP concentrations. In contrast, the regulatory subunit-cyclic AMP complex from R kinase is stable to heat inactivation only when formed in the presence of high cyclic AMP concentrations (10(-5)M). The findings indicate that the transition from a cyclic AMP-sensitive to a cyclic AMP-resistant lymphoma cell phenotype is related to a structural alteration in the regulatory subunit of the cyclic AMP-dependent protein kinase which has affected the protein's affinity for cyclic AMP and its interaction with the catalytic subunit.     

Isolation of multiple classes of mutants of CHO cells resistant to cyclic AMP

From mutagenized Chinese hamster ovary (CHO) cells we have isolated, in a single step, 11 independent mutants resistant to the growth-inhibitory effects of 8-Br-cyclic AMP, cholera toxin, and methylisobutylxanthine. Two major classes and several subclasses of mutants were obtained. Mutants from all classes have a normal doubling time. None of the mutants respond to cyclic AMP treatment with increased flattening and elongation as do the parental cells. Members of the first class have an altered protein kinase activity which has either an increased Ka for cyclic AMP or an absent response to cyclic AMP. Most of those mutations which result in a protein kinase with increased Ka for cyclic AMP (6/11) are dominant in somatic cell hybrids. Those mutations which result in a protein kinase with little or no response to cyclic AMP (3/11) are recessive. Members of the second major class (2/11) have normal levels of basal and cyclic AMP-dependent protein kinase activity. One is recessive and one is dominant by genetic tests. The basis for the defect in this second class of mutants has not been determined.     

Mapping endpoints of partial proteolysis fragments from regulatory subunit of type I cyclic AMP-dependent protein kinase

Methods for mapping endpoints of partial proteolysis fragments from regulatory subunit of type I cyclic AMP-dependent protein kinase are described with a view to using such data for fine-structure analysis of mutations and/or modifications affecting the protein's electrostatic charge. Peptides generated from [35S]methionine-labeled regulatory subunit were separated by high-resolution two-dimensional gel electrophoresis. Sites of papain cleavage in denatured regulatory subunit were deduced from the kinetics of the appearance, molecular weights, and relative isoelectric points of the fragments produced. These sites and sites of chymotrypsin digestion in the native protein were confirmed by studying peptide overlaps. Carboxy-terminal peptides were identified both by overlaps with cyclic AMP-binding chymotryptic fragments and by their preferential labeling during polysome runoff mediated by pactamycin, an inhibitor of protein initiation. Since peptides containing modifications or mutations that alter protein charge can be identified by shifts in first-dimension isoelectric focusing gel positions, knowledge of fragment endpoints will permit rapid mapping of sites of such alterations by two-dimensional gel analysis of partial proteolytic digests. Such a mapping procedure is inexpensive, can be applied to partially purified proteins or to proteins eluted from polyacrylamide gels, requires only nanogram amounts of the protein of interest, and does not require sequence data to determine relative positions of peptides. Therefore, it provides an attractive alternative to more classical peptide analysis for studying point mutations in cellular proteins of low abundance.     

The characterization and cloning of a gluconate (gnt) operon of Bacillus subtilis

The enzymes involved in gluconate utilization in Bacillus subtilis seemed to be gluconate permease and gluconate kinase. Several mutants unable to grow on gluconate were isolated. The mutations they harboured (gnt) were clustered between iol-6 and fdp-74 on the B. subtilis chromosome (a tentative map order of gnt-10, gnt-4, gnt-26, gnt-23 and gnt-9 was obtained). The gnt-10 mutation seemed to be located within the structural gene of the kinase, and the gnt-23 and gnt-26 mutations seemed to be within that of the permease. An EcoRI fragment (4.5 MDal) containing an intact gluconate (gnt) operon consisting of these two structural genes was cloned in phage phi 105 by prophage transformation and was mapped physically. The physical location of the mutations coincided with their order on the genetic map. The HindIII-A fragment (2.4 MDal), which corrects all the gnt mutations, was subcloned in plasmid pC194. The fragment contained the structural genes for the gluconate permease and kinase, but not the regulatory region of the gluconate operon.     

Long-range genomic map of the Duchenne muscular dystrophy (DMD) gene: Isolation and use of J66 (DXS268), a distal intragenic marker

By cloning the endpoints of a DMD-associated deletion, we have "jumped" 1100 kb from pERT87-1 (DSX164) to a new locus designated J66 (DXS268), mapping distally within the Duchenne muscular dystrophy (DMD) gene. Both J66 and JBir are mapped by field-inversion gel electrophoresis and detect abnormal SfiI fragments in DMD patients and distal DMD-associated X; autosome translocations. Our long-range map extends the physical map of the DMD gene from 800 to 2000 kb (2 Mb) and increases the mapped portion of Xp21 to approximately 8 Mb. The position of the glycerol kinase gene and the adrenal hypoplasia locus are further confined to the region between J66 and the nearest distal probe L1-4. This region spans at least 1.5 Mb. The multiallelic J66 polymorphism has immediate application in the diagnosis of DMD and generally appears to be distal to DMD mutations.     

Fine-structure mapping of herpes simplex virus type 1 temperature-sensitive mutations within the short repeat region of the genome.

Cloned herpes simplex virus type 1 (HSV-1) DNA fragments were used to fine-structure map the temperature-sensitive (ts) lesions from four mutants, ts T, D, c75, and K, by marker rescue. These mutants all overproduced immediate-early viral polypeptides at the nonpermissive temperature. Although one of these viruses, ts K, gave a more restricted infected-cell polypeptide profile under these conditions than the other three, no complementation was detected between pairwise crosses of these mutants in the yield test. Recombination, however, was obtained between all mutant pairs except ts T and D. In physical mapping experiments, ts+ virus was recovered from cells coinfected with DNA of ts T, D, or c75 and BamHI fragment k from wild-type strain 17 HSV-1 DNA cloned in pAT153, whereas ts K was rescued by cloned HSV-1 BamHI-y. Both of these cloned DNA fragments contained sequences from the short repeat region of the HSV-1 genome. The ts mutations were more precisely mapped by marker rescue, using restriction enzyme fragments within BamHI-k and -y from cloned DNA. The smallest fragment able to rescue a mutant was 320 base pairs long. The order of the four mutations derived from these studies was consistent with the assignment by genetic recombination. All four lesions mapped within the coding sequences of the immediate-early polypeptide Vmw IE 175 (ICP4) which lie outside the "a" sequence. The results showed that mutations in different regions of the gene encoding Vmw IE 175 could produce similar phenotype effects at the nonpermissive temperature.     

Differential effects of histamine, vasoactive intestinal polypeptide, prostaglandin E2 and somatostatin on cyclic AMP-dependent protein kinase activation in gastric glands isolated from the guinea pig fundus and antrum

Histamine, vasoactive intestinal polypeptide (VIP), secretin and prostaglandin E2 (PGE2) stimulate cyclic AMP-dependent protein kinase activity in gastric glands isolated from the guinea pig fundus and antrum. The effects are observed in the absence of any cyclic AMP phosphodiesterase inhibitor and maximal stimulation of the protein kinases occurs within 0.5 min of incubation at 20 degrees C. As shown by dose-response studies, VIP is equally potent in the antrum as in the fundus (identical values of the activation constant are found in both types of gland, Ka = 2.5 . 10(-9) M); a similar situation occurs for PGE2 action (but with Ka = 2.0 . 10(-8) M), whereas the potency of histamine is higher in the fundus (ka = 8.0 . 10(-6)M) than in the antrum (Ka = 5.0 . 10(-5) M). Secretin also increases the protein kinase activity ratio but with a 1000 times lower potency than VIP. In fundic glands, histamine (10(-3) M) is the activator of by far the greatest efficacy (increasing protein kinase activity at 4 times of the basal value) as compared with the effect obtained with 10(-6) M PGE2 (2.7 times) and 10(-7) M VIP (1.4 times). In contrast, VIP has greater efficacy (2.3 times) than histamine (2.1 times) in antral glands, whereas PGE2 is equally active in the two parts of the gastric mucosa. In addition, somatostatin (10(-6) M) inhibits partially (30%) and specifically the protein kinase activation stimulated by histamine, whereas it has no effect on VIP- and PGE2-induced activation. The results are consistent with increased cyclic AMP levels in response to these effectors in this system. A physiological role of histamine on acid-secreting parietal cells, of VIP on nonparietal cells and of PGE2 on both cell types, mediated by the cyclic AMP/protein kinase system is proposed.     

Cyclic AMP-dependent phosphorylation of filamin in mammalian smooth muscle.

Filamin is a high molecular weight actin-binding protein found in large quantities in smooth muscle and other non-muscle cells. We have studied the phosphorylation of filamin in a mammalian smooth muscle, the guinea pig vas deferens. Intact vas deferens incorporated [32P]orthophosphate into filamin. Incubation of particulate fractions of vas deferens with [gamma-32P]ATP resulted in 32P-labeling of filamin. Cyclic AMP stimulated this phosphorylation, whereas cyclic GMP and Ca2+ had no effect. Purified vas deferens filamin can be phosphorylated by purified cyclic AMP-dependent protein kinase. We have compared cyclic AMP and cyclic GMP effects on phosphorylation in smooth muscle. Cyclic GMP stimulated phosphorylation of two particulate proteins, G-I (Mr = 130,000) a protein previously described by Casnellie, J. E., and Greengard, P. (1974) Proc. Natl. Acad, Sci. U.S.A. 71, 1891-1895 and G-III (Mr = 240,000). Both proteins and the kinase responsible for their phosphorylation appear to be membrane-bound. Phosphorylation of both proteins is stimulated by cyclic GMP (Ka = 3 x 10(-8) M), cyclic AMP (Ka = 3 x 10(-7) M), and to a lesser degree by Ca2+. In contrast, filamin phosphorylation is due to a soluble kinase stimulated only by cyclic AMP (Ka = 3 x 10(-7) M) and not by cyclic GMP or Ca2+.     

Photoaffinity labeling of three renal cyclic 3',5'-adenosine monophosphate-binding proteins.

Evidence is presented for the presence of multiple cyclic AMP binding components in the plasma membrane and cytosol fractions of porcine renal cortex and medulla. N6-(Ethyl-2-diazomalonyl)-3',5'-adenosine monophosphate, a photoaffinity label for cyclic AMP binding sites, exhibits non-covalent binding characteristics similar to cyclic AMP in membrane and soluble fractions. Binding data for either compound to the plasma membrane fraction yields biphasic Scatchard plots while triphasic plots are obtained with the dialyzed cytosol. When covalently labeled fractions are separated on SDS-polyacrylamide gel electrophoresis, the cyclic AMP photoaffinity label is found on 49 000 and 130 000 dalton components in each kidney fraction. DEAE-cellulose and gel filtration chromatography of the labeled cortical cytosol fraction establishes that the three components suggested by the binding data correspond to two 49 000 dalton species and a 130 000 component. The 49 000 species have higher affinities for cyclic AMP than the 130 000 component (Ka(1) = 2.0 . 10(9), Ka(2) = 1.7 . 10(8), Ka(3) = 1.0 . 10(7)). The 49 000 components are associated with protein kinase activity while the 130 000 component does not exhibit protein kinase, adenosine deaminase, or cyclic nucleotide phosphodiesterase activity. Immunologic results and effects of phosphorylation and cyclic GMP on cyclic AMP binding further suggest that the 49 000 components are regulatory subunits of cyclic AMP-dependent protein kinases. Cyclic AMP binding to the 130 000 component is markedly inhibited by adenosine and adenine nucleotides, but not cyclic GMP. Thus, this component may reflect an aspect of adenosine control or metabolism which may or may not be a cyclic AMP-related cellular function.     

Two mutations in the dispensable part of alanine tRNA synthetase which affect the catalytic activity

Two previously described chromosomal mutant alleles, alaS4 and alaS5, of Escherichia coli Ala-tRNA synthetase have been analyzed. Each causes a sharp diminution in aminoacylation activity and disrupts the alpha 4 tetramer structure of identical chains of 875 amino acids; neither mutation significantly disturbs the activity for synthesis of alanyladenylate. The location of each mutation within the structural gene has been mapped by marker rescue with specific gene fragments. Each mutant allele was cloned from the genome by reciprocal recombination with a multicopy plasmid that contains segments of alaS which flank the respective mutations. Further analysis established: 1) a single G----A transition results in a Gly----Asp change for each mutant allele at codon 674 (alaS4) and at codon 677 (alaS5). 2) The mutations are in the oligomerization domain, about 200 amino acids beyond the C-terminal side of the catalytic domain that previously was mapped by deletion analysis; the mutations are, thus, in a part of the polypeptide which is dispensable for catalytic activity. 3) For both mutant enzymes, there is little effect of the mutation on the Km for tRNAAla; kcat for aminoacylation is decreased by an order of magnitude. These point mutations reveal a subtle integration of the catalytic core with parts of the polypeptide that are not essential for catalytic activity.     

Retinoic acid increases cyclic AMP-dependent protein kinase activity in murine melanoma cells

The influence of all trans-retinoic acid on cyclic AMP metabolism was examined in B16-F1 mouse melanoma cells. Treatment of these cells with retinoic acid resulted in a dose-dependent inhibition of cell growth which was accompanied by a concentration-dependent increase in both basal and cyclic AMP-stimulated protein kinase activity, Intracellular levels of cyclic AMP, however, were not altered by retinoid treatment. A protein kinase-deficient variant of B16-F1 (MR-4) did not exhibit decreased growth or increased protein kinase activity in response to retinoic acid treatment. At least 24 h of incubation was required before increased protein kinase activity could be detected in treated B16-F1 cells. Retinoic acid treatment increased the Vmax of protein kinase, but the Ka for cyclic AMP activation was not altered. These findings suggest that in B16 mouse melanoma cells, cyclic AMP-dependent protein kinase may be a target for the growth inhibitory effects of the retinoid.     

Adenosine 3'':5''-cyclic monophosphate-dependence of protein kinase isoenzymes from mouse liver.

Conditions influencing the cyclic AMP-dependence of protein kinase (ATP-protein phosphotransferase, EC 2.7.1.37) during the phosphorylation of histone were studied. Protein kinase from mouse liver cytosol and the two isoenzymes [PK (protein kinase) I and PK II] isolated from the cytosol by DEAE-cellulose chromatography were tested. A relation between concentration of enzyme and cyclic AMP-dependence was observed for both isoenzymes. Moderate dilution of isoenzyme PK II decreased the stimulation of the enzyme by cyclic AMP. Isoenzyme PK I could be diluted 200 times more than isoenzyme PK II before the same decrease in cyclic AMP-dependence appeared. Long-term incubation with high concentrations of histone increased the activity in the absence of cyclic AMP relative to the activity in the presence of the nucleotide. This was more pronounced for isoenzyme PK II than for isoenzyme PK I. The cyclic AMP concentration needed to give half-maximal binding of the nucleotide was the same as the cyclic AMP concentration (Ka) at which the protein kinase had 50% of its maximal activity. The close correlation between binding and activation is also found in the presence of KCl, which increased the apparent activation constant (Ka) for cyclic AMP. With increasing [KCl], a progressively higher proportion of the histone phosphorylation observed in cytosol was due to cyclic AMP-independent (casein) kinases, leading to an overestimation of the degree of activation of the cyclic AMP-dependent protein kinases present. The relative contributions of cyclic AMP-dependent and -independent kinases to histone phosphorylation at different ionic strengths was determined by use of heat-stable inhibitor and phospho-cellulose chromatography.     

Localization of symbiotic mutations in Rhizobium meliloti

A total of 5 Nod- and 57 Fix- symbiotic mutants of Rhizobium meliloti strain 41 have been isolated after either nitrosoguanidine or Tn5 transposition mutagenesis. Chromosomal locations of mutations in 1 Nod- and 11 Fix- derivatives were ascertained by transferring the chromosome (mobilized by plasmid R68.45), in eight fragments, into symbiotically effective recipients and testing the recombinants for symbiotic phenotype. Alternatively, the kanamycin resistance marker of Tn5 was mapped. In five mutants the fix alleles were localized on different chromosomal regions, but six other fix mutations and one nod mutation tested did not map onto the chromosome. It was shown that the chromosome-mobilizing ability (Cma+) of R68.45 was not involved in the mobilization of genes located extrachromosomally. Moreover, Cma- derivatives of R68.45 could mobilize regions of the indigenous plasmid pRme41b but not chromosomal genes. Thus, mobilization of a marker by Cma- R68.45 indicates its extrachromosomal location. With a 32P-labeled DNA fragment carrying Tn5 as a hybridization probe, it was shown that in five extrachromosomally located Tn5-induced fix mutants and one nod mutant Tn5 was localized on plasmid pRme41b. This is in agreement with the genetic mapping data.     

Physical mapping of DNA-temperature-sensitive mutations of vaccinia virus

Four DNA-temperature-sensitive (ts) mutations were mapped in the genome of vaccinia virus (VV). Physical mapping of these mutations was performed by restriction analysis of the genomes of recombinants between VV DNA- ts mutants and ectromelia virus as well as by the marker rescue with cloned restriction fragments of VV DNA. One of the mutations was mapped on the HindIII-E-fragment. Biochemical studies of this mutant indicate that the mutation is not in the DNA polymerase gene which is located on the same fragment. The other three mutations were mapped in a 10 kilobase region in the middle of the HindIII-D-fragment. As shown previously, these mutations inactivate different genes, and the products of these genes participate directly in the DNA synthesis. Thus, at least three proteins involved in the VV DNA synthesis are encoded by neighboring genes in the central part of the viral genome.     

Genetic Mapping of the Amino-Terminal Domain of Bacteriophage T4 DNA Polymerase

The DNA polymerase of bacteriophage T4 is a multifunctional enzyme that harbors DNA-binding, DNA-synthesizing and exonucleolytic activities. We have cloned in bacterial plasmids about 99% of the structural gene for this enzyme (T4 gene 43). The gene was cloned in six contiguous 5'-terminal DNA fragments that defined seven intragenic mapping regions. Escherichia coli hosts harboring recombinant plasmids carrying the gene 43 subsegments were used in marker-rescue experiments that assigned a large number of ts and nonsense polymerase mutations to different physical domains of the structural gene. Conspicuously, only one missense mutation in a large collection of mutants mapped in the 5'-terminal 450 base-pair segment of the approximately 2700 base-pair gene. To test if this indicated a DNA polymerase domain that is relatively noncritical for biological activity, we mutagenized a recombinant plasmid carrying this 5'-terminal region and generated new conditional-lethal mutations that mapped therein. We identified five new ts sites, some having mutated at high frequency (nitrosoguanidine hot spots). New ts mutations were also isolated in phage genes 62 and 44, which map upstream of gene 43 on the T4 chromosome. A preliminary examination of physiological consequences of the ts gene 43 mutations showed that they exhibit effects similar to those of ts lesions that map in other gene 43 segments: some were mutators, some derepressed gene 43 protein synthesis and they varied in the severity of their effects on T4-induced DNA synthesis at nonpermissive temperatures. The availability of the gene 43 clones should make it possible to isolate a variety of lesions that affect different activities of the T4 DNA polymerase and help to define the different domains of this multifunctional protein.     

Physical and genetic mapping of the protein A gene in the chromosome of Staphylococcus aureus 8325-4

The gene coding for protein A (spa) has been mapped close to nov on the genetic map of the chromosome of Staphylococcus aureus 8325-4. A rapid mapping procedure has been developed which first allowed the region of the chromosome carrying the spa gene to be identified by blot +hybridization of large DNA fragments which had been separated by pulsed-field gel electrophoresis. Restriction endonuclease SmaI fragment G was shown to carry the spa gene. An insertion mutation in spa was constructed by in vitro insertion of a fragment of DNA expressing resistance to kanamycin and neomycin. A spa::Kan(r)Neo(r) mutation was isolated in S. aureus 8325-4 by allele replacement. This provided a selectable marker which allowed the spa gene to be mapped by transformation analysis.     

A method for rapid mapping of mutations by plasmid rescue strategy inSaccharomyces cerevisiae

The products ofPRP17 andPRP18 genes are required for the second step of pre-mRNA splicing reactions inSaccharomyces cerevisiae. Temperature-sensitive mutants at either of these loci accumulate products of the first splicing reaction at nonpermissive temperature. To characterize functional regions in these proteins the mutations in three temperature-sensitive alleles ofPRP17 and two temperature-sensitive alleles ofPRP18 were mapped by the plasmid rescue strategy, One of the procedures adopted in the past is plasmid rescue of the mutant allele followed by sequencing of the entire gene. In this work we describe an adaptation of the above procedure that allows, first, rapid mapping of chromosomal segments bearing the mutations, followed by sequence characterization of the minimal segment. The strategy adopted was to integrate a wild-type copy of the gene at the homologous mutant chromosomal locus, followed by recovery of the chromosomal fragments from these integrants as plasmids inE. coli. The recovered plasmids were screened by a complementation assay for those that contained in them the chromosomal mutation. The mutations in all the three alleles ofPRP17 map to a small region in the N-terminal half of the protein, whereas the temperature-sensitive mutations in the two alleles ofPRP18 map to different regions of the PRP18 protein. The recovered mutant plasmids from all five alleles at the two loci were sequenced and the nucleotide changes were found to result in missense mutations in each case. Our strategy is therefore a rapid method to map chromosomal mutations and is of general use in structure-function analysis of cloned genes.     

Somatic genetic analysis of cyclic AMP action: selection of unresponsive mutants.

Dibutyryl cyclic AMP and theophylline kill S49.1 mouse lymphoma tissue culture cells. When cells are grown in soft agar with these drugs, the few clones that survive are resistant to cytolysis. The rate of mutation to resistance is 1-3 times 10-7/cell/generation in both diploid and tetraploid cells. The incidence of mutants is increased by treatment with a chemical mutagen, ICR 191. The mutation is consistently associated with greatly reduced or absent cytoplasmic cyclic AMP binding protein. These results suggest that a somatic mutation leads to a defect of the protein kinase regulatory subunit and that activity of this kinase is required for induction of cell death by cyclic AMP.