Identification of four candidate cGMP targets in Dictyostelium

Jonathan M. Goldberg*†, Leonard Bosgraaf†‡, Peter J. M. Van Haastert‡, and Janet L. Smith*§

*Boston Biomedical Research Institute, 64 Grove Street, Watertown, MA 02472-2829; and ‡Department of Biochemistry, University of Groningen, Nijenborgh 4, 9747 AG, Groningen, The Netherlands

Communicated by James A. Spudich, Stanford University School of Medicine, Stanford, CA, March 22, 2002 (received for review November 13, 2001) In Dictyostelium, a transient increase in intracellular cGMP is signaling is mediated by cGMP-dependent kinases important for cytoskeletal rearrangements during chemotaxis. (PKGs), cGMP-gated ion channels, a cAMP͞cGMP-regulated There must be cGMP-binding in Dictyostelium that regu- RasGEF (6), and cGMP-regulated phosphodiesterases (PDEs). late key cytoskeletal components after treatment with chemoat- The cGMP-binding sites in these proteins are homologous to the tractants, but to date, no such proteins have been identified. Using Escherichia coli catabolite gene activator protein (CAP), except a bioinformatics approach, we have found four candidate cGMP- for the PDEs, which bind cGMP by using GAF domains (cGMP- binding proteins (GbpA–D). GbpA and -B have two tandem cGMP- specific and stimulated PDEs, Anabaena adenylate cyclases, and binding sites downstream of a metallo ␤-lactamase domain, a E. coli FhlA; ref. 7). Remarkably, no cGMP-binding proteins superfamily that includes cAMP phosphodiesterases. GbpC con- have been cloned from Dictyostelium. Biochemical evidence for tains the following nine domains (in order): leucine-rich repeats, such proteins comes from studies quantitating cGMP-binding Ras, MEK kinase, Ras guanine nucleotide exchange factor N- sites in cellular lysates (8), and the observations that myosin light terminal (RasGEF-N), DEP, RasGEF, cGMP-binding, GRAM, and a chain kinase-A (MLCK-A) and cGMP PDE are activated by second cGMP-binding domain. GbpD is related to GbpC, but is cGMP in crude lysates (9, 10). much shorter; it begins with the RasGEF-N domain, and lacks the We have taken a bioinformatics approach to identifying such DEP domain. Disruption of the gbpC gene results in loss of all proteins. The Dictyostelium genome is well advanced and partial high-affinity cGMP-binding activity present in the soluble cellular or complete sequence for Ն95% of genes is publicly available. (L. fraction. GbpC mRNA levels increase dramatically 8 h after starva- Eichinger, personal communication). We have mined this data tion is initiated. GbpA, -B, and -D mRNA levels show less dramatic resource and discovered four proteins predicted to bind cGMP. changes, with gbpA mRNA levels highest 4 h into starvation, gbpB These proteins have novel domain structures compared with mRNA levels highest in vegetative cells, and gbpD levels highest at cGMP targets that have been identified in metazoans. 8 h. The identification of these genes is the first step in a molecular approach to studying downstream effects of cGMP signaling in Methods Dictyostelium. Sequence Databases. The Dictyostelium genome is being se- quenced at the following centers: genome.imb-jena.de͞ ictyostelium discoideum is a small, motile eukaryote that is dictyostelium (Genome Sequencing Center Jena); www.uni- Dremarkably similar to mammalian cells in morphology and koeln.de͞dictyastelium (University of Cologne); dictygenome. biology. Dictyostelium cells have a highly developed chemo- bcm.tmc.edu (Baylor College of Medicine); and www. tactic response. During vegetative growth they respond to bac- sanger.ac.uk͞Projects͞D࿝discoideum (Sanger Centre). Data terial compounds as a mechanism for finding food; upon star- from a large-scale cDNA Project in Tsukuba, Japan can be vation they secrete and chemotax toward cAMP, thus using obtained at www.csm.biol.tsukuba.ac.jp. cAMP as a beacon for aggregation so that multicellular fruiting bodies are formed. In mammals, chemotaxis plays key roles in Identification and Sequencing of the gbp Genes. We searched the CELL BIOLOGY defense against infection, wound healing, angiogenesis, embry- genomic and cDNA databases of individual sequence reads by ogenesis, and metastasis of cancer cells. Overall, the signal using the BLAST search engines available at each of the sequenc- transduction pathways associated with chemotaxis are conserved ing centers’ web sites, as well as a central repository of sequences between Dictyostelium and higher organisms (1). With its rela- available at the San Diego Supercomputer Center (www. tively small genome (34 MB; Ϸ10,000 genes), Dictyostelium sdsc.edu͞mpr͞dicty͞). Query sequences included all known offers the virtue of a simple setting in which the cellular cGMP-binding sites from ion channels, PKGs, and CNrasGEF, processes and pathways associated with as well as cAMP-binding sites from CAP, Rap1-GEF (11, 12), chemotaxis can be understood in detail. and PKA regulatory subunits (PKA-Rs). In addition, we Migration of cells up a chemotactic gradient involves asym- searched using 11 different GAF domain sequences. metric remodeling of the cytoskeleton during the first minute For each gene, sequences were assembled from individ- after stimulation, and therefore requires that cytoskeletal com- ual reads downloaded from the sequencing centers, using ponents be tightly regulated. In Dictyostelium, cGMP transiently DNASTAR (Lasergene, Madison, WI). For uncertain portions, increases when cells are exposed to chemoattractants (2, 3). Cells with two guanylyl cyclase genes disrupted do not increase cGMP in response to chemoattractant, and have a partial chemotaxis Abbreviations: CAP, E. coli catabolite gene activator protein; MBL, metallo ␤-lactamase; defect, indicating that cGMP signaling is important for the MEKK, mitogen-activated͞extracellular-signal regulated protein kinase kinase kinase; PDE, phosphodiesterase; PKA-R, cAMP-dependent protein kinase regulatory subunit; Ras- chemotactic response (4). GEF, Ras guanine nucleotide exchange factor. cGMP signaling in mammals is important in physiological Data deposition: The sequences reported in this paper have been deposited in the GenBank processes such as smooth muscle tone, epithelial electrolyte database (accession nos. AF481921–AF481924 and AY099028). transport, neuronal excitability, and phototransduction (5). Be- †J.M.G. and L.B. contributed equally to this work. cause cGMP signaling is not well documented in yeasts, Dictyo- §To whom reprint requests should be addressed. E-mail: [email protected]. stelium is the simplest model organism in which cGMP signaling The publication costs of this article were defrayed in part by page charge payment. This pathways can be studied. For cGMP to function, it must bind to article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. and regulate the activity of key proteins. In metazoans cGMP §1734 solely to indicate this fact.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.102167299 PNAS ͉ May 14, 2002 ͉ vol. 99 ͉ no. 10 ͉ 6749–6754 Downloaded by guest on October 1, 2021 additional data were obtained by sequencing cDNAs and presented in Fig. 1A, and an annotated version of their genomic genomic PCR products. All introns were verified from cDNAs sequences is shown in Figs. 5–8, which are published as sup- or by reverse transcription (RT)-PCR. For gbpB and -D, the porting information on the PNAS web site, www.pnas.org. Each 3Ј ends of the ORF were deduced based on sequences from the protein has two potential cGMP-binding sites with homology to Japanese cDNA project. For gbpC, the 3Ј end was clear based on CAP, as well as one or more other domains that are discussed a TAA stop codon, followed by an A-rich sequence typical of in more detail below. We also found one GAF domain protein Dictyostelium terminators. A long ORF on the opposite strand (GafA), but based on the crystal structure and mutagenesis ending 567 bp downstream of the stop codon (data not shown) studies of a yeast GAF domain (16), this Dictyostelium protein makes an additional downstream exon unlikely. For gbpA, the 3Ј is not predicted to bind cGMP. end was determined by 3ЈRACE (Invitrogen). The 5Ј ends of the An alignment of the putative cGMP-binding sites from the gbpA and -B ORFs were determined from cDNA clones. The 5Ј Dictyostelium Gbp proteins with known cAMP and cGMP- ends of the gbpC and -D ORFs were assigned based on the fact binding sites is presented in Fig. 1B. Overall, the Dictyostelium that they are immediately preceded by at least 470 bp (gbpC) and sites are quite divergent from those previously described, al- 220 bp (gbpD) of AT-rich sequences typical of intergenic regions. though all except the first site in GbpD are readily recognized as cNMP-binding sites in an NCBI Conserved Domain (CD) search Total RNA Preparation and Northern Analysis. One hundred million against the Pfam collection (17). The E-values log-phase cells were allowed to develop on 42.5-mm Whatman from these CD searches are reported in Fig. 1B. Lower values No. 50 filters, as described (13). At 4 h intervals, a filter was indicate a better match to the consensus sequence; by this placed in 10 ml of TRIzol Reagent (Invitrogen) and gently measure the first site in GbpA and the second site in GbpB are agitated for 5 min to dissolve the cells. The filter was removed, the best fits to the cNMP-binding motif, but the other scores also and the RNA was isolated according to the manufacturer’s indicate significant similarity. The first site in GbpD was iden- protocol. The yield decreased during development; 1.12 mg was ␮ tified by eye, and this sequence represents a good match to the recovered at 0 h, and 0.45 mg at 24 h. For Northern blots, 6.2 g consensus for approximately the first half of the domain; the of total RNA was fractionated by agarose-formaldehyde gel second half is more degenerate. The second site in GbpD has a electrophoresis, transferred to nitrocellulose, and probed as large insertion, but inspection of the PKA-R crystal structures described (14). As probes, cloned DNA fragments (0.7–1.2 kb) reveals that a large loop can be accommodated at this position from each gene were radioactively labeled using a rediprime II (18, 19). kit (Amersham Pharmacia). For eukaryotic cAMP and cGMP-binding proteins, specificity is thought to be determined by the highly conserved ‘‘FGE... Gene Disruption of gbpC. A 874-bp portion of gbpC was amplified R(T͞A)A’’ motif, where typically an ‘‘RTA’’ or occasionally an from genomic DNA using the primers ATCATTGATTTCGG- ‘‘RSA’’ is present in cGMP-binding sites, and an ‘‘RAA’’ is TACAAG and TCTTCCTCTGTGAAAATATC. This frag- present in cAMP-binding sites (20). Crystal structures for type ment was cloned, and the Bsr selection cassette (15) was inserted I and II PKA-Rs have been determined (18, 19); these structures, in the same orientation as the gbpC gene into the BglI site as well as earlier modeling studies (20), provide a structural basis between the kinase and the RasGEF-N domains. The resulting for cyclic nucleotide specificity. When cGMP is superimposed on insert was amplified with the primers described above, and 4 ␮g of the PCR product was used to transform DH1 (‘‘wild type’’) cAMP, and the RAA is replaced by RTA or RSA, the Ser or Thr cells by electroporation. Transformants were selected in medium side chain can accept a hydrogen bond from the amino nitrogen with 10 ␮g͞ml blasticidin. Knockouts were screened by PCR and of the C2 carbon of cGMP. In the CAP structure, the confor- confirmed by Southern analysis. Ninety-five percent of the mation of the bound cAMP is anti, rather than syn as in PKA transformants had a disrupted gbpC locus. (21). As a result, the residues likely to be important in specificity differ in these two evolutionarily related classes of proteins. CAP Cyclic Nucleotide Binding Assays. Cells were starved for4hin10 is a dimer, and T127 from one subunit, together with S128 from mM sodium phosphate (pH 6.5), washed two times with lysis the other are implicated in defining specificity. The regions buffer (40 mM Hepes/NaOH (pH 7.0)͞0.5 mM EDTA), resus- corresponding to T127 and S128 are not conserved in GbpA–D, pended at 108 cells per ml in lysis buffer containing 0.25 M or in the other eukaryotic proteins (Fig. 1B), suggesting that for sucrose, and passed through a 0.45-␮m Nuclepore filter. The them this region in not involved in cNMP recognition. Also, resulting lysate was precleared by centrifugation for 2 min at 8-Br-cGMP, which strongly favors the syn conformation, acti- 14,000 ϫ g, followed by centrifugation for1hat48,000 ϫ g. vates cGMP targets in Dictyostelium (9, 22). Binding reactions (200 ␮l) contained 100 ␮l of high-speed Based on these criteria, we expect that the cGMP binds to the supernatant, 0.5ϫ lysis buffer, 0.125 M sucrose, 50 mM phos- Dictyostelium proteins in a syn conformation, and have chosen 3 phate (pH 6.1), 3 mM MgCl2, 5 mM DTT, and 10 nM [ H]cGMP PKA, rather than CAP, in evaluating specificity in our proteins. (with or without 250 nM cold cGMP) or 10 nM [3H]cAMP. We predict that GbpA.1, both sites in GbpB and -C, and GbpD.2 Nonspecific binding was determined by including 30 ␮M cGMP bind cGMP preferentially to cAMP, because they have a Ser or or cAMP. Reactions were incubated for 15 min at 0°C and Thr at the position marked with an asterisk in Fig. 1B. GbpA.2 filtered through 0.45 ␮m of nitrocellulose, and the filters then and GbpD.1 have very degenerate FGE...RTAmotifs, and thus washed two times with 50 mM phosphate (pH 6.1), dried, and may not bind either cyclic nucleotide with high affinity. analyzed by liquid scintillation counting. GbpA and -B. GbpA and -B have a metal hydrolase domain Results upstream of two tandem cGMP-binding sites (Fig. 2). These Identification of the gbp Genes. We took a bioinformatics approach metal hydrolase domains are in the metallo ␤-lactamase (MBL) to identify cyclic nucleotide (cNMP)-binding proteins in the superfamily, which includes a broad range of enzymes that use Dictyostelium genome, using low-threshold BLAST searches with metal catalytically in hydrolysis reactions (23). MBL itself has ϩ known cAMP- and cGMP-binding proteins. We have identified two bound Zn2 ions that are coordinated by His and Asp four novel cNMP-binding proteins. We have designated these residues (24). These residues are conserved in GbpA and -B, genes gbpA–D, and the proteins GbpA–D, for cGMP-binding presumably reflecting shared metal-binding properties. In MBL proteins, because, as discussed below, they have a higher likeli- the Zn2ϩ ions, together with two serines and a tyrosine, are hood of binding cGMP than cAMP. Their domain structures are involved in catalysis. These residues are not conserved in GbpA

6750 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.102167299 Goldberg et al. Downloaded by guest on October 1, 2021 Fig. 1. Four candidate cGMP-binding proteins. (A) The domain structures of the GbpA–D proteins are shown approximately to scale. The putative cGMP-binding sites are labeled ‘‘cG.’’ The amino acid present at a position known to correlate with cNMP specificity is indicated; this position is also shaded pink and marked with an asterisk in B.(B) The putative cGMP-binding sequences from the Gbp proteins are aligned with known cAMP-binding sites from type I bovine PKA-R (GenBank accession no. P05987) and CAP (P03020), the cGMP-binding domains from mouse PKGII (L12460), bovine cGMP-gated ion channel (X51604), and a domain from human CNrasGEF, which binds both cAMP and cGMP (KIAA0313). The two cNMP sites found in most of the proteins are distinguished by a numerical suffix. Residues are shaded black if they are identical to the consensus for each position in the alignment, and shaded gray if they are similar to the consensus. The residues thought to be involved in cNMP specificity in CAP are shaded yellow and indicated with open circles. The E-values for a Pfam cNMP domain were obtained from a Conserved Domain Search (NCBI). For GbpD.1, the E-value was obtained after manually removing the 63-aa insert indicated in the alignment. CELL BIOLOGY

or -B, suggesting that these proteins have a different catalytic from Dictyostelium (25). Because a cGMP PDE activity that is function. stimulated by cGMP has been identified in Dictyostelium lysates GbpA and -B do not show particularly strong similarity to any (9), we suggest that GbpA and͞or -B are cGMP-activated cGMP characterized member of the MBL superfamily, so it is not PDEs. The metal hydrolase domains in GbpA and -B are only possible to predict substrate specificity on this basis. However, 39% identical, leaving open the possibility that they catalyze several cAMP PDEs are in the MBL superfamily, including one different reactions.

Fig. 2. The metal hydrolase domain from GbpA and -B. The domains from GbpA and -B are aligned with DPDE, the Dictyostelium cAMP PDE (J02628), and MBL, the L1 MBL from Stenotrophomonas maltophilia (CAB63489). Residues are shaded black if they are identical to the consensus for each position in the alignment, and shaded gray if they are similar to the consensus. The conserved His and Asp residues that bind Zn2ϩ in the MBL structure are shaded pink and marked with an asterisk, and the Ser and Tyr residues involved in ␤-lactam hydrolysis by MBL are shaded yellow and indicated by a plus sign.

Goldberg et al. PNAS ͉ May 14, 2002 ͉ vol. 99 ͉ no. 10 ͉ 6751 Downloaded by guest on October 1, 2021 Fig. 3. Sequence analysis of GbpC and -D. Residues are shaded black if they are identical to the consensus for each position in the alignment, and shaded gray if they are similar to the consensus. (A) Alignment of the Ras sequences from GbpC, the Drosophila CG5483 protein (D.m., GenBank accession no. AAF55793), a human protein encoded by the KIAA1790 cDNA (H.s., BAB47419), and human H-Ras (P01112). Key residues involved in GTP hydrolysis are highlighted and marked with an asterisk. (B) Alignment of the protein kinase sequences from GbpC, transforming growth factor ␤-activated protein kinase 1 (Tak1, O43318), a RAF homolog from Arabidopsis thaliana (Ctr1, A45178), and mixed lineage kinase 1 (Mlk1, AAG44591). The twelve nearly invariant protein kinase residues (41) are highlighted and indicated with an asterisk. For positions that show a conserved residue in Ͼ90% of Tyr kinases, but not Ser͞Thr kinases (42), the preferred amino acid is indicated above the alignment. (C) Alignment of the RasGEF catalytic domains from GbpC and -D with those from human Sos-1 (Q07889) and Saccharomyces cerevisiae cdc25 (P04821). The bar above the alignment in C indicates the helical hairpin involved in nucleotide exchange. (D) Alignment of the DEP domain from GbpC with mouse protein (Dv1, P51141), mouse pleckstrin (Plk., AAG29513), and C. elegans Egl-10 (Egl10, P49809). Residues implicated in forming a ␤-hairpin arm and electric dipole are highlighted and indicated by asterisks.

N-Terminal Half of GbpC. GbpC is a large (2631 aa), multidomain responsible for the basal GTPase activity of Ras are present in protein. The first three domains—i.e., the leucine-rich repeats the Dictyostelium protein, suggesting that this domain functions (LRRs), Ras, and MEK kinase domains—are unique to GbpC, as a GTPase. whereas the C-terminal half, containing the RasGEF domain For GbpC’s protein kinase domain, all 12 of the nearly and cGMP-binding sites, is very similar to GbpD. The six invariant protein kinase residues are conserved (asterisks in Fig. 22–23-aa-long LRRs in GbpC are aligned in Fig. 7B. LRRs occur 3B), strongly suggesting that it has catalytic activity. Overall the in a wide variety of bacterial and eukaryotic proteins, and GbpC kinase domain is most similar (31–34% identity) to function by forming complexes with other proteins (26, 27). Arabidopsis CTR1, mammalian transforming growth factor ␤ Immediately downstream of the LRRs, GbpC has a domain with (TGF-␤)-activated kinase 1, and mixed lineage kinase 1, which homology to Ras (Fig. 3A). This domain has highest homology are all MEKKs (mitogen-activated͞extracellular signal- to a region of the Drosophila CG5483 gene product (28). The regulated protein kinase kinase kinase). MEKKs activate the second best match in a BLAST search is KIAA1790, a protein MEK family of protein kinases, which in turn activates ERKs. deduced from a human partial cDNA clone (29). An alignment Although GbpC and the other sequences in Fig. 3B are hybrid of the Ras domains from these proteins, together with the human between those of Tyr and Ser͞Thr kinases, MEKKs are specific H-Ras sequence, is shown in Fig. 3A. The catalytic residues for Ser͞Thr residues, and GbpC presumably shares this feature.

6752 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.102167299 Goldberg et al. Downloaded by guest on October 1, 2021 GbpC appears to be unique to Dictyostelium; no protein with a similar domain architecture exists in the current sequence database. However, proteins are found in other organisms that are similar to the N-terminal half of GbpC ending after the kinase domain. The Drosophila CG5483 gene product, the GenBank entry most similar to GbpC’s Ras domain, has a series of LRRs upstream of the Ras domain, and a protein kinase domain downstream, making its layout very similar to the N-terminal half of GbpC, and suggesting evolutionary conser- vation. This suggestion is strengthened by the fact that CG5483’s Fig. 4. Northern analysis. (A) Total RNA from vegetative wild-type cells was kinase domain, like that of GbpC, is most similar to mixed probed with radioactive DNA fragments from the indicated genes. The indi- lineage kinase. The human protein KIAA1790 also has a Ras cated size of the transcript was calculated from RNA markers (Promega). (B) domain very similar to that of GbpC with a mixed lineage protein Total RNA prepared at 4-h intervals during development was probed with the kinase domain downstream. KIAA1790 is a partial cDNA clone same DNA fragments as in A. The number of hours of development is indicated that begins just upstream of the Ras domain, so despite the fact for each lane. that LRRs are not found in the reported sequence, they could be present in the uncharacterized 5Ј end of the gene. It appears that gbpC-Nulls Lack a cGMP-Binding Protein. GbpC binding to cyclic GbpC arose as a duplication of GbpD, followed by fusion with ͞ nucleotides was evaluated by assaying the soluble fraction of a CG5483 KIAA1790-like gene. wild-type and gbpCϪ cells. Disruption of gbpC had no effect on binding of 10 nM cAMP: 242 Ϯ 12 fmol͞mg protein for GbpD and the C-Terminal Half of GbpC. Ϫ GbpD and the C-terminal wild-type, and 222 Ϯ 20 fmol͞mg protein for gbpC . In contrast, half of GbpC have similar domain structures and are predicted binding of 10 nM cGMP was reduced from 221 Ϯ 12 fmol͞mg to have a cGMP-regulated RasGEF activity. RasGEFs activate for wild type to 17.7 Ϯ 1.8 fmol͞mg for gbpCϪ, indicating that Ras by promoting the exchange of GDP for GTP. This catalytic GbpC binds cGMP, as predicted from the sequence. The residual activity is carried out by the C-terminal RasGEF domain, binding activity in gbpCϪ cytosol indicates that one or more whereas the RasGEF-N domain is thought to play a structural additional cGMP-binding proteins are present. At 260 nM role (30). GbpC and -D both have RasGEF domains (aligned in cGMP, binding increased 4-fold to 853 Ϯ 88 fmol͞mg for wild Fig. 3C) and RasGEF-N domains (aligned in Fig. 7C). The type and 22-fold to 419 Ϯ 60 fmol͞mg for gbpCϪ. The greater domains found in GbpD are better matches to the consensus relative increase in binding in the null extract suggests that one than are those found in GbpC. In particular, it is noteworthy that or both GbpC-binding sites becomes saturated at approximately the GbpC RasGEF domain does not appear to have the helical 10 nM, while the remaining Gbps have higher Kd values. hairpin (indicated above the alignment in Fig. 3C) that is thought to play an important role in catalytic activity (30). However, Developmental Expression of the gbp Genes. Dictyostelium devel- GbpC’s RasGEF domain could potentially act on GbpC’s own opment is a process rich in signaling events and cytoskeletal Ras domain, and this missing helical hairpin may reflect an changes, and changes in mRNA levels for the gbp genes may adaptation for intramolecular signaling. reflect specific roles for these proteins in vegetative growth or The DEP domain found immediately upstream of the Ras- during development. We used Northern blotting to determine GEF catalytic domain in GbpC is not present in GbpD, even changes in expression during development. In addition, the though it is in the portion of GbpC that is similar to GbpD (Fig. mRNA size determined by this analysis provides an independent 1A). DEP domains are found in a wide variety of proteins, and measure for evaluating the gbp gene sequences presented in Fig. are named for dishevelled, Egl-10, and pleckstrin—three well 5–8. gbpA is maximally expressed at 4 h (Fig. 4B). The mRNA studied proteins in which they are found (31). DEP domains is 2.9 kb (Fig. 4A), consistent with our mapping of the transcript target proteins to the membrane, but the residues involved in this by using cDNAs and 3Ј RACE—a 2,607-bp ORF and 300 bp of CELL BIOLOGY function are not well defined. DEP domains also mediate untranslated sequence. gbpB is maximally expressed in vegeta- protein–protein interactions, and for the dishevelled DEP do- tive cells as a 3.4-kb mRNA (Fig. 4), consistent with the 3,291-bp main this occurs via an electric dipole that is formed in part by ORF and 122 bp of untranslated sequence that we observed from ␤ ␤ a -hairpin arm (32). The two glycines that flank the -hairpin sequencing cDNA clones. For both gbpA and -B, the changes in arm, and the one basic and two acidic residues that form the gene expression levels are fairly small. In contrast, gbpC expres- electric dipole, are conserved the GbpC DEP domain (Fig. 3D). sion increases dramatically by 8 h development. The size of the GbpC and -D both have a GRAM domain between their two gbpC mRNA is Ϸ8 kb, consistent with the 7,896-bp ORF that we cGMP-binding sites (highlighted in Figs. 7 and 8). This motif was have determined. The level of gbpD mRNA is also maximal at named after glucosyltransferases, Rab-like GTPase activators, 8 h, although the increase is less dramatic than is observed with and myotubularins—the most well characterized proteins that gbpC. The mRNA is Ϸ4.1 kB (Fig. 4A), consistent with a contain this domain (33). The function of GRAM domains is 3,929-bp ORF for gbpD and 24–68 bp of 3Ј untranslated unknown, but they are proposed to mediate protein–protein sequence, as indicated by two clones from the cDNA project. interactions (33). GbpD and the C-terminal half of GbpC do not appear to have Discussion a direct homolog in metazoans. Although Ras and Rap1 GEFs Relationship of the Gbp Proteins to cGMP Targets in Higher Eu- containing one or two cNMP-binding sites have been found in karyotes. We are confident that we have identified all of the mammals, Caenorhabditis elegans and Drosophila melanogaster, candidate cGMP-binding proteins (both in the CAP and GAF in each of these cases the cNMP-binding site(s) are upstream of families) that are represented in the current Dictyostelium se- the RasGEF domain, rather than downstream as in GbpC and quence databases. Given the advanced stage of the genome -D. Each of the metazoan proteins also has a DEP domain, as sequencing effort, all cGMP signaling in Dictyostelium may well is the case for GbpC, but its position in the primary sequence be mediated by GbpA–D. Remarkably, we did not find direct differs as well. The metazoan proteins also lack the GRAM homologs of any of the cGMP targets that have been identified domain found in GbpC and -D. Thus it appears that cyclic in higher eukaryotes. Rather, GbpA–D represent four previously nucleotide-regulated GEFs evolved separately in these lineages. uncharacterized proteins that in turn do not have clear homologs

Goldberg et al. PNAS ͉ May 14, 2002 ͉ vol. 99 ͉ no. 10 ͉ 6753 Downloaded by guest on October 1, 2021 in higher eukaryotes. Our findings indicate that cGMP signaling based on sequence analysis, that GbpA and͞or -B are cGMP evolved independently in these two lineages. PDEs, and preliminary data indicates that gbpA nulls do indeed have reduced cGMP PDE activity (L.B. and P.J.M.V.H., un- Proposed Regulatory Mechanism for GbpC and -D. GbpC and -D are published results). Thus GbpA, and possibly GbpB, may be implicated in Ras signaling, because they each contain RasGEF involved in temporally and spatially controlling the cGMP signal domains and GbpC contains a Ras domain. Ras signaling in in response to chemoattractants, whereas GbpC and -D would be Dictyostelium is complex, as evidenced by the fact that Dictyo- responsible for relaying the cGMP signal downstream so that the stelium contain at least ten Ras genes (34)—a disproportionately cytoskeletal changes responsible for chemotactic movements large number compared with its genome size. Dictyostelium occur. GbpC and -D likely activate a Ras during chemotaxis; appears to have a similarly high number of RasGEFs; two have perhaps GbpC’s own Ras domain, or one of the other Dictyo- been described (35, 36), GbpC and -D are described here, and stelium Ras proteins. Given the homology of GbpC’s kinase our searches of the Dictyostelium genomic sequence databases domain to MEKKs, it is likely that GbpC phosphorylates a MEK readily revealed at least 18 more (data not shown). in a MAP kinase cascade. Dictyostelium appears to have at least The juxtaposition of RasGEF and cGMP-binding domains in three MAP kinase pathways (37–39). The Dictyostelium proteins GbpD implies that the RasGEF activity is regulated by cGMP; ERK2 and Mek1 are both important for chemotaxis, and are in and given that GbpC and -D likely have a common origin, our separate pathways. Given the role of cGMP signaling in chemo- working model is that the RasGEF activity of GbpC is also taxis, it follows that GbpC might activate ERK2 through an regulated by cGMP. We propose that the MEKK activity of unidentified MEK, or it might phosphorylate Mek1. GbpC may GbpC is directly regulated by its upstream Ras domain; this also phosphorylate other types of targets. For example, myosin suggestion is based on the fact that Raf, another MEKK, is well II activity is tightly regulated during chemotaxis, and the acti- known to be directly activated by Ras. Assuming that GbpC’s vation of myosin light chain kinase-A (MLCK-A) is cGMP- own RasGEF domain activates its Ras domain, then the MEKK dependent in crude lysates (10, 40). Identification of GbpA–Das activity of GbpC would require cGMP, where this dependence candidate downstream players in cGMP signaling paves the way is transmitted from the cGMP-binding domains through the to a molecular approach to studying cGMP signaling during RasGEF and Ras domains on the same polypeptide chain. chemotaxis in this important model organism. However, in light of the multiplicity of Ras and RasGEF proteins in Dictyostelium, it is important to consider the possibility that We are indebted to all of the teams involved in the Dictyostelium GbpC is involved in a network of Ras signaling, rather than sequencing projects. We thank Ana C. Urbin-Reyes for technical functioning as a signal transduction pathway within one assistance, Eugene Koonin for helpful discussions, and L. Aravind for polypeptide. help in identifying GAF proteins. This work was supported by grants from the National Science Foundation and the American Heart Asso- Potential Roles for GbpA–D in Chemotaxis. cGMP signaling plays an ciation–New England Affiliate (to J.L.S.), and a grant from the Neth- important role in chemotaxis in Dictyostelium. We propose, erlands Organization of Scientific Research (to P.J.M.V.H.).

1. Parent, C. A. & Devreotes, P. N. (1999) Science 284, 765–770. 22. Kuwayama, H., Ecke, M., Gerisch, G. & Van Haastert, P. J. M. (1996) Science 2. Mato, J. M., Van Haastert, P. J. M., Krens, F. A., Rhijnsburger, E. H., Dobbe, 271, 207–209. F. C. P. M. & Konijn, T. M. (1977) FEBS Lett. 79, 331–336. 23. Aravind, L. (1999) In Silico Biol. 1, 69–91. 3. Wurster, B., Schubiger, K., Wick, U. & Gerisch, G. (1977) FEBS Lett. 76, 24. Ullah, J. H., Walsh, T. R., Taylor, I. A., Emery, D. C., Verma, C. S., Gamblin, 141–144. S. J. & Spencer, J. (1998) J. Mol. Biol. 284, 125–136. 4. Roelofs, J., Meima, M., Schaap, P. & Van Haastert, P. J. M. (2001) EMBO J. 25. Lacombe, M. L., Podgorski, G. J., Franke, J. & Kessin, R. H. (1986) J. Biol. 20, 4341–4348. Chem. 261, 16811–16817. 5. Ruth, P. (1999) Pharmacol. Ther. 82, 355–372. 26. Andrade, M. A., Perez-Iratxeta, C. & Ponting, C. P. (2001) J. Struct. Biol. 134, 6. Pham, N., Cheglakov, I., Koch, C. A., de Hoog, C. L., Moran, M. F. & Rotin, 117–131. D. (2000) Curr. Biol. 10, 555–558. 27. Kobe, B. & Deisenhofer, J. (1994) Trends Biochem. Sci. 19, 415–421. 7. Aravind, L. & Ponting, C. P. (1997) Trends Biochem. Sci. 22, 458–459. 28. Adams, M. D., Celniker, S. E., Holt, R. A., Evans, C. A., Gocayne, J. D., 8. Kuwayama, H. & Van Haastert, P. J. M. (1996) J. Biol. Chem. 271, 23718– Amanatides, P. G., Scherer, S. E., Li, P. W., Hoskins, R. A., Galle, R. F., et al. 23724. (2000) Science 287, 2185–2195. 9. Van Haastert, P. J. M. & Van Lookeren Campagne, M. M. (1984) J. Cell Biol. 29. Nagase, T., Nakayama, M., Nakajima, D., Kikuno, R. & Ohara, O. (2001) DNA 98, 709–716. Res. 8, 85–95. 10. Silveira, L. A., Smith, J. L., Tan, J. L. & Spudich, J. A. (1998) Proc. Natl. Acad. 30. Boriack-Sjodin, P. A., Margarit, S. M., Bar-Sagi, D. & Kuriyan, J. (1998) Nature Sci. USA 95, 13000–13005. (London) 394, 337–343. 11. de Rooij, J., Zwartkruis, F. J., Verheijen, M. H., Cool, R. H., Nijman, S. M., 31. Ponting, C. P. & Bork, P. (1996) Trends Biochem. Sci. 21, 245–246. Wittinghofer, A. & Bos, J. L. (1998) Nature (London) 396, 474–477. 32. Wong, H. C., Mao, J., Nguyen, J. T., Srinivas, S., Zhang, W., Liu, B., Li, L., Wu, 12. Kawasaki, H., Springett, G. M., Mochizuki, N., Toki, S., Nakaya, M., Matsuda, D. & Zheng, J. (2000) Nat. Struct. Biol. 7, 1178–1184. M., Housman, D. E. & Graybiel, A. M. (1998) Science 282, 2275–2279. 33. Doerks, T., Strauss, M., Brendel, M. & Bork, P. (2000) Trends Biochem. Sci. 25, 13. Sussman, M. (1987) in Dictyostelium discoideum: Molecular Approaches to Cell Biology, ed. Spudich, J. A. (Academic, Orlando), Vol. 28, pp. 9–30. 483–485. 14. Brown, T. & Mackey, K. (1997) in Current Protocols in Molecular Biology, eds. 34. Wilkins, A. & Insall, R. H. (2001) Trends Genet. 17, 41–48. Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith, 35. Insall, R. H., Borleis, J. & Devreotes, P. N. (1996) Curr. Biol. 6, 719–729. J. A. & Struhl, K. (Wiley, New York). 36. Wilkins, A., Chubb, J. R. & Insall, R. H. (2000) Curr. Biol. 10, 1427–1437. 15. Sutoh, K. (1993) Plasmid 30, 150–154. 37. Gaskins, C., Clark, A. M., Aubry, L., Segall, J. E. & Firtel, R. A. (1996) Genes 16. Ho, Y.-S. J., Burden, L. M. & Hurley, J. H. (2000) EMBO J. 19, 5288–5299. Dev. 10, 118–128. 17. Bateman, A., Birney, E., Durbin, R., Eddy, S. R., Howe, K. L. & Sonnhammer, 38. Ma, H., Gamper, M., Parent, C. & Firtel, R. A. (1997) EMBO J. 16, 4317–4332. E. L. (2000) Nucleic Acids Res. 28, 263–266. 39. Chung, C. Y., Reddy, T. B., Zhou, K. & Firtel, R. A. (1998) Genes Dev. 12, 18. Diller, T. C., Madhusudan, Xuong, N. & Taylor, S. S. (2001) Structure (London) 3564–3578. 9, 73–82. 40. Smith, J. L., Silveira, L. A. & Spudich, J. A. (1996) EMBO J. 15, 6075–6083. 19. Su, Y., Dostmann, W. R., Herberg, F. W., Durick, K., Xuong, N. H., Ten Eyck, 41. Hanks, S. K. & Hunter, T. (1995) in The Protein Kinase FactsBook: Protein- L., Taylor, S. S. & Varughese, K. I. (1995) Science 269, 807–813. Serine Kinases, eds. Hardie, G. & Hanks, S. (Academic, San Diego), pp. 7–47. 20. Shabb, J. B. & Corbin, J. D. (1992) J. Biol. Chem. 267, 5723–5726. 42. Hubbard, S. R., Wei, L., Ellis, L. & Hendrickson, W. A. (1994) Nature (London) 21. Weber, I. T. & Steitz, T. A. (1987) J. Mol. Biol. 198, 311–326. 372, 746–753.

6754 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.102167299 Goldberg et al. Downloaded by guest on October 1, 2021