The evolutionary emergence of cell type-specific genes inferred from the gene expression analysis of Hydra

Abstract

Cell lineages of cnidarians including Hydra represent the fundamental cell types of metazoans and provides us a unique opportunity to study the evolutionary diversification of cell type in the animal kingdom. Hydra contains epithelial cells as well as a multipotent interstitial cell (I-cell) that gives rise to nematocytes, nerve cells, gland cells, and germ-line cells. We used cDNA microarrays to identify cell type-specific genes by comparing gene expression in normal Hydra with animals lacking the I-cell lineage, so-called epithelial Hydra. We then performed in situ hybridization to localize expression to specific cell types. Eighty-six genes were shown to be expressed in specific cell types of the I-cell lineage. An additional 29 genes were expressed in epithelial cells and were down-regulated in epithelial animals lacking I-cells. Based on the above information, we constructed a database (http://hydra.lab.nig.ac.jp/hydra/), which describes the expression patterns of cell type-specific genes in Hydra. Most genes expressed specifically in either I-cells or epithelial cells have homologues in higher metazoans. By comparison, most nematocyte-specific genes and approximately half of the gland cell- and nerve cell-specific genes are unique to the cnidarian lineage. Because nematocytes, gland cells, and nerve cells appeared along with the emergence of cnidarians, this suggests that lineage-specific genes arose in cnidarians in conjunction with the evolution of new cell types required by the cnidarians.

Hydra is a member of the phylum Cnidaria, which branched >500 million years ago from the main stem leading to all bilaterian animals. Nevertheless, all Cnidaria including Hydra contain multiple cell types that represent the fundamental building blocks of all multicellular animals. Early investigations of Hydra cell morphology have revealed seven basic cell types (1, 2) including interstitial cells (I-cells) and ectodermal and endodermal epithelial cells that are capable of proliferation and self-renewal. I-cells are multipotent stem cells, which proliferate and continuously differentiate into nematocytes, nerve cells, gland cells, and germ-line cells. In Hydra, big I-cells occur as single cells, in pairs or in clusters of four, whereas small I-cells are found in clusters of eight or more cells (1). I-cells are a continuously proliferating, self-renewing population in the body column (3, 4, 5). Nematocytes and nerve cells continuously differentiate from this population (6, 7). At a late stage of differentiation, most precursors migrate to the hypostome/tentacles and the basal disk, whereas only a few remain in the body column (8). Gland cells and mucous cells differentiate from I-cells, which migrate from the ectoderm into the endoderm and give rise to at least two major types of secretory cells interspersed among the endodermal epithelial cells (9, 10). Mucous cells are localized in the hypostome and upper body column. Gland cells or zymogen cells are distributed throughout the body column (11). I-cells also give rise to germ-line cells, which differentiate to sperm or egg during sexual reproduction (12–16).

Previous results have shown that Hydra do not require the I-cell lineage because, when hand fed, Hydra can carry out metabolic activities and grow after removal of the I-cell lineage (17, 18). Such epithelial Hydra contain only two epithelial layers and a limited number of gland cells. These animals are unable to move and catch or ingest food but can be force-fed by inserting food through the mouth into the gastric cavity (19). Such epithelial Hydra can be maintained in the laboratory for years and remain capable of normal growth, budding, and regeneration.

In the present study, we identified genes that are expressed specifically in the I-cell lineage of Hydra by comparing gene expression in intact and I-cell-free Hydra. We constructed a 6.6 thousand cDNA microarray and performed competitive hybridization by using probes from epithelial Hydra and normal Hydra to identify genes that are expressed in normal animals but lacking in epithelial Hydra. We identified 151 genes expressed differentially in normal Hydra. Studies of these genes indicate that many encode “novel” proteins. Thus, the evolutionary diversification of differentiated cell types in Hydra was accompanied by the evolution of lineage-specific genes.

Results and Discussion

cDNA Microarray Analysis.

The cDNA microarray that we constructed in this study includes 6,528 individual genes derived from a normal Hydra strain, H. magnipapillata 105. This array was screened with RNA probes from epithelial and normal animals. The differential gene expression between normal and epithelial Hydra was evaluated by using a statistical algorithm, SMA (Statistical Microarray Analysis; http://stat-www.berkeley.edu/users/terry/zarray/Software/smacode.html). Genes whose expression was >2-fold higher in normal Hydra than in epithelial Hydra were provisionally identified as I-cell specific. By using this criterion, 151 genes were identified from five independent arrays as candidates for I-cell-specific genes. An additional 19 genes were identified that had 2-fold greater expression in epithelial Hydra than in normal Hydra. These genes will be described elsewhere.

We examined the expression patterns of all 151 selected genes by whole-mount in situ hybridization. One hundred twenty-four genes gave significant signals above background. By searching the Hydra genome assembly that has been annotated by the DOE Joint Genome Institute (Berkeley, CA), we were able to identify full-length or nearly full-length coding sequences for 115 genes. In summary, 51 were nematocyte-specific, 21 were gland cell-specific, 9 were nerve cell-specific, and 5 genes were I-cell/germ-line cell-specific. Surprisingly, 29 of the selected genes were expressed specifically in epithelial cells (see below). All data including gene sequences and expression patterns can be viewed in the database as a primary resource for the biological community to study the genetic characteristics of different cell types in Hydra (see Molecular Database of Hydra Cells, http://hydra.lab.nig.ac.jp/hydra/).

Nematocyte-Specific Genes.

Hydra contains four types of nematocytes (stenotele, holotrichous isorhiza, atrichous isorhiza, and desmoneme) (20). In general, the differentiation of nematocytes proceeds as follows: growth (capsule wall and tubule formation), maturation (tubule invagination, spine assembly, capsule wall hardening and poly-γ-glutamate polymerization), migration (cnidocil apparatus formation), and deployment (septate junction formation and cyst positioning) (21, 22). Using the array technique and in situ hybridization, we have identified 51 nematocyte-specific genes that are expressed at distinct differentiation stages or in a particular type of nematocyte (Table 1). It is not surprising that a large proportion of the genes identified here are nematocyte-specific, because nematocytes comprise approximately half of the total cell population in Hydra (23). Many of the encoded proteins have N-terminal signal peptides, suggesting that they are synthesized on the endoplasmic reticulum (ER) and transported to the post-Golgi vacuole in which capsules are formed. Some of these, e.g., minicollagens and toxin proteins, also contain a short “nematocyte sorting signal” that is removed from the mature protein in capsules (24). Some of the signal-peptide-containing genes may not be incorporated into capsules but sorted to the cell membrane. Finally, there are nematocyte-specific genes, which lack N-terminal sorting signals and appear to encode cytoplasmic proteins specifically expressed in nematocytes.

Table 1.

Nematocyte-specific genes of hydra identified by cDNA microarray and whole-mount in situ hybridization

Clone name	UniGene	| Accession no. | homologous gene of BLASTP search	E-value	SP*	Cnidarian-specific†	Expression localization
Capsule gene
hmp_09910	Hma.940	| AAC39121 | Hydra magnipapillata]	0.0	+	+	Differentiating nematocyte
hm_03768	Hma.6702	| CAA43379 | mini-collagen [H. magnipapillata]	1e-11	+	+	Differentiating nematocyte
hm_02081	—	| CAA43380 | collagen-related protein 2 (fragment) [H. magnipapillata]	2e-10	+	+	Differentiating nematocyte
hydmg014 secrev_095	—	| CAA43381 | collagen-related protein 3 precursor [H. magnipapillata]	4e-12	+	+	Differentiating nematocyte
hmp_01875	Hma.5797			+	+	Differentiating nematocyte
hmp_07679	Hma.1712	| BAA06407 | mini-collagen [Acropora donei]	8e-18	+	+	Differentiating nematocyte
hydmg_008 secrev_016	Hma.2229			+	+	Differentiating nematocyte
hmp_18133	Hma.5361			+	+	Differentiating nematocyte
hmp_11623	Hma.933			+	+	Differentiating nematocyte
hmp_01703	—	| CAA43380 | collagen-related protein 2 (fragment) [H. magnipapillata]	2e-14	+	+	Differentiating nematocyte
hmp_17396	—			+	+	Differentiating nematocyte
hmp_02656	Hma.8906			+	+	Differentiating nematocyte
hm_02844	Hma.1803			+	+	Differentiating nematocyte
hm_02181	Hma.3707	| AAF78087 | protein disulfide isomerase ER-60 [Takifugu rubripes]	1e-16	+	−	Differentiating nematocyte
hm_02634	Hma.3432	| AAH75461.1 | γ-glutamyltranspeptidase 1 [Xenopus tropicalis]	9e-29	+	−	Differentiating nematocyte
hydmg_006 secrev_019	Hma.1619	| NP_801012.1 | carbonic anhydrase [Vibrio parahaemolyticus RIMD 2210633]	7e-22	+	−	Differentiating stenotele
hm_02290	Hma.1831	| XP_548098.2 | similar to FK506-binding protein 10 precursor (peptidyl-prolyl cis-trans isomerase) [Canis familiaris]	8e-34	+	−	Differentiating nematocyte and epithelial cells around tentacle base
hm_04121	Hma.4708	| AAV41095 | physcomitrin [Physcomitrella patens]	3e-06	+	+	Differentiating isorhiza
hmp_11024	—	| BAC45007 | PsTX-20A [Phyllodiscus semoni]	9e-08	+	+	Differentiating stenotele
hmp_21268	Hma.1520	| BAB85217 | PR-1 like protein [Volvox carteri f.nagariensis]	6e-15	+	+	Late differentiating nematocyte
Novel capsule gene
hmp_08654	Hma.1833			+	+	Nematoblast
hm_04447	Hma.8919			+	+	Nematoblast
hmp_15565	Hma.1043			+	+	Nematoblast
hmp_08480	Hma.4149	| BG924950.1 | HyTSRp1 protein [Hydra vulgaris]	1e-10	+	+	Differentiating nematocyte
hmp_22241	Hma.4697			+	+	Differentiating nematocyte
hmp_02020	Hma.4931			+	+	Differentiating nematocyte
hydmg002bw_25	—			+	+	Differentiating nematocyte
hmp_09895	Hma.1654			+	+	Differentiating stenotele
hm_02137	Hma.1900	| ABK35136.1 | rendezvin [Lytechinus variegatus]	3e-10	+	+	Differentiating isorhiza
hm_02450	Hma.3102	| XP_971465.1 | similar to tubulin tyrosine ligase-like protein 2 [Ribolium castaneum]	4e-18	+	+	Differentiating isorhiza
hmp_08269	Hma.4045			+	+	Late differentiating nematocyte
hmp_07910	Hma.9335			+	+	Stenotele
hmp_17528	Hma.9348			+	+	Holotrichous isorhiza
hmp_20434	Hma.3901			+	+	Holotrichous isorhiza
Sensory apparatus
hmp_08523	Hma.4022	| NP_523742.2 | lamin C CG10119-PA [Drosophila melanogaster]	2e-21	−	+	Nematocyte
hm_04087	Hma.8965	| NP_523742.2 | lamin C CG10119-PA [D. melanogaster]	2e-20	−	+	Nematocyte
hmp_00406	Hma.1413	| AAG15328.1 | β-tubulin [Chionodraco rastrospinosus]	e-100	−	−	Migrating nematocyte
hydmg019bw_14	—	| NP_055771 | centrosomal protein 164kDa [Homo sapiens]	2e-35	−	−	Late differentiating nematocyte
hmp_08051	Hma.1566	| AAD00259.1 | calbindin D28k [Xenopus laevis]	2e-20	−	−	Migrating stenotele and desmoneme
hmp_10958	Hma.1341	| Q27052 | flagellar calcium-binding protein [Trypanosoma rangeli]	8e-22	−	−	Migrating nematocyte
Unknown
hm_03857	Hma.624	| AAA65698.1 | ADP-ribosyl cyclase precursor [Aplysia californica]	1e-10	−	+	Subpopulation of nematoblast
hmp_20753	Hma.7753			−	+	Subpopulation of nematoblast
hmp_07718	—			−	+	Nematoblast at early stage
hmp_13584	—			−	+	Nematoblast at early stage
hm_02549	—			−	+	Differentiating nematocyte
hmp_06864	Hma.4958	| ZP_01459247 | inositol 2-dehydrogenase [Stigmatella aurantiaca DW4/3–1]	1e-62	−	−	Differentiating nematocyte
hmp_16781	—			−	+	Differentiating nematocyte
hmp_02578	—			−	(+)	Differentiating nematocyte
hm_04093	—			−	+	Stenotele and differentiating desmoneme
hmp_07578	Hma.3700			−	+	Stenotele at body column
hm_02330	—			−	+	Holotrichous isorhiza

*Signal peptide (SP): +, gene encoding a signal peptide; −, gene without a signal peptide.

^†Cnidarian-specific: +, gene with homologues (E-value <1e-10) only in cnidarian species; −, gene with homologues (E-value <1e-10) in other phyla. (+), homologue is not present in the current nr protein database but is detected in invertebrate genomes.

A striking feature of the list of nematocyte-specific genes is that only 18% (9 of 51) of nematocyte-specific genes have orthologues in other species based on BLASTP (Table 1). These include putative capsule-specific enzymes like protein disulfide isomerase (PDI) and γ-glutamyltranspeptidase as well as known cytoplasmic proteins such as two calcium-binding EFh domain proteins, tubulin and the centrosomal protein [supporting information (SI) Table 2]. The great majority of nematocyte-specific genes (42 of 51) lacked orthologues in other metazoan phyla, although in some cases, they contained sequence motifs found in known proteins. For example, minicollagens contain GlyXY collagen tripeptide domains, hm_02450 encodes a galactose lectin domain and a GlyXY collagen tripeptide domain, and hmp_21268 encodes an SCP domain. Repeating these searches with more sensitive PSI-BLAST (SI Table 2) did not identify a significant number of additional orthologues in the protein database. Because not all protein-coding genes are compiled in the nr proteins database, we also carried out additional BLAST searches on the completed genomes of two fungi (Aspergillus and Candida), the ciliate Tetrahymena, the parasitic protozoa Trypanosoma, the slime mold Dictyosterium, the choanoflagellate Monosiga, the sponge Reneira, the flatworm Schmidtea, the mollusk Lottia, the annelid Capitella, and two deuterosomes (Strongylocentrotus and Branchiostoma). These searches (summarized in SI Table 3) identified one additional homologue not yet compiled in the nr protein database. This, however, does not significantly alter the conclusion: the great majority (41 of 51) of nematocyte-specific genes identified in the array appear to be cnidarian-specific. Of these, 17 lack a clear homologue in Nematostella and thus appear to be Hydra-specific.

Nematocyte-specific genes encode structural proteins that are essential for assembling the rigid capsule wall, the internal tubule, the spines, the toxin proteins, and poly-γ-glutamate required for the explosive extrusion as well as genes encoding the sensory apparatus (cnidocil) of mature nematocytes. Genes involved in capsule formation are expressed in nests of differentiating nematocytes during the first half of the differentiation process. In addition to genes for known minicollagens and spinalin (25, 26), our screen identified nine minicollagen genes (hmp_01875, hmp_07679, hydmg_008secrev_016, hmp_18133, hmp_11623, hmp_01703, hmp_17396, hmp_02656, and hm_02844; Fig. 1A and Table 1). All these genes had signal peptides and short nematocyte sorting motifs. We identified a protein disulfide isomerase homologue (hm_02181) with a signal peptide and nematocyte sorting motif that may be required for disulfide bond isomerization during hardening of the capsule wall (Fig. 1B) (27, 28). Finally, we identified a γ-glutamyltranspeptidase (GGT) homologue (hm_02634), which may be involved in synthesis of poly-γ-glutamate in the capsule matrix (29).

Fig. 1.

Expression of nematocyte-specific genes using whole-mount in situ hybridization. Whole animals are shown on the left, and high-magnification images showing the expression signal in individual cells are shown on the right. (A) hmp_11623, a previously uncharacterized minicollagen expressed in differentiating nematocytes. (B) hm_02181, a protein disulfide isomerase expressed in differentiating nematocytes. (C) hm_04121, a toxin protein expressed in differentiating isorhizas. (D) hmp_21268, a toxin protein expressed in differentiating nematocytes. (E) hmp_17528, a cnidarian-specific capsule protein expressed in mature holotrichous isorhizas. (F) hmp_08523, a lamin-like protein expressed in all mature nematocytes. (G) hmp_08051, a calcium-binding protein expressed in migrating nematocytes. (H) hmp_07578, a cnidarian-specific protein with unknown function expressed only in stenoteles and desmonemes. All clone names refer to Table 1. S, stenotele; HI, holotrichous isorhiza; I, isorhiza; D, desmoneme; dS, differentiating stenotele; dI, differentiating isorhiza; dD, differentiating desmoneme. (Scale bars, 50 μm.)

Toxins are commonly found in the nematocysts of cnidarians. We found two homologues (hm_04121, hmp_11024) of equinotoxin (30), a member of the actinoporin family from sea anemones as well as a pathogen-related protein (hmp_21268). One toxin (hm_04121) is localized in the isorhizas, which were previously thought to contain no toxins at all (Fig. 1C and Table 1). A second toxin (hmp_11024) is present in stenoteles. Both toxins are believed to be cytolysins and have hemolytic activity (31). A third gene (hmp_21268) encodes a well conserved SCP domain and an ShK toxin domain that was first found as a short polypeptide in sea anemones (Fig. 1D and SI Table 2) (32).

In addition to these genes with known functions, we identified 14 more genes that had signal peptides and were expressed in differentiating nematocytes, but which had no significant homologue in the nr protein database and the genomic databases (Table 1 and SI Table 3). These appear to be “novel” capsule proteins, although some could represent nematocyte membrane proteins. Several had short proline-rich stretches, which are also present in minicollagens, and thus may constitute novel structural proteins in the capsule wall. Two genes were expressed specifically in stenoteles, four genes were expressed in isorhizas (Fig. 1E), and the rest were expressed in all types of nematocytes. This indicates that each type of nematocyte has some unique proteins, which presumably contribute to the particular function of that nematocyte type, as well as some proteins that are shared with other nematocyte types.

In addition to the proteins listed above, all of which contain signal peptides and, hence, can be incorporated into nematocyte capsules, we also identified 17 genes (sensory apparatus and unknown) encoding proteins that lack signal peptides but which are nematocyte specific (Table 1). These proteins are presumably cytoplasmic. Several are expressed in migrating nematocytes after capsule formation is complete and also in mature nematocytes mounted in the tentacles. Formation of the cnidocil apparatus takes place at this stage of differentiation (22). The cnidocil is a mechanosensory organelle extending from the apical surface of nematocytes mounted in the ectoderm. The two lamin-like genes (hmp_08523 and hm_04087) (Fig. 1F) and the β-tubulin gene (hmp_00406) encode proteins that participate in cnidocil formation. The centrosomal protein homologue (hydmg019bw_14) may be involved in formation of the basal body at the base of the cnidocil. Finally, two genes (hmp_08051 and hmp_10958) encode small calcium-binding proteins, which may be involved in regulating intracellular calcium levels and exocytosis of the mature nematocyte (Fig. 1G). Interestingly, all BLAST matches of hmp_10958 occurred as proteins in protists (SI Table 3), possibly suggesting that either this gene is originated from protist or that the orthologues of hmp_10958 have evolved divergently in other metazoans. Most of the remaining 11 nematocyte-specific genes had no homologues in nr protein database or in other phyla, and we are unable to determine their functional role in nematocyte differentiation (Fig. 1H, Table 1, and SI Table 3).

Gland Cell-Specific Genes.

Gland cells and mucous cells are secretory cells located in the endoderm of Hydra. They secrete proteolytic enzymes and mucopolysacchrides (33) into the gastric cavity of Hydra. In this study, a total of 21 genes were expressed in either gland or mucous cells (SI Table 4). Almost all these genes had clearly recognizable signal peptides, suggesting that they encode secreted proteins. Five gland cell-specific genes (hmp_01803, hmp_03838, hmp_17122, hm_02146, and hmp_14705) showed clear homology to annotated metazoan genes. These encoded extracellular proteases. Hmp_08169 encodes antistasin (34), a gland cell-specific protease inhibitor. Three more genes encoded homologues of nonenzymatic lysosomal proteins (hmp_09905, hmp_22174, and hmp_00993), yet they are relatively divergent from many others including those from Nematostella (SI Table 3). Finally, 11 of 21 gland- and mucous cell-specific genes encoded sequences with no known homologues in the nr protein database (SI Table 4). Half of these genes also lacked a clear Nematostella homologue and thus appear to be Hydra-specific (SI Table 3).

The gland cell-specific genes can be further grouped into subtypes according to their gene expression patterns. For instance, the expression of hmp_14705 (a homologue of metalloproteinase) was strong in the upper body column but gradually less toward the peduncle (Fig. 2A); whereas, hm_02146 (a homologue of serine protease) was expressed primarily in the peduncle (Fig. 2B). This indicates that gland cells are not identical but establish a position-specific expression pattern along the body axis. Recent reports also showed similar expression patterns in gland cells (35, 36).

Fig. 2.

Expression of gland cell-specific genes using whole-mount in situ hybridization. Whole animals are shown on the left, and high-magnification images showing the expression signal in individual cells are shown on the right. (A) hmp_14705. (B) hm_02146. All clone names refer to Table 2. (Scale bars, 50 μm.)

Nerve Cell-Specific Genes.

Nine genes isolated from the microarray were nerve cell-specific. Six of them encoded neuropeptides and were expressed in ganglion cells (SI Table 5). Various neuropeptides have been isolated from Hydra by using systematic screening for peptide signaling molecules (37–39), and four sequences reported in this study encode such known neuropeptides (SI Fig. 3D). However, additional paralogues of RFamide (hmp_01824) and hym-176 (hmp_10112) neuropeptides were identified here. So far, all nerve cell-specific genes isolated from Hydra encode neuropeptides, and more than half of the genes identified in this study also encode neuropeptides, indicating neuropeptides play a key role in Hydra neurotransmission. A few of the neuropeptides in Hydra such as RFamide and LWamide/MWamide have homologues in metazoans. However, other neuropeptides such as hym-176 (KVamide) and hym-355 (RGamide) have no metazoan homologues so far.

Three additional nerve cell-specific genes, hmp_09790, hmp_11958, and hmp_13646, were identified in this study (SI Table 5). Expression of hmp_09790 was localized in sensory cells, and the tentacles of polyps were highly stained by whole-mount in situ hybridization (SI Fig. 3A). No homologue for hmp_09790 could be identified in the protein databases, and a search against the genome databases also resulted in no match (SI Table 3). Hmp_11958 was expressed in both sensory and ganglion cells (SI Fig. 3B). Its predicted protein sequence can be cleaved at KR residues into putative peptide signaling molecules, although such peptides have not yet been identified in Hydra. We suspect that both genes above may function as neurotransmitters involved in signal transduction or as neurohormones involved in developmental processes. Hmp_13646 encodes an unknown protein of 222 aa. The first 123 aa show homology to the ligand-binding domain of the ligand-gated ion channel pHCl-C (AAX11177) of Drosophila melanogaster, indicating hmp_13646 may also function in neurotransmission. Expression of hmp_13646 is localized in ganglion cells (SI Fig. 3C).

The apparent enrichment of neuropeptide genes among nerve cell-specific genes suggests that neuropeptide signaling systems play an important role in the neural network of Hydra. RFamide and LWamide neuropeptides are known from several other metazoan phyla. The genes encoding the precursors to these neuropeptides are not well conserved, and, hence, no homologues in other phyla are listed in SI Table 3. This is also true for the Nematostella homologues. Although initially surprising, because both Hydra and Nematostella are cnidarians, this is probably because of the fact that these two cnidarian lineages have been separated for >500 million years. In addition to the evidence for neuropeptide signaling shown here, there is genetic evidence that Hydra possess glutamatergic and GABAnergic systems signaling systems (40).

Interstitial Cell-Specific Genes.

Five genes identified in the array experiment were expressed specifically in the I-cells and germ-line cells. They are hmp_06154, hm_03505, hmp_06895, hmp_22205, and hydmg_005secrev_068 (SI Table 6 and SI Fig. 4). Hmp_06154 is a putative homologue of cyclin C that is associated with RNA polymerase II and involved in the control of cell cycle (41). Hm_03505 is a nuclear localized protein containing a domain called nucleoplasmin (SI Fig. 4A). Human nucleoplasmin 2 is a chromatin decondensation protein that is required for nucleolar organization and embryonic development (42). Hmp_06895 is related to another nuclear protein, thymocyte nuclear protein 1, that is highly conserved among vertebrates, plants and bacteria (SI Fig. 4B). Hmp_22205 (SI Fig. 4C) has no significant homologue in the current database. Hydmg_005secrev_068 (SI Fig. 4D and SI Table 2) is a cytoskeletal protein that may be required for the elongation of germ-line cell during mitosis because it contains a spectrin domain and a growth-arrest-specific protein 2 domain (43). It is conserved throughout metazoans. The conservation of stem cell-specific genes between Hydra and human suggests that they may play an important role in metazoan stem cell behavior.

Epithelial Cell-Specific Genes.

Unexpectedly, some genes selected in the screen were not expressed in the I-cell lineage but in epithelial cells (see below). Close examination of the array data indicated that these genes were indeed less strongly expressed in animals lacking the I-cell lineage. This suggests that these genes are up-regulated in animals containing the I-cell lineage and hence that they could play a role in establishing a suitable environment for I-cells in such intact animals. These genes were expressed in either ectodermal or endodermal epithelial cells or both (SI Table 7). The expression pattern in ectodermal epithelial cells was not uniform along the body column. For example, tyrosine kinase receptor homologues exhibited various regional expression patterns: in one case expression was localized in the bottom half of body column and the peduncle (SI Fig. 5A), whereas in another case it was localized around the base of tentacle and in the hypostome (SI Fig. 5B). Expression patterns in endodermal epithelial cells also were specific and regional. For example, endodermal epithelial cells of the hypostome expressed hmp_02857 and hmp_06947 (SI Fig. 5C). Similarly, the expression of hmp_09678 was strongly localized to both the hypostome and the peduncle (SI Fig. 5D). Although the number of epithelial cell specific genes selected in the screen was not large, there appears to be a bias for secreted and transmembrane proteins. Several of these are novel receptor tyrosine kinases that lack high sequence similarity to other metazoan receptor tyrosine kinases. Interestingly, none of them were detected in the genome of Nematostella, a cnidarian lacking the I-cell lineages (SI Table 3) (44). It is interesting to speculate that these receptors are involved in cross talk between the I-cell lineage and epithelial cells to coordinate the behavior of these two cell populations in intact animals.

Emergence of Cnidarian-Specific Genes and Specialized Cell Types.

A striking feature of the present results is the high proportion of cell type-specific genes, which appear to lack homologues in other metazoan phyla. Some of these genes may even be Hydra-specific because no homologues were found in Nematostella. This suggests that the emergence of specialized cell types in Hydra was accompanied by the evolution of “novel” proteins. Although this was the expected result in the case of nematocytes, which are a novel cell type unique to Cnidaria, it was not expected for gland cells and nerve cells, which appear to be similar to the corresponding cells types in higher metazoans. This latter observation suggests that these cell types in cnidarians have features that are unique to cnidarians, despite their morphological similarity to equivalent cell types in other phyla.

The high proportion of cnidarian-specific genes in Table 1 raises the question of the origin of these genes. To answer this question, we searched the complete genomes or genome reads for two fungi (Aspergillus and Candida), four protists (Tetrahymena, Trypanosoma, Dictyosterium, and Monosiga), and the sponge Reneira. For most nematocyte genes (41 of 51), approximately half the gland cell genes (11 of 21), and all of the nerve cells genes (9 of 9), no homologues could be found (SI Table 3). Thus, the origin of more than half the lineage-specific genes is presently not clear.

It is important to note that the conclusions cited above are based on highly expressed genes. The mRNA for the screen was isolated from whole animals. Weakly expressed transcripts are poorly represented in this total mRNA population. Similarly, transcripts from less-abundant cell types such as nerve cells are also not well represented in total mRNA preparations. Hence, most genes identified in the screen represent strongly expressed genes. This may be the reason most of the identified genes encoded secretory proteins. Capsule proteins in nematocytes, secreted enzymes in gland and mucous cells, and neuropeptides in nerve cells are all secreted cell products, which are synthesized in large quantities in their respective cell types.

Smaller screens for cell type-specific genes have been carried out previously in Hydra (25, 26, 34). The nematocyte-specific minicollagens and spinalin were identified in such screens, as was the gland cell-specific protease inhibitor antistasin. A large-scale screen for embryogenesis-specific genes also revealed large numbers of previously unidentified genes (45) and many of these encoded secreted proteins as found here.

A second striking feature of the results was the isolation of genes expressed specifically in epithelial cells. Although the screen was designed to identify genes specific to the I-cell lineage, 25% of the selected genes turned out to be epithelial cell-specific when tested by in situ hybridization. Closer examination of the array results showed that expression of these genes was indeed up-regulated in intact animals compared with epithelial Hydra. This suggests that expression of these genes is required in animals containing the I-cell lineage. In this context, it is interesting to note that many of these epithelial cell-specific genes are involved in intracellular signaling and signal transduction. Another example of such signaling is the release from epithelial cells of peptide signals regulating nerve cell differentiation (37). Both observations are consistent with the idea that, in multicellular animals, there is cross talk between different cell lineages to maintain constant proportions. This is particularly important for Hydra tissue, which is constantly expanding during asexual growth and where it has been shown that cell type proportions remain constant during growth (2). Raff (46) has proposed that cellular homeostasis in multicellular animals is mediated by growth/survival factors secreted by one cell type, i.e., epithelial cells, and required for survival by a second cell type, i.e., I-cells. Suggestive evidence that signaling via receptor tyrosine kinases may play a role in cellular homeostasis in Hydra has come from specific inhibition experiments. Treatment of Hydra with wortmannin, an inhibitor of PI-3-kinase, which blocks receptor tyrosine kinase signaling, leads to massive apoptosis (47).

Materials and Methods

Hydra Strains.

Hydra magnipapillata strain 105 were cultured at 18°C and fed with Artemia larvae three times per week. Epithelial Hydra were derived from strain 105 by treatment with colchicine (48) and have been kept in the laboratory for >10 years without changing their phenotype.

Construction of 6.6 Thousand cDNA Array.

6,528 clones were obtained from nonredundant ESTs and subjected to PCR. One of the two primers that were used to amplify the clones had an amino-link modification at the 5′ site. All PCR products were checked by agarose gel electrophoresis, and only those consisting of a single band were used for the array. PCR products were precipitated with 0.3 M sodium acetate and 100% isopropanol and then resuspended in 50% DMSO. Slides were printed in-house with an SPBIO spotter (Hitachi Software Engineering, Yokohama, Japan) on γ-aminopropyl silane-coated slides (CMT GAPS; Corning, Corning, NY). After printing, all arrays were incubated in an oven at 80°C for 3 h. Four sets of 15 controls were printed at four locations on the slides. Positive controls included Hydra collagen, 60s ribosomal RNA, elongation factor, choline transporter (weakly expressed gene), and nicotinic acetylcholine receptor α-like-1 (weakly expressed gene). Negative controls included poly(dA)₄₅, luciferase gene, GFP gene, and water. In addition, two exogenous genes, Cab (photosystem I chlorophyll a/b-binding protein) and RCP1 (root cap 1) of Arabidopsis thaliana, were purchased from Stratagene (La Jolla, CA) and used to monitor the signal intensity differences between Cy3 and Cy5.

cDNA Labeling and Hybridization.

mRNAs of epithelial and intact Hydra were extracted and amplified by using a MessageAmp II aRNA kit (Ambion, Austin, TX). By using random hexamers, 2 μg of amplified RNAs were labeled in a reverse-transcriptase reaction with Cy3-dUTP/Cy3-dCTP and Cy5-dUTP/Cy5-dCTP (CyDye; Amersham Biosciences, Piscataway, NJ), respectively, or vice versa. The mRNAs of the external controls (1 ng of each), Cab, and RCP1 was also added to each Cy3- and Cy5-labeling reaction. After reverse transcription, Cy3- and Cy5-labeled cDNAs were pooled, purified by using Microcom YM-30 (Amicon; Millipore, Billerica, MA), and denatured at 95°C for 2 min. Hybridization was carried out in 5× SSC and 0.5% SDS at 65°C for 40 h. The array slide was treated with a series of washes as follows: 2× SSC, 0.1% SDS; 0.2× SSC, 0.1% SDS four times; 0.2× SSC. To monitor the background intensity and verify the integrity of the array, the self–self hybridization was carried out by using labeled cDNAs of intact Hydra.

Scanning of the Arrays.

Cy3 and Cy5 intensity signals from five independent arrays (in which two were performed by a dye-swapping experiment) were scanned by using ScanArray BS2000 (GSi Lumonica, Wilmington, MA) and quantified by using Quantarray (GSi Lumonica). To identify genes that were differentially expressed in normal Hydra as compared with epithelial Hydra, all array data were normalized and processed by a Lowess print-tip normalization method computed by using the SMA (Statistics for Microarray Analysis) package (49). The final expression differences between intact and epithelial Hydra were denoted as log₂ intensity ratio, M = log₂(R/G), where R and G were the two different dyes.

Sequence Homology Search and Signal Peptide Identification.

One hundred fifteen Hydra ESTs (accession nos. listed in SI Table 8) were queried by using BLASTN against the Hydra genome at the DOE Joint Genome Institute (Berkeley, CA). Either a full-length or longer sequence detected from the genomic sequence was translated into amino acid sequence. By using BLASTP (50), amino acid sequences were searched against the GenBank nr protein database (February, 2007). Signal peptide of protein sequence was predicted by using SignalP 3.0 (www.cbs.dtu.dk/services/SignalP/).

Whole-Mount in Situ Hybridization.

The protocol used for all in situ hybridization experiments was originally described by Grens et al. (51). The concentration of riboprobe used for the hybridization varied from 50 ng/ml to 200 ng/ml.