INTRODUCTION

Immune recognition is widespread in invertebrates (1), and allograft rejection has been described in ascidians (2, 3), earthworms (4, 5), molluscs (6), coelenterates (7-9), sponges (10), echinoderms (11-13), and insects (14, 15). Allogeneic reactions have also been observed in vitro with isolated cells, at least in urochordates (16), and in sponges (17-19), indicating that invertebrate allorecognition is a direct cellular response.

Ig domains are found in a vast number of extracellular components, generally involved in intercellular recognition, including MHC¹ and cell adhesion molecules (20, 21). It is commonly accepted that the Ig superfamily has been derived by gene duplication and divergence from a primordial ancestor (22, 23), whose descendants are most likely to be found on the surface of invertebrate cells involved in cell/cell recognition in development or in allorecognition phenomena (24). The existence of Ig-like domains has been discovered in squid (25), insects (26-28), tunicates (29), and sponges (30). A receptor tyrosine kinase from the marine sponge Geodia cydonium has been described to contain a highly polymorphic Ig-like domain (31), although its involvement in sponge immunity has not been demonstrated. Proteins related to the Ig superfamily have been found even in slime molds (32) and yeast (33). However, as reasonably argued by several authors (34, 35), the differences in physiology between vertebrates and invertebrates demand extreme prudence before attempting to force invertebrate induced recognition into mammalian models of immunity (36). The existence of Ig domains in invertebrates does not imply the existence of an Ig domain-based histocompatibility system like the vertebrate MHC. Indeed, no Ig and MHC sequences have been described in any invertebrate, suggesting that these molecules, and therefore the vertebrate-like immune system, are late evolutive events.

Since invertebrates have well developed histocompatibility systems and since all evidence at this point in time suggests that they do not have the same molecular components that define the histocompatibility systems of vertebrates, then which molecules do invertebrates use? Are these molecules or their descendants still present in more evolved animals, or did they disappear when the vertebrate MHC arose? If present, are they still involved in some aspect of histocompatibility, or did they evolve to perform other functions in the cell? Whether vertebrate and invertebrate immunity share a common ancestry or are the result of convergent evolution remains an open question, and despite the scattered knowledge about primitive immune systems described above, there is an absolute lack of protein sequence information concerning the molecules involved in invertebrate histoincompatibility reactions, as has been pointed out in several reviews (23, 24, 37, 38).

Rothenberg (39) proposed that the primordial genes of the self-recognition system are closely related to the genes responsible for species-specific cell recognition. Sponges represent the lowest metazoan phylum, and although they have a primitive organization, they are capable of rapid allogeneic recognition (40-42). Curtis and Van de Vyver, working with the freshwater sponge Ephydatia fluviatilis, first demonstrated that the aggregation factor promoted aggregation with autogeneic cells but inhibited that of allogeneic cells (43), acting in fact as a histocompatibility molecule. In xenograft rejection studies with sponges of the Geodia genus, an indirect involvement of the AF in histoincompatibility reactions has also been suggested (44). Sponge AFs are extracellular proteoglycan-like complexes involved in species-specific cell recognition and adhesion (45-49). Recently, the sequence of the main protein component of the aggregation factor from the red beard sponge, Microciona prolifera, was deduced from cDNA (50), and Southern blot analysis suggested the existence of several related genes. We have studied this polymorphism in the context of the evolution of invertebrate histocompatibility systems and their possible divergence from primitive cell-cell interaction molecules.

EXPERIMENTAL PROCEDURES

Live specimens of the red beard sponge, M. prolifera, were collected in the Woods Hole area. AF was isolated as described by Misevic et al. (51).

In general, the protocol described by Sajdera and Hascall (52) was followed. To a MAF solution in H₂O, guanidinium hydrochloride and CsCl were added to final concentrations of 4 M and 50% (w/v), respectively. The resulting volume was centrifuged at 10 °C in Beckman quick seal tubes for 12 h at 48,000 rpm in a VTi 65 vertical rotor (Beckman). Fractions were collected, dialyzed against H₂O before electrophoresis, and finally stored at 20 °C.

SDS-PAGE was performed as described by Laemmli (53). Purified AF and dissociative gradient fractions were analyzed in 5% gels stained with combined Alcian blue/silver (54) to visualize glycosaminoglycan-containing bands. Western blots were transferred to a polyvinylidene difluoride membrane (Immobilon, Millipore Corp.) with a semidry electroblotter (JKA-BIOTECH), blocked in 0.1 M Tris-HCl, pH 7.5, 0.5% Tween 20, 1% Triton X-100, 3% bovine serum albumin, and incubated in the presence of 1 µg/ml block 1 or block 2 monoclonal antibodies in blocking solution. The ECL Western blotting detection system (Amersham Corp.) was used to visualize the decorated bands.

Enzyme treatments were done at 37 °C during 5 h in solutions containing 0.1 mg/ml substrate protein. Digestions with peptide-N-glycosidase F (Boehringer, recombinant N-Glycosidase F from Escherichia coli, 20 units/ml final concentration) were performed in 25 mM potassium phosphate, pH 7.5, 12.5 mM EDTA, 1% 2-mercaptoethanol. To denature the molecules prior to deglycosylation, the samples were heated (100 °C, 3 min) in the presence of 0.1% SDS. Nonidet was then added to a final concentration of 1% to avoid denaturation of the enzyme by SDS. Treatments with proteinase K (Boehringer Mannheim, 20 µg/ml) were in 10 mM Tris-HCl, pH 7.5, 5 mM EDTA, 0.5% SDS. Trypsin (Promega, sequencing grade, 10 µg/ml) was used in 0.1 M NH₄HCO₃, pH 7.7. Treatments with cyanogen bromide (Merck) were according to Hanson and Bentley (55). Amino acid sequencing was performed as described by Fernàndez-Busquets et al. (50).

One-cm-long sponge branches (mean section, 4 mm) were impaled together on a 0.5-mm-thick stainless steel insect pin, and incubated in running sea water at 20 °C for the times indicated. The contact surfaces were about 0.3 cm². The sponge individuals were kept in running sea water cooled to 14 °C for as long as they looked healthy.

Digoxigenin-labeled DNA probes were obtained by PCR and amplified from regions lacking internal DraI sites. Primers I-S1 (5-TTTACTGTTCCGATTTACGT-3, sense) and I-A (5-CGGTATTGAGGTGGTTGGTT-3, antisense) were used with the cloned cDNA of form MAFp3A1 as template to generate the MAFp3-specific probe I (~1 kb). The primer I-S2 (5-TACAAAGTGGACAGCGCCTTTTCAG-3, sense) was used in combination with I-A in the PCR amplifications shown in Fig. 6. Primers II-S1 (5-CGTGGCCGAAGAATGGAAAACTTC-3, sense) and II-A (5-ACTTTCGTTACGGGTCACATTGGG-3, antisense) were designed to generate probe II (~1.9 kb), using as template a cloned fragment previously amplified from genomic DNA. Due to the existence of numerous introns and relatively short exons in region II, cDNA-derived probes were found to give a poor signal. The extension of sequence E in the 5 direction was amplified from total cDNA with primers II-S2 (5-GGCTGGACCTTCAGCACCTGACTC-3, sense) and E-A (5-TGATCCTCCAGTATTCAAAGCAAT-3, antisense). The ~1-kb collagen probe was amplified from M. prolifera total cDNA and spanned the region between Gly-41 and Pro-380 of the sequence reported by Aho et al. (56). Alkali-labile DIG-dUTP (Boehringer Mannheim) was included in the reaction mixture with a DIG-dUTP:dTTP of 1:3. The resulting probes were analyzed in a 1% agarose gel, and after the presence of a single labeled band of the expected size was confirmed, the PCR reaction volume was stored at 20 °C and used at a working dilution of 1:2000 in hybridization buffer. The MAFp3-specific probe for Northern blots was prepared as described by Fernàndez-Busquets et al. (50).

[View Larger Version of this Image (13K GIF file)]

Total RNA from freshly dissociated sponge cells was prepared according to Chomczynski and Sacchi (57). Poly(A)⁺ RNA selection was done according to Sambrook et al. (58). cDNA libraries were made using the ZAP Express^TM cDNA Gigapack II Gold cloning kit from Stratagene. Plaques (10,000/14-cm plate) were transferred to Hybond-N+ membranes (Amersham Corp.), and screened with probe I. Hybridization, stringency washes, and detection were performed following the instructions in the DIG DNA labeling kit from Boehringer Mannheim. From the positive plaques, pBK-CMV was excised from the ZAP Express^TM vector using the in vivo excision protocol of the ExAssist/XLOLR system (Stratagene). DNA inserts were sequenced with an A.L.F. automatic sequencer (Pharmacia), using fluorescein-labeled primers.

Taq DNA polymerase, dNTPs, and PCR buffer were purchased from Boehringer Mannheim. PCR reaction conditions were those specified by the manufacturer. The products were analyzed in 0.5 × TBE-agarose gels containing 0.5 µg/ml ethidium bromide and visualized under UV light. Sponge genomic DNA was isolated as described by Fernàndez-Busquets et al. (50). When needed as template for PCR, ~60 pg of genomic DNA or ~20 pg of poly(A)⁺ RNA-derived cDNA were used per 1 µl of reaction. Human genomic DNA from placenta, chorionic membrane, and umbilical cord of different individuals was purchased from Sigma. For Southern blots, genomic DNA was digested with DraI or HaeIII (Boehringer Mannheim), under the conditions specified by the supplier, for 4 h with constant gentle shaking. The digested DNA was electrophoresed in 1% agarose gels, transferred to a positively charged nylon membrane (Boehringer Mannheim) by capillary transfer, and hybridized with the DIG-labeled DNA probe overnight at 42 °C. Human tissue Northern blots (CLONTECH) were hybridized to the DIG-labeled RNA probes overnight at 68 °C. Stringency washes and detection were performed as described above, with a final stringency wash of 0.1 × SSC, 0.1% SDS at 68 °C for both Southern and Northern blots. Intron positions were determined from PCR-amplified fragments using sponge genomic DNA as template and primers derived from the cDNA sequences. The PCR products were analyzed in 1% agarose gels and excised, and the DNA was extracted (QIAEX II Gel Extraction Kit, QIAGEN), subcloned in pCR^TMII (TA Cloning Kit, Invitrogen), and sequenced as described. The intron positions were then deduced after comparison with the cDNA sequences.

RESULTS

A cDNA library constructed from a mixture of poly(A)⁺ RNAs from at least two different sponge individuals was screened with the MAFp3-specific probe I. After sequencing some of the positive clones, it became evident that there existed several MAFp3 forms. To investigate whether each sponge individual possessed only one or several forms, poly(A)⁺ RNA was then separately isolated from six M. prolifera individuals and used to construct six oligo(dT)-cDNA libraries, each representing a single individual. The libraries were screened with the MAFp3-specific probe I, and a total of 25 positive clones were selected among those containing the longest inserts, identifying 10 different sequences whose cDNA-deduced protein sequences are presented in Fig. 1. In a previous work (50), we described that MAFp3 is a 35-kDa protein that seems to be translated from a variety of transcripts ranging from 1.2 to ~9 kb, assuming that the different RNA species are extended in the 5-end and that the Northern blot results do not reflect unspliced introns. Although the longest clone presented in this work has ~2.5 kb, we have specified in Fig. 1A that region II can have up to ~8 kb. The open reading frames of the MAFp3-coding region and its extension in the 5 direction are continuous, but we decided to represent them separately in Fig. 1A as region I (MAFp3) and region II (5-ORF) because we have not found evidence of a functional protein containing the whole continuous sequence of both regions. In fact, it is most likely that they are translated as a single polypeptide, which is then processed into two proteins. Moreover, as we show later, the products of regions I and II have characteristics that make them different enough to be considered independently.

Different protein sequences derived from ORFs in the cDNAs of MAFp3. A, schematic drawing of the mRNA of MAFp3, indicating the relative positions of regions I (coding for MAFp3) and II, both within a single uninterrupted ORF. The positions of the sequences corresponding to probes I and II are shown. B-D, amino acid sequences deduced from incomplete cDNA clones. The represented sequences are the longest obtained for each form in the screening of the cDNA libraries. In B, MAFp3 forms B, C, D, and E are compared with form A with dots indicating identity and dashes indicating gaps. Variants of A, B, C, and E are identified as A1, A2, A3, B1, B2, C1, C2, C3, E1, and E2. Sequence A1 corresponds to the MAFp3 cDNA reported by Fernàndez-Busquets et al. (50). To simplify the figure, only those residues that change are indicated in the variants. In panel C, forms B and C are so similar to form A that they are treated as variants. The arrowhead indicates the position where region EIII (panel D) is inserted into form E according to the scheme. Amino acid residues are numbered on the right, separately for regions I and II. The arrows indicate intron positions. For region II, the introns are only indicated on sequence A, although they have also been determined for sequence D and found to be identically placed. Stars and squares mark cysteine residues and stop codons, respectively. The binding positions of the primers used in this work are indicated above sequence MAFp3A (I-S1, I-S2, I-A, II-S1, II-S2, and II-A) and below sequence MAFp3E (E-A). The underlined and double underlined sequences in panel D are likely to have originated as duplications of the sequence underlined in panel C. In boldface type is shown the peptide identified after cyanogen bromide/trypsin digestion of MAF.

[View Larger Version of this Image (30K GIF file)]

According to their length, five different MAFp3 forms have been identified (MAFp3A, -B, -C, -D, and -E), and variants for some of them have also been found (Fig. 1B). The only clone containing sequence E was short, starting with Val-59 in region I. For a better comparison, we decided to extend it, performing PCR on a total cDNA template using primers E-A (antisense), specific for MAFp3E, and II-S2 (sense), derived from a region conserved in forms A and D. MAFp3A1 corresponds to the sequence described by Fernàndez-Busquets et al. (50) and goes beyond it in the 5 direction. Assuming that the amino termini for the mature proteins of all the forms begin at the same position as that of MAFp3A (position 1 in region I), their sizes range from 290 to 333 amino acids (31.4-36.8 kDa). While MAFp3A, -B, and -C lack introns, MAFp3D and -E have a single intron centrally placed, of ~350 bp in MAFp3D and ~600 bp in MAFp3E, as revealed in PCR amplifications with sponge genomic DNA as template.

Forms MAFp3A and -B have been isolated from one of the individual libraries, and three libraries were found to contain forms B and C, thus indicating that a single sponge individual has several MAFp3 forms. Polymorphic MAFp3-related bands were observed in Southern blots of sponge genomic DNA isolated from different individuals (Fig. 2A, lanes 1-7), suggesting the existence in the population of different sets of MAFp3 forms. Data base searches did not find any MAFp3-related DNA or protein sequences, but Southern blot analysis revealed the presence of homologous regions in the human genome (Fig. 2A, lane 8), although in this case we did not observe polymorphism in the bands of seven different individuals analyzed. Northern blots of human poly(A)⁺ RNA showed that the MAFp3-specific probe strongly hybridized with a ~4.5-kb transcript found in different tissues, after a stringency wash of 0.1 × SSC, 0.1% SDS at 68 °C (Fig. 2B), suggesting the existence in humans of still unknown proteins with a considerable degree of homology to MAFp3.

[View Larger Version of this Image (44K GIF file)]

The regions II of forms A, B, and C are very similar (Fig. 1C), while MAFp3D and -E are more related to each other than to any other form. Not all of the clones chosen for sequencing were long enough to provide long stretches of region II; therefore, its variability might not be well reflected in Fig. 1. The presence of numerous introns (with sizes between 300 and 650 bp) suggests the existence of recombination events that can provide a source of variability. The region EIII (Fig. 1D) has been found only in sequence E and seems to have originated as a duplication. A long stretch in region II shares 30% identity with the cytoplasmic domain of the Na⁺-Ca²⁺ exchanger protein (50), although in human Southern and Northern blots we did not detect any bands hybridizing with the region II-specific probe II. However, we have found evidence of the existence in M. prolifera of a protein with the sequence of region II, since cyanogen bromide treatment followed by trypsin digestion of nondeglycosylated AF yielded, among other peptides, the peptide XGATSEPFTVR, whose sequence corresponds to residues 61-71 in the region II of forms A, B, and C. This finding supports the idea that other proteins besides MAFp3 will participate in the assembly of the Microciona aggregation factor (50) and indicates that at least two of those proteins are processed together from a common mRNA. We have not identified an initiation codon, and, presumably, the ORF will continue in the 5 direction, suggesting that the protein from region II is a sizable MAF component that escaped detection until now, probably because of its tendency to aggregate after being deglycosylated (50).

The fact that 10 different sequences have been found in the 25 positive clones isolated from the cDNA libraries indicates that MAFp3 is a highly variable molecule. The development of variable region molecules is thought to have been a crucial event in the evolution of primordial immune systems (59), followed by gene rearrangement to provide more diversity. Bodmer (60) suggested that histocompatibility polymorphism may have been an evolutionary byproduct of the need to develop a cell-cell recognition system when multicellular organisms appeared. Since sponges are the most primitive extant multicellular animals, the involvement of a polymorphic protein in sponge cell/cell recognition prompted us to consider a possible relation of MAFp3 or closely related molecules with sponge histocompatibility reactions.

We have used a population of 23 M. prolifera individuals to perform a total of 243 grafting combinations, including allografts and control autografts. Each graft was checked after 12 h and classified as fusion or rejection (Table I). In rejecting grafts, the bright red tissues of two opposing individuals were separated by a ~0.5-mm-wide yellowish line (Fig. 3A, center), formed by the deposition of a collagen layer and the massive migration of gray cells, which are suspected to be involved in the sponge immune response (42). The speed of this reaction reproducibly varied depending on the individual combination, and while in some cases rejection was clearly visible after 6 h, other pairings did not react before 24 h. Such a range of responses has also been observed in grafting experiments with the tropical sponge Callyspongia diffusa (41) and with the coral Montipora verrucosa (61). Occasionally, in some pairings one of the partners became necrotic, whereas the other was perfectly healthy (Fig. 3B), indicating the existence in sponges of a good system to fight off infection after long exposures to allogeneic decaying tissue. Replicate combinations of the same alloparabiosed individuals yielded identical directional rejection reactions in both qualitative (collagen barrier or necrosis) and quantitative (timing) terms, an observation consistent with the presence of a hierarchical system based on the genetic constitutions of different clones.

Table I. Outcome of grafting combinations within the population used for this study Numbers represent sponge individuals, crosses indicate rejection, and circles indicate fusion. Empty spaces correspond to missing grafts, result of the death or severe deterioration of one or both individuals before the excision of tissue to be grafted.

[View Larger Version of this Image (57K GIF file)]

In fusing grafts, the furrow dividing the two pieces was indistinguishable from the surrounding tissue (Fig. 3A, sides), and microscopical observation revealed that the epithelium originally separating the fragments disappeared while new tissue was rapidly being built up, making it difficult after 24 h to trace the original border (not shown). After that period, a physical resistance was clearly felt when trying to pull apart the two fused pieces. In our experiments, only one out of 220 alloparabionts was found to be compatible, although both sponges (numbers 14 and 15) were collected the same day and looked morphologically similar, suggesting that they might actually be a single genetical individual. As first pointed out by Van de Vyver (10), since sponges do also reproduce asexually, either regenerating viable individuals from fragments or through the production of gemmules or buds, they can generate clones of genetically identical but anatomically distinct individuals. The results presented above confirm an extensive polymorphism of histocompatibility molecules in sponges already observed by other authors (18, 41).

Far from being a laboratory artifact, both fusion and rejection phenomena can also be observed in the field, where the arms of an individual fuse wherever they touch each other for a long enough time (Fig. 3C), whereas the bodies of two individuals living close to each other will keep growing separately, leaving a cleft between them (Fig. 3D). In the population used for this study, the Microciona clumps 8 and 10 included two individuals each (the smaller ones being labeled individuals 8s and 10s). Clumps 4 and 5 also included more than one individual, but their bodies were so much entangled that we decided to exclude them from our experiments. In the M. prolifera population around the Woods Hole area, such side-by-side cohabitation seems to be common, since about 20% of the clumps collected contained more than one individual.

Defined genetic crosses would be the best approach to study sponge allogeneic reactions, but for M. prolifera the required methods have not been described. Spontaneous rejection occurring between sponges is difficult to study in the field. In our hands, M. prolifera individuals freshly collected from the wild and immediately transferred to a tank of abundant running sea water cooled to 14 °C did not survive in optimal conditions for more than 4 weeks, often starting to decay a few days after collection. Therefore, as a preliminary approach to investigate the possible involvement of MAFp3 or related proteins in sponge graft rejection, we decided to study the polymorphism of MAFp3-related molecules in the population used for the grafting experiments.

Genomic DNA isolated from each individual used in the grafting experiments was subjected to RFLP analysis (62). We have found 99.5% correlation between fusion/rejection behavior within the population used (Table I) and the identical/dissimilar RFLP pattern observed with the MAFp3-specific probe I. Seventeen distinct bands can be distinguished in Fig. 4A, some of them difficult to discern, like the 1.4- and 0.9-kb double bands better seen in short exposures. Exactly the same bands and relative intensities were obtained by using probe I derived by PCR either from total cDNA or from the cloned sequence MAFp3A. A 435-bp probe corresponding to the carboxyl-terminal half of MAFp3A comprised between the intron site and the stop codon also detected all major bands (not shown), despite its short length on the limit of detection for hybridized DIG-labeled probes. The multiple-banding pattern implies that each single sponge individual possesses multiple MAFp3-related sequences, thus confirming the result obtained in the sequencing of different positive clones from a single individual cDNA library described above. Of several restriction enzymes assayed, DraI was the only one that completely digested the genomic DNA of all of the individuals. Using HaeIII, we also observed polymorphism, as shown in Fig. 2A, although we do not present HaeIII-digested Southern blots of the population used in this work because several samples could not be completely digested even using higher enzyme concentrations and extended incubation times. Besides the nonrejecting sponge pair 14/15, individuals 8s/13 and 17/23 also exhibited an identical band pattern. Unfortunately, we lack the grafting combination 17/23 because 17 died soon. Nevertheless, the intensity of the lower 0.9-kb band in individuals 17 and 23 is clearly different, suggesting even in this case the existence of different MAFp3-related sequences. When the DraI-digested Southern blot was hybridized to a probe specific for region II (Fig. 4B), four main polymorphic bands around 4.3 kb could be identified. Individual 8s had two DraI fragments binding to probe II while individual 13 had one. Therefore, in the population used for this work, any two rejecting individuals exhibited a dissimilar RFLP pattern using AF-related probes.

[View Larger Version of this Image (58K GIF file)]

To study if this polymorphism was also observed in other sponge proteins we performed an RFLP analysis of the population used for the rejection experiments with a collagen-specific probe lacking internal DraI sites, amplified from the cDNA of the only protein besides MAFp3 cloned so far in M. prolifera (56). The result presented in Fig. 5 indicates a considerable degree of polymorphism also for collagen.

[View Larger Version of this Image (54K GIF file)]

As mentioned above, the screening of the different libraries revealed that each single individual possessed several of the MAFp3 sequences from Fig. 1, but it also suggested that not all of the forms were present in all of the individuals. Since mere sequencing is unable to detect the absence of a given form, we decided to perform PCR amplifications of the genomic DNA from the population used in the rejection experiments with specific primers common to forms A, B, and C, flanking the stretch between positions 287-300 in region I of forms B and C, which is missing from form A. The expected 156- and 198-bp bands corresponding to forms A and B/C, respectively, were found to segregate in the population (Fig. 6), and although most of the individuals had both forms, MAFp3A was not present in individuals 1, 2, 17, and 23, while MAFp3B and -C were absent from individual 22. Both amplified fragments were subcloned and sequenced to confirm that they corresponded to the sequences of forms A and B/C. Forms D and E were amplified from all the individuals, but we could not design primers specific enough to distinguish between forms B and C. This result indicates that at least forms MAFp3A, -B, and -C represent different alleles rather than members of a multigene family.

We described a glycoprotein subunit of ~210 kDa termed S1 that is a component of MAF (50). When the AFs purified from the 23 specimens were analyzed in an SDS-polyacrylamide gel, a remarkable polydispersity in glycosaminoglycan-containing bands could be observed (Fig. 4C), particularly clear in the case of S1 because of its relatively smaller size, which allows mass variations to be detected as changes in electrophoretical mobility. S1 is individual-specific, since its electrophoretic mobility remained unchanged between AFs isolated either from different parts of the same individual or at different times, e.g. the day of collection and 4 weeks later. Occasionally, S1 included two or even three bands, although protein sequencing yielded for all of them the same amino terminus already reported by us (50). Again, individuals 14 and 15 were identical in this assay.

S1 subunits purified from individual sponges were digested with peptide-N-glycosidase F or proteinase K. The results shown in Fig. 7A indicate the presence of N-linked glycans and the existence of individual variability at the glycosaminoglycan level, since the products resulting from protease digestion still had different mobilities. The same result was obtained after trypsin digestion (not shown). Other carbohydrate-hydrolyzing enzymes (chondroitinase ABC, heparinase, keratanase, -glucosidase, endoglycosidase F), had no effect on S1 under conditions in which the natural substrates were completely digested. Monoclonal antibodies block 1 and block 2, which inhibit the AF-mediated aggregation of sponge cells through their binding to MAF carbohydrate epitopes (63, 64), strongly recognized most of the subunits observed after Alcian blue/silver staining (Fig. 7B), suggesting their involvement in the aggregation process. S1 is likely to be the same entity described by Varner as an extracellular matrix molecule able to inhibit sponge cell adhesion (65), thus supporting the hypothesis of its role in intercellular interactions.

[View Larger Version of this Image (56K GIF file)]

DISCUSSION

Using RFLP analysis and SDS-PAGE, we have revealed three polymorphic components involved in sponge cell adhesion. MAFp3, the protein from region II, and the S1 subunit exhibit individual variability, and their combination seems to produce unique aggregation factor-related molecular systems for each genetically distinct individual. Since virtually any two genetically distinct sponge individuals will reject each other (Refs. 18 and 41; this work), it is tempting to describe the existence of a correlation between sponge graft rejection and AF variability, thus supporting the observations of Curtis and Van de Vyver (43), which suggested a direct involvement of sponge AFs in sponge self/non-self recognition. We have also found in the population used for this work an elevated correlation between histocompatibility behavior and the RFLP band pattern using a collagen-specific probe, which might indicate an active involvement of collagen in allorecognition, given its accumulation in rejection grafts. However, despite the 100% concordance observed between a dissimilar RFLP pattern using AF-related probes and histoincompatibility, we must consider the fact that sponges are extraordinarily diverse (66). Sequence data suggest that the diversity of gene families found in higher organisms already existed in sponges (56). Multiple homologous genes have been found for nonfibrillar collagen and for src-related kinase genes in the freshwater sponges Ephydatia mülleri (67) and Spongilla lacustris (68), respectively. Allozyme variation in the tropical sponge Niphates erecta showed 84% correlation with the observed graft response between different individuals (69), although the enzymes used are not suspected to have any direct relation with cell/cell interaction systems. Along the same lines, the results of Curtis et al. (70), reporting dissimilar plasma membrane proteins in graft-accepting pairs of sponges, are in agreement with the remarkable genetic variability of invertebrates (71, 72) without being in conflict with the presence of a highly polymorphic histocompatibility system in sponges. Nevertheless, not all sponge genes are polymorphic, and single-copy polyubiquitin and fibrillar collagen genes have already been identified in G. cydonium (73) and E. mülleri (67), respectively.

The differences between regions I and II within the MAFp3 gene are evidenced by the higher variability of region I detected in RFLP analysis. Under certain circumstances, MAFp3 seems to be translated from a short ~1.2-kb mRNA, which would represent the lower limit of the transcript size (50). This small transcript might be the result of alternative splicing, but it also could have originated from a different gene. When screening the cDNA libraries, we always selected the longest clones for sequencing, and therefore any MAFp3-related sequences originated from the small 1.2-kb transcript are probably not represented in Fig. 1, although they might be responsible for some of the bands in Southern blots. The variability detected for region II is restricted to four polymorphic bands around 4.3 kb in the DraI-digested DNA, which suggests that region II, at least as represented by probe II, is a single copy sequence with some alleles apparent. Using the probe specific for region I, the strongest band corresponds precisely to a 4.3-kb restriction fragment present in all the individuals, which probably is the same fragment detected with the region II-specific probe, since probes I and II are separated in the genomic DNA by about 3 kb according to the PCR amplifications performed to place the introns. In Southern blots using the MAFp3-specific probe I, we have detected in each individual between four and six strongly hybridizing bands both with DraI- and HaeIII-digested sponge genomic DNA. If we assume that the weak bands reflect binding to short target sequences resulting from the presence of internal restriction sites either in the cDNA sequences or in introns that can appear in the region spanned by the probe, our result would suggest that each sponge has between four and six MAFp3 copies, indicating the existence in the haploid genome of three loci coding for MAFp3. The interpretation of the Southern result becomes more difficult if we consider that some of the weak bands can also represent related sequences sharing low homology with MAFp3.

Minor changes in heparan sulfate structure dramatically affect the binding of syndecan to collagen (74). Biglycan, a dermatan sulfate proteoglycan from articular cartilage, also exists as a nonproteoglycan form especially abundant in adults (75), suggesting different functional properties of biglycan during development, which would be linked to the extent of glycosylation. Differences in the carbohydrate structures of neural cell adhesion molecules modulate their binding properties during development (76, 77). Carbohydrates are relevant in lymphocyte recognition (78, 79), and their role is especially important in natural killer cell functions (80). Of particular interest are the glycosylation microheterogeneities found on plant self-incompatibility glycoproteins that might provide structural diversity crucial for allelic specificity (81, 82). We have shown the existence of different core proteins and of individual variability at the carbohydrate level that might account for MAF diversity, thus explaining the differences observed in monosaccharide and amino acid composition between different MAF preparations (50, 51, 83). Therefore, changes in protein and/or glycosaminoglycan composition could modulate the AF activity in intercellular interactions. In our previous work (50), we showed that the protein subunits of MAF interact through disulfide bonds, whose structural importance is again revealed by the conservation of cysteines in all of the sequences obtained so far (Fig. 1) even without new cysteines appearing in other positions. This suggests that the precise positioning of disulfide bonds is essential for maintaining the aggregation factor structure.

Three lines of evidence contribute to distinguish two groups within the different forms presented in Fig. 1. First, the primary sequences of D and E are more related to each other than to the other forms. Second, the gene structures are different in region I, where the intron present in MAFp3D and -E is missing in forms A, B, and C. Third, while MAFp3A, -B, and -C seem to be of an allelic nature, forms D and E have been found in all of the individuals tested so far. Taken together, these data indicate that the evolutive branching point between the two groups is previous to the differentiation of forms within each group, suggesting that the cellular functions of the MAFp3 forms A, B, and C might be different from those of D and E. Amino acid sequencing has provided three peptide sequences of MAFp3 (50), although all of them fit with the relatively unvariable amino-terminal half of forms MAFp3A, -B, and -C, and two of them even fit with forms D and E, thus not revealing which of these forms (if not all of them) actually constitute aggregation factor components. We have sequenced several times the amino terminus of the deglycosylated MAFp3 protein from several individuals, never obtaining the sequence of forms D or E, but of course this absence of proof does not imply proof of absence.

Work with the tunicate Botryllus schlosseri has experimentally demonstrated that allorecognition in this species is regulated by a highly polymorphic fusion/histocompatibility (Fu/HC) locus, comparable in complexity with the vertebrate MHC (84-87). Fuke and Nakamura (3) showed that alloreactivity in the solitary ascidian Halocynthia roretzi was controlled by two highly polymorphic loci from the histocompatibility gene. Intriguingly, the Fu/HC locus and the histocompatibility gene seem to control intercolonial compatibility and also fertilization (3, 85, 88). In other words, the same gene product acting as an immune system molecule might mediate cell-cell interactions in general, but a biochemical basis for these phenomena has yet to be discovered. A view of MHC evolution is that the polymorphism of cell surface MHC-like proteins preceded the ability of these proteins to present peptides and that a diverse set of receptors was educated to tolerate self but reject polymorphic ligands (89). Danska et al. (90) biochemically characterized from Botryllus a cell surface molecule of unknown function that resembled T cell receptors, of which several isoforms were detected, although their individual distribution did not correlate with Fu/HC allelic diversity. To explain the evolution of the primitive immune system, Burnet (91) formulated the self-recognition hypothesis, according to which invertebrate allorecognition would be based on the positive recognition of common structures originating from common alleles, rejection resulting from the failure to recognize kin. Indeed, rejection in Botryllus and Halocynthia occurs when the interacting individuals do not share alleles (3, 87). This system, according to Burnet, would detect similarity, not difference, unlike the vertebrate immune system. However, vertebrate natural killer cells have been found to destroy self cells if these do not express self MHC class I molecules (92-94), thus acting as a first line defense, perhaps evolved from a primitive self-recognition system similar to that found in present day invertebrates and kept when T cell-based vertebrate adaptive immunity appeared. Despite the presence in all MAFp3 sequences of eight conserved cysteines, a correctly placed tryptophan in position 35, and stretches of alternating hydrophobic and hydrophilic residues that might structure -sheets according to secondary structure predictions (not shown), we have not identified any Ig domains. Detailed genetic studies such as those performed in tunicates have not been done with sponges yet, and therefore no loci have been described that might be involved in regulating the accurate sponge histoincompatibility system.

All evidence to date, much of it derived from former work done in our laboratory, indicates that MAF isolated from one individual has equal activity on the cells of all individuals of the species. In fact, the molecular components of the AF cause aggregation in a relatively nonspecific manner by weak molecular association based on repeating glycosylated subunits. However, the aggregation activity was usually measured with glutaraldehyde-fixed cells and without care for their individual or collective origin (51, 83). We have not observed any significant differences when testing AF on homologous and heterologous glutaraldehyde-fixed cells, thus confirming the lack of individual specificity of the MAF-mediated aggregation process. The first hint of a possible involvement of the AF in sponge allorecognition came from studies done with E. fluviatilis live cells by Curtis and Van de Vyver (43), suggesting that allorecognition and species-specific aggregation might ultimately be based on similar, if not the same, molecular components. At present, we are establishing the assay conditions to reproduce with Microciona cells the experiments done by Curtis and Van de Vyver.

Self/non-self discrimination requires polymorphic recognition molecules. In this work, we describe a highly polymorphic gene system that matches the high level of sponge alloincompatibility. So far, we have found no exception to the fact that any two rejecting M. prolifera individuals exhibit different AF-related components. We have also demonstrated the allelic nature of several MAFp3 forms. The existence of AF components that are allelic forms of a polymorphic system might represent an early form of self-recognition and therefore suggests an evolutionary relationship between cell adhesion and histocompatibility systems.