We hypothesize that tgrB1 and tgrC1 encode matching transmembrane proteins (TgrB1 and TgrC1, respectively) that mediate self-recognition by heterotypic interactions between the extracellular protein domains on adjacent cells. If the hypothesis were correct, we would expect that deleting the resident alleles would render the cells unable to self-recognize. Furthermore, replacing the resident alleles with a matching pair of alleles from a different strain would restore self-recognition but abolish cooperation with the parental strain.

We deleted the tgrB1-tgrC1 locus (Fig. S1) and mixed the mutants with the parental AX4 cells (6). The strains started to segregate within the streams of cells that form at the onset of aggregation and eventually formed segregated multicellular structures (Fig. 1A). In pure populations, the tgrB1⁻tgrC1⁻ cells were arrested at the aggregate stage (Fig. 1F) and exhibited compromised spore production (Fig. 1J). These results suggest that an intact pair of tgrB1-tgrC1 alleles is necessary for self-recognition, consistent with our previous findings that tgrB1 and tgrC1 are individually necessary for self-recognition (5). We then reintroduced a pair of tgrB1-tgrC1 alleles from AX4 into the deletion strain, restoring the locus to its original structure with minor changes (Fig. S1). Both alleles were fully functional because they complemented all the tgrB1⁻tgrC1⁻ phenotypes. This double-gene replacement strain (tgrB1^AX4tgrC1^AX4) mixed with the parental AX4 strain as seen in the even distribution of red and green-labeled cells (Fig. 1B). The strain also produced fruiting bodies (Fig. 1G) and spores (Fig. 1J) in pure populations.

The tgrB1 and tgrC1 sequences are polymorphic (5). We cloned them from four wild strains, which are different from one another and from AX4 (Fig. S2), and recombined them into the resident locus of the tgrB1⁻tgrC1⁻ cells. We then mixed the cells at equal proportions and found that the double gene replacement strains carrying alleles from QS4 (tgrB1^QS4tgrC1^QS4) or QS38 (tgrB1^QS38tgrC1^QS38) segregated from AX4 (Fig. 1C,D) and eventually formed separate fruiting bodies (Fig. S3). Segregation was observed at other proportions as well (Fig. S4). The double-gene replacement cells developed and produced spores in pure populations (Fig. G–I), as did strains carrying either tgrB1^QS31tgrC1^QS31 or tgrB1^QS45tgrC1^QS45 (Fig. 1J). We also tested a mixture between the tgrB1^QS4tgrC1^QS4 and the tgrB1^QS38tgrC1^QS38 cells and observed that the two strains segregated (Fig. 1E).

In the models (Fig. 1) we propose that cells lacking both tgrB1 and tgrC1 segregate from AX4 because they are not recognized as self. They also fail to develop in pure populations because they do not recognize one another as self. Reintroducing a matching pair of alleles from AX4 restores self-recognition and allows the cells to cooperate with AX4. Introducing a pair of polymorphic alleles from any wild strain restores self-recognition, as reflected in the cells’ ability to develop in pure populations, but these double-gene replacement strains do not cooperate with strains with non-matching tgrB1-tgrC1 alleles, including the parental AX4. We also conclude that self/non-self-recognition is not a result of compromised development – the double gene replacement cells develop well as pure populations, but segregate from non-self cells in mixes.

The only differences between the double gene-replacement strains are the sequences of tgrB1-tgrC1. Therefore, matching pairs of these alleles are both necessary and sufficient for self-recognition in the AX4 background, suggesting direct involvement in recognition.

The donor strains of tgrB1 and tgrC1 are genetically distinct from AX4 and from each other, and segregation between them is directly correlated with their genetic distances (4). Moreover, segregation and tgrB1-tgrC1 polymorphism are correlated (5), suggesting that tgrB1 and tgrC1 are major determinants in self-recognition. To test that possibility directly, we examined recognition between the double gene-replacement strains and the donor strains (Fig. 2). In control experiments, the tgrB1^AX4tgrC1^AX4 cells segregated from the wild strains (Fig. 2A and 2D). Segregation was evident at the streaming stage and later in development when the cells formed segregated fingers. Mixing QS4 with the tgrB1^QS4tgrC1^QS4 strain resulted in an even distribution of cells during early aggregation and the two strains remained in the same multicellular structures although sorting was observed within the slugs (Fig. 2B). We observed similar patterns when we mixed QS31 and tgrB1^QS31tgrC1^QS31 cells (with less homogeneous mixing during early aggregation, but with both strains present in the same slug; Fig. 2E). The reciprocal non-matching mixes resulted in segregation – QS4 segregated from tgrB1^QS31tgrC1^QS31 (Fig. 2C) and QS31 segregated from tgrB1^QS4tgrC1^QS4 (Fig. 2F). In the models (Fig. 2), two cells recognize each other via matching tgrB1 and tgrC1 alleles, despite other genetic differences between them. These observations suggest that tgrB1 and tgrC1 are major determinants in self-recognition. The segregation observed at the finger stage may reflect physiological differences, because the wild strains are different from AX4 at various genetic loci (4).

TgrC1 is a heterotypic cell-cell adhesion protein (7) and TgrB1 is its potential binding partner (5), suggesting that self-recognition is mediated by differential binding of matching TgrB1 and TgrC1. However, tgrC1 also affects cAMP-signaling and gene regulation (8, 9), so self-recognition could be mediated by differential signaling. To test the role of differential cell-cell adhesion directly, we developed AX4 and double gene-replacement strains separately until they aggregated, disaggregated the structures into single-cell suspensions, mixed the cells, and tested cell-cell adhesion in suspension (10). Control experiments showed even mixing between AX4 and the tgrB1^AX4tgrC1^AX4 cells (Fig. 3A). However, AX4 and tgrB1^QS4tgrC1^QS4 cells adhered into distinct aggregates (Fig. 3B), whereas differentially labeled tgrB1^QS4tgrC1^QS4 cells formed mixed aggregates (Fig. 3C). We also tested single gene-replacement strains in pure populations and in mixes and found that tgrB1 and tgrC1 interact “in trans”, such that TgrB1 on one cell interacts with a matching TgrC1 on the other cell (Fig. S5). In the models (Fig. 3) we propose that TgrB1 on one cell binds the matching TgrC1 on a neighboring cell such that cells with matching tgrB1-tgrC1 alleles adhere to one another.

Self-recognition can be mediated by attraction of self or by rejection of non-self. To test which mechanism is utilized, we tested the effect of an extra tgrB1-tgrC1 pair. AX4 cells are haploid, so introducing extra alleles would make them merodiploid for these genes. If self-recognition were mediated by attraction of self, then introducing an extra pair of foreign alleles would allow the merodiploids to aggregate with two different strains, each sharing one pair of tgrB1-tgrC1 alleles. In the case of non-self rejection, an extra pair of foreign tgrB1-tgrC1 alleles would lead to self-rejection, resulting in aberrant aggregation. We constructed three merodiploid strains and found that they aggregated well and produced normal fruiting bodies and spores in pure populations. Mixing the tgrB1^+/QS4tgrC1^+/QS4 merodiploid strain with either AX4 (Fig. 4A) or with the tgrB1^QStgrC1^QS4 double gene-replacement strain (Fig. 4B) resulted in even cell mixing, suggesting that self-recognition is mediated by attraction. Mixing the tgrB1^+/QS4tgrC1^+/QS4 merodiploid with the tgrB1^QS31tgrC1^QS31 double gene-replacement strain (Fig. 4C) resulted in segregation, suggesting that the merodiploid genotype did not result in recognition promiscuity. The tgrB1^+/QS31tgrC1^+/QS31 merodiploid behaved in the same way (Fig. 4D–F). Finally, the tgrB1^+/AX4tgrC1^+/AX4 merodiploid co-aggregated with AX4 but segregated from the two other double gene-replacement strains (Fig. 4G–I), suggesting that doubling the tgrB1-tgrC1 dosage is not a confounding factor. In the models (Fig. 4) we propose that any matching pair of tgrB1-tgrC1 alleles in one cell is sufficient for recognition of a respective pair in another cell. In addition, the presence of a non-matching pair of tgrB1-tgrC1 alleles does not interfere with this recognition. These data suggest that tgrB1 and tgrC1 mediate self-recognition through an inclusive attraction mechanism.

Our data implicate tgrB1 and tgrC1 as central components in Dictyostelium self-recognition, because a matching pair of tgrB1-tgrC1 alleles is both necessary and sufficient for self-recognition in the AX4 genetic background and because the double-gene replacement strains cooperate with the genetically different donor strains. TgrC1 is a known heterotypic cell-cell adhesion protein (7), so we propose that TgrB1 is its binding partner and that TgrB1 on the surface of one cell binds a matching TgrC1 on the surface of another cell.

D. discoideum has other cell-cell adhesion genes, mainly cadA and csaA (11), that have no direct effect on self-recognition (5), but may explain the adhesion of tgrB1⁻tgrC1⁻ cells to each other. tgrB1 and tgrC1 could mediate repulsion between cells with non-matching alleles by modifying other adhesion complexes (11), but the inclusive attraction of merodiploid cells to one another, to the parental strains, and to the double-gene replacement strains, suggests that tgrB1 and tgrC1 mediate attraction rather than repulsion.

Other self-recognition systems rely on polymorphic membrane proteins with extracellular immunoglobulin-like folds (1), including the mammalian MHC (12), Botryllus FuHC (13), Drosophila dendrite self-avoidance (14), and Hydractinia allorecognition (15, 16). These proteins are not related to tgrB1-tgrC1 or to each other, suggesting convergent evolution in which similar properties arose despite the lack of extensive sequence similarity. The important properties are sequence polymorphism, which provides unique identities, and extracellular domains, which mediate cell-cell interactions. The immunoglobulin fold is common in these systems probably because it is tolerant of sequence variation (17).

01