The fester family

The first highly polymorphic gene identified in the fuhc locus was fester, which was a 350 amino acid type I transmembrane protein with a signal sequence, an extracellular Ig domain, a Sushi domain, followed by three transmembrane domains, and a short intracellular tail (Nyholm et al. 2006; Rodriguez-Valbuena et al. 2024). Fester is both polymorphic, there are ca. 100 allotypes in the US, and polygenic, as two haplotypes were found in our lab-reared strains, one with a single fester locus, and another with two loci. However, fester is not as polymorphic as fuhc-sec, nor did the polymorphisms correlate exactly with fusibility outcomes. In addition, it appeared that multiple exons of fester can be alternatively spliced, with each individual analyzed expressing a full-length form of fester, along with a unique repertoire of alternative splice variants. Importantly, genotype-specific alternative splice repertoires were maintained for the life of the individual—the ampullae can be ablated and will regenerate with no loss of specificity, and the fester splice repertoire was maintained, even following multiple ablation/regeneration cycles (Nyholm, et al. 2006). Together, these genotype-specific characteristics suggested that fester was an allorecognition receptor responsible for discrimination of fuhc ligands, and we speculated that avidity from binding of different splice variants could be responsible for specificity. As described below, this is turning out to be a much more complex story (Rodriguez-Valbuena et al. 2024).

Fester is expressed on the ampullae as well a subset of blood cells, and its role in allorecognition was assessed using two methods—monoclonal antibody (mAb) interference and siRNA knockdown. When a monoclonal antibody raised to an allele (festerA) carried in our lab-reared lines was injected into individuals in a compatible pairing—both festerA/A homozygotes, there was no effect, and fusion occurred normally. However, when injected into individuals that were incompatible, but both expressed the festerA allele, a rejection could be turned into a fusion, suggesting that the antibody was stimulating fester, which could override the rejection response. This result was allele-specific, but more importantly we found that one of the two individuals in the preceding experiment had to be a festerA/A homozygote—if the antibody was injected into two incompatible festerA/B heterozygotes, the rejection response was not blocked. This indicated a dose-dependent response to mAb binding was required to override the rejection response, and that ampullae from both individuals had to be stimulated for fusion to occur. In contrast, siRNA knockdown of fester gave an unexpected phenotype. Given the mAb-blocking results, we expected that knockdown of fester would cause compatible pairings to reject, as there would be no recognition of the fuhc ligand. However, we found that knockdown of fester blocked both fusion and rejection responses—when ampullae of either compatible or incompatible pairings came into contact, there was no reaction, equivalent to that seen for BHF (discussed above). As the siRNA was complementary to non-spliced exons of fester and should have ablated all splice variants, we hypothesized that while the spliced repertoire provided specificity, either the full-length form common to all individuals or a subset of splice variants could be partitioned into a receptor controlling rejection and another controlling fusion.

Subsequently, we identified another genomic locus that appeared to be a duplication of the region encoding fester exons 6–11, but with no homology to the 5′ end of the gene, and discovered it encoded a related protein, uncle fester. The NH2 terminal half of uncle fester had no DNA or amino acid homology with fester, but an equivalent architecture, including a signal sequence, Ig and sushi domain, followed by three TM domains, and an intracellular tail. In contrast to fester, uncle fester was monomorphic, and we only identified two alleles that differed by a single amino acid in populations from the USA and Europe. In addition, uncle fester did not show extensive alternative splicing; however, we found that there were genotype-specific expression levels that differed by over tenfold between individuals, and were stable over time (Taketa et al. 2024).

Uncle fester was also expressed in the ampullae and a subset of blood cells, and we carried out equivalent functional studies as described above. First, siRNA knockdown of uncle fester in incompatible pairings resulted in a no-reaction phenotype, blocking the rejection response, equivalent to fester. In contrast, knockdown of uncle fester in compatible pairings had no effect, and the ampullae fused normally, indicating that uncle fester is solely involved in the rejection pathway, and that fusion and rejection are not mutually exclusive. In addition, these studies also revealed that uncle fester expression level correlated with the rejection phenotype. For example, partial knockdowns resulted in a slower, less intense rejection response vs control pairings, and genotypes with higher expression would have robust rejection responses (McKitrick et al. 2011; Taketa et al. 2024). We also carried out mAb interference experiments and found that the presence of an anti-uncle fester mAb had no effect on incompatible pairings, which rejected normally. In contrast, in compatible colonies the presence of the antibody would override a fusion in process, resulting in a rejection. Finally, by conjugating the mAb to magnetic beads we could manipulate an individual ampullae, which initiated a rejection response, but had no effect on adjacent ampullae, consistent with the spatial segregation of the reaction, described above. In summary, uncle fester is necessary and sufficient to activate the rejection response.

Together, these studies revealed that allorecognition outcomes are not mutually exclusive, rather a quantitative response due to the integration of signals from two independent pathways, each of which is tunable. When juxtaposed ampullae come into contact and form a synapse, a rejection response is initiated at the immune synapse via a receptor that includes uncle fester (Fig. 1A). This recruits a blood cell type called a morula cell to the tips of the ampullae in contact, and recent scRNA seq data suggests this inflammatory signal is mediated by IL-17 (in preparation). If no other signal is received, the ampullae lose barrier function, and the morula cells leak into the periphery and lyse, discharging the precursors of the prophenoloxidase pathway, which form melanin scars in situ, called points of rejection (POR; Fig. 1B). This reaction also produces ROS that can kill the nearby the ampulla (Ballarin et al. 1998, 2002). As described below, there are a range of rejection responses that can be classified by the number, size, and time of formation of the POR, as well as the integrity of the interacting ampullae. However, if a self fuhc allele is recognized by fester, this overrides the rejection response and triggers fusion of the ampullae, which transdifferentiate into a vessel (Fig. 1C).

Functionally, this is equivalent to responses by mammalian NK cells, which integrate signals from activating and inhibitory receptors via integration of ITIM and ITAM signaling; however, neither fester nor uncle fester encodes any known signaling domains. Nonetheless, we had already found that proteins encoding cytoplasmic ITIM and ITAM domains—as well as nearly all associated signal transduction molecules used in mammals—were present in ascidian and other invertebrate genomes (Azumi et al. 2003; Nicotra 2019), and are expressed in the ampullae (see below), so hypothesized that signaling was transmitted via promiscuous adaptor molecules encoded activating (e.g., a DAP12 homolog) and inhibitory motifs. In addition, while the knockdown phenotypes were robust and consistent in different genotypes, it was not clear if other receptors contributed to specificity. As described below, we have now found the signal transduction partner, which has revealed another level of complexity to the effector system.

HECT family E3-ubuiqutin ligase (HE3L)

Finally, in our initial studies, two haplotypes from our mapping populations were physically mapped (Nyholm et al. 2006). Haplotype A had a single fester and single uncle fester locus, while haplotype B had two fester loci and one uncle fester locus. We found that a unique HE3L locus was encoded next to each fester and uncle fester locus in both haplotypes. The amino acid sequences were ca. 95% identical, revealing another unexpectedly polymorphic protein encoded in the fuhc locus (Rodriguez-Valbuena et al. 2024; Nath and Isakov 2024). Multiple E3-ligases, such as c-cbl, cbl-b, and NEDD4, play critical roles in tuning of both T-cell and NK signaling pathways, regulating signal strength and setting activation thresholds for both cells (O’Leary et al. 2015; Nath and Isakov 2024). For example, in naïve peripheral T-cells, ablation of cbl-b allows naïve T-cells to be activated without co-stimulation (Chiang et al. 2000), while in NK cells, c-cbl sets a threshold that can only be overcome by two independent activation signals (Kim et al. 2010). Moreover, in peripheral T-cells, the HECT family member NEDD4 positively regulates signal strength by interacting with both cbl-b and PTEN (O’Leary et al. 2015). The genomic redundancy, polymorphism, and linkage of the HE3L genes within the fuhc locus of Botryllus suggest a role in allele-specific tuning and discrimination.

Expanded fester and the fester co-receptor (FcoR) families

We have recently found that the fester family is much more diverse than previously described, and have identified over 37 new loci. These are encoded in polymorphic haplotypes with gene content variation, and each individual expresses an average of 10 loci (Rodriguez-Valbuena et al. 2024). Similar to fester and uncle fester, the new loci share little amino acid homology but equivalent domain architecture, and we are renaming these genes the fester family—fester is now FF1, and uncle fester is FF3. We have also identified the signal transduction partners, which are another large multigene family thus far consisting of 53 loci, that we are calling the fester co-receptors (FcoRs). The FcoRs are type I TM proteins related to fester and encoded in the fuhc locus (Fig. 2). They range in size from 520 to 610 aa, and encode a signal sequence, a variable type Ig domain, sushi domain, two C2-type Ig domains, a transmembrane domain, and a long intracellular tail which can encode canonical ITIMs, hemITAMs, and a core tyrosine motif (Y-x-x-I/L/V) of unknown function (Bauer and Steinle 2017). The core motif can be found alone or in combination with the ITIM or hemITAM domains, and many loci also encode SH2 domains. Each FF family member is encoded next to and expressed with a cognate FcoR family member. For example, fester (FF1) is paired with FcoR7, while uncle fester (FF3) is paired with FcoR1 (Fig. 2). There is also a link between polymorphism of each gene in a pair, and the putative functional role of the FcoR partner. For example, FcoR7 is as polymorphic as fester, and also encodes an ITIM domain, consistent with its role as an inhibitory receptor. By contrast, uncle fester (FF3) is paired with FcoR1, which is also monomorphic and encodes a tyrosine core motif domain, matching its role as an activating receptor.

These studies have also revealed several other surprising results. First, while only several haplotypes have been sequenced, the FF/FcoR are encoded as pairs in two polymorphic haplotypes—one is in the fuhc locus on chromosome 11 (Fig. 2), and another is on chromosome 5—and gene content variation in both haplotypes is of the gene pairs. These pairs are also co-expressed in transcriptomes from multiple individuals (n = 25), suggesting they assemble into specific heterodimers. To our knowledge, haplotype variation of gene pairs has never been described before, and if these genes do heterodimerize, the binding site will consist of two IgV-like domains. Finally, there is a third polymorphic haplotype found on chromosome 9 that encodes only FcoR loci.

In addition, there is only one highly polymorphic locus of both genes—FF1 and FcoR7, and these are paired. There are also two oligomorphic pairs, with four alleles discovered of each partner thus far, and a number of monomorphic pairs that were identical in > 5 individuals were also found. However, about 40% of the loci have only been identified in a single individual, so this is likely an underestimate. Nevertheless, given their role in allelic discrimination, we would have predicted that there would be more polymorphic, inhibitory gene pairs. The presence of multiple monomorphic pairs is intriguing, as the FcoR partner always encodes a hemITAM or core motif, but never an ITIM. The oligomorphic pairs are particularly interesting, as the polymorphisms in the FF partner are concentrated in the ectodomain, but the FcoR partner alleles differ in both the ectodomain and signaling motifs, and encode either an ITIM or a hemITAM. In addition, each oligomorphic FcoR locus can undergo alternative splicing that swaps one motif for the other, suggesting that the link between binding and signal output (activation or inhibition) is plastic, and required for allelic discrimination. However, the big surprise was that the uncle fester pair (FF3/FcoR1), which we thought was the sole activating receptor, is not ubiquitous. In summary, our previous hypotheses that allorecognition is due to the integration of activating and inhibitory pathways initiated by two receptors was oversimplified; this is a much more complicated interaction (discussed below).

Analysis of individual transcriptomes reveals that nearly all FF and FcoR genes show evidence of genotype-specific expression and alternative splicing that can differ by orders of magnitude, as we had found previously (Nyholm et al. 2006; Taketa et al. 2024). However, similar to the original results for cfuhc (described above; Nydam et al. 2013b), the extent of alternative splicing of fester (FF1) assessed by PCR does not match that found by either short read or long read bulk transcriptome sequencing from multiple individuals. This is intriguing, because all of the 64 splice variants of fester originally identified by PCR were in frame, while exons randomly mixed and analyzed computationally were not. Moreover, we had done several qPCR studies on individual splice variants that were entirely consistent, but the same variants are not found in bulk sequence data. Finally, several of these either extremely rare or non-existent splice variants were expressed in mammalian cells and targeted to the plasma membrane, indicating they fold correctly (Taketa and De Tomaso 2015). This is the second time that splice variants identified by PCR—and found at the plasma membrane when expressed ectopically—were not found in bulk transcriptome studies, and it is interesting that these are the two most highly polymorphic proteins in the fuhc locus. By contrast, this is not the case in the other 30 + genes encoded in the fuhc locus which were initially characterized using old-school RACE and PCR techniques, including uncle fester, which has two splice variants that were equivalent using both techniques. Finally, one consistent finding between all studies is that the vast majority of genotype-specific alternative splicing in all FF and FcoR is of the exons between the TM domain and the Ig/Sushi region, which would reposition the NH2 half of the protein relative to the plasma membrane.

ITAM and ITIM signal transduction

As described above, ITAM and ITIM domains were identified on candidate allorecognition genes in the cnidarian, Hydractinia, and the genome encodes associated signal transduction molecules (Rodriguez-Valbuena et al. 2022). In Botryllus, all fuhc-encoded proteins, as well as a homolog of nearly every protein in the ITAM and ITIM signaling pathway used by mammalian T- and NK cells, are expressed in the Botryllus ampullae. This includes receptor tyrosine phosphatases with homology to CD45, Src and syk family kinases, shp-1, SHIP, VAV, PI(3)K, PLC-γ, ITK, GADS Grb2, Crk, RasGRP, PKC-θ, the MAPK pathway, CARMA1, calcineurin, calmodulin, NFAT, and NFkB. However, there are two notable exceptions: homologs of LAT and slp-76 are not found in either Botryllus or other ascidian genomes (Rodriguez-Valbuena et al. 2024; Azumi et al. 2003).