The developing male gonad is a collective cell migration consisting of a chain of passively migrating somatic and germ cells led by a migratory somatic cell, the linker cell (LC) (Figure 1A–C). After migration, the interconnected somatic cells behind the LC differentiate during the transition from the fourth larval (L4) stage to the adult into a mature gonad structure, a tube comprising the vas deferens and seminal vesicle. Behind the somatic gonad are the proliferating germ cells, arranged from the newest in the distal region to the most developed closest to the somatic gonad. Capping the distal end of the gonad are the two male distal tip cells, which maintain the mitotic germ cells. To form this gonad shape during the L2 through L4 stages of the larval development, the LC leads the elongating gonad from the mid-body region to the cloaca opening in the posterior body, where it dies after completing the migration. During the L3 stage of the migration, the somatic cells of the vas deferens and seminal vesicle precursors divide from seven to 53 cells to form the elongating gonad. The developing somatic gonad has epithelial-like characteristics consisting of strong intercellular connections and a developing apical domain running down the core of the gonad (Figure 1A). The somatic cells have a radial symmetry around this core, and as they proliferate, the daughter cells are incorporated into the chain while maintaining this configuration.

Figure 1

Download asset Open asset

We examined the requirement of different gonadal cell types for migration. By ablating the LC (15/15), we confirmed the findings of Kimble (1981) that the LC is necessary for migration and is not replaced by a follower cell taking on its migratory role. Without the LC, the gonad stopped elongating but continued to balloon through cell proliferation. The LC, however, was capable of migrating alone if the somatic cells around it were ablated in the L3 stage (8/10). In this case, the LC alone migrated along its normal course while the gonad stopped elongating at the point of LC detachment (Figure 1D). Since LC migration was slower than normal, the LC often did not complete its migration by the L4-to-adult transition. The cause of the slower migration may be due to the missing contribution of gonadal cells or other factors such as drag from scarred tissue. The germ cells add to gonadal mass but are not necessary for migration, as the LC reaches the cloaca even when germ cell precursors are ablated.

We discovered the lnkn-1 mutant during a process of identifying new genes involved in LC migration by utilizing a database of expression patterns reported by the Genome BC C. elegans Gene Expression Consortium (Hunt Newberry et al., 2007). Since the consortium only reports hermaphrodite expression patterns, we searched their database for genes expressed in the migratory leader cells for the hermaphrodite gonad, the distal tip cells (DTCs), which functionally are the closest cells to the male LC (Kimble and Hirsh, 1979). We reasoned that the gonadal leader cells of both sexes may use partially overlapping genes for their migrations. One of the genes reported to be expressed in the hermaphrodite DTCs was ZK637.3 (WBGene00014023), a conserved gene of unknown function that had an available deletion mutant, gk367. After obtaining this mutant and observing unusual male gonad defects, we decided to investigate this gene further. We have renamed this gene from tag-256 (temporarily assigned gene-256, ZK637.3) to lnkn-1 (LiNKiNg-1).

In males homozygous for lnkn-1(gk367), gonadal cells near the LC became detached during gonad migration (n = 29/30) such that the LC continued to migrate, either alone or with a few remaining follower cells, but the rest of the gonad did not follow. This detachment resulted in a partially elongated gonad and, further ahead along the normal path, a detached LC alone or with a few adherent follower cells (Figure 1E,F). This phenotype is similar to that of the gonad with ablated follower cells behind the LC (Figure 1D). The position of detachment was variable but usually occurred within a few cell lengths behind the LC, suggesting that the pulling force generated by the LC may have caused detachment. Although the LC continued to migrate along its normal course after detachment and occasionally completed the migration, the male was sterile since the gonad did not connect to the cloaca opening. In lnkn-1(gk367) mutant hermaphrodites, the gonadal leader DTCs remained connected but migrated a shorter distance than the wild type and their shape appeared elongated and strained. The mutant also recessively caused maternal effect lethality (n = 30/30). The male gonad phenotype in lnkn-1(gk367) mutants suggested that lnkn-1 is required for cell–cell interaction rather than LC migration.

LNKN-1 was a conserved, poorly characterized protein predicted to be a type I single-pass transmembrane protein of 599 amino acids (AA), consisting of a 19 AA signal sequence, 533 AA extracellular domain, 23 AA transmembrane domain, and 24 AA intracellular domain (Figure 2). We were able to identify homologs of LNKN-1 back to an early branching animal phylum, Placozoa, as well as in fungi, and have called this protein family LINKIN. The presence of LINKIN in Plasmodium falciparum has previously been noted (Kaczanowski and Zielenkiewicz, 2003). LINKIN is conserved in Metazoa from Placozoa Trichoplax adhaerens, a basal metazoan, to vertebrates including human. A protein alignment of LINKIN from Homo sapiens (ITFG1/TIP, 612 AA), Mus musculus (ITFG1/TIP, 610 AA), Drosophila melanogaster (CG7739, 596 AA), and C. elegans (LNKN-1, 599 AA) revealed that all orthologs have similar protein lengths and domain organizations (Figure 2B).

Figure 2

Download asset Open asset

Overall, the protein sequence between H. sapiens and C. elegans excluding the signal sequence is 26% identical (154 AA) and 61% similar (365 AA). However, LINKIN has a highly conserved intracellular domain, which is 62.5% (15/24 AA) identical and 87.5% similar (21/24 AA, clustalo analysis). In particular, the last eight amino acids are identical in all four species (Figure 2B). Despite its high conservation, a BLAST search of the intracellular domain alone did not identify strong similarities with domains in other proteins.

Protein motifs in LINKIN were largely unknown, with the only ascribed motif being an FG–GAP domain found in one copy in H. sapiens and three in M. musculus (uniprot.org). The FG–GAP is a domain that occurs in seven copies in α-integrins and folds into a seven-bladed β-propeller structure that serves as its ligand-binding domain (Springer, 1997; Xiong et al., 2002). We have identified seven atypical FG–GAP domains in the N-terminal of the extracellular domain, based on sequence similarities to the annotated LINKIN FG–GAP domains from H. sapiens and M. musculus and to α-integrin FG–GAP domains (Figure 2, ‘Materials and methods’). FG–GAP domains have a loosely conserved Phe-Gly and Gly-Ala-Pro sequence, which are separated by sequence that can include a calcium-binding motif. A comparison of all human α-integrin FG–GAP domains showed that their calcium-binding motif has a strong DxxxDxxxD signature (D = Asp, x = AA; Chouhan et al., 2011). We found a strong DxxxDxxxD signature in all seven calcium-binding domains of LINKIN, suggesting that these are similar to FG–GAP domains of α-integrins (Figure 2). The significance of finding seven FG–GAP domains in the N-terminal of LINKIN is the possibility that LINKIN, like α-integrins, uses a seven-bladed β-propeller structure to bind ligand. We also identified a highly conserved extracellular region adjacent to the transmembrane domain, which is a yet unrecognized protein domain (Figure 2). LINKIN has a dozen predicted N-linked glycosylation sites (uniprot.org). Taken together, our investigation shows that LINKIN is a conserved transmembrane glycoprotein pre-dating Metazoa and potentially containing a seven-bladed, β-propeller, ligand-binding domain.

We examined the expression pattern and subcellular localization of LNKN-1 in C. elegans, particularly in the male gonad. Previously, lnkn-1 localization was categorized to be in cell membrane, when examined by GFP-tagging in a screen of muscle-related genes (Meissner et al., 2011). The only other investigation into LINKIN observed that in mammals the extracellular domain functions as a secreted protein that modulates T-cell dependent immune response (Fiscella et al., 2003). We made both extracellularly and intracellularly YFP-tagged versions of LNKN-1 expressed under its natural regulatory specific promoter (Figure 3E,F, Figure 3—figure supplement 2M,N). Both YFP::LNKN-1 and LNKN-1::YFP are similarly localized to the plasma membrane of many cells. LNKN-1 begins to be expressed in all somatic gonadal cells of the male, including the LC, the vas deferens precursor cells, and seminal vesicle precursor cells, starting in the early L3 stage and continuing through adulthood (Figure 3E,F). It is also expressed in all somatic gonadal cells of the hermaphrodite, including the distal tip cells, anchor cell, uterine precursor cells, and spermatheca precursor cells (Figure 3—figure supplement 2). Other expression occurs in pharynx, pharyngeal-intestinal valve, intestine, excretory cell and canal, seam cells, a specialized subset of hypodermal cells, the vulval precursor cells of the hermaphrodite, and hook precursor cells in the male (Figure 3—figure supplement 2). YFP-tagged LNKN-1 is localized to the plasma membrane, exhibiting stronger localization to the sides of cell–cell contact in tissues such as the intestine, seam, and gonad. This broad expression is consistent with our observations that lnkn-1 is expressed in the LC but not enriched (Schwarz et al., 2012).

Figure 3 with 3 supplements see all

Download asset Open asset

To examine the localization of the native protein, two polyclonal antibodies were raised in rabbit against the entire extracellular domain (533 AA) of LNKN-1 and against a peptide derived from the last 17 AA of the short intracellular domain and affinity-purified with the respective antigens. Immunofluorescence staining was performed on dissected gonads and intestines, which greatly improves antibody penetration over whole animals (Figure 3A–D). Staining of other dissected parts of the worm confirmed expression in the pharynx, excretory canal, and seam cells, indicating that the tissue expression pattern of LNKN-1::YFP is accurate. However, there was an important difference in subcellular localization between the native and YFP-tagged proteins: the antibodies showed stronger localization of LNKN-1 to the apical and lateral domain of the gonad (Figure 3A,B) and intestine, while YFP-tagged LNKN-1 is uniformly distributed in the plasma membrane. While both antibodies show heavier localization to the apical and lateral plasma membrane, the extracellular domain antibody shows additional staining in large cytoplasmic puncta (Figure 3A), and the intracellular domain antibody shows nuclear staining (Figure 3B, Figure 3—figure supplement 1G). In lnkn-1 mutants and lnkn-1 RNAi-treated animals, we do not see plasma membrane staining with antibodies against either the extracellular or intracellular domain, but we do see non-specific staining in the cytoplasm and nucleus (Figure 3—figure supplement 1). This indicates that the membrane staining is due to LNKN-1 localization, but the cytoplasmic and nuclear stainings are due to non-specific binding of other proteins. The antibodies also stain the plasma membrane in the germ cells, where YFP expression could not be determined because germ cells do not readily express transgenes. The two antibodies reveal the true localization of native LNKN-1 to be in the plasma membrane with a preference for apical and lateral domains.

Since this was the first reported use of the lnkn-1(gk367) deletion allele, we characterized it molecularly. The genomic locus of lnkn-1 spans 3640 nucleotides and is the second gene in an operon. The gk367 deletion of lnkn-1 excises 393 base pairs (bp) of genomic DNA starting soon after the signal sequence and ending in an intron. Since a possibility existed that an in-frame mRNA could be transcribed from the deletion, we characterized the truncated mRNA through RT-PCR and sequencing. The lnkn-1 cDNA sequence resulting from the gk367 deletion is missing the last 18 bp of the signal sequence and the first 258 bp of the extracellular domain (Figure 2). The larger size of the cDNA deletion than would be predicted based on the genomic lesion indicates that an alternate splice site was used when the lesion removed the usual splice donor; however, the product cDNA is still in-frame.

Since mRNA was being transcribed in lnkn-1 mutants, we investigated whether protein was being expressed. We generated a lnkn-1 promoter::lnkn-1(mutant cDNA)::yfp construct, in which yfp was fused to mutant lnkn-1 cDNA and expressed using its 5′ genomic region. We found that LNKN-1(mutant)::YFP is in fact expressed but shows cytoplasmic rather than plasma membrane expression (Figure 3G,H). This confirms that a truncated protein is being produced from the lnkn-1 deletion locus but is mislocalized and unlikely to have its normal function.

To test whether the lnkn-1 deletion mutant functions as a null despite producing a truncated protein, we examined the phenotype produced by lnkn-1 RNAi silencing. Males treated with lnkn-1 RNAi produced gonadal defects that were milder than the mutant (Figure 3—figure supplement 3A). While only 11% (4/35 animals) had detached gonadal cells, an additional 17% (6/35 animals) had ‘stringy’ gonads, in which fewer cells remained attached and were stretched from pulling by the LC. We also performed lnkn-1 RNAi on animals expressing LNKN-1::YFP to ensure that the RNAi was effective (Figure 3—figure supplement 3B–E). LNKN-1::YFP was absent from the gonad in all animals (n = 28), but was retained in a few tissues including the pharynx and excretory canal, likely because these tissues produce higher levels of LNKN-1 or were more resistant to the effects of RNAi. Since RNAi effectively reduces but does not eliminate the function of LNKN-1, we interpret the similar but stronger phenotype of the mutant to indicate that the mutant is in fact a loss-of-function allele.

We investigated the requirement of various domains of LNKN-1 for rescuing the mutant phenotype. We were able to completely rescue the lnkn-1 mutant, including maternal effect lethality and gonad adhesion defects, using a genomic construct (Figure 4A). Since lnkn-1 is the second gene in an operon, this construct contains 4.5 kb of genomic region upstream of lnkn-1 start site, the lnkn-1 gene, and the lnkn-1 3′ UTR (Figure 4A). We also made a cDNA construct using the same 5′ region of lnkn-1 fused to lnkn-1 cDNA and unc-54 3′ UTR, a common 3′ UTR for C. elegans constructs. The lnkn-1 cDNA was able to rescue the gonad defect but not maternal effect lethality (Figure 4B), possibly because it does not contain all regulatory elements for complete tissue expression or is silenced in the germline.

Figure 4

Download asset Open asset

We also investigated whether other lnkn-1 constructs can rescue the mutant (Figure 4). Full-length lnkn-1 with yfp fused to either the intracellular or extracellular domain did not rescue the mutant (Figure 4C,D). This is not surprising considering that YFP tagging prevents correct localization of LNKN-1 to the apical domain (see above). While both YFP-tagged proteins were expressed well in wild-type animals, intracellularly fused LNKN-1::YFP is not well-tolerated in lnkn-1 mutant heterozygotes and is lost within a few generations.

Partial domains of lnkn-1 were also not able to rescue the mutant; we attempted rescue with constructs containing only the secreted extracellular domain, the extracellular and transmembrane domain, the transmembrane and intracellular domain, and the cytoplasmic intracellular domain (Figure 4E–H). Lastly, we expressed full-length lnkn-1 under a LC promoter to test a gonadal non-cell autonomous effect, but this also did not rescue the mutant (Figure 4I). We conclude that the intact protein, with intact extracellular and intracellular domains, is required for the function of LNKN-1.

Since no interactors were previously known, we used a proteomics approach to identify binding partners in order to better understand the function of LNKN-1 through its interactions. Based on the hypothesis that the highly conserved intracellular sequence suggests both a required function for this domain and a potential for its binding partners to be conserved, we decided to identify LINKIN interactors by mass spectrometry using a human cell line. The advantage of using a human cell line compared to whole C. elegans is that it is a homogeneous cell type and the conditions for SILAC mass spectrometry are established. By using SILAC (stable isotope labeling by amino acids in cell culture), experiment and control immunoprecipitates can simultaneously be analyzed by mass spectrometry, enabling both quantitation and background subtraction. SILAC mass spectrometry was performed on immunoprecipitates from human LINKIN-Myc-expressing HEK 293T cells without isotopic labeling and from non-transfected cells with heavy isotopic labeling. Based on the ratio of ‘light’ to ‘heavy’ isotopes for each protein, enrichment through specific binding to LINKIN over background non-specific binding was determined. 484 proteins with at least two unique peptides were identified by LC/MS/MS from the immunoprecipitate with LINKIN, excluding common contaminants (Supplementary file 1). As validation for successful immunoprecipitation, LINKIN was itself one of the proteins with highest enrichment in LINKIN-Myc-expressing cells over control cells (35-fold enrichment; Figure 5A).

Figure 5

Download asset Open asset

Our aim was to identify, among the many LINKIN interactors, those functioning in maintaining gonadal cell attachment. We performed an assay in C. elegans based on the hypothesis that some of the binding partners of human LINKIN would also be conserved in C. elegans. Our approach was to assign C. elegans homologs to the ITFG1-interacting genes and perform an RNAi screen in C. elegans, seeking genes with a similar gonadal cell detachment phenotype as the lnkn-1 mutant. There were 68 proteins (excluding LINKIN) identified by mass spectrometry that were >fivefold enriched in ITFG1-Myc immunoprecipitates over control, and 45 of them had at least one C. elegans homolog (Supplementary file 2). 40 genes were available in existing RNAi libraries and screened. The silencing of three genes, ruvb-1/RUVBL1 (also known as Pontin), ruvb-2/RUVBL2 (also known as Reptin), and tba-2/α-tubulin, caused a similar gonadal defect to lnkn-1, (Figure 5B–E). ruvb-1/RUVBL1 and ruvb-2/RUVBL2 are highly conserved members of the AAA+ ATPase superfamily of proteins that often function together in a hexameric ring complex (Matias et al., 2006; Gorynia et al., 2011). α-tubulin together with β-tubulin forms microtubules, which as part of the cell cytoskeleton have roles in cell mechanics and transport of cellular components (Etienne-Manneville, 2013). β-tubulin was also among the highly enriched human gene interactors, but the homologous C. elegans β-tubulin did not produce a gonadal phenotype by RNAi. We tested the other five C. elegans β-tubulin genes and found that tbb-2 has a gonadal detachment defect (Figure 5F).

Having identified RUVBL1, RUVBL2, α- and β-tubulin as potential interactors of LINKIN that were also involved in the same biological process as LNKN-1 in cell adhesion, we wanted to confirm their physical interaction. Binding between RUVBL1 and RUVBL2 into heteromeric multimers (Gorynia et al., 2011) and binding between RUVBLs and microtubules have been reported (Gartner et al., 2003; Dobreva et al., 2008; Ducat et al., 2008). To test physical interaction between LINKIN and each of RUVBL1, RUVBL2, α- and β-tubulin, we performed Western blots on co-immunoprecipitates from ITFG1-Myc-expressing HEK 293T cells. Probing with antibodies against RUVBL1, RUVBL2, α-tubulin, and β-tubulin, we found that RUVBL1, RUVBL2, and α-tubulin bound LINKIN (Figure 5G–I). However, we did not observe β-tubulin binding LINKIN (Figure 5J). As a control for binding specificity of abundant cytoskeletal proteins, we also probed with an anti-β-actin antibody and found that β-actin did not bind LINKIN (Figure 5K).

We next investigated whether these interactions occur at the plasma membrane. We isolated the membrane and cytoplasmic fractions from HEK293T cells transfected with plasmids expressing ITFG1-Myc, Flag-RUVBL1, and HA-RUVBL2. By Western blot analysis, we showed that RUVBL1, RUVBL2, and α-tubulin were present in both cytoplasmic and membrane fractions, but ITFG1 was only present in the membrane fraction. We then immunoprecipitated with Flag-RUVBL1 from both cytoplasmic and membrane fractions and determined by Western blot analysis which proteins interacted with RUVBL1 (Figure 5L–S). ITFG1, RUVBL2, and α-tubulin interacted with RUBVL1 in the membrane fraction (Figure 5P–S). In the cytoplasmic fraction, RUVBL1, RUVBL2 and α-tubulin interact even in the absence of ITFG1 (Figure 5L–O). These results showed that the interaction between ITFG1, RUVBL1, RUVBL2, and α-tubulin occurs at the membrane, likely the plasma membrane based on C. elegans LINKIN localization.

We next investigated whether LNKN-1 interactors also localize to the plasma membrane. To examine RUVB-1 and RUVB-2 localization in the gonad, we generated polyclonal rabbit antibodies against full-length C. elegans RUVB-1 and RUVB-2 proteins, which were affinity-purified using the respective full-length proteins (Figure 6A–B). Strong staining for both RUVB-1 and RUVB-2 was observed in the cytoplasm and nucleus of gonadal cells (Figure 6A,B). We demonstrated the specificity of our anti-RUVB-1 and anti-RUVB-2 antibodies by comparing their staining in dissected gonads from wild-type animals (n = 6) and ruvb-1 (n = 6) or ruvb-2 (n = 8) RNAi-treated animals (Figure 6—figure supplement 1). We observed a consistent decrease in staining in the cytoplasm and nucleus of the gonads from RNAi-treated animals, indicating that RUVB-1 and RUVB-2 actually localize to both locations. Anti-α-tubulin antibody stained a dense network of microtubule fibers throughout the cytoplasm but particularly in the cell cortex of all gonadal cells (Figure 6C). The microtubules were more densely aligned along the developing apical domain of the gonad. Based on the Western blot analysis of membrane fractionated cells described above and the antibody staining results in C. elegans, the localization of LINKIN, RUVBL1, RUVBL2, and α-tubulin intersect at the cytoplasmic face of the plasma membrane, which agrees with the adhesion function of LINKIN.

Figure 6 with 1 supplement see all

Download asset Open asset

Since the known functions for RUVB-1 and RUVB-2 include transcriptional regulation (Jha and Dutta, 2009) and for α-tubulin include protein transport (Miller et al., 2009), we investigated whether RUVB-1, RUVB-2, or α-tubulin is required for either LNKN-1 expression or localization. We used RNAi to reduce ruvb-1, ruvb-2, and tba-2 function in LNKN-1::YFP animals and found that RNAi silencing of these genes did not affect LNKN-1::YFP function or localization (Figure 7). For ruvb-1 and ruvb-2 RNAi-treated animals, we also examined LNKN-1 expression by immunofluorescence staining and found no difference from untreated animals.

Figure 7

Download asset Open asset

RUVBLs are members of many complexes (Rosenbaum et al., 2013), but their only known interaction with microtubules has been at the mitotic spindle (Gartner et al., 2003; Ducat et al., 2008). We therefore investigated whether LINKIN or cleaved domains of LINKIN may also localize to the spindle during cell division. There are precedents for membrane-associated proteins involved in cell adhesion, such as ILK and β-catenin, to localize to the spindle with RUVBLs and microtubules (Kaplan et al., 2004; Fielding, et al., 2008; Dobreva et al., 2008). We first examined microtubule localization, which is known to redistribute during mitosis to a formation that radiates out from the mitotic spindle and attaches to the kinetochore and cell cortex. In fixed dissected gonads from L3 stage animals, by using an antibody against α-tubulin to identify microtubules and DAPI staining to identify condensed chromosomes of dividing cells, we observed the expected microtubule redistribution to the mitotic spindle (Figure 8A,B). The usually dense network of microtubules found in the cytosol and cell periphery of non-mitotic cells was replaced in mitotic cells by radial microtubules. We next examined whether LNKN-1 or domains of LNKN-1 were redistributed during cell division. Both antibodies against LNKN-1 extracellular (Figure 8C,D) and intracellular domains (Figure 8E,F) showed that localization of LNKN-1 remains unchanged at the plasma membrane during mitosis. These observations imply that there is a temporary loss of interaction between α-tubulin, RUVBL proteins, and LNKN-1 at the membrane and the interactions must be reestablished after cell division.

Figure 8

Download asset Open asset

To determine whether gonad detachment in lnkn-1 mutants involves the dissolution of mature cell–cell junctions, we examined cell junction formation using the apical junction marker, AJM-1::GFP, and gap junction marker innexin, INX-5::GFP (Figure 9). AJM-1 localizes to apical adhesion junctions in a complex with cadherins (Köppen et al., 2001). In the tube-shaped adult gonad, we observed AJM-1::GFP lining the entire lumen composed of the apical surfaces of the surrounding somatic gonad cells (Figure 9C,D). This apical accumulation begins in the mid-L4 stage as puncta (Figure 9A,B), but it is not present in the L3 stage when gonad detachment usually occurs in the lnkn-1 mutant. INX-5 shows expression and localization to cell–cell junctions in a few cells in the distal somatic gonad also starting in the L4 stage and becoming stronger in the adult (Figure 9E–H). Results based on both adhesion markers are consistent with the timeline of gonadal development, which dictates that cell division and rearrangement occurs during the L3 stage, while differentiation into the mature structure occurs during the latter half of the L4 stage up until the transition into adulthood. Cell dissociation in lnkn-1 mutants occurs at a stage when cell junctions are being rearranged; during this growth phase, adhesion-promoting genes like lnkn-1 may have a greater effect than later after the establishment of other more secure junctions.

Figure 9

Download asset Open asset

We have demonstrated that LINKIN is a unique transmembrane glycoprotein that functions as an adhesion molecule. This approximately 600 AA protein has a large extracellular domain of approximately 530 AA and a short 22 AA intracellular domain. We found a number of notable features about LINKIN. First, it is a protein pre-dating metazoans. Second, the size of the protein and organization of the domains have not changed during its evolution. The seven atypical FG–GAP domains, the extracellular region proximal to the transmembrane domain, and the intracellular sequence are all conserved. The identical sequence at the C-terminal of LINKIN suggested that some of its intracellular interactors may also be conserved proteins. In fact, LINKIN interactors, RUVBL1, RUVBL2, and α-tubulin are all highly conserved proteins that likely pre-date LINKIN. Third, among the genomes we searched, including those of H. sapiens, M. musculus, D. melanogaster, C. elegans, and T. adhaerens, LINKIN has no paralog. Even the highly conserved short 22 AA intracellular domain is not found in other proteins. The lack of paralogs is surprising considering that other highly conserved adhesion molecules like integrins and cadherins have numerous paralogs and have expanded into superfamilies (Angst et al., 2001; Takada et al., 2007).

The presence of seven atypical FG–GAP domains in the extracellular domain suggests that LINKIN might fold into a seven-bladed β-propeller. The significance of this prediction is that this structure resembles the ligand-binding domain of the adhesion molecule α-integrin. As with α-integrins, the β-propeller structure for LINKIN is located towards the amino-terminal of the extracellular domain. LINKIN is expressed on the surface of each adherent cell and might use the β-propeller domain to bind ligands on a neighboring cell to promote adhesion.

The C. elegans gonad shape develops through a collective migration of epithelial-like cells lead by the LC. One of the striking phenotypes of the lnkn-1 mutant was cell detachment in the migrating gonad. In lnkn-1 mutants, gonadal cell dissociation occurs in the L3 stage. At this stage, although the adhesions are strong enough that these cells cannot be mechanically dissociated without lysing, they have not yet formed mature adherens junctions. The expression of adhesion junction markers, AJM (apical junction molecule)-1::GFP, and gap junction marker, INX(innexin)-5::GFP, was absent in the L3 stage, when gonad cells begin detaching in lnkn-1 mutants; their expression only begins in the L4 stage and grows stronger in the fully differentiated adult gonad. The role of adhesion molecules, such as LINKIN, may therefore be more important before other, possibly stronger, adhesions are formed. The gonad detachment defect of lnkn-1 mutants likely results from the combined effects of the absence of a functional LINKIN adhesion molecule, the lack of other permanent adhesion structures in the L3 stage, and the force generated by the migrating LC.

LNKN-1 was present in many C. elegans tissues that possess apical/basal polarity, including the intestine, excretory canal, vulva, hook cells, and gonad; its localization was stronger on the apical and lateral sides for the tissues examined. YFP-tagged to either the extracellular or intracellular domain of LNKN-1 changes its localization from apical- and lateral-biased to uniform plasma membrane localization, indicating that apical localization is an active process requiring the function of both extracellular and intracellular domains of the protein. YFP-tagging also disrupts the function of LINKIN, as neither YFP-tagged LINKN-1 construct rescued the mutant phenotype, suggesting that apical localization may be necessary for LNKN-1 function. Known adhesion molecules have preferential localization at the plasma membrane. Integrins are often enriched in the basal domain since they bind extracellular matrix (Schoenenberger et al., 1994), while cadherins are enriched in lateral domains (Halbleib and Nelson, 2006). Specialized adhesion molecules like claudins and occludins localize to tight junctions (González-Mariscal et al., 2003) and connexins to gap junctions (Evans and Martin, 2002). LINKIN may be an adhesion molecule for cell–cell contacts and apical junctions.

Our rescue experiments also showed that secreted forms and partial domains of LNKN-1 do not provide function. The only previous study of LINKIN showed that the extracellular domain of human LINKIN functions as a secreted protein to modulate T-cell activation in cell culture and graft-versus-host disease model (Fiscella et al., 2003). Our experiments indicate that in the context of C. elegans collective migration, a secreted extracellular domain is insufficient to rescue the detachment defect. Our results suggest that, in the many C. elegans and human tissues expressing LINKIN, its function could originally have been as a transmembrane adhesion molecule, which in vertebrates conceivably has expanded to include a secreted form.

Although we do not yet know the binding partner on the extracellular side, we have made progress in identifying conserved interactors on the intracellular side—RUVBL1, RUVBL2 and α-tubulin. RUVBL1 and RUVBL2 are highly conserved members of the AAA+ ATPase superfamily, which use the energy harvested from ATP hydrolysis to perform mechanical functions on macromolecules. In bacteria where they were first identified, RUVBs have a function as a DNA helicase at holiday junctions; but in more complex organisms, the RUVBLs have additional diverse functions (Jha and Dutta, 2009). Among its non-nuclear roles, RUVBLs function in R2TP co-chaperone complex assembly (Kakihara and Houry, 2012) and in spindle assembly by nucleating microtubules and localizing components to the mitotic spindle (Gartner et al., 2003; Ducat et al., 2008). We have shown through limited cell fractionation that LINKIN is present in the membrane, where it interacts with RUVBL proteins and α-tubulin. Although we showed that LINKIN does not localize to the mitotic spindle, the latter role of RUVBLs as regulators of microtubule assembly and complex formation may be most relevant to its function with LINKIN and α-tubulin. Based on previously known RUVBL functions, we propose that RUVBL1 and RUVBL2 form a heterometric ring structure that promotes assembly of a LINKIN complex and nucleation of microtubules (Figure 10).

Figure 10

Download asset Open asset

Most adhesion receptors interact with the cell cytoskeleton through their intracellular domain (Juliano, 2002); LINKIN interacts with microtubules. Microtubules are known to play important functions in cell migration and tissue organization (Gauthier-Rouvière et al., 2004; Etienne-Manneville, 2013). They provide structure and rigidity to tissues so that they can withstand high compression forces (Brangwynne et al., 2007), and the bundled cortical microtubules of gonadal cells likely provide such a structure. Microtubules are also involved in creating polarity and trafficking components along polarized tracks. The subunits have a plus and minus end polarity growing out from the centrosome, which in turn can create a front–back polarization in migratory cells through selective stabilization of fibers (Wadsworth, 1999) and polarized trafficking to the membrane (Miller et al., 2009). Microtubules serve as tracks to transport cadherin-containing vesicles to specific areas of the plasma membrane to establish cell adhesion (Mary et al., 2002). LINKIN may serve to anchor the microtubule cytoskeleton to particular domains of the plasma membrane. Conversely, microtubules and microtubule-associated motor proteins may transport LINKIN to select domains of the plasma membrane, helping to establish LINKIN localization to apical domains and cell–cell contacts.

As investigations into other cell adhesion molecules have revealed numerous important functions in animal development (Thiery, 2003; Halbleib and Nelson, 2006), LINKIN's expression in many human and C. elegans tissues suggests that future studies will demonstrate its involvement in many processes. We have demonstrated an adhesion function for the transmembrane glycoprotein LINKIN and have identified interactors RUVBL1, RUVBL2, and α-tubulin, which support a model for LINKIN regulating the microtubule cytoskeleton. Considering that the interactions between LINKIN, RUVBL1, RUVBL2, and α-tubulin were identified using a human cell line and were also required for gonad cell adhesion in a nematode worm, these interactions may have a conserved molecular function in Metazoa including human. We are proposing that LINKIN functions in maintaining tissue integrity through cell–cell adhesion and apically polarization, but further studies are necessary to show the generality of this function and to elucidate differences between the roles of LINKIN and other adhesion receptors.

C. elegans strains were cultured at room temperature using standard protocols unless indicated otherwise (Brenner, 1974). Strain VC877 tag-256(gk367)/hT2 was obtained from the Caenorhabditis Genetics Center (CGC). Other alleles and transgenes used in this study are him-5(e1490) (Hodgkin et al., 1979) | PS4730 syIs128 [lag-2::YFP]; him-5(e1490) (Kato and Sternberg, 2009) | syIs78 [ajm-1::GFP] (Gupta et al., 2003) | inx-5::GFP | PS6018 unc-119(ed4); him-5(e1490) syEx1130[tag-256::TAG-256::YFP(20 ng/µl) + unc-119(+) (70 ng/µl) + Bluescript] | PS6372 tag-256(gk367); him-5(e1490); syEx1184[tag-256::TAG:256(5 ng/µl) + myo-2::mCherry].

Total RNA was extracted from 25 lnkn-1(gk367) mutant animals using a TRIzol extraction method (Chomczynski and Sacchi, 1987) but modified for small quantities of worms (Morimoto group, http://groups.molbiosci.northwestern.edu/morimoto/research/Protocols/IX.%20C.%20elegans/B.%20Extraction/2.%20TotalRNA.pdf). This was followed by a reverse transcription reaction using SuperScriptIII first-strand synthesis supermix (Invitrogen, Waltham, MA) following manufacturer's instructions and PCR using lnkn-1 primers to the beginning and end of the gene. The amplified product was submitted for DNA sequencing (Laragen).

See Supplementary file 3.

For lnkn-1 mutant rescue experiments, a DNA mixture of 5 ng/μl or 1 ng/μl of lnkn-1 rescuing construct, 7 ng/μl of myo-2::dsRed, and 150 ng/μl of 1 kb ladder (NEB) as carrier DNA, was injected into adult gonads of PS6372 lnkn-1(gk367)/hT2; him-5(e1490) hermaphrodites.

Genecards.org was used to identify the C. elegans homologs of the human genes identified by mass spectrometry. C. elegans genes were screened using an RNAi protocol previously described (Kamath et al., 2001), with a few modifications. Single RNAi bacterial colonies were grown for 6–8 hr in LB with carbenicillin selection (100 μg/ml). Carbenicillin (25 μg/ml) and IPTG (1 mM) were spread on NGM plates just prior to adding 200 μl of RNAi bacterial culture, and plates were dried at RT overnight. The following day, eggs were harvested from gravid adults by bleaching and placed on plates containing RNAi bacteria. Animals were grown at RT and L4 stage males were scored by Nomarski and fluorescence microscopy for detached gonad. The RNAi bacteria were obtained from the Vidal library (Rual et al., 2004) when the gene was available, and the Ahringer library (Kamath et al., 2003) otherwise.

Equal amounts of cell lysate were mixed with SDS sample buffer (4×) and heated at 90°C for 5 min. After centrifugation at 3000×g for 30 s, sample (25 μg) was loaded on a 4–12% or 4–20% SDS-PAGE. Proteins were transferred to nitrocellulose membranes and stained with Ponceau S. The nitrocellulose membranes were blocked with 5% milk/TBST for 5 min to remove Ponceau S and for additional 30 min. Primary antibodies were prepared in 3% milk/TBST and incubated for 1 hr at room temperature on a shaker. After removing the primary antibody, membranes were washed three times with 3% milk/TBST for 5 min each. Primary antibodies were used including rabbit polyclonal antibodies against human RUVBL1 and RUVBL2 (Proteintech), anti-β-tubulin and anti-β-actin. Secondary antibodies were incubated for 1 hr at room temperature on a shaker. Membranes were washed three times with TBST with 5 min each time. ECL Plus (GE healthcare) was used to detect signal.

HEK 293T cells were transfected with plasmids encoding ITFG1-Myc (RC204773, Origene), Flag-RUVBL1 (51635, Addgene, Cambridge, MA), and HA-RUVBL2 (51636, Addgene) or with control Myc-vector. Mem-PER Plus Membrane Protein Extraction Kit (Thermo Fisher Scientific) was used to isolate membrane and cytoplasmic fractions from harvested cells. Flag-RUVBL1 was immunoprecipitated from the two fractions using Flag-beads (Sigma-Aldrich, St. Louis, MO). Immunoprecipitation/Western blot analysis was performed as described above.

A soluble LNKN-1 protein containing the entire extracellular domain was expressed and purified from S2 cells by the Caltech protein expression facility. Rabbit polyclonal antibodies were generated against this LNKN-1 extracellular domain protein and a 17 AA intracellular domain peptide, Ac-C(Ahx)DRYERQQQSHRFHFDAM-OH (QCB, Hopkinton, MA). The antibodies were affinity-purified using their antigens, either the extracellular domain protein or the intracellular domain peptide. Rabbit polyclonal antibodies were also generated against the entire RUVB-1 and RUVB-2 proteins (Proteintech) and affinity-purified using the RUVB-1 and RUVB-2 proteins. Specificity of all antibodies was tested by staining tissues from C. elegans that were either treated with RNAi against the antigen gene or mutant for the gene. Reduction in staining was observed for each of these antibodies (Figure 3—figure supplement 1, Figure 6—figure supplement 1).

Gonads were dissected from larval stage males following Chan and Meyer's ‘Protocol 21: Antibody staining of C. elegans gonads’ (Shaham, 2006). A final concentration of 2% paraformaldehyde was used and phosphate buffered saline was substituted for sperm salts. Primary antibodies were used at 1:250 dilution for antibody against LNKN-1 extracellular domain, RUVB-1, and RUVB-2 and at 1:200 dilution for LNKN-1 intracellular domain. Mouse monoclonal antibodies against α-tubulin (12G10, supernatant, Developmental Studies Hybridoma Bank, Iowa City, IA) and DLG-1 (DLG-1, supernatant, DSHB) were used at a 1:100 dilution. Secondary antibodies against rabbit (Alexa Fluor 594 Goat Anti-Rabbit IgG, Life Technologies, Carlsbad, CA) and mouse (Alexa Fluor 594 Goat Anti-Mouse IgG, Life Technologies) were used at 1:500 dilution. Tissues were mounted on slides using Vectashield mounting media containing DAPI (Vector Laboratories, Burlingame, CA).

LINKIN sequences from H. sapiens, M. musculus, D. melanogaster (uniprot.org), C. elegans (wormbase.org) were aligned using clustalw (http://www.genome.jp/tools/clustalw/) and clustalo (http://www.ebi.ac.uk/Tools/msa/clustalo/). T. adhaerens LINKIN is TRIADDRAFT_52570 but the gene prediction is imperfect (metazoa.ensembl.org).

The FG–GAP domain, first identified in α-integrin, contains a loosely conserved sequence of Phe-Gly and Gly-Ala-Pro. A motif frequently found between the FG and GAP sequences is a DxDxDG calcium-binding motif (D = Asp, G = Gly, x = AA; Chouhan et al., 2011). While DxDxDG is a common calcium-binding motif, variations on this motif depend on the particular protein family (Rigden and Galperin, 2004). A comparison of all human α-integrin FG–GAP domains showed that their calcium-binding motif has a strong DxxxDxxxD signature, which contains a more weakly conserved DxDxDG sequence as its first 6 AAs (Chouhan et al., 2011). This DxxxDxxxD signature is different from DxDxDG calcium-binding regions of other proteins like the EF-hand protein family (Rigden and Galperin, 2004). The FG–GAP sequence was only loosely conserved and not always identifiable, but the DxxxDxxxD calcium-binding domain was highly conserved in LINKIN.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We are grateful to Amir Sapir, Hillel Schwartz, and Srimoyee Ghosh for critical reading of the manuscript and to TFC's postdocs, Xiaoyi Zhaang and Lin Gui, for assistance with IP/Western blot analysis experiments. Caltech protein expression facility expressed and purified LNKN-1 extracellular domain. Mass spectrometry experiments were performed at the Caltech protein exploration laboratory directed by Dr Sonja Hess and Prof Ray Deshaies. Wormbase (wormbase.org) was a valuable resource for C. elegans information as was the British Columbia C. elegans Gene Expression Consortium (http://elegans.bcgsc.ca/home/ge_consortium.html) GFP expression pattern database. The monoclonal antibody developed by Frankel J and Nelsen EM was obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH, and maintained at The University of Iowa. Nematode strains were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources. CZY was funded by a CIRM pre-doctoral training grant. PWS is an investigator with the Howard Hughes Medical Institute, which supported this work.

1. Received: August 21, 2014
2. Accepted: November 28, 2014
3. Accepted Manuscript published: December 1, 2014 (version 1)
4. Version of Record published: December 24, 2014 (version 2)

© 2014, Kato et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.