To understand the mechanism of lipoprotein exchange we identified mutants universally defective in cgl
stimulation and protein transfer. Interestingly, these TraAB proteins were required in both donor and recipient cells. To our knowledge, this is the first bacterial transfer system where the same gene products are required in both donor and recipient cells. This finding and the ability of M. xanthus
cells to rapidly and homogeneously exchange lipoproteins 
implies that lipoproteins are transferred in a bidirectional manner. A bidirectional transfer mechanism is distinct from known secretion and conjugative systems 
, where proteins or DNA are transferred unidirectionally from donor to recipient cells.
Since OM lipoprotein exchange occurs efficiently and involves a form of bulk transfer 
, we hypothesized that OM lipids may also be exchanged. This hypothesis was supported by the finding that a lipophilic fluorescent dye was readily exchanged between cells. Importantly, transfer of lipophilic dye and hence membrane lipids, have the same stringent requirements in transfer as OM lipoproteins 
. That is, dye transfer only occurred when cells were motile within structured biofilms; no detectable dye transfer occurred in liquid or between nonmotile (non-stimulatable) cells on a solid surface. In addition, dye transfer required the TraAB proteins in donor and recipient cells. We thus conclude that dye exchange does not occur by diffusion or by diffusible OM vesicles, but instead requires specific cell-cell contacts mediated by cell motility. Based on earlier observations that OM, but not IM, lipoproteins are transferred 
, we surmise that only OM lipids are exchanged bidirectionally. Presumably transfer consists of the outer leaflet lipopolysaccharide (LPS) and the inner leaflet phospholipids. In this respect it is interesting to note that species of Borrelia
have been directly observed to fuse their OMs, a process apparently mediated by cell motility 
, and Bacillus subtilis
reportedly transfers proteins in biofilms via membrane enclosed nanotubes 
Based on sequence, domain architecture and functional similarities to eukaryotic proteins, we propose that TraA serves as a cell surface receptor. In particular, TraA has similarities to the Saccharomyces cerevisiae
FLO1 and FLO5 cell surface receptors/adhesions 
(). These FLO proteins have domain architecture consisting of a SS, N-terminal PA14 domain, a central tandem repeat region and a C-terminal protein sorting tag (GPI site; glycosylphosphatidylinositol anchor) for cell surface attachment 
. Thus, by analogy, we suggest that in TraA the SS serves to transport the protein to the periplasm followed by SS cleavage. The processed N-terminal PA14 domain would serve as a receptor for ligand binding, presumably a glycan. The cysteine-rich tandem repeats could serve as a rigid stalk for PA14 presentation on the cell surface. The MYXO-CTERM motif could function, analogous to a GPI site, in protein sorting to the cell surface. Recent reports suggest the MYXO-CTERM and related C-terminal tags () are widely distributed in bacteria and archaea, where they are proposed to be posttranslationally modified and direct protein sorting to the cell surface 
. Although initial attempts to generate TraA antibodies or fluorescent protein fusions were unsuccessful, TraAB overexpression was found to dramatically increase the ability of cells to adhere to one another (Figure S3
). This result is consistent with TraA serving as a cell surface adhesin. Furthermore, the identification of the traA227P→L
missense mutation within PA14 highlights the importance of this domain for function (). We also note that Dictyostelium discoideum
, a eukaryotic soil slime mold that exhibits similar multicellular behaviors as M. xanthus
, produces two secreted signals, called DicA1 (PsiF) and PsiA, whose proteins contain PA14 domains followed by cysteine-rich repeats (Pfam00526) of various lengths that show some resemblance to TIGR04201 
. Thus, M. xanthus
and other microbes, including eukaryotes, appear to utilize PA14 encoding proteins as extracellular signaling and recognition molecules to mediate social interactions.
Recent bioinformatic analysis suggests gram-negative bacteria encode C-terminal protein sorting tags that function analogously to the well-characterized gram-positive LPXTG/sortase system 
. In the case of MYXO-CTERM, we postulate that this motif forms a transmembrane α-helix and anchors pre-TraA into the IM 
. Here the Arg rich C-terminal tail would reside in the cytoplasm, while the remainder of the protein would be in the membrane or periplasm (). Thus analogous to lipoprotein processing 
, an acyl transferase could attach a lipid moiety via a thioether bond to the invariant Cys ( and Figure S7
). Subsequently, an endoprotease would cleave the TIGR03901 motif downstream of the aforementioned Cys residue. Once processed a system analogous to the Lol pathway could transport these proteins to the cell surface.
As the traB gene overlaps in a bicistronic operon with traA () and mutations in each gene elicit identical phenotypes, suggests that TraAB likely function in the same transport pathway. Since the C-terminal region of TraB contains an OmpA-like domain (Pfam00691), it likely binds non-covalently to the cell wall. The N-terminal region constitutes the majority of this protein (~400 amino acids) and has no ascribed function (), but theoretically could interact with the OM and even traverse the OM to interact with TraA. It is also plausible that TraB may facilitate TraA's localization to the cell surface.
A working model for the mechanism of cell contact-dependent exchange is outlined in . First, cell-cell recognition is postulated to be mediated by TraA serving as a cell surface receptor. We suggest that the distant PA14 domain may function in ligand binding to neighboring cell surfaces. Glycans found in LPS or glycoproteins are possible ligands. In a variation of this model TraA may function as a homophilic receptor. Similar to the FLO1 system, a key component of this model involves reciprocal TraA binding by both cells. A ‘donor’ cell was arbitrarily assigned and its OM (mCherry) lipoproteins were symbolized as red lollipops. Upon aligned cell-cell contact and docking the OM membranes of adjoining cells fuse. Although not directly depicted, TraAB may facilitate membrane fusion by bringing OMs into close proximity and perhaps causing local membrane perturbations that help catalyze OM fusion. Membrane fusion may also be facilitated at cell poles where the membranes have high tip curvatures and thus are more fusogenic 
. Once cells are adhered cell motility could also stress the membrane. Upon OM fusion, lipids and lipoproteins rapidly exchange bidirectionally; a process presumably driven by lateral diffusion. Integral and associated OM proteins are also likely transferred as the CglE and CglF proteins encode type I signal sequences 
. It is unknown whether soluble periplasmic proteins are transferred. Prior studies clearly indicate inner membrane lipoproteins and cytoplasmic proteins are not transferred 
. Following fusion cells physically separate, a process likely facilitated by gliding motility.
Working model for OM and lipoprotein transfer between M. xanthus biofilm cells.
The exchange of OM lipoproteins has phenotypic consequences to the cell, including complementation of mutational defects (). Whether the restoration of mutation defects is ecologically important is unknown; however population heterogeneity within biofilms, especially from an environmental setting are significant 
, and consequently some individuals within a population are less fit. Thus, we hypothesize that the ability to exchange and share the OM proteome allows some individuals to gain fitness and for the population to establish OM homeostasis. In turn, homeostasis may increase population fitness by normalizing intercellular signal output and reception by reducing population heterogeneity. Thus community behaviors, such as swarming and development might be better coordinated. In this respect, our findings that a mixture of nonmotile cells with motile cells inhibits the latter cells from swarming in a TraAB and time dependent manner (), suggests these cells are communicating and coordinating their behaviors via OM exchange. Similarly, OM exchange can regulate development behaviors between genetically distinct strains (). The use of strain mixtures to study cell-cell interactions in motility and development is ecologically relevant, as diverse M. xanthus
isolates are found in close proximity in nature 
. The mechanism for developmental inhibition by nonmotile cells on motile cells is unknown, but may simply reflect a block in motility () 
. Alternatively or in addition, OM exchange with nonmotile cells may transmit a signal that blocks development. Currently, we are investigating the nature of these putative signals.
Our results indicate that myxobacteria exchange and thus share a significant amount of their cellular resources. This has led us to hypothesize that cell contact-dependent OM exchange represents a form of cooperative social behavior that may involve kin recognition. A kin recognition mechanism avoids the theoretical and ecologically relevant concern that ‘cheater’ cells could exploit or disrupt this social behavior to obtain resources 
. This problem is highlighted by observations that environmental M. xanthus
populations arise from diverse origins 
. Thus unlike artificial laboratory settings where multicellular behaviors are typically studied with a single homogenous culture, natural myxobacteria isolates must recognize kin from non-kin cells as they vacillate between single cell and multicellular life.
The data presented here provide three lines of evidence that cell contact-dependent OM exchange involves kin recognition. First, TraAB proteins are required in both ‘donors’ and ‘recipients.’ Thus if one cell does not express TraAB, exchange cannot occur. Second, exchange appears bidirectional, thus both cells are giving and receiving. Therefore, there is no inherent advantage one cell type has over another, unless one cell is starving and has depleted resources. Third, TraA contains a PA14 domain, with features resembling PA14 domains in yeast flocculin proteins involved in kin recognition and social behaviors. More specifically, flo1
and other genes within this group were classified as ‘greenbeard’ genes, which by molecular definition are cell surface receptors that recognize other cells carrying the same gene to provide social preferential treatment 
. In the case of FLO1 the protein allows yeast cells to enter the protective domain of a floc, where cells are so tightly joined they become deformed. Within flocs cells are protected from environmental stresses and cheater cells (flo1−
) cannot enter 
. In another greenbeard example, the Dictyostelium csA
gene, which encodes a homophilic cell surface receptor, plays a discrimination role in partitioning cells to desirable locations within fruiting bodies 
. Current experiments are testing whether TraA plays such a role.