The protein Sam68 is a well-known SH3-domain binder comprising an exceptionally large number of seven potential PxxP ligand motifs. To comprehensively characterize the SH3 binding potential in an unbiased manner, we conducted a phage-display-based screening of Sam68 against a library containing the entire human SH3 proteome. Thereby we identified twelve high-confidence binders, five of which are described for the first time to our knowledge. Furthermore, we identified a set of 48 SH3 domains, which might contain lower-affinity interactors, among them again some already known Sam68-binders such as Grb-2 or Vav1. Extension of the analysis would presumably have led to the classification of more domains as high-confidence binders, and to the identification of more lower-affinity binders, as even some of the already known binders remain undetected. Moreover, in the case of proteins with more than one SH3 domain, cooperative binding to different PxxP motifs might be necessary for a high-affinity interaction 
. As these domains are presented separately on different phages, such proteins might elude identification in the bio-panning, thus possibly explaining why e.g. Grb-2, which has been shown to bind to Sam68 via both of its SH3 domains 
, was only among the lower-affinity binders. Finally, we cannot rule out that the structure of individual SH3 domains is compromised on the phage surface.
For an SH3 domain subset consisting of the highest-affinity binders, we confirmed the Sam68-interactions in independent assays, i.e. in vitro
by GST-SH3-pull-down-assays and in vivo
by FRET-analysis using fluorescent-protein-fusions. However, differences in the relative interaction strengths were observed between the various assays for some pairs (compare e.g. affinity of Src in the ELISA with the band intensity in the pull-down assay, or Fyn in the ELISA vs. the FRET-analysis). Most likely these differences are due to post-translational modifications of Sam68, which influence its interaction capacities. As mentioned in the introduction, Sam68 is subject to S/T-phosphorylation 
, Y-phosphorylation 
, acetylation 
, methylation 
, or sumoylation 
. Furthermore, the affinites may be influenced by assay-specific constraints. For instance, the FRET efficiency also depends on the spatial orientation of both fluorophors towards each other, which might vary for the different SH3-YFP fusion proteins despite very high similarity in the overall structure.
Our analysis of the very PxxP-motifs engaging SH3 domains shows a delicate selectivity of certain motifs and, considering the three intertwined but formally separable PxxPs of P5, or the basic aa in P0 and P4, even the importance of individual amino acids. Only P0, P3, P4, and P5 constitute SH3-domain target sites. Mutations in Sam68 inactivating these four motifs suppressed any interactions with SH3 domains, thus ruling out functionality of P1, P2 and P6 as SH3 ligands. The absence of SH3-interactions of P1, P2, and P6 suggests SH3-independent functions of these motifs, i.e. interactions with other domains recognizing proline-rich sequences, like WW-domains 
. The various SH3 domains have special preferences to the four motifs concerning selectivity and affinity. The recognition pattern of Src-kinase-family SH3 domains is quite similar, with major preference for P5, while it is completely different to the pattern of e. g. intersectin 2 or the osteoclast stimulating factor.
In conclusion, the diverse preferences of the different SH3 domains for certain PxxP-motifs constitutes a prime example for the high selectivity of SH3 domains for their target sequences. Moreover, in the cellular context, it is conceivable that yet an increase in specificity is achieved for proteins with more than one SH3 domain (i.e. Intersectin 2 (5 SH3s), Nck1 (3 SH3s), CIN85 (3 SH3s)) by cooperative binding to different PxxP motifs of Sam68, as it has been suggested for the interaction of Nck1 with its binding partner Cbl 
To rank SH3 domain affinities towards Sam68 a phage-ELISA analysis was performed. As outlined in the results section, calculation of Kd-values relies on the estimation of the mean number of SH3-domains present on one phage particle. This number was deduced from a comparison with the Nef-Hck-SH3 pair, for which a Kd-value of 250 nM has been determined by surface plasmon resonance measurements 
. Thereby, we obtained a value of 420 pVIII-SH3 proteins (SH3 domain ≈7 kDa), corresponding to 26 % of the approx. 1600 pVIII surface proteins. This number is plausible when compared to values from the literature: Short 15-meric peptides (≈1.7 kDa) are incorporated as pVIII-fusions at 30–40 % 
, while antibody-Fab-fragments (≈50 kDa) are only incorporated at less than 1% 
Remarkably, Kd values calculated for the Sam68-SH3-interactions (considering the aforementioned correction value) lie in the low nanomolar range (cf. ). This is unexpected for SH3 domains, whose affinities are considered to lie in the low micromolar range 
. However, critical examination of the literature challenges the generality of the latter proposition. Several examples can be found for much better SH3-interactions (e.g. Pak2 with β-Pix-SH3 at 59 nM 
), and Kd values for SH3-domains have often been determined only for short peptide-ligands and not the whole proteins. This can have a significant influence on binding-strength, as illustrated for instance for the Abp1-SH3 domain, comprising a Kd-value of 100 µM to a 14-mer ligand-peptide, and 40 µM after elongation to a 17-mer peptide 
. Nevertheless, some values obtained for Sam68-SH3 interactions still are one order of magnitude lower than even the best reported in the literature. Likely, this is due to an artifical avidity effect resulting from the use of the SH3-phages. As the SH3 domains bind to more than one of the PxxP-motifs, it is conceivable that one phage-particle docks to two or more PxxP-motifs of an individual Sam68 molecule via multiple SH3 domains. Consequently, even after dissociation of one SH3-PxxP-pair, the phage would still be retained by the protein. Kinetically, this corresponds to a decrease in the off-rate and concomitantly to a decrease in the Kd value. The affinity gain of the interaction is not due to cooperativity, as is evident from the Hill-transformed ELISA data yielding Hill-coefficients α of 1.0. Rather, the increase can simply be attributed to enhancement as defined by Mammen et al.
due to the polyvalent nature of the interaction. In fact, binding curves from phage-ELISAs with the PxxP-peptides instead of full-length Sam68 indicate weaker interactions, supporting the above observation of binding enhancement. In conclusion, the given data represent the apparent Kd-values of the interaction between SH3-phages and Sam68, which nevertheless allow for comparison of SH3-domain binding stengths on a relative scale.
Apart from those SH3 domains binding Sam68 with high affinity that have already been described in the literature, we identified five new ones: Intersectin 2 (IS2), nephrocystin, sorting nexin 9, Cbl-interacting protein of 85 kDa (CIN85), and Osteoclast stimulating factor 1 (OSF).
Intersectins 1 and 2 are implicated in Clathrin-dependent endocytosis 
. They comprise a number of protein-interaction domains, among others five SH3 domains each. Intersectins are considered as scaffold-proteins organizing components of the endocytosis machinery. A similar function is ascribed to the Cbl-interacting protein CIN85, which facilitates endocytosis of receptor tyrosine kinases after activation by ligands 
. Sortin nexin 9 is involved in endocytosis as well, likely by linking the key GTPase dynamin to the actin cytoskeleton 
. Notably, some Sam68-binding SFKs are implicated in endocytotic processes as well, like Hck, which is involved in the regulation of actin-dependent processes during phagocytosis 
. Taken together, the identification of several Sam68-binders that are involved in endocytosis strongly suggests a so far unknown function of Sam68 in this central biological process. Endocytosis plays an important role in many signalling processes such as activation of the MAP-kinase cascade 
, and Sam68 might be engaged in cross-talk of these processes.
As implicit in the name, the osteoclast stimulating factor (OSF) plays an important role in osteoclast differentiation. It has been shown that expression of osf
leads to secretion of a so-far unknown factor, which induces differentiation of hematopoietic stem cells into osteclasts in cell culture 
. Furthermore, OSF interacts with Src, which is a noteworthy connection, as Src-/-
knock-out mice exhibit major bone deformations due to impaired osteoclast function leading to osteopetrosis 
. Integrating the observation that the Sam68−/−
knock-out mouse exhibits an osteopetrosis-phenotype as well 
, and the interaction between Sam68 and OSF, suggests a picture of an osteoclast-specific signal transduction pathway containing Src, OSF, and Sam68. The latter possibly facilitates phosphorylation of OSF by Src, functioning as a platform that brings both proteins close together. This view might help to understand the osteopetrosis phenotype of the Sam68−/−
knock-out mouse on a molecular level. Interaction of OSF with Src is in principle still possible, but maybe only occurs inefficiently, presumably translating into the milder bone-related phenotype for knock-out of Sam68 than for Src.
Similar roles in facilitating certain steps of signal transduction pathways are often carried out by scaffold proteins, a heterogeneous group of unrelated proteins. Classical scaffold proteins are defined by three criteria according to Zeke et al.
: (i) They possess no signalling-related catalytic activity by themselves, but (ii) directly interact with at least two proteins of a signalling pathway, that (iii) form a pair of a catalytically active protein and its corresponding target. Sam68 lacks catalytic activity and binds to a multitude of proteins even when putting the numerous SH3 domains aside, thus complying with the first two criteria. Regarding the third criterion, the here described OSF-Src-interaction is satisfactory. In principal, this characteristic has already been recognized by Richard et al.
for a different protein-pair, namely an SFK-member and phospholipase C gamma 1 (PLCG1). In their proposed model, the SFK phosphorylates PLCG1 after both proteins made contact to Sam68 
. Thus, our findings support their original farsighted proposition, and together the findings suffice to formally consider Sam68 as a bona fide
classical scaffold protein. However, Sam68 is unique in this group in two regards: First, it is predominantly located in the nucleus rather than in the cytoplasm like common scaffold proteins, and second, it is capable of binding RNA, thus adding another degree of complexity to the scaffolding-property. Hopefully, this view will help to better understand the multiple roles that Sam68 plays in the many different biological processes it is involved in. This demands the identification and characterization of the relevant Sam68-ligands, which actually mediate a certain function that is facilitated by Sam68.