S100A4 is a small (approximately 12 kDa) acidic calcium-binding protein that has been associated with a range of biological functions, such as cell migration, invasion and angiogenesis, potentially contributing to higher metastatic capacity of tumor cells [1-4]. In line with this, increased expression of S100A4 has been correlated with adverse prognosis in patients with various types of cancer [5].

S100A4 belongs to the S100 protein family comprising at least 20 members, of which all are exclusively expressed in vertebrates. The human variant of S100A4 consists of 101 amino acids and is characterized by two calcium-binding EF-hands connected by an intermediate region referred to as the hinge region, and a distinct C-terminal extension. The S100 protein family shows a high degree of sequence homology, especially in the calcium binding EF-hand regions, while the composition of the C-terminal extension and the hinge region is more diversified and thus characterize each member [6,7]. Numerous studies indicate that S100A4 is arranged as homodimers held together by non-covalent bonds, and that this dimerization is important for the biological function. Upon calcium binding the homodimer undergo a conformational change that leads to exposure of the hydrophobic regions in the C-terminal end, initially buried in the complex [8].

S100A4 is located both in the cytoplasm, extracellularly [9], and also in the nucleus of tumor cells [10], but the mechanisms for transport and homing to subcellular compartments remains largely unexplored. The protein has also been found expressed in a variety of different normal cells [11]and the release of S100A4 into the extracellular space may thus originate both from tumor and/or stromal cells [12].

Most of the reported intracellular effects of S100A4 are associated with cytoskeleton rearrangements that may influence cellular motility [13-17], and extracellulary added S100A4 has also been shown to boost migration of astrocytic tumor cells [18]. In addition, the extracellulary added protein may sensitize osteosarcoma cells to INF-γ mediated apoptosis [19], and provoke degradation of the extracellular matrix (ECM) by augmenting the levels of matrix metalloproteinases [20]. The fact that S100A4 induces remodeling of the ECM suggests that the protein may also affect angiogenesis [21]. The mechanism by which S100A4 exerts its many reported and partly contradictory biological functions is not well understood, but one hypothesis could be that the protein exhibits different functions depending on its subcellular localization and/or posttranslational modifications.

A vast number of posttranslational protein modifications (PTMs) have been described [22]. PTMs may result in physiochemical changes of the protein with respect to mass, charge, structure and conformation, and thereby alter functional properties of the protein, such as binding affinity, enzyme activity and protein hydrophobicity. As a consequence, PTMs may target a protein for compartmentalization, degradation and protein-protein interactions. During the last few years progression in genomic and proteomic technology has pushed the limits of protein knowledge and one has realized the functional importance of several novel PTMs other than the more common. As an example, human p53 has been reported to be potentially post-translationally modified on at least 18 sites [23] including phosphorylation, acetylation and sumoylation, with implications for DNA binding, stability, oligomerization, nuclear import/export and ubiquitination [24].

The localization of S100A4 in the different cellular compartments suggests some type of subcellular targeting and transport. The present study was undertaken to investigate whether PTM of S100A4 could cause specific localization of the protein. Such modifications could, through increased calcium binding potential or altering the functional C-terminal extension, also have impact on the biological function of the protein when localized in the particular compartments. To study this we isolated S100A4 from several tissues, cellular compartments and also from recombinant sources and analyzed the protein by 2D-PAGE and western blotting. The S100A4 identity of several observed 2D spots were subsequently confirmed by the use of mass spectrometry (MALDI-TOF).

In the present work we have demonstrated that endogenous human S100A4 distributes in a characteristic pattern when separated by 2D-PAGE, and irrespective of source or subcellular compartment used for immunoisolation of the protein, the observed pattern was almost identical. Interestingly, a similar pattern, but with different quantitative distribution, was observed when analyzing different batches of recombinant S100A4. The identity of several examined spots was confirmed for the first time by mass spectrometric methods, verifying the existence of at least three charge variants of S100A4. The most likely interpretation of the presence of these different charge variants is that S100A4 is posttranslationally modified, since changes introduced during sample processing was largely excluded.

Immunoprecipitated human S100A4 was separated by 2D-PAGE in at least three (in some cases four) distinct spots of equivalent mass but with different charge. A comparable observation, although verified only by immunological methods, was made when S100A4 from tumor interstitial fluid from breast cancer patients was separated by 2D-PAGE [11]. Furthermore, in a recently published study on p53-related changes, two different PTMs of S100A4 were indicated by SELDI-TOF-MS [36]. These modifications, glutathionylation and cysteinylation, could theoretically have caused the distribution pattern observed in this study, however preliminary experiments indicate that this is not the case.

The finding that human S100A4 isolated from primary colorectal carcinomas, cell lines and normal cells resulted in similar patterns is remarkable and suggests that the observed 2D-PAGE distribution is a general phenomenon. This was also underscored by the identification of an almost identical pattern when the protein was isolated from nuclear, cytoplasmic and extracellular compartments. This is unanticipated since PTM often is transiently present and the occurrence is frequently time-, location- and site specific [37].

Surprisingly, all four recombinant S100A4 proteins exhibited more than one charge variant when separated by 2D-PAGE. Because of its relative ease in manipulation, E. coli has emerged as the most common hosts for heterologous protein production, however, one major disadvantage hamper its use; the inability to introduce important PTMs such as glycosylation, phosphorylation and acylation, which may be essential for biological activity [38]. In contrast to the pattern observed for endogenous S100A4, none of the recombinant proteins clearly displayed the characteristic picture of one main spot followed by gradually less intense spots, which would be expected for proteins altered by sequential PTMs. The choice of cloning strategy may result in the addition of extra amino acids and affinity tags on recombinant proteins that could influence 2D-PAGE migration patterns, potentially complicating the interpretation of such data; however, one might expect that the final protein distributed in one single spot. The observed variable distribution could possibly be linked to differences in biological effect when using different recombinant proteins. In theory, the observed distribution pattern of the recombinant variants could be caused by heterogeneous cleavage of the selection tags (GST/His) or by unspecific degradation, but examination of undigested samples by MALDI-MS did not support such an explanation (data not shown).

Since immunological methods depend on antibody specificity and the S100 protein family exhibits high sequence homology, the confirmation of the present results by non-immunological methods was warranted. Using MALDI-TOF mass spectrometry to analyze tryptic peptides from excised gel spots, the identity of several spots reproducibly confirmed the results obtained by western blotting. Unfortunately, we have so far not been able to identify and further characterize the present PTMs. One possible explanation could be that the modified fragments were undetectable since the sequence coverage of this small protein was only between 30 and 50%. Also, trypsin digestion might cleave potentially interesting parts of the molecule into fragments that are too small for further analysis. We are in the process of establishing protocols using other digestive enzymes and MALDI-matrixes ensuring better coverage of the protein sequence.

Although the actual PTMs have not been identified, the observed 2D pattern gives several indications with respect to the type of the putative modifications. Since the apparent mass did not change significantly between the spots, all PTMs introducing heavy residues or major deletions in the amino acid sequence can be excluded. Furthermore, the visual impression of the 2D pattern suggests the presence of one quantitatively dominating variant, with less abundant variants distributing towards lower pH, which could indicate either successive loss of positive charges or gain of negative charges, resulting in gradually more acidic forms of the protein. Even with these limitations a large number of PTMs are possible from a physicochemical point of view, and although a comprehensive analysis cannot be made, it is tempting to speculate on a few of the more likely alternatives. Phosphorylation introduces a negative charge on Serine, Threonine or Tyrosine residues while only changing the mass by 80 Da. Analyzing the S100A4 amino acid sequence in NetPhos revealed three Serine residues as likely phosphorylation sites, and the observed 2D pattern match well with the predicted pattern if such modifications were present. Another possible modification that would introduce a negative charge with minimal mass change is deamidation, in which Asparagine or Glutamine is converted to Aspartic or Glutamic acid. A prediction algorithm for deamidation of S100A4 suggested the presence of three labile Asparagines which, if sequentially deamidated, theoretically would produce a 2D pattern similar to the one observed in our experiments. Another frequently occurring PTM involves cleavage of peptide residues resulting in loss of amino acids either from the N- or C-terminal end. Such loss of the three most C-terminal amino acids in S100A4 might also explain the observed pattern. All three PTMs suggested above could potentially have an impact on biological function, subcellular targeting or protein interaction of S100A4.

It is becoming increasingly evident that PTMs may be involved in the modulation of protein synthesis, degradation, activity, and molecular interactions, of which all may have implications for multiple cellular events [39,40]. S100A4 has been associated with many biological functions, some of which are probably related to localization in specific tissues and cellular compartments, and it may be possible that PTMs could be a permissive requirement for the exertion of some of the functions attributed to this protein.

Given the high sequence homology within the S100 family, modifications observed in other S100-proteins might also be relevant for S100A4, even if individual family members have discrete functions and tissue distribution. The S100A11 protein has been found translocated to the nucleus upon phosphorylation [41], and we therefore hypothesized that phosphorylation could also be involved in translocating S100A4 to the nucleus. The observed comparable distribution pattern of S100A4 from different compartments would in principle argue against such hypothesis. However, even with apparent qualitative similarity in observed 2D patterns subtle quantitative differences, not apparent by western blot or silver staining analysis, might exist between corresponding spots from different sources, reflecting stochiometric variation of unmodified and modified S100A4. Recently, the modification of S100A8 and S100A9 by introduction of N^ε-carboxymethyllysine was demonstrated to be associated with sustained intestinal inflammation through NFκB activation [42]. The addition of extracellular S100A4 has been shown to induce NFκB activity in a low S100A4 expressing osteosarcoma cell line [19], but the possibility of a PTM being involved in such activation has yet to be explored.

At present we do not know which PTMs are present in S100A4 or the location in the amino acid sequence, but speculations regarding the purpose of possible PTMs can be made. PTMs in the hinge region could affect steric properties, while changes to the C-terminal end, which is exposed upon calcium binding, could directly affect the proteins interacting capabilities. Both hinge and C-terminal regions have, based on their heterogeneity in the S100-family [8], been proposed to be important for the biological activity of each S100-protein [43], which further suggests that PTMs in these regions might also be specific to each S100-protein. One of the main features of the S100-proteins is their calcium binding motifs, the EF-hands, and which have also been shown to be important for dimerization and biological functions [6], is also the most conserved region among the S100 family members. PTMs introducing a negative charge or neutralizing an initially positive charge could potentially increase the calcium binding affinity [8], and the 2D pattern observed for S100A4 with gradually more acidic variants could thus suggest increased calcium binding capacity. For S100A9 and S100A12, phosphorylation of at least two different amino acids was shown to induce augmented calcium binding, which subsequently led to plasma membrane translocation of the proteins [44]. Since the calcium-binding domains are highly conserved, a general PTM that affects several of the S100-members could be possible. However, posttranslational modifications made directly to the EF-hands in the family of EF-hand proteins have to our knowledge not yet been described, and would therefore be of great interest in an even broader perspective.

Endogenously expressed S100A4 was demonstrated to exist in several charge variants, indicating the presence of posttranslationally modified forms of the protein. These charge variants appeared to be universially present in S100A4 isolated from different subcellular compartments and tissues. As part of the emerging field of oncoproteomics, PTMs are proving to be increasingly relevant for the understanding of cancer biology and as potential drugable targets [45,46]. S100A4 has in numerous studies already been established as an important determinant of cancer metastasis, and potential PTMs may be of relevance for determining the biological processes resulting in increased metastatic capacity. Further efforts will be made to discover the nature of the PTMs present in S100A4 by optimizing mass spectrometry strategies and to further screen for eventual differential presence of the PTM using 2-D Fluorescence Difference Gel Electrophoresis (DIGE).

MHH carried out the molecular work, mass spectrometric study and drafted the manuscript. KF conceived the study, participated in its design and coordination, and helped to draft the manuscript. S–OM participated in the design of the study and training MHH in mass spectrometry techniques. GMM conceived the study, participated in its design and coordination, and helped to draft the manuscript. All authors read and approved the final manuscript.

The pre-publication history for this paper can be accessed here: