Our experiments demonstrate that the 70 kDa heat shock protein DnaK is able to bind to 13-mer peptide segments of the preproinsulin molecule and that the binding is limited to four distinct preproinsulin regions, amino acids S7-23 of the signal peptide, B9-25 of the B-chain, C15-31 of the C-peptide and A6-21 of the A-chain. These peptide regions exhibit the leucine-rich motif found essential for binding of DnaK , usually a hydrophobic core of four to five residues enriched in L, but also in I, V, F and Y, and two flanking regions enriched in basic residues. Of the proinsulin molecule, strongest binding occurred to peptide 18 (B11-23), and this is probably mediated by a distinct hydrophobic motif formed by the three amino acid residues L-Y-L in the central positions B15, B16 and B17. This conclusion is supported by the finding that the B-chain-derived peptide B18-30, which comprises the six C-terminal residues of B11-23 but lacks the central L-Y-L sequence or a corresponding hydrophobic motif exhibited no significant DnaK affinity.
Native insulin exhibited only low affinity to DnaK. This observation indicates that steric hindrance prevents effective interaction with the chaperone, possibly due to the conformation of the polypeptide heterodimer or the aggregation of insulin molecules into dimers, trimers or hexamers, rendering the regions B11-23 and A6-21 less accessible. As deduced from the crystal structure of the insulin molecule http://swissmodel.expasy.org/repository, the B11-23 epitopes are buried at the interface between two mature insulin heterodimers, thus preventing access to the peptide binding site of DnaK. Similarly, the A-chain region A6-21 comprises a number of amino acid residues localized at the interfaces of the insulin oligomers. In particular, the surface-exposed E-residue in position A17 contributes to the stabilization of the dimer-dimer interfaces by its interaction with the F-residue in position 1 of the B-chain. Moreover, the C-residue in position A7 which is located at the surface of the insulin molecule, forms a stabilizing interchain disulfide bond with the C-residue in position 7 of the B-chain [24, 25].
To further elucidate a potential impact of steric conditions on DnaK-insulin interactions we investigated the DnaK binding of a proinsulin variant with an inversion of the K-P sequence in positions B28 and B29 causing a decreased tendency to oligomerization. This variant indeed exhibited a much higher DnaK affinity than mature insulin. However, the Kd obtained for the mature insulin molecule may only represent an estimate for the (low) chaperone affinity of the hormone, as the fluid phase system applied in our approach is optimised for studies on the interaction of chaperones with peptides rather than larger proteins. Nevertheless, previous studies which applied experimental systems designed to investigate the binding of DnaK to polypeptides demonstrate weak, but significant interactions between DnaK and full-length, native proteins . In those studies, DnaK binding was found to depend on the accessible hydrophobic area of a protein and was characterized by dissociation constants of 1 - 50 μM, a range comparable to the Kd of 67.8 ± 20.8 μM found for insulin in our current study. Taken together, these observation suggest that not only the (pro)insulin peptides but also mature proinsulin monomers can interact with DnaK and that there is little interaction between DnaK and native oligomeric forms of insulin.
In our current approach we selected insulin as the predominant β-cell specific (auto-) antigen. DnaK was selected as a representative member of the dominant and phylogenetically high conserved hsp70 family with the best characterized chaperone activity [27, 28]. Although the members of the hsp70 family were found to exhibit some differences in peptide binding specificities  extensive comparative analyses revealed a common binding specificity for peptide stretches of at least seven amino acid residues with a hydrophobic core [7, 28] thus largely resembling the hydrophobicity distribution pattern of the insulin-derived peptide B11-23. Therefore, our analyses on the interaction of potentially autoantigenic insulin peptides with the chaperone DnaK was performed in a suitable model system which only partially reflects the situation in vivo, in human patients. Further detailed studies of the interaction of insulin and proinsulin with human hsp70 isoforms will be needed to fully understand the rules of (pro-)insulin-chaperone interaction.
Besides the chaperoning function, one other property of hsp70 family chaperones is the ability to induce T-cell immune responses to complexed peptides [14, 15, 29–31], and a role of this pathway has been postulated for the induction of autoimmune disease [12, 32, 33]. It therefore is noteworthy that the three DnaK-binding regions of proinsulin, amino acids B9-25 of the B-chain, C15-31 of the C-peptide and A6-21 of the A-chain superimpose with the three major target regions for T-cell autoimmune reactivity in the NOD mouse model or in humans with β-cell autoimmunity [18, 34–36]. In the mouse model, the B-chain region around B16 was identified as critical for disease development, i.e., mutating the native E in position B13 to Q abolished T-cell reactivity to the peptide B9-23 . Moreover, replacement of Y to A in position B16 abrogated the response of B9-23 - reactive T-cell clones as well as the development of diabetes in vivo . I.e., the peptide B11-23 is closely associated with diabetes pathogenesis, and its diabetogenic potential depends on an intact core sequence, comprising the region which is important for the binding of this peptide to DnaK.
As the insulin molecules with the mutated B-chains completely retain their metabolic activity [34, 37] it has to be assumed that a single amino acid-exchange does not impair the formation of the correct tertiary structure, a complex process that most likely involves proper binding and assistance by chaperones. The findings from the mutated insulin molecules therefore further support the view, that the peptide binding motifs of chaperones are not determined by strictly defined amino acid sequences but by more roughly defined peptide properties, e.g. by the hydrophobicity distribution pattern within a stretch of amino acids . In contrast, T-cell receptor-mediated peptide recognition is known to be a highly selective and largely amino acid sequence specific process. Based on these considerations it may be speculated that chaperones are able to bind defined sets of insulin peptides characterized by hydrophobic motifs. Chaperone-peptide complexes may be recognized by antigen presenting cells which perform the ultimate decision on the pathogenicity of a peptide by applying stringent, highly sequence-sensitive selection criteria during processing and re-presentation of the peptide in the context of MHC structures and co-stimulatory signals.
Further studies are warranted to analyze whether enhanced binding of proinsulin monomers or misfolded (pro)insulin to hsp70 family chaperones during periods of β-cell stress contributes to the pathogenesis of autoimmune diabetes.