We hypothesized that if San1 uses its disordered N- and C-terminal domains for substrate interactions, then each of the predicted binding sites within those domains should help provide distinct specificity for the differently shaped misfolded substrates that San1 encounters. We also considered the possibility that some of these sites might mediate the interaction between San1 and cofactors. Notably, both prediction methods accurately identified the ordered segments of the RING domain as potential binding sites; these segments contain the zinc-coordinating cysteine and histidine residues required for interaction with ubiquitin conjugating enzymes. Aside from identifying substrate and cofactor interaction sites, there remained the possibility that some or all of the binding sites predicted by the PONDR and ANCHOR algorithms do not exist, or that these sites mediate interactions completely unrelated to San1's function in PQC degradation.
To resolve these ambiguities, we needed to evaluate experimentally the importance of these predicted sites for San1-substrate interactions. We decided to conduct a systematic deletion analysis of San1 and test the resulting deletions for their ability to bind substrates using the two-hybrid assay. If San1 contains multiple substrate interaction sites, then no single deletion should eliminate San1-substrate interaction. Moreover, if these sites provide distinct specificities for different abnormally folded substrates, then each deletion would have varying effects on San1's interaction with different substrates. However, if San1 recognizes substrates through a single binding site, each deletion would have identical effects on San1's interaction with each substrate.
To rigorously test our hypothesis that San1 binds each of its substrates distinctly, we needed a larger collection of substrates. Because the two-hybrid assay faithfully reported San1's interaction with known substrates, we used another two-hybrid selection to identify more substrates. In this selection, we tested the interaction between San1 and a yeast-derived cDNA library and identified 28 unique interactors. Notably, no interacting chaperones or other adaptors came out of the library used in our genetic selection. To verify that the San1 interactors were in fact substrates, we tested the identified interactors for San1-dependent degradation using cycloheximide-chase assays. Most of the identified interactors (25 of 28) underwent San1-mediated degradation to some degree, confirming that we had identified San1 substrates and expanding our pool of substrates to 31 distinct, abnormally folded proteins.
Having developed a sufficiently sized catalog of San1 substrates, we were able to conduct an exhaustive San1 deletion and interaction analysis. We made 20 small deletions in San1, covering all of the conserved predicted binding sites and non-conserved disordered segments in San1's sequence. We then tested the San1 deletions for interaction with each substrate in our collection. We conducted our two-hybrid assays on three types of media, each with varying stringency for two-hybrid interaction, to distinguish more clearly the effects of each San1 deletion on each substrate interaction. In the results of our extensive two-hybrid analysis, we observed that each of the San1 deletions had distinct effects on San1's interaction with each substrate, supporting the hypothesis that San1 uses its multiple binding sites to specifically interact with each of its differently shaped substrates. We also observed that many of the San1 deletions only had observable effects on substrate interaction when tested on higher stringency media for two-hybrid interaction, indicating variable affinities for each substrate. We also tested a deletion of San1's RING domain in the two-hybrid and observed no effect on San1-substrate interaction, demonstrating that ubiquitin conjugating enzymes do not contribute to San1's substrate recognition.
We then tested the effects of San1 deletions on its ability to degrade its substrates in vivo. Because of the difficulty involved in testing each San1 deletion against our collection of substrates using conventional cycloheximide-chase assays, we opted to measure the effects of San1 deletions on the steady-state levels of green fluorescent protein-tagged (GFP) variants of representative substrates using flow cytometry. Using this assay, we observed distinct effects of San1 deletions on the steady-state levels of each substrate tested. Our results were analogous, but not identical to the results obtained using the two-hybrid, with the discrepancy largely attributable to changes in the constructs used and the differences in what the assays report. Taken together, the results of the two-hybrid and steady-state experiments confirmed the hypothesis that San1 uses multiple binding sites embedded in its disordered regions to distinctly interact with each of its differently shaped substrates.
By using multiple sites embedded in disordered regions to bind substrates, San1 is unusual for PQC-involved ubiquitin ligases. For example, the ubiquitin ligases CHIP, Hrd1, Doa10, PML-IV and UHRF-2 do not possess the extent of disorder seen in San1 (our unpublished observations). Intriguingly, San1's disordered binding regions do bear remarkable similarities to the disordered N-terminal regions of sHSPs. Notably, the disordered N-terminal domain of the Pisum sativum
(garden pea) sHsp 18.1 contains at least six residues involved in substrate interactions, as demonstrated by p
-benzoyl-L-phenylalanine (Bpa) crosslinking.36
These residues are dispersed throughout the disordered domain, indicating the presence of multiple substrate-binding sites in the PsHsp18.1 N-terminal region.