To identify proteins that are induced by DR, we analysed extracts from BY4741 yeast cells grown under standard (2% glucose) or DR (0.5% glucose) conditions by 2-D gel electrophoresis. Initial experiments using wide range (pH 3–10) gels revealed no obvious reproducible changes in protein spot abundance, indicating that DR does not cause gross proteomic alterations (). However, using pH 5.3–6.5 and pH 3–5.6 zoom gels, we could resolve several proteins that were barely detectable in control conditions, yet clearly induced in DR conditions (). Mass spectrometry was used to identify these differentially expressed proteins as Eno1, Hxk1, Hsp12, Rtc3, Rgi1, Sbp1 and Yef3 (Table S1
DR induces expression of a relatively small number of proteins.
To help pinpoint DR-induced proteins that play a causal role in mediating lifespan extension, as opposed to those whose expression patterns are merely coincidental, we then analysed proteins that were induced by high osmolarity (Fig. S1
). Our rationale was based on the observation that lifespan extension by DR and high osmolarity act via a common downstream pathway 
; and hence the crucial effector proteins are likely to be common to both interventions. Proteins induced by high osmolarity comprised Ctt1, Eno1, Fba1, Hsp12, Hsp26, Hsp31, Lys9, Rtc3, Rgi1 and Oye2 (Table S1
). To validate these protein expression changes, we prepared extracts from yeast containing chromosomally-tagged GFP fusion constructs 
and performed western blots using a GFP antibody. Examples of proteins selectively induced by DR (GFP-Hxk1) or high osmolarity (GFP-Ctt1) are shown in Fig. S1B
. Of the proteins identified as being induced by both interventions, specific bands of the predicted size could not be reproducibly detected for GFP-Rtc3 or –Rgi1; whereas GFP-Eno1 expression was not altered by DR. However, GFP-Hsp12 was confirmed to be induced by both DR and high osmolarity (Fig. S1B
). To rule out any artefactual effect of the GFP tag, we raised an antiserum against the N-terminus of Hsp12 and used this in western blots of wild type cells. This revealed a band of the expected size (~12 kDa), which was increased in intensity upon growth in DR and high osmolarity conditions and which was not present in an isogenic hsp12
deletion strain (Fig. S1C
), thus confirming the increased expression of Hsp12 under conditions of enhanced longevity.
To determine if Hsp12 is causally linked to DR-induced lifespan extension, we performed replicative lifespan analysis. This was done by determining the number of daughter cells removed by micromanipulation from individual virgin mother cells 
. Wild type BY4741 cells exhibited a mean lifespan of 21 (95% CI: 19–23) on standard (2% glucose) media, which was significantly increased to 31 (95% CI: 28–35) under DR (0.5% glucose) conditions (). Deletion of HSP12
did not reduce longevity under standard conditions, but rather resulted in a small increase in mean lifespan to 26 (95% CI: 23–29). Strikingly, however, the ability of DR to increase longevity was abolished in the hsp12
Δ strain, which exhibited a mean lifespan of 25 (95% CI: 22–28) under DR conditions (). These data suggest that the impact of Hsp12 on cellular ageing is complex: the low-level expression observed in standard media has a small negative effect on lifespan, whereas high Hsp12 levels induced by DR are essential for the increase in longevity caused by this intervention.
HSP12 is essential for lifespan extension by DR.
Stress resistance correlates positively with lifespan in various model organisms and DR may represent a mild stress that extends lifespan via a hormesis-like mechanism 
. We therefore investigated if deletion of HSP12
reduced resistance to environmental stresses. For comparison, we also included sir2
deletion strains in this analysis, as deletion of SIR2
is known to decrease and increase replicative lifespan respectively 21
. There was no detectable difference between the ability of BY4741 wild type and the deletion mutants to grow under a wide variety of stress conditions, including DR and other stresses that increase Hsp12 expression (Fig. S2
). We therefore conclude that Hsp12 does not contribute to general stress resistance.
Various small heat shock proteins have been shown to be ‘holdase’ molecular chaperones that bind to denaturing proteins and prevent their aggregation. To determine if Hsp12 had such activity, we investigated the ability of recombinant purified Hsp12 to prevent aggregation of the model substrate, insulin, using the method of Haslbeck et al.
. Addition of DTT reduces the disulphide bonds between the A and B chains of insulin, causing aggregation; whereas in the absence of DTT, insulin remains stable (Fig. S3A
). DTT-induced insulin aggregation was greatly reduced by recombinant GST-fusion proteins of the known chaperones, yeast Hsp26 
and mammalian cysteine string protein (CSP) 
; but not by CaBP1s, used as a control for a protein of similar size to Hsp12 with no known or predicted chaperone functions (Fig. S3B
). However, GST-Hsp12 was similar to GST-CaBP1s in terms of ability to prevent insulin aggregation. The differences in chaperone activity for GST-Hsp12 and GST-Hsp26 were then assessed in a dose-dependent manner. This revealed that GST-Hsp26 has approximately 100-fold higher anti-aggregation activity than GST-Hsp12 (Fig. S3C
), indicating that Hsp12 has very low, if any, intrinsic chaperone activity.
In addition to possessing anti-aggregation properties, small heat shock proteins are often large homo-oligomeric assemblies of folded subunits. To further investigate the possible function of Hsp12, we determined its solution structure using NMR. Recombinant Hsp12 expressed in E. coli
was monomeric. The 15
H HSQC spectrum showed poor resonance dispersion in the proton dimension, which suggested that Hsp12 is intrinsically disordered in aqueous buffer (). Recently published circular dichroism studies have shown that Hsp12 gains significant helical content upon binding to lipid or SDS micelles 
, we therefore examined the effect of varying SDS concentrations. The 15
H HSQC spectra of Hsp12 showed a dose-dependent increase in dispersion in response to SDS, indicating that Hsp12 adopts a folded conformation upon micelle binding (). Having determined the optimal SDS concentration for NMR, we then characterised the temperature-dependence of Hsp12 in the presence (Fig. S4A
) and absence (Fig. S4B
) of SDS. This resulted in linear resonance dispersion until 45°C, above which some resonances deviated from a straight line in the presence of SDS, indicating heat-induced unfolding. These optimised conditions were then used to assign the residues of SDS-bound 15
C-labelled Hsp12 ().
Hsp12 is unstructured in solution, but folds in the presence of SDS.
Analysis of the backbone dynamics of Hsp12 in the presence of SDS revealed relatively long T1
relaxation values compared to T2
( A,C,E), suggesting restricted mobility in the majority of the polypeptide. In contrast, T1
values were similar in the absence of SDS ( B,D,F), suggesting that the protein is highly dynamic in solution, but is structured on micelles. Consistent with this, analysis of the assigned chemical shifts in Hsp12 using CSI 
suggested that micelle binding induces the formation of four α-helices (). These α-helices cover the majority of the polypeptide and comprise residues F9-A16 (Helix I), Q22-A41 (Helix II), V52-G63 (Helix III) and L74-E94 (Helix IV). Helix III is not as stable as the other four helices, as revealed by the lower number of dαNi,i+3
connectivities for this helix and more variation in its length compared with the other three helices together with a high RMSD value of 0.465 (Table S2
). The experimentally-determined structural data correspond well with prediction using the AGADIR programme 
, which shows that the region between 52–63 has a lower helical propensity compared with the other three helical regions. Extensive analysis of residual dipolar couplings using stretched acrylamide gels revealed no evidence of long-range interactions between the individual helices, indicating that Hsp12 does not form a stably-folded structure.
Backbone dynamics and chemical shift-based secondary structure of Hsp12.
We generated a model of the tertiary structure of Hsp12 using CYANA. The ensemble presented ( and Fig. S5
) highlights the flexibility of the α-helices relative to one another. The four α-helices can be more clearly identified in the representative model in , with the 4th
and most C-terminal helix represented in yellow/red. Analysis of the charge distribution reveals each α-helix to be broadly amphipathic, with hydrophobic (green) residues lying on one face and charged (red) residues on the opposite face ( B,C). In addition, the residues flanking each α-helix also tend to be charged. This suggests that hydrophobic residues of Hsp12 insert into the lipidic component of membranes, while the charged (mainly positive) residues interact with negatively charged head groups and project away from the membrane. A Ramachandran plot of the data is presented in Fig. S6
. Overall, the NMR data indicate that Hsp12 is intrinsically unstructured in aqueous solution, but switches to a dynamic 4-helical conformation upon membrane binding.
Ensemble of structures calculated for micelle-bound Hsp12 overlaid on each of the four helices.
Helical properties of micelle-bound Hsp12.