Proteomic analysis of aging and the heat shock response in C. elegans
We set out to examine the relationship between proteostasis and aging in C. elegans using quantitative proteomics. In addition to profiling the aging proteome by investigating three age groups, we examined the heat shock response during aging, since the heat shock pathway is not only important for aging, but because heat shock is also a well-established experimental paradigm in organisms ranging in complexity from yeast to mammals [36, 37]. Alongside a global proteomic analysis, we were particularly interested in the pool of newly synthesized proteins, and how this de novo proteome changes with age and under conditions of stress.
To profile the aging proteome and enable a comparison between early adulthood and aged animals, three representative time points were selected: day 1 of adulthood (young adult/YA), day 5 (D5) and day 10 (D10). For each time point, two populations of nematodes were examined; one grown at 20 °C and another subjected to heat shock for 2 h at 34 °C followed by 4-h recovery at 20 °C (Fig. 1).
De novo synthesized proteins can be visualized by bio-orthogonal labeling in vivo
To enable the identification of de novo synthesized proteins, we established protocols for bio-orthogonal tagging of proteins in C. elegans using the methionine homologue AHA. Metabolic labeling of proteins with AHA has previously been reported in cultured mammalian cells [25] and in larval zebrafish [38], but not C. elegans. We therefore first delineated suitable conditions for AHA incorporation into proteins in adult nematodes, growing them for 6 h at a range of concentrations (0–2 mM). After fixation and permeabilization, fluorescent Chromeo-546-alkyne was applied and reacted with AHA using ‘click chemistry’, and subsequently detected by confocal microscopy. Compared with untreated controls, where the alkyne tag was added, AHA was not added, and no ‘click chemistry’ was performed, increased fluorescence was detected in all tissues of nematodes grown in the presence of AHA, indicating successful incorporation (Fig. 2a). Thus, such an approach enables the visualization of de novo protein synthesis at a cellular level in response to any insult or stimulus. Toxicity as determined by impaired thrashing was only found at higher AHA concentrations and for prolonged incubation times (Online resource 1a and data not shown). To specifically assess whether AHA exposure might initiate the mitochondrial UPR, a hsp-6::gfp reporter [21] was monitored. No change in expression of this reporter following a 6-h incubation in 2 mM AHA was observed (Online resource 1b).
We next tested whether AHA can be incorporated at the relevant life stages and under conditions of heat stress. We cultured YA, D5 and D10 nematodes with and without 0.5 mM AHA and subjected them either to control conditions or heat shock. Following reaction with Chromeo-546-alkyne, increased fluorescence was detected in all samples grown with AHA (Fig. 2b, c).
As a complementary approach to visualize AHA incorporation, Western blotting was used. We established suitable conditions by obtaining protein lysates from nematodes that had been grown for 6 h in 0–8 mM AHA. ‘Click chemistry’ was used to react the incorporated AHA with biotin that was, in contrast with the BONCAT protocol [25], directly coupled to an alkyne-reactive group. The samples were analyzed by Western blotting, using an anti-biotin antibody to visualize AHA-tagged proteins. This established that AHA incorporation is detectable at all tested concentrations (data not shown). We next examined incorporation into YA, D5 and D10 nematodes under both control and heat shock conditions. For all conditions tested, AHA was incorporated into proteins over a wide range of molecular weights (Fig. 2d). A 70-kDa band was observed in all samples even in the absence of AHA labeling and ‘click chemistry’, indicating cross-reactivity with an unlabeled nematode protein. While incubation time- and AHA concentration-dependent changes in relative AHA incorporation can be visualized using Western blotting (data not shown), this method is not sufficiently sensitive to detect age-dependent or heat shock-dependent changes in AHA incorporation (Fig. 2d and data not shown). We therefore proceeded to identifying de novo synthesized proteins by direct detection of the incorporated AHA using proteomics.
iTRAQ analysis identifies 3,387 unique proteins
To prepare samples for quantitative proteomic analysis, nematodes were cultured in bulk until the relevant time points, and then split into ‘control’ and ‘heat shock’ samples. Two mM of AHA was added before heat shock commenced. Following a 2-h heat shock and 4-h recovery (6 h at 20 °C for the control), nematode samples were harvested. Fifteen samples including a minimum of two biological replicates for each of the six tested conditions were analyzed in four iTRAQ 4-plex experiments. To enable comparisons, one sample (‘young adult, heat shocked’, YAHS) served as an internal reference and was included in all four mixtures. iTRAQ-labeled samples were fractionated and subjected to nanoLC-ESI–MS/MS mass spectrometry. The data were analyzed using ProteinPilot V4.2 software with reference to the UniProt database collection from both C. elegans and E. coli.
To determine the global proteome, AHA modification was not included in the database search parameters. A total of 3,387 unique proteins from 20,182 unique peptides were identified, with an estimated protein identification false discovery rate of 0.15 %. Less than 0.2 % of all identified proteins were derived from the nematodes’ bacterial food source.
For the nematode proteins, the relative abundance for each condition was determined as the geometric mean of the iTRAQ ratios from replicate samples and the p value of this combined ratio was calculated according to Stouffer’s z test method. Proteins for which the combined ratio was <0.83 or >1.2, with p values <0.05, were considered to be either significantly less or more abundant. The changes observed were in the order of up to 25-fold. The thresholds were set with reference to the literature suggesting that iTRAQ in fact underestimates fold changes such that, for example, an iTRAQ ratio of 1.2 may reflect an actual twofold difference [39].
The global proteome of aged animals is characterized by a relative abundance of vitellogenins and diminished levels of ribosomal, mitochondrial, and myosin-related proteins
We first considered the aging proteome. When comparing D5 aged adults with YA, 221 proteins were increased in abundance (Online Resource 2a). When we used Wormbase to classify the Gene Ontologies of these 221 increased proteins at the level of subcellular structures and macromolecular complexes (i.e., cellular components) we found nuclear and extracellular as the two most represented categories, with 41 and 26 proteins, respectively (Fig. 3a; Table 1). Among the identified nuclear proteins were replication licensing factors (MCM-2–7), DNA topoisomerase TOP-2, nuclear lamin LMN-1 and several histone proteins including HIS-1, -4, -35, and -71, and the histone H1 variant, HIS-24. The latter is involved in the regulation of immune-related genes [40]. Increased were also several extracellular proteins that belong to the Transthyretin-Related family (TTR-2, -15, -16, -45, and -51). A major increased group of proteins within the extracellular category are the yolk proteins, vitellogenins, which are required for oocyte development. Importantly, levels of all six vitellogenins (VIT-1–6) were elevated in the aged animals.
Table 1 Proteins increased in abundance in day 5 animals compared with young adults
We next considered the group of proteins that is decreased in abundance with age. A total of 327 proteins were reduced at D5 compared with YA (Online Resource 2b). Among these decreased proteins, the two most represented cellular components are mitochondrial and ribosomal, with 79 and 56 proteins, respectively (Fig. 3b; Table 2). The decreased mitochondrial proteins include 20 components of the ETC and 14 enzymes of the tricarboxylic acid (TCA) cycle. Also decreased were several factors that play roles in mitochondrial proteostasis. These include Hsp70 (HSP-6), the co-chaperone grpE (C34C12.8), Tim44 (T09B4.9), Hsp60 (HSP-60), Hsp90 (R151.7a), clpX (D2030.2), two mitochondrial translation factors (GFM-1 and TSFM-1), a mitochondrial ribosomal protein (MRPL-12) and structural proteins (ATAD-3 and the prohibitins PHB-1 and PHB-2) (reviewed in [24]).
Table 2 Proteins decreased in abundance in day 5 animals compared with young adults
The decreased ribosomal proteins at D5 compared with YA include 22 components of the small (40S) ribosomal subunit and 28 components of the large (60S) ribosomal subunit. This striking decrease in abundance of numerous ribosomal subunits in aged animals is consistent with the observation of a decrease in protein synthesis during aging [31]. Other decreased proteins that contribute to the regulation of protein synthesis include RACK-1, a scaffolding component of the 40S ribosomal subunit (reviewed in [41]) and C08H9.2, the nematode ortholog of vigilin, which associates with 80S ribosomes and is proposed to regulate the translocation of tRNAs from the nucleus to the cytoplasm for association with ribosomes [42]. Among the decreased proteins were also a translation initiation factor (eIF5A homologue, IFF-2), a translation elongation factor (EF-2 homologue, EEF-2) and a polyA binding protein (PAB-1).
Outside these two major categories of decreased proteins, 12 myosin-related proteins were also decreased in D5 aged animals, including myosin heavy chain isoforms (MYO-1, -2, -3, UNC-54), myosin light chain isoforms (MLC-1, MLC-3), troponin T (MUP-2), tropomyosin (LEV-11), and paramyosin (UNC-15).
Since 42 % (94/221) of proteins that were increased and 32 % (104/327) of proteins that were decreased in D5 aged animals did not have a Gene Ontology (cellular component) term listed in WormBase (identified as “Unclassified” in Fig. 3a, b), we extended our analysis by identifying the human orthologs of these proteins and classifying the proteins based on the cellular component Gene Ontologies of these orthologs (Online Resource 2a and 2b). Similar to the primary analysis described above, the most represented categories among the increased proteins were nuclear, cytoplasmic, extracellular and membrane, while cytoplasmic and mitochondrial were most represented among the decreased proteins (Online Resource 2c).
How does the proteome change when we analyze even older worms? We next compared D10 adults with YA. Here, relative protein abundance was calculated indirectly since these conditions were not assayed in the same iTRAQ 4-plex experiment. That is, the D10 versus YAHS and YA versus YAHS ratios were calculated directly and then the former were divided by the latter to yield the D10 versus YA ratio. A p value was calculated using a Student’s t test and p < 0.05 was considered significant. Using this method, 33 proteins were identified as increased in abundance at D10 compared with YA, while 36 proteins were decreased in abundance (Online Resource 3a–3c). Although the number of proteins showing changes in abundance at D10 is much smaller than those identified at D5, this most likely reflects the more stringent statistics applied to these indirectly calculated data, rather than suggesting that the proteome of D10 adults is more similar to YA than that of D5 adults is. Importantly, Wormbase analysis of the Gene Ontologies of the proteins that were increased and decreased at D10 compared with YA revealed enrichment of the same cellular components as had been observed in the D5 proteome. That is, nuclear and extracellular proteins are most prominent among those increased at D10, while mitochondrial and ribosomal proteins are most prominent among those decreased at D10 (Online Resource 4a and 4b).
The majority of proteins identified at D10 were also identified at D5 (28/33 and 29/36) (Fig. 3c, d). Interestingly, among those proteins in the overlapping dataset, 25 of the increased proteins showed higher abundance at D10 compared with D5, while 20 of the decreased proteins showed lower abundance at D10 compared with D5, supporting the notion that these proteins are regulated with age (Fig. 3e, f).
Given that the indirect comparison of protein abundance described above found relatively few proteins to be changed at D10 compared with YA, we complemented these analyses of the aging proteome by also comparing the proteome of YA and D10 animals following heat shock. To identify only those changes that characterize the aging proteome, rather than the response to heat shock, we excluded from this analysis any protein that changed in abundance in response to heat stress at either of the examined life stages (67 proteins in total). This revealed 381 proteins as increased in abundance at D10 compared with YA and 474 proteins as decreased in abundance (Online Resource 3d–3f). Wormbase analysis of the Gene Ontologies of the increased proteins identified nuclear and extracellular as the two most represented cellular component categories, with 48 and 34 proteins, respectively. Among the decreased proteins, mitochondrial and ribosomal were the two most represented categories, with 92 and 73 proteins, respectively (Online Resource 4c and 4d).
Furthermore, these proteomic changes identified in D10 animals overlap substantially with those described above in D5 animals. That is, 160 proteins are increased at both D5 and D10 compared with YA and 247 proteins are decreased at both D5 and D10 compared with YA. Within the set of proteins increased at both D5 and D10 are the replication licensing factors, topoisomerase, nuclear lamin, histone proteins, Transthyretin-Related family proteins and vitellogenins that were described above. Similarly, among the proteins decreased at both D5 and D10 are the myosin-related proteins and proteins involved in mitochondrial proteostasis and ribosomal protein synthesis. In these latter categories, additional proteins were identified as decreased in abundance at D10, including three mitochondrial ribosomal proteins (MRPS-9, MRPS-22, and MRPS-26), translation initiation factors (EIF-1.A, EIF-3.H, IFFB-1, IFG-1, INF-1) and a translation elongation factor (EEF-1G).
Heat shock proteins and intermediate filaments increase in abundance following heat shock while P granule-associated proteins decrease
We next determined proteomic changes in response to heat shock. At YA stage, 40 proteins were increased and 36 decreased in heat shocked nematodes compared with controls (Online Resource 5a and 5b). Analysis of Gene Ontologies revealed an association of the increased proteins with the following cellular components: cytoplasmic, endoplasmic reticulum, Golgi apparatus, intermediate filament and extracellular (Fig. 4a). The intermediate filament proteins, consisting of IFA-1, MUA-6, and IFC-2, are of particular interest since such proteins are critical in the formation of aggresomes that form in response to protein misfolding [43]. Decreased upon heat shock were proteins associated with the cellular components: P granule, extracellular and mitochondrial (Fig. 4b). The P granule-associated proteins were PGL-1, CGH-1, CAR-1, and GLH-1, and the extracellular proteins included vitellogenins VIT-2 and VIT-6.
While the analysis of cellular component Gene Ontologies had successfully classified the majority of proteins identified as changed in abundance with age, 44 of the 76 proteins changed upon heat shock were not classified using this ontology. Among them were HSPs belonging to several families: hsp70 (HSP-4, HSP-70, F44E5.4), hsp110 (C30C11.4 [44]), hsp90 (DAF-21) and the small HSPs (HSP-16.1, HSP-16.41), which increased in abundance up to 25-fold upon heat shock. Interestingly, several stress-related proteins decreased, including catalases CTL-1 and CTL-2 and small heat shock protein SIP-1.
The heat shock response diminishes with age
We next compared the proteomic response to heat shock across the three age groups. In this case, since all comparisons were made directly between the proteomes of the heat shocked animals and non-heat shocked counterparts of the same age, and the same statistical thresholds were applied, the scale of the response at each time point can be directly compared. Interestingly, when compared with young adults, fewer proteins changed in abundance in response to heat shock in the aged animals (Fig. 4c, d). Specifically, while at YA 40 proteins were increased and 36 decreased, at D5 only 11 proteins were increased and 8 decreased (Online Resource 5c and 5d). Similar to the profile at D5, at D10, only 16 proteins increased and 6 decreased following heat shock (Online Resource 5e and 5f). Comparing the three time points, we found no proteins that uniformly decreased in abundance after heat shock in both young and old animals. We did, however, identify four HSPs that were increased in both young and old animals: two hsp70s (HSP-70 and F44E5.4) and two small HSPs (HSP-16.1 and HSP-16.41) (Fig. 4c). When their abundance at each time-point following heat shock is expressed relative to the age-matched control, there is a striking reduction in the scale of the increase of these proteins in response to heat shock in D10 animals and this reduction is most marked for the small HSPs (Fig. 4e). We additionally expressed the abundance of these proteins at each time-point relative to the YA control sample and found that all 4 proteins were increased in D5 and D10 control samples (Fig. 4f). However, the diminished scale of induction in response to heat shock means that the abundance of small HSPs following heat shock in D10 animals is reduced to <50 % of the levels attained in young adulthood. Together, these observations suggest that the heat shock response diminishes with age.
AHA-containing peptides are identified and relative abundance quantified by iTRAQ
We next identified the subset of MS/MS-identified peptides that contain AHA by including ‘AHA modification’ in the ProteinPilot search parameters. This identified 323 AHA-modified peptides corresponding to 205 proteins. It is not surprising that this number is small relative to the number of peptides identified in our global proteomic analysis, since peptides synthesized de novo during the 6-h AHA incubation would represent only a small portion of the total protein pool. Furthermore, the low charging rate and relatively low abundance of methionine mean that not all de novo peptides will be AHA-tagged. The relative abundance of the AHA-tagged peptides was computed as the geometric mean of the iTRAQ ratios from all measurements of a given peptide across the replicate experiments, and <0.83 or >1.2 were fixed as thresholds for decreased and increased abundance, respectively.
Analysis of AHA-tagged peptides reveals increased vitellogenin synthesis and decreased synthesis of distinct ribosomal, mitochondrial and myosin-related proteins in aged animals
We first considered those AHA-tagged peptides that were detected as increased in abundance in D5 aged animals. When we classified these 33 peptides using the cellular component ontology, the most represented components were extracellular and nuclear, mirroring the components that were enriched in our global analysis of proteins with increased abundance at D5 (Fig. 5a). We next examined all AHA-tagged peptides that were detected at D5 and/or D10 (Table 3 and Online Resource 6a) and compared these with the proteins identified in our global analysis. 42 AHA-containing peptides were increased at D5 and/or D10 compared with YA and of these, 31 peptides correspond to proteins that were also increased in abundance in aged animals in our global analysis. These include 23 peptides derived from vitellogenins as well as peptides corresponding to other extracellular proteins such as the transthyretin-like protein TTR-15 and to nuclear proteins such as the histone H4 HIS-1. For these proteins, our analysis of the AHA-tagged protein pool indicates that the observed increased abundance with age is not solely due to accumulation, but rather reflects a relative increase in synthesis.
Table 3 AHA-labeled peptides increased in abundance in day 5 and/or day 10 animals compared with young adults
We next considered those AHA-tagged peptides that were detected as decreased in abundance in D5 aged animals. When we classified these 38 peptides using the cellular component ontology, the most represented components were mitochondrial, myosin-related, ribosomal and cytoplasmic, mirroring the components that were enriched in our global analysis of proteins showing decreased abundance at D5 (Fig. 5b). We then examined all AHA-tagged peptides that were detected as decreased at D5 and/or D10 (Table 4 and Online Resource 6b) and compared these with the proteins identified in our global analysis. Fifty-three AHA-containing peptides were decreased at D5 and/or D10 compared with YA and of these, 42 peptides are derived from proteins that were also decreased in abundance in aged animals in our global analysis. These include 6 from ribosomal (including RPS-17, RPL-23, UBL-1), 15 from mitochondrial (including SDHA-1, MDH-1, FUM-1, and ATP-2) and 8 from myosin-related proteins (MYO-1, MYO-2, MLC-1/2/3, UNC-15, UNC-54). For these proteins, their relative decrease in abundance with age is not solely due to degradation but additionally reflects a relative decrease in their synthesis.
Table 4 AHA-labeled peptides decreased in abundance in day 5 and or day 10 animals compared with young adults
Analysis of AHA-tagged peptides affirms a decrease in the heat shock response with age
To complement our global analysis of the heat shock response, we secondly examined changes to AHA-labeled peptides in response to heat shock and identified 12 increased and 17 decreased peptides in YA animals (Fig. 6a, b and Online Resource 7a and 7b). Increased were AHA-labeled peptides corresponding to the categories mitochondrial (ZK836.2 and GAS-1) and myosin-related (UNC-54 and UNC-15), and HSPs [HSP-3 (hsp70) and C30C11.4 (hsp110)]. The nascent polypeptide-associated complex (NAC) alpha subunit, ICD-2 was also increased in response to heat shock. This is particularly interesting because this complex is a key regulator of proteostasis [31]. Not detected were other HSPs that were dramatically increased in response to heat shock through our global proteome analysis, possibly because of the timing of the addition of AHA and a lag before incorporation. Among the AHA-labeled proteins decreased after heat shock were vitellogenins (VIT-2, -5, -6), P granule components (PGL-1 and GLH-1) and cytoskeletal proteins (ACT-1/2/4, TBA-2 and TBB-2).
Examining the AHA-labeled peptides that changed in abundance in D5 and D10 heat shocked nematodes revealed fewer labeled peptides in aged animals, 25 at D5 and 14 at D10, compared with 29 at YA (Online Resource 7c-f). Interestingly, no AHA-labeled peptides corresponding to heat shock proteins were detected as increased in abundance relative to control at either D5 or D10. As indicated by our global proteomic analysis, these analyses of de novo synthesized proteins suggest that the response to heat shock is diminished in aged animals.