That Which Does Not Kill You Makes You Stronger: A Molecular Mechanism for Preconditioning

See allHide authors and affiliations

Science's STKE  05 Jul 2005:
Vol. 2005, Issue 291, pp. pe34
DOI: 10.1126/stke.2912005pe34


Preconditioning by sublethal stress can protect a cell from subsequent injury and apoptosis through a mechanism that has been unclear. Many such stresses stimulate the formation of stress granules: transient cytoplasmic foci that contain heat shock protein as well as translationally stalled mRNA and various mRNA-binding proteins. Recent research suggests that sequestration in stress granules of TRAF2, an adaptor protein that is required for tumor necrosis factor receptor 1 signaling, may underlie preconditioning by sublethal stresses.

Preconditioning by sublethal stress can protect a cell from subsequent injury and apoptosis. Although this phenomenon has been observed for a number of stressors and stress signals [for instance, tumor necrosis factor–α (TNF-α) (1), heat shock (2), and ischemia-reperfusion (3)], a mechanism for protection has not been identified. Several rational genetic approaches have been explored in unsuccessful attempts to identify the cellular processes involved in preconditioning. For example, researchers suggested that heat shock factor 1 (HSF1)–mediated transcription was a potential source of preconditioning (2); however, the hsf1 knockout mouse was subsequently shown to have an intact preconditioning response (4). The variety of stressors that elicit a preconditioning response provides a clue that preconditioning is a general cellular response to stress and is not necessarily transcriptionally mediated.

This physiological paradox has been the subject of intense research over the past two decades in the hope that clinically active, nontoxic agents could be developed to exploit the protective cellular effect. For example, prophylactic preconditioning might provide benefit before anticipated severe stresses such as military combat or major surgical procedures, such as cardiopulmonary bypass or organ transplantation. An important molecular clue to the mechanism of preconditioning was reported recently by Kim et al. (5), who found that the appearance and persistence of punctate cytoplasmic foci, so-called stress granules (SGs) (6), are required for heat- or arsenite-induced TNF desensitization. More important, Kim et al. demonstrated that TNF receptor–associated factor 2 (TRAF2), an obligate participant in TNF receptor 1 (TNFR1)–mediated signaling, localizes to SGs during the initial stress response.

TRAF2 and TNF-Stimulated Transcription

TNF is a proinflammatory cytokine produced by immune cells (macrophages and T cells) upon xenobiotic stimulation. Soluble TNF is recognized primarily by TNFR1, the constitutive TNF receptor, and signaling through this receptor leads to both c-Jun– and nuclear factor-κB (NF-κB)–mediated transcription [recently reviewed in (7, 8)]. Proteins derived from these nascent transcripts determine the cell’s ability to respond to the stress by either adapting or undergoing apoptosis. The TRAF family of proteins is required for signaling through members of the TNFR superfamily (9). Structural studies have suggested that TRAFs assemble into trimers using their coiled-coil domain and that different TRAF homo- and heterotrimers mediate cellular responses to signaling through specific TNFRs (10). TRAF2 is critical for NF-κB–mediated transcription after ligation of TNFR1 (Fig. 1).

Fig. 1.

Proposed mechanism for preconditioning. (A) TNF binds to TNFR1 and, without preconditioning, forms a competent signaling complex through recruitment of a number of adaptor proteins, including TRAF2, which binds to the SxxE motif of TNFR1. What follows is a multistep process resulting in IκB phosphorylation, release of NF-κB, translocation of NF-κB into the nucleus, and initiation of NF-κB-mediated transcription. PM, plasma membrane; IκB, inhibitor of nuclear factor κB. (B) Various stressors activate one or more of the eIF2α kinases (PERK, HRI, GCN2, and PKR). These kinases phosphorylate eIF2α Ser51, resulting in a stalled preinitiation complex. Unless acted on by protein phosphatase 1α (PP1α), the mRNA-binding protein TIA-1 recognizes the stalled complex and shuttles it to a cytoplasmic organizing center that becomes the SG. Within the SG, the eIF4GI SxxE motif becomes exposed and a putative direct interaction with TRAF2 occurs, effectively sequestering TRAF2 during the persistence of the SG. The small molecule salubrinal inhibits PP1α and leads to a preconditioning phenotype under basal conditions, presumably by increasing the intracellular concentration of stalled preinitiation complexes and initiating SG assembly. Subsequent challenge of preconditioned cells by TNF does not lead to NF-κB–mediated transcription. This is presumably due to the unavailability of TRAF2. AAA, polyadenylated mRNA.


SGs were originally reported over 20 years ago as transient microscopic foci found within heat-stressed Peruvian tomato cells (6). (The term "stress granule" is something of a misnomer. Many researchers prefer the term "punctate cytoplasmic foci" to distinguish these macromolecular assemblies from granules, because there is no evidence that SGs are membrane-encapsulated. For historical reasons, the term "stress granule" is commonly used and will be used here throughout.) Similar structures were identified in heat-stressed mammalian cells shortly thereafter and were reported to contain small inducible heat shock proteins (HSPs) and mRNA (11). In addition to HSPs and mRNA, recent studies have found that SGs contain a subset of proteins from the translation preinitiation complex (12) and other RNA-binding proteins [for instance, T cell internal antigen–1 (TIA-1) and TIA-1–related protein (TIAR)] (Fig. 1). Cells expressing mutant TIA-1 unable to bind mRNA (TIA-1ΔRRM) were unable to form SGs, suggesting that the interaction between components of the preinitiation complex and TIA-1 is an obligate step in SG assembly (13). Using fluorescence recovery after photobleaching analysis, TIA-1 labeled with green fluorescent protein (TIA-1–GFP) was found to ferry in and out of SGs with a half-time of 2 s. These results suggest that SG assembly occurs by a TIA-1–dependent mRNA-shuttling mechanism (14), with the implication that SGs are not junk heaps of precipitated protein and mRNA waiting to be degraded, but potential way stations for cellular mediators.

The eukaryotic initiation factor 2 (eIF2)–guanosine triphosphate (GTP)–initiator tRNA (tRNAiMet) ternary complex (the molecular assembly that joins with the preinitiation complex and aligns tRNAiMet onto the mRNA start site) plays an active role in SG assembly as well. Phosphorylation of the alpha subunit of eIF2 (eIF2α) Ser51, a modification that abrogates protein synthesis (15), stimulates SG formation. Cells expressing a nonphosphorylatable eIF2α mutant (S51A) do not decrease protein synthesis in response to arsenite (13) (a preconditioning agent that promotes eIF2α phosphorylation, see below). Further, cells expressing an eIF2α mutant that mimics constitutive phosphorylation (S51D) appear to have SGs under basal (nonstressed) conditions (16). One consequence of eIF2α phosphorylation is a 150-fold increase in the affinity of eIF2α for eIF2B, the eIF2α guanine nucleotide exchange factor (17), and thus inhibition of eIF2B function. Inhibition of guanosine diphosphate (GDP) exchange for GTP does not allow cycling of the ternary complex and results in the accumulation of eIF2-GDP, effectively halting translation. The SG phenotype also appears in response to pharmacologic depletion of cellular energy stores (ATP and GTP), corroborating the role of the ternary complex in SG formation.

Stress-induced depletion of the ternary complex, either by energy starvation or eIF2α phosphorylation, appears to be upstream of TIA-1 recruitment to the preinitiation complex (13). Presumably, TIA-1 is a poor competitor for cytoplasmic mRNA as compared with the ternary complex; however, when the cytoplasmic ternary complex is depleted, the preinitiation complex stalls, allowing TIA-1 to bind and initiate SG formation (Fig. 1).

There are four known eIF2α kinases: PERK (PKR-like endoplasmic reticulum kinase), HRI (heme-regulated inhibitor), GCN2 (general control nonrepressed 2), and PKR (double-stranded RNA-dependent protein kinase). Each of these kinases is activated by a different type of cellular stress, but the common result is the reprioritization of protein synthesis. Most mRNAs are translationally silent during stress, although some transcripts (HSPs) contain sequence elements in their flanking untranslated regions that promote translation under stress (18, 19). The stimuli that activate these four kinases are endoplasmic reticulum (ER) stress (20) and ischemia-reperfusion (21) (PERK), arsenite (HRI) (15), starvation (GCN2) (22), and double-stranded (ds) RNA [PKR (also referred to as DAI, for dsRNA-activated inhibitor)] (23). No eIF2α kinase is responsible for preconditioning by heat shock; instead, HSP27-mediated sequestration of eIF4G into punctate cytoplasmic foci appears to account for the heat shock response (24). Although PKR is a dsRNA-responsive kinase, protein synthesis does not stop and SGs do not appear after viral challenge in vitro. This anomaly is thought to be due to acquired viral evasion of SG formation, and viral products have been identified that interfere with SG assembly {namely, inhibition of eIF2α kinases or stimulation of eIF2α phosphatase activity [(25) and references therein]}. None of the eIF2α kinases are specifically activated during bacterial or fungal infections.

Tying SGs to Preconditioning

Using yeast two-hybrid analysis, Kim et al. identified TRAF2 as a likely binding partner for eIF4GI, a subunit of the preinitiation complex (5). They identified a requisite peptide within eIF4GI that contains the putative TRAF2-binding domain SxxE, a motif that occurs in other TRAF2-binding proteins, including CD30 and TNFR1. Kim et al. validated the TRAF2-eIF4GI interaction in a pulldown assay with glutathione S-transferase–labeled eIF4GI and confirmed TRAF2 specificity by showing that, out of a panel of FLAG-tagged TRAFs (TRAF1 through TRAF6), only TRAF2 bound eIF4GI. Colocalization of TRAF2 with transfected full-length eIF4GI indicated that TRAF2-eIF4GI binding is weak but occurs even at 37°C. The functional significance of TRAF2 sequestration after arsenite stress was demonstrated, because TRAF2 no longer associated with TNFR1, resulting in decreased NF-κB DNA binding by electrophoretic mobility shift assay. These results suggest a physical basis for the preconditioning response; namely, that stress-induced aggregation of eIF4GI sequesters TRAF2, thereby diminishing its ability to participate in NF-κB activation. Sequestration of TRAF2 in a stress-induced cytoplasmic compartment eliminates the possibility of dysfunctional TNF signaling during certain stresses. The kinetics of the recovery of signaling fidelity and hence the ability to survive a subsequent lethal threat may be a secondary benefit of immediate protection.

Taken together with studies of translation regulation in response to numerous stressors, and the observation that eIF2α phosphorylation stimulates SG formation, it is tempting to speculate that the mechanism described by Kim et al. accounts for preconditioning from a number of stimuli (arsenite, heat shock, ER stress, and ischemia-reperfusion). Moreover, the data suggest that other stimuli (stimuli leading to eIF2α phosphorylation, eIF4G sequestration, or energy starvation) may result in cells developing a preconditioned phenotype.

Future Prospects

A number of new questions are raised by this report. For example, TRAF2 associates with caveolin-1, a protein that is localized in specialized membrane domains (caveolae) that perform a number of signaling functions (26), in the absence of TNFR engagement, but some cell lines (like the 293 cells used in Kim's experiments) have very low levels of endogenous caveolin-1 (27). Where then is TRAF2 in their system before either stress or TNF ligation? Is there a trafficking protein that shepherds TRAF2 to eIF4GI upon SG assembly? Are SGs linked to the cytoskeleton? In addition, Kim et al. show that not all TRAF2 goes into the insoluble compartment. They speculate, but do not show, that this soluble material is nuclear TRAF2 associated with TRAF4. More generally, their data beg the question, how do other stresses, such as H2O2, produce a preconditioned phenotype without inducing SGs? And how are we to interpret the finding that adenoviral delivery of HSP70 alone provides protection against sepsis-induced organ dysfunction (28)? Are the punctate cytoplasmic foci formed after HSP27 expression functionally equivalent to SGs formed after eIF2α phosphorylation?

The identification of a molecular mechanism for the preconditioning response provides not only a basis for resolving a long-standing paradox in cell physiology but also a target for drug development. Boyce et al. recently identified salubrinal, a small molecule that selectively inhibits eIF2α dephosphorylation (29). Salubrinal protects cells from ER stress, enhances the accumulation of pSer51 eIF2α, and likely stimulates SG formation. Randomized controlled trials of other pharmacologic preconditioning agents have shown promise for the treatment of advanced cancer (30) and ischemic heart disease (31). Whether these agents stimulate eIF2α phosphorylation and SG accumulation is as yet undetermined, and whether there are other tissue-specific or stress-specific processes that result in a preconditioned phenotype remains to be seen. Although SGs have not been found in response to infectious stress, the application of a pharmacologic preconditioning agent may provide the same protection to the whole organism.

Kim et al. provide insight into the mechanism by which the inflammatory and stress responses interact. Not only do their data provide a molecular basis for resolving a long-standing physiologic paradox, but they also provide a paradigm for screening pharmacologic agents. The generality of the mechanism will determine the extent to which pharmacologic agents can take advantage of the protective effects of preconditioning. With the recent results from Boyce et al., there appears to be good reason to be optimistic.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
View Abstract

Navigate This Article