A temperature-dependent conformational shift in p38α MAPK substrate–binding region associated with changes in substrate phosphorylation profile

Febrile-range hyperthermia worsens and hypothermia mitigates lung injury, and temperature dependence of lung injury is blunted by inhibitors of p38 mitogen-activated protein kinase (MAPK). Of the two predominant p38 isoforms, p38α is proinflammatory and p38β is cytoprotective. Here, we analyzed the temperature dependence of p38 MAPK activation, substrate interaction, and tertiary structure. Incubating HeLa cells at 39.5 °C stimulated modest p38 activation, but did not alter tumor necrosis factor-α (TNFα)-induced p38 activation. In in vitro kinase assays containing activated p38α and MAPK-activated kinase-2 (MK2), MK2 phosphorylation was 14.5-fold greater at 39.5 °C than at 33 °C. By comparison, we observed only 3.1- and 1.9-fold differences for activating transcription factor-2 (ATF2) and signal transducer and activator of transcription-1α (STAT1α) and a 7.7-fold difference for p38β phosphorylation of MK2. The temperature dependence of p38α:substrate binding affinity, as measured by surface plasmon resonance, paralleled substrate phosphorylation. Hydrogen–deuterium exchange MS (HDX-MS) of p38α performed at 33, 37, and 39.5 °C indicated temperature-dependent conformational changes in an α helix near the common docking and glutamate:aspartate substrate-binding domains at the known binding site for MK2. In contrast, HDX-MS analysis of p38β did not detect significant temperature-dependent conformational changes in this region. We observed no conformational changes in the catalytic domain of either isoform and no corresponding temperature dependence in the C-terminal p38α-interacting region of MK2. Because MK2 participates in the pathogenesis of lung injury, the observed changes in the structure and function of proinflammatory p38α may contribute to the temperature dependence of acute lung injury.

Humans usually maintain body temperature within a narrow range, but regulated deviations occur as a result of normal circadian rhythm and in response to exogenous or endogenous pyrogens. Unregulated temperature deviations occur when normal thermoregulation fails or thermoregulatory effector mechanisms are overwhelmed by intentional or accidental environmental exposures. The biological effects of hypothermia have been harnessed to improve neurologic outcome after cardiac arrest (1,2) and more recently to reduce acute lung injury (3,4). Hyperthermia has proven to be effective as adjuvant treatment against some forms of cancer (5), but it may worsen organ injury, including acute lung injury (6,7). Changes in temperature within this clinically relevant hypo-to-hyperthermia range exert critical biological effects that impact cell survival, endothelial barrier function, coagulation, leukocyte trafficking, and inflammation (8 -15). Within the lungs, temperature elevations accelerate pathogen clearance but also increase permeability pulmonary edema and neutrophil-mediated lung inflammation and injury (7). Within tumors, temperature elevations enhance recruitment of lymphocytes with antitumor activity (14). Despite the significant biological impact of hypo-and hyperthermia, gene expression analyses have shown that temperature shifts between 30 and 41°C modify Յ4% of expressed genes (16 -20) and Յ1% of miRNAs (21), suggesting that temperature shifts in this range affect only a narrow subset of signaling events.
The p38 MAPKs are classically activated through a canonical three-kinase module that dually phosphorylates p38 on the threonine and tyrosine in the TXY motif common to all MAPKs, but under certain conditions, p38 activation can occur through autophosphorylation of the threonine in the TXY motif (46,47). Once activated, p38 translocates to both the nucleus and the cytoskeleton and can phosphorylate a broad range of substrates that have both pro-and anti-inflammatory actions (48). Substrate and function specificity of p38 kinase activity is achieved through selective binding of substrates to specific motifs on p38, including a substrate-docking groove located between the C-and N-terminal lobes and opposite the catalytic site (49). The nature of substrate-specific interactions with the p38 substrate-docking domains likely determines the phosphorylation efficiency for each substrate and the downstream biological effects.
The purpose of this study was to determine how shifts in temperature within a clinically relevant range modify p38 signaling. We show that changing the reaction temperature between 33 and 39.5°C causes a conformational change within the substrate-binding groove of p38␣ but not p38␤ with an associated substrate-specific change in p38␣ substrate binding and kinase function. Binding affinity and phosphorylation rate of p38␣ for MK2 relative to ATF2 and STAT1␣ increased at 39.5°C and decreased at 33°C.

Temperature dependence of p38 activation
We previously showed that exposure to FRH (ϳ39.5°C) causes a modest increase in levels of activated p38 MAPK in whole-lung tissue homogenates in mice in vivo (11) and human microvascular lung endothelial cells (HMVECs) in vitro (10,11). In this study, we found that incubating HeLa cells at 39.5°C stimulated an increase in phosphorylated p38 that peaked ϳ2-fold above baseline after 30 -60 min and returned to baseline by 2-4 h ( Fig. 1 (A and B) and Fig. S4). THP1 and BEAS2B cell lines and primary cultured HMVECs exhibited a similar transient activation of p38 in response to hyperthermia ( Fig. 1C and Fig. S4). Unlike p38, there was no detectable ERK activation in the hyperthermic HeLa cells (Fig. 1A and Fig. S4). In contrast with hyperthermia alone, stimulating HeLa cells with 10 ng/ml TNF␣ activated a rapid, transient ϳ25-fold increase in phosphorylated p38 that was similar at 33, 37, and 39.5°C ( Fig. 1D and Fig. S4).

p38␣ activation at febrile temperature does not occur via autophosphorylation
To determine the potential contribution of autophosphorylation to the p38 activation observed in hyperthermia-stimulated cells, we first determined whether p38 catalytic activity was required for p38 activation in HeLa cells incubated at 39.5°C by pretreating with the p38 catalytic inhibitor, SB203580 ( Fig. 1E and Fig. S4). Pretreating HeLa cells with 10 M SB203580 caused only a slight decrease in p38 activation after 30-min exposure to 39.5°C that was similar in magnitude to its effect on basal and TNF␣-stimulated p38 activation. To further assess the potential for p38 to activate via autophosphorylation, we measured phosphorylation of recombinant p38␣ and p38␤ incubated in cell-free in vitro kinase reactions at various temperatures between 33 and 41°C in the absence of other kinases ( Fig. 1F and Fig. S4). Neither p38␣ nor p38␤ exhibited temperature-dependent phosphorylation during a 30-min incubation at temperatures between 33 and 41°C.

Temperature dependence of p38:substrate interactions
We next analyzed the temperature dependence of p38␣ and p38␤ kinase activity for specific substrates in in vitro kinase assays performed at 33, 37, or 39.5°C (Fig. 2 (A-C) and Fig. S5). The initial phosphorylation rate of MK2 by p38␣, measured over the first 10 min of the reaction, was 14.5-fold greater at 39.5°C than at 33°C compared with a 3.1-fold temperaturedependent difference for ATF2 and 1.9-fold difference for STAT1␣. For p38␤, the phosphorylation rate of MK2, ATF2, and STAT1␣ was 7.7-, 2.4-, and 2.1-fold greater at 39.5°C than at 33°C. For p38␣, the phosphorylation rate of MK2 relative to ATF2 was 4.7-fold greater at 39.5°C than at 33°C. For p38␤, the ratio of MK2/ATF2 phosphorylation rate was 3.2 greater at 39.5°C than at 33°C.
We used SPR to quantify substrate binding affinity for immobilized, unphosphorylated recombinant p38␣ and p38␤ at 33, 37, and 39.5°C (Fig. 3 (A-D) and Fig. S1 (A and B)). Based on the dissociation constant (K D ) values derived from the SPRs, binding affinity was approximately 3 orders of magnitude higher for ATF2 than MK2 for both p38␣ and p38␤, but the temperature dependence of p38 binding was very different for the two substrates. The binding affinity of p38␣ to MK2 was 2.7-fold greater at 39.5°C than at 33°C, whereas its binding affinity for ATF2 was 33% lower at 39.5°C versus 33°C. For p38␤, binding affinity for MK2 was 2-fold greater at 39.5°C than at 33°C, whereas its binding affinity for ATF2 was similar at all three temperatures. The temperature-dependent changes in p38 binding to and phosphorylation of MK2 occurred well below the melting temperatures for p38␣ and p38␤, which were 47 and 48.3°C, respectively, compared with 53.1°C for ERK2 (Fig. 3F).

Analysis of temperature-dependent p38 conformational changes using HDX-MS
To understand how temperature-dependent conformational changes in p38 might explain the substrate-specific effect of temperature on kinase function, we performed HDX-MS for unphosphorylated p38␣ and p38␤ at 33, 37, and 39.5°C. The analysis yielded 149 peptides with 97.8% sequence coverage for p38␣ and 85 peptides with 96.7% sequence coverage for p38␤ (Fig. S2). The HDX-MS analysis of p38␣ structure at 33°C was in overall agreement with previous studies (50,51), recapitulating the exchange-protected core (␣E and ␣F for the C-terminal lobe, ␤3-␣C-␤4 for the N-terminal lobe) and the high deute- Figure 1. Hyperthermia-induced p38 activation. A, HeLa cells were incubated at 37 or 39.5°C for the indicated time and then immunoblotted for phosphorylated (p) and total (t) p38 and ERK. Results representative of four p38 and two ERK immunoblots are shown. B, mean Ϯ S.D. (error bars) of band densities from four p38 blots; p value for difference between 37°C and 39.5°C by MANOVA is indicated. C, the same analysis was performed in THP1, BEAS2B, and HMVECs, and one of two sets of blots is shown. D, TNF␣ (10 ng/ml)-stimulated p38 activation was analyzed in HeLa cells at 33, 37, and 39.5°C. A representative blot is shown. E, HeLa cells were pretreated with or without 10 M SB203580 at 37°C for 30 min and then incubated for 10 min with 10 ng/ml TNF␣ at 37°C or at 39.5°C without TNF␣ for 30 min, and p38 activation was analyzed by immunoblotting. F, IVK with unphosphorylated recombinant human p38␣ or p38␤ was incubated at the indicated temperature for 1, 3, or 10 min, separated, and immunoblotted using an antibody that recognized both monophosphorylated and dually phosphorylated p38␣ and p38␤. A control (C) without ATP is included. One of four representative blots is shown for each.

Temperature-dependent p38␣ MAPK structure and function
rium uptake of loops and peripheral regions (glycine-rich loop or the MAPK insert) (Fig. S2). Moreover, the qualitative agreement extends to catalytically important motifs, such as the activation lip, which appears largely labile and readily exchangeable; the DFG motif; and the substrate Pϩ1 binding pocket within the active site. HDX-MS analysis at 33°C demonstrated regions displaying a bimodal isotopic envelope typical of EX1 kinetics that is indicative of large, slow, and cooperative conformational changes. Bimodal isotopic envelopes such as those that are clearly seen in Fig. 4A for p38␣ at 39.5°C arise when a contiguous stretch of residues unfold in a concerted manner and remain unfolded long enough for all of the exposed amide hydrogens to exchange nearly completely before refolding (52,53). This feature, which was not reported in a previous HDX-MS analysis of p38␣ performed at 10° (50,51), became more pronounced as the temperature increased from 33 to 39.5°C. Four peptide regions in p38␣ exhibited temperaturedependent EX1-type kinetics, residues 130 -145, 207-215, 281-288, and 300 -306. The most intense and fastest EX1-type kinetics were located around residues 130 -145 and involved ϳ7 amide hydrogens encompassing the C-terminal end of ␣E. Fig. 4A shows the stacked spectra with the bimodal isotopic envelopes for peptide 130 -145 (␣E) of p38␣ (left) and p38␤ (right). For p38␣, the exchange-prone species was not detected until 2 h of deuterium incubation at 33°C, when it accounted for ϳ33% of the overall isotopic envelope. At 37 and 39.5°C, the exchange-prone envelope was initially detected by 10 min, and by 2 h it had reached 64 and 83% of the total isotopic envelope, respectively. For p38␤, the EX1 kinetics were much slower and less temperature-sensitive; the exchange-prone envelope was significantly populated only after a 2-h incubation at 39.5°C and not at all at lower temperatures. The EX1 kinetics of ␣E are summarized in Fig. 4B, which shows decay of the exchangeprotected envelope as a function of deuterium incubation time for both p38␣ and p38␤ at each of the three temperatures. For p38␣, the half-life of exchange on ␣E was 85 and 60 min at 37 and 39.5°C, respectively, and Ͼ2 h at 33°C. For p38␤, the halflife was Ͼ2 h at all three temperatures. Mapping peptide 130 -145 (in red) to the known structure of p38␣ in Fig. 4C shows it to span ␣E within the substrate-binding groove between the CD and ED motifs. The C terminus of MK2 (yellow) is shown binding to the same region of the substrate-binding groove (PDB entry 2OKR) (54).
The conformational changes observed in peptide 130 -145 of p38␣ were accompanied by somewhat slower but similarly temperature-dependent EX1 kinetics of conformational unfolding in adjacent ␣-helical segments spanning residues 207-215 (␣F), 281-288 (␣H), and 300 -306 (␣I) of p38␣, each involving ϳ6 -8 amide hydrogens (Fig. S3). The half-life for deuteration at 39.5°C was ϳ60 min for residues 207-215, 281-  288, and 300 -306, respectively, and was Ͼ2 h at 33 and 37°C for all three regions. For regions 207-215 and 281-288 in p38␤, Ͻ20% of the protected species had decayed by 2 h of incubation, and for region 300 -306, deuterium exchange did not follow EX1 kinetics. Fig. 4D summarizes percent deuteration after 2 h for each of these four regions in p38␣ and the corresponding regions in p38␤ as a function of incubation temperature. In Fig.  4E, the largest difference in EX1 deuterium exchange rate between 33 and 39.5°C occurred in the four peptides of interest in p38␣, but not in the corresponding p38␤ peptides. Although other p38␣ peptides demonstrated temperature-dependent differences in deuterium exchange, the changes did not follow EX1 kinetics (indicated by black arrows). Fig. 4F demonstrates the extensive contacts formed between ␣E, ␣F, ␣H, and ␣I to form a structural motif near the substrate-binding groove.
In contrast with the temperature-dependent conformational changes in p38␣, HDX-MS analysis of MK2 failed to detect any temperature-dependent conformational changes in the C-terminal p38␣-interacting region of MK2 (Fig. 5).

Discussion
This study addressed the molecular mechanisms underlying the temperature dependence of endothelial permeability and acute lung injury. We focused on p38 signaling because it is known to participate in the regulation of endothelial barrier function (23,26,55), inflammation, and neutrophil recruitment (11,30,56) and because previous studies suggested that p38 signaling is suppressed by hypothermia (13) and enhanced by hyperthermia (10,11). We studied temperatures representative of currently used therapeutic hypothermia protocols (33°C) and those that typically occur in response to serious infections (39.5°C). In agreement with our past report (11), we found that exposing cell cultures to 39.5°C stimulated a modest and transient p38 activation in multiple cell types. Based on a previous report that a p38␣ mutant with increased activation loop flexibility underwent spontaneous autophosphorylation in Escherichia coli grown at 32°C but not at 21°C (57), we analyzed whether WT p38␣ might exhibit similar conformational changes that allow autophosphorylation at temperatures Ͼ37°C. Whereas the p38 catalytic inhibitor, SB203580, can inhibit other kinases besides p38 (58), its inability to block FRHstimulated p38 phosphorylation when present at a concentrations above the IC 50 for both p38␣ and p38␤ suggests that p38 catalytic activity is not required for its phosphorylation in these conditions. Furthermore, neither WT p38␣ nor p38␤ exhibited autophosphorylation in in vitro kinase reactions at temperatures up to 41°C even with p38 protein present at 5.3 M and the ATP concentration much greater than the reported K m for ATP (59). Finally, HDX-MS analysis failed to show temperature dependence of the p38␣ activation loop. Canonical p38 activation following stimulation of HeLa cells with TNF␣ was not significantly affected by shifting temperature between 33 and 39.5°C.
Having shown that clinically relevant shifts in temperature had only modest effects on p38 activation, we analyzed whether clinically relevant hypo-or hyperthermia might alter p38 kinase function. We focused this analysis on p38␣ and p38␤, the two most widely expressed isoforms, because SB203580, a p38 catalytic inhibitor with activity limited to the ␣ and ␤ isoforms of p38, blocks hyperthermia effects on endothelial permeability and lung injury (10,11). To avoid ambiguity caused by overlap and redundancies among multiple p38 isoforms and other MAPK signaling pathways (60), we analyzed the effect of temperature shifts on structure and function of purified recombinant proteins in cell-free reactions. We focused on two proinflammatory p38 substrates, MK2 and ATF2, which differ in cell location, function, and their molecular interaction with p38␣. MK2 is a cytoplasmic kinase that modifies posttranscriptional regulation of proinflammatory cytokine expression and regulates endothelial permeability and neutrophil extravasation by phosphorylating HSP27 and modifying the actin cytoskeleton (61). ATF2 is a transcription factor that activates expression of multiple genes that regulate inflammation, cell cycle progression, and cell death (62). The substrate-docking domains of p38␣ confer substrate selectivity by increasing the phosphorylation efficiency of bound substrates (49). Structural analysis of p38␣:MK2 complexes has shown that the C terminus of MK2 binds to the substrate-docking groove stretching between the ED and CD domains of the p38␣ (54). The structure of the p38␣:ATF2 complex has not yet been solved, but mutational analysis suggests that ATF2 binds to the DEF pocket of p38␣, which is distinct from the MK2-binding site (57).
In cell-free IVK reactions with recombinant dually phosphorylated p38␣, the maximal rate of MK2 phosphorylation was 14.5-fold greater at 39.5°C than at 33°C, which was 4.7-fold greater than the temperature dependence of ATF2 phosphorylation. In these reactions, the ATP concentration was below the ϳ25 M K m for ATP (59), and the initial reaction rates were measured during the first 10 min of the reaction. Even if substrate limitation caused slowing of substrate phosphorylation prior to the 10-min sample collection, the more rapid reactions would be most impacted, thereby underestimating the effect of temperature on MK2 phosphorylation by p38␣. For p38␤, MK2 phosphorylation was 7.7-fold greater at 39.5°C than at 33°C, which was 3.2-fold greater than the difference for ATF2 phosphorylation. The narrow physiologically and clinically relevant temperature range studied is below the denaturation temperature for p38 as detected by the differential scanning fluorimetry (DSF) assay and below the optimal temperature for p38␣ and p38␤ catalytic activity for all three substrates tested (63). The similar modest increase in phosphorylation rate for STAT1␣

Temperature-dependent p38␣ MAPK structure and function
and ATF2 observed over the 6.5°C experimental temperature range is consistent with previously described temperature dependence of enzyme activity and with the classical and equilibrium models (63) and the macromolecular rate theory (64) of temperature dependence. However, these models of temperature-dependent enzyme catalytic activity do not explain the disproportionately large temperature dependence for MK2 phosphorylation. Therefore, we reasoned that the excess effect of hyperthermia on MK2 phosphorylation might be due to enhanced MK2 binding to its substrate-specific docking domain on p38␣ (54). We used SPR to directly measure the temperature dependence of p38 substrate-binding affinity. Based on prior studies demonstrating that unphosphorylated and dually phosphorylated p38␣ have a similar binding affinity for MK2 (65), we immobilized unphosphorylated recombinant p38␣ or p38␤ to CM5 chips within the SPR flow cell and adjusted the temperature of the circulating buffer containing full-length MK2 or ATF2 to 33, 37, or 39.5°C. Affinity of p38␣:MK2 binding increased, whereas p38␣:ATF2 binding affinity decreased as the reaction temperature was increased from 33 to 39.5°C. The ratio of p38␣:MK2 to p38␣:ATF2 binding affinity was 4-fold greater at 39.5°C than at 33°C compared with a 2-fold difference for binding of the two substrates to p38␤. The temperature dependence of selective p38:substrate binding is consistent with the results of the in vitro kinase reactions and with the known contribution of the p38 substrate-docking domains to substrate phosphorylation efficiency and specificity (49). We note that a prior study using SPR, stopped-flow fluorescence, and isothermal calorimetry found K D values for p38:MK2 binding to range from 1 to 106 nM, depending on the method used and buffer salt concentration, compared with our observed K D values of 1.27 and 0.474 M at 33 and 39.5°C, respectively (65). The NaCl concentration used in our SPR assay, 150 mM, was intermediate between the concentrations used in the prior study, and we used the same full-length MK2 and p38␣ variants. Our recombinant p38␣ and MK2 proteins could be activated in vitro and phosphorylated MK2 and HSP27, respectively (data not shown), suggesting that they were properly folded. The major methodologic difference between the prior and present studies was the use of a subphysiologic (25°C) assay temperature in the prior study. To the best of our knowledge, our study is the first to measure p38:substrate binding affinity at physiologic temperatures.
Because the significant temperature-dependent changes in p38␣:MK2 interaction occurred at temperatures ϳ8°C below the melting temperature for p38␣, we searched for localized, temperature-dependent conformational changes in p38␣ consistent with its functional changes using HDX-MS. We performed HDX-MS reactions at physiologic pH and over the physiologic temperature range of interest, 33-39.5°C. We observed regions demonstrating EX1 kinetics, which we identified by the presence of a typical signature bimodal isotopic envelope with the progressive appearance of a distinct exchange-prone species concomitant with the progressive decay of the exchange-protected species. The slow kinetics and the cooperative nature (i.e. spanning multiple amide hydrogens) of the conformational interconversion has generally led to the phenomenon being ascribed to large changes in secondary (local unfolding), tertiary (local and global unfolding), or quaternary structure (subunit dissociation, protein/protein interfaces) (52,53).
The prediction that the chemical HDX rate is temperature-dependent (66) has been experimentally confirmed and shown to modify the EX2 exchange, in which the chemical HDX rate is limiting. Analytical methods for HDX-MS have been developed to distinguish between temperature-dependent conformational information and direct effects of temperature on the chemical HDX rate over large temperature ranges (67,68). In contrast with EX2 kinetics, for EX1 kinetics, the rate of conformational changes/fluctuations is ratelimiting rather than the temperature dependence of the chemical exchange rate, thereby preserving the characteristic behavior of EX1 kinetics if the rate of the underlying cooperative unfolding/conformational change is not heavily temperature-dependent.
The results of the HDX-MS analysis of p38 are consistent with a localized temperature-dependent cooperative unfolding of four helices in p38␣ that form a tertiary structure element located within a region of known direct contact between p38␣ and MK2 (54). Importantly, this analysis did not show conformational changes in any other p38␣ domain, including the DEF pocket, which is required for ATF2 phosphorylation (57), or the p38␣ catalytic site. In contrast with the temperature-dependent changes in localized unfolding of p38␣, there was no significant temperature-dependent difference in percentage of deuterium uptake in p38␤.
The results of the SPR, HDX-MS, and IVK assays are consistent with a temperature-dependent conformational change in p38␣ localized to one of its substrate-docking domains that confers efficiency and specificity to substrate phosphorylation (49) and with the previously demonstrated structure of the p38␣:MK2 complex (54). The absence of corresponding temperature dependence within the C-terminal p38␣-interacting  , red) to the CD (yellow) and ED (brown) motifs and the MK2 C-terminal regulatory domain (orange) lying within the substrate-docking groove of p38␣. D, the relative proportion of the percentage exchange-protected envelope after 2 h of deuterium incubation is plotted as a function of temperature for p38␣ (solid) and p38␤ (dashed) for segment 207-215 (green), 281-288 (yellow), and 300 -306 (blue). E, difference plots are plotted for p38␣ (top) and p38␤ (bottom). The differences in percentage deuteration between 39.5 and 33°C after 10 s (orange), 10 min (blue), and 2 h (purple) deuteration incubation were plotted as a function of the peptide segments from the N to C terminus based on the first residue of the segment. Vertical color bars, peptide segment that displayed EX1 bimodal behavior and are color-coded according to D. Segments with significant differences that do not display EX1 bimodal behavior are denoted by vertical back arrows. F, structural representation of temperature-dependent peptide segments of p38␣ (top; PDB entry 5UOJ) and corresponding segments from p38␤ (bottom; PDB entry 3GC8) using the same color coding as in D and E.

Temperature-dependent p38␣ MAPK structure and function
region of MK2 indicates that the temperature dependence of p38␣:MK2 binding and MK2 phosphorylation is attributable to temperature-dependent changes in the conformation of p38␣ itself.
To our knowledge, the only other signaling pathway that has been shown to contribute to lung injury and exhibit temperature dependence in the clinically relevant range is transient receptor potential cation channel subfamily V member 4 Temperature-dependent p38␣ MAPK structure and function (TRPV4), which is expressed on pulmonary endothelium and is responsive to both mechanical stretch and clinically relevant shifts in temperature (69).
In summary, we showed that increases in temperature in a range between clinically relevant hypothermia and febrilerange hyperthermia cause localized conformational changes within the p38␣ substrate-binding groove known to interact with the proinflammatory substrate MK2 and are associated with increased binding and phosphorylation of MK2. Considering the known contribution of the p38-MK2-HSP27 pathway to the pathogenesis of acute lung injury (23), the observed temperature responsiveness of p38␣ might explain the previously described temperature dependence of pulmonary endothelial barrier dysfunction, cytokine expression, and acute lung injury (7,11,70). These results also demonstrate the importance of analyzing signaling processes at temperatures at which they occur in vivo and demonstrate that localized conformational changes in certain kinases may modify the pattern of substrate phosphorylation without substantially altering global catalytic function.

Immunoblotting
Cells were lysed in radioimmunoprecipitation buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 0.1% SDS, and protease and phosphatase inhibitors (Roche Applied Science), and lysates were resolved by SDS-PAGE, transferred to polyvinylidene fluoride membrane, blocked with 5% nonfat dry milk, and probed with primary Abs against an antibody that recognizes monophosphorylated and dually phosphorylated p38␣/␤ and total p38␣/␤. Bands were detected using secondary Abs conjugated to IR fluorophores (LI-COR) and IR fluorescence imaging (Odyssey; LI-COR), and the data were quantified using the LI-COR image analysis software.

In vitro kinase assay
The temperature dependence of p38 substrate phosphorylation was analyzed by in vitro kinase assay in 20 l of reaction volume containing 75 ng of active dually phosphorylated p38␣ or p38␤ (ϳ100 nM), 85-170 ng of substrate proteins (ϳ100 nM), and 5 Ci of [␥-32 P]ATP with final ATP concentration 83 nM in kinase buffer (50 mM Tris-HCl, 10 mM MgCl 2 , 0.1 mM EDTA, 2 mM DTT, 0.01% Brij 35, pH 7.5). Reactions were performed at 33, 37, or 39.5°C for 10, 30, 60, or 120 min and then terminated by adding an equal volume of 2ϫ SDS-PAGE sample buffer. The reaction mixtures were separated by SDS-PAGE. The gels were dried, and the radioactive phosphate incorporation into substrates was measured by phosphorimaging on a Typhoon FLA 7000 (GE Healthcare Life Sciences). The data were quantified using ImageQuant TL (GE Healthcare Life Sciences) and expressed as relative values. Autophosphorylation of p38␣ and p38␤ was measured in the same in vitro kinase reactions as described for substrate phosphorylation, except each reaction contained 2 g of unphosphorylated p38 protein (2.6 M) and nonradioactive ATP at a final concentration of 5 mM and then Temperature-dependent p38␣ MAPK structure and function were immunoblotted using an antibody that detects both monophosphorylated and dually phosphorylated p38␣/␤.

Surface plasmon resonance analysis of protein-protein interactions
Protein-protein interactions of p38 with substrates were investigated by SPR analysis utilizing a Biacore T200 instrument (GE Healthcare). Unphosphorylated p38␣ or p38␤ was covalently bound to the surface of flow cell 2 and flow cell 4 of a CM5 chip to a final level of 1200 and 1400 RU, using the NHS-EDC kit from GE Healthcare. Flow cells 1 and 3 were treated as blanks. MK2 (0 -3 M in 120 l of HBS-EP buffer consisting of 10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, and 0.05% (v/v) Surfactant P20) and ATF2 (0 -1 M in 120 l of HBS-EP buffer) were injected into flow cells 1-4 at 33, 37, and 40°C. The surface was then washed with buffer at the same temperature for 3 min, and the dissociation of analyte-ligand complexes was followed over time. The surfaces of the flow cells were regenerated by injecting 15-l aliquots of 10 mM glycine, pH 1.75, and the process was repeated. Curve fitting and calculation of K D was performed using Biacore BIAeval version 2.0 software as we have described previously (72). Values from the reference flow cell were subtracted to obtain the values for specific binding. For ATF2, the K D was calculated using the Langmuir 1:1 binding model using Equation 1, where k d and k a are the dissociation and association rate constants, respectively ( Fig. 3 (C and D) and Fig. S1). Because p38: MK2 association and dissociation were rapid, we determined K D values for MK2 binding using a steady-state affinity model and Equation 2, where the value of K A is the inverse of K D , R eq is the equilibrium concentration at the analyte concentration (C), and R max is the maximal binding response. The value of K A was calculated by fitting a plot of the equilibrium response (R eq ) against the analyte concentration (C) to this equation, and K D was calculated as the inverse of K A . The experimental data were fitted using numerical integration methods with an iterative approximation algorithm to find the best solution to the equation. The closeness of the fitted and the experimental data were evaluated using a 2 defined in Equation 3, where r f is the fitted value, r x is the corresponding experimental value, n is the number of data points, and p is the number of fitted parameters. Both 1:1 binding curves (Fig. 3 (C and D) and Fig. S1 (A and B)) and steady-state binding curves (Fig. 3, A and  B) are shown.

DSF
The melting temperatures of recombinant p38␣, p38␤, and ERK2 were determined using DSF (73). SYPRO orange (Invitrogen; diluted 1:1000 in 10 mM HEPES, 150 mM NaCl, pH 7.5) and 1 M unphosphorylated recombinant human p38␣, p38␤, or ERK2 were added to 96-well PCR plates. The plates were mixed, sealed, and centrifuged at 1000 rpm for 1 min, and a melting curve was performed using an Applied Biosystems realtime PCR instrument. The melting point was determined from the first derivative curve as described (73).
Duplicate hydrogen:deuterium exchange reactions were performed for each condition by diluting 20 M p38␣ or p38␤ in 10 mM HEPES, pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , and 1 mM DTT 1:10 deuterium buffer (10 mM HEPES, 99.99% D 2 O, 150 mM NaCl, 10 mM MgCl 2 , and 1 mM DTT); incubating in a sand block set to 33, 37, or 39.5°C; and monitored with a digital temperature probe. All reaction components were equilibrated for 5 min at the respective temperatures prior to initiation of the reactions. Reactions were quenched at various times (10 s, 1 min, 10 min, 1 h, and 2 h) with an equal volume of ice-cold 100 mM glycine buffer, 10 mM tris(2-carboxyethyl)phosphine, pH 2.5. Back-exchange correction was performed against fully deuterated controls acquired by incubating 3 l of p38 protein in 27 l of 10 mM HEPES, 99.99% D 2 O, 150 mM NaCl, 10 mM MgCl 2 , and 1 mM DTT containing 4 M deuterated guanidine DCl for 2 h prior to quenching. Fully deuterated control proteins did not exhibit any temperature-dependent deviations when incubated at 33, 37, and 39.5°C. For 38␤, the deuterium uptake reactions and fully deuterated controls were performed as follows. 5 l of protein was diluted with 45 l of D 2 O buffer and quenched with 50 l of quench buffer. All other experimental acquisition parameters were similar to those of p38␣.
The deuterium uptake by the identified peptides over time and for the fully deuterated controls was analyzed using Water's DynamX 3.0 software. The normalized percentage of deuterium uptake (%D) at an incubation time t for a given peptide is given by Equation 4,

Temperature-dependent p38␣ MAPK structure and function
where m t is the centroid mass at incubation time t, m 0 is the centroid mass of the undeuterated control, and m t is the centroid mass of the fully deuterated control. Percentage deuteration difference plots (⌬%D) were generated using the percentage deuteration calculated. Confidence intervals for the ⌬%D plots were determined using the method outlined by Houde et al. (74) using the duplicate runs at 10 s, 10 min, and 2 h, adjusted to percentage deuteration using the fully deuterated controls. Confidence intervals (98%) were plotted on the ⌬%D plots as horizontal dashed lines. EX1-type cooperative unfolding was analyzed using HX-Express2 (52). MK2 controls were performed as described for p38 with the following differences. Coverage maps were obtained from 5 l of 2.5 mg/ml MK2 diluted 1:20 in ice-cold quench buffer with 1.5 M guanidine HCl. Data acquisition was performed on a Synapt G2Si equipped with HDX technology and temperature-controlled sample chambers. Peptides were identified using ProteinLynx Global Server 3.0.3. Deuterium incubation reactions were performed for 10 s, 10 min, and 2 h at 33, 37, and 39.5°C, the same as for p38 with the following exceptions. Sample pre-equilibration and hydrogen exchange reactions were performed in the temperature-controlled sample chambers. 5 l of MK2 was diluted with 45 l of 1ϫ PBS (99.99% D 2 O), pH 7.4, and quenched with 70 l of ice-cold quench (100 mM glycine buffer, 2.5 M guanidine HCl) prior to manual injections.

Statistical methods
Data are presented as mean Ϯ S.E. Differences among more than two groups were analyzed by applying a Tukey honestly significant difference test to a one-way analysis of variance (ANOVA). Differences in the phosphorylation rates between substrates and between p38 isoforms were analyzed by multivariate ANOVA (MANOVA). Differences with p Ͻ 0.05 were considered significant.