p38 Ser/Thr kinases are mitogen-activated protein kinases (MAPKs) involved in the regulation of multiple cellular processes, including cell proliferation, differentiation, senescence, and death (Cuadrado and Nebreda, 2010). The p38 subgroup of MAPKs comprises four members (α-δ) where only p38α is expressed ubiquitously at high levels (Cuadrado and Nebreda, 2010). p38 signaling is strongly activated by environmental stresses (e.g. osmotic shock, ionizing radiation), and biological stimuli, such as growth factors and inflammatory cytokines (Trempolec et al., 2013). Furthermore, p38α has been implicated in several pathological conditions, for example, chronic inflammatory diseases, cancer, and heart and neurodegenerative diseases (Yong et al., 2009, Denise Martin et al., 2012), which is why elucidation of its activation mechanism is of therapeutical importance (Yong et al., 2009; Hammaker and Firestein, 2010).

Like most kinases, p38α is composed of two lobes: the smaller N-terminal lobe, consisting mostly of β-sheets, and the α-helical C-terminal lobe. The lobes are linked by a flexible hinge that forms the ATP-binding site together with structural elements from both lobes (marked regions in Figure 1). In the canonical activation pathway, MAPK kinases (MAP2Ks) dually phosphorylate the TGY sequence in the activation loop (A-loop; Figure 1a and Video 1) which, according to X-ray crystallography (Zhang et al., 2011a), triggers its large conformational rearrangements and the formation of a characteristic β-sheet motif away from the ATP- and substrate-binding sites (Figure 1b). The new position of the A-loop brings N- and C-lobes closer and facilitates the reorientation of key residues which participate in the stabilization of ATP and the catalytically crucial Mg²⁺ ions (Taylor and Kornev, 2011) (i.e. the universally conserved K53, E71, and D168). Moreover, p38α and other MAPKs possess two distinct structural elements (Figure 1): (1) the MAPK insert (MKI), which forms a lipid-binding site (Diskin et al., 2008) and (2) the L16 loop that extends from the C-lobe to the N-lobe where it folds into the C-terminal L16 helix. At its C-lobe segment, the L16 loop contains an acidic patch, called the common docking (CD) motif (Tanoue et al., 2000), which together with the hydrophobic groove, formed by the linker between α_D and α_E helices, and the β₇-β₈ reverse turn (also termed the ED site) (Chang et al., 2002) (Figure 1a), defines the peptide docking site recognized by the conserved docking motif present in activators, substrates, and regulators of MAPKs. The docking site enhances the specificity of MAPKs’ interactions, as well as its activity, albeit through an unknown molecular mechanism (Tanoue et al., 2000). All the structural elements are detailed in the Video 1.

Figure 1

Download asset Open asset

Video 1

Download asset

The canonical activation mechanism outlined above, derived from the analysis of phosphorylated and unphosphorylated crystal structures of p38α (Zhang et al., 2011a) has recently been challenged by an NMR study of p38α in solution (Tokunaga et al., 2014), which surprisingly showed that the dual phosphorylation of the A-loop induces only small conformational changes to p38α, raising doubts about the reliability of conformational states associated with the canonical activation. To understand the changes that take place upon p38α phosphorylation and to explain the observed discrepancies between the NMR and X-ray results, we employed state-of-the-art simulation techniques. Specifically, we used unbiased molecular dynamics (MD) simulations and parallel tempering metadynamics (PT-metaD) (Bussi et al., 2006; Bonomi and Parrinello, 2010; Sutto et al., 2012), a method that combines an enhanced sampling technique with a multi-replica approach and is thus able to converge very complex conformational free-energy surfaces as a function of a few relevant variables, as has been shown previously for several protein kinases (Berteotti et al., 2009; Lovera et al., 2012; Sutto and Gervasio, 2013; Lovera et al., 2015; Marino et al., 2015) and other proteins (Bonomi et al., 2007; Papaleo et al., 2014; Lambrughi et al., 2016). In total, we obtained a high level of sampling (>115 μs) which allowed us to explore the molecular mechanism of p38α canonical activation in unprecedented detail and reconcile the apparently contradictory X-ray and NMR data.

We performed PT-metaD in the well-tempered ensemble to obtain free-energy surfaces (FESs) for human p38α in four different states: (1) the unphosphorylated, (2) the apo dually phosphorylated, (3) the dually phosphorylated state with bound ATP-Mg²⁺, and (4) the dually phosphorylated state with bound ATP-Mg²⁺ and MK2 docking peptide (Figure 2 and Figure 2—figure supplement 1), obtaining total sampling times of 24.25, 24.85, 10.35, and 9.86 μs, respectively. To analyze the systems, we projected the FESs along two collective variables (CVs) - CV₁ and CV₂, both of which are based on the A-loop contacts that measure the distance from the crystallographically observed inactive (CV₁) and active (CV₂) conformations (see Materials and methods), and which were successfully used previously to study protein kinases (Sutto et al., 2012; Sutto and Gervasio, 2013; Marino et al., 2015). When CV₁ < 0.2 and CV₂ > 0.8, then the system assumes an inactive conformation (as in Figure 1a). On the other hand, when CV₁ > 0.6 and CV₂ < 0.2, the A-loop has formed the characteristic β-sheet motif (as in Figure 1b) and the enzyme has adopted the canonical active state. As the representative structure of each highlighted minimum, we chose the central structure of the most populated cluster present in the minimum (see Materials and methods for further details).

Figure 2 with 6 supplements see all

Download asset Open asset

Unphosphorylated p38α

FES of the unphosphorylated state (upper left panel, Figure 2) and the representative structures of the explored minima (Figure 2 and Figure 2—figure supplement 1, enlarged structures in Figure 2—figure supplement 2) show that the unphosphorylated protein neither explores the active conformation, nor does it populate a state able to accommodate ATP. Specifically, in the more inactive of the two free-energy minima - (CV₁, CV₂) = (0.2, 0.7) (upper left panel in Figure 2, Figure 2—figure supplement 2a), the A-loop is positioned over the C-lobe forming contacts with residues involved in the substrate binding and phosphotransfer (Table 1). The P-loop is quite flexible and the K53-E71 salt bridge, which stabilizes ATP, is displaced from the C-lobe due to the broken regulatory spine that is otherwise formed with D168 upon ATP binding (Taylor and Kornev, 2011). Furthermore, E71 interacts with the MAPK-conserved R67 from the α_C helix that stabilizes pT180 in the active form (Figure 1b, present in >90% of structures in the most populated cluster). On the other hand, the A-loop of the second minimum contains a short helix which places the conserved R173 in direct contact with the K53-E71 salt bridge, as well as D168 of the conserved DFG motif (Figure 2, Figure 2—figure supplement 2b), while the P-loop is partly collapsed into the ATP-binding site. We observe similar autoinhibitory interactions for the other two higher energy minima explored by p38α (Supplementary file 1, Figure 2—figure supplement 2c,d). In both cases, R173 is in contact with D168 which otherwise stabilizes ATP upon binding (Kumar et al., 1995; Bellon et al., 1999). Furthermore, in the more inactive minimum (Figure 2—figure supplement 2c), the P-loop is collapsed into the ATP-binding site; while the other explores a conformation reminiscent of the Src-like inactive conformation (Xu et al., 1999) (Figure 2—figure supplement 2d) in which the A-loop forms a short helix stacked against the α_C helix. This configuration has been previously observed also in CDK (De Bondt et al., 1993) and EGFR kinases (Sutto and Gervasio, 2013). These findings are in perfect agreement with experimental studies reporting that the unphosphorylated p38α shows almost no affinity for ATP (Tokunaga et al., 2014; Frantz et al., 1998), confirming the strong regulation of the enzyme by phosphorylation.

Table 1

System	Minimum	Residue pairs
p38α	0–2 kcal mol−1 (CV1, CV2 = 0.2, 0.7)	D176-R189, D177-R186, E178-K152, E178-R173
	0–2 kcal mol−1 (CV1, CV2 = 0.2, 0.6)	R173-E71, D168-R173, D177-R186
p38α-pTpY	0–2 kcal mol−1	E178-R149, E178-R189, D177-R67, pY182-R173
	4–6 kcal mol−1 (CV1, CV2 = 0.45, 0.4)	D176-R149, D176-R189, D177-R70, pT180-R67, R173-E328
p38α-pTpY (ATP-Mg2+)	0–2 kcal mol−1	pT180-R67, pT180-R70, pY182-R186, pY182-R189, E178-R70
p38α-pTpY (ATP-Mg2+ + MK2 peptide)	0–2 kcal mol−1 (CV1, CV2 = 0.7, 0.1)	pT180-R67, pT180-R173, D112-ATP
	0–2 kcal mol−1 (CV1, CV2 = 0.4, 0.4)	pY182-R67, pY182-R70, D177-R173, D112-ATP

To confirm that the extended conformation of the A-loop attributed to the active state is highly unstable in the unphosphorylated protein, we used unbiased MD simulations that were started from the conformation observed in the X-ray structure of p38α-pTpY (PDB ID: 3PY3) (Figure 1b) (see Materials and methods). We performed ten 500-ns-long independent runs at 380 K, all of which quickly showed a change in the A-loop conformation toward the inactive state and sampled conformations that fall in the regions explored in metadynamics simulations. Furthermore, in one of the independent runs the A-loop managed to break the β-sheet motif in the simulated time, so we extended this run to 2.675 μs to allow the protein to transition to the inactive state (Figure 3a). Indeed, the protein mostly remained in the global minima observed in metadynamics simulations, and the final conformation showed very similar features to the representative inactive structures (Figure 3b). Root-mean-square fluctuations (RMSF) reveal a high flexibility of several structural elements besides the A-loop (i.e. α_C, α_D, and the MKI helices, Figure 3b) that occurs due to the opening of the lobes and the local unfolding (in the case of α_D helix). Both these events have been previously linked to large conformational changes occurring in other kinases (Sutto and Gervasio, 2013; Whitford et al., 2007; Shan et al., 2013) but have been so far unobserved for p38α.

Figure 3

Download asset Open asset

Dually phosphorylated p38α

Surprisingly, once p38α is dually phosphorylated, it explores conformations that are not largely different from the ones observed for the unphosphorylated protein (upper right panel in Figure 2 and Figure 2—figure supplement 1, enlarged structures in Figure 2—figure supplement 3), contrary to the reported X-ray data (Zhang et al., 2011a). In the global minimum, the A-loop moves away from the ATP-binding site, mostly due to the electrostatic interactions formed by the phosphorylated residues which subsequently also stabilize the P-loop and prevent it from collapsing into the ATP binding pocket (Table 1). However, the sampled conformation is unlikely to bind ATP due to the formation of a salt bridge between R49 (positioned beneath the P-loop) and D112 (found at the beginning of the α_D helix) (Figure 2, Figure 2—figure supplement 3a; distance histogram is shown in Figure 2—figure supplement 4). This interaction, which seems to be specific for the phosphorylated protein, locks the hinge and occludes the ATP-binding site. However, the protein also explores a higher energy minimum that comes closer to the active one (Figure 2, Figure 2—figure supplement 3b): the A-loop extends toward the L16 loop, yet does not form the β-sheet motif associated with the active conformation. Regardless, the higher energy conformation allows the A-loop to form a number of electrostatic interactions with the α_C helix and the L16 loop (Table 1), while pT180 forms a salt bridge with the highly conserved R67 preventing the P-loop from collapsing into the ATP-binding site. Furthermore, the α_D helix partly unfolds allowing the ATP-binding pocket to assume a more open conformation with the key residues aligned for ATP binding. This observation explains the protein’s low affinity toward ATP in the phosphorylated state (Tokunaga et al., 2014; Frantz et al., 1998) and demonstrates that the dual phosphorylation can help define a receptive ATP-binding site but is not sufficient for the protein to adopt the conformation captured by X-ray crystallography (Figure 1b). In fact, to populate that state, we would need to add 16–18 kcal mol⁻¹ of free energy to the phosphorylated system (minimum located at (CV₁, CV₂) = (0.6, 0.2) in the upper right panel of Figure 2, Figure 2—figure supplement 3d).

To further confirm the instability of the conformation observed by X-ray crystallography, we again used unbiased MD simulations that were started from the said conformation (PDB ID: 3PY3) (Figure 1b) (see Materials and methods). We performed ten 1-μs-long independent runs at 380 K, most of which showed a change in the A-loop conformation toward the regions explored by the protein in metadynamics simulations (upper right panel in Figure 2). Furthermore, in one of the independent runs the A-loop broke the β-sheet motif, so we extended this run to 2 μs during which the protein settled in the aforementioned higher energy minima that seems likely to bind ATP (Figure 4a). The final conformation shows similar features to the representative structure of this minimum, that is, the A-loop is positioned away from the ATP-binding site and forms several electrostatic interactions with the α_c helix and the L16 loop (Figure 4b). RMSF values reveal a high flexibility of several structural elements besides the loops (i.e. α_D, and the MKI helices, Figure 4b) that occurs due to local unfolding.

Figure 4

Download asset Open asset

Dually phosphorylated p38α with bound ATP-Mg²⁺

The FES plot for the phosphorylated system with bound ATP-Mg²⁺ complex shows a clear shift toward a more active conformation (i.e. a displacement toward right and bottom in the lower left panel of Figure 2 and Figure 2—figure supplement 1, enlarged structures in Figure 2—figure supplement 5). While the representative structure of the global minimum does not form the β-sheet motif associated with the active state (Figure 2—figure supplement 5a), its A-loop extends toward the L16 loop and connects to the α_C helix via the coordination of pT180 by R67 and R70 (as observed in the active X-ray structure in Figure 1b). The β-sheet motif does form in the higher energy minimum centered at (CV₁, CV₂) = (0.6, 0.1); however, its representative structure (lower left panel in Figure 2, Figure 2—figure supplement 5b) shows pT180 forming less contacts with the α_C helix and the L16 loop (Supplementary file 1). The lack of these interactions, together with the disorder present in the P-loop, breaks the contact of E71 with Mg²⁺ ions resulting in a less stable complex with ATP. Overall, albeit the binding of ATP stabilizes an active-like structure, it is not enough for the protein to display the active conformation observed in the crystal structure (Figure 1b). While the A-loop does adopt an extended conformation and forms the β-sheet motif at the higher energy state, it still remains quite mobile and the phosphorylated residues do not necessarily form all the contacts observed in the X-ray structures.

Once again, we used unbiased MD simulations to give further support to our free-energy surfaces. We performed ten 1-μs-long independent runs at 380 K of the dually phosphorylated p38α with bound ATP-Mg²⁺ starting with the A-loop in the extended conformation (as in Figure 1b) (see Materials and methods). Most of the simulations explored only the local minimum close to the starting structure, while one of the runs showed more conformational diversity, so we extended it to 1.55 μs which allowed the protein to transition to the global minimum (Figure 5a). The final conformation shows the disruption of the β-sheet motif in the A-loop and the calculated RMSFs reveal its high flexibility (Figure 5b). Furthermore, pT180 remains in contact with the conserved arginine residues of the α_C helix, while pY182 moves away from H228, similarly to the representative structure of the global minimum in the free-energy surface. Overall, RMSF values show a lower flexibility of the protein upon the addition of the ATP compared to the simulations of the apo systems (Figures 3b and 4b).

Figure 5

Download asset Open asset

The active conformation

While the dual phosphorylation and the ATP-binding drive the enzyme toward an active conformation, there is an obvious need for additional elements to complete the transition. Particularly, we wondered whether the binding of the MK2 docking peptide could provide the necessary stability to form the β-sheet motif. The FES for this system indeed shows a deepening of the minimum at (CV₁, CV₂) = (0.7, 0.1) which is associated with the formed β-sheet motif (lower right panel in Figure 2, enlarged structures in Figure 2—figure supplement 6). The representative structure of the minimum (Figure 2—figure supplement 6a) confirms this observation and shows pT180 near the three conserved arginines (R67, R70, and R173), while pY182 again displays higher mobility and moves away from H228 (Table 1, Figure 1b). Furthermore, the addition of the MK2 peptide does not eliminate the global minimum at (CV₁, CV₂) = (0.4, 0.4), observed also in the presence of ATP (two lower panels in Figure 2). However, it flattens the FES which would indicate an easier transition between the two described minima. The representative structure of the minimum (Figure 2—figure supplement 6b) shows a conformation in which the A-loop is extended away from the ATP-binding site, yet without the formed β-sheet motif and with switched positions of phosphorylated residues compared to the X-ray structure (Table 1, Figure 1b).

In addition, we performed unbiased MD simulations of the dually phosphorylated p38α with bound ATP-Mg²⁺ and the MK2 docking peptide. We generated ten 1-μs-long independent runs at 380 K, obtaining a total accumulated sampling time of 10 μs (see Materials and methods). The protein remains mostly in the regions associated with the extended conformation of the A-loop with formed β-sheet motif, but it also explores the conformational space closer to the other global minimum (also the global minimum in the system without the MK2 peptide) (Figure 6a). Interestingly, the N-terminal part of the MK2 docking peptide binds the ED site and stabilizes the α_D-α_E linker hindering the flexibility of the α_D helix and allowing the D112 to reorient and stabilize ATP by forming hydrogen bonds with ribose (Figure 2—figure supplement 6, Figure 6b). While we also observe this interaction in the absence of the MK2 peptide, it is not as stable as in its presence which explains the phosphotransfer enhancement upon MK2 peptide docking (Tokunaga et al., 2014).

Figure 6

Download asset Open asset

Our results suggest that p38α predominantly samples conformations with the fully extended A-loop only upon double phosphorylation and the binding of ATP and the docking peptide. This might seem in contrast to the model obtained from X-ray crystallography (Figure 1b). However, careful analysis of the crystal contacts (which form between the molecules due to their orientation and packing within the crystal, Figure 7) reveals numerous interactions formed between functionally important structural elements, such as the L16 loop. These contacts mimic the stabilizing effects of p38α binders. Particularly, we found an atypically long expression His-tag of a neighboring molecule bound to the hydrophobic docking groove which, upon closer inspection (see below), has the same effect on the protein as the MK2 docking peptide did in our simulations.

Figure 7

Download asset Open asset

To prove that this is the case, we simulated a crystal supercell of the apo dually phosphorylated p38α composed of 4 unit cells and 16 protein molecules (see Materials and methods), which allowed us to maintain crystal contacts and obtain 8 μs of sampling for ~170,000 atoms. As shown by Figure 8, the A-loop is very stable in the conformation captured by the X-ray model - the β-sheet motif is unperturbed during the simulation, as well as the contacts made by pT180 with the three conserved arginines (R67, R70, and R173). However, pY182 is more mobile and at times breaks away from H228 and the conserved arginines of the P+1 loop (R186 and R189). This is consistent with the electron density which is not so well defined for pY182 as it is for the rest of the A-loop. Furthermore, we could also qualitatively match the B-factors reported for the X-ray structure (PDB ID: 3PY3) (Figure 8b,d), while the most mobile regions in our simulations (mainly, the N- and C-terminal tails) are missing from the deposited electron density. We would like to stress that it is impossible for our MD simulations to quantitatively match the reported B-factors for the following reasons: (1) While the crystals were grown at 293 K, the data were collected at 100 K. Since MD simulations and their algorithms have not been developed for such low temperatures, we performed our crystal simulations at 293 K which increased the overall atomic motions; (2) The experimental data collected for the crystals in general are time and ensemble averages of 10²³ molecules which are in the end represented by a single model, while its B-factors then reflect both the static and the dynamic disorder of the crystal. Finally, two independent studies had shown that the refinement process underestimates the disorder present in the crystal thereby producing lower B-factors (Janowski et al., 2013; Kuzmanic et al., 2014).

Figure 8

Download asset Open asset

Thus, our analysis and simulations strongly suggest that the deposited crystal structure does not fully capture the apo conformation that p38α assumes under physiological conditions and it most likely suffers from artifacts due to the crystal contacts.

In the clever paraphrase of Tolstoy’s Anna Karenina, Noble et al. (Noble et al., 2004) stated how all active kinases are alike but an inactive kinase is inactive after its own fashion - referring to the attractiveness of the inactive form in the everlasting search for inhibitor selectivity. Kinases, however, seem to have a few surprises left within their folds, even when it comes to the seemingly uniform active state. In the case of p38α, the protein requires dual phosphorylation of the TGY motif for its activation and subsequent ATP binding. Based on the apo X-ray structure of p38α-pTpY (Zhang et al., 2011b), the phosphorylation event has been perceived as the trigger for the conformational change in which the A-loop moves away from the ATP-binding site enabling the N- and C-lobes to reorient and align the residues necessary for the ATP stabilization. Unlike CDK5 (Berteotti et al., 2009), EGFR (Sutto and Gervasio, 2013), and Src (Shukla et al., 2014; Foda et al., 2015), p38α never explores the α_C-out conformation (neither in any of the deposited X-ray structures - Figure 9a nor in our simulations), in which the helix rotates by >45° (breaking the K53-E71 salt bridge in the process). This likely occurs due to the extensive range of contacts formed between the α_C and the L16 helix, which is unique to MAPKs, indicating that the observed feature could be family-specific. Furthermore, MAP kinases contain two conserved arginine residues (R67 and R70) in the α_C helix that readily form salt bridges with numerous negatively charged residues in the A-loop (Figure 10), thereby stabilizing various A-loop conformations which are not necessarily observed by X-ray crystallography. Despite an overwhelming amount of structural data (with 206 structures of human p38α in the PDB), just 47 structures have complete A-loops (i.e. without any missing or zero-occupancy residues), out of which only one structure contains the A-loop devoid of any crystal contacts (PDB ID: 2BAJ). Unfortunately, this structure has been solved in the DFG-out conformation with a bound inhibitor and it cannot be representative of the apo unphosphorylated state. However, the mere lack of information indicates just how flexible the A-loop is and why it is almost impossible to resolve it unless it forms crystal contacts. Our extensive simulations also indicate that the unphosphorylated p38α never explores a conformation that could possibly bind and stabilize ATP - mostly due to A-loop interactions with the DFG-Asp which is crucial for ATP stabilization (Kumar et al., 1995). It is worth noting that related proteins, like ERK2, are able to bind ATP in the unphosphorylated state (Lee et al., 2005; Sours et al., 2008; Canagarajah et al., 1997; Khokhlatchev et al., 1998) most likely due to longer A-loops (Figure 10), which are then easier to stabilize away from the ATP-binding site.

Figure 9

Download asset Open asset

Figure 10

Download asset Open asset

Phosphorylation, however, does change the free-energy surface and enables the A-loop to form salt bridge interactions with the autoinhibitory arginine residues - R67 and R173. The former of the two, as mentioned above, is MAPK-specific and its importance is highlighted further by the R67A mutation that decreases p38α kinase activity (Gum et al., 1998), and the R67I somatic mutation that was found in colon adenocarcinoma (COSMIC: COSM1444062). The newly formed interactions keep the A-loop away from the ATP-binding site which is occluded by the formation of the R49-D112 salt bridge. ATP binding is therefore only possible in less populated conformations of the protein, explaining the experimentally observed low ATP-binding affinity (Tokunaga et al., 2014; Frantz et al., 1998). If one looks solely at X-ray structures where R49 is ~15 Å away from D112 (Figure 9b), the formation of this particular salt bridge might seem like an artifact arising from force field or methodological imperfections. However, R49 is a part of the loop positioned below the very flexible P-loop (whose residues are often unresolved) and we have often observed its interaction with E160 of the ED site in our simulations (both unbiased and biased), regardless of the protein’s phosphorylation state (R49-E160 is ~9 Å in the analyzed X-ray structures - Figure 9c). We believe that the reason for these discrepancies lies in the fact that the loop containing R49 forms crystal contacts in all the 206 deposited X-ray structures which would most likely restrict its mobility and prevent the interactions observed in MD simulations.

While the binding of ATP-Mg²⁺ helps p38α come closer to the active form, it is insufficient to lead to a complete shift, as proposed by the NMR model (Tokunaga et al., 2014). The full transition, however, does occur in our simulations upon binding of the MK2 docking peptide, which has been previously shown to enhance the catalytic steps (Tokunaga et al., 2014). We hypothesize that the docking peptide prevents the formation of the R49-D112 salt bridge in the phosphorylated protein and brings in the L16 loop. This allows the A-loop to form additional stabilizing contacts, as suggested in the case of ERK2 (Zhou et al., 2006) and causes the observed 20-fold increase in the ATP-binding affinity by the MK2 docking peptide (Tokunaga et al., 2014). The changes that the docking peptide introduces into the protein structure are particularly interesting when it comes to the crystal contact analysis of the dually phosphorylated p38α X-ray structure with the active A-loop conformation. We found an atypically long expression tag of the neighboring molecule bound in the docking site which, together with other crystal contacts, induces the active state stabilized within the crystal. These observations show how important it is to carefully analyze symmetry-related molecules and they call for caution in the interpretation of deposited X-ray structures, as they can be misleading.

Overall, our extensive PT-metaD calculations converge to free-energy surfaces that together with the unbiased MD simulations offer unparalleled insight into the molecular mechanism of p38α canonical activation elucidating a complex conformational behavior that is consistent with experimental observations. Despite the initial discrepancies between previous structural studies, our findings reconcile the deposited models from X-ray crystallography with the NMR data. Considering the importance of p38α for both normal physiology and pathological processes, we hope the knowledge obtained in this study will help to target the protein with more specificity.

1. 1. Abagyan R
   2. Totrov M
   3. Kuznetsov D

   (1994) ICM - a new method for protein modeling and design: applications to docking and structure prediction from the distorted native conformation

   Journal of Computational Chemistry 15:488–506.

   https://doi.org/10.1002/jcc.540150503

   • Google Scholar
2. 1. Anandakrishnan R
   2. Aguilar B
   3. Onufriev AV

   (2012) H++ 3.0: automating pK prediction and the preparation of biomolecular structures for atomistic molecular modeling and simulations

   Nucleic Acids Research 40:W537–W541.

   https://doi.org/10.1093/nar/gks375

   • PubMed
   • Google Scholar
3. 1. Bellon S
   2. Fitzgibbon MJ
   3. Fox T
   4. Hsiao HM
   5. Wilson KP

   (1999) The structure of phosphorylated p38gamma is monomeric and reveals a conserved activation-loop conformation

   Structure 7:1057–1065.

   https://doi.org/10.1016/S0969-2126(99)80173-7

   • PubMed
   • Google Scholar
4. 1. Berendsen HJC
   2. Postma JPM
   3. van Gunsteren WF
   4. DiNola A
   5. Haak JR

   (1984) Molecular dynamics with coupling to an external bath

   The Journal of Chemical Physics 81:3684–3690.

   https://doi.org/10.1063/1.448118

   • Google Scholar
5. 1. Berteotti A
   2. Cavalli A
   3. Branduardi D
   4. Gervasio FL
   5. Recanatini M
   6. Parrinello M

   (2009) Protein conformational transitions: the closure mechanism of a kinase explored by atomistic simulations

   Journal of the American Chemical Society 131:244–250.

   https://doi.org/10.1021/ja806846q

   • PubMed
   • Google Scholar
6. 1. Bonomi M
   2. Gervasio FL
   3. Tiana G
   4. Provasi D
   5. Broglia RA
   6. Parrinello M

   (2007) Insight into the folding inhibition of the HIV-1 protease by a small peptide

   Biophysical Journal 93:2813–2821.

   https://doi.org/10.1529/biophysj.107.106369

   • PubMed
   • Google Scholar
7. 1. Bonomi M
   2. Branduardi D
   3. Bussi G
   4. Camilloni C
   5. Provasi D
   6. Raiteri P
   7. Donadio D
   8. Marinelli F
   9. Pietrucci F
   10. Broglia RA
   11. Parrinello M

   (2009) PLUMED: a portable plugin for free-energy calculations with molecular dynamics

   Computer Physics Communications 180:1961–1972.

   https://doi.org/10.1016/j.cpc.2009.05.011

   • Google Scholar
8. 1. Bonomi M
   2. Parrinello M

   (2010) Enhanced sampling in the well-tempered ensemble

   Physical Review Letters 104:190601.

   https://doi.org/10.1103/PhysRevLett.104.190601

   • PubMed
   • Google Scholar
9. 1. Bussi G
   2. Gervasio FL
   3. Laio A
   4. Parrinello M

   (2006) Free-energy landscape for beta hairpin folding from combined parallel tempering and metadynamics

   Journal of the American Chemical Society 128:13435–13441.

   https://doi.org/10.1021/ja062463w

   • PubMed
   • Google Scholar
10. 1. Bussi G
    2. Donadio D
    3. Parrinello M

    (2007) Canonical sampling through velocity rescaling

    The Journal of Chemical Physics 126:014101.

    https://doi.org/10.1063/1.2408420

    • PubMed
    • Google Scholar
11. 1. Canagarajah BJ
    2. Khokhlatchev A
    3. Cobb MH
    4. Goldsmith EJ

    (1997) Activation mechanism of the MAP kinase ERK2 by dual phosphorylation

    Cell 90:859–869.

    https://doi.org/10.1016/S0092-8674(00)80351-7

    • PubMed
    • Google Scholar
12. 1. Chang CI
    2. Xu BE
    3. Akella R
    4. Cobb MH
    5. Goldsmith EJ

    (2002) Crystal structures of MAP kinase p38 complexed to the docking sites on its nuclear substrate MEF2A and activator MKK3b

    Molecular Cell 9:1241–1249.

    https://doi.org/10.1016/S1097-2765(02)00525-7

    • PubMed
    • Google Scholar
13. 1. Cuadrado A
    2. Nebreda AR

    (2010) Mechanisms and functions of p38 MAPK signalling

    Biochemical Journal 429:403–417.

    https://doi.org/10.1042/BJ20100323

    • PubMed
    • Google Scholar
14. 1. Daura X
    2. Gademann K
    3. Jaun B
    4. Seebach D
    5. van Gunsteren WF
    6. Mark AE

    (1999) Peptide folding: when simulation meets experiment

    Angewandte Chemie International Edition 38:236–240.

    https://doi.org/10.1002/(SICI)1521-3773(19990115)38:1/2<236::AID-ANIE236>3.0.CO;2-M

    • Google Scholar
15. 1. De Bondt HL
    2. Rosenblatt J
    3. Jancarik J
    4. Jones HD
    5. Morgan DO
    6. Kim SH

    (1993) Crystal structure of cyclin-dependent kinase 2

    Nature 363:595–602.

    https://doi.org/10.1038/363595a0

    • PubMed
    • Google Scholar
16. 1. Denise Martin E
    2. De Nicola GF
    3. Marber MS

    (2012) New therapeutic targets in cardiology: p38 alpha mitogen-activated protein kinase for ischemic heart disease

    Circulation 126:357–368.

    https://doi.org/10.1161/CIRCULATIONAHA.111.071886

    • PubMed
    • Google Scholar
17. 1. Diskin R
    2. Engelberg D
    3. Livnah O

    (2008) A novel lipid binding site formed by the MAP kinase insert in p38 alpha

    Journal of Molecular Biology 375:70–79.

    https://doi.org/10.1016/j.jmb.2007.09.002

    • PubMed
    • Google Scholar
18. 1. Dolinsky TJ
    2. Nielsen JE
    3. McCammon JA
    4. Baker NA

    (2004) PDB2PQR: an automated pipeline for the setup of Poisson-Boltzmann electrostatics calculations

    Nucleic Acids Research 32:W665–W667.

    https://doi.org/10.1093/nar/gkh381

    • PubMed
    • Google Scholar
19. 1. Essmann U
    2. Perera L
    3. Berkowitz ML
    4. Darden T
    5. Lee H
    6. Pedersen LG

    (1995) A smooth particle mesh Ewald method

    The Journal of Chemical Physics 103:8577–8593.

    https://doi.org/10.1063/1.470117

    • Google Scholar
20. 1. Foda ZH
    2. Shan Y
    3. Kim ET
    4. Shaw DE
    5. Seeliger MA

    (2015) A dynamically coupled allosteric network underlies binding cooperativity in Src kinase

    Nature Communications 6:5939.

    https://doi.org/10.1038/ncomms6939

    • PubMed
    • Google Scholar
21. 1. Foloppe N
    2. MacKerell, Jr. AD

    (2000) All-atom empirical force field for nucleic acids: I. parameter optimization based on small molecule and condensed phase macromolecular target data

    Journal of Computational Chemistry 21:86–104.

    https://doi.org/10.1002/(SICI)1096-987X(20000130)21:2<86::AID-JCC2>3.0.CO;2-G

    • Google Scholar
22. 1. Frantz B
    2. Klatt T
    3. Pang M
    4. Parsons J
    5. Rolando A
    6. Williams H
    7. Tocci MJ
    8. O'Keefe SJ
    9. O'Neill EA

    (1998) The activation state of p38 mitogen-activated protein kinase determines the efficiency of ATP competition for pyridinylimidazole inhibitor binding

    Biochemistry 37:13846–13853.

    https://doi.org/10.1021/bi980832y

    • PubMed
    • Google Scholar
23. 1. Gum RJ
    2. McLaughlin MM
    3. Kumar S
    4. Wang Z
    5. Bower MJ
    6. Lee JC
    7. Adams JL
    8. Livi GP
    9. Goldsmith EJ
    10. Young PR

    (1998) Acquisition of sensitivity of stress-activated protein kinases to the p38 inhibitor, SB 203580, by alteration of one or more amino acids within the ATP binding pocket

    Journal of Biological Chemistry 273:15605–15610.

    https://doi.org/10.1074/jbc.273.25.15605

    • PubMed
    • Google Scholar
24. 1. Hammaker D
    2. Firestein GS

    (2010) 'Go upstream, young man': lessons learned from the p38 saga

    Annals of the Rheumatic Diseases 69:i77–i82.

    https://doi.org/10.1136/ard.2009.119479

    • PubMed
    • Google Scholar
25. 1. Hess B
    2. Bekker H
    3. Berendsen HJC
    4. Fraaije JGEM

    (1997) LINCS: a linear constraint solver for molecular simulations

    Journal of Computational Chemistry 18:1463–1472.

    https://doi.org/10.1002/(SICI)1096-987X(199709)18:12<1463::AID-JCC4>3.0.CO;2-H

    • Google Scholar
26. 1. Janowski PA
    2. Cerutti DS
    3. Holton J
    4. Case DA

    (2013) Peptide crystal simulations reveal hidden dynamics

    Journal of the American Chemical Society 135:7938–7948.

    https://doi.org/10.1021/ja401382y

    • PubMed
    • Google Scholar
27. 1. Jorgensen WL
    2. Chandrasekhar J
    3. Madura JD
    4. Impey RW
    5. Klein ML

    (1983) Comparison of simple potential functions for simulating liquid water

    The Journal of Chemical Physics 79:926–935.

    https://doi.org/10.1063/1.445869

    • Google Scholar
28. 1. Khokhlatchev AV
    2. Canagarajah B
    3. Wilsbacher J
    4. Robinson M
    5. Atkinson M
    6. Goldsmith E
    7. Cobb MH

    (1998) Phosphorylation of the MAP kinase ERK2 promotes its homodimerization and nuclear translocation

    Cell 93:605–615.

    https://doi.org/10.1016/S0092-8674(00)81189-7

    • PubMed
    • Google Scholar
29. 1. Kumar S
    2. McLaughlin MM
    3. McDonnell PC
    4. Lee JC
    5. Livi GP
    6. Young PR

    (1995) Human mitogen-activated protein kinase CSBP1, but not CSBP2, complements a hog1 deletion in yeast

    The Journal of Biological Chemistry 270:29043–29046.

    https://doi.org/10.1074/jbc.270.49.29043

    • PubMed
    • Google Scholar
30. 1. Kuzmanic A
    2. Pannu NS
    3. Zagrovic B

    (2014) X-ray refinement significantly underestimates the level of microscopic heterogeneity in biomolecular crystals

    Nature Communications 5:3220.

    https://doi.org/10.1038/ncomms4220

    • PubMed
    • Google Scholar
31. 1. Lambrughi M
    2. De Gioia L
    3. Gervasio FL
    4. Lindorff-Larsen K
    5. Nussinov R
    6. Urani C
    7. Bruschi M
    8. Papaleo E

    (2016) DNA-binding protects p53 from interactions with cofactors involved in transcription-independent functions

    Nucleic Acids Research 44:gkw770–9109.

    https://doi.org/10.1093/nar/gkw770

    • PubMed
    • Google Scholar
32. 1. Lee T
    2. Hoofnagle AN
    3. Resing KA
    4. Ahn NG

    (2005) Hydrogen exchange solvent protection by an ATP analogue reveals conformational changes in ERK2 upon activation

    Journal of Molecular Biology 353:600–612.

    https://doi.org/10.1016/j.jmb.2005.08.029

    • PubMed
    • Google Scholar
33. 1. Li H
    2. Robertson AD
    3. Jensen JH

    (2005) Very fast empirical prediction and rationalization of protein pKa values

    Proteins: Structure, Function, and Bioinformatics 61:704–721.

    https://doi.org/10.1002/prot.20660

    • PubMed
    • Google Scholar
34. Book

    1. Lide DR

    (2005)

    CRC Handbook of Physics and Chemistry, Internet Version 2005

    Boca Raton, Florida: CRC Press.

    • Google Scholar
35. 1. Lovera S
    2. Sutto L
    3. Boubeva R
    4. Scapozza L
    5. Dölker N
    6. Gervasio FL

    (2012) The different flexibility of c-Src and c-Abl kinases regulates the accessibility of a druggable inactive conformation

    Journal of the American Chemical Society 134:2496–2499.

    https://doi.org/10.1021/ja210751t

    • PubMed
    • Google Scholar
36. 1. Lovera S
    2. Morando M
    3. Pucheta-Martinez E
    4. Martinez-Torrecuadrada JL
    5. Saladino G
    6. Gervasio FL

    (2015) Towards a molecular understanding of the link between imatinib resistance and kinase conformational dynamics

    PLOS Computational Biology 11:e1004578.

    https://doi.org/10.1371/journal.pcbi.1004578

    • PubMed
    • Google Scholar
37. 1. MacKerell AD
    2. Bashford D
    3. Bellott M
    4. Dunbrack RL
    5. Evanseck JD
    6. Field MJ
    7. Fischer S
    8. Gao J
    9. Guo H
    10. Ha S
    11. Joseph-McCarthy D
    12. Kuchnir L
    13. Kuczera K
    14. Lau FT
    15. Mattos C
    16. Michnick S
    17. Ngo T
    18. Nguyen DT
    19. Prodhom B
    20. Reiher WE
    21. Roux B
    22. Schlenkrich M
    23. Smith JC
    24. Stote R
    25. Straub J
    26. Watanabe M
    27. Wiórkiewicz-Kuczera J
    28. Yin D
    29. Karplus M

    (1998) All-atom empirical potential for molecular modeling and dynamics studies of proteins

    The Journal of Physical Chemistry B 102:3586–3616.

    https://doi.org/10.1021/jp973084f

    • PubMed
    • Google Scholar
38. 1. MacKerell AD
    2. Banavali NK

    (2000) All-atom empirical force field for nucleic acids: II. application to molecular dynamics simulations of DNA and RNA in solution

    Journal of Computational Chemistry 21:105–120.

    https://doi.org/10.1002/(SICI)1096-987X(20000130)21:2<105::AID-JCC3>3.0.CO;2-P

    • Google Scholar
39. 1. Mahoney MW
    2. Jorgensen WL

    (2000) A five-site model for liquid water and the reproduction of the density anomaly by rigid, nonpolarizable potential functions

    The Journal of Chemical Physics 112:8910–8922.

    https://doi.org/10.1063/1.481505

    • Google Scholar
40. 1. Margreitter C
    2. Petrov D
    3. Zagrovic B

    (2013) Vienna-PTM web server: a toolkit for MD simulations of protein post-translational modifications

    Nucleic Acids Research 41:W422–W426.

    https://doi.org/10.1093/nar/gkt416

    • PubMed
    • Google Scholar
41. 1. Marino KA
    2. Sutto L
    3. Gervasio FL

    (2015) The effect of a widespread cancer-causing mutation on the inactive to active dynamics of the B-Raf kinase

    Journal of the American Chemical Society 137:5280–5283.

    https://doi.org/10.1021/jacs.5b01421

    • PubMed
    • Google Scholar
42. 1. Noble ME
    2. Endicott JA
    3. Johnson LN

    (2004) Protein kinase inhibitors: insights into drug design from structure

    Science 303:1800–1805.

    https://doi.org/10.1126/science.1095920

    • PubMed
    • Google Scholar
43. 1. Papaleo E
    2. Sutto L
    3. Gervasio FL
    4. Lindorff-Larsen K

    (2014) Conformational changes and free energies in a proline isomerase

    Journal of Chemical Theory and Computation 10:4169–4174.

    https://doi.org/10.1021/ct500536r

    • PubMed
    • Google Scholar
44. 1. Parrinello M
    2. Rahman A

    (1981) Polymorphic transitions in single crystals: a new molecular dynamics method

    Journal of Applied Physics 52:7182–7190.

    https://doi.org/10.1063/1.328693

    • Google Scholar
45. 1. Peacock CJ
    2. Nickless G

    (1969) The dissociation constants of some phosphorus(V) acids

    Zeitschrift Für Naturforschung A 24:245–247.

    https://doi.org/10.1515/zna-1969-0210

    • Google Scholar
46. 1. Pronk S
    2. Páll S
    3. Schulz R
    4. Larsson P
    5. Bjelkmar P
    6. Apostolov R
    7. Shirts MR
    8. Smith JC
    9. Kasson PM
    10. van der Spoel D
    11. Hess B
    12. Lindahl E

    (2013) GROMACS 4.5: a high-throughput and highly parallel open source molecular simulation toolkit

    Bioinformatics 29:845–854.

    https://doi.org/10.1093/bioinformatics/btt055

    • PubMed
    • Google Scholar
47. 1. Sali A
    2. Blundell TL

    (1993) Comparative protein modelling by satisfaction of spatial restraints

    Journal of Molecular Biology 234:779–815.

    https://doi.org/10.1006/jmbi.1993.1626

    • PubMed
    • Google Scholar
48. 1. Schrodinger LLC

    (2015)

    The PyMOL Molecular Graphics System

    The PyMOL Molecular Graphics System, Version 1.8.

    • Google Scholar
49. 1. Shan Y
    2. Arkhipov A
    3. Kim ET
    4. Pan AC
    5. Shaw DE

    (2013) Transitions to catalytically inactive conformations in EGFR kinase

    PNAS 110:7270–7275.

    https://doi.org/10.1073/pnas.1220843110

    • PubMed
    • Google Scholar
50. 1. Shukla D
    2. Meng Y
    3. Roux B
    4. Pande VS

    (2014) Activation pathway of Src kinase reveals intermediate states as targets for drug design

    Nature Communications 5:3397.

    https://doi.org/10.1038/ncomms4397

    • PubMed
    • Google Scholar
51. 1. Sours KM
    2. Kwok SC
    3. Rachidi T
    4. Lee T
    5. Ring A
    6. Hoofnagle AN
    7. Resing KA
    8. Ahn NG

    (2008) Hydrogen-exchange mass spectrometry reveals activation-induced changes in the conformational mobility of p38alpha MAP kinase

    Journal of Molecular Biology 379:1075–1093.

    https://doi.org/10.1016/j.jmb.2008.04.044

    • PubMed
    • Google Scholar
52. 1. Sutto L
    2. Marsili S
    3. Gervasio FL

    (2012) New advances in metadynamics

    Wiley Interdisciplinary Reviews: Computational Molecular Science 2:771–779.

    https://doi.org/10.1002/wcms.1103

    • Google Scholar
53. 1. Sutto L
    2. Gervasio FL

    (2013) Effects of oncogenic mutations on the conformational free-energy landscape of EGFR kinase

    PNAS 110:10616–10621.

    https://doi.org/10.1073/pnas.1221953110

    • PubMed
    • Google Scholar
54. 1. Tanoue T
    2. Adachi M
    3. Moriguchi T
    4. Nishida E

    (2000) A conserved docking motif in MAP kinases common to substrates, activators and regulators

    Nature Cell Biology 2:110–116.

    https://doi.org/10.1038/35000065

    • PubMed
    • Google Scholar
55. 1. Taylor SS
    2. Kornev AP

    (2011) Protein kinases: evolution of dynamic regulatory proteins

    Trends in Biochemical Sciences 36:65–77.

    https://doi.org/10.1016/j.tibs.2010.09.006

    • PubMed
    • Google Scholar
56. 1. Tokunaga Y
    2. Takeuchi K
    3. Takahashi H
    4. Shimada I

    (2014) Allosteric enhancement of MAP kinase p38α's activity and substrate selectivity by docking interactions

    Nature Structural and Molecular Biology 21:704–711.

    https://doi.org/10.1038/nsmb.2861

    • PubMed
    • Google Scholar
57. 1. Trempolec N
    2. Dave-Coll N
    3. Nebreda AR

    (2013) SnapShot: p38 MAPK substrates

    Cell 152:924–924.e1.

    https://doi.org/10.1016/j.cell.2013.01.047

    • PubMed
    • Google Scholar
58. 1. Tribello GA
    2. Bonomi M
    3. Branduardi D
    4. Camilloni C
    5. Bussi G

    (2014) PLUMED 2: new feathers for an old bird

    Computer Physics Communications 185:604–613.

    https://doi.org/10.1016/j.cpc.2013.09.018

    • Google Scholar
59. 1. Whitford PC
    2. Miyashita O
    3. Levy Y
    4. Onuchic JN

    (2007) Conformational transitions of adenylate kinase: switching by cracking

    Journal of Molecular Biology 366:1661–1671.

    https://doi.org/10.1016/j.jmb.2006.11.085

    • PubMed
    • Google Scholar
60. 1. Xu W
    2. Doshi A
    3. Lei M
    4. Eck MJ
    5. Harrison SC

    (1999) Crystal structures of c-Src reveal features of its autoinhibitory mechanism

    Molecular Cell 3:629–638.

    https://doi.org/10.1016/S1097-2765(00)80356-1

    • PubMed
    • Google Scholar
61. 1. Yong HY
    2. Koh MS
    3. Moon A

    (2009) The p38 MAPK inhibitors for the treatment of inflammatory diseases and ancer

    Expert Opinion on Investigational Drugs 18:1893–1905.

    https://doi.org/10.1517/13543780903321490

    • PubMed
    • Google Scholar
62. 1. Zhang YY
    2. Wu JW
    3. Wang ZX

    (2011a) Mitogen-activated protein kinase (MAPK) phosphatase 3-mediated cross-talk between MAPKs ERK2 and p38alpha

    Journal of Biological Chemistry 286:16150–16162.

    https://doi.org/10.1074/jbc.M110.203786

    • PubMed
    • Google Scholar
63. 1. Zhang YY
    2. Wu JW
    3. Wang ZX

    (2011b) A distinct interaction mode revealed by the crystal structure of the kinase p38α with the MAPK binding domain of the phosphatase MKP5

    Science Signaling 4:ra88.

    https://doi.org/10.1126/scisignal.2002241

    • PubMed
    • Google Scholar
64. 1. Zhou T
    2. Sun L
    3. Humphreys J
    4. Goldsmith EJ

    (2006) Docking interactions induce exposure of activation loop in the MAP kinase ERK2

    Structure 14:1011–1019.

    https://doi.org/10.1016/j.str.2006.04.006

    • PubMed
    • Google Scholar

Author details

1. Antonija Kuzmanic

   Institute for Research in Biomedicine (IRB Barcelona), The Barcelona Institute of Science and Technology, Barcelona, Spain

   Contribution

   AK, Conceptualization, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0003-2815-5605

2. Ludovico Sutto

   Department of Chemistry, University College London, London, United Kingdom

   Contribution

   LS, Conceptualization, Investigation, Methodology

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-4084-8562

3. Giorgio Saladino

   Department of Chemistry, University College London, London, United Kingdom

   Contribution

   GS, Formal analysis, Investigation, Methodology

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-3234-5762

4. Angel R Nebreda

   1. Institute for Research in Biomedicine (IRB Barcelona), The Barcelona Institute of Science and Technology, Barcelona, Spain
   2. Catalan Institution for Research and Advanced Studies (ICREA), Barcelona, Spain

   Contribution

   ARN, Conceptualization, Supervision, Funding acquisition, Project administration, Writing—review and editing

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-7631-4060

5. Francesco L Gervasio

   Department of Chemistry, University College London, London, United Kingdom

   Contribution

   FLG, Conceptualization, Resources, Supervision, Investigation, Methodology, Writing—review and editing

   For correspondence

   f.l.gervasio@ucl.ac.uk

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0003-4831-5039

6. Modesto Orozco

   1. Institute for Research in Biomedicine (IRB Barcelona), The Barcelona Institute of Science and Technology, Barcelona, Spain
   2. Joint BSC-CRG-IRB Program in Computational Biology, Barcelona, Spain
   3. Department of Biochemistry, University of Barcelona, Barcelona, Spain

   Contribution

   MO, Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing—review and editing

   For correspondence

   modesto.orozco@irbbarcelona.org

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-8608-3278

• 3,859

  views

• 653

  downloads

• 62

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.