Dynamic NHE1-Calmodulin complexes of varying stoichiometry and structure regulate Ca2+-dependent NHE1 activation

Calmodulin (CaM) engages in Ca2+-dependent interactions with numerous proteins, including human Na+/H+-exchanger NHE1. Using nuclear magnetic resonance (NMR) spectroscopy, isothermal titration calorimetry, and fibroblasts expressing wildtype and mutant NHE1, we discovered multiple accessible states of this important complex existing in different NHE1:CaM stoichiometries and structures. We solved the NMR solution structure of a ternary complex in which CaM links two NHE1 cytosolic tails. In vitro, stoichiometries and affinities were tunable by variations in NHE1:CaM ratio and calcium ([Ca2+]) and by phosphorylation of S648 in the first CaM-binding α-helix. In cells, Ca2+-CaM-induced NHE1 activity was reduced by mimicking S648 phosphorylation or mutating the first CaM-binding helix, whereas Ca2+-induced NHE1 activity was unaffected by inhibition of Akt, one of several kinases phosphorylating S648. Our results reveal the diversity of NHE1:CaM interactions and suggest that CaM may contribute to NHE1 dimerization. We propose that similar structural diversity is relevant to other CaM complexes.


Introduction 36
Calmodulin (CaM) is a ubiquitously expressed EF-hand Ca 2+ -binding hub protein, which regulates a plethora 6

201
However, signals of the C-lobe now overlaid with the spectrum of CaM bound to H2 (Fig. 3d-f). Thus, at this 202 molar ratio, the major state is the N-lobe of CaM bound to H1 and the C-lobe bound to H2, as depicted in Fig. (CaM:H1H2), two sets of signals originated from each residue of H1, while signals from H2 closely resembled 206 those of free H2 (Figure 2 -Figure supplement 2c). This means that H1 was bound to both CaM lobes, while 207 H2 did not partake in the interaction. At a stoichiometry of 1:1, signals from H1 overlapped with the N-lobe 208 bound state, while signals from H2 were invisible or closely resembled the C-lobe bound state (Figure 2 -209 Figure supplement 2d). For both H1H2:CaM stoichiometries analyzed, the described complexes were the 210 major states as judged by the relative NMR signal intensities. It is, however, important to note that at 1:1 as 211 well as 1:2 (CaM:H1H2), the NMR signals were drastically broadened compared to those of free CaM or of the 212 1:2 complex saturated with H1 (Fig. 3a,d), indicating a dynamic equilibrium of states with different 213 stoichiometries. Higher order complexes are also possible.

214
These results show that in vitro, NHE1 and CaM can form multiple high affinity complexes of different

218
(a) The 1:2 CaM:NHE1622-692 (H1H2) complex (molar ratio 1:2). Three regions of the 1 H, 15 N HSQC spectra of transport indicating that dimerization contributes to the regulation of NHE1 activity [40][41][42][43]. To obtain a more detailed understanding of this interaction, we solved the structure of the ternary complex of CaM bound to two the linker region remains flexible in the bound state ( Fig. 4b-c). In both lobes, the interaction surface is defined 259 by contacts between the hydrophobic face of the amphipathic H1 (L623, I631, I634, L635, N638,

265
[27], inspection of the neighboring molecules in the crystal lattice reveals that the mode of interaction between 266 the CaM C-lobe and H1, which we describe here by NMR, was also present in the crystal, but was translated

284
The NHE1 C-terminal tail contains several phosphorylation sites some of which are located in the CaM-binding 285 region [26]. S648, which can be phosphorylated by Akt [35,36] is located at the C-terminal distal end of H1. In 286 cardiomyocytes, the S648A mutation increased CaM binding to NHE1 in vitro, and S648 phosphorylation by 287 Akt was proposed to inhibit NHE1 by preventing CaM binding [35]. In contrast, in fibroblasts, S648 288 phosphorylation by Akt was required for NHE1 activation by platelet-derived growth factor [36], indicating that 289 the impact of this residue may be cell-type-and/or context-specific.

290
Having the structural information described above at hand, we revisited this issue. To determine the 291 effect of Ca 2+ -CaM interaction and S648 phosphorylation on NHE1 localization and function, we generated

305
To determine the effect of CaM binding and S648 phosphorylation status on NHE1 ion transport 306 activity, cells were loaded with the pH sensitive fluorescent probe BCECF-AM and subjected to real-time induced acid load. Since PS120 cells lack endogenous NHE1 activity ( Fig. 5f-i, untransf.), pHi recovery in the 310 nominal absence of HCO3reflects the activity of exogenously expressed NHE1 [18]. Following an acid load 311 without further stimulation, recovery rates were not significantly different between cells expressing WT and 312 variant NHE1s ( Fig. 5f-g). In the presence of the Ca 2+ -ionophore ionomycin, which causes elevated [Ca 2+ ]I,

317
These results show that mutations within the CaM binding region (specifically H1) or mimicking S648 318 phosphorylation reduce Ca 2+ -stimulated NHE1 activation in PS120 fibroblasts.  To study the interaction between CaM and NHE1 in situ, we visualized and quantified their close proximity 352 using proximity ligation assay (PLA), which detects close proximity (≤ 40 nm) of two proteins of interest in situ 353 [49]. Previous work has established the feasibility of detecting CaM interactions specifically using PLA and 354 demonstrated that PLA signals between two binding partner proteins can be abrogated by mutation of one 355 protein partner [50,51]. Fig. 6a shows the localization of NHE1:CaM PLA puncta (magenta) and NHE1 (green).

361
To determine the impact of S648 phosphorylation status on NHE1:CaM interaction in a cellular 362 context, PS120 cells stably expressing WT NHE1, H1-CR, and the two phosphorylation variants, S648A and 363 S648D, were also probed for NHE1:CaM proximity by PLA. In contrast to their inhibitory effect on Ca 2+ -induced 364 NHE1 activation ( Fig. 5h-i), none of the three variants abolished NHE1:CaM PLA in situ ( Fig. 6a-b). We 365 therefore examined the solution NMR structure of the 1:2 (CaM:H1) complex and the positions of the different 366 mutated residues. In the complex, the four residues K641, R643, R645 and R648 are all solvent exposed both 367 when bound to the N-lobe or the C-lobe of CaM, with some hydrophobic contacts from K641 (Fig. 6c,d).

368
Likewise, in either of the lobes, S648 does not form direct contacts with CaM ( Fig. 6c,d). While several other

402
Akt activity is not solely responsible for the S648 phosphorylation-dependent NHE1 regulation

403
In addition to its phosphorylation by Akt [35,36], S648 is phosphorylated or predicted to be phosphorylated by 404 numerous other kinases, including protein kinase C (PKC), protein kinase A, cyclin-dependent kinase, Aurora 405 kinase, and CDC-like kinase-2 [26]. We therefore asked whether Akt was an essential regulator of Ca 2+ -406 dependent NHE1 activity. A 10 min stimulation with ionomycin had no detectable effect on Akt activation 407 (phosphorylation) in PS120 cells (Fig. 6f). Furthermore, Akt activity was ablated under both control and 408 ionomycin conditions by the Akt inhibitor Akti-1/2 (Fig. 6f), yet Akti-1/2 had no significant effect on ionomycin-409 stimulated NHE1 activity after an acid load (Fig. 6e). These observations suggest that other kinases than Akt 410 might regulate S648 phosphorylation in the context of an increase in [Ca 2+ ]i. To investigate this, and to assess 411 how phosphorylation of S648 impacts NHE1:CaM interaction in vitro, we phosphorylated the H1H2 peptide 412 using either recombinant Akt or PKC in an in vitro phosphorylation assay. Phosphorylation of a single residue 413 by Akt was confirmed by mass spectrometry, and S648 was identified by NMR, while PKC gave rise to mono-

460
The indices c), d), e) and f) refer to the states shown in cartoon below. c) Interaction between apo-CaM and 461 NHE1 is seen, but the structure of it is unknown (indicated by question mark). d) At low Ca 2+ -levels, NHE1-H1 462 is bound to the Ca 2+ -loaded C-lobe of CaM, whereas the N-lobe is in its apo-state. e) at high Ca 2+ -levels, and 463 high NHE1:Ca 2+ -CaM ratio, a 1:1 complex is formed with H1 bound to the N-lobe and H2 bound to the C-lobe

464
of CaM. f) at high Ca 2+ -levels, and low NHE1:Ca 2+ -CaM ratio, a 1:2 complex is formed, where NHE1-H1 is 465 bound to both the N-lobe and the C-lobe of CaM.

467
CaM-mediated oligomerization has been reported for both soluble and membrane proteins.

468
Examples of the former include the ER, two of which bind to one Ca 2+ -CaM, one ER in each lobe, to 469 maximally activate transcription [13,55]. The affinity of ER for one CaM lobe is almost 100 fold lower than 470 what we see here for NHE1, further supporting a unique role of the 1:1 NH1:CaM complex. For membrane 471 proteins, both CaM-mediated multimerization and apo-CaM binding are well studied for tetrameric, voltage-also associates with the channel, triggering dimerization and channel opening [56][57][58]. Another example is the proposed that CaM acts as a molecular clamp interacting with domains from neighboring subunits, thereby 477 "twisting" the pore closed [14].

478
NHE1 proteins are functional homodimers in a manner involving several incompletely understood 479 mechanisms [18]. Based on our findings, it could be speculated that CaM binding contributes to NHE1 C-tail 480 dimerization in a similar way as for ER, by CaM binding to H1 on two separate NHE1s in a 2:1 (NHE1:CaM) 481 stoichiometry (Fig. 7f), as shown in the ternary complex (Fig. 4). Conceivably, the 1:1 stoichiometry observed 482 for CaM:NHE1622-692, with C-lobe:H2 and N-lobe:H1 interaction, could also be possible through a 2:2 complex.
cross-over structure (not shown). Whether this is feasible in a given cellular context will depend on the 485 structural organization of the C-tail complexed with its binding partners, and on the interaction of the NHE1 Since the discovery that NHE1 is a CaM binding protein [28,29], interaction with CaM has emerged 491 as important for its regulation in response to a wide range of stimuli [32][33][34]. In most cases, however, the 492 mechanisms involved have remained unaddressed. Initially, release of an autoinhibitory interaction with an 493 allosteric "pH sensor site" on NHE1 by CaM interaction was proposed [29]. However, detailed kinetic analyses 494 have questioned the existence of such a site and indicated that NHE1 activity is regulated by its dimerization 495 state [18,40,42,43]. Our finding that CaM may contribute to stabilization of the NHE1 dimer state would 496 reconcile these models. Finally, given the key role of anionic lipids including phosphatidyl-inositol(4,5)-497 diphosphate (PI(4,5)P2) in regulating NHE1 by interacting with the C-tail [60,61], Ca 2+ -CaM could also regulate 498 NHE1 through electrostatic tuning of the NHE1:PI(4,5)P2 interaction. Such mechanisms are reported for The role of phosphorylation of S648 in NHE1 regulation has been a conundrum, with apparently 502 conflicting findings in different cell types [35,36]. Here, we show that phosphorylation of this site which is observation mirrors a recent study of CaM binding to smoothelin-like 1 (SMTNL1) protein, where, CaM-

514
In congruence with the reduced CaM affinity of the S648-phosphorylated H1H2 in vitro, Ca 2+ -induced 515 NHE1 activity in PS120 cells was reduced by the S648D phospho-mimicking variant. Inhibiting the kinase 516 responsible for S648 phosphorylation would therefore be expected to increase Ca 2+ -induced NHE1 activity. In 517 our hands, an increase in [Ca 2+ ]i did not activate Akt, one obvious candidate kinase [35,36], and inhibition of Akt had no significant effect on Ca 2+ -mediated NHE1 activity. Furthermore, S648 was phosphorylated in vitro 519 by both Akt and PKC, which have overlapping consensus sites, strongly indicating that other kinases than Akt 520 can act via S648 to regulate NHE1. Conventional PKCs, which require an increase in [Ca 2+ ]i for activation, 521 may be particularly relevant in this context. R643, R645, and R647 which are mutated in NHE1 H1-CR,

522
constitute key residues of the Akt and PKC consensus recognition site [65], which could very well influence 523 phosphorylation in the H1-CR variant. T653 has not been studied in the context of CaM but has been implicated

562
Endocrinology grant (S.F.P.), and a Novo Nordisk Scholarship supported L.S.F. We thank Villumfonden for 563 generous support for the NMR instruments. The funding sources were not involved in study design, data 564 collection and interpretation, or the decision to submit the work for publication.

567
The authors declare that they have no conflicts of interests.

709
After induction the cells were grown for 4h at 37°C and harvested by centrifugation at 6000 g for 10 min. The

825
The following datasets were generated: