Calmodulin as Ca2+-Dependent Interactor of FTO Dioxygenase

FTO is an N6-methyladenosine demethylase removing methyl groups from nucleic acids. Several studies indicate the creation of FTO complexes with other proteins. Here, we looked for regulatory proteins recognizing parts of the FTO dioxygenase region. In the Calmodulin (CaM) Target Database, we found the FTO C-domain potentially binding CaM, and we proved this finding experimentally. The interaction was Ca2+-dependent but independent on FTO phosphorylation. We found that FTO–CaM interaction essentially influences calcium-binding loops in CaM, indicating the presence of two peptide populations—exchanging as CaM alone and differently, suggesting that only one part of CaM interacts with FTO, and the other one reminds free. The modeling of FTO–CaM interaction showed its stable structure when the half of the CaM molecule saturated with Ca2+ interacts with the FTO C-domain, whereas the other part is disconnected. The presented data indicate calmodulin as a new FTO interactor and support engagement of the FTO protein in calcium signaling pathways.


Introduction
The FTO (FaT mass and obesity-associated) protein is an alpha-ketoglutarate and iron-dependent dioxygenase [1]. This two-domain protein is responsible for the removal of methyl groups from certain types of RNA [2]. FTO has been shown to be able to demethylate N 3 -methyluridine from RNA oligomers [3], N 6 -methyladenosine (N 6 -meA) from mRNA and snRNA, N 1 -methyladenosine from tRNA [4], N 6 ,2-O-dimethyladenosine from both snRNA [4] and, the most efficiently, from the mRNA cap [5]. Importantly, the substrate specificity of FTO may be dependent on its localization within the cell [4]. The cap demethylation takes place in the cytoplasm only, snRNA demethylation occurs exclusively in the nucleus while the removal of the methyl group from mRNA happens regardless of its location [4].
The cellular FTO localization depends on its interaction with other proteins. It has been shown that FTO molecules actively move between the nucleus and the cytoplasm. The phosphorylation state of FTO in the position T150, introduced by casein kinase II, determines FTO presence in a particular cell compartment [6]. Phosphorylation of T150 maintains cytoplasmic location of FTO, while non-phosphorylated molecules are directed to the nucleus. Exportin-2 (XPO2) is one of the proteins involved in FTO transport through recognition of the NLS sequence present at the start of the FTO N-domain [7].
In the last 10 years, a number of other interactions influencing FTO parameters and regulating its functions in the cell have been identified. Lin and colleagues [8] have shown that

FTO Protein Forms Homo-and Heterocomplexes In Vivo
We have already shown that overexpression of the FTO protein occurs in head and neck cancer (HNSCC) and that FTO level was positively correlated with the tumor size, indicating the involvement of this protein in cancer metabolism [25]. This raises the question whether FTO interactions with various proteins, already reported for CaMKII [8], GSK-3 [20] or SFPQ [18], explain its engagement in the metabolic processes other than cancerogenesis, such as adipogenesis [4], osteogenesis [26] or neural development [27] in normal cells. To answer the question of whether regulation of FTOs occurs at the gene or protein level, it must be first determined if in vivo FTO exists as monomer, dimer or in protein complexes of higher molecular mass.
Here, size exclusion chromatography (SEC) was performed for the purified recombinant FTO obtained from the baculovirus expression system ( BES FTO), as well as for samples from the HNSCC tissues and specific cell lines, both cancerous (HeLa and U87) and non-cancerous (HEK293). In the absence of other proteins, FTO (58 kDa) showed the ability to form dimers [28]. This form predominates in vitro outside the cell. There were no higher oligomeric forms of the protein, even at high concentrations of BES FTO, that were clearly visible in the Coommasie stained SDS-PAGE gel (Figure 1a). On the other hand, FTO derived from cellular lysates behaved differently. Western blot analysis of fractions eluted from the column showed that FTO appeared in almost all the fractions, suggesting that in cancer tissues and cancer cell lines the protein exists in highly diverse forms, ranging from monomers up to large complexes (Figure 1b-j). Interestingly, in human embryonic kidney cells (HEK293), under the tested conditions, FTO appeared rather as a mixture of dimers and predominating monomers. This behavior strongly suggests that FTO may interact with other proteins, which is clearly evidenced by changes in the elution profile. Moreover, the data obtained in the HEK293 cell line suggest that other proteins or small compounds may interfere with FTO homodimerization.

FTO Protein Forms Homo-and Heterocomplexes In Vivo
We have already shown that overexpression of the FTO protein occurs in head and neck cancer (HNSCC) and that FTO level was positively correlated with the tumor size, indicating the involvement of this protein in cancer metabolism [25]. This raises the question whether FTO interactions with various proteins, already reported for CaMKII [8], GSK-3 [20] or SFPQ [18], explain its engagement in the metabolic processes other than cancerogenesis, such as adipogenesis [4], osteogenesis [26] or neural development [27] in normal cells. To answer the question of whether regulation of FTOs occurs at the gene or protein level, it must be first determined if in vivo FTO exists as monomer, dimer or in protein complexes of higher molecular mass.
Here, size exclusion chromatography (SEC) was performed for the purified recombinant FTO obtained from the baculovirus expression system ( BES FTO), as well as for samples from the HNSCC tissues and specific cell lines, both cancerous (HeLa and U87) and non-cancerous (HEK293). In the absence of other proteins, FTO (58 kDa) showed the ability to form dimers [28]. This form predominates in vitro outside the cell. There were no higher oligomeric forms of the protein, even at high concentrations of BES FTO, that were clearly visible in the Coommasie stained SDS-PAGE gel (Figure 1a). On the other hand, FTO derived from cellular lysates behaved differently. Western blot analysis of fractions eluted from the column showed that FTO appeared in almost all the fractions, suggesting that in cancer tissues and cancer cell lines the protein exists in highly diverse forms, ranging from monomers up to large complexes (Figure 1b-j). Interestingly, in human embryonic kidney cells (HEK293), under the tested conditions, FTO appeared rather as a mixture of dimers and predominating monomers. This behavior strongly suggests that FTO may interact with other proteins, which is clearly evidenced by changes in the elution profile. Moreover, the data obtained in the HEK293 cell line suggest that other proteins or small compounds may interfere with FTO homodimerization.

The Manner of FTO and CaM Interaction
Bearing in mind the elution profiles ( Figure 1) suggesting interaction with other proteins, we looked at FTO sequences signatures determining interactions with other proteins.
Inspection of the Calmodulin Interactors Database (http://calcium.uhnres.utoronto.ca/ctdb; accessed on 2 April 2021) showed that the Cterminal domain of FTO is likely to interact with CaM ( Figure 2a). Analysis of ten FTO

The Manner of FTO and CaM Interaction
Bearing in mind the elution profiles ( Figure 1) suggesting interaction with other proteins, we looked at FTO sequences signatures determining interactions with other proteins. Inspection of the Calmodulin Interactors Database (http://calcium.uhnres.utoronto.ca/ ctdb; accessed on 2 April 2021) showed that the C-terminal domain of FTO is likely to interact with CaM ( Figure 2a). Analysis of ten FTO sequences from distinct species pointed sequences from distinct species pointed out two regions putatively involved in this interaction: the first one corresponds to residues 375-395 of human FTO sequence and is common for Danio rerio, Xenopus laevis, Gallus gallus, Mus musculus and Rattus norvegicus; the second one, covering residues 430-455, is common for Rattus norvegicus, Canis familiaris, Pongo pygmaeus abelii and Sus scrofa. Taking into consideration that CaM is an important signaling molecule, we decided to investigate this interaction experimentally.  Only the sequences with a significant score (>7) are highlighted in green. Such regions are present mostly in FTO C-terminal domain. On the right, human FTO structure with marked sequences of interest (blue) presented on the protein surface. Annotations from Clustal Omega software: "*"-identical amino acids; ":"-high similarity between amino acids; "."-low similarity between amino acids; " "-no identity. (b-e) FTO-CaM interaction in the solution containing 0.5 mM Fe 2+ , 1 mM 2-OG, and 0.5 mM Ca 2+ analysed with the use of MST (a,b). Representative MST pseudo-titration data for the binding of BT CaM to a HisTag-labeled EC FTO (a) or BES FTO (b). Both FTO preparations showed one visible transition, corresponding to the CaM-binding site to FTO, at the nanomolar level. (d,e) Titration of the unlabeled EC FTO (d), BES FTO (e) to a lysine-labeled BES CaM. Both FTO preparations showed one visible transition, corresponding to the CaM-binding site to FTO, at the nanomolar level. Circles represent experimental data, blue lines follow the resulting model of the one binding site, and red lines represents 95% confidence limits for this model. K D are shown as estimated value ± standard deviation. n-number of repetitions.
We use MicroScale Thermophoresis (MST) to follow FTO-CaM interaction. The fluorescently labeled FTO sample was titrated with unlabeled bovine testicle calmodulin ( BT CaM) in the presence of FTO cofactors (Fe 2+ and 2-OG) and Ca 2+ . FTO from both expression systems displayed one binding site (Figure 2b,c), suggesting that the protein may form a heterodimer with CaM (FTO-CaM), with the affinity of the first one at the nanomolar level (K D = 26 ± 10 nM for EC FTO and K D = 34 ± 18 nM for BES FTO). To confirm the interaction between FTO and CaM, we also used another type of fluorescent tag directed to surface-exposed lysine residues. However, with this approach, labeling of BES FTO perturbed the FTO-CaM interaction (Supplementary Figure S1; for details, see Materials and Methods). To study the interaction at the nanomolar level, we changed the labeling strategy and titrated fluorescently labeled BES CaM with unlabeled FTO (in the presence of cofactors and Ca 2+ ). In this case, K D was 9 ± 1 nM and 6 ± 1 nM for EC FTO and BES FTO, respectively (Figure 2d,e). The dissociation constants determined for FTO from both expression systems were similar.

Both CaM Lobes Show the Ability to Interact with FTO
To identify the protein regions taking a part in the FTO-CaM interaction, we used the hydrogen-deuterium exchange (HDX) technique. Taking into consideration that there were no substantial differences in binding affinity of EC FTO and BES FTO to calmodulin, we tested recombinant proteins BES FTO and BES CaM. By investigating the BES CaM HDX profiles, we found that the regions responsible for Ca 2+ binding tended to show a lower rate of exchange relative to residues located in the central part. This is consistent with the supplier's application note for holo calmodulin [29]. This effect was evident for the three detectable calcium-binding sites ( Figure 3). While the addition of the BES FTO did not alter this relationship, in its presence changes in the accessibility of almost all of the BES CaM regions were as early as after 60 s of exposure to D 2 O, with a strong increase in exchange rate for the three observed calcium-binding loops: residues 21-32, 94-105 and 130-141 ( Figure 3). The effect of the BES FTO presence was also observed for the second calcium-binding loop in a second experiment, in which peptides covering the previously mentioned sequence were identified (Supplementary Figure S2). In the reverse experiment, BES CaM only slightly destabilized the FTO structure, mostly in the C-terminal domain, with the effect becoming noticeable after 1 h of incubation (Supplementary Figure S3). These results again confirm FTO-CaM interaction. Further analyses were conducted to obtain more insight into FTO-CaM complex formation.
Detailed analysis of BES CaM HDX data pointed to the specific behavior of particular peptides ( Figure 4a) belonging to the regions responsible for calcium binding. These peptides were observed simultaneously in the both lobes of the BES CaM and they seem to be present in two populations of ions (Figure 4b,c): one with the profile similar to free BES CaM, while the other displaying an increased exchange rate. This effect is detectable at two BES CaM and BES FTO molar ratios: 1 to 1.5 ( Figure 4a) and 1 to 2 (Supplementary Figure S4), clearly suggesting that BES CaM-FTO interaction affects only one lobe of the calmodulin (either N or C terminal), with the other one remaining unchanged, but never both simultaneously. Moreover, there is no specific preference for which of the two lobes is affected more by BES    two BES CaM and BES FTO molar ratios: 1 to 1.5 ( Figure 4a) and 1 to 2 (Supplementary Figure  S4), clearly suggesting that BES CaM-FTO interaction affects only one lobe of the calmodulin (either N or C terminal), with the other one remaining unchanged, but never both simultaneously. Moreover, there is no specific preference for which of the two lobes is affected more by BES FTO presence.

FTO-CaM Interaction is Ca 2+ -Dependent
Keeping in mind that the interaction of CaM with many proteins is Ca 2+ -dependent, we studied the effect of Ca 2+ on the interaction with FTO, using the MST technique for EC FTO labeled on the HisTag. As mentioned above, in the presence of the Ca 2+ , the stronger KD for the FTO-CaM complex was 26 ± 10 nM (

FTO-CaM Interaction Is Ca 2+ -Dependent
Keeping in mind that the interaction of CaM with many proteins is Ca 2+ -dependent, we studied the effect of Ca 2+ on the interaction with FTO, using the MST technique for EC FTO labeled on the HisTag. As mentioned above, in the presence of the Ca 2+ , the stronger K D for the FTO-CaM complex was 26 ± 10 nM (

FTO Interacts with CaM via C-Terminal Domain
A model of the FTO-CaM complex based on the obtained results is presented on Figure 6a. The N-terminal pair of EF-hand motifs of CaM was found to interact efficiently with FTO, while the C-terminal was oriented unfavorably relative to FTO. In order to avoid the steric hindrances between the C-terminal CaM domain and FTO, various conformations of the CaM interdomain linker were tested, while keeping coordinates of the N-terminal pair of EF-hands fixed. The best model was obtained for the linker in a helical conformation, adopted from the holo form of CaM (1CLL), with the C-terminal domain of CaM remaining calcium free. This model is consistent with the experimental results indicating that the presence of calcium is crucial for interaction and that half of the calmodulin displays different properties than the other half.
After detailed investigation of CaM behavior in the presence of the FTO, we experimentally verified FTO C-domain interaction with calmodulin. First, the interaction between bovine testicle calmodulin ( BT CaM) and FTO C-domain (hereafter called CD-FTO) expressed in the E. coli system was tested by a pull-down assay where the presence of His-tagged CD-FTO delayed the removal of BT CaM from nickel beads (Figure 6b). Next, complex formation between these proteins was monitored by gel filtration (Figure 6c). The CD-FTO (22.2 kDa) elution profile showed three clearly visible peaks corresponding to large aggregates (at 11.0 mL), a homodimer (at 13.8 mL), and monomer (at 14.9 mL). BT CaM (17 kDa) was eluted as the singular peak at 14.5 mL. As indicated by its molecular mass, it should have eluted at 15.2 mL, similarly to the myoglobin, but because of its nonglobular shape, BT CaM was eluted earlier. After gel filtration of the CD-FTO-BT CaM mixture, the elution profile showed three distinct peaks. The first likely corresponds to large aggregates (at 11.0 mL), it was also present in the CD-FTO elution profile. The The plot represents complex level of labeled protein at a given BT CaM concentration for each separate sample (data points) and for modeled equilibrium between monomer and dimer (straight lines).

FTO Interacts with CaM via C-Terminal Domain
A model of the FTO-CaM complex based on the obtained results is presented on Figure 6a. The N-terminal pair of EF-hand motifs of CaM was found to interact efficiently with FTO, while the C-terminal was oriented unfavorably relative to FTO. In order to avoid the steric hindrances between the C-terminal CaM domain and FTO, various conformations of the CaM interdomain linker were tested, while keeping coordinates of the N-terminal pair of EF-hands fixed. The best model was obtained for the linker in a helical conformation, adopted from the holo form of CaM (1CLL), with the C-terminal domain of CaM remaining calcium free. This model is consistent with the experimental results indicating that the presence of calcium is crucial for interaction and that half of the calmodulin displays different properties than the other half.
After detailed investigation of CaM behavior in the presence of the FTO, we experimentally verified FTO C-domain interaction with calmodulin. First, the interaction between bovine testicle calmodulin ( BT CaM) and FTO C-domain (hereafter called CD-FTO) expressed in the E. Coli system was tested by a pull-down assay where the presence of His-tagged CD-FTO delayed the removal of BT CaM from nickel beads (Figure 6b). Next, complex formation between these proteins was monitored by gel filtration (Figure 6c). The CD-FTO (22.2 kDa) elution profile showed three clearly visible peaks corresponding to large aggregates (at 11.0 mL), a homodimer (at 13.8 mL), and monomer (at 14.9 mL). BT CaM (17 kDa) was eluted as the singular peak at 14.5 mL. As indicated by its molecular mass, it should have eluted at 15.2 mL, similarly to the myoglobin, but because of its non-globular shape, BT CaM was eluted earlier. After gel filtration of the CD-FTO-BT CaM mixture, the elution profile showed three distinct peaks. The first likely corresponds to large aggregates (at 11.0 mL), it was also present in the CD-FTO elution profile. The second (at 12.4 mL) may result from the initial stages of protein aggregation. The CD-FTO-BT CaM mixture was eluted mainly as a peak at 14.1 mL, with the location corresponding to a putative heterodimer. The fact that this peak eluted between the two latter peaks of pure CD-FTO (13.8 mL for the homodimer and 14.9 mL for the monomer) is consistent with the lower molecular mass of BT  second (at 12.4 mL) may result from the initial stages of protein aggregation. The CD-FTO-BT CaM mixture was eluted mainly as a peak at 14.1 mL, with the location corresponding to a putative heterodimer. The fact that this peak eluted between the two latter peaks of pure CD-FTO (13.8 mL for the homodimer and 14.9 mL for the monomer) is consistent with the lower molecular mass of BT CaM compared with CD-FTO (17 kDa vs. 22.2 kDa).

Discussion
The FTO demethylase is a member of the ALKBH family of proteins connected with both adipogenesis, osteogenesis, maintaining bone mass and heart regeneration, as well as civilization diseases such as obesity, type 2 diabetes, cancer and others [28]. Many studies have demonstrated that FTO is involved in crucial cell processes; however, not much is known regarding the regulation of its activity. Here, we found that FTO-CaM protein-protein interaction plays a key role in understanding the role of FTO in cellular homeostasis. The amino acid sequence of FTO comprises a fairly conserved motif predicted to be recognized by CaM. Using several complementary approaches, we verified that FTO can indeed interact with CaM at nanomolar concentrations in the presence of Ca 2+ . Moreover, the interplay between FTO and CaM was only slightly affected by the type of the expression system used and only in the part of the experiments. Thus, the presence of posttranslational modifications, especially serine phosphorylation [28], occurring mostly in the FTO N-terminal domain did not significantly affect FTO-CaM binding. The obtained results showing nanomolar K D are likely to be biologically relevant, because FTO in tissues occurs at nanomolar-or even lower-concentrations that can be detected by sensitive methods, such as Western blot [30]. On the other hand, total CaM concentration can be much higher. In certain states, it can reach several µM [31,32] or even 40 µM [33]. However, the free pool of CaM in a cell may be much lower due to the interactions of CaM with other protein targets, e.g., GAP-43 or RC3 [34,35]. However, it is important to note that the CaM/Ca 2+ affinity to FTO is stronger than many other CaM interactors measured by similar biophysical methods. Here, we estimated K D for CaM/Ca 2+ /FTO complexes to equal ca. 10-30 nM. In contrast, for the CaM/Ca 2+ /CaMKII-specific peptide complex (residue 290-309), the K D measured by MST technique was much higher (190 nM). Meanwhile, the neurogranin protein binds the apo form of calmodulin with even lower affinity: K D = 890 nM [36]. Finally, another newly identified CaM interactor, phosphoinositideinteracting regulator of transient receptor potential (PIRT), shows Ca 2+ -dependent affinity with K D of 350 nM and 60 µM in the absence and presence of Ca 2+ , respectively [37].
Having confirmed FTO-CaM interaction, we then investigated the one binding site mode, in comparison with other known calmodulin interactors. According to the Calmodulin Target Database, 350 protein sequences have been shown to bind CaM. The process of CaM-targeted amino acid sequence binding is known to occur both in the presence, as well as in the absence, of Ca 2+ .
Earlier, Nissen and co-workers [38] analyzed a number of CaM-targeted complexes and stored them in the Protein Data Bank (PDB). Unsurprisingly, the presence of four Ca 2+ within the CaM-targeted complex was found in the majority of studied PDB structures, because CaM is known to be able to interact with the target proteins/peptides upon binding four Ca 2+ . Multiple examples of four Ca 2+ -bound CaM-targeted complexes have been reported so far, predominantly featuring CaM-dependent protein kinases and phosphatases [39]. The two groups are characterized by a number of CaM-binding modes, namely 1-14, 1-5-10 and 1-10-16. However, CaM has also been shown to form Ca 2+ -free complexes. This includes, for example, the CaM-NaV1.5 (IQ-motif) interaction that occurs within the C-lobe of CaM and IQ-motif in α6 of NaV1.5 or the previously mentioned interactions with neurogranin and PITR. Notably, in these cases, the CaM C-lobe adopts a semi-open conformation [40]. Another Ca 2+ -free complex is CaM-Myosin-5A with the following topology: the C-lobes of the two CaM molecules bind to the N-terminals of the IQ motifs of NaV1.5 in an antiparallel orientation [41]. The last type of CaM-targeted complexes features partial CaM saturation with Ca 2+ . The CaM-binding domain (CaMBD) of small-conductance Ca 2+ -activated potassium channel, SK2-a, forms a complex with a single CaM molecule at each of the two ends of the SK-CaMBD dimer consisting of two helix-loop-helix motifs. Interestingly, only the N-lobes of CaM molecules bound Ca 2+ wrapped around three α-helices of CaMBDs [42]. The ApoCaM-SK2-a binding mode is different: a single SK CaMBD binds to the C-lobe of CaM, resulting in 1:1 stoichiometry. A helical fragment of the CaMBD was shown to interact with one of the CaM C-lobes [43].
Here, our structural analysis showed that the interaction with FTO destabilized the CaM, especially the calcium-binding regions; nevertheless, the reciprocal effect was negligible. This may be explained by FTO homodimerization, when interfaces of the FTO-FTO homodimer and the FTO-CaM heterodimer are similar [28]. Possibly, both interactions require FTO to maintain the same particular structure. Further, FTO-CaM interaction on the nanomolar level was also Ca 2+ -dependent, with negligible K D in the absence of Ca 2+ . This feature is typical for CaM interaction with numerous partners [44]. Importantly, this indicates that the proteins are likely to form a complex when the concentration of free Ca 2+ is sufficiently high and that they may be regulated by calcium concentration. For example, such complexes may be present in neural cells where the CaM level influences synapse development [36] and FTO activity regulates neuronal development and the proliferation and differentiation of adult neural stem cells [27].
Finally, we modelled the structure of the newly discovered FTO-CaM complex. The generated structure indicates similarity between the FTO-CaM-binding mode and that of the previously mentioned SK2-a-CaM. Indeed, FTO-CaM interaction at nanomolar concentrations is Ca 2+ -driven, but only two Ca 2+ binding to the single lobe of CaM participate in the complex formation, while the second lobe remain free of Ca 2+ (Figure 6a). This model is also consistent with the HDX results. The predicted stoichiometry of the biological FTO-CaM complex is 1:1, with the CaM molecule wrapped around the FTO C-terminal domain. Importantly, this type of interaction suggests that CaM-FTO may mediate FTO interaction with other partners.
The discovery of this exclusive CaM-FTO interaction type sheds a new light on the results of previous FTO studies [8]. It is worth mentioning that both proteins interact with numerous other proteins, including three isoforms of CaMKII. Li and coworkers have shown that an increased FTO expression delays forskolin-mediated dephosphorylation of one of CaMKII's cellular targets: cAMP response element-binding protein (CREB) in human neuroblastoma cells [8]. However, direct influence of FTO on CaMKII enzymatic activity as well as FTO phosphorylation by this kinase were not observed. Due to the fact that CaM is a CaMKII activator, our results show new connections between CaM, CaMKII and FTO phosphorylation by CaMKII. Non-physiological FTO concentration of 25 µM in the performed experiment and the finding that the affinity of CaM/Ca 2+ to FTO seems to be higher than to CaMKII likely results in a shortage of free CaM necessary for CaMKII activation. Clearly, this relationship indicates that interplay between FTO, CaM and CaMKII needs further study [8]. Finally, it cannot be excluded that FTO-CaM interaction may affect the binding of other FTO interactors. For example, as previously mentioned, the FTO C-terminal domain interacts with SFPQ protein. This interaction affects substrate specificity of the FTO protein, thus the regulatory role of calmodulin has to be taken into consideration.

•
For investigation of the molecular state of endogenous human FTO, fast protein liquid chromatography (FPLC) was performed. Recombinant FTO from baculovirus expression system ( BES FTO) was expressed and purified as described previously [28]. Protein extracts from selected cancer tissues and HeLa, HEK293, and U87 cell lines were prepared as described previously [25]. A total of 900 µg of BES FTO and 200 µg of each protein lysate sample of a final volume of 500 µL were loaded onto a SEC650 column (#780-1650, BIO-RAD, Hercules, CA, USA) equilibrated and eluted with the PBS buffer. The eluate was monitored with NGC system ChromLab™ (BIO-RAD, Hercules, CA, USA) at 215, 260, and 280 nm. Protein standards (#1511901, BIO-RAD, Hercules, CA, USA) for calibration were analyzed using the same method. Next, fractions of 1 mL were precipitated by the addition of 20% pyrogallol redmolybdate (PRM, #611A, Sigma-Aldrich, Darmstadt, Germany) solution, subsequently re-dissolved in Laemmli buffer and analyzed by Western blot as before [25].

In Silico Analysis of FTO Sequence
The amino acid sequence of human FTO was analyzed to identify possible protein interactors with the set of software: ProSite (https://prosite.expasy.org/; accessed on 1 April 2021), The Calmodulin Target Database (http://calcium.uhnres.utoronto.ca/ctdb; accessed on 2 April 2021), and MyHits (https://www.expasy.org/resources/myhits/; accessed on 3 April 2021). After detection of the region potentially binding calmodulin in C-terminal domain of human FTO, the amino acid sequence of FTO C-domain from selected species were investigated by The Calmodulin Target Databes. Next, the similarity between sequences was evaluated by the Clustal Omega software (https://www.ebi.ac. uk/Tools/msa/clustalo/; accessed on 4 April 2021).

cDNA Cloning and Plasmid Constructs
For use in the Baculovirus Expression System (BES), the coding sequence of CaM was PCR-amplified with suitable primers and inserted into the pENTRY-IBA5 donor vector between Xbal and HindIII sites. The insert was then moved to the pLSG-IBA35 destination vector (StarGate®cloning system, IBA Life Science, Göttingen, Germany).
To obtain the FTO C-terminal domain, the reaction was performed in a 50 µL of a final concentration of CON1. The cDNA was amplified in an Eppendorf Mastercycler ® according to the program: 95 • C for 2 min, followed by 30 cycles of 95 • C for 20 s, 61.5 • C for 30 s, 70 • C for 1 min and finally 70 • C for 10 min. The primers used for the FTO C-terminal domain construct generation are shown in Supplementary Table 1. Next, the PCR product was incorporated into empty pET28a(+) plasmid with the use of In-Fusion ® HD Cloning Plus (Takara Bio, #638909), according to protocol. Then,~100 ng of plasmid was mixed with 200 µL of E. coli competent DH5α bacteria solution, incubated for 30 min on ice, heat shocked for 30 s at 42 • C, 800 µL of SOC buffer was added and the bacterial suspension was incubated for 1 h at 37 • C. The bacteria were plated on LB-agar plates and incubated overnight at 37 • C. Single colonies were cultured in liquid LB overnight at 37 • C. Bacteria were then harvested and plasmid was isolated with the use of GeneJET Plasmid Miniprep Kit (#K0503, Thermo Scientific™, Waltham, MA, USA).
All plasmids were verified by sequencing.

Expression and Purification of EC FTO and BES FTO
The procedure was performed as before [28].

Expression and Purification of BES CaM in Baculovirus Expression System (BES)
The recombinant baculoviruses were generated directly in Sf21 insect cells by cotransfection with plasmid containing the human CaM coding sequence and FlashBacUltra virus DNA following the manufacturer's protocol (FlashBAC™ system, Oxford Expression Technologies, Oxford, United Kingdom). Insect cells were infected with a recombinant baculovirus at a MOI = 4. After 72 h, the cells were harvested, resuspended in PRP4 lysis buffer and homogenized. After centrifugation (20,000× g, 10 min.), proteins His-tagged at the N-terminus were purified and verified by size exclusion and SDS-PAGE. The purified proteins were flash-frozen in liquid nitrogen and stored at −80 • C.

Expression and Purification of FTO C-Terminal Domain in E. coli
Plasmids harboring cDNA encoding C-terminally His-tagged C-terminal domain of human FTO, were introduced into E. coli BL-21. Bacteria were cultured at 37 • C to OD 600 = 0.8 in 2 L of LB supplemented with kanamycin (50 µg/mL) and chloramphenicol (25 µg/mL), induced with 1 mM IPTG, cultured for 16 h at 16 • C, harvested and handled as described for full length EC FTO [28].

Preparation of Protein Samples
Procedure was performed as before [28].

Pull-Down Assay
Protein samples: 67 pmol of BT CaM (1.3 µM) and/or 130 pmol of CD-FTO (2.6 µM) were incubated 10 min at 4 • C in PDA1 buffer of final volume 50 µL. Next, 50 µL of Ni-charged resin suspension (Profinity™ IMAC Ni-Charged Resin, BIO-RAD, #156-0133) was added and incubated overnight at 4 • C with mixing. The resin was then washed with increasing concentrations of imidazole in the PDA1 buffer: five times with 10 mM imidazole (50 µL per wash), three times with 150 mM and one with 0.5 M. Subsequently, the samples were diluted by 4x SDS-PAGE loading buffer, boiled 5 min and separated on the Mini-PROTEAN TGX 4-15% gradient gels (Bio-Rad). Proteins were visualized by Coomassie staining.

•
HisTag labeling: For each labeling, 125 pmol of protein (50 µL of 2.5 µM solution) and 125 pmol of the dye were used. After labeling, the protein was transferred to experimental buffer MST2 alone or supplemented with 0.5 mM CaCl 2 (MST3). • Lysine labeling: The procedure was performed as before [28]. Briefly, for each labeling,  Figure S1). This suggests that protein labeling may occur on lysines proximal to the FTO and CaM interface and such labeling strategy for this type of experiment should be avoided.
Pseudo-titration experiments were performed on a Monolith NT.115 RED/BLUE device equipped with holder allowing for parallel measurement of 16 samples placed in capillaries. Emission of the dye was monitored at 670 nm (excitation at 650 nm by LED laser) at room temperature. The LED laser power and MST power were optimized for each system individually, according to the labeling efficiency. A total of 2-3 repetitions of each experimental set-up with different MST powers were performed.
The data were analyzed using R 3.3.3 software [45] according to a standard two-state model describing a 1:1 equilibrium between the unbound and bound forms of the labeled protein [46], as well as using Origin software (www.originlab.com; accessed on 28 April 2021) according to the three-state model describing coexistence of 1:1 and 1:2 complexes [47].

Hydrogen-Deuterium Exchange (HDX)
Protein samples at two different ratios: 1-3 nmol of BES FTO (30 µM) and/or 2 nmol BES CaM (20 µM); 2-4 nmol of BES FTO (40 µM) and/or 2 nmol BES CaM (20 µM), were incubated for 30 min at 4 • C in an incubation buffer supplemented with 0.5 mM CaCl 2 (HDX1). The samples were then diluted 10-fold with HDX2 reaction buffer (prepared in D 2 O). The hydrogen-deuterium exchange was terminated at: 0 s, 10 s, 1 min, 20 min, 1 h, and 24 h. The exchange was terminated by decreasing pH to approx. 2-3 with HDX3 stop buffer and the samples were immediately frozen in liquid nitrogen. The extent of hydrogen-deuterium exchange was determined using a nanoACQUITY UPLC system (Waters Corporation, Milford, MA, USA). The samples, after initial protein digestion on a trypsin column, were analyzed by liquid chromatography coupled with mass spectrometry (LC-MS). Each experiment was repeated four times. The data were analyzed with the DynamX HDX Data Analysis Software 3.0 (Waters Corporation, Milford, MA, USA).

Molecular Modeling
The structure of the FTO-CaM heterodimer was initially modeled based on the structure of CaM interacting with a fragment of small conductance potassium channel SK2-A (4qnh; A), whose CaM-interacting helix2 (D445-T486) was aligned iteratively with helix12 of FTO (P427-S458) using varying sets of successive residues located in the two helices. The best model was obtained by aligning residues 476 QANTLVDLAKT 486 of SK2-A with residues 439 LASLTARQNLR 449 of FTO.
The final model was obtained after several rounds of simulated annealing, in which the coordinates of FTO and the N-terminal domain of CaM together with the local fold of the C-terminal CaM domain were fixed, and the linker was kept in helical form.

Conclusions
The last 10 years of intensive research into the role of the FTO protein in metabolism have identified a number of transcripts as substrates of this demethylase, affecting cell metabolism and mediating FTO specificity. Our discovery of CaM as another FTO interactor supports the thesis that this demethylase is regulated by other proteins. We discovered the ability of FTO, especially its C-terminal domain, to interact in vitro with the single lobe of a CaM molecule in a Ca 2+ -dependent manner. Moreover, only one part of CaM interacts with FTO, whereas the other one remains free. This interaction may be a part of regulatory mechanisms that FTO is subjected to, due to the fact that other FTO interactors, such as FTO itself and SFPQ protein, also bind to the FTO C-domain. The recognition of the role of CaM in the FTO interaction helps broaden the understanding of the function of this RNA modification demethylase. This finding is important because the multiplicity of FTO interactors may explain the involvement of this protein in a number of different metabolic pathways.