αKlotho is a Non-Enzymatic Molecular Scaffold for FGF23 Hormone Signaling

Soluble αKlotho^ecto acts as a co-receptor for FGF23

To determine whether soluble αKlotho^ecto can support FGF23 signaling, αKlotho-deficient HEK293 cells – which naturally express FGFRs – were incubated with a concentration of αKlotho^ecto sufficient to drive all available cell surface cognate FGFRs into binary complexed form. Following brief rinses with PBS, the cells were stimulated with increasing concentrations of FGF23. In parallel, a HEK293 cell line overexpressing membrane-bound αKlotho (HEK293-αKlotho™) was treated with increasing concentrations of FGF23. The dose-response for FGF23-induced ERK phosphorylation in αKlotho^ecto-pretreated untransfected HEK293 cells was similar to that observed in HEK293-αKlotho™ cells (Extended Data Fig. 1b, upper panel), suggesting that αKlotho^ecto can serve as a co-receptor for FGF23. Pre-treatment of HEK293-αKlotho™ cells with αKlotho^ecto did not result in any further increase in FGF23 signaling, implying that all cell surface FGFRs in this cell line were in binary FGFR-αKlotho™ form (Extended Data Fig. 1b, lower panel). We conclude that soluble and transmembrane forms of αKlotho possess a similar capacity to support FGF23 signaling. Consistent with these results, injection of wild-type mice with αKlotho^ecto protein led to an increase in renal phosphate excretion and a decrease in serum phosphate (Extended Data Fig. 1c). Notably, it also led to a 1.5-fold increase in Egr1 transcripts in the kidney (Extended Data Fig. 1d), demonstrating that αKlotho^ecto can serve as a bona fide co-receptor to support FGF23 signaling in renal proximal tubules. In light of these data, we propose that the pleiotropic anti-aging effects of αKlotho are all dependent on FGF23.

Structural basis for αKlotho co-receptor function

We solved the crystal structure of a human 1:1:1 FGF23-FGFR1c^ecto-αKlotho^ecto ternary complex at 3.0 Å resolution (Extended Data Table 1). In this complex, αKlotho^ecto serves as a massive scaffold, tethering both FGFR1c and FGF23 to itself. In doing so, αKlotho^ecto enforces FGF23-FGFR1c proximity and thus augments FGF23-FGFR1c binding affinity (Fig. 1). The overall geometry of the ternary complex is compatible with its formation on the cell surface (Extended Data Fig. 2a).

The binary FGF23-FGFR1c^ecto complex adopts a canonical FGF-FGFR complex topology in which FGF23 is bound between the receptor’s D2 and D3 domains, engaging both these domains and a short interdomain linker (Extended Data Fig. 3a). However, compared to paracrine FGFs, FGF23 makes fewer/weaker contacts with the D3 domain and D2-D3 linker, explaining the inherently low affinity of FGF23 for FGFR1c (Extended Data Fig. 3b, c). Notably, analysis of the binding interface between FGF23 and FGFR1c D3 in the crystal structure reveals specific contacts between FGF23 and a serine residue uniquely present in the “c” splice isoforms of FGFR1-3 and FGFR4 (Extended Data Fig. 4a). Indeed, replacing this “c”-isoform specific serine residue with a “b”-isoform specific tyrosine impaired FGF23 signaling (Extended Data Fig. 4b, c). We conclude that the FGFR binding specificity inherent to FGF23 operates alongside that of αKlotho (Extended Data Fig. 4d, e) to restrict FGF23 signaling to the “c” splice isoforms and FGFR4^11,12.

In the ternary complex, αKlotho^ecto exists in an extended conformation. Consistent with their sequence homology to the glycoside hydrolase A (GH-A) clan⁸, αKlotho KL1 (Glu-34 to Phe-506) and KL2 (Leu-515 to Ser-950) domains each assume a (βα)₈ TIM barrel fold consisting of an inner eight-stranded parallel β-barrel and eight surrounding α-helices (Fig. 2a and Extended Data Fig. 5a). The two KL domains are connected by a short, proline-rich and hence stiff linker (Pro-507 to Pro-514) (Fig. 1a, b). KL1 sits atop KL2, engaging it via a few interdomain contacts involving the N-terminus preceding the β1 strand, the α7 helix of KL1, and the β5α5, β6α6 loops and the α7 helix of KL2 (Extended Data Fig. 2b). Intriguingly, one of the interdomain contacts is mediated by a Zn²⁺ ion (Fig. 3c and Extended Data Fig. 2b, c). These contacts stabilize the observed elongated conformation of αKlotho^ecto, creating a deep cleft between the two KL domains. This merges with a wide-open central β-barrel cavity in KL2, and forms a large binding pocket that tethers the distal C-terminal tail of FGF23 past the ¹⁷⁶Arg-His-Thr-Arg¹⁷⁹ proteolytic cleavage site (Fig. 1b). Meanwhile, the long β1α1 loop of KL2 (Fig. 2a) protrudes as much as 35 Å away from the KL2 core to latch onto the FGFR1c D3 domain, thus anchoring the receptor to αKlotho (Fig. 1b). Accordingly, we have named this KL2 loop the “Receptor Binding Arm” (RBA; residues 530-578; Extended Data Fig. 5a).

We superimposed the TIM barrels of KL1 and KL2 onto that of Klotho Related Protein (KLrP; also known as GBA3), the cytosolic member of the Klotho family with proven glycosylceramidase activity²⁶. This comparison revealed major conformational differences in the loops surrounding the entrance to the catalytic pocket in KL1 and KL2 (Fig. 2b and Extended Data Fig. 5b–d). Moreover, both KL domains lack one of the key catalytic glutamates deep within the putative catalytic pocket. These substantial differences are incompatible with an intrinsic glycosidase activity for αKlotho^22,23. Indeed, αKlotho^ecto failed to hydrolyze substrates for both sialidase and β-glucuronidase in vitro (Fig. 2c). Together, our data define αKlotho as the only known example of a TIM barrel protein that serves purely as a non-enzymatic molecular scaffold.

Binding interface between αKlotho and FGFR1c

The interface between αKlotho RBA and FGFR1c D3 (Fig. 3a) buries over 2,200 Ų of solvent-exposed surface area, which is consistent with the high affinity of αKlotho binding to FGFR1c (K_D = 72 nM)¹⁰. At the distal tip of the RBA, residues ⁵⁴⁷Tyr-Leu-Trp⁵⁴⁹ and ⁵⁵⁶Ile-Leu-Arg⁵⁵⁸ form a short β-strand pair (RBA-β1:RBA-β2) as their hydrophobic side chains are immersed in a wide hydrophobic groove between the four-stranded βC’-βC-βF-βG sheet and the βC-βC’ loop of FGFR1c D3 (Fig. 3b, upper panel). The RBA-β1:RBA-β2 strand pair forms an extended β sheet with the βC’-βC-βF-βG sheet of D3 as the backbone atoms of RBA-β1 and D3 βC’ make three hydrogen bonds which further augment the interface (Fig. 3b, lower panel). Residues at the proximal end of the RBA engage a second smaller binding pocket at the bottom edge of D3 next to the hydrophobic groove (Extended Data Fig. 6a, b). Both αKlotho binding pockets in the receptor D3 domain differ between “b” and “c” splice isoforms. Leu-342, for example, is strictly conserved in the “c” splice isoforms of FGFR1-3 and FGFR4. This explains the previously described binding selectivity of αKlotho for this subset of FGFRs (Extended Data Fig. 4a)^11,12,27.

Consistent with the crystal structure, soluble αKlotho lacking the RBA (αKlotho^ecto/ΔRBA) failed to form a binary complex with FGFR1c^ecto in solution (Fig. 4a) and hence could not support FGF23 signaling (Fig. 4b). Likewise, membrane-bound αKlotho lacking the RBA (αKlotho^TM/ΔRBA) was also disabled in acting as a FGF23 co-receptor (Fig. 4b). Importantly, αKlotho^ecto/ΔRBA did not exhibit any phosphaturic activity in vivo (Extended Data Fig. 7a). On the contrary, the αKlotho^ecto/ΔRBA mutant antagonized the activity of native αKlotho by sequestering FGF23 into functionally inactive binary complexes, i.e. by acting as an FGF23 ligand trap (Extended Data Fig. 7). These data refute the concept that αKlotho^ecto functions as an FGF23-independent phosphaturic enzyme²⁴. Our conclusion is supported by a gene knockout study which compared the phenotypes of Fgf23^−/−, Klotho^−/−, and Fgf23^−/−/Klotho^−/− mice²⁸.

Binding interface between αKlotho and FGF23

Regions from both KL domains act together to recruit FGF23 (Fig. 1b), thus explaining why only an intact αKlotho ectodomain is capable of supporting FGF23 signaling^12,29. The interactions between FGF23 and αKlotho result in the burial of a large amount of solvent-exposed surface area (2,732 Ų), of which nearly two-thirds (1961 Ų) are buried between the FGF23 C-terminal tail and αKlotho, with the remaining one-third buried between the FGF23 core and αKlotho (Fig. 3a). At the interface between αKlotho and FGF23 C-terminal tail, FGF23 residues ¹⁸⁸Asp-Pro-Leu-Asn-Val-Leu¹⁹³ adopt an unusual cage-like conformation (Fig. 3a, c) which is tethered by residues from both KL domains via hydrogen bonds and hydrophobic contacts deep inside the KL1-KL2 cleft (Fig. 3c). Further downstream, the side chains of Lys-194, Arg-196, and Arg-198 of the FGF23 C-terminal tail dip into the central barrel cavity of KL2, making hydrogen bonds with multiple αKlotho residues (Fig. 3c). At the interface between the FGF23 β-trefoil core and αKlotho, residues from the β5–β6 turn and the αC helix of FGF23 make hydrogen bonds and hydrophobic contacts with residues in the short β7-α7 and β8-α8 loops at the upper rim of the KL2 cavity (Extended Data Fig. 6a, c).

To test the biological relevance of the observed contacts between αKlotho and FGF23 C-terminal tail, we introduced multiple mutations into αKlotho™ and FGF23 in order to disrupt αKlotho-FGF23 binding (Fig. 4c). Consistent with our structure-based predictions, all αKlotho™ mutants showed an impaired ability to support FGF23 signaling (Fig. 4c). The FGF23 mutants also exhibited a reduced ability to signal, regardless of whether soluble or membrane-bound αKlotho served as co-receptor (Fig. 4d). Remarkably, the FGF23^D188A mutant (which eliminates the intramolecular hydrogen bonds that support cage conformation) was totally inactive, underscoring the importance of the cage-like conformation in the tethering of FGF23 to αKlotho. Notably, tethering of this cage-like structure requires a precise alignment of residues from both KL domains deep within the KL1-KL2 cleft (Fig. 3c), implying that their correct apposition is critically important for αKlotho co-receptor activity. These structural observations suggest that the bound Zn²⁺ ion serves as a prosthetic group in αKlotho by minimizing interdomain flexibility and hence promoting co-receptor activity. Consistent with such a role, mutants of membrane-anchored αKlotho™ carrying alanine in place of two, three, or all four Zn²⁺ coordinating amino acids (Fig. 3c) showed a reduced ability to support FGF23 signaling (Fig. 4e). Together with our data on the impact of RBA deletion, these results corroborate the biological relevance of the crystallographically-deduced mode by which αKlotho implements FGF23-FGFR1c proximity and thus confers high binding affinity.

FGF23 signaling is αKlotho and HS-dependent

Both FGF23 and FGFR1c have a measurable (albeit weak) binding affinity for HS. Because HS is ubiquitously expressed, we wondered whether it participates in the apparent αKlotho^ecto-mediated FGF23-FGFR dimerization in our cell-based and in vivo experiments. We therefore analyzed the molecular mass of the ternary complex in the absence and presence of increasing molar equivalents of homogenously sulfated heparin hexasaccharide (HS6). Consistent with our previous observations, in the absence of HS6, the ternary complex migrated as a monomeric species¹⁰ with an apparent molecular mass of 150 kDa, in good agreement with the theoretical value for a 1:1:1 complex (160 kDa) (Fig. 5a). With increasing molar ratios of HS6 to ternary complex, the peak for monomeric ternary complex diminished, while a new peak with a molecular mass of 300 kDa (corresponding to a 2:2:2 FGF23-FGFR1c^ecto-αKlotho^ecto dimer) appeared and increased in prominence. Excess HS6 beyond a 1:1 molar ratio of HS6 to ternary complex did not lead to any further increase in the amount of dimer complex formed, as judged by the integrated area of the dimer complex peak (Fig. 5a). We conclude that HS is required for the dimerization of 1:1:1 FGF23-FGFR1c^ecto-αKlotho^ecto complexes, and that at least a 1:1 molar ratio of HS6 to ternary complex is required for complete dimerization of the complex in solution (Fig. 5a). To further confirm the HS-dependency of dimerization, we introduced mutations into the HS-binding sites of FGFR1c (K160Q/K163Q, FGFR1c^ΔHBS, and K207Q/R209Q, FGFR1c^ΔHBS’) and FGF23 (R140A/R143A; FGF23^ΔHBS). Neither mutating the HS-binding site in FGFR1c nor mutating that site in FGF23 impacted the formation of a monomeric 1:1:1 FGF23-FGFR1c-αKlotho complex in solution, demonstrating that αKlotho-mediated stabilization of the FGF23-FGFR complex is HS-independent. However, ternary complexes containing any of these three mutants failed to dimerize in the presence of HS6 (Fig. 5b).

Reconstitution experiments in the context of BaF3 cells (an FGFR, αKlotho, and HS triple deficient cell line³⁰) showed that both soluble αKlotho^ecto and membrane-bound αKlotho™ required HS to support FGF23-mediated FGFR1c activation in a more physiological context (Fig. 5c). We also examined the impact of the HS-binding site mutations in FGFR1c and FGF23 on FGFR1c activation by FGF23 in BaF3 cells (Fig. 5d). In agreement with our solution binding data, activation by FGF23 of HS-binding site mutants of FGFR1c in BaF3 cells was markedly impaired, regardless of whether soluble or membrane-bound αKlotho served as the co-receptor (Fig. 5d). Similarly, the HS-binding site mutant of FGF23 showed a significantly reduced ability to activate FGFR1c (Fig. 5e). These in vitro and cell-based analyses unequivocally demonstrate that whereas HS fulfills a dual role in paracrine FGF signaling – enhancing 1:1 FGF-FGFR binding and promoting 2:2 FGF-FGFR dimerization – it shares this task with αKlotho in FGF23 signaling. Thus, αKlotho primarily acts to promote 1:1 FGF23-FGFR1c binding, whereas HS induces dimerization of the resulting FGF23-FGFR1c-αKlotho complexes.

Based on the crystallographically-deduced 2:2:2 (PDB ID: 1FQ9)⁴ and 2:2:1 (PDB ID: 1E0O)³¹ paracrine FGF-FGFR-HS dimerization models, two distinct HS-induced 2:2:2 endocrine FGF23-FGFR1c-αKlotho quaternary dimers can be envisioned that differ dramatically in the composition of the dimer interface (Extended Data Fig. 8). Specifically, in the 2:2:2:1 model, there would be no protein-protein contacts between the two 1:1:1 FGF-FGFR-αKlotho protomers (Extended Data Fig. 8a). By contrast, in the 2:2:2:2 model, FGF23 and FGFR from one 1:1:1 FGF-FGFR-αKlotho protomer would interact with the D2 domain of FGFR in the adjacent 1:1:1 FGF-FGFR-αKlotho protomer across a two-fold dimer interface (Extended Data Fig. 8b). Based on the fundamental differences in the composition of the dimer interface between these two models, we introduced mutations into the secondary-receptor-binding site (SRBS) in FGF23 (M149A/N150A/P151A; FGF23^ΔSRBS) and into the corresponding secondary-ligand-binding site (SLBS) in FGFR1c D2 (I203E, FGFR1c^ΔSLBS, and V221D, FGFR1c^ΔSLBS’), both of which are unique to the 2:2:2:2 quaternary dimer model. The direct receptor-receptor binding site in FGFR1c D2 (A171D; FGFR1c^ΔRRBS), another binding site unique to the 2:2:2:2 model, was also mutated (Extended Data Fig. 8b). While all these FGF23 and FGFR1c mutants were able to form ternary complexes with αKlotho^ecto, the ternary complexes containing any of the mutated proteins were impaired in their ability to dimerize in the presence of HS6 in solution (Fig. 5f). Moreover, FGF23^ΔSRBS mutant showed a markedly diminished ability to activate FGFR1c in BaF3 cells (Fig. 5e). The loss-of-function effects of these mutations are consistent with a 2:2:2:2 quaternary dimer model (Extended Data Fig. 8b). Hence, we envision that HS engages the HS-binding sites of FGFR1c and FGF23 in two stabilized 1:1:1 FGF23-FGFR1c-αKlotho ternary complexes to promote the formation of a two-fold symmetric 2:2:2:2 FGF23-FGFR1c-αKlotho-HS dimer (Fig. 5g). In doing so, HS enhances reciprocal interactions of FGFR1c D2 and FGF23 from one ternary complex with FGFR1c D2 in the other ternary complex, thereby buttressing the dimer (Extended Data Fig. 8b). This replicates the role that HS plays in paracrine FGF signaling⁴. In contrast to HS, αKlotho molecules do not directly participate in the dimer interface (Fig. 5g), but rather indirectly support HS-induced dimerization by enhancing 1:1 FGF23-FGFR1c binding affinity. Hence, it appears that FGF23 strikes a fine balance between losing a large amount of HS-binding affinity to enable its endocrine mode of action and retaining sufficient HS-binding affinity to allow HS-mediated dimerization of two 1:1:1 FGF23-FGFR1c-αKlotho complexes. These considerations do not formally exclude the possibility that 2:2:2:2 and 2:2:2:1 quaternary dimers might co-exist as a higher order cluster on the cell surface, as has been proposed for paracrine 2:2:2 and 2:2:1 FGF-FGFR1-HS dimers³².

FGF19 and FGF21, the other two endocrine FGFs, require βKlotho as an obligate co-receptor to bind and activate cognate FGFRs^33,34 so as to mediate effects that regulate, for example, metabolic pathways involved in bile acid biosynthesis or fatty acid oxidation^35,36. Based on the structural analysis and supporting cell-based data shown in Extended Data Fig. 9 and 10, we propose that βKlotho, similar to αKlotho, functions as a non-enzymatic molecular scaffold to promote signaling by these two FGF hormones.

Extended Data

Supplementary Material