Designing receptor agonists with enhanced pharmacokinetics by grafting macrocyclic peptides into fragment crystallizable regions

Short half-lives in circulation and poor transport across the blood–brain barrier limit the utility of cytokines and growth factors acting as receptor agonists. Here we show that surrogate receptor agonists with longer half-lives in circulation and enhanced transport rates across the blood–brain barrier can be generated by genetically inserting macrocyclic peptide pharmacophores into the structural loops of the fragment crystallizable (Fc) region of a human immunoglobulin. We used such ‘lasso-grafting’ approach, which preserves the expression levels of the Fc region and its affinity for the neonatal Fc receptor, to generate Fc-based protein scaffolds with macrocyclic peptides binding to the receptor tyrosine protein kinase Met. The Met agonists dimerized Met, inducing biological responses that were similar to those induced by its natural ligand. Moreover, lasso-grafting of the Fc region of the mouse anti-transferrin-receptor antibody with Met-binding macrocyclic peptides enhanced the accumulation of the resulting Met agonists in brain parenchyma in mice. Lasso-grafting may allow for designer protein therapeutics with enhanced stability and pharmacokinetics.

Short half-lives in circulation and poor transport across the blood-brain barrier limit the utility of cytokines and growth factors acting as receptor agonists. Here we show that surrogate receptor agonists with longer half-lives in circulation and enhanced transport rates across the bloodbrain barrier can be generated by genetically inserting macrocyclic peptide pharmacophores into the structural loops of the fragment crystallizable (Fc) region of a human immunoglobulin. We used such 'lasso-grafting' approach, which preserves the expression levels of the Fc region and its affinity for the neonatal Fc receptor, to generate Fc-based protein scaffolds with macrocyclic peptides binding to the receptor tyrosine protein kinase Met. The Met agonists dimerized Met, inducing biological responses that were similar to those induced by its natural ligand. Moreover, lasso-grafting of the Fc region of the mouse anti-transferrin-receptor antibody with Met-binding macrocyclic peptides enhanced the accumulation of the resulting Met agonists in brain parenchyma in mice. Lasso-grafting may allow for designer protein therapeutics with enhanced stability a nd p ha rmacokinetics.
The clinical use of cytokines and growth factors as therapeutics has been approved by the US Food and Drug Administration 1,2 , and their potential application in emerging areas, such as neurogenesis and brain repair 3,4 , is a topic of intense research. However, their inherent structural properties render them challenging to engineer for improved physicochemical stability and pharmacokinetics, in particular, for better half-life or blood-brain barrier (BBB) penetrance. Existing methods that control the pharmacokinetics of protein therapeutics include poly(ethylene glycol) conjugation and fusion with the fragment crystallizable (Fc) region of immunoglobulin [5][6][7] to extend their half-lives; and fusion with Fab of the anti-transferrin receptor (TfR) antibody to enhance their penetration across the BBB [8][9][10] . However, the Article https://doi.org/10.1038/s41551-022-00955-6 (EC 50 ) comparable to that of ungrafted aMD4-dimer peptide (EC 50 : 4.5 ± 0.2 nM vs 5.1 ± 0.6 nM, respectively), and a slightly reduced maximum activation relative to HGF-induced phosphorylation (74.8% ± 0.4% vs 101.3% ± 0.1%, respectively). In contrast, grafting aMD4 into other loops resulted in lower Met activation relative to aMD4-dimer peptide. In the case of aMD5, B1 was the best grafting site, where Fc(aMD5)B1 activated Met with a 3.7-fold higher EC 50 compared with ungrafted aMD5-dimer peptide, and a slightly lower maximum activation (EC 50 : 16.6 ± 1.6 nM vs 4.5 ± 0.2 nM, respectively; maximum activation: 76.2% ± 1.4% vs 105.0% ± 4.2%, respectively; Fig. 1c). Grafting aMD5 into other loops either reduced or abolished Met activation. Notably, the agonistic activities of lasso-grafted Fc correlated well with their binding strengths to cell surface Met ( Supplementary Fig. 1a) or Met ectodomain fragment (Met ECD ) (Extended Data Fig. 2 and Supplementary Fig. 1b), indicating that their efficacies are determined largely by their affinity to Met. The difference in the graft-site dependency of the two peptides is probably due to the steric effects imposed by the difference in their binding site on the Met ectodomain and their divergent spatial arrangements in the Fc structure ( Supplementary Fig. 2).
To improve agonistic activity, we appended a Cys residue at each end of the aMD4 sequence to promote a tightly closed macrocyclic conformation via disulfide linkage ( Fig. 2a and Supplementary Fig. 3). The disulfide-bonded derivatives aMD4 (hereafter prefixed with 'ds') at T1 and B3 showed greater maximum Met activation than the control (T1, P = 0.039; B3, P < 0.0001), while no effects were observed in the EC 50 (Supplementary Fig. 3f) and dissociation constant (K D ) values of Met ECD binding (Extended Data Fig. 2). In contrast, dsB2 showed no improvement in agonistic activity, maximum activation or EC 50 values ( Supplementary Fig. 3f), suggesting that the reduced Met affinity after grafting into these two sites was not due to suboptimal conformation of the grafted peptide motif per se. This was further supported by the similar conformations of aMD4 or aMD5 when grafted at the B1 or B3 position ( Supplementary Fig. 2).
The cellular responses induced by Fc(aMD4)B3 in EHMES-1 cells or in hepatocytes were highly comparable to those of HGF ( Fig. 2 and Extended Data Fig. 3). Fc(aMD4)B3 at 1 nM induced phosphorylation of Met at Y1234/1235 and activated subsequent downstream signalling with phosphorylation of Akt and Erk1/2 at levels comparable to those induced by HGF ( Fig. 2b and Supplementary Fig. 4). The levels of phosphorylation induced by Fc(aMD4)B3 and HGF at three different tyrosine residues of Met, as well as the induction kinetics of Met, Akt and Erk1/2 phosphorylation, were comparable (Extended Data Fig. 3). The selectivity of Fc(aMD4)B3 for Met was confirmed by a survey of tyrosine phosphorylation of 49 receptors (Fig. 2c and Supplementary Fig. 5). The changes in gene expression induced by Fc(aMD4) B3 or HGF were examined in primary human hepatocyte spheroid culture by RNA-seq ( Fig. 2d-g). A scatterplot shows the correlation of gene expression changes induced by HGF and Fc(aMD4)B3 relative to untreated spheroids (Fig. 2e). Lastly, a heat map of differentially expressed genes (DEGs) relative to untreated spheroids indicates that both Fc(aMD4)B3 and HGF similarly altered the expression of more than 2,000 transcripts (Fig. 2f), with significant enrichment of gene ontology (GO) terms related to wound healing and liver functions (Fig. 2g).

Mechanism of Met dimerization by Fc(aMD4)B3
As peptides generated de novo could differentially bind to the target receptors in a way completely different from that of the native ligands, we could theoretically generate agonists that do not compete with native ligands or are capable of activating mutated receptors that lack ligand binding. To investigate this potential, we first mapped the putative Fc(aMD4)B3 binding site (shown in red in Fig. 3a) using chimeric Met ECD fragments that variably fused human and mouse Met ECD (Extended Data Fig. 4), exploiting the specificity of aMD4 for human Met 40 . The putative binding site is mapped to the lower face of the extent to which these methods can improve the pharmacokinetics of the proteins without adversely affecting their bioactivities is dependent on the properties of each protein [5][6][7][8] . To overcome the inherent structural limitations of cytokines and growth factors, there has been progress in the development of surrogate agonists that are structurally unrelated to native ligands 11,12 . Despite these recent advances, there remains a large unmet need for more robust and versatile methods to design protein therapeutics with desired physicochemical stability and pharmacokinetics.
Macrocyclic peptides have emerged as a promising novel class of drug candidates boasting a number of desirable features, such as antibody-like binding affinity and specificity 13,14 , the ability to target unique chemical spaces [15][16][17][18] and efficient discovery through both rational/computational design and in vitro display [18][19][20][21] . In general, macrocyclic peptides display greater affinities for targets than linear peptides due to their constrained cyclic structures. An intriguing possibility in extending the applicability of these macrocyclic peptides is to 'engraft' them onto protein scaffolds to allow for functional combinations of peptide and protein 18,[22][23][24][25][26] . However, grafting de novo identified peptides to protein loops has been challenging due to the potential misfolding of both the grafted peptide and the host protein 26 .
The development of the RaPID (random non-standard peptides integrated discovery) system, which integrates messenger RNA display and genetic code reprogramming, has enabled the discovery of thioether-based cyclized macrocyclic peptides with exquisite binding specificity against target proteins [16][17][18] . In previous studies, we have shown that RaPID-derived pharmacophore sequences can be readily implanted onto surface-exposed loops of proteins and maintain both functions of the guest peptide and the host protein, which we termed 'lasso-grafting' [27][28][29] . The observed exceptional grafting compatibility is probably due to the intrinsic property of the pharmacophore motifs to self-fold into target-binding conformation similar to the parental macrocycle, even in the context of unrelated loop structure of the scaffold proteins 18,30 .
In this study, we leverage the desirable scaffolding properties of the Fc fragment, its long half-life, versatility in combination with Fab 5,6,31,32 and ease of production to show the feasibility and application of lasso-grafting. Using the Fc as a scaffold, we generated Met receptor agonists that are characterized by markedly improved half-life and BBB penetrance.

Met agonists generated by lasso-grafting Fc
Met is a receptor tyrosine kinase activated upon dimerization triggered by its ligand, hepatocyte growth factor (HGF). Met-HGF is centrally involved in morphogenesis during development, wound repair and organ homoeostasis by stimulating cell growth, survival and migration 33,34 . Recombinant HGF protein displays therapeutic efficacy in preclinical models 34,35 and in patients with specific diseases 36,37 . However, its short half-life and inability to be delivered beyond the BBB have long limited its potential in medical applications 38,39 . To address these shortcomings, we inserted by lasso-grafting the linearized pharmacophore sequence of aMD4 or aMD5, two macrocyclic peptides that bind to the ectodomain of Met 40 into each of the eight loops of Fc (Fig. 1a). These two series of aMD4-or aMD5-grafted Fc fragments (hereafter Fc(aMD4) and Fc(aMD5), respectively), each displaying two Met binders within 31-42 Å of each other, have the potential to proximate two Met receptors on the cell surface. The expression levels of Fc(aMD4) and Fc(aMD5) in Expi293F cells were similar to that of the control Fc, with the exception of Fc(aMD5)T3 (Extended Data Fig. 1), suggesting that lasso-grafting generally preserved the high expression of Fc.
The activation of cellular Met by purified Fc(aMD4) or Fc(aMD5) was highly dependent on the grafting site (Fig. 1b,c), where B3 was the optimal site for aMD4 (Fig. 1b). Our analyses show that Fc(aMD4) B3 activated Met with an half maximal effective concentration Article https://doi.org/10.1038/s41551-022-00955-6 Sema domain of Met, which does not overlap with the HGF binding site 41 , hence explaining why aMD4 peptide does not compete with HGF-induced Met activation 40 .
We next examined the structure of the Met dimer induced by Fc(aMD4). Complexes between Met ECD and Fc(aMD4) were stabilized by crosslinking using bis(sulfosuccinimidyl)suberate (BS 3 ) and analysed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) ( Fig. 3b and Extended Data Fig. 5a,b). The results indicated the formation of complexes between Met ECD and Fc(aMD4) in 1:1 or 2:1 stoichiometry. The relative efficiency of various Fc(aMD4) in forming a 2:1 signalling complex was in the order of B3 > B1 > B2, which correlated well with their ability to activate cellular Met. The BS 3 -crosslinked 2:1 complex of Met ECD and Fc(aMD4)B3 was purified by size exclusion chromatography (Extended Data Fig. 5c) and subjected to single-molecule examination by high-speed atomic force microscopy (HS-AFM) 42 , with Fc and Met ECD as controls (  Video 4). The flexibility of IPT domains probably allows for the proper association of the intracellular kinase domains of two Met molecules, this association being essential for Met activation 33,34 . These results show that while Fc(aMD4)B3 binds to Met on a site different from that in HGF, its recruitment of two Met receptors in close proximity enables it to activate Met to the same extent as HGF.

FcRn binding, half-life and in vivo activity
The long half-lives of Fc and antibodies are mainly maintained through their binding to the FcRn [43][44][45][46] . As such, we have selectively lasso-grafted into loop locations that are distal from the FcRn binding site (Fig. 4a). Accordingly, our analysis of binding kinetics between Fc(aMD4) and immobilized FcRn showed that lasso-grafting into most loops did not affect the affinity of Fc to FcRn, although Fc(aMD4)T2 and M2 exhibited slightly higher K D values than control Fc ( Fig. 4b and Extended Data Fig. 6). This suggests that lasso-grafting on Fc preserves the affinity of Fc to FcRn.
We next evaluated the serum half-life of Fc(aMD4)B3 in comparison to those of control Fc and HGF in wild-type mice after a single intravenous (i.v.) administration at equimolar amount per kg (Fig. 4c). The serum concentration of HGF decreased to below 0.01 nM within 1 h, which was unable to activate Met. In contrast, Fc(aMD4)B3 showed an extended half-life comparable to that of control Fc (Fig. 4c). To accurately determine the pharmacokinetics of our human Fc-based variants, we measured the serum half-lives of Fc(aMD4)B3 and control Fc in mFcRn −/− hFcRn + mice 47 (Fig. 4d). The serum half-life of Fc(aMD4)B3 in these mice was 49.4 h and comparable to that of control Fc (46.6 h). The serum concentration of Fc(aMD4)B3 remained above 1 nM, the minimum concentration for activating Met (Fig. 2a,b), for up to 200 h after a single i.v. administration at 0.74 mg kg −1 (Fig. 4d).   Article https://doi.org/10.1038/s41551-022-00955-6 The in vivo activity of Fc(aMD4)B3 was next examined using chimeric mice transplanted with human hepatocytes with approximately 80% of hepatocytes being humanized (PXB-mice) 48 , as Fc(aMD4)B3 is incapable of activating murine Met 40 . We first confirmed the extended serum half-life of Fc(aMD4)B3 in PXB-mice after a single i.v. or subcutaneous (s.c.) injection at 5 mg kg −1 ( Supplementary Fig. 6). Subsequently, the proliferation and gene expression of human hepatocytes 46 h after a single s.c. injection of Fc(aMD4)B3 or PBS in PXB-mice ( Fig. 4e-j) were analysed. Fc(aMD4)B3 significantly increased replicative DNA synthesis of hepatocytes compared with those of control mice treated with PBS (3.73% ± 0.31% vs 0.37% ± 0.07%, respectively, P < 0.0001) (Fig. 4f,g). A heat map of DEGs between Fc(aMD4)B3-treated and PBS-treated livers indicated that Fc(aMD4)B3 altered the expression of more than 3,200 transcripts (Fig. 4h). Significantly enriched GO terms among them were primarily involved in mitotic DNA replication/cell cycle and metabolic processes (Fig. 4i,j). Taken together, these results show that lasso-grafting generated a Met-activating Fc that is bioactive in vivo, with a markedly improved half-life that is comparable to that of unmodified Fc.

Met agonists with BBB penetration
It has been reported that HGF-induced Met activation exerts beneficial neuroprotective effects in preclinical models of cerebral ischaemia, spinal cord injuries and neurological pathologies, such as Alzheimer's disease, amyotrophic lateral sclerosis and multiple sclerosis 35 . In line with earlier studies, we confirmed that Met is expressed in neurons in mouse brains ( Supplementary Fig. 7). To generate Met agonists capable of crossing the BBB, we applied lasso-grafting of aMD4 on the Fc of the anti-mouse TfR (mTfR) antibody 49 in either single-or double-Fab configuration (sTfR(aMD4) and dTfR(aMD4), respectively) (Fig. 5a,b and Supplementary Fig. 8). Lasso-grafting did not significantly alter the previously reported 9,10 affinity of these antibodies to mTfR, suggesting that Fab affinity is largely preserved (Fig. 5c). In contrast, both sTfR(aMD4) and dTfR(aMD4) showed reduced Met activation compared with Fc(aMD4), with 9.6-and 18.7-fold reduction in EC 50 values (20.2 ± 3.4 nM vs 39.2 ± 22.3 nM vs 2.1 ± 0.2 nM, respectively) (Fig. 5d). This was possibly due to the presence of large Fab arm(s) on sTfR(aMD4) and dTfR(aMD4), which may have interfered with their binding and/or dimerization of Met on the cell surface.
Lastly, we examined the BBB penetration of the three aMD4-grafted variants in wild-type mice after a single i.v. administration at therapeutically relevant doses of the proteins at equimolar amount per kg ( Fig. 5e-j). Their concentrations in PBS-perfused brains or serum at 24 h post i.v. injection were determined by ELISA for human IgG Fc. Of the three variants, sTfR(aMD4) showed the highest percentage injected dose per gram brain (0.96% ± 0.07%), brain-to-serum ratio (8.3% ± 0.7%), as well as the highest brain concentration (10.7 ± 1.2 nM) ( Fig. 5f-i), which is sufficient for Met activation (Fig. 5d). These results are in line with previous reports of low-affinity anti-TfR antibody accumulation in the brain 9,10,50 . In comparison, these values were significantly lower for dTfR(aMD4) and were nearly negligible in Fc(aMD4) (Fig. 5f-i). Concordant with the higher brain concentration determined by ELISA, sTfR(aMD4) showed prominent parenchymal staining co-localized with NeuN, GFAP and Iba1, indicating its distribution to neuron, astrocytes and microglia, respectively ( Fig. 5j(middle right) and Supplementary  Fig. 9). In comparison, dTfR(aMD4) showed less extensive neuronal staining and more endothelial staining (Fig. 5j(right)), in line with previous reports of high-affinity anti-TfR antibody accumulation in brain endothelial cells 9,10 . Lastly, Fc (aMD4)  Article https://doi.org/10.1038/s41551-022-00955-6 brain parenchyma (Fig. 5j(middle left)). Further evaluation of detailed pharmacokinetic profiles, including time course, will be needed for future therapeutic evaluation of sTfR(aMD4) in preclinical models. Taken together, these observations showed that we have generated BBB-penetrating Met agonists by lasso-grafting of Met-binding macrocyclic peptide on the Fc of anti-mTfR antibody.

Discussion
The grafting of de novo identified peptides to improve their stability and bioavailability has been attempted on a variety of scaffolds [23][24][25][26] . These past studies revealed that successful grafting is highly dependent on the combination and compatibility between grafted peptides and scaffold proteins 25,26 . As such, the selection of scaffold protein     e-j, BBB penetration of aMD4-grafted anti-TfR antibodies. Schematic (e), brain concentration (f), percentage injected dose per gram of brain (g), serum concentration (h) and brain-to-serum ratio (i) at 24 h after a single injection of Fc(aMD4), sTfR(aMD4) or dTfR(aMD4) at equimolar amount per kg (12, 20 and 26.5 mg kg −1 , respectively) via the tail vein. Results are shown as mean ± s.e.m. (n = 4 mice per group; unpaired two-tailed t-test). j, Representative images of immunohistochemical staining of brain sections from mice at 24 h after a single injection of Fc(aMD4), sTfR(aMD4) or dTfR(aMD4) via the tail vein. Note the broad staining for sTfR(aMD4) in the brain parenchyma and localization around NeuN-positive neuronal cell bodies. Staining for dTfR(aMD4) was more prominent in endothelial cells than in neuronal cell bodies. Scale bars, 50 µm. is a key consideration, as it determines the success of peptide grafting while conferring desirable properties such as metabolic stability and cell permeability to the engineered products [24][25][26] . To cater for this broad spectrum of desirable features, disulfide-stabilized peptides and miniproteins have been engineered to act as specialized scaffolds for peptide grafting [23][24][25][26] . However, these specialized scaffolds are small non-human proteins with no particular bioactivities of their own. Although natural human proteins are desirable scaffolds for peptide grafting, there has been very limited attempts at grafting de novo identified peptides onto natural human scaffolds so far. A major reason behind this is that grafting synthetic de novo identified peptides into an independently folded domain of a natural scaffold would often result in the misfolding and inactivation of both entities. It is therefore remarkable that our data thus far indicate that the functional grafting of de novo identified macrocyclic peptides into the protein loops of natural scaffolds is far more feasible than previously expected [27][28][29] . As a protein scaffold, the Fc is highly desirable because of its high expression, ease of manufacture, long half-life, and Fab compatibility 5,6,31,32 . In this study, we report that lasso grafting Fc with macrocyclic peptides yielded Met receptor agonists with extended half-life and BBB penetration. Our application of lasso-grafting on Fc has revealed three key strengths to this approach: Firstly, lasso-grafting preserved the overall biochemical characteristics of Fc in terms of expression level, FcRn affinity, and Fab compatibility. Secondly, lasso-grafted macrocyclic peptide largely preserved parental macrocyclic conformation even without disulfide linkage. Thirdly, there is a wide choice of suitable insertion positions in Fc. Our study further showed that the target affinity of grafted peptides is site-dependent. For example, the T-loops were less effective in our study, probably due to the presence of a N-terminus hinge region that could interfere with the binding of grafted peptides to the target protein, Met. Yet, these loops acted as suitable grafting positions for other receptor binders 27,29 . Notwithstanding this, the robust graft compatibility between macrocyclic peptides and Fc, in combination with a wide choice of insertion positions, renders lasso-grafting a particularly effective approach of Fc engineering.
An important implication of our findings is that lasso-grafting protein scaffolds with specific properties can serve as a framework for engineering designer receptor agonists with the desired properties based on the choice of protein scaffolds. As a demonstration of this, we generated BBB-permeable agonists by lasso grafting an anti-TfR antibody, thereby leveraging an established BBB-permeation technique [8][9][10] . At present, there are several efforts in developing anti-TfR Fab fusion antibodies with blocking properties, but only few that attempt to develop agonist antibodies 51 . Lasso-grafting could be used to design agonist BBB-permeating antibodies, in addition to antagonistic ones. Furthermore, one major drawback of current BBB-permeating antibodies is that, owing to their low brain-to-serum ratio and long half-life in circulation, systemic concentrations often exceed the desired range for extended periods. This could cause unwanted neo-immunogenicity or side effects, especially if the targeted molecule is functionally important in other tissues. With lasso-grafting, this could be avoided by grafting functional peptides directly into scaffold proteins with a short half-life, such as the anti-TfR Fab.
Contrary to existing methods that improve the pharmacokinetics of protein therapeutics by modifying the protein, lasso-grafting endows novel functions derived from small peptides to well-understood protein scaffolds [27][28][29] , hence avoiding potential pitfalls such as reduced biological activity, unfavourable protein characteristics for manufacturing or neo-immunogenicity arising from protein fusion or from poly(ethylene glycol) [5][6][7] . Although Fc serves as an excellent scaffold because of its immunosuppressive property 52 , the immunogenicity, safety and efficacy of lasso-grafted proteins await future evaluation as part of drug development. In the same vein, HGF has shown therapeutic efficacies in preclinical models of various diseases such as non-alcoholic steatohepatitis 53,54 or neurological pathologies 35,37 . The therapeutic advantages of the lasso-grafted Met agonists generated in this study should likewise be further evaluated in disease models.
In view of the rapid advances in potent bioactive and graft-compatible macrocyclic peptides, lasso-grafting holds great promise for the generation of designer protein therapeutics with the desired physicochemical stability and pharmacokinetics, as well as for engineering multifunction proteins and antibodies.

Protein design, construction and purification
To design peptide-grafted Fc protein, Fc protein was structurally inspected to find an appropriate insertion point where the approximate distance between the two anchor point residues located within an exposed loop was less than 7 Å. For the construction of control or Met-binding macrocyclic peptide (aMD4 or aMD5)-grafted Fc, human IgG1 Fc (residues 104-330, Uni-Prot P01857) was used without any tags. The internal sequence of aMD4 or aMD5 appended with or without Cys and two spacer residues of Gly at both ends were inserted into these identified sites (T1 to B3). For the construction of anti-mouse TfR antibodies with aMD4ds insertion at the B3 site, the Fab fragment of a rat anti-mouse TfR monoclonal antibody (clone 8D3) 49 was used. The VH to CH1 fragment of 8D3 was fused to the N terminus of Fc(aMD4) to yield heavy chain (HC) of dTfR(aMD4). The VL to CL fragment of 8D3 was used to yield light chain (LC). For the construction of sTfR(aMD4), a flag tag was added at the N terminus of Fc(aMD4). The amino acid sequences used for s/dTfR(aMD4) can be found in Supplementary  Fig. 8. All expression constructs were made using pcDNA3.1-based backbone with appropriate signal peptide and the coding region of all expression constructs was verified by DNA sequencing. Protein expressions were performed using the Expi293 expression system (Thermo Fisher). Secreted proteins were purified on a HiTrap Protein A HP column (Cytiva) using elution with 0.1 M citrate acid-sodium citrate buffer (pH 3.5), followed by neutralization, dialysis against phosphate-buffered saline (PBS) and size exclusion chromatography on a Superdex 200 Increase 10/300GL column (Cytiva) equilibrated with PBS. SDS-PAGE analysis and purity of purified Fc variants used in the assay can be found in Supplementary Fig. 10. For the expression and purification of sTfR(aMD4), three DNAs (TfR(aMD4) HC, TfR(aMD4) LC and Fc(aMD4)) were used to co-transfect Expi293F cells, yielding the mixture of sTfR(aMD4), dTfR(aMD4) and Fc(aMD4). The sTfR(aMD4) was purified from the latter two components by sequential anti-Flag affinity purification and size exclusion chromatography on a Superdex 200 Increase 10/300GL column equilibrated with PBS.

Flow cytometry
The binding of peptide-grafted Fc proteins to Met-knockout CHO cells and Met-reconstituted CHO cells 27 was detected using flow cytometry. Article https://doi.org/10.1038/s41551-022-00955-6 Cells were detached from dishes by brief treatment with 0.25% trypsin and 1 mM ethylenediaminetetraacetic acid, plated at 200,000 cells per well and incubated with peptide-grafted Fc proteins diluted at 10 mg ml −1 in 100 ml Ham's F-12 medium containing 5% FBS on ice for 1.5 h. After washing twice with ice-cold PBS, cells were incubated with Alexa Fluor 488-labelled goat anti-human IgG (1:400 dilution in 100 µl Ham's F-12 medium containing 5% FBS, Thermo Fisher, A11013) on ice for 30 min, then analysed on an EC800 system (Sony). Gate on forward scatter vs side scatter was set to include all cell populations but excluding debris and dead cells as shown in Supplementary Fig. 11. The data were analysed with FlowJo software (Tomy Digital Biology).

Surface plasmon resonance analysis
The ectodomain fragment of human Met (Met ECD , 1-931) was C-terminally fused with a biotin acceptor sequence (SSLRQILDSQK-MEWRSNAGG) and co-expressed with a biotin ligase (BirA) in Expi293F cells to achieve BirA-mediated biosynthetic biotinylation and immobilized onto a Series S Biotin CAPture chip (Cytiva) at a surface density of ~210 resonance units (RU). The binding of Fc and peptide-grafted Fcs was evaluated by injecting the analyte solutions serially diluted using the running buffer (20 mM HEPES-NaOH (pH 7.5), 150 mM NaCl, 0.05% Tween 20). The runs were conducted in single cycle kinetics mode employing the following parameters: flow rate of 30 µl min −1 , contact time of 120 s and dissociation time of 300 s. After each run, the surface was regenerated by injecting the regeneration buffer (6 M guanidine-HCl, 250 mM NaOH) until the response returned to the original baseline level. The binding curves were obtained by subtracting the RU of the reference cell (un-immobilized) from that of the measurement cell (immobilized with Met ECD ) and used to derive kinetic binding values. Data were obtained using a Biacore T200 instrument (Cytiva) at 25 °C, and the results were analysed using Biacore T200 evaluation software v3.2 (Cytiva). Biotinylated human FcRn/β2M (ACROBiosystems) was captured on the surface of streptavidin-coated chips (Cytiva) at 180~220 RU. Analytes were diluted in running buffer (50 mM phosphate (pH 6.0), 150 mM NaCl, 0.01% Tween 20) and injected at 30 µl min −1 for 1 min, followed by dissociation for 0.7 min. The binding curves were obtained by subtracting the RU of the reference cell (un-immobilized) from that of the measurement cell (immobilized with FcRn/2M) and following axis-zeroing, sensorgrams were fit locally to a 1:1 Langmuir binding model. Data were obtained using a Biacore 3000 instrument (Cytiva) at 25 °C, and the results were analysed using BIAevaluation software v4.1 (Cytiva).

Predicted structures of the CH3 domain of Fc with aMD4 or aMD5 insertion
Structures of the CH3 domain of Fc with aMD4 or aMD5 inserted in the B1 or B3 loop were predicted using ColabFold and AlphaFold2 in MMseqs2 55 .

Trypsin digestion and liquid chromatography-mass spectrometry (LC-MS/MS)
Trypsin digestion and LC-MS/MS analysis were performed by KANEKA TECHNO RESEARCH. One hundred µg of Fc(aMD4ds)T1 or Fc(aMD4ds)B3 in PBS were treated with 100 mM ammonium bicarbonate (Sigma-Aldrich) containing 8 M urea (Fujifilm Wako Pure Chemicals). After desalting using an Amicon Ultra device (MWCO: 10 K, Millipore), 10 µg of each protein was digested with trypsin (Promega) for 12 h at 37 °C under non-reducing conditions. The reaction was stopped by the addition of formic acid. The samples were injected into a Nexera X2 UHPLC system (Shimadzu corporation) connected to a SCIEX TripleTOF 6600 system (AB SCIEX) equipped with an ESI ionization source. Peptides were chromatographically separated on a C18 reverse-phase column with linear-gradient of mobile phase A (water containing 0.1% formic acid) and mobile phase B (acetonitrile (Fujifilm Wako Pure Chemicals) containing 0.1% formic acid). MS and MS/MS data were collected using IDA, positive ion mode. The LC-MS raw data were processed using SCIEX BioPharmaView software. The following criteria were used to identify peptides: mass accuracy for the matched precursor must be within 5 ppm of mass error, and peptide matching score was set to at least 3 for peptide identification.

Crosslink and isolation of Fc(aMD4)/Met ECD complex
Met ECD (60 nM) and Fc(aMD4)B1/B2/B3 (20 nM) were mixed in 100 µl PBS with 0.01% Triton X-100 for 1 h. The samples were divided into two and treated with 1 mM BS 3 (Thermo Fisher) or left untreated for 1 h, quenched by 50 mM Tris, subjected to SDS-PAGE (7.5% or 3-10% gel) and silver stained. For the HS-AFM observation, Met ECD (2.6 µM) and Fc(aMD4)B3 (1.3 µM) were mixed in 100 µl PBS with 0.01% Triton X-100 for 1 h, treated with 1 mM BS 3 for 1 h, quenched by 50 mM Tris and analysed by size exclusion chromatography on a Superose 6 Increase 10/300GL column (Cytiva) equilibrated with PBS. The separated fractions were analysed by SDS-PAGE and stained with Coomassie brilliant blue. A single fraction was subjected to HS-AFM analysis.

HS-AFM observations
HS-AFM measurements were operated in tapping mode. Cantilever deflection was detected with an optical beam deflection detector using an infrared (IR) laser (0.7 mW, 780 nm). The IR laser beam was focused onto the backside of a cantilever covered with gold film (Olympus, BL-AC10DS-A2) through a ×60 objective lens (CFI S Plan Fluor ELWD ×60; Nikon). Reflection of the IR laser from the cantilever was detected using a two-segmented PIN photodiode (MPR-1; Graviton). The free-oscillation amplitude was approximately 1 nm and the set-point amplitude was approximately 90% of the free amplitude for feedback control of HS-AFM. For the AFM probe, an amorphous carbon tip with a length of approximately 500 nm grown by electron beam deposition was used. For the AFM substrate, a mica surface treated with 0.01% 3-aminopropyl-triethoxysilane (Shin-Etsu Silicone) was used. All HS-AFM observations were performed in a buffer solution containing 50 mM Tris-HCl (pH 7.4) and 100 mM NaCl at room temperature.

RNA-seq
Total RNA of hepatocyte spheroids or RNAlater (Thermo Fisher)-treated livers from PXB-mice was prepared using the QIAzol lysis reagent (Qiagen) and shipped to the Bioengineering Lab (Kanagawa, Japan) for RNA quality control and RNA-seq on a DNBSEQ-G400 sequencer (MGI Tech). Complementary DNA libraries for RNA-seq were prepared using MGIEasy RNA directional library prep set (MGI Tech) and evaluated using AATI fragment analyzer (Advanced Analytical Technologies). The cDNA libraries were circularized using MGIEasy circularization kit (MGI Tech), prepared as DNA nano balls by DNBSEQ-G400RS high-throughput sequencing kit (MGI Tech), and followed by massively parallel sequencing (2×100 bp) on a DNBSEQ-G400 sequencer (MGI Tech). After excluding adaptor sequences using cutadapt (ver. 1.9.1) and sequences with low-quality score (<20) or less than 40 nucleotides using sickle (ver 1.33), reads were aligned to the reference human genome (GRCh38.p13) using the hisat2 software (v2.2.0). The alignment data were counted using featureCounts (ver. 2.0.0). Transcript abundances were measured in reads per kilobase of exon per million mapped reads (RPKM) and transcripts per million (TPM). DEG analysis was performed using DEGES in TCC (v1.18.0) and DESeq (v1.30.0) (FDR < 0.05 or <0.01, Benjamini-Hochberg, two-sided). All DEGs identified were subjected to cluster analysis using Gene Cluster 3.0. Significantly enriched GO terms were identified using GO Enrichment Analysis by PANTHER.

Half-life in serum
Half-life in serum up to 9 h was determined in wild-type C57BL/6 mice (female, 9 weeks, 18-21 g, obtained from Japan SLC) given a single tail-vein injection at 0.7 mg kg −1 for Fc, 0.74 mg kg −1 for Fc(aMD4) B3 and 1.0 mg kg −1 for human recombinant HGF (Kringle Pharma), diluted in 100 µl PBS. Half-life in serum up to 12 d was determined in mFcRn −/− hFcRn Tg 276 heterozygote mice (female, 9-12 weeks, 18-24 g) produced in-house by mating mFcRn −/− mice and hFcRn Tg 276 homozygote mice obtained from Jackson Laboratory. Mice were given a single tail-vein injection at 0.7 mg kg −1 for Fc or 0.74 mg kg −1 for Fc(aMD4)B3 diluted in 100 µl PBS. Half-life in serum up to 12 d was also determined in chimeric mice with humanized liver (PXB-mice, PhoenixBio, male, >12 weeks, 19-24 g) given a single tail-vein injection or subcutaneous injection at 5.0 mg kg −1 for Fc(aMD4)B3, diluted in 100 µl PBS. Blood was drawn from the submandibular vein using a Goldenrod animal lancet (Bio Research Center) or from the orbital plexus at each time point, processed to serum and stored at −80 °C until analysis. The serum concentrations of Fc and Fc(aMD4)B3 were determined using a human immunoglobulin recognition immunoassay (human IgG ELISA quantitation set, Bethyl Laboratories). The serum concentrations of HGF were determined using an in-house developed ELISA. All animal experimental procedures were conducted in accordance with the guidelines provided by the Proper Conduct of Animal Experiments (1 June 2006, Science Council of Japan). The procedures were approved by the Institutional Review Board of Kanazawa University or the Laboratory Animal Ethics Committee of PhoenixBio.

BrdU labelling in chimeric mice with humanized liver
Chimeric mice with humanized livers (PXB-mice, PhoenixBio) were generated from urokinase-type plasminogen activator-cDNA/ severe combined immunodeficiency mice transplanted with human hepatocytes, with approximately 80% of hepatocytes being humanized 48 . PXB-mice (male, >12 weeks, 15-21 g) were given a single subcutaneous injection at 5 mg kg −1 for Fc(aMD4)B3 in 200 µl PBS or control PBS. Forty-four hours after administration of Fc(aMD4)B3 or control PBS, 5-bromo-2'-deoxyuridine (BrdU) (Sigma Chemical) was intraperitoneally injected into chimeric mice at a dose of 100 mg kg −1 for 2 h before killing. Livers isolated from PXB-mice were treated with RNAlater for RNA-seq or prepared for BrdU staining. Formalin-fixed paraffin sections of chimeric livers were incubated with a rat anti-BrdU antibody at 3 µg ml −1 (Abcam, ab6326), followed by an Alexa Fluor 488-conjugated anti-rat IgG goat polyclonal antibody at 1 µg ml −1 (Abcam, ab150157). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI, Thermo Fisher) at 300 nM. Fluorescent images were obtained using Keyence BZ-X810. The number of BrdU-positive nuclei was determined by counting at least 15,000 cells in 20 randomly selected visual fields in sections from the lateral left lobe and medial right lobe per animal. The procedures involving humanized liver tissues were approved by the Utilization of Human Tissue Ethical Committee of PhoenixBio. All animal experimental procedures were conducted in accordance with the guidelines provided by the Proper Conduct of Animal Experiments (1 June 2006, Science Council of Japan). The procedures were approved by the Laboratory Animal Ethics Committee of PhoenixBio.

BBB penetration
Wild-type C57BL/6 mice (female, 7-8 weeks, 17-20 g, obtained from SLC) were given a single tail-vein injection at equimolar amount per kg: 26.5 mg kg −1 for dTfR(aMD4), 20.0 mg kg −1 for sTfR(aMD4) or 12.0 mg kg −1 for Fc(aMD4), diluted in 200 µl PBS. Twenty-four hours after injection, mice were anaesthetized and blood was drawn from the right ventricle and processed to serum. Brains were isolated from mice after perfusion with PBS at a rate of 2 ml min −1 for 8 min. Each half of the brain was homogenized in 1% NP-40 (Calbiochem) in PBS containing CompleteMini EDTA-free protease inhibitor cocktail tablets (Roche Diagnostics), rotated at 4 °C for 1 h, spun at 14,000 r.p.m. for 20 min, and supernatants collected. The serum and the brain lysates were subjected to antibody measurement using a human immunoglobulin recognition immunoassay (human IgG ELISA quantitation set, Bethyl Laboratories). To calculate the concentrations in the brain, ng g −1 brain was considered to be ng ml −1 . Each half of the brain was embedded in optimal cutting temperature compound (Sakura Finetek Japan), frozen in liquid nitrogen and sectioned at 10 µm thickness. After fixation in 4% paraformaldehyde in PBS for 30 min, the sections were blocked in 5% BSA, 0.3% Triton X-100 in PBS and stained with a goat anti-human IgG Fc antibody (1:500 dilution, Bethyl Laboratories) and a mouse anti-mouse NeuN antibody (1:500 dilution, Millipore), a mouse anti-GFAP antibody (1:500 dilution, GA5, Cell Signaling Technologies) or a mouse anti-Iba1 antibody (1:500 dilution, GT10312, GeneTex) in 1% BSA and 0.1% Triton X-100 in PBS at 4 °C overnight. Anti-human IgG Fc and anti-NeuN were visualized with an Alexa Fluor 488-conjugated anti-goat secondary antibody or an Alexa Fluor 594-conjugated anti-mouse secondary antibody, respectively (1:200 dilution, Thermo Fisher). Fluorescent images of cortical regions were obtained with the Keyence BZ-X810. This study was approved by the Institutional Review Board of Kanazawa University and performed in accordance with the guidelines and regulations.

Statistical analysis
Graphpad Prism 6.0d was used for graphing and statistical analysis. Relevant statistical methods for individual experiments are detailed in figure legends.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

March 2021
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy The main data supporting the findings of this study are available within the Article and its Supplementary Information. The RNA-seq data are available at the DDBJ Sequence Read Archive under accession numbers DRA014557 (hepatocyte spheroids) and DRA014558 (livers of PXB-mice). Source data for the figures are provided with this paper. The raw data generated during the study are available from the corresponding authors on reasonable request.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
No statistical method was used to predetermine sample sizes. Sample sizes were chosen to establish statistical significance on the basis of similar experiments reported in the literature or of data from pilot experiments. Sample sizes were chosen as large as practically possible, and adequate statistics have been applied.