Article https://doi.org/10.1038/s41467-022-34391-6 Interaction of a viral insulin-like peptide with the IGF-1 receptor produces a natural antagonist Received: 23 March 2022 Francois Moreau 1,11, Nicholas S. Kirk 2,3,11, Fa Zhang4, Vasily Gelfanov5, Edward O. List6, Martina Chrudinová7, Hari Venugopal8, Accepted: 19 October 2022 Michael C. Lawrence 2,3, Veronica Jimenez9,10, Fatima Bosch 9,10, John J. Kopchick 6, Richard D. DiMarchi 4, Emrah Altindis 7 & C. Ronald Kahn 1 Check for updates 1234567890():,; 1234567890():,; Lymphocystis disease virus-1 (LCDV-1) and several other Iridoviridae encode viral insulin/IGF-1 like peptides (VILPs) with high homology to human insulin and IGFs. Here we show that while single-chain (sc) and double-chain (dc) LCDV1-VILPs have very low affinity for the insulin receptor, scLCDV1-VILP has high affinity for IGF1R where it can antagonize human IGF-1 signaling, without altering insulin signaling. Consequently, scLCDV1-VILP inhibits IGF-1 induced cell proliferation and growth hormone/IGF-1 induced growth of mice in vivo. Cryo-electron microscopy reveals that scLCDV1-VILP engages IGF1R in a unique manner, inducing changes in IGF1R conformation that led to separa- tion, rather than juxtaposition, of the transmembrane segments and hence inactivation of the receptor. Thus, scLCDV1-VILP is a natural peptide with specific antagonist properties on IGF1R signaling and may provide a new tool to guide development of hormonal analogues to treat cancers or metabolic disorders sensitive to IGF-1 without affecting glucose metabolism. The insulin/IGF superfamily of hormones includes insulin and contain C-terminal extensions denoted as the D- and E-domains. In insulin-like growth factors-1 and −2 (IGF-1 and 2), three isoforms of insulin, the C- domain is flanked by dibasic amino acid motifs (lysine relaxin (RLN1, 2, and 3) and four insulin-like proteins (INSL3, 4, 5, and arginine) that are proteolytically processed, allowing secretion and 6), which are expressed in development of specific tissues1–8. All of the mature insulin double-chain disulfide-linked hormone and of the members of this superfamily are synthesized as single chain C-peptide10,11. IGF-1 and IGF-2, on the other hand, have short precursors that share a similar primary structure consisting of a C-domains lacking these paired basic amino acids and thus are signal peptide followed by a prohormone with three domains: the A- secreted as single chain hormones. All superfamily members also and B-domains are localized at the C- and N-termini, respectively, contain six highly conserved cysteine residues that form two with C-domain in between9. In addition, IGF-1, IGF-2 and INSL6 interchain bonds between the A-domain and B-domain, as well as 1 Section of Integrative Physiology and Metabolism, Joslin Diabetes Center, Harvard Medical School, Boston, MA, USA. 2WEHI, Parkville, VIC, Australia. 3 Department of Medical Biology, Faculty of Medicine, Dentistry and Health Sciences, University of Melbourne, Parkville, VIC, Australia. 4Department of Chemistry, Indiana University, Bloomington, IN, USA. 5Novo Nordisk, Indianapolis Research Center, Indianapolis, USA. 6Edison Biotechnology Institute and Heritage College of Osteopathic Medicine, Ohio University, Athens, OH, USA. 7Boston College Biology Department, Chestnut Hill, MA, USA. 8Ramaciotti Centre for Cryo-Electron Microscopy, Monash University, Clayton, VIC, Australia. 9Department of Biochemistry and Molecular Biology, School of Veterinary Medicine and Center of Animal Biotechnology and Gene Therapy, Universitat Autonoma de Barcelona, Bellaterra, Spain. 10CIBER de Diabetes y Enfermedades Metabólicas Asociadas (CIBERDEM), 28029 Madrid, Spain. 11These authors contributed equally: Francois Moreau, Nicholas S. Kirk. e-mail: c.ronald.kahn@joslin.harvard.edu Nature Communications | (2022)13:6700 1 Article https://doi.org/10.1038/s41467-022-34391-6 one intrachain disulfide bond in the A-domain11. In addition, all the non-helical regions at both the N- and C- termini whereas the other members, for which the tertiary structure was determined, exhibit members have a similar predicted structure11 (Fig. 1a, c). similar features in which the A-chain forms a peptide with two α- Insulin, IGF-1, and IGF-2 signal through two highly homologous helices and the B-chain forms a peptide containing one α-helix with receptor tyrosine kinases, the insulin and IGF-1 receptors (IR and IGF1R, Fig. 1 | Structural homology of LCDV1-VILP with insulin and IGFs. a Sequence (AlphaFold2 prediction, pLDDT 78.23) and LCDV1-VILP as a single- (sc) and double alignment of LCDV1-VILP with the human insulin and IGF-1. Strictly conserved chain (dc) molecules (AlphaFold2 predictions, pLDDT scores: scLCDV1 = 71.47, cysteine residues are in red. The yellow square represents the C-domain, while the dcLCDV1 = 80.58). The A- and B-chains are represented in blue and green, respec- A- and the B-domains are in blue and green respectively. Cleavage sites composed tively, and the C-peptide is in yellow. Residues involved in binding to site 1 and site 2 of dibasic residues are underlined. Conserved amino acids are represented by an of the insulin/IGF-1 receptor are respectively in red and blue. The percentage of asterisk (*), and conservatively substituted residues are noted as a dot (.). b The conserved binding residues with the sites 1 and 2 are indicated in red and blue, LCDV1-VILP sequence was aligned with the 10 sequences showing the highest respectively. sequence identity. c Three-dimensional representation of insulin (PDB: 3I40), IGF-1 Nature Communications | (2022)13:6700 2 Article https://doi.org/10.1038/s41467-022-34391-6 Fig. 2 | LCDV1-VILP as a single chain peptide is more potent to bind on IGF1R detection of IGF1R, IR, AKT, IRS1, ERK1, and ERK2 phosphorylation in lysates of and on IR. Competition binding assays on human IGF1R (a) and human IR (b) murine preadipocytes overexpressing the human IGF1R (c) or the human IR (d). ectodomains. Data are expressed as the mean ± SEM, normalized to the baseline Quantitative analysis of phosphorylated IGF1R, IR and AKT. Data, expressed as and expressed as % of the maximal binding of 125I-IGF-1 (a) and 125I-Insulin (b) mean ± SEM (n = 3 independent experiments), were normalized by phosphoryla- incubated alone (n = 3 (IGF1R) and 4 (IR) independent wells). Western blot tion intensity induced by 10 nM of IGF-1 (e) or insulin (f). respectively)12–14, whereas the relaxins and INSL peptides act through for the IGF1R and a low affinity for the IR19, whereas IGF-2 has relatively G-protein coupled receptors15. The human IR exists in two isoforms (A high affinity for both the IGF1R and the A-isoform of IR17. Structural and B), which differ by 12 amino acids at the C-terminus of the A-chain studies have revealed that in their ligand-free (apo) form, the extra- due to alternative splicing of exon 11 (A isoform: -exon 11; B isoform: cellular domain of these receptors form an inverted “V” (or inverted +exon11), whereas human IGF1R exists as a single isoform16. Insulin “U”)20 that holds their transmembrane and intracellular elements binds with high affinity to both the A- and B-isoforms of IR but exhibits apart, about 67 Å in the case of IGF-1 receptor and 115 Å in the case of much lower affinity for IGF1R17,18. Conversely, IGF-1 shows a high affinity IR. The binding of IGF-1 to the IGF-1 receptor (or insulin to IR) then Nature Communications | (2022)13:6700 3 Article https://doi.org/10.1038/s41467-022-34391-6 leads to a conformational change that permits the receptor “legs” to Table 1 | Receptor binding IC50 of human IGF-1, human insulin, come together, enabling transphosphorylation of the intracellular and LCDV1-VILPs tyrosine kinase domains and activation of the receptor21. IC50 (nM) Insulin and IGFs have been found in animal species ranging from Ligands IGF1R IR insects to humans22. Recently, we extended the scope of the insulin/ IGF-1 0.4 – IGF family of hormones by identifying viral insulin/IGF-1 like peptides Insulin – 0.5 (VILPs) in four members of the Iridoviridae family of double-stranded- DNA viruses23. While Iridoviridae primarily infect fish, reptiles and scLCDV1-VILP 3.3 188.7 insects24, DNA of these VILP-carrying viruses have been found in the dcLCDV1-VILP 36.3 >1000 human fecal virome25,26 and in human blood27, indicating potential exposure of humans to these VILPs. These VILPs show high sequence homology to insulin, IGF-1, and IGF-223. They also possess short IGF-like Single-chain (sc) and double-chain (dc) LCDV1-VILP are predicted C-peptide regions, but in contrast to human IGF-1/2, some VILPs have to share similar tertiary structures to IGF-1 and insulin (Fig. 1c). potential sites for cleavage, raising the possibility that they may act as Sequence alignment between human insulin and dcLCDV1-VILP single- and/or double chain molecules depending on the infected cells revealed that 7 out of 15 residues involved in insulin’s binding to site expressing them. In previous studies we have shown that when syn- 1 in the insulin receptor (shown in red) and 4 out of 8 residues involved thesized as single chain (sc) molecules, VILPs can bind to and activate insulin’s binding to site 2 (shown in blue) are conserved in LCDV1-VILP. both human IR (hIR) and human IGF1R (hIGF1R), with higher affinity for Similarly, 7 out of 16 and 2 out of 5 of the residues involved in IGF-1’s hIGF1R. Single-chain VILPs can also stimulate proliferation of human binding to IGF1R sites 1 and 2, respectively, are conserved in scLCDV1- fibroblasts, increase glucose uptake into mouse adipocytes in vitro, VILP23. and, at high dose, decrease blood glucose in vivo in mice23. In addition, VILPs can have unique properties not observed with insulin or IGF1. For scLCDV1-VILP is more potent in binding to IGF1R than IR example, when synthesized as double-chain (dc) peptides, the VILPs To understand the differences in the two potential forms of LCDV1- from Singapore Grouper Iridovirus (SGIV) and Grouper Iridovirus (GIV) VILP, we synthesized both the single chain (more IGF-1-like) and the not only act as insulin analogs in vitro and in vivo, but dcGIV-VILP also double chain (more insulin-like) peptides and assessed their ability to exhibits preferential tissue-specific action to stimulate glucose uptake compete with 125I-insulin and 125I-IGF-1 for binding to the ectodomains in white adipose tissue28. Also recently, we have shown that Lympho- of the hIR (A-isoform) and hIGF1R. As expected, unlabeled IGF-1 was cystis disease virus 1 (LCDV1) VILP can inhibit hIGF1R phosphorylation able to compete with labeled IGF-1 for binding to IGF1R with an IC50 in vitro29 suggesting it may have antagonistic actions on this receptor. (i.e., the concentration of unlabeled ligand that inhibits labeled ligand In the present study, we have characterized the role of sc- and binding by 50%) of 0.38 nM (Fig. 2a and Table 1). Compared to hIGF-1, dcLCDV1-VILPs as ligands of the hIR and hIGF1R in vitro and in vivo. We scLCDV1-VILP was about an order of magnitude less potent in com- find that both sc- and dcLCDV1-VILPs are very weak IR agonists. peting for binding to the IGF-1 receptor, but >10-fold more potent than scLCDV1-VILP, on the other hand, is a weak partial agonist and potent dcLCDV1-VILP (IC50 = 3.3 nM vs. 36.6 nM). scLCDV1-VILP was also competitive antagonist of IGF1R. More interestingly, scLCDV1-VILP can more potent in competing with 125I-Insulin for binding to IR than inhibit IGF-1-induced signaling, reduce IGF-1-induced cell proliferation dcLCDV1-VILP with IC50s of ~188 nM and >1000 nM, respectively, in vitro, and slow body weight gain of transgenic mice with high cir- although both VILPs had very low affinity compared to insulin itself, culating levels of IGF-1 due to over-expression of bovine growth hor- which had an IC50 of 0.55 nM (Fig. 2b and Table 1). mone (bGH). Using cryo-electron microscopy (cryoEM), we find that To assess the ability of sc- and dcLCDV1-VILPs to activate the IR scLCDV1-VILP in complex with the IGF1R ectodomain forms a unique and IGF1R signaling pathways, we used murine brown preadipocyte interaction at the ligand-binding site and a dramatically different cell lines in which the endogenous IR and IGF1R genes had been overall receptor conformation, explaining scLCDV1-VILP’s antagonism. inactivated to create a double knockout cell, after which the cells were Thus, scLCDV1-VILP presents a new naturally occurring antagonist of reconstituted with either hIGF1R alone or hIR (B isoform) alone. In the IGF1R and understanding this ligand-receptor interaction may preadipocytes over-expressing hIGF1R, IGF-1 stimulated receptor also assist design of analogues that could serve as treatment for autophosphorylation in a dose-dependent manner with half-maximal metabolic and growth disorders linked to hyperactivation of IGF1R stimulation at 3.8 nM (Fig. 2c, e). On the other hand, despite its ability signaling, with a minimal impact on glucose metabolism. to compete for binding to IGF1R with about 10% of the affinity of IGF-1, scLCDV1-VILP only minimally induced autophosphorylation of IGF1R, Results and this peaked at 10 nM ligand and then decreased with further LCDV1-VILP shares a similar structure with vertebrate and increases in ligand concentration. Consistent with its lower affinity for invertebrate IGFs and insulins IGF1R, dcLCDV1-VILP produced no measurable increase in IGF1R All members of the insulin/IGF superfamily, including VILPs, exhibit autophosphorylation. comparable domain architecture (Fig. 1a)11,23. Sequences of the ten Similar dose response curves were seen for stimulation of IRS-1 insulin/IGFs showing the highest percentage of similarity to LCDV1- Tyr895 phosphorylation and AKT Ser473 phosphorylation for both sc- VILP are aligned in Fig. 1b and can be found in the Supplementary and dcLCDV1-VILPs. Again, with the single-chain ligand, IRS-1 and AKT Data 1. Among these sequences, starfish insulin-like peptide, herring phosphorylation were much lower than with IGF-1, reaching a max- IGF-2 and mammal insulin/IGFs showed the highest sequence iden- imum at 10 nM ligand then decreasing at higher concentrations, sug- tity (about 50%) with LCDV1-VILP. Like human IGF-1 and −2, LCDV1- gesting potential inhibition of signaling at these higher concentrations. VILP has a short C-peptide (10 amino acids); however, unlike these Interestingly, in cells expressing hIGF1R, neither sc- nor dcLCDV1-VILP IGFs, this C-peptide contains only one site of potential cleavage stimulated any detectable ERK1/2T202/Y204 phosphorylation, indicating a represented by three successive arginines at the B-domain/C-domain unique post-receptor pathway bias in the signaling by these ligands. junction (underlined in Fig. 1b), which might be recognized by Since IGF1R undergoes internalization following its stimulation, we enzymes such as the proprotein convertases 1 and 230. Thus, also compared the ability of scLCDV1-VILP and IGF1 to stimulate IGF1R depending on the cellular host and its potential for processing and endocytosis, using cells in which we had biotinated surface proteins31. secretion, LCDV1-VILP could be produced as a single chain or double As expected, stimulation of cells expressing hIGF1R with 10 nM IGF1 chain peptide. lead to a significant decrease by 57% of IGF1R at the membrane by Nature Communications | (2022)13:6700 4 Article https://doi.org/10.1038/s41467-022-34391-6 Fig. 3 | scLCDV1-VILP is a potent IGF1R antagonist. Western blot detection of were normalized to total protein, and data expressed as mean ± SEM (two-way IGF1R, IR, AKT, ERK1 and ERK2 phosphorylation in lysates of murine preadipocyte ANOVA followed by a Šídák’s multiple comparison test; Graphpad Prism V.9; n = 4 overexpressing the human IGF1R (a) or the human IR isoform B (b) as described in (IGF1R) and 6 (IR) independent experiments; data were normalized by phosphor- Methods. Quantitation of phosphorylated IGF1R, IR, AKT and ERK1. Scan intensities ylation intensity induced by 10 nM of IGF-1 (c) or insulin (d). 45 min (Supplementary Fig. 1a, b). At the same time point, scLDV1-VILP using both sc- or dcLCDV1-VILP along with IGF-1 or insulin and murine at 10 and 1000 nM were able to reduce the level of membrane IGF1R by preadipocytes overexpressing the hIR or hIGF1R. In cells expressing 40% and 47% respectively, in line with its lower potency to bind and only the hIGF1R, 10 nM IGF-1 produced a robust (>40-fold) stimulation activate IGF1R. of IGF1RY1135/Y1136 autophosphorylation, and this was accompanied by In double knockout preadipocytes expressing only the human IR, increased downstream signaling with >10-fold increases in phosphor- insulin produced an expected dose-dependent increase in receptor ylation of AKTS473 and ERK1/ERK2T202/Y204 (Fig. 3a). Interestingly, when autophosphorylation, but neither sc- nor dc-LCDV1-VILP had a mea- IGF-1 (10 nM) was co-incubated with an equal concentration of surable effect. In these cells, single chain-, but not double chain-, scLCDV1-VILP, autophosphorylation of IGF1RY1135/Y1136 and phosphor- LCDV1-VILP did produce a very low level of IRS-1Y895 and AKTS473 ylation of AKTS473 and ERK1/ERK2T202/Y204 were reduced by ~40%. Fur- phosphorylation at the highest concentrations tested, and again nei- ther increasing the concentration of scLCDV1-VILP to 100 and 1000 nM ther had an effect on ERK1/2T202/Y204 phosphorylation (Fig. 2d, f). increased this inhibition, with almost a complete blockade IGF-1 sti- mulated receptor autophosphorylation, as well as phosphorylation of scLCDV1-VILP is a potent IGF-1 receptor antagonist AKTS473 and ERK1/ERK2T202/Y204 at the highest VILP concentration The above data indicate that while scLCDV1-VILP can bind to hIGF1R (Fig. 3a, c). By contrast, co-incubation with dcLCDV1-VILP had no effect with relatively high affinity, it has very little signaling capacity, sug- on IGF-1 stimulation of receptor autophosphorylation or downstream gesting it may be a competitive antagonist. To examine this hypothesis signaling. This antagonistic effect of scLCDV1-VILP was specific to in more detail, we performed a series of co-stimulation experiments IGF1R. Thus, using DKO cells expressing the human IR, addition of Nature Communications | (2022)13:6700 5 Article https://doi.org/10.1038/s41467-022-34391-6 10 nM insulin marked increased IRY1150/Y1151 autophosphorylation and (GS-FSC 0.143 ca 4.6 Å), sufficient to permit unambiguous placement stimulated AKTS473 and ERK1/ERK2T202/Y204 phosphorylation (Fig. 3b). In of the scLCDV1-VILP moieties and the domains of the receptor. contrast to the effects on IGF1R signaling, however, co-incubation of Whereas the resolution varied significantly across the map (local these cells with either sc- or dcLCDV1-VILP at concentrations from 0.1 resolution: min 2.4 Å, median 6.8 Å, 75th percentile 8.4 Å), highest to 1000 nM had no significant effect on phosphorylation on insulin resolution was achieved at one scLCDV-1 VILP binding site, permitting stimulation of IRY1150/Y1151 or its downstream kinases (Fig. 3b, d). Insulin/ conservative morphing of an AlphaFold236 (AF2) model of scLCDV1- IGF-1 signaling in response to either IGF-1, sc- or dcLCDV1-VILP, was VILP to fit the density. The resultant model and associated cryoEM assessed in MCF-7 cells, expressing endogenous levels of both hIR and potential density are shown in (Fig. 5a–c, Supplementary Fig. 4e). hIGF1R. Consistent with our data shown in (Fig. 2a, c) IGF-1 was more Within our atomic model, the structural changes within each potent than sc- and dcLCDV1-VILP in inducing phosphorylation of the receptor monomer upon scLCDV1-VILP binding (Supplementary IGF1R, AKT and ERK1/ERK2 (Supplementary Fig. 2a). In line with the Fig. 4a) are seen to be considerably less marked than those induced by competition binding assay performed on hIGF1R (Fig. 2a), scLCDV1- IGF-1 binding to the IGF-1R ectodomain (Supplementary Fig. 4b). The VILP was more potent than dcLCDV1-VILP in stimulating AKT phos- L1 domain is displaced from its apo location adjacent to domain FnIII-2’ phorylation. Consistent with the Fig. 3a, c, only the scLCDV1-VILP was by only ∼13 Å, compared to ∼42 Å in the IGF-1 bound case (Fig. 5d, e). able to antagonize IGF1R signaling induced by 10 nM of IGF1. Also, although difficult to quantify due to heterogeneity in the relative disposition of domains FnIII-3 and FnIII-3′, the points of membrane scLCDV1-VILP inhibits IGF-1 induced cell proliferation and entry are judged here to be approximately 90 Å apart, based on a linear mouse growth extrapolation from domains FnIII-1 and FnIII-1′. Such separation is A major effect of IGF-1 is stimulation of cell cycle progression and cell greater than that in the apo IGF1R ectodomain and likely incompatible replication32. In murine brown preadipocyte over-expressing hIGF1R, with signaling activation21. hIGF-1 produced a dose-dependent increase in cell number with a half- In contrast to what is seen within the IGF-1 bound receptor37, the maximal effect at ~0.07 nM IGF-1 and a maximal effect at 1 nM (Fig. 4a). scLCDV1-VILP moieties are not translocated to the receptor head At low concentrations, scLCDV-VILP also induced cell proliferation, but module and do not engage the membrane-distal regions of the FnIII-1 the curve was shifted one-order of magnitude to the right, peaking at or FnIII-1′ domains (Fig. 5f, g vs Fig. 5h). Instead, the scLCDV1-VILP 10 nM and then decreasing to basal levels with further increases in bound to domain L1 makes a sparse contact with domain FnIII-1′ in ligand, suggesting a self-inhibitory effect. This antagonistic effect of the vicinity of residue His539′. This residue lies towards the scLCDV1-VILP was even clearer when preadipocytes expressing hIGF1R C-terminus of canonical strand E of domain FnIII-1′ and, based on the were incubated with 10 nM of IGF-1 and increasing concentrations of homology of IGF1R and IR, may form part of IGF-1’s binding site 2 scLCDV1-VILP. Thus, consistent with the signaling data, co-incubation (equivalent to IR Arg554′)38. The [domain-L1´]-bound scLCDV1-VILP of IGF-1 with increasing concentrations of scLCDV1-VILP produced a moiety approaches, but however does not engage, domain FnIII-1 dose dependent inhibition of IGF-1 stimulated cell proliferation (Fig. 5f, g). (Fig. 4b). By contrast, dcLCDV1-VILP had no effect on IGF-1 mediated The C peptide of scLCDV1-VILP appears more ordered in our cell proliferation (Fig. 4a, b). The lack of increase in cell number was structure than that of IGF-1 in its receptor complex37 and is displaced not related to increased apoptosis, since cell survival was not changed more towards the C-terminal end of the domain-L1 β barrel (Supple- by high dose of either scLCDV1-VILP or dcLCDV1-VILP (Fig. 4c). mentary Fig. 4c vs Supplementary Fig. 4d). We note the substitution of To assess the functional consequence of LCDV1-VILP antagonism Gly7 in IGF-1’s B-chain (a highly conserved residue in the insulin/IGF- of IGF1R signaling in vivo, we utilized Adeno-Associated Vector ser- 1 superfamily) by serine in LCDV1-VILP (Fig. 1a) as the presence of otype 8 (AAV8) viral vectors carrying the sequence of LCDV1-VILP and a serine (rather than glycine) at this location is important for the transgenic mouse model expressing bGH. These mice have been pre- antagonistic properties of LCDV1-VILP29. Glu9 in the B-chain of IGF-1 is viously shown to have elevated circulating levels of IGF-1, an increase in a key residue involved in the binding with IGF1R39. In LCDV1-VILP, Glu9 lean mass, accelerated growth, and reduced lifespan compared to is replaced by histidine, which would alter this interaction. wildtype littermates33,34. As expected, transfection of mice with AAV- LCDV1-VILP led to a dramatic increase in the levels of Lcdv1-vilp mRNA Discussion in liver compared to undetectable levels in mice receiving the control Viral insulin-like peptides (VILPs) represent a unique branch of the AAV (Fig. 4d). This was associated with a slight, but not significant, insulin/IGF-1 superfamily tree, being the only insulin-like molecules in increase in the expression of endogenous Igf-1 in liver, as well (Fig. 4d). non-eukaryotic organisms. In this study we have explored the actions Nonetheless, over the following seven weeks, body weight gain in the of one of these unique insulin-like molecules, LCDV1-VILP, as a single- bGH transgenic mice that received the AAV-LCDV1-VILP was sig- chain (sc) IGF-1-like and double-chain (dc) insulin-like peptide, as well nificantly reduced versus those receiving control AAV (Fig. 4e, f). as interaction of scLCDV1-VILP with IGF1R receptor at a structural level. Nuclear magnetic resonance revealed that this reduction of weight We find that despite its insulin-like structure, dcLCDV1-VILP has no gain was accompanied by a reduction in lean body mass, but no change ability to interact with the human insulin receptor and no insulin-like in fat mass (Supplementary Fig. 3a–c). There was also no change in the actions. The IGF-1-like single-chain LCDV1-VILP, on the other hand, has levels of fasting blood glucose (Fig. 4g). relatively high affinity for hIGF1R (~12% that of hIGF-1). However, in contrast to IGF-1, scLCDV1-VILP has very low agonist activity (<1%), but CryoEM modeling of the IGF1R ectodomain in complex with instead acts as a potent competitive antagonist of IGF1R both in vitro scLCDV1-VILP and in vivo. This action of scLCDV1-VILP is due to a different mode of To determine how LCDV1-VILP effects antagonism of hIGF1R, we per- interaction with IGF1R. Thus, in contrast to IGF-1 that binds in a 1:1 formed cryo-electron microscopy (cryoEM) imaging of an IGF1R ligand to receptor ratio, two molecules of LCDV1-VILP bind to the ectodomain construct IGF1Rzip35 prepared in the presence of a 30-fold receptor dimer, but they are unable to induce the conformational stoichiometric ratio of scLCDV1-VILP (∼30-fold molar excess calcu- change to allow site 1b interaction, and as a result, the extracellular lated per receptor αβ monomer). The initial reconstruction was readily juxtamembrane domains, and hence the associated transmembrane interpretable, displaying a pseudo-two-fold symmetric arrangement of and intracellular domains, move further apart rather than closer IGF1Rzip domains, with each L1-CR + αCT′ module engaging a together, leading to an inability of the IGF1R kinase to undergo scLCDV1-VILP moiety. Further ‘non-uniform’ refinement (NU-refine- transphosphorylation and activation of the receptor toward its ment) in cryoPARC (v3.2) led to a final map with improved resolution downstream substrates. Nature Communications | (2022)13:6700 6 Article https://doi.org/10.1038/s41467-022-34391-6 Fig. 4 | scLCDV1-VILP reduces cell proliferation and growth in mice over- a Dunn’s multiple comparisons test (Graphpad Prism V.9); n = 5 (scLCDV1-VILP) and secreting bovine growth hormone. a Murine preadipocytes overexpressing the 6 (dcLCDV1-VILP) independent wells). d Relative expression of the LCDV-1 and IGF-1 human IGF1R were plated at a density of 20,000 cells/well then counted after a 48-h mRNA expression in liver samples from mice that received either an AAV control incubation with the indicated concentrations of IGF-1, scLCDV1-VILP and dcLCDV1- (n = 6 mice) or the AAV LCDV-1 (n = 8 mice). Data are expressed as the mean ± SEM VILP. Data are expressed as the mean ± SEM (Two-way ANOVA followed by a Dun- (ns, not significant; two-sided Mann–Whitney test; Graphpad Prism V.9). e Body nett’s multiple comparison test (Graphpad Prism V.9); n = 3 independent experi- weight gain at week 2 and 6 following AAV injection. Data are normalized by the ments). b The number of murine preadipocytes expressing the hIGF1R were basal weight measured at t = 0 and expressed as mean ± SEM (two-sided counted after a 48-h incubation with various concentrations of scLCDV1-VILP and Mann–Whitney test; Graphpad Prism V.9; AAV control = 6 mice, AAV LCDV-1 = 8 dcLCDV1-VILP and a fixed concentration of IGF-1 (10 nM). Data are expressed as the mice). f The increment (delta) of body weight gain was measured weekly for seven mean ± SEM (two-way ANOVA followed by a Šídák’s multiple comparison test weeks following the AAVs injection. Data are express as mean ± SEM (AAV con- (Graphpad Prism V.9); n = 3 independent experiments). c Viable murine pre- trol=6 mice, AAV LCDV-1 = 8 mice). g Fasting blood glucose was assessed 2 and adipocytes overexpressing the hIGF1R were stained with crystal violet after a 48-h 5 weeks after the AAVs injection. Data are express as mean ± SEM (ns, not sig- incubation with various concentrations of IGF-1, scLCDV1-VILP and dcLCDV1-VILP. nificant; two-way ANOVA followed by a Šídák’s multiple comparison test; Graphpad Data are express as mean ± SEM (ns, not significant; Kruskal-Wallis test followed by Prism V.9; AAV control=6 mice, AAV LCDV-1 = 8 mice). Interestingly, while viruses are very diverse and include both RNA shares ~50% homology with the IGFs of certain species of fish and and DNA carrying species in over 180 families, all of the VILPs dis- reptiles, which might suggest horizontal gene transfer, the sequence covered thus far are encoded by double-stranded DNA viruses conservation of LCDV1-VILP is almost as high with human IGFs and belonging to the Iridoviridae family23. This family of viruses is known to insulin (~47%), making this hypothesis uncertain. The exact role of infect primarily fish, reptiles and insects24. Although, LCDV1-VILP these VILPs in viral function remains to be determined, infected fish Nature Communications | (2022)13:6700 7 Article https://doi.org/10.1038/s41467-022-34391-6 Fig. 5 | Relative conformations of the apo- and ligand-bound IGF1R ectodo- Domains are labeled L1, L2: first and second leucine-rich repeat domains; FnIII-1, -2, main. a CryoEM density plus the atomic model of scLCDV-1VILP bound to IGF1R.zip -3: first, second and third fibronectin Type III domains; αCT: C-terminal α-helical shown in ribbon representation (map contour level 0.149). b CryoEM density segment of IGF1R α-chain (prime symbol ´ denotes domains from alternate associated with scLCDV1-VILP, domain FnIII-1′, αCT´ and domain L1 (map contour monomer). d Apo IGF-1R ectodomain crystal structure20 (PDB: 5U8R). Domain level 0.149). c CryoEM structure of IGF1R ectodomain in complex with two copies of representation and nomenclature is as in panel c. e CryoEM structure of a single scLCDV1-VILP with one receptor monomer in low-resolution surface representation IGF-1 bound to IGF1R (PDB: 6PYH)37. Domain representation and nomenclature as in and the other in ribbon representation. Note domains FnIII-3 and FnIII-3´ and c. f, g Relative disposition of scLCDV1-VILPs and (h) IGF-1 upon binding to receptor, upstream regions of the insert domains are positioned based on weak potential illustrating (*) the comparative lack of interaction of the L1´-bound scLCDV1-VILP density from an initial consensus cryoEM map; these entities are unmodelled in the with domain FnIII-1. final deposited structure as they are not discerned in the final potential map. develop wart-like nodules on their fins, skin, or gills, which represent molecules, in cells infected with these viruses remains unknown. Two clusters of fibroblasts massively enlarged due to viral production. of the major differences between vertebrate insulins and IGFs are in Although not believed to be a human pathogen, it is important to note the length of C-peptide region that links the B- and A-domains (30-33 that sequences of the LCDV-1 virus have been found in the human fecal amino acids in insulins versus 8-12 amino acids in IGFs) and the pre- virome25,26 and in human blood26, suggesting that humans are likely sence of dibasic amino acids at the C-peptide junctions with the A- and exposed to these viral insulin-like peptides. Furthermore, scLCDV1- B-chains in proinsulin, allowing the C-peptide to be cleaved out, leav- VILP has a high affinity for both the human and murine IGF-1 ing the mature two-chain insulin molecule11. LCDV1-VILP is a hybrid of receptors23, indicating that if humans are exposed to this viral pep- these features, having a short IGF-like C-peptide (10 amino acids), but tide, it may have biological consequences40. having basic amino acids at the theoretical B-chain/C-peptide junction, Whether LCDV1-VILP or other VILPs are processed to double which potentially allows for proteolytic processing to a two-chain chain, insulin-like molecules or remain as single chain, IGF-1-like molecule in LCDV1-infected cells. To understand the differences Nature Communications | (2022)13:6700 8 Article https://doi.org/10.1038/s41467-022-34391-6 between these two products, in the present study we have compared ligand-bound L1-CR module detaches from the FnIII-2´ domain, par- the actions of scLCDV1-VILP with that of dcLCDV1-VILPs both in vitro tially freeing up the three-domain FnIII′ module and permitting it to and in vivo. move into proximity of the FnIII domain module in the opposing While the differences between sc- and dcLCDV1-VILPs are modest monomer. As the fully stabilized conformation of an IGF-1-liganded on insulin receptor signaling, since they have very low affinity for the ectodomain is not achieved, such movement may reduce agonism IR, they are dramatic on the IGF-1 receptor. Thus, scLCDV1-VILP has an compared to the IGF-1 occupied receptor (“partial agonism”). Also, in affinity for IGF1R that is almost 12% that of hIGF-1, as compared to the the absence of the conformational transition that enables the bound double chain hormone, which has <1% of the affinity of the native scLCDV1-VILP to engage site 1b, a second scLCDV1-VILP can bind to the ligand. This is likely due to its more IGF-1-like structure. In addition, the alternate [L1-CR]′ + αCT element, yielding the double-liganded struc- C-domains of LCDV1-VILP and IGF-1 are of similar length (10 vs 12 ture observed in our study. The overall conformational flexibility of the amino acids) and share 33% sequence identity, including Arg37, one of ectodomain is also reduced by the constraints imposed by the the residues in the IGF-1 C-domain known to be involved in the disulfide-linked αCT elements that cross-link the two scLCDV1-VILP- receptor binding interaction41. Removal of the C-peptide from LCDV1- bound L1-CR modules, resulting in the ectodomain’s points of mem- VILP decreases the affinity of LCDV1-VILP for the IGF1R by tenfold, brane entry and the associated intracellular domains becoming further consistent with the important role for the C-peptide in the receptor separated, rather than closer together as they do when IGF-1 binds. As a binding, similar to what is observed with IGF-1 itself37,42,43. scLCDV1- result, transphosphorylation of the intracellular beta-subunits does VILPs exhibits very low affinity for binding to the IR and activating IR not occur, and the receptor is unable to activate its downstream sig- post-receptor signaling (about 1/1000 as active as insulin), and this is naling pathways. not increased with dc-LCDV1-VILP, despite having a more insulin-like The potency of scLCDV1-VILP to activate IGF1R and its down- structure. Consistent with this, glycemia in mice expressing LCDV1- stream signaling may vary depending on the concentration of ligand. VILP in liver was not changed compared to controls. Whereas higher concentrations (100–1000 nM) of scLCDV1-VILP are LCDV1-VILP does have one unique and important property, associated with the inactivation of the receptor, lower levels (1–10 nM) namely its ability to antagonize IGF-1 at the IGF1R. This effect of LCDV1- of ligand could potentially lead to partial IGF1R activation. Since the VILP is restricted to the single chain peptide and is seen only with the cryoEM was performed in the presence of a large excess of the ligand, IGF1R, and not with the IR. The fact that scLCDV1-VILP has the potential we hypothesize that at low concentrations, it may be possible for one to serve as an IGF-1 receptor antagonist is unique among naturally scLCDV1-VILP to bind to the receptor and induce a conformational occurring insulins and IGFs. Furthermore, our data show that the change leading to the partial activation of the receptor, further studies potency of scLCDV1-VILP to inhibit IGF1R signaling is sufficient to have will be required to know if this actually occurs. functional consequences in vitro and in vivo. Thus, scLCDV1-VILP is As noted above, studies of chimeric and mutant IGFs show that able to reduce IGF-1 induced cell proliferation in vitro and partially the antagonistic properties of scLCDV1-VILP are associated with its block the effects of endogenous IGF-1 in transgenic mice over- unique C-peptide sequence and with a serine at position 8 in the expressing bovine growth hormone in vivo, as manifested by B-domain (equivalent to a glycine in IGF-1 at position 7)29. Within the reduced lean body mass and reduced body weight gain. Since this AF2 predicted structure, scLCDV1-VILP Ser8 lies at the N-terminus of antagonism is restricted to the single chain version of the hormone, the B-domain helix in scLCDV1-VILP which would, compared to glycine, this suggests that when over-expressed in liver, LCDV1-VILP is not impose a Ramachandran constraint on the positioning of residues B1- processed, but secreted as a single chain molecule, similar to IGF-1. B7. Within our model, Ser8 could interact with the nearby residues Consistent with the current study, our groups have shown that Glu91 and Lys115 within domain L1 and Glu693′ within αCT′, though we LCDV1-VILP can also serve as an inhibitor of IGF1R in HEK293 cells, and acknowledge the resolution of the cryoEM map precludes confirma- using chimeric peptides of IGF-1 and LCDV1-VILP, shown that the tion. These differences may impact the ability of scLCDV1-VILP to C-peptide domain of the VILP is essential for its antagonistic action29. engage key Site 1b residues, Lys530, Tyr487 and Arg488 from FnIII-1´37 In addition, one amino acid, a glycine in position 7 in the B-domain of that compete for the same intra-receptor interactions. Differences in IGF-1 also appears to be required for the antagonist effects. This gly- the structure of the respective C domains may also contribute to the cine is highly conserved in most members of the insulin/IGF- unique interaction. Replacement of the scLCDV1-VILP C-domain resi- 1 superfamily44 and is replaced by a serine in LCDV1-VILP. This residue due Arg32 by its critical hIGF-1 counterpart Tyr31 does not reduce also appears to be required for the antagonist effects, since substitu- antagonism, but replacement of the scLCDV1-VILP Arg31-Arg32-Arg33 tion of the serine by a glycine decreases the potency of LCDV1-VILP to triplet by its Gly-Tyr-Gly IGF-1 equivalent does29. In our structure, the C inhibit IGF-1 signaling and increases its agonist property29. Also of domain of scLCDV1-VILP appears relatively displaced towards IGF1R interest is the finding that, despite its low agonist activity, scLCDV1- domain L1, which may hinder it forming a native-like interaction with VILP has unique post-receptor signaling properties, such that at a domain L2. Moreover, the residue Glu9 in the B-chain of IGF-1 forms a concentration of 10 nM, it can activate the AKT pathway, but does not salt bridge with the Tyr448 of the site 1b35,37. Glu9 is highly conserved in appear to activate the ERK pathway. Since scLCDV1-VILP is capable of insulin superfamily members. However, this residue is replaced by a inducing IGF1R internalization similar to IGF-1 itself, this selective sig- histidine in LCDV1-VILP. The absence of this salt bridge may also naling seems to be independent of IGF1R endocytosis. Which struc- reduce affinity for the Site 1b. tural features account for this biased signaling remain to be IGFs and IGF signaling have a key role in growth and cell pro- determined. liferation, and deregulation of these processes contributes to the In addition to demonstrating its unique signaling and biological pathogenesis of multiple diseases45. Overproduction of IGF-1 can lead activities in vitro and in vivo, in the present study we were also able to to excessive growth in patients with acromegaly or gigantism46, and determine the unique nature of the LCDV1-VILP receptor interaction at high expression of IGF-1 can also promote growth of many types of a structural level using cryoEM. While previous studies have tumors47. Excessive secretion of incompletely processed IGF-2 has shown that IGF1R binds only a single IGF-1 or IGF-2 molecule35,37, even been shown to be a cause of non-islet cell tumor hypoglycemia48. when prepared with molar excesses of these ligands, we find that the Overexpression of IGF1R has also been shown to promote a number of IGF1R is able to bind two molecules of scLCDV1-VILP. This two-ligand- cancers, including breast, gastrointestinal and prostate cancer49. bound structure appears to arise as a consequence of the [L1-CR + αCT Conversely, inhibitors of IGF1R have been shown to have beneficial ′]-bound scLCDV1-VILP being unable to engage productively with site effects in some cancers50, alleviate the inflammatory process asso- 1b. We propose that when a single scLCDV1-VILP binds to the IGF1R, the ciated with diabetic kidney disease in mice51, and are already in use in Nature Communications | (2022)13:6700 9 Article https://doi.org/10.1038/s41467-022-34391-6 humans in the treatment of hyperthyroidism associated eye disease52. (Parkville, Australia), with the protein being expressed and purified as Thus, developing inhibitors of IGF1R signaling could provide impor- previously described56,57. Competitive binding assays was then per- tant therapeutics. Because of the strong homology between IGF1R and formed as previously described with slight modification20. Briefly, wells IR, the development of IGF1R inhibitors has been challenging and most of a 96-well plate were coated with 250 ng of an IGF1R monoclonal did not achieve their therapeutic goals due to a lack of specificity and/ antibody (24-31) or an IR monoclonal antibody (83-7), gifts from Pro- or only modest efficacy53. scLCDV1-VILP is the first natural peptide that fessor Ken Siddle (University of Cambridge). After an overnight incu- exhibits a specific antagonist potency on IGF1R. Thus, this peptide or bation at 4 °C, the antibody solution was removed, and the wells were peptides based on its structure may serve as a guide for development blocked with 0.5 % bovine serum albumin (BSA) in 20 nM Tris pH 7.4, of IGF1 antagonists. 150 nM NaCl and 0.1% tween-20 (TBS-T) for 2 h at room temperature. Ectodomain protein was then added for 2–3 h at room temperature in Methods TBS-T (0.5 μg/well). I125 labeled IGF-1 or insulin (0.01 nM) was added in Bioinformatics each well, in addition of an increasing concentrations of unlabeled IGF- The sequence of LCDV1-VILP was compared to those of insulin and IGF- 1, insulin, or LCDV1-VILPs for 16 h at 4 °C. Wells were washed three 1 by performing a multiple sequence comparison by log-expectation times with TBS-T, and 100 μL of a solution containing NaOH (0.1 nM) (MUSCLE)54 using Molecular Evolutionary Genetics Analysis software and SDS (0.1%) was added. After 10 min, 80 μL of the lysis solution was (MEGA version 10.1.8). The whole LCDV1-VILP sequence was used on added to a 4 mL of scintillation liquid (Perkin Elmer), and the radio- BlastP to identify the top 10 sequences that exhibit the highest identity. activity was measured in a liquid scintillation counter (Beckman). AlphaFold2 was used to predict three-dimensional structures of scLCDV1-VILP and dcLCDV1-VILP36. Structures were visualized using IR and IGF1R signaling ChimeraX55. MCF-7 cells, a breast cancer cell line (SL017, Genecopoeia) and brown preadipocytes knocked-out for the endogenous insulin and IGF-1 Peptide synthesis receptors and overexpressing either the human IR or the human IGF-1 Single chain LCDV1-VILP. Single chain viral insulins were synthesized receptor31 were plated into six-well plates. Cells were cultured in on 0.1 mmol Rink amide ChemMatrix® (PCAS BioMatrix Inc.) resin DMEM supplemented with 10% FBS (Sigma), 100 U/mL penicillin, using an ABI-433A peptide synthesizer and Fmoc/6-Cl-HOBt/DIC cou- 100 μg/mL streptomycin (Gibco) and 100 μg/mL Normocin (Invivo- pling protocols29. Fmoc-Asp-OtBu was employed to introduce the gen) at 37 C with 5% CO2 and 90% humidity. When cells reached C-terminal Asn. Cleavage was conducted by treatment with 10 mL of a 70–80% confluence, the medium was removed and replaced by a TFA solution containing 2.5% TIS, 2.5% 2-mercaptoethanol, 2.5 % ani- starvation medium, i.e., without FBS, for 6 h. Cells were then incubated sole, and 2.5% H2O at room temperature with gentle agitation for 1.5 h. for 15 min with insulin, IGF-1 or LCDV1-VILP at concentrations from 0.1 The resin was filtered, and the peptide precipitated by addition of cold to 1000 nM. In the case of co-incubation experiments, cells were pre- ether (50 mL). The peptide precipitate was collected by centrifugation treated with the LCDV1-VILP for 30 min and then stimulated with 10 nM then washed with cold ether (3 × 50 mL). The crude peptide was insulin or IGF-1 for 15 min. Cells were washed with ice cold PBS and solubilized in mixture of ammonium acetate (0.0429 M), urea snap frozen in liquid nitrogen. Proteins were extracted with a RIPA (2.571 M) and acetic acid (0.01 M). I2 in methanol was added dropwise, buffer (MilliporeSigma) supplemented with 0.1% SDS and a cocktail of and the solution was further stirred for 20 min before quenching with protein phosphatase (#B15001-A and B15001-B, Bimake) and protease ascorbic acid. The intermediate with two disulfide bonds was obtained inhibitors (#B14001, Bimake) at 1x. Protein concentrations were after preparative HPLC purification and lyophilization. The lyophilized assessed using a BCA protein assay kit (Pierce). Proteins (10–20 μg) peptide was dissolved in TFA with the addition of DMSO (5%). The were loaded in a 4–12% Bis-Tris gel (Invitrogen) and transferred to a mixture was stirred at room temperature for 2 h. After cold ether PVDF membrane. The membrane was incubated at room temperature precipitation, the peptide was solubilized in 20% CH3CN/H2O and for 1 h in a block solution (ThermoFisher) before an overnight incu- adjusted to pH 8.0. The solution was acidified after 30 min at room bation at 4 oC with primary antibodies (1:1000). Membranes were temperature. Another preparative HPLC purification and lyophiliza- washed in a TBS-T buffer then incubated for 4 h at room temperature tion to yield the final product. with HRP conjugated secondary antibodies (1:1000). The following antibodies were used in this study: IRS-1 (catalog 611394) from BD Double chain LCDV1-VILP. The A-chain was assembled on Rink amide Bioscience, p-IRS-1Y895 (catalog 3070), IR (catalog 3025), IGF1R (catalog ChemMatrix resin with Fmoc-Asp-OtBu to introduce the C-terminal 3027), p-IRY1150/1151/IGF1RY1135/1136 (catalog 3024), AKT (catalog 4685), Asn. The B-chain was constructed on Thr-preloaded HMPB ChemMa- p-AKTSer473 (catalog 4060), ERK1/2 (catalog 9102) and p-ERK1/2T202/Y204 trix resin. Both peptides were cleaved and purified as described (catalog 4370) from CellSignaling, goat anti-rabbit HRP conjugated previously29. A-chain (0.0048 mmol) and B-chain (0.0048 mmol) were (catalog 1706515) from BioRad and sheep anti-mouse HRP conjugated then mixed in 2.8 mL of ammonium acetate (0.0143 M) and urea (catalog NA931) from Sigma-Millipore. Quantification of protein bands (0.0857 M, pH 6.9). The thiolysis chain combination was monitored by intensity were analyzed on ImageJ (v2.3.0/1.53q) and Microsoft Excel LC-MS analysis. Upon completion, the mixture was diluted with acetic (v16.64). All the uncropped and unprocessed pictures of each blot are acid (0.01 M) and H2O. I2 (30 mg) in 0.8 mL methanol was added, and available in the source data file. the solution was stirred for 20 min. Aqueous ascorbic acid solution was added to quench the reaction to yield a colorless solution. The IGF1R internalization assay resulting dimer was purified with preparative HPLC. The A-B hetero- Murine brown preadipocytes overexpressing hIGF1R were seeded into dimer with two Cys(tBu) residues was dissolved in TFA (1.9 mL) fol- six-well plates. When cells reached 90% of confluence the complete lowed by addition of DMSO (0.1 mL). The mixture was stirred at room medium was removed and replaced by serum-free medium for 3 h. temperature for 2 h. The crude product was recovered by ether pre- Cells where then incubated for 45 min with IGF1 or scLCDV1-VILP at 10 cipitation. The final product was obtained after prep-HPLC and or 1000 nM. Cells were washed with ice-cold PBS (pH = 8) and labeled lyophilization. with 0.3 mg/ml of sulfo-NHS-biotin (#21335, ThermoFischer) for 30 min at 4 °C. The labeling solution was removed and replaced by an Competition binding assay ice-cold solution of 100 mM of glycine for 10 min at 4 °C.Cells were Stable CHO Lec8 cells line expressing respectively the IGF1R ectodo- washed two times with ice cold PBS and lysed in a RIPA buffer con- main (Δβ mutant) and the IR-A ectodomain were obtained from CSIRO taining a cocktail of protein phosphatase (#B15001-A and B15001-B, Nature Communications | (2022)13:6700 10 Article https://doi.org/10.1038/s41467-022-34391-6 Bimake) and protease inhibitors (#B14001, Bimake) at 1x. Protein same as described previously for bovine GH antagonist transgenic concentrations were assessed using a BCA protein assay kit (Pierce). mice60. Nineteen-day old female mice were injected via the tail vein Biotinylated membrane proteins were enriched by incubating 15 μl of with the AAV8 construct above designed to overexpress LCDV1 in liver. streptavidin-agarose beads solution (#20357, ThermoFisher) with 120 Each mouse received 1 × 1012 viral genomes (vg) of either the AAV8 μg of protein lysate in 800 μL final volume overnight at 4 °C.Agarose vectors carrying the LCDV1-VILP sequence (n = 8) or an AAV8 control beads were pelleted by centrifugation (5 min, 9000 × g, 4 °C) and (n = 6) that only carry the hAAT promoter and a poly (A) sequence. washed three times in RIPA lysis buffer. Bound proteins were detached Body composition was assessed weekly using the Bruker Minispec from agarose beads by heating for 5 min at 95 °C in 1× loading buffer. (Woodlands, TX), which uses NMR technology to estimate the fat and Biotinylated membrane proteins and total proteins were loaded in a 4- lean mass of the animals33. Tail vein blood glucose was measured 2 and 12% Bis-Tris gel (Invitrogen) and transferred to a PVDF membrane. 5 weeks after the AAV injection using Onetouch Ultra test strips and Membranes were immunoblotted with the following antibody: p-IRY1150/ glucometers (Lifescan, Inc, Milpitas, California). Mice were euthanized 1151 /IGF1RY1135/1136 (catalog 3024), IGF1R (catalog 3027) from CellSignal- 8 weeks after the AAV injection, body length was measured and tissues ing and goat anti-rabbit HRP conjugated (catalog 1706515) from (liver, skeletal muscles, brown and white adipose tissues) were col- BioRad. lected, weighed, snap frozen and stored at −80 °C until further analy- sis. Mice were housed 2–4 per cage and given ad libitum access to Proliferation and survival assays water and rodent chow (ProLab RMH 3000; 14% of energy from fat, Proliferation and survival assays were performed on murine brown 60% from carbohydrates, and 26% from protein; PMI Nutrition Inter- preadipocyte overexpressing the human IGF1R31. For the proliferation national, Brentwood, Missouri). The cages were maintained in a tem- assay, cells were seeded on a 48-well plate (20,000 cells/well) and perature (22 °C) and humidity-controlled room and exposed to a 14-h incubated with IGF-1 (Peprotech), scLCDV1-VILP and dcLCDV1-VILP at light, 10-h dark cycle. All procedures were approved by the Ohio Uni- concentrations from 0.1 to 1000 nM. After a 48-h incubation at 37°C, versity Institutional Animal Care and Use Committee and fully com- cells were counted using a cell counter (Cellometer). For the survival plied with all federal, state, and local policies. assay, cells were seeded into a 96-well plate. Once they reached 100% confluence, the cells were starved overnight. The cells were incubated Cloning and production of IGF1Rzip with 1000 nM of scLCDV1-VILP or dcLCDV1-VILP. After 48 h, the cells The construct IGF1Rzip comprises a 30-residue signal peptide, fol- were washed with PBS and stained with a 0.5% solution of crystal violet lowed in order by residues 1–905 of the intact IGF1 holoreceptor, a 33- (Sigma) in 20% methanol for 10 min at room temperature. Each well residue leucine-zipper motif61, a three-residue spacer t, and an eleven- was washed three times with PBS, and the cells were dissolved in residue c-myc tag. The leucine-zipper motif is proposed to act as a 100 μL of methanol. Absorbance was read at 560 nm. “soft restraint” on the spatial separation of the FnIII-3 domains and to mimic membrane embedding35,62,63. Briefly, the construct was pro- Quantitative reverse transcription PCR duced by stable expression and secretion from CHO-K1 cells and then Messenger RNAs were extracted from liver biopsies using a mix of purified by a sequential combination of 9E10 antibody-affinity chro- TRIzol® reagent (Invitrogen) and chloroform and then precipitated with matography and size-exclusion chromatography as previously 70% of ethanol. Messenger RNA purification was performed using described35, with the final sample being prepared at a concentration of RNeasy Mini Kit columns (QIAGEN) according to the manufacturer’s 0.2 mg mL−1 in a 24.8 mM Tris-HCl (pH 8.0), 137 mM NaCl, 2.7 mM KCl protocol. Complementary DNA were synthesized by reverse transcrip- plus 0.02% NaN3 (“TBSA”). tion of 2 μg mRNA using a high-capacity cDNA synthesis kit (Applied Biosystems). The qPCR was performed using a Sybr Green Supermix CryoEM sample preparation (Biorad) on a CFX384 real time PCR detection system (Biorad). The scLCDV1-VILP was prepared at a concentration of 150 μM in 10 mM HCl following primers were used: Igf-1 forward: 5′-GTGGATGCTCTT- (=1 mg mL−1) then diluted 2:5 (v/v) in TBSA (described above) to give a CAGTTCGTGTG-3′, Igf-1 reverse: 5′-TCCAGTCTCCTCAGATCACAGC-3′, 60 μM solution in TBSA plus 4 mM HCl. The resultant solution was the Lcdv1-vilp forward: 5’AGAAGAAGCACCAG-AAACGG-3′ and Lcdv1-vilp combined 1:1 with IGF1Rzip sample prepared as above to provide a final reverse: 5′-TTTCCAGGTCGTCGGTTGTAC-3′. sample concentration of ∼0.1 mg mL−1 IGF1Rzip and 30 μM scLCDV1- VILP in TBSA plus 2 mM HCl (∼30-fold molar excess calculated per Recombinant AAV vectors receptor αβ monomer). The AAV expression cassette was obtained by cloning, between the inverted terminal repeats (ITRs) of AAV2, the coding sequence of CryoEM imaging LCDV1-VILP under the control of the liver-specific human α1- Quantifoil Cu 1.2/1.3 grids (200 mesh) were glow discharged using a antitrypsin promoter (hAAT). A non-coding cassette carrying the 30 mA current for 30 s. The sample was then applied to the glow- hAAT promoter, but no transgene, was used to produce control null discharged grids and blotted to achieve desired sample film thickness vectors (AAV-null). Single-stranded AAV vectors of serotype 8 (AAV8) (Thermo Fisher Scientific Vitrobot; settings: 3 s blot time, -3 blot force, were produced by triple transfection in HEK293 cells and purified 100% humidity, 4 °C) before freeze plunging. Movie data were col- using an optimized CsCl gradient-based purification protocol that lected using a K2 camera (Gatan) attached to a Titan Krios cryo- renders vector preps of high purity and devoid of empty capsids58. electron microscope (Thermo Fisher Scientific) operating at 300 kV Viral genome titers were determined by Picogreen using phage lambda (C2 aperture: 50 mm; magnification: 130,000×; energy-filtered zero- DNA as standard curve. loss imaging mode with slit width 20 eV; pixel size: 1.06 Å; total exposure 50 e− Å−2; 50 frames per movie; dose rate: 6 e− px−1 s−1) (Sup- bGH mice plementary Fig. 5a–c). Bovine growth hormone transgenic mice were generated and identi- fied as previously described59. Briefly, the bGH were generated using a CryoEM 3D reconstruction metallothionein transcriptional regulatory element linked to the first 3025 movies underwent motion-correction and dose-weighting using exon and intron of the bGH cDNA. This construct was injected into the the patch-motion correction job within cryoSPARC v3.264, followed by pronucleus of C57BL/6J embryos. Mice containing the transgene were CTF parameter estimation using the patchCTF job (Supplementary identified with PCR analysis of DNA from tail biopsy specimens Fig. 6). Micrograph quality was then assessed, with only micrographs obtained 4 weeks after birth. The primers used for genotyping are the comprising thin vitreous ice, CTF fit better than 4.3 Å and low motion Nature Communications | (2022)13:6700 11 Article https://doi.org/10.1038/s41467-022-34391-6 curvature retained, giving a set of 2426 good micrographs. Particle 6. Best, C. H. The internal secretion of the pancreas. Can. Med. Assoc. coordinates were determined using crYOLO 1.8.0-beta65, giving 1.36 M J. 87, 1046–1051 (1962). particles, which were re-imported, binned 4× and extracted with a 64 7. Rinderknecht, E. & Humbel, R. E. Primary structure of human insulin- px box size. The binned particles were 2D classified into 200 classes, like growth factor II. FEBS Lett. 89, 283–286 (1978). and 410k particles from 45 good 2D classes were retained. Junk parti- 8. Rinderknecht, E. & Humbel, R. E. The amino acid sequence of cles were removed by heterogeneous refinement into two classes, with human insulin-like growth factor I and its structural homology with a noisy initial model employed as a sink for poor particles. 255k par- proinsulin. J. Biol. Chem. 253, 2769–2776 (1978). ticles corresponding to the good class were extracted at full resolution 9. Kasik, J. W., Lu, C. & Menon, R. K. The expanding insulin family: with a box size of 360 px, and 3D classified into three classes using the structural, genomic, and functional considerations. Pediatr. Dia- ab initio reconstruction job. The most complete class, containing 89k betes 1, 169–177 (2000). particles, was then reconstructed using non-uniform refinement to 10. Malide, D., Seidah, N. G., Chretien, M. & Bendayan, M. Electron give a nominally 4.6 Å resolution map. All classification and recon- microscopic immunocytochemical evidence for the involvement of struction steps were carried out within cryoSPARC v3.2. Further pro- the convertases PC1 and PC2 in the processing of proinsulin in cessing (viz., 3D classification in cryoSPARC or Relion 3, per- pancreatic beta-cells. J. Histochem. Cytochem. 43, 11–19 (1995). micrograph and per-particle CTF refinement and local refinements) 11. Shabanpoor, F., Separovic, F. & Wade, J. D. The human insulin did not result in improved reconstructions. superfamily of polypeptide hormones. Vitam. Horm. 80, 1–31 (2009). 12. Massague, J., Pilch, P. F. & Czech, M. P. Electrophoretic resolution of Model building three major insulin receptor structures with unique subunit stoi- An atomic model of the scLCDV1-VILP-bound IGF1Rzip was built and chiometries. Proc. Natl Acad. Sci. USA 77, 7137–7141 (1980). refined as follows. First, individual copies of receptor modules L1-CR, 13. Kasuga, M., Karlsson, F. A. & Kahn, C. R. Insulin stimulates the L2, FnIII-1, and FnIII-2 were extracted from PDB entry 5U8R and docked phosphorylation of the 95,000-dalton subunit of its own receptor. manually into the map using ChimeraX55. The model was then rigid- Science 215, 185–187 (1982). body fit using Phenix version 1.19.2-4158-00066, treating each of the 14. Chernausek, S. D., Jacobs, S. & Van Wyk, J. J. Structural similarities above structural modules as individual rigid entities. Two copies of the between human receptors for somatomedin C and insulin: analysis extended αCT segment (extracted from PDB entry 6PHY) and of the by affinity labeling. Biochemistry 20, 7345–7350 (1981). scLCDV1-VILP moieties (generated with AlphaFold236) were then added 15. Bathgate, R. A. et al. Relaxin family peptides and their receptors. to the model based on their discerned location within the map. Manual Physiol. Rev. 93, 405–480 (2013). editing with secondary structure, torsion and Ramachandran restraints 16. Seino, S., Seino, M., Nishi, S. & Bell, G. I. Structure of the human was undertaken using Coot version 0.967 to regularize inter-domain insulin receptor gene and characterization of its promoter. Proc. junctions and to generate a plausible conformation for the C peptide Natl Acad. Sci. USA 86, 114–118 (1989). of each scLCDV1-VILP. A final rigid-body fit with the two receptor 17. Frasca, F. et al. Insulin receptor isoform A, a newly recognized, high- monomers and two scLCDV1-VILP molecules as four rigid entities was affinity insulin-like growth factor II receptor in fetal and cancer cells. then undertaken. Refinement statistics are in supplementary Table 1. Mol. Cell Biol. 19, 3278–3288 (1999). Map density in Fig. 5a, b was generated from the final reconstruction 18. Sciacca, L. et al. Insulin analogues differently activate insulin using the ‘volume zone’ tool in ChimeraX 1.3 with an atomic distance of receptor isoforms and post-receptor signalling. Diabetologia 53, 2.5 Å. Map vs model FSC curve can be found in supplementary Fig. 4f. 1743–1753 (2010). 19. Pandini, G. et al. Insulin/insulin-like growth factor I hybrid receptors Reporting summary have different biological characteristics depending on the insulin Further information on research design is available in the Nature receptor isoform involved. J. Biol. Chem. 277, 39684–39695 (2002). Research Reporting Summary linked to this article. 20. Xu, Y. et al. How ligand binds to the type 1 insulin-like growth factor receptor. Nat. Commun. 9, 821 (2018). Data availability 21. Kavran, J. M. et al. How IGF-1 activates its receptor. Elife 3, CryoEM map and atomic model of scLCDV1-VILP bound to IGF1R.zip e03772 (2014). generated in this study have been deposited in the EMDB and PDB 22. Lagueux, M., Lwoff, L., Meister, M., Goltzene, F. & Hoffmann, J. A. databases, respectively, under the accession codes EMD-26306 and cDNAs from neurosecretory cells of brains of Locusta migratoria 7U23. The modules L1-CR, L2, FnIII-1, and FnIII-2 were extracted from (Insecta, Orthoptera) encoding a novel member of the superfamily PDB entry 5U8R. The two copies of the extended αCT segment were of insulins. Eur. J. Biochem. 187, 249–254 (1990). extracted from PDB entry 6PHY. Source data are provided with 23. Altindis, E. et al. Viral insulin-like peptides activate human insulin this paper. and IGF-1 receptor signaling: a paradigm shift for host-microbe interactions. Proc. Natl Acad. Sci. USA 115, 2461–2466 (2018). References 24. Chinchar, V. G. et al. ICTV virus taxonomy profile: iridoviridae. J. 1. Chassin, D., Laurent, A., Janneau, J. L., Berger, R. & Bellet, D. Cloning Gen. Virol. 98, 890–891 (2017). of a new member of the insulin gene superfamily (INSL4) expressed 25. Norman, J. M. et al. Disease-specific alterations in the enteric virome in human placenta. Genomics 29, 465–470 (1995). in inflammatory bowel disease. Cell 160, 447–460 (2015). 2. Hsu, S. Y. Cloning of two novel mammalian paralogs of relaxin/ 26. Breitbart, M. et al. Viral diversity and dynamics in an infant gut. Res. insulin family proteins and their expression in testis and kidney. Mol. Microbiol. 159, 367–373 (2008). Endocrinol. 13, 2163–2174 (1999). 27. Dinakaran, V. et al. Elevated levels of circulating DNA in cardio- 3. Lok, S. et al. Identification of INSL6, a new member of the insulin vascular disease patients: metagenomic profiling of microbiome in family that is expressed in the testis of the human and rat. Biol. the circulation. PLoS ONE 9, e105221 (2014). Reprod. 62, 1593–1599 (2000). 28. Chrudinova, M. et al. Characterization of viral insulins reveals white 4. Conklin, D. et al. Identification of INSL5, a new member of the adipose tissue-specific effects in mice. Mol. Metab. 44, insulin superfamily. Genomics 60, 50–56 (1999). 101121 (2021). 5. Bathgate, R. A. et al. Human relaxin gene 3 (H3) and the equivalent 29. Zhang, F., Altindis, E., Kahn, C. R., DiMarchi, R. D. & Gelfanov, V. A mouse relaxin (M3) gene. Novel members of the relaxin peptide viral insulin-like peptide is a natural competitive antagonist of the family. J. Biol. Chem. 277, 1148–1157 (2002). human IGF-1 receptor. Mol. Metab. 53, 101316 (2021). Nature Communications | (2022)13:6700 12 Article https://doi.org/10.1038/s41467-022-34391-6 30. Bennett, D. L. et al. Identification of the type 2 proinsulin processing 52. Neumann, S., Krieger, C. C. & Gershengorn, M. C. Targeting TSH endopeptidase as PC2, a member of the eukaryote subtilisin family. and IGF-1 receptors to treat thyroid eye disease. Eur. Thyroid J. 9, J. Biol. Chem. 267, 15229–15236 (1992). 59–65 (2020). 31. Cai, W. et al. Domain-dependent effects of insulin and IGF-1 53. Simpson, A., Petnga, W., Macaulay, V. M., Weyer-Czernilofsky, U. & receptors on signalling and gene expression. Nat. Commun. 8, Bogenrieder, T. Insulin-like growth factor (IGF) pathway targeting in 14892 (2017). cancer: role of the IGF axis and opportunities for future combination 32. Gross, S. M. & Rotwein, P. Unraveling growth factor signaling and studies. Target Oncol. 12, 571–597 (2017). cell cycle progression in individual fibroblasts. J. Biol. Chem. 291, 54. Edgar, R. C. MUSCLE: multiple sequence alignment with high 14628–14638 (2016). accuracy and high throughput. Nucleic Acids Res. 32, 33. Palmer, A. J. et al. Age-related changes in body composition of 1792–1797 (2004). bovine growth hormone transgenic mice. Endocrinology 150, 55. Goddard, T. D. et al. UCSF ChimeraX: meeting modern challenges 1353–1360 (2009). in visualization and analysis. Protein Sci. 27, 14–25 (2018). 34. Ding, J., Berryman, D. E. & Kopchick, J. J. Plasma proteomic profiles 56. Whitten, A. E. et al. Solution structure of ectodomains of the insulin of bovine growth hormone transgenic mice as they age. Transgenic receptor family: the ectodomain of the type 1 insulin-like growth Res. 20, 1305–1320 (2011). factor receptor displays asymmetry of ligand binding accompanied 35. Xu, Y. et al. How IGF-II binds to the human type 1 insulin-like growth by limited conformational change. J. Mol. Biol. 394, factor receptor. Structure 28, 786–798.e786 (2020). 878–892 (2009). 36. Jumper, J. et al. Highly accurate protein structure prediction with 57. Elleman, T. C. et al. Mutational analysis of the N-linked glycosylation AlphaFold. Nature 596, 583–589 (2021). sites of the human insulin receptor. Biochem. J. 347, 37. Li, J., Choi, E., Yu, H. & Bai, X. C. Structural basis of the activation of 771–779 (2000). type 1 insulin-like growth factor receptor. Nat. Commun. 10, 58. Ayuso, E. et al. High AAV vector purity results in serotype- and 4567 (2019). tissue-independent enhancement of transduction efficiency. Gene 38. Gutmann, T. et al. Cryo-EM structure of the complete and ligand- Ther. 17, 503–510 (2010). saturated insulin receptor ectodomain. J. Cell Biol. 219, 59. Berryman, D. E. et al. Comparing adiposity profiles in three mouse e201907210 (2020). models with altered GH signaling. Growth Horm. IGF Res. 14, 39. Gauguin, L. et al. Alanine scanning of a putative receptor binding 309–318 (2004). surface of insulin-like growth factor-I. J. Biol. Chem. 283, 60. Coschigano, K. T. et al. Deletion, but not antagonism, of the mouse 20821–20829 (2008). growth hormone receptor results in severely decreased body 40. Yang, K. et al. Alterations in the gut virome in obesity and type 2 weights, insulin, and insulin-like growth factor I levels and diabetes mellitus. Gastroenterology https://doi.org/10.1053/j. increased life span. Endocrinology 144, 3799–3810 (2003). gastro.2021.06.056 (2021). 61. O’Shea, E. K., Klemm, J. D., Kim, P. S. & Alber, T. X-ray structure of 41. Zhang, W., Gustafson, T. A., Rutter, W. J. & Johnson, J. D. Positively the GCN4 leucine zipper, a two-stranded, parallel coiled coil. Sci- charged side chains in the insulin-like growth factor-1 C- and ence 254, 539–544 (1991). D-regions determine receptor binding specificity. J. Biol. Chem. 62. Hoyne, P. A. et al. High affinity insulin binding by soluble insulin 269, 10609–10613 (1994). receptor extracellular domain fused to a leucine zipper. FEBS Lett. 42. Geddes, S. et al. Structure-function studies of an IGF-I analogue 479, 15–18 (2000). that can be chemically cleaved to a two-chain mini-IGF-I. Protein 63. Weis, F. et al. The signalling conformation of the insulin receptor Eng. 14, 61–65 (2001). ectodomain. Nat. Commun. 9, 4420 (2018). 43. Gill, R. et al. Engineering the C-region of human insulin-like growth 64. Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoS- factor-1: implications for receptor binding. Protein Eng. 9, PARC: algorithms for rapid unsupervised cryo-EM structure deter- 1011–1019 (1996). mination. Nat. Methods 14, 290–296 (2017). 44. Guo, Z. Y., Tang, Y. H., Wang, S. & Feng, Y. M. Contribution of the 65. Wagner, T. et al. SPHIRE-crYOLO is a fast and accurate fully auto- absolutely conserved B8Gly to the foldability of insulin. Biol. Chem. mated particle picker for cryo-EM. Commun. Biol. 2, 218 (2019). 384, 805–809 (2003). 66. Afonine, P. V. et al. Real-space refinement in PHENIX for cryo-EM 45. Forbes, B. E., Blyth, A. J. & Wit, J. M. Disorders of IGFs and and crystallography. Acta Crystallogr. D. Biol. Crystallogr. 74, IGF-1R signaling pathways. Mol. Cell Endocrinol. 518, 531–544 (2018). 111035 (2020). 67. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and 46. Clemmons, D. R. et al. Evaluation of acromegaly by radio- development of Coot. Acta Crystallogr. D. Biol. Crystallogr. 66, immunoassay of somatomedin-C. N. Engl. J. Med. 301, 486–501 (2010). 1138–1142 (1979). 47. Grimberg, A. Mechanisms by which IGF-I may promote cancer. Acknowledgements Cancer Biol. Ther. 2, 630–635 (2003). This work was supported by the National Institute of Diabetes and 48. Shapiro, E. T. et al. Tumor hypoglycemia: relationship to high Digestive and Kidney Diseases of the National Institutes of Health under molecular weight insulin-like growth factor-II. J. Clin. Invest. 85, award number R01DK031036 (to C.R.K.) and K01DK117967 (to E.A.). We 1672–1679 (1990). thank Ms Mai Margetts and Dr Yibin Xu (WEHI, Parkville) for the provision 49. Werner, H. Tumor suppressors govern insulin-like growth factor of an aliquot of IGF1Rzip for use in these studies and Prof. Ken Siddle signaling pathways: implications in metabolism and cancer. (University of Cambridge) for providing MCL’s laboratory with the Oncogene 31, 2703–2714 (2012). hybridomas used to express mAb 26-60 and mAb 83-7. Stable cell lines 50. Qu, X. et al. Update of IGF-1 receptor inhibitor (ganitumab, dalotu- expressing receptor ectodomains were obtained from CSIRO (Parkville, zumab, cixutumumab, teprotumumab and figitumumab) effects on Australia) by WEHI under a Technology Transfer Agreement. WEHI cancer therapy. Oncotarget 8, 29501–29518 (2017). receives Victorian State Government Operational Infrastructure Support 51. Li, J. et al. Inhibitor of IGF1 receptor alleviates the inflammation and funding from the Australian NHMRC Independent Research Insti- process in the diabetic kidney mouse model without activating tutes Infrastructure Support Scheme. The authors acknowledge the use SOCS2. Drug Des. Devel. Ther. 12, 2887–2896 (2018). of instruments and assistance at the Monash Ramaciotti Centre for Cryo- Nature Communications | (2022)13:6700 13 Article https://doi.org/10.1038/s41467-022-34391-6 Electron Microscopy, a Node of Microscopy Australia and the use of Peer review information Nature Communications thanks the anon- equipment funded by the Australian Research Council grant ymous reviewer(s) for their contribution to the peer review of this LE120100090. work. Peer reviewer reports are available. Author contributions Reprints and permissions information is available at F.M., E.A., and C.R.K. design the study, F.M. and M.C. carried out the http://www.nature.com/reprints experiments, F.M. and C.R.K. analyzed the data, F.M. and C.R.K. wrote the manuscript with the help of M.C.L. and N.S.K. for the part dedicated Publisher’s note Springer Nature remains neutral with regard to jur- to the cryoEM experiment., F.Z., V.G., and R.D.D. purified the VILPs, isdictional claims in published maps and institutional affiliations. E.O.L. and J.J.K. undertook experiments on BGH mice, V.J. and F.B. produced the AAV8, N.S.K. and H.V. undertook cryoEM experiments, Open Access This article is licensed under a Creative Commons N.S.K. undertook cryoEM reconstruction, M.C.L. and N.S.K. undertook Attribution 4.0 International License, which permits use, sharing, molecular model building, C.R.K. supervised the study. F.M. and N.S.K. adaptation, distribution and reproduction in any medium or format, as contributed equally to this manuscript. All authors discussed the data long as you give appropriate credit to the original author(s) and the and contributed to the final manuscript. source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Competing interests indicated otherwise in a credit line to the material. If material is not M.C.L.’s laboratory has a funded agreement with Eli Lilly and Company included in the article’s Creative Commons license and your intended to conduct research not connected to this publication. The remaining use is not permitted by statutory regulation or exceeds the permitted authors declare no competing interests. use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Additional information licenses/by/4.0/. Supplementary information The online version contains supplementary material available at © The Author(s) 2022 https://doi.org/10.1038/s41467-022-34391-6. Correspondence and requests for materials should be addressed to C. Ronald Kahn. Nature Communications | (2022)13:6700 14