Evaluation of anti insulin receptor antibodies as potential novel therapies for human insulin receptoropathy

Biallelic loss-of-function mutations in the insulin receptor (INSR) commonly cause extreme insulin resistance and early mortality. Therapeutic options are limited, but anti-INSR antibodies have previously been shown to activate two mutant receptors. This study evaluated four murine anti-INSR monoclonal antibodies, each capable of stimulating wildtype INSR signaling, against ten known pathogenic INSR mutants and one novel mutation, F248C. All the mutant INSR bound antibody. Eight mutants showed antibody-induced autophosphorylation while co-treatment with antibody and insulin increased maximal phosphorylation. After simultaneous knockdown of mouse Insr and expression of mutant INSR in 3T3-L1 adipocytes two antibodies activated signaling via AKT preferentially over signaling via ERK1/2 for seven mutants. Two antibodies (83-7 and 83-14) stimulated glucose uptake via P193L, S323L, F382V, and D707A mutant INSR, with antibody response greater than insulin response for D707A. These findings suggest that selected monoclonal anti-INSR antibodies elicit sufficient signaling by some mutated INSR to be clinically significant. One sentence summary Anti-insulin receptor monoclonal antibodies can activate selected naturally occurring mutated human insulin receptors, raising the prospect of novel therapy for lethal recessive insulin resistance.


Introduction
Insulin downregulates catabolic and activates anabolic pathways, suppresses apoptosis and promotes mitosis in response to nutrient ingestion by activating a homodimeric receptor tyrosine kinase (1,2). Human loss-of-function mutations in the insulin receptor (INSR) were first reported in 1988 (3,4), since when more than 100 alleles causing severe insulin resistance have been described (5). Biallelic INSR mutations produce extreme congenital insulin resistance, clinically described as Donohue or Rabson Mendenhall syndromes (OMIM#246200 or 262190), both featuring impaired linear growth, soft tissue overgrowth and severe metabolic derangement, with demise usually in the first 3 years of life in Donohue syndrome.
Some INSR mutations impair receptor processing and cell surface expression. Many mutations, however, are well expressed, but exhibit impaired insulin binding, impaired signal transduction, perturbed recycling kinetics, or a combination of these (6). It has been suggested that the signaling defect of mutant receptors at the cell surface might be circumvented by binding of anti-receptor antibodies, and proof of this principle was provided for two mutations, one in a cell culture model and one as solubilised receptor (7,8).
Therapeutic use of antibodies has now become well established. In particular, antibodies have been developed as cancer therapeutics to block receptor signaling (9), although therapeutic antibodies are also increasingly being developed for non-cancer indications (10). Interest in the potential of therapies targeting the INSR has recently rekindled, with inhibitory antibodies in Phase 1 human trials (11), and stimulatory antibodies shown to ameliorate diabetes in rodents (12)(13)(14)(15) and primates (16).
Given these developments, and the high clinical need in recessive insulin receptoropathy, we assessed the effect of monoclonal anti-INSR antibodies (17)(18)(19)(20)(21) on signaling by a series of diseasecausing mutant INSR in cell-based models. We demonstrate that INSR antibodies bind many mutant receptors, triggering autophosphorylation and downstream signaling. For mutant receptors with selective defects in insulin binding, the signaling response to antibody may exceed that elicited by immunoblot in Figures 4 and S2 contained a sample of 3T3-L1 MmINSRKD hINSR WT treated with 10nM insulin.

Western Blot Image Densitometry
Pixel density of grey-scale 16-bit TIFs was determined in ImageJ 1.47v (NIH, USA). The rectangle tool was used to select lanes and the line tool to enclose the peak of interest and subtract background. The magic-wand tool was used to select the peak area and obtain the raw densitometry value. Mean band intensity of total INSRβ, myc-tagged INSRβ, AKT, ERK1/2, GSK3α/β, p70S6K, and calnexin was used to normalise raw densitometry values for phospho-INSRβ, pAKT, pERK1/2, pGSK3α, pp70S6K, and pAS160. Normalised values for phosphorylated targets were scaled to the mean WT INSR response to insulin.

Statistical Analysis
One-way ANOVAs with Tukey's multiple comparisons test were performed with GraphPad Prism 6 (GraphPad Software). Error bars represent SEM or SD as indicated. All experiments were performed at least three times.

Assessment of mutant INSR cell surface expression and antibody binding
Eleven INSR mutations were selected for study (Table S6). Eight were based on evidence of cell surface expression, prioritizing mutations identified in multiple reports to maximize likely Page 6 of 31 availability of patients for future trials. A previously unpublished F248C mutation, identified in a child with Rabson Mendenhall Syndrome, was included opportunistically. The well-studied P1178L tyrosine kinase mutation (25,26), and the L62P mutation which severely impairs processing (27), were added as controls. Figure 1A displays the extracellular INSR mutations mapped onto the crystal structure of the INSR (28). Four mouse monoclonal anti-human INSR antibodies were used, which had previously all been shown to have partial agonist activity at wild type (WT) receptors, but different effects on kinetics and affinity of insulin binding ( Table 1). As Fab fragments of 83-7 and 83-14 were used in determining the crystal structure of insulin-bound INSR (29) and their binding epitopes are known and are displayed in Figure 1B.
Mutations were introduced into the B isoform of the INSR, believed to be the more important isoform for insulin's metabolic actions (30). To enable discrimination of endogenous INSR and human INSR mutants a C-terminal myc-tag was used in the mutant constructs. Tagged mutants were expressed in Chinese hamster ovary (CHO) cells using the Flp-In system, ensuring differences in protein expression are due to differential processing or stability of receptor protein rather than differential mRNA expression. The mutants were well processed to mature β-subunits with the exception of L62P, for which beta subunit was barely detectable. More modest reductions were seen for the previously unstudied F248C and for the P1178L mutation ( Figure 1C). was in keeping with prior reports(31) and L62P was not studied further. The rightward shift for mutants corresponded to expression of mature β-subunits seen by immunoblotting, suggesting that relative shifts reflected differences in receptor expression rather than antibody affinities. Although some mutations are close to the epitope for antibody 83-7, none of the affected residues provide critical antibody contacts. Indeed, no difference in binding of 83-7 to the mutant panel was seen compared to 83-14, which binds to a surface unaffected by the mutations (Figure 1A Table S7, data not shown for non-responsive P1178L). R118C, I119M, and K460E however showed autophosphorylation comparable to WT and so were not studied further. Altered insulin EC50 was discernible only for S323L (Table S7) Table S7), however for S323L the maximal autophosphorylation response to 83-7 and 83-14 was similar to that with insulin, while D707A was activated by antibodies but not insulin.
We next evaluated responses to insulin plus 10nM antibody, based on evidence suggesting this concentration elicits the maximal response (19,21). In the presence of antibodies 83-7 and 83-14 the maximal response of WT and mutant INSRs to insulin was increased without affecting potency (although EC50 values were not precisely determined) ( Figure 2, Table S7). This was observed across all mutant receptors except the kinase-dead P1178L (25,26). Again, antibodies 18-44 and 18-146 elicited smaller effects.

Generation of a novel adipocyte cell model of insulin receptoropathy
To assess antibody-induced signaling downstream from the INSR, an adipocyte model of insulin receptoropathy was generated. A Tet-responsive miR-shRNA selectively targeting murine Insr was transduced into 3T3-L1 pre-adipocytes to generate a stable clone ( Figure 3A). This was transduced with lentiviruses encoding C-terminal myc-tagged WT or mutant hINSR, also controled by Tet-responsive elements ( Figure Figure 3E). The preadipocyte cell lines generated differentiated efficiently into mature adipocytes as evidenced by Oil-red-O staining ( Figure 3F).

Activation of signaling downstream from mutant INSRs by insulin and antibody
Plasma insulin concentration in human insulin receptoropathies usually lies between 0.3 and 3 nM in the fasting state (Table S6), and at least an order of magnitude higher when fed. We focused on an insulin concentration of 10nM, mimicking the fed disease state. WT INSR autophosphorylation was strongly induced by insulin, but was undetectable after receptor knockdown alone (Figure 4). Across the panel of mutants studied several patterns were seen. In D707A receptor-expressing cells, phosphorylation of AKT and its substrates was severely attenuated in response to insulin, while for S323L, F382V, and F248C mutants progressive "escape" from signaling impairment was seen, with lesser impairment of AKT phosphorylation than of receptor autophosphorylation, and only partial inhibition at downstream substrates. P193L and R252C demonstrated similar insulin-induced AKT and AKT substrate phosphorylation to WT receptor.

Effect of Insulin and/or antibody on glucose uptake.
Stimulation of glucose uptake is a key outcome of INSR activation, and this was assessed in the 3T3-L1 model. Parent 3T3-L1 cells and cells harboring the Insr knockdown construct but not treated with DOX displayed similar high levels of insulin-stimulated glucose-uptake ( Figure S1A, B), but insulin did not stimulate uptake in conditional Insr knockdown cells treated with doxycycline ( Figure 5). Cells with endogenous mouse Insr knockdown and WT hINSR re-expression, in contrast, demonstrated only a 1.8 fold increase in glucose uptake upon insulin stimulation ( Figure S1A). The apparently poor response to insulin was due to increased basal glucose uptake in WT receptoroverexpressing cells ( Figure S1B). Basal uptake significantly differed among mutant receptorexpressing cell lines, appearing to reflect mutant receptor function ( Figure S1C).
Despite this loss of dynamic range in the assay, insulin stimulated glucose uptake via WT, P193L, F248C, R252C, and F382V receptors ( Figure 5). Insulin-stimulated uptake was similar in Page 11 of 31 cells expressing P193L, R252C or F382V receptor and those expressing WT receptor, but was significantly reduced in cells expressing the F248C mutant. No significant stimulation of glucose uptake was seen in cells expressing S323L, D707A or P1178L receptors ( Figure 5).
Antibodies 83-7 and 83-14 alone stimulated glucose uptake via P193L, S323L, F382V and D707A receptors, while antibodies 18-44 and 18-146 were again less effective across the full range of mutants ( Figure 5). While the magnitude of antibody-stimulated uptake was less than seen with insulin via WT, P193L, F248C, R252C, and F382V receptors, antibodies 83-7, 83-14 and 18-44 were more effective than insulin at stimulating glucose uptake via D707A. Dual treatment with antibodies plus insulin did not enhance glucose uptake compared to insulin alone acting via WT, P193L, F248C, R252C, and F382V receptors, or antibody alone when acting via S323L and D707A receptors.

Discussion
Autosomal recessive insulin receptoropathies feature failure to thrive, extreme metabolic derangement, and childhood mortality, and response to therapy is poor. Longitudinal studies of genetic receptoropathies suggest a steep relationship between residual INSR function and clinical outcome: loss of 50% human INSR function, as in the parents of infants with Donohue syndrome, does not produce insulin resistance in lean people. Heterozygous dominant negative mutations produce severe insulin resistance, diagnosed peripubertally in females and later in males, and reduce receptor function to 25% or less of WT. The severe recessive receptoropathies that this study focuses on confer greater loss of function, however even with 0-25% residual function a dramatic range of phenotypes is seen, with complete or near complete loss of function producing Donohue syndrome and lethality in infancy, and less extreme loss of function producing Rabson Mendenhall syndrome, in which survival to the second or third decades is normal. Collectively these observations suggest that even modest improvements in receptor signaling in recessive disease may have decisive clinical benefit.

Page 12 of 31
Many pathogenic INSR mutations are known, including more than 100 missense mutations. A subset are expressed at the cell surface but show impaired insulin binding, signal transduction, or internalization and recycling. This subset may be amenable to non conventional activation by antibody. Proof of this principle came from demonstration that two bivalent antibodies stimulated kinase activity of a single solubilized mutant receptor (F382V (7)), and, independently, that one bivalent antibody increased glycogen synthesis acting via a mutant receptor expressed in intact cells (S232L (18)). We extend these limited previous findings with systematic characterization of multiple receptor mutants and antibodies in two cellular systems, with several readouts of physiologically important responses including adipocyte glucose uptake.
One of the mutants assessed, F248C, is novel. It lies close to the well-studied R252C mutant, which is expressed but exhibits impaired internalization after insulin exposure. F248C shows only minor reduction in cell surface expression, but insulin-stimulated receptor autophosphorylation and downstream signaling are severely impaired. Across known mutants our data generally agree with prior studies. Assay of receptor autophosphorylation in CHO cells using immunocapture of myctagged receptor prior to immunoassay demonstrated signaling defects more clearly than phosphotyrosine immunoblotting in the 3T3-L1 overexpression system. This is likely due to the inherently greater dynamic range of immunoassay allied to use of a generic anti-phosphotyrosine antibody, with different degrees of receptor overexpression between the two models possibly also contributing.
We confirmed that S323L and F382V receptors can be partially activated by antibodies, and extended these observations to a wider range of mutants. Previous studies suggest that receptor activation by antibody depends on receptor cross-linking rather than reaction at specific epitopes (20). exert insulin-like activity on primary human adipocytes (21). Differences among antibodies are likely to reflect differences of affinity and/or steric constraints on cross-linking receptors in intact cells.
The mutants showing the largest antibody response were S323L and D707A, both being activated by antibodies similarly to WT receptor, and to a greater extent than by insulin. Such mutants with 'pure' insulin binding defects are particularly attractive therapeutic targets. Other mutants studied in both cell systems (P193L, F248C, R252C, and F382V) all showed some activation of AKT, GSK3, AS160 and glucose uptake by antibodies. In these cases responses were less than for WT receptor or those induced by insulin although it is unclear that the conditions studied optimally model the in vivo environment, and receptor overexpression is likely to have mitigated some functional defects. Testing the therapeutic potential of antibodies against such mutants is warranted in vivo, where antibody signalling may be prolonged compared to insulin signalling due to slower receptor internalisation. Indeed, a previously studied anti insulin receptor antibody showed markedly greater hypoglycemic effects in vivo in wild type animals than had been apparent in cell culture models (14).
Antibodies would be a particularly appealing therapeutic proposition were they to exhibit synergy with insulin in receptor stimulation, amplifying insulin action rather than simply imposing a tonic signal. The current studies have not addressed this in detail, although suggestive evidence for synergic stimulation of WT receptor and some mutant receptors is seen. This was not mirrored by detectable synergistic activation of downstream signaling or metabolic endpoints, possibly because maximal downstream signaling requires only submaximal receptor autophosphorylation. It remains possible that insulin-antibody synergy does exist but was obscured under the conditions of the experiments undertaken, which pragmatically employed relatively high concentrations of insulin and antibody.
Early cellular studies of antibody-induced insulin receptor activation were interpreted as suggesting that antibodies elicit greater downstream responses than expected from low levels of receptor autophosphorylation (17,(34)(35)(36). These observations were later argued to have a methodological basis, hinging on lower sensitivity in detecting tyrosine phosphorylation than Page 14 of 31 downstream signaling (37,38). This is in part because signal amplification is an inherent property of signal transduction cascades. Our observation of apparent "escape" from signaling inhibition in the face of efficient Insr knockdown in 3T3-L1 adipocytes supports this contention, as activation of residual receptors is undetectable directly but is observable downstream due to signal amplification. Importantly, receptor activation by antibodies leads to selective AKT phosphorylation, which is critical for metabolic actions of insulin, with little or no ERK phosphorylation. As activation of the RAS/RAF/MEK/ERK pathway is mitogenic, this is an encouraging property of antibodies for translational purposes, suggesting that they may exert metabolic benefits without undue mitogenic activity. Similar dissociation between activation of AKT and ERK has also been observed following INSR activation by the peptide ligand S597 (39) and in previous studies with anti-receptor antibodies (40). The mechanism underlying such pathway-specific activation is poorly understood, although IRS proteins may be preferentially phosphorylated by plasma membrane-associated receptor (41,42), whereas receptor internalization is required for full ERK activation (42,43). This suggests that differential signaling may relate to cell surface retention of antibody-bound receptors.
In summary, multiple monoclonal antibodies can bind and activate a range of mutated cell surface insulin receptors to a potentially highly clinically significant degree. Prior experience in wild type animals (14) and theoretical considerations argue that effects of anti-INSR antibodies in vivo may be greater than those in cells, and so further studies of mutant human INSR in animal models are warranted. The murine antibodies studied are particularly amenable to this, as their effect will not be attenuated by host immune responses to human antibodies. Later translation to humans will require humanisation of the antibodies, however many current therapeutic antibodies are humanised or chimeric versions of murine antibodies (9), and moreover humanisation has already been reported for a different purpose for 83-14, with no deleterious consequences for wild type receptor binding (44,45). This suggests a practical route to future testing of anti-INSR antibodies in severely unwell infants with recessive lethal receptoropathy where the potential clinical benefits are transformational, and also raises the prospect of later evaluation of antibodies as novel ultralong-acting insulin mimetics in

Conflict of Interest
The authors declare no conflict of interests.

Author Contributions
GVB, KS and RKS designed the experiments. GVB performed the experiments. GVB, KS, and RKS analysed the data. GVB, KS, and RKS and wrote the manuscript. RKS is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation
This study was presented as a poster at the Diabetes UK Professional Conference, Manchester, UK