Engineering Glucose Responsiveness Into Insulin

Diabetes Volume 67, February 2018 299

Niels C. Kaarsholm, Songnian Lin, Lin Yan, Theresa Kelly, Margaret van Heek, James Mu, Margaret Wu, Ge Dai, Yan Cui, Yonghua Zhu, Ester Carballo-Jane, Vijay Reddy, Peter Zaﬁan, Pei Huo, Shuai Shi, Valentyn Antochshuk, Aimie Ogawa, Franklin Liu, Sandra C. Souza, Wolfgang Seghezzi, Joseph L. Duffy, Mark Erion, Ravi P. Nargund, and David E. Kelley

Diabetes 2018;67:299–308 | https://doi.org/10.2337/db17-0577

Insulin has a narrow therapeutic index, reflected in a small modulate its action in the face of descending (or ascending) margin between a dose that achieves good glycemic control plasma glucose and thus do not change the intrinsically and one that causes hypoglycemia. Once injected, the narrow therapeutic index of exogenously administered insulin. THERAPEUTICS AND PHARMACOLOGY clearance of exogenous insulin is invariant regardless of To address the latter, the notion of a glucose-responsive blood glucose, aggravating the potential to cause hypo- insulin (GRI) was proposed four decades ago (8). Attempts glycemia. We sought to create a “smart” insulin, one that at creating a GRI have sought to incorporate insulin within can alter insulin clearance and hence insulin action in re- a subcutaneous depot matrix containing glucose-sensitive sponse to blood glucose, mitigating risk for hypoglyce- chemical “triggers” that release insulin (9–12), but there is mia. The approach added saccharide units to insulin to a substantial challenge in sufficiently modulating insulin create insulin analogs with affinity for both the insulin re- release across the narrow physiological range of plasma glu- ceptor (IR) and mannose receptor C-type 1 (MR), which cose. An alternative strategy for a GRI, to exploit endogenous functions to clear endogenous mannosylated proteins, a lectin–based clearance, was reported by Zion and Lancaster principle used to endow insulin analogs with glucose responsivity. Iteration of these efforts culminated in the (13), and is the focus of this report. discovery of MK-2640, and its in vitro and in vivo preclinical Lectins recognize and bind carbohydrate domains of properties are detailed in this report. In glucose clamp glycoproteins. Circulating and cell-based lectins function in experiments conducted in healthy dogs, as plasma glu- immune surveillance and as clearance of glycoproteins cose was lowered stepwise from 280 mg/dL to 80 mg/dL, (14,15). Mannose receptor C-type 1 (MR) is the prototypical progressively more MK-2640 was cleared via MR, reduc- member of the mannose receptor family of transmembrane ing by ∼30% its availability for binding to the IR. In dose lectins. Its main function is to recognize endogenous senes- escalations studies in diabetic minipigs, a higher thera- cent proteins tagged for destruction and pathogens iden- peutic index for MK-2640 (threefold) was observed versus tified by their surface glycan and to deliver these for regular insulin (1.3-fold). lysosomal degradation without eliciting an immune response or cytokine release (16–19). MR is an abundant lectin in hepatic sinusoidal endothelial cells and on certain macro- Nearly a century has passed since the discovery of insulin, yet phages and dendritic cells, with a well-conserved homology the remarkable success of insulin therapy has been tempered across species (14). Glucose has a low affinity for MR but by the risk of iatrogenic hypoglycemia, which stands as a nevertheless can compete with binding of other ligands. major barrier to the achievement of tight glycemic control A concept of a GRI that could be engineered into an (1–4). To mitigate risk for hypoglycemia, insulin analogs insulin analog by targeting MR-based clearance is illustrated have been developed with improvements in pharmaco- in Fig. 1. This concept involves creating a chimeric insulin kinetics (PKs) (5–7). However, efforts to improve insulin analog that undergoes a substantial fraction of its clearance PKs do not enable exogenous insulin to autonomously through MR when plasma glucose is within the euglycemic

Merck Research Laboratories, Merck & Co., Inc., Kenilworth, NJ © 2017 by the American Diabetes Association. Readers may use this article as ﬁ Corresponding author: Ravi P. Nargund, [email protected], or David E. long as the work is properly cited, the use is educational and not for pro t, and the Kelley, [email protected]. work is not altered. More information is available at http://www.diabetesjournals .org/content/license. Received 29 May 2017 and accepted 30 October 2017. This article contains Supplementary Data online at http://diabetes .diabetesjournals.org/lookup/suppl/doi:10.2337/db17-0577/-/DC1. 300 Glucose-Responsive Insulin Diabetes Volume 67, February 2018

Figure 1—A concept for GRI is illustrated that is based on adding saccharide modifications to the insulin molecule. The resulting analog has affinity for the IR as well as for an MR, which clears insulin via endocytosis and lysosomal degradation. Glucose competes for binding to the MR. At high ambient glucose, less of the analog is cleared by MR and a larger proportion becomes available for IR pharmacology; conversely, at low glucose, a proportionally larger fraction of the analog is cleared via MR, lowering availability for IR interaction. ICC, insulin-carbohydrate conjugate. or hypoglycemic range, thus lessening availability for interac- of mice was cannulated for perfusion with MK-2640, tion with the insulin receptor (IR), with the opposite direc- and varying concentrations of glucose were added to tion effect at a progressively higher ambient glucose. We the infusion buffer. Liver caudate lobe samples were will describe our initial steps in creating a GRI using lectin- taken and fixed(10%bufferedformalin),andimmuno- based clearance (J. Mu et al., unpublished observations), whereas histochemistry was performed with guinea pig anti- the current report focuses on a single GRI, MK-2640. We insulin antibody (Invitrogen #180067). Digital images describe the preclinical development of MK-2640, includ- were captured, and the percentage of immunoreactive cells ing dog and minipig in vivo studies designed to assess glu- was determined via instrument software (Membrane v9, cose responsivity. Aperio Technologies).

RESEARCH DESIGN AND METHODS In Vivo Evaluations of MK-2640 In Vitro Binding Assays Allanimalprocedureswerereviewedandapprovedbythe MK-2640 binding to IR was determined in a scintillation Merck Research Laboratories (Merck & Co., Kenilworth, NJ) proximity assay with membranes prepared from CHO cells Institutional Animal Care and Use Committee. The Guide overexpressing human, minipig, or dog IR(B), and insulin for the Care and Use of Laboratory Animals was followed in as the competitive ligand (J. Mu et al., unpublished the conduct of the animal studies. Veterinary care was given observations). MK-2640 direct binding to human or dog to any animals requiring medical attention. MR was assessed via surface plasmon resonance using His- Male Yucatan minipigs from the Sinclair Research Center, tagged purified receptors immobilized onto a Biacore CM5 healthy (nondiabetic [ND]) or rendered type 1 diabetic (D) chip. To appraise glucose inhibition of MR binding, the by alloxan injections, underwent placement of two jugular surface plasmon resonance assay was modified to include venous vascular access ports for administration of MK-2640 varying concentrations of glucose or a-methylmannose or recombinant human insulin (RHI; Humulin R, Lilly) (a-MM) in the presence of a fixed concentration of MK- intravenously (i.v.) and for blood sampling. The study used 2640, approximating its Kd value, to determine the glucose minipigs rendered diabetic by alloxan that attained plasma inhibitory IC50 for MK-2640 binding to human or dog MR. glucose levels of ;300 mg/dL. RHI or MK-2640 was coad- Competition binding assays for MR, DC-SIGN, and MBL ministered with PBS or a-MM (9.4 mg/kg/min), the latter a were conducted using mannosylated-BSA, labeled with high-affinityligandforMRthatcanactasachemical europium for fluorescent detection, and varying concentra- blockade of MR (20). Insulin levels were measured by ELISA tions of MK-2640 to determine affinity for the human (RHI) or by liquid chromatography–mass spectrometry lectin receptors. An ex vivo assay of glucose inhibition of for MK-2640. To elucidate glucose-responsiveness of MK-2640 binding in the liver was conducted by perfusing RHI or MK-2640, respective clearances were assessed the compound into the portal vein of mice with varying in D minipigs and compared with corresponding values concentrations of glucose and then imaging liver sections obtained in ND minipigs that had been coinfused with with an anti-insulin antibody. Briefly, the portal vein PBS or a-MM. To appraise therapeutic index, MK-2640 or diabetes.diabetesjournals.org Kaarsholm and Associates 301

RHI was administered subcutaneously (s.c.) to D minipigs reflecting glucose responsive PKs. MK-2640 was tested for with dose escalation (successive 30% dose increments), activation of cytokine release in the plasma of dogs receiv- and plasma glucose monitoring for 8 h (AU480 Clinical ing MK-2640 or RHI infusions. Cytokine levels (interleukin Chemistry Analyzer, Beckman Coulter) until a dose that [IL]-6, IL-10, IL-8, IL-2, and tumor necrosis factor-a)were caused hypoglycemia was identified. Plasma epineph- evaluated by ELISA (Meso Scale Diagnostics, Rockville, MD), rine was assessed by liquid chromatography–mass spec- and dog blood exposed ex vivo to lipopolysaccharide trometry to monitor the hypoglycemic counterregulatory (10 pg/mL) was a positive control. response. Statistical Analysis MK-2640 was examined in healthy 4- to 6-year-old male Data analyses were performed in GraphPad Prism (GraphPad beagle dogs with euglycemic and hyperglycemic clamps to Software, San Diego, CA). Calculations of P values were examine its PK and pharmacodynamic (PD) effects across a based on ANOVA and the unpaired Student t test, which- range of steady-state plasma glucose concentrations. Dogs ever was applicable. Statistical significance was defined as were fasted overnight before a clamp and had previously P , 0.05. been prepared with dual-ported femoral artery and femoral vein cannulations, respectively, for sample collection and RESULTS infusions. Animals received an infusion of somatostatin Design and Synthesis of MK-2640 (0.8 mg/kg/min) during 6 target glucose levels (80, 120, Chemistry efforts focused on conjugating RHI at the amino 160, 200, 240, and 280 mg/dL) to suppress endogenous termini, including A1, B1, and the B29 Lys amino groups, insulin secretion. Each dog completed all clamps. RHI at individually or as bis-functionalization with linkers bearing an infusion rate of 3 pmol/kg/min, was used as a compar- mono-, di-, or trisaccharide units, as shown in Fig. 2A. ator to MK-2640. Pilot studies found an infusion rate for Glucose, mannose, and fucose were selected for exploring MK-2640 of 45 pmol/kg/min achieved PD equipoise with the identification of potential lead GRI candidates, and RHI when clamps were conducted at 200 mg/dL, and these through studies of the structure activity relationship, a respective infusion rates were used in all clamps. PK read- lysine-based, nonsymmetric, mannose based–linker structure outs were RHI and MK-2640 concentrations and clear- was identified that began to demonstrate modest glucose ance. A change in MK-2640 clearance was interpreted as modulation of its PKs. This recognition also helped to define

Figure 2—A: Chemical ligation strategy to synthesize insulin oligosaccharide conjugates is outlined, together with the specific structure of MK-2640. B: The structure of compound 2 (MK-2640 analog with A1 and B29 positions conjugated with moieties possessing non-MR binding morpholines instead of saccharide units). IUPAC, International Union of Pure and Applied Chemistry; NHS, N-hydroxysuccinimide. 302 Glucose-Responsive Insulin Diabetes Volume 67, February 2018 an appropriate range for respective affinities for IR and MR. MR was similar to that observed for the minipig and dog MR, We subsequently observed that the use of fucose in place of as reported in Table 2. mannose increased MR binding affinity and with iterations MK-2640 did not bind to MBL at the highest concen- led to the identification of MK-2640 as an analog with tration tested (10 mmol/L) and had relatively weak affinity promising glucose responsiveness, and its structure is shown forDC-SIGN,withanIC50 of 1,800 nmol/L. Consistent with in Fig. 2A. Additional information on the chemistry effort these findings, MK-2640 did not evoke cytokine release can be accessed in a recently published patent application from human differentiated macrophages (Supplementary and references therein (21). To support the mechanism of Fig. 1). By using increasing concentrations of glucose, as action studies in minipigs, compound 2 (Fig. 2B)wassyn- showninFig.3A, the ability of glucose to compete with thesized by modifications at both A1 and B29 positions the binding of MK-2640 to MR was assessed in vitro, and with morpholines replacing saccharide units that render it an IC50 of 8 mmol/L glucose was determined; a correspond- with similar IR affinity but greatly reduced MR affinity ing value for a “high-affinity” MR–binding small molecule, compared with MK-2640. a-MM, is 0.62 mmol/L. An ex vivo mouse liver perfusion PK Properties of MK-2640 model was used to assess whether glucose affected partitioning of selected analogs between IR and MR in the pres- The PK properties of MK-2640 were evaluated in beagle B dogs and Yucatan minipigs after i.v. and s.c. dosing. As a ence of both receptors. As shown in Fig. 3 , a large fraction preliminary formulation effort, directed at achieving drug of MK-2640 was taken up by MR at a low glucose concen- solubility, MK-2640 was formulated in sodium phosphate tration, and this fraction progressively declined as the glucose concentration was raised to hyperglycemia, with (26.5 mmol/L) buffer (pH = 7.4) containing sodium chloride ; (87 mmol/L) and phenol (21 mmol/L). Dogs were dosed an IC50 value of 9mmol/Lglucose,quitesimilartothe m m IC50 estimated from in vitro binding experiments using i.v. at 2.5 g/kg and s.c. at 15 g/kg. Minipigs were dosed B i.v. at 5 mg/kg and s.c. at 45 mg/kg. After i.v. dosing, immobilized MR (Fig. 3 ). MK-2640 exhibited moderate to high systemic clearance Glucose Responsiveness in Minipigs in dogs and minipigs (8 and 39 mL/min/kg, respectively). MK-2640 PK was evaluated in ND and D Yucatan minipigs The steady state volume of distribution was low (;0.04–0.2 after i.v. dosing in the absence and presence of coinfused L/kg), and the elimination half-life was short (3–14 min) in a-MM. The i.v. administration of RHI led to a short-lived these species. After s.c. administration of MK-2640, absorp- fall of plasma glucose; its PKs and PDs were unaffected by tion was faster in the dogs than in the minipigs (t max a-MM,asshowninFig.4A and C. The PKs and PDs of 18 min vs. 36 min), and the bioavailability varied between MK-2640 in ND minipigs were strongly affected by a-MM 26% and 44%. The biodistribution of MK-2640 was exam- coinfusion. A dose of MK-2640 (0.35 nmol/kg) that had ined in healthy rats using s.c. dosing of [3H]MK-2640, and rapid clearance and only a modest glucose-lowering effect this showed significant amounts of [3H]MK-2640–related in ND minipigs when administered alone produced transient radioactivity detected in liver, kidney, muscle, and fat. hypoglycemia in the presence of a-MM (Fig. 4B), because its Targeting of MR clearance was markedly protracted (Fig. 4D). The MK-2640 The binding affinity of MK-2640 for the IR, as presented in PK differences observed with and without a-MM served Table 1, was substantially lower than RHI (;4% in com- to demarcate potential boundaries of a corresponding max- parison), which suggests that the potential for full clinical imal glucose responsiveness because these PK differences development of MK-2640 might be limited due to low po- delineate the contribution of MR-mediated clearance (at tency. However, the potential for glucose responsiveness was euglycemia). Glucose-responsive PKs were then explored regarded as important to interrogate in vivo. In vitro, at the by giving RHI and MK-2640, each as an i.v. bolus to D highest concentrations of MK-2640 examined, the compound minipigs. MK-2640 has been shown to dose-dependently completely blocked binding of native insulin. Binding to human lower glucose in D minipigs when administered i.v. (see Supplementary Fig. 2), and a dose (0.35 nmol/kg) with minimal glucose-lowering was used to evaluate its PKs dur- — fi Table 1 Cell membrane IR-binding af nities ing stable hyperglycemia. As shown for RHI (Fig. 4C)and IR binding MK-2640 (Fig. 4D), hyperglycemia in D minipigs did not Human Minipig Dog change clearance of RHI versus ND, whereas the clearance

IC50 IC50 IC50 for MK-2640 in D minipigs was decreased. This difference (nmol/L)SD (nmol/L)SD (nmol/L) SD in clearance of MK-2640 in D versus ND minipigs was inter- RHI 0.48 1.5 0.39 1.8 0.44 1.3 preted as highly similar to the PKs change observed with MK-2640 7.0 2.1 7.4 1.2 8.5 1.1 blockade of MR clearance by a-MM in ND minipigs. Ad- ditional studies demonstrated that a-MM–induced PK and Geometric mean IC50 and geometric SD are from four or more individual MK-2640 titrations for all species. Compound 2 is an PD changes for MK-2640 (Fig. 4B) cannot be ascribed to the analog of MK-2640 wherein the saccharides are replaced by reduced IR potency of MK-2640 per se because compound morpholino groups. The IR human binding activity of compound 2 (0.69 nmol/L i.v. dose), with comparable IR potency as 2 was 5.7 nmol/L. MK-2640 but minimal MR binding afﬁnity, demonstrated diabetes.diabetesjournals.org Kaarsholm and Associates 303

Table 2—MR-binding afﬁnities MR binding (Biacore) MR binding (DELFIA) Human Dog Human Minipig Dog

Kd (nmol/L) SD Kd (nmol/L) SD IC50 (nmol/L) SD IC50 (nmol/L) SD IC50 (nmol/L) SD MK-2640 3.4 1.4 2.7 1.3 28 1.2 23 1.3 16 1.2

Geometric mean Kd and IC50 and geometric SD are from four or more individual MK-2640 titrations for all species. Compound 2 was inactive in the human IR binding assay. DELFIA, dissociation-enhanced lanthanide fluorescent immunoassay. negligible PK differences when dosed in ND minipigs with reduced by ;33% compared with euglycemic conditions, versus without a-MM. and with infusion of a-MM during hyperglycemia, clearance by MR still accounted for a large fraction (;66%) of its Glucose Clamps clearance. To assess glucose responsiveness of MK-2640, glucose clamp studies were performed in healthy dogs clamped at target plasma glucose concentrations of 80, 120, 160, 200, 240, and 280 mg/dL, one level per study, with complete crossover among dogs. Doses of MK-2640 (45 pmol/kg/min) and RHI (3 pmol/kg/min) were used that required similar glucoseinfusionrate(GIR)during clamps conducted at 200 mg/dL (11 6 1vs.106 1 mg/kg/min, respectively). These high rates of matched GIR, approximately fivefold above fasting rates of glucose flux, denote stimulation of glucose utilization within peripheral tissues consistent with the rat biodistribution data. Stepwise increases in plasma concentration of MK-2640 and corresponding decreases in clearance were observed with increased plasma glucose during clamps. The change in PKs of ;30% across this glycemic range is shown in Fig. 5A. RHI plasma concentrations remained unaffected by glycemic level, and RHI clearance (23 6 3 mL/kg/min) remained constant (Fig. 5B). For MK-2640, the reduction of GIR required to maintain target levels of glycemia at less than 200 mg/dL declined to a greater extent than in the corresponding clamps conducted with RHI. At a plasma glucose of 80 mg/dL, the steady state GIR was 4 6 1mg/kg/minforRHIand26 1mg/kg/min for MK-2640 (P , 0.05). Two-hour infusions of MK-2640 did not evoke changes in cytokines and were not different from the responses observed with RHI, as shown in Supplementary Fig. 3. Partitioning of MK-2640 to Uptake by MR in Dogs When compared during clamp studies conducted at a plasma glucose of 200 mg/dL, the infusion rate of MK-2640 (45 pmol/kg/min) needed to achieve PD equipoise with a 3 pmol/kg/min infusion of RHI was 15-fold greater. This difference can partly be attributed to the lower in vitro potency of MK-2640 for the IR. However, it is potentially important to take into account the percentage of MK-2640 Figure 3—Glucose modulation of MK-2640 binding to MR. A:Data partitioned to degradation via MR-mediated uptake. Clamp from an MR Biacore binding assay are shown in which titrating studies were conducted in dogs with and without a con- glucose concentrations are used in the constant presence of 4 nmol/L MK-2640 (;K concentration added to the running buffer). comitant infusion of a-MM (9.65 mg/kg/min). At a plasma d A glucose IC50 of ;8 mmol/L was determined (compared with IC50 glucose of 80 mg/dL, infusion of a-MM reduced the clear- ;0.62 mmol/L for a-MM). B: Ex vivo mouse liver perfusion assay ance of MK-2640 by 81%, delineating that a high fraction of where an increasing concentration of glucose was added to the buffer fi its clearance was via MR, leaving the remaining 19% to be containing a xed concentration of MK-2640. The MK-2640 taken up by the liver was quantified by immunohistochemistry using an anti- cleared by other pathways, presumably by IR. At a plasma insulin antibody capable of detecting MK-2640. IOC, insulin oligosac- glucose of 300 mg/dL, the overall clearance of MK-2640 was charide conjugate. 304 Glucose-Responsive Insulin Diabetes Volume 67, February 2018

Figure 4—PDs and PKs after an i.v. bolus of RHI (0.17 nmol/kg) (A and C) and MK-2640 (0.35 nmol/kg) (B and D) given to ND and D minipigs with and without coinfusion of a-MM (ND only), which provides a short-term chemical blockade of MK-2640 clearance by MR. A and B:The effect of the two insulins is shown in ND minipigs with and without a-MM. C and D: The PK differences and the effects of a-MM in ND and D minipigs are compared. For panels A–D, data is presented as mean 6 SEM. Areas under the curve0-∞ for the different MK-2640 groups are indicated in panel D (as mean 6 SD).

Dose Escalation to Hypoglycemia dose (Fig. 6D). By these criteria, the therapeutic index for Dose escalations (30% successive increases) with s.c. ad- MK-2640 can be estimated as 3:1 relative to its MED. ministration of MK-2640 and RHI in D minipigs were Subchronic Dosing of MK-2640 in D Minipigs performed to determine the effect of upward titrations of Proof of concept that sustained insulin therapy can dose to lower plasma glucose and to thereby identify a dose maintain reduction of hyperglycemia is well established. that caused hypoglycemia. To estimate the therapeutic However, given the novelty of a mannosylated insulin such index of RHI and MK-2640, the ratio of a dose that caused as MK-2640, subchronic studies in D minipigs were under- hypoglycemia (definedasavalueof55mg/dL),relativeto taken to assess stability of glycemic control. Thrice-daily the minimally efficacious dose (MED), was calculated; MED injections of MK-2640 or RHI were administered to D was defined as the dose at which at least 50% of the ani- minipigs for 2 weeks after animals were withdrawn from mals reach glucose levels ,120 mg/dL. As shown in Fig. 6A, their usual long-acting insulin therapy. Repeated dosing of s.c. administration of RHI led to a rapid decrease in plasma MK-2640 achieved control of hyperglycemia during the glucose levels, with a MED of 0.41 nmol/kg. Further 2 weeks of dosing that was comparable to that attained increases above the MED for RHI resulted in a fall of blood with RHI, as shown in Supplementary Fig. 4. Repeated glucose levels into the hypoglycemic range. At a dose of RHI dosing of MK-2640 was well tolerated, did not cause local of 0.53 nmol/kg, a robust counterregulatory epinephrine reactions at injection sites, and did not elicit the generation response occurred, with a 150% increase over baseline val- of anti-drug antibodies or anti-insulin neutralizing anti- ues observed at hypoglycemia (Fig. 6C); this RHI dose was bodies (Supplementary Table 1). only a 1.3-fold increment over the MED. MK-2640 caused a dose-dependent decrease in plasma glucose with an MED of Formulation Efforts to Achieve a Basal Insulin Profile 2.1 nmol/kg (Fig. 6B), and hypoglycemia caused by MK-2640 The reduced in vitro potency of MK-2640 and its consequent was not observed until the 8.0 nmol/kg dose, a nearly four- relatively slow onset of action, even when administered fold therapeutic index. A counterregulatory response of epi- i.v., fitaprofile that seemed better suited as a long- nephrine was, however, observed with an MK-2640 dose of acting basal insulin, rather than as a rapid-acting insulin, 6.1 nmol/kg and was more pronounced at the 8.0 nmol/kg because the latter requires high in vitro potency. However, diabetes.diabetesjournals.org Kaarsholm and Associates 305

glucose-responsive manner and hence its availability for insulin action. The innovation was based on creating an insulin chimera by attaching saccharides to native insulin that would imbue capacity to bind to the lectin receptor MR yet retain capacity for alternatively binding to the IR. Upon binding to MR, the insulin analog MK-2640, like other ligands for MR, is taken up and degraded in lysosomes. This fate precludes interaction with the IR. Glucose responsiveness does not ensue per se from a capacity for binding to MR; rather, MR binding opens a pathway of insulin egress that does not evoke insulin action. Creating glucose responsiveness required that the interaction with MR could in turn be modulated by the level of plasma glucose. MK-2640 binding by MR in vitro is increased at low concentrations of glucose and vice versa. Liver perfusion studies also revealed that uptake by the MR pathway is modulated by ambient glucose in a linear response and with an IC50 of ;9mmol/L glucose very similar to that identified from in vitro binding studies. The results of these assays suggested that the clearance and degradation of MK-2640 by the MR pathway could have potential to be the greatest when ambient glucose is reduced, when it will be desirable to lower circulating insulin and thereby to hopefully mitigate risk for insulin- induced hypoglycemia. From the outset of the protein engineering effort, the balance of the affinity for IR relative to MR would clearly be an essential parameter guiding construction of GRI candi- date molecules. Attachment of saccharides to the insulin backbone enabled interaction with MR, and with iterations of saccharide selection, this interaction with MR could be tuned to respond to ambient glucose. Attachment of saccharides to the insulin backbone, the site of attachments (e.g., A1, B1, and/or B29 of insulin), and the length and Figure 5—Plasma drug concentrations for infused insulins were composition of the linker joining a saccharide to insulin also measured during steady-state conditions of a glucose clamp pro- cedure performed in healthy dogs using a constant rate of infusion affected analog potency on the IR (21). MK-2640 was iden- of 45 pmol/kg/min for MK-2640 (A) and an infusion of 3 pmol/kg/min tified as an analog that might have a desirable balance of for RHI (B). The target plasma glucose concentrations for each of the respective binding affinities for IR and MR, and the glucose x clamp studies are shown on the -axis. A crossover study design responsiveness of MK-2640 derives from modulation of its was used, by which dogs participated in each of the clamps, and a glucose-dependent increase in MK-2640 plasma concentration was PKs because changes in ambient glucose do not modify the observed, consistent with a glucose-responsive modulation of its intrinsic potency of MK-2640 for the IR. The diminished clearance, whereas no effect on the clearance of RHI was observed. in vitro potency of MK-2640 for the IR likely would posi- 6 Data are mean SEM. tion this or similar GRI analogs to find usefulness as a basal rather than as a rapid-acting insulin because the latter to achieve a sustained PK profile, it was necessary to advance needs full in vitro potency so that its onset of action is beyond the preliminary formulation used in the PK and PD rapid. Although the reduced IR potency of MK-2640 posi- studies described above and explore development of a formu- tions it as a potential basal insulin, additional effort to lation of MK-2640 that could manifest slow absorption achieve a formulation that would slow its absorption kinet- kinetics over hours rather than minutes. Exploratory formu- ics would ultimately need to be achieved. In the current lation work was begun to assess whether slow absorption report, preliminary formulations efforts with MK-2640 in- kinetics could be achieved for MK-2640. The preliminary dicated that slowing its absorption kinetics might be feasi- results are shown in Supplementary Fig. 5, suggesting that ble, but the in vivo data presented for MK-2640 in dogs and with further refinement, a basal insulin preparation could minipigs, whether administered s.c. or by i.v. injection, used be achieved for MK-2640 or other similar GRI candidates. asimpler“fit for purpose” formulation that achieved good solubility of the analog and that was associated with rela- DISCUSSION tively prompt absorption kinetics after s.c. administration. The synthesis of MK-2640 was undertaken in an effort The hypothesis that MK-2640 might have a desirable to create an analog with a capacity to alter its PKs in a balanceofrespectivebindingaffinities for IR and MR 306 Glucose-Responsive Insulin Diabetes Volume 67, February 2018

Figure 6—Plasma glucose levels (A and B) and changes in plasma epinephrine levels (C and D) in D minipigs after s.c. administration of RHI (A and C) and MK-2640 (B and D) in escalating doses of a 30% increment over the preceding dose until the dose yielding a hypoglycemic response was observed. The shadowed area in A and B (55–120 mg/dL) deﬁnes the euglycemia range for the minipigs. The shadowed areas in C and D represent the normal range of variance for plasma concentrations of epinephrine under resting conditions in the D minipigs. Data are mean 6 SEM.

was tested in the dog and minipig, two preclinical species minipigs. The clearance of MK-2640 in ND minipigs was commonly used in the development of novel insulins. In greatly impeded by an infusion of a-MM, a finding recapit- both species, MK-2640 was shown to demonstrate glucose- ulated in the setting of marked hyperglycemia of D mini- responsive PKs and PDs. The MK-2640 infusion studies pigs. Importantly, dose escalation studies in D minipigs conducted in healthy dogs, using glucose clamp methods suggested a safer therapeutic index for MK-2640 compared conducted across a range of target glucose concentrations, with regular insulin. The working definition of therapeutic focused on PK changes and demonstrated that ambient index for insulin used in our studies was simple: it was to glucose modulates in a linear manner the clearance of compare the dose that induced hypoglycemia relative to one MK-2640, increasing its clearance at lower plasma glucose. that reduced fasting hyperglycemia to a desirable target of Collateral studies conducted during coinfusion of a-MM, a 120 mg/dL. In the comparison of MK-2640 to regular insulin, high-affinity ligand for MR that provides chemical blockade the MK-2640 dose could be increased by threefold above the of MK-2640 clearance by MR, revealed that at euglycemia, MED before hypoglycemia ensued or hypoglycemic counter- ;80% of MK-2640 clearance is accounted for by MR-based regulation was evoked. For regular insulin, the dose escalation clearance. Thus, a substantial portion of the MK-2640 dose was an order of magnitude smaller, a 30% increase above the was cleared via MR in the euglycemic clamp. On one hand, MED caused hypoglycemia and a counterregulatory response. the considerable flux of MK-2640 through MR offers the Insulin-induced hypoglycemia is the most serious acute potential for meaningful plasma PK shifts upon glucose complication of using this medicine (2,3). Innovations in modulation of MR binding. On the other hand, whether insulin chemistry have yielded basal insulin preparations it is practical to have an insulin analog that undergoes with more consistent duration of the absorptive phase this proportion of degradation independent of evoking in- and flatter PK profiles, while efforts on prandial insulin sulin action and whether it is desirable to chronically engage continue to strive for rapid onset of action (6,7). Yet, these the MR pathway to this extent by a drug administered daily insulin analogs continue to manifest a narrow therapeutic clearly will require more investigation. index and do not per se contain a capacity to modify insulin Glucose responsiveness was also clearly revealed in action once the compound has been administered. The as- the studies using MK-2640 in ND and alloxan-induced D piration of a GRI is to engineer into an insulin preparation a diabetes.diabetesjournals.org Kaarsholm and Associates 307

“postinjection” flexibility that can modulate insulin avail- to these, a third modulus of control, that of glucose- ability based on changing needs, as dictated by changes in responsive variance of insulin elimination kinetics once a blood glucose. Some of the approaches that have been ex- GRI is already circulating in plasma, that a potential for plored or are still being explored for engineering GRI center increased efficacy and safety can be introduced into insulin on introducing glucose-dependent flexibility in the release therapy. of insulin from its injected depot (9,10,12). The approach we have undertaken with the compound MK-2640 is to introduce a capacity for glucose responsiveness onto the Acknowledgments. The authors thank many Merck colleagues for their backbone of native insulin itself to enable its uptake and fi contributions to this project, including Raul Camacho, Carlos G. Rodriguez, and degradation by the MR pathway. In this speci c regard, it is Joel Mane for technical and scientific support during in vivo studies, Xiaoping interesting to consider that a major site for expression of Zhang and Steven Williams for IR-binding data, and Xun Shen for the MR is on hepatic sinusoidal endothelial cells that are closely MR-binding data. The authors wish to acknowledge Hsuan-shen Chen, Huaibing adjacent to underlying hepatocytes, cell types separated He, Tina Santos, Ling Xu, Xiaofang Li, Xinchun Tong, and Bernard Choi, of the narrowly by the space of Disse (14,15). More studies are Discovery Bioanalytics Group, Department of Pharmacokinetics, Pharmacody- needed to understand not only the systemic changes in namics and Drug Metabolism, Merck Research Laboratories, Rahway, NJ, for the PKs and PDs of a glucose-responsive analog such as the quantitative analysis of MK-2640, insulin, and epinephrine. The authors also MK-2640 but also the tissue- and organ-specific contribu- acknowledge the valuable contributions of Merck colleagues Nancy Thornberry, tions to glucose responsiveness, notably the contribution of Natarajan Sethumaran, Michael Meehl, and Gus Gustafson for early contribu- the hepatic bed. Although a key site of drug disposal for tions to the project, Paul Carrington for editorial comments, and Alan Cherrington, of Vanderbilt University, for insightful discussions throughout the course of this MK-2640 within hepatic sinusoidal endothelial cells is likely project. due to the high expression of MR in these cells, biodistri- Duality of Interest. All of the authors were full-time employees of Merck bution studies and relatively high rates of GIR stimulated Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, and held during dog clamp studies indicate that MK-2640 is not a stock or stock options in the company during the conduct of these studies. Merck “liver-selective” insulin analog per se. Research Laboratories was the sponsor of this research. No other potential conflicts The intent of the reported studies is to focus on proof of of interest relevant to this article were reported. pharmacology that engineering a novel GRI analog using a Author Contributions. N.C.K., S.L., T.K., M.v.H., J.M., V.R., M.E., R.P.N., strategy of lectin-based glucose tunable clearance is feasible. and D.E.K. interpreted the data. N.C.K., T.K., M.v.H., J.M., E.C.-J., V.R., S.S., and Across the range of plasma glucose from 80 to 300 mg/dL, D.E.K. designed the studies conducted at Merck Research Laboratories. N.C.K., L.Y., the clearance of MK-2640 changed by 30%, which is un- T.K., M.v.H., J.M., M.W., G.D., Y.C., Y.Z., E.C.-J., V.R., P.Z., S.S., A.O., F.L., S.C.S., precedented, but how much glucose-based PK modula- and W.S. researched data. N.C.K., J.M., R.P.N., and D.E.K. wrote the manuscript. S.L., P.H., S.S., V.A., J.L.D., and R.P.N. led the medicinal chemistry research that tion of a GRI analog must be achieved to attenuate risk discovered and synthesized MK-2640. R.P.N. and D.E.K. are the guarantors of this for hypoglycemia is unknown. In the D minipig studies of work and as such had full access to all the data in the study and take responsibility escalating dose until hypoglycemia ensued, a substantially for the integrity of the data and the accuracy of the data analysis. greater therapeutic index was observed for MK-2640 than Prior Presentation. Parts of this study were presented in abstract form at the for RHI (threefold vs. 1.3-fold, respectively). In general, the 25th American Peptide Symposium, Whistler, BC, Canada, 17–22 June 2017. greater the amount of glucose-responsive PK change and the greater the therapeutic index, the greater would seem References a potential to mitigate the risk for hypoglycemia. Whether the degree of glucose responsiveness achieved by MK-2640 in 1. Frier BM. Hypoglycaemia in diabetes mellitus: epidemiology and clinical im- – preclinical studies has sufficient potential to attain a meaning- plications. Nat Rev Endocrinol 2014;10:711 722 2. The Diabetes Control and Complications Trial Research Group. Hypoglycemia in ful reduction risk for hypoglycemia remains unknown. the Diabetes Control and Complications Trial. Diabetes 1997;46:271–286 In summary, proof of pharmacology preclinical studies, 3. UK Hypoglycaemia Study Group. Risk of hypoglycaemia in types 1 and 2 di- together with relevant in vitro assays, are presented that abetes: effects of treatment modalities and their duration. Diabetologia 2007;50: demonstrate preliminary feasibility in engineering glucose 1140–1147 responsivity onto native insulin. The PKs of MK-2640 can 4. Seaquist ER, Anderson J, Childs B, et al. Hypoglycemia and diabetes: a report change by nearly one-third across a physiologically rele- of a workgroup of the American Diabetes Association and the Endocrine Society. vant range of plasma glucose values that would be experi- Diabetes Care 2013;36:1384–1395 enced by an individual with insulin-requiring diabetes. The 5. Bolli GB, DeVries JH. New long-acting insulin analogs: from clamp studies to glucose-dependent decrease in PKs at euglycemia compared clinical practice. Diabetes Care 2015;38:541–543 with hyperglycemia was mediated by increases in the 6. Heinemann L, Muchmore DB. Ultrafast-acting insulins: state of the art. J Di- – clearance of MK-2640 by MR and was associated with abetes Sci Technol 2012;6:728 742 7. Zaykov AN, Mayer JP, DiMarchi RD. Pursuit of a perfect insulin. Nat Rev Drug reduced insulin action on glucose metabolism. The first Discov 2016;15:425–439 locus of control in insulin therapy is cognitive and re- 8. Brownlee M, Cerami A. A glucose-controlled insulin-delivery system: semi- sides with a physician and patient selecting an appropriate synthetic insulin bound to lectin. Science 1979;206:1190–1191 dosage, and the second locus of control with currently 9. Chou DH, Webber MJ, Tang BC, et al. Glucose-responsive insulin activity by available insulin resides in its absorption kinetics because covalent modification with aliphatic phenylboronic acid conjugates. Proc Natl Acad elimination kinetics are invariant. We hope that by adding Sci U S A 2015;112:2401–2406 308 Glucose-Responsive Insulin Diabetes Volume 67, February 2018

10. Guo J, Sun H, Alt K, et al. Boronate-phenolic network capsules with dual re- 16. Kerrigan AM, Brown GD. C-type lectins and phagocytosis. Immunobiology sponse to acidic pH and cis-diols. Adv Healthc Mater 2015;4:1796–1801 2009;214:562–575

11.HuX,YuJ,QianC,etal.H2O2-responsive vesicles integrated with trans- 17. Lee RT, Hsu TL, Huang SK, Hsieh SL, Wong CH, Lee YC. Survey of immune- cutaneous patches for glucose-mediated insulin delivery. ACS Nano 2017;11:613–620 related, mannose/fucose-binding C-type lectin receptors reveals widely divergent 12. Yu J, Zhang Y, Ye Y, et al. Microneedle-array patches loaded with hypoxia- sugar-binding speciﬁcities. Glycobiology 2011;21:512–520 sensitive vesicles provide fast glucose-responsive insulin delivery. Proc Natl Acad Sci 18. Lee SJ, Evers S, Roeder D, et al. Mannose receptor-mediated regulation of U S A 2015;112:8260–8265 serum glycoprotein homeostasis. Science 2002;295:1898–1901 13. Zion TC, Lancaster TL. Conjugate based systems for controlled drug delivery. 19. Martinez-Pomares L. The mannose receptor. J Leukoc Biol 2012;92:1177–1186 Patent application WO2010088294 A1. 5 August 2010 20. Firon N, Ashkenazi S, Mirelman D, Ofek I, Sharon N. Aromatic alpha-glycosides 14. Drickamer K, Taylor ME. Recent insights into structures and functions of C-type of mannose are powerful inhibitors of the adherence of type 1 ﬁmbriated Escherichia lectins in the immune system. Curr Opin Struct Biol 2015;34:26–34 coli to yeast and intestinal epithelial cells. Infect Immun 1987;55:472–476 15. Taylor ME, Drickamer K. Convergent and divergent mechanisms of sugar 21. Lin S, Yan L, Kekec A, et al. Glucose-responsive insulin conjugates. Patent recognition across kingdoms. Curr Opin Struct Biol 2014;28:14–22 WO/2015/051052 A3. 2 July 2015.