Insulin prodrugs with programmed chemical release of an intramolecular cross-link

doi:10.21203/rs.3.rs-4891729/v1

Download PDF

Research Article

Insulin prodrugs with programmed chemical release of an intramolecular cross-link

https://doi.org/10.21203/rs.3.rs-4891729/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Basal insulin therapy is conventionally achieved by one of two broad medicinal approaches that entail protracted absorption at an injection site, or through chemical modification that reduces systemic clearance. We report an alternative approach that employs novel insulin prodrug chemistry to slow the onset of glucose lowering in a controlled manner. Reversible attenuation of insulin action is achieved through a series of chemical linkers that structurally constrain the hormone until the point of their programmed chemical degradation. Under physiological conditions, dipeptide-based linkers convert at varying speed to a corresponding diketopiperazines (DKP) to initiate the transition from a constrained insulin to a flexible conformation characteristic of bioactive insulin. The experimental findings establish a path toward improved insulin-mediated glycaemic control through enhanced precision in time of action and compatible with conventional approaches to extend time action, or other therapeutic purposes.

pharmacokinetics

insulin

prodrug

diabetes

diketopiperazine

Insulin remains indispensable as a drug in the treatment of multiple forms of diabetes. Over the course of a century its source has evolved from animal gland extraction to rDNA-based methods. It continues to progress through chemical analogues designed to provide two types of temporal glucose management (Sims et al. 2021; Hirsch et al. 2020; Zaykov, Mayer, and DiMarchi 2016). Meal-time insulin is optimized for rapid absorption and action to address prandial glycaemic demands, whereas basal insulin largely focuses upon endogenous daily glucose production. The design of basal insulin relies on two basic principles that include slowing insulin absorption from the site of injection and reducing systemic insulin clearance. Decreased rate of insulin absorption following subcutaneous injection is a hallmark of zinc-based insulin formulation, such as NHP. Glargine was a precedent-setting analogue that lessened solubility at site of injection in association with zinc to prolong the duration of action. The molecular dynamics at the injection site is also an important component in the sustained action of insulin degludec. To reduce clearance, insulin has been chemically modified with various fatty acids or increased in molecular size by PEGylation or extension with the Fc portion of an antibody. The successful use of these chemical modifications to extend the duration of insulin action is individually exemplified by analogues peg-lispro, efsitora alfa, icodec and degludec (Hansen et al. 2012; Sinha et al. 2014; Rosenstock et al. 2024; Rosenstock and Del Prato 2022; Kjeldsen et al. 2021; Kurtzhals et al. 2023).

Binding of insulin to its receptor and the subsequent biological signalling associated with receptor-mediated endocytosis is a major mechanism in its metabolic clearance from circulation (Tokarz, MacDonald, and Klip 2018; Duckworth 1988; Sonne 1988). The process is directly dependent on insulin affinity for the receptor, and insulin analogues are now intentionally designed to reduce this association, which serves to minimize in vitro potency and in vivo clearance (Moyers et al. 2022; Kjeldsen et al. 2021). Non-covalent association with albumin or covalent fusion with an Fc can counter receptor-mediated clearance by engaging the recycling capability of FcRn receptors (Rosenstock et al. 2024; Liu 2018; Martins et al. 2016; Kurtzhals et al. 2023). Increased functional size of such insulin analogues further decreases clearance, as it serves to inhibit loss via glomerular filtration (Chamberlain and Stimmler 1967; Valera Mora et al. 2003). In the present study, we explored a chemical mechanism that would control insulin clearance through an internal switch from an inactive to active conformation. We employed the inherent tendency of dipeptides to cyclize to diketopiperazines (DKP) under physiological conditions when suitably configured (Figure 1). This is a well characterized reaction in peptide chemistry that can lead to undesired dipeptide deletion during synthesis in organic solvents (Fischer 2003; Wang et al. 2022). It relies upon the propensity of certain dipeptides to assume the cis conformation, rather the more common trans conformation characteristic of canonical amino acids. Shifting of cis-trans equilibrium to the former conformer under aqueous conditions induces dipeptide cyclization, a feature that was exploited in our prodrug design (Fischer 2003; Santos et al. 2005; Gomes, Vale, and Moreira 2007; De and DiMarchi 2008; Lin and Wang 2012; DiMarchi and Kou 2015; Thalluri et al. 2018). The conversion results in the formation of a DKP and the release of the active drug, and in this instance insulin, or an analogue thereof.

The application of dipeptide-based, DKP-forming prodrugs to insulin has been previously explored with fatty acylation of the N-terminal B-chain residues or the side chain of A19 to reversibly control potency (DiMarchi, Kou, and Cheng 2014). Herein we use DKP formation in a different approach to control the onset of insulin action, as the prodrug forces a constrained conformation of insulin. Productive interaction of insulin with its receptor requires a conformational change, where the C-terminal fragment of the B-chain is transposed to bind the receptor (Menting et al. 2014). Preventing such a shift can deactivate the insulin molecule, as exemplified in single-chain insulin structures where the C-terminal of the B-chain is connected to the N-terminus of the A-chain (see Fig. S1 in Supporting Information for relevant examples)(Smith et al. 2022; Glidden et al. 2018; Derewenda et al. 1991). To mimic such a covalent constraint, we envisioned reversibly crosslinking two juxtaposed positions on the insulin molecule with a dipeptide moiety.

In an analogous report, we explored the incorporation of an additional disulfide bond as a constraint and identified several positions that substantially decreased insulin activity (Brunel et al. 2019). However, there are practical challenges in using cysteine to implement such constraints. These analogues required chemical synthesis to allow the orthogonal introduction of the fourth, non-native disulfide. Furthermore, there is no physiologic mechanism to facilitate the selective reduction of the non-native disulfide enabling its use as a molecular switch controlling insulin action. A more pragmatic approach is to employ a lysine sidechain and an N-terminal amine as anchor points to intramolecularly constrain insulin in an inactive conformation. Such crosslinking was previously reported by Brandenburg et al., where the N-terminus of the A-chain and the side chain of lysine at B29 were bridged by a dicarboxylic spacer (Brandenburg and Wollmer 1973; Brandenburg 1972; Brandenburg and Puls 1975; Freychet, Brandenburg, and Wollmer 1974; Jones et al. 1976; Nakagawa and Tager 1989). We employed a similar approach for several analogues, including desB30 (LysB29) and DesDi (LysB28) (Zaykov et al. 2014). Alternatively, we utilized a second lysine side chain, strategically placed to crosslink to a native insulin amine.

Since a prodrug susceptible to DKP formation requires a dipeptide sequence, a self-cleavable linker requires two connected dipeptides retaining the ability to cyclize at either end. The three positions considered for this purpose are side chains of the two amino acids (R1 and R2) and the N-alkylation between the two (R3) as depicted in the Fig. 2. To achieve a low insulin potency, the length between the two dipeptide modules is ideally short, such that it produces a rigid, closed insulin conformation. The shortest distance is found in structures such as B1, C2 and C3, but their preparation proved to be challenging due to increased steric demand. We were not successful after limited initial attempts.

We utilized cyclization with thiols in designs A, B2, and C1 as a more straightforward and efficient method for producing macrocyclic insulin analogues. The first synthetic step involved preparation of the bis-modified insulin intermediates followed by joining the two prodrug moieties by oxidation to a disulfide (Fig. 3B). Formation of the macrocycle in two steps avoids complications associated in using bifunctional reagents. Selective modification of insulin at a lysine sidechain amine typically involves acylation with an active ester in a high-pH aqueous buffer, which can lead to partial hydrolysis of the ester (Jonassen et al. ; Jonassen et al. 2012). Therefore, such bifunctional crosslinkers ideally require alternative acylation methods, typically under anhydrous conditions that necessitate protecting the N-terminal amines of insulin.

Table 1 A list of insulin analogues and prodrugs generated in this work. Activity of insulin was determined in insulin receptor Y1158 phosphorylation assay in HEK293 cells overexpressing B-isoform of insulin receptor (hIR-B), reported as EC50 values and percent potency relative to human insulin (HI). The second value allow side-by-side comparison of insulin analogues that account for variation between experiments. DesDi is desB28,B30 insulin; ^SS indicates a disulfide bond; FA stands for fatty acid. Exact structures of all compounds can be found in Supporting Information.

pep#	Insulin	Crosslinked Residue	Linker / Modification	EC₅₀(nM) at hIR-B	Potency, % rel. HI
1	DesDi			0.27	152%
2	DesDi	B28, A1	Boc-Cys(Acm)-Sar x2	3.0	21%
3	DesDi	B28, A1	[Cys^SS-Sar]₂	5.8	11%
4	DesDi, B9K	B28, B9		1.9	53%
5	DesDi, B9K	B28, B9	Boc-Cys(Acm)-Sar x2	3.3	30%
6	DesDi, B9K	B28, B9	[Cys^SS-Sar]₂	19	5.2%
7	DesDi	B28, A1	[Lys-(N-Et) Cys^SS]₂	18	4.8%
8	DesDi, A8K			0.51	171%
9	DesDi, A8K	B28, A8	[Lys-(N-Et) Cys^SS]₂	58	1.5%
10	DesDi, A14K			0.88	115%
11	DesDi, A14K	B28, A14	[Lys-(N-Et) Cys^SS]₂	1.9	53%
12	DesDi, B17K			0.65	74%
13	DesDi, B17K	B28, B17	[Lys-(N-Et) Cys^SS]₂	3.4	29%
14	DesDi	B28, A1	Lys-(N-C₂H₄SH)Gly x2	1.5	50%
15	DesDi	B28, A1	[Lys-(N-C₂H₄S^SS)Gly]₂	1.5	49%
16	DesB30	B29, A1	Gly₂-EDDA	1.3	71%
17	DesB30	B29, A1	Lys₂-EDDA	1.5	59%
18	DesDi	B28, A1	[Cys(Acm)]₂-EDDA	1.3	32%
19	DesDi	B28, A1	(Cys^SS)₂-EDDA	9.9	4.0%
20	DesDi	B28, A1	[Boc-Lys(Boc)]₂-EDDA	66	1.0%
21	DesDi	B28, A1	[Lys(FA)]₂-EDDA	>10000	0%

Design A. With Cys at the first position of the dipeptide we selected sarcosine (Sar) as the second residue. The dipeptide was assembled by coupling Boc-Cys(Acm)-OH to the Sar, preloaded on a chlorotrityl resin, followed by soft cleavage (30%HFIP in THF) and subsequent conversion to the NHS ester (Fig. 3A). As a test case, we used DesDi insulin analogue pep1 (desB28,B30-insulin), which is a high potency insulin analogue (Table 1). Using DMSO as the solvent enabled selective modification of the N-terminus of Gly-A1 and Lys-B28 residues (Fig. 3B). The resulting pep2 had significantly reduced potency relative to the starting insulin, likely due to acylation at A1. The macrocyclization was completed via iodine oxidation, followed by TFA deprotection to yield the product pep3. We observed only a modest reduction in potency upon macrocyclization to pep3, which while 10-fold less potent than native insulin was only 2-fold less potent than the analogous open-form, pep2 (Table 1). To test DKP conversion, we dissolved pep3 in PBS and incubated it at 37°C. The starting peptide first converted to the mixture of intermediates with one DKP at the respective ends of the linker, and then to the free insulin pep1. The release half-life was determined to be approximately 60 hours (Fig. 4).

This prodrug design was also evaluated with an insulin analogue incorporating a SerB9-Lys mutation. This mutation did not substantially lower the potency of insulin in the in vitro assay, with insulin pep4 having only two-fold lower potency than native insulin (Table 1). The crosslinking of the two lysines required selective reaction conditions to avoid modification of the N-termini. The NHS ester of the prodrug peptide was conjugated to insulin pep4 in a pH 11 Na₂CO₃buffer. As anticipated, it was difficult to achieve full conversion, and a mixture of free insulin and mono-substituted byproducts were recovered alongside the bis-modified product pep5 (Supporting Information). The disulfide cyclization and deprotection were performed as previously, yielding prodrug insulin pep6. The potency of bis-modified pep5 was only slightly lower than that of starting insulin pep4, and macrocyclic insulin pep6 was approximately ten-fold less potent than pep4.

Design B2. A cysteine at the second position of the dipeptide requires employing N-alkylated cysteine analogues to achieve DKP formation. While an Fmoc derivative of (NMe)Cys is commercially available, we elected to use Trt-protected cysteine with a free amine, which was N-alkylated by reductive amination with acetaldehyde (Fig. 5). This route also allows additional modifications to be placed at this position to control the speed of DKP formation or to provide a handle for a pharmacokinetic protractor. The N-ethylated cysteine product was purified before amidation with the terminal amino acid, lysine in this case. Finally, the prodrug dipeptide was converted to the NHS ester for subsequent use in insulin modification.

As previously, we assembled a DesDi insulin prodrug by crosslinking the lysine at B28 with the N-terminus of the A-chain. In comparison to the previous example (pep3), a 30-fold reduction in potency was observed in the crosslinked pep7, relative to DesDi insulin. This is likely attributable to a shorter spacer provided by the B2-type crosslinker and additional steric hindrance arising from the lysine residues. Inspired by these results, we expanded this design to three other insulin analogues: pep8 with A8Lys, pep10 with A14Lys, and pep12 with B17Lys site-specific mutations.

The pep8 analogue displayed potency comparable to DesDi, while the two other analogues exhibited lower potency similar to human insulin (Table 1). Crosslinking produced a different degree of change in potency for the three analogues, with the largest change observed in the case of the A8Lys mutant. Macrocyclic prodrug pep9 had a potency of 1.5% relative to native insulin, which represents a hundred-fold shift from its reference peptide, pep8. Prodrug insulins pep11 and pep13, on the other hand, showed only a marginal decrease upon macrocyclization. While we attempted DKP conversion for the other insulins, these resulted in precipitation, and the results could not be readily interpreted (see example with pep7 in the Supporting Information). We suspect that the stability of the dipeptide-based disulfide is compromised during incubation at neutral pH and elevated temperatures.

Design C1. Prodrug dipeptides require an N-alkylated amide between the two amino acids to promote DKP formation. Consequently, we employed a thiol group at the terminal end of the alkyl chain as a site for modification and macrocyclization. To prepare an N-alkylated glycine, we reacted bromoacetic acid with trityl-protected 3-aminopropanethiol (Fig. 6). The secondary amine was then coupled with the terminal amino acid of the prodrug moiety, once again lysine. This design was only tested with pep1, crosslinking the N-terminal A1 and Lys B28 amines. The bis-functionalized intermediate pep14 had reduced potency, but only 50% relative to human insulin, and macrocyclization (pep15) did not produce any further reduction in potency (Table 1). Incubation of pep15 in PBS resulted in relatively fast conversion to the single DKP intermediate, followed by a much slower formation of the second DKP and release of pep1 (Fig. 6).

Design C2. Synthesis of this prodrug moiety starts with ethylenediamine-N,N-diacetic acid (EDDA), which, due to its zwitterionic nature, was found to be insoluble in organic solvents commonly used for amino acid coupling. To overcome this limitation, EDDA was treated with trimethylsilyl chloride (TMS-Cl) to produce the more soluble TMS ester. TMS protection could be readily removed at a later point by exposure to an acidic aqueous buffer during HPLC purification. Using TMS-protected EDDA, we explored the coupling with HATU-activated amino acids, which worked well with a few amino acids such as Gly, Lys(Boc), or Cys(Acm). In an attempt to increase the size of the crosslinker to drive additional reduction in activity of crosslinked insulin, we examined several bulky amino acids, such as Ile, Trp(Boc), Chg (cyclohexylglycine), Tle (tert-leucine) and Dip (3,3-diphenyl-L-alanine). However, use of these amino acids resulted in increased steric hindrance during coupling, and in most cases, only the mono-substituted products were obtained.

We first evaluated these crosslinkers on DesDi pep1 and the readily available desB30 insulin. Pep16 and pep17 are the macrocyclic products of the reaction between insulin desB30 and EDDA-based crosslinkers containing Gly and Lys, respectively. No significant change in potency was observed in these cases, demonstrating that the applied constraints were not sufficient to inactivate insulin. Using the [Cys(Acm)]₂-EDDA crosslinker and pep1, we produced two structures: pep18 with Acm-protected thiols and pep19 with an additional disulfide formed within the prodrug structure itself. While pep18 showed only a 3-fold reduction in potency, the additional disulfide in the case of pep19 lowered potency to only 4% of native insulin (a 25-fold reduction).

Lastly, we explored the application of additional modifications to the prodrug structure to further increase steric bulk to help disrupt insulin-receptor interactions. We used Lys₂-EDDA to crosslink B28 and A1 amines on pep1. Intermediate pep20 with four Boc-protecting groups was tested and showed potency of 1% relative to human insulin, an encouraging result. The Boc groups were removed, and lysine residues were functionalized with fatty acid protractors at the side-chain amines. The resulting pep21 was almost completely inactive (Table 1). When incubated at neutral pH, DKP conversion could be observed by LCMS with nearly complete release of insulin pep1 after a week (Fig. 7).

Insulin SAR with EDDA-based prodrugs. We utilized Gly₂-EDDA crosslinker to further explore bridging alternative positions on insulin. To do so, we mutated select residues in DesDi insulin (pep1) that were expected to impose minimal impact on insulin potency: A8, A10, A14, B10, B17, and B22. Furthermore, these residues provide coverage across the three-dimensional insulin structure and offer guidance as to which crosslinking sites may prove superior (Fig. 8). Insulin analogues were prepared using Fmoc SPPS, folded as single-chain structures, and processed to two-chain analogues by treatment with Lys-C protease (see Methods and Supporting Information).

For crosslinking two lysine residues, we could have used high-pH reaction conditions as previously demonstrated, but it is not ideal due to concomitant hydrolysis of the active ester. To perform modification in anhydrous organic solvent, additional steps were required to protect the N-terminal amines of insulin analogues. For this purpose, we employed a simple carbamoylation step, where insulins were treated with sodium isocyanate overnight (Stark 1965; Oimomi et al. 2008). Mutations combined with N-terminal protection produced various effects on the potency of the analogues, with pep24 (A8Lys) and pep32 (B22Lys/B28Arg) demonstrating increases, while pep22 (B10Lys) and pep30 (B17Lys) resulted in reduction (Table 2).

Crosslinked insulins were prepared by reacting the bis-NHS ester of Gly₂-EDDA spacer with carbamoyl insulins in DMSO. Many constrained insulins had significantly reduced potency compared to the free peptides ranging from a 5-fold reduction with B28-B10 crosslinking to a 74-fold reduction for the B28-A14 analogue (Table 2).

To test the DKP conversion, peptides were incubated in PBS at 37°C. In most cases, the starting constrained peptide intermediates and the released active insulins were resolved by HPLC chromatography (example in Fig. 9B), enabling measurement of the reaction kinetics. Although the DKP kinetics varied somewhat between different constrained insulins, they exhibited similar values (Table 2). The half-life for the first DKP formation was ~70-90 hours, followed by the second DKP-cyclization to release free insulin release at an additional 50 hours. Longer reaction times were found to correlate with increased distance between the residues as well as a larger difference between the potencies of the open and closed conformations.

Table 2 A list of carbamoyl insulin analogues (X) with site-specific mutation to lysine and corresponding prodrugs with the Gly₂-EDDA spacer. Activity was assessed and reported as in Table 1. DKP conversion was measured by incubating 50µM insulin analogues in PBS, pH7.4 at 37°C. Two separable half-life values were obtained: one from disappearance of the starting material (SM) and the other from appearance of the unmodified insulin analogue (n.d.- could not be determined).

pep#	Insulin	Crosslinked positions	EC50 (nM)	Potency, % rel. HI	Fold-change	SM conv. t1/2, h	Free Ins t1/2, h
22	DesDi (A0,B0-X), B10K		19	6.2%
23	DesDi (A0,B0-X), B10K	B28, B10	91	1.3%	5 x	73 h	47 h
24	DesDi (A0,B0-X), A8K		0.6	203%
25	DesDi (A0,B0-X), A8K	B28, A8	23	5.2%	39 x	n.d.	54 h
26	DesDi (A0,B0-X), A10K		2.1	90%
27	DesDi (A0,B0-X), A10K	B28, A10	96	1.9%	47 x	94 h	50 h
28	DesDi (A0,B0-X), A14K		1.9	96%
29	DesDi (A0,B0-X), A14K	B28, A14	143	1.3%	74 x	92 h	45 h
30	DesDi (A0,B0-X), B17K		2.8	37%
31	DesDi (A0,B0-X), B17K	B28, B17	83	1.3%	29 x	79 h	47 h
32	DesDi, B0Ac, B22K, B28R		0.7	243%
33	DesDi, B0Ac, B22K, B28R	B22, A1	30	5.7%	43 x	71 h	51 h

DKP Chemistry. A conventional prodrug employing DKP-formation entails a dipeptide extension that can excise from the parent drug under first-order kinetics. In this instance, it offers a preferred approach for a traceless conversion of an inactive insulin to an active one. By combining two DKP-susceptible motifs in a single structure, we created a series of linkers that lock insulin into a constrained (closed) conformation, compromising its interaction with its receptor. We initially focused on cysteine or other thiol containing dipeptides, as these offer a more modular approach to assembling insulin prodrugs. The examples we studied did not produce a sufficient change (>100-fold) in inherent potency, but it is conceivable that additional optimization in linker size and crosslinking positions may lead to enhanced performance.

The preferred bifunctional prodrug linker we studied is based on ethylenediamine-N,N-diacetic acid (EDDA). While two other designs were envisioned, we had limited success in their synthesis. The EDDA-based spacers produced some of the more impressive potency results when comparing the drug (open) versus prodrug (closed) forms. The measurement of DKP formation kinetics in the glycine-EDDA prodrugs indicates a half-life of ~3-4 days to form the first cyclic dipeptide. The intermediate produced after the first DKP formation is likely to be of greater potency than the bridged precursor and possibly as active as the final drug. This needs to be confirmed through kinetic measurements and assessment of glucose lowering capacity. Unless the native insulin chemical structure is necessary, a single dipeptide cyclization is sufficient to achieve the objective of prodrug conversion to an active insulin and may be a preferred design utilizing conformational constraints. PK modelling could also help determine an ideal release rate for such a drug, but several days is likely a reasonable half-life for a once-weekly insulin. Additional optimizations of the DKP construct can be achieved by substitution of the individual residues and the N-alkyl group, allowing for fine-tuning of the insulin release.

Insulin Chemistry. Using lysine and N-terminal amines for crosslinking is an obvious choice in a disulfide-rich molecule such as insulin. The single lysine at the C-terminal end of the B-chain has been extensively utilized for modification of insulin analogues, and the conditions for selective lysine modification are well-developed. There is also a good understanding of the reactivity of the two N-termini, making their selective chemical modification possible.

Under high-pH aqueous conditions, selective acylation of the lysine side chain can be performed due to the relative nucleophilicity of this amine. We have developed several insulin analogues incorporating two lysines, one at position B28 and another positioned in one of the two helices found in the A- or B-chain. Such a selection provides increased proteolytic resistance given the stability derived from the secondary structure. This allows for the conversion of a single-chain insulin precursor to an active two-chain molecule without proteolytic cleavage at the newly introduced lysine (Wu et al. 2015).

Crosslinking of the two lysine residues is more straightforward when the macrocyclization follows the acylation step, as high-pH conditions cause undesirable NHS ester hydrolysis. Using a bifunctional active ester in such a manner results in the formation of significant amounts of byproducts caused by partial hydrolysis. Anhydrous conditions are preferred to achieve improved conversion but require additional protecting groups to be installed at the N-termini.

Connecting the lysine side chain with the N-terminus of the A-chain offers a compromise, as both can be selectively functionalized using DMSO as a solvent. From the perspective of inactivating insulin, however, the main disadvantage is the proximity of the B28 or B29 lysine to A1. Crosslinking of these positions does not produce sufficient conformational strain to effectively reduce the activity for prodrug application, as seen in multiple examples. To overcome this, one option is to utilize a bulkier spacer to increase steric interference with receptor binding, as shown in pep20 and pep21. Another option is to replace B28 lysine with an arginine and introduce another lysine elsewhere in the hormone. Such an analogue, pep32, with lysine at position B22, has shown a significant loss in potency (43-fold) when crosslinked by a suitable prodrug spacer (pep33). Presumably, combining increased spacer size with crosslinking of the N-terminal A1 to B22 will offer the best opportunity in design of an insulin activity switch (Hoeg-Jensen 2021).

We have demonstrated a strategy for reversibly inactivating insulin by conformational constraint using a prodrug crosslinker. The crosslinkers were designed by connecting two dipeptides susceptible to DKP formation, each capable of independent release from insulin. Constrained insulins were generated using four types of linkers which produced different degrees of change in potency. Longer linkers and crosslinking of spatially close residues (i.e. A1-B28) typically did not produce sufficient inactivation. This limitation can be resolved by introducing additional steric hindrance in the linker.

An alternative strategy was explored by crosslinking distal residues on the insulin surface. Using a Gly₂-EDDA crosslinker we demonstrated that a 50-fold and 70-fold change in potency can be achieved by bridging lysine B28 with A10, or A14 respectively. The B22Lys, B28Arg-DesDi insulin analogue appears as an ideal substrate for creating insulin switches without significantly changing the parent insulin sequence. With a sterically small Gly₂-EDDA crosslinker, a 43-fold potency difference between open and closed conformation was observed, which we believe can be further enhanced by increasing the overall bulk of the spacer. The observed 70-hour conversion half-life to the open form renders this design suitable for a once-weekly insulin therapy.

This general concept can be directly expanded for multiple uses. Ideally, the DKP release assembly would be attached to a carrier system (such as Fc, lipidation, albumin, etc.) to ensure the drug remains in circulation until the time when the insulin is released. Alternatively, the system can be coupled with a polymer or nanoparticle support for extended-release applications. The design principles seem applicable to the identification of glucose-sensitive insulins, where the differential interaction with glucose at a suitable plasma concentration remains the greatest challenge. Finally, this prodrug strategy can certainly be applied to other peptides and proteins where inactivation can be achieved via crosslinking.

General methods. All reagents and solvents were purchased from Sigma-Aldrich or VWR. Protected amino acids and H-Rink amide Chemmatrix resin were obtained from Gyros Protein Technologies. Analytical HPLC was performed on an Agilent’s 1260 Infinity/ 6120 Quadrupole instrument with Kinetex C8 2.6m 100A (75V60 mm) column using a flow rate of 1 mL/min^-1 and a gradient 10%-80% B over 10 min. Eluent A is water with 0.05% TFA and eluent B is 10% water, 90%MeCN, 0.05% TFA. MS collected in 200–2000 m/z range. HPLC purifications were performed on a Waters instrument (Waters Controller model 600, Waters dual wavelength detector 2487, ProStar model 701 fraction collector and Kipp & Zonen BD41 chart recorder) with Luna 10m C8(2) 100A AXIA (250V21.2 mm) column and flow rate of 10–15 mL/min.

Synthesis of insulin analogues. Insulin peptides were assembled on Symphony X instrument (Gyros Protein Technologies) using 30-minute coupling cycles with 0.3M AA activated by DIC/ Oxyma and 20% piperidine in DMF for deprotection. Peptides were cleaved with TFA cocktail (15 mL per 0.1 mmol of resin) containing 2.5% v/v of each: β-mercaptoethanol, triisopropylsilane, anisole and water. Insulins were folded by first dissolving crude peptide in 200mL of 100mM Na₂CO₃ buffer, pH11 with 100 mg Cys-HCl. 500 mg of cystine was dissolved in 1N NaOH and added to the stirring peptide solution. After 20 minutes, solution was acidified using TFA and folded peptide purified by preparative HPLC. Lys-C cleavage was done by adding 0.001% w/w of enzyme (produced internally at NovoNordisk) to purified insulin peptide dissolved in 20%MeCN - 50mM NH₄HCO₃ buffer, pH 8. Final insulin purified by prep HPLC.

Synthesis of crosslinker type A. 1 mmol of chlorotrityl resin (1.3 mmol/g) was loaded with Fmoc-Sar-OH (2 mmol) overnight in 2 mL DCM, 300uL DIEA. Solution was drained and resin washed with DMF and DCM. Capping was done by incubating the resin with 9 mL DCM, 1mL MeOH, 300uL DIEA for 1 hour. For coupling, 2 mmol of Boc-Cys(Acm) and 2 mmol DEPBT were added in 3mL DMF with 320 uL DIEA. Reaction was allowed to proceed overnight, before resin was washed with DCM and MeOH. Dipeptide can be soft-cleaved by either treatment with 30%HFIP in THF for 60 minutes or 1% TFA in THF for 15 min. Cleaved peptide was diluted into 100 mL of water containing 20%MeCN, 0.1%TFA and lyophilized. To obtain NHS ester, crude peptide was dissolved in 3 mL DMF and 1 mmol of NHS-OH (115mg) was added, followed by 1 mmol of DIC (155uL). After 1 hour, reaction mixture was diluted with 20 mL water with 20%MeCN, 0.1%TFA and purified by HPLC. 37.6mg of NHS ester was isolated. After reaction with insulin (see procedure below), 4.9 mg of bis-modified peptide (example pep2) was dissolved in 100 uL of AcOH with 5 uL of 1M HCl. 50 uL of 100mM iodine in MeOH was then added. Reaction was analysed by LCMS after 5 minutes. Peptide was precipitated in ether after 20 minutes and then purified by HPLC. 1.5 mg of product was isolated. To remove Boc groups, peptide was dissolved in 50uL TFA and precipitated in ether after 5 minutes.

Synthesis of crosslinker type B2. 0.95 mmol H-Cys(Trt)-OH was suspended in 2 mL of MeOH and 2.4 mmol acetaldehyde was added. Reaction started turning yellowish. After 5 minutes 100 mg of NaCNBH₃ in 1 mL of MeOH was added to the reaction, which was allowed to go overnight. MeOH was evaporated with a jet of nitrogen and crude was dried under vacuum for 30 minutes. For coupling step, 1.1 mmol of Boc-Lys(Boc)-OH was preactivated for 30 minutes with 1 mmol HATU in 2 mL DMF, 300 uL DIEA and then added to the crude from the previous step. Reaction was purified by HPLC after 6 hours. Purified dipeptide was dissolved in 4 mL DMF with 2 mmol NHS-OH and 2 mmol DIC was then added. After 3 hours reaction was purified by HPLC. 72mg of product was isolated.

Synthesis of crosslinker type C1. 1 mmol bromoacetic acid was dissolved in 1 mL of DMF and 1 mmol STrt-mercaptoethylamine was added, followed by 350uL DIEA. Reaction was stirred for 3 hours at room temperature. In the second step, 2.2 mmol Boc-Lys(Boc)-OH was preactivated with 2.2 mmol HATU in 2mL DMF with 350 µL DIEA for 30 minutes before being added to the crude reaction mixture from the first step. After 3 hours, reaction was diluted into 50 mL of water with 20%MeCN, 0.1%TFA and purified by HPLC. After lyophilization clean protected dipeptide was dissolved in 5 mL THF. 2 mmol NHS-OH was added followed by 2 mmol of DIC. After 2 hours reaction was diluted into 50 mL of water with 20%MeCN, 0.1%TFA and purified by HPLC. 27.4 mg of NHS ester was obtained after lyophilization.

Synthesis of crosslinker type C2. 2.2 mmol of chlorotrimethylsiane was slowly added to 1 mmol of ethylenediamine-N,N-diacetic acid and mixture was stirred overnight. Mixture was dried under vacuum and used as crude. About quarter of material (appears wet) was dissolved in 2 mL of DMF. 2 mmol of Boc-Gly-OH was preactivated for 20 minutes with 2 mmol HATU/DIEA in 2 mL DMF and added to the EDDA(TMS) solution, followed by an additional 250 uL of DIEA. After 4 hours reaction was diluted with 100 mL of water with 20%MeCN, 0.1%TFA and purified by HPLC. After lyophilization peptide was dissolved in 2mL DMF. 1 mmol NHS-OH was added followed by 1 mmol of DIC. After 2 hours reaction was diluted into 50 mL of water with 20%MeCN, 0.1%TFA and purified by HPLC.

Example of procedure for derivatization of N-terminal amine at A1 and lysine side chain. 20mg DesDi two chain insulin dissolved in 1 mL DMSO, 100 uL DIEA and added to 3.3 mg NHS ester. Reaction frozen after 5 minutes while performing LCMS analysis. Then it is diluted in water with 20%MeCN, 0.1%TFA and purified by HPLC. 9.9mg of bis-modified product was isolated, together with 3.6mg of mono-modified byproduct.

Example of procedure for selective derivatization of lysine side chains. 5.1 mg of insulin was dissolved in 1 mL of 100mM Na2CO3 buffer, pH11. 5.2 mg of NHS ester was dissolved in 500 µL of MeCN and added to the peptide solution. After 10 minutes reaction was purified by HPLC. 1.8 mg of bis-modified product was isolated.

Insulin Receptor In Vitro Phosphorylation. Insulin receptor phosphorylation was measured in an indirect ELISA assay. Briefly HEK293 cells (ATCC CRL-1573) overexpressing B-isoform of human insulin receptor (hIR-B) were plated in 96- well tissue culture plates and cultured in DMEM supplemented with 100 IU/mL penicillin, 100 μg/mL streptomycin, 10 mM HEPES, and 0.25% bovine growth serum (HyClone SH30541) for 16–20 h at 37°C, 5% CO2, and 90% humidity. Serial dilutions of biosynthetic human insulin and test peptides prepared in DMEM supplemented with 1% ovalbumin (Millipore-Sigma, A5503) were added to the plate wells. After 15 min incubation at 37°C in humidified atmosphere with 5% CO2, the cells were fixed with 5% paraformaldehyde for 20 min at room temperature, washed twice with phosphate buffered saline (PBS, pH 7.4) and blocked with 2% bovine serum albumin (BSA, Millipore-Sigma 810533) in PBS for 1 h. The plate was washed three times, filled with anti-phospho-IR/IGF1R (Tyr1158) antibody (Life Technologies, 710101) and incubated for 3 h at room temperature, after which the plate was washed four times and filled with goat anti-rabbit secondary antibody (#A16110, ThermoFisher). Following 30 min incubation and another four washes, 0.1 mL of TMB One Solution substrate (Life Technologies, #00-2023) was added to each well. Colour development was stopped 10 min later by adding 0.05 mL 1M hydrochloric acid. Absorbance at 450 nm was measured on Envision multimode reader (Perkin-Elmer). Absorbance vs. Peptide concentration dose–response curves were plotted, and EC50 values were determined using logistic nonlinear three-parameter regression in GraphPad Prism 8 (GraphPad Software). Each experiment was performed at least three times and the error bars represent standard deviation.

Competing interests

There are no conflicts to declare. A.N.Z., V.M.G., and R.D.D. were employees of Novo Nordisk at the time this specific work was conducted, and the work was funded by Novo Nordisk.

Author Contributions

A.N.Z. conceptualization, investigation, writing – original draft, writing – review & editing; V.M.G. investigation; R.D.D. writing – review & editing.

Acknowledgements

We thank Jay Levy for assistance with peptide SPPS, and John P. Mayer for constructive critic of the manuscript.

Brandenburg, Dietrich. 1972. 'Preparation of NαA1, NεB29-Adipoylinsulin, an Intramolecularly Crosslinked Derivative of Beef Insulin', Hoppe-Seyler´ s Zeitschrift für physiologische Chemie, 353: 869-73.
Brandenburg, Dietrich, and Walter Puls. 1975. "Insulin derivatives crosslinked by a dicarboxylic acid moiety." In.: Google Patents.
Brandenburg, Dietrich, and Axel Wollmer. 1973. 'The effect of a non-peptide interchain crosslink on the reoxidation of reduced insulin', Hoppe-Seyler´ s Zeitschrift für physiologische Chemie, 354: 613-27.
Brunel, Florence M., John P. Mayer, Vasily M. Gelfanov, Alexander N. Zaykov, Brian Finan, Diego Perez-Tilve, and Richard D. DiMarchi. 2019. 'A Disulfide Scan of Insulin by [3 + 1] Methodology Exhibits Site-Specific Influence on Bioactivity', ACS Chemical Biology, 14: 1829-35.
Chamberlain, M. J., and L. Stimmler. 1967. 'The Renal Handling of Insulin', The Journal of Clinical Investigation, 46: 911-19.
De, Arnab, and Richard D DiMarchi. 2008. 'Investigation of the feasibily of an amide-based prodrug under physiological conditions', International Journal of Peptide Research and Therapeutics, 14: 255-62.
Derewenda, U., Z. Derewenda, E. J. Dodson, G. G. Dodson, Xiao Bing, and Jan Markussen. 1991. 'X-ray analysis of the single chain B29-A1 peptide-linked insulin molecule: A completely inactive analogue', Journal of Molecular Biology, 220: 425-33.
DiMarchi, Richard D, and Binbin Kou. 2015. "Amide based glucagon and superfamily peptide prodrugs." In.: Google Patents.
DiMarchi, Richard D, Binbin Kou, and Shujiang Cheng. 2014. "Amide based insulin prodrugs." In.: Google Patents.
Duckworth, William C. 1988. 'Insulin Degradation: Mechanisms, Products, and Significance*', Endocrine Reviews, 9: 319-45.
Fischer, P. M. 2003. 'Diketopiperazines in peptide and combinatorial chemistry', J Pept Sci, 9: 9-35.
Freychet, P., D. Brandenburg, and A. Wollmer. 1974. 'Receptor-binding assay of chemically modified insulins', Diabetologia, 10: 1-5.
Glidden, Michael D., Yanwu Yang, Nicholas A. Smith, Nelson B. Phillips, Kelley Carr, Nalinda P. Wickramasinghe, Faramarz Ismail-Beigi, Michael C. Lawrence, Brian J. Smith, and Michael A. Weiss. 2018. 'Solution structure of an ultra-stable single-chain insulin analog connects protein dynamics to a novel mechanism of receptor binding', Journal of Biological Chemistry, 293: 69-88.
Gomes, Paula, Nuno Vale, and Rui Moreira. 2007. 'Cyclization-activated prodrugs', Molecules (Basel, Switzerland), 12: 2484-506.
Hansen, Ryan J, Gordon B Cutler Jr, Andrew Vick, Anja Koester, Shun Li, Angela M Siesky, and John M Beals. 2012. "LY2605541: leveraging hydrodynamic size to develop a novel basal insulin." In Diabetes, A228-A28. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.
Hirsch, Irl B, Rattan Juneja, John M Beals, Caryl J Antalis, and Eugene E Wright, Jr. 2020. 'The Evolution of Insulin and How it Informs Therapy and Treatment Choices', Endocrine Reviews, 41: 733-55.
Hoeg-Jensen, Thomas. 2021. 'Review: Glucose-sensitive insulin', Molecular Metabolism, 46: 101107.
Jonassen, I, T Hoeg-Jensen, S Havelund, U Ribel-Madsen, TM Tagmose, and P Madsen. 'New insulin derivative comprising a naturally occurring insulin with a side chain attached to the alpha-amino or epsilon-amino in the B chain of the parent insulin, useful in composition for treating diabetes', Patent registered by Novo Nordisk A/S and published in the Derwent World Patents Index (DWPI). Patent no. WO2005012347-A2.
Jonassen, Ib, Svend Havelund, Thomas Hoeg-Jensen, Dorte Bjerre Steensgaard, Per-Olof Wahlund, and Ulla Ribel. 2012. 'Design of the Novel Protraction Mechanism of Insulin Degludec, an Ultra-long-Acting Basal Insulin', Pharmaceutical Research, 29: 2104-14.
Jones, R. H., D. I. Dron, M. J. Ellis, P. H. Sönksen, and D. Brandenburg. 1976. 'Biological properties of chemically modified insulins. I. Biological activity of proinsulin and insulin modified at A1-glycine and B29-lysine', Diabetologia, 12: 601-08.
Kjeldsen, Thomas B., František Hubálek, Claudia U. Hjørringgaard, Tina M. Tagmose, Erica Nishimura, Carsten E. Stidsen, Trine Porsgaard, Christian Fledelius, Hanne H. F. Refsgaard, Sanne Gram-Nielsen, Helle Naver, Lone Pridal, Thomas Hoeg-Jensen, Claus Bekker Jeppesen, Valentina Manfè, Svend Ludvigsen, Inger Lautrup-Larsen, and Peter Madsen. 2021. 'Molecular Engineering of Insulin Icodec, the First Acylated Insulin Analog for Once-Weekly Administration in Humans', Journal of Medicinal Chemistry, 64: 8942-50.
Kurtzhals, Peter, Søren Østergaard, Erica Nishimura, and Thomas Kjeldsen. 2023. 'Derivatization with fatty acids in peptide and protein drug discovery', Nature Reviews Drug Discovery, 22: 59-80.
Lin, Shan-Yang, and Shun-Li Wang. 2012. 'Advances in simultaneous DSC–FTIR microspectroscopy for rapid solid-state chemical stability studies: Some dipeptide drugs as examples', Advanced Drug Delivery Reviews, 64: 461-78.
Liu, Liming. 2018. 'Pharmacokinetics of monoclonal antibodies and Fc-fusion proteins', Protein & Cell, 9: 15-32.
Martins, João Pedro, Patrick J. Kennedy, Hélder A. Santos, Cristina Barrias, and Bruno Sarmento. 2016. 'A comprehensive review of the neonatal Fc receptor and its application in drug delivery', Pharmacology & Therapeutics, 161: 22-39.
Menting, John G., Yanwu Yang, Shu Jin Chan, Nelson B. Phillips, Brian J. Smith, Jonathan Whittaker, Nalinda P. Wickramasinghe, Linda J. Whittaker, Vijay Pandyarajan, Zhu-li Wan, Satya P. Yadav, Julie M. Carroll, Natalie Strokes, Charles T. Roberts, Faramarz Ismail-Beigi, Wieslawa Milewski, Donald F. Steiner, Virander S. Chauhan, Colin W. Ward, Michael A. Weiss, and Michael C. Lawrence. 2014. 'Protective hinge in insulin opens to enable its receptor engagement', Proceedings of the National Academy of Sciences, 111: E3395-E404.
Moyers, Julie S., Ryan J. Hansen, Jonathan W. Day, Craig D. Dickinson, Chen Zhang, Xiaoping Ruan, Liyun Ding, Robin M. Brown, Hana E. Baker, and John M. Beals. 2022. 'Preclinical Characterization of LY3209590, a Novel Weekly Basal Insulin Fc-Fusion Protein', Journal of Pharmacology and Experimental Therapeutics, 382: 346-55.
Nakagawa, S. H., and H. S. Tager. 1989. 'Perturbation of insulin-receptor interactions by intramolecular hormone cross-linking', Journal of Biological Chemistry, 264: 272-79.
Oimomi, Munetada, Hiroshi Hatanaka, Yukio Yoshimura, Koichi Yokono, Shigeaki Baba, and Yoshihisa Taketomi. 2008. 'Carbamylation of Insulin and Its Biological Activity', Nephron, 46: 63-66.
Rosenstock, Julio, and Stefano Del Prato. 2022. 'Basal weekly insulins: the way of the future!', Metabolism, 126: 154924.
Rosenstock, Julio, Rattan Juneja, John M. Beals, Julie S. Moyers, Liza Ilag, and Rory J. McCrimmon. 2024. 'The Basis for Weekly Insulin Therapy: Evolving Evidence With Insulin Icodec and Insulin Efsitora Alfa', Endocrine Reviews: bnad037.
Santos, Cledir, Maria Luísa Mateus, Ana Paula dos Santos, Rui Moreira, Eliandre de Oliveira, and Paula Gomes. 2005. 'Cyclization-activated prodrugs. Synthesis, reactivity and toxicity of dipeptide esters of paracetamol', Bioorganic & Medicinal Chemistry Letters, 15: 1595-98.
Sims, Emily K., Alice L. J. Carr, Richard A. Oram, Linda A. DiMeglio, and Carmella Evans-Molina. 2021. '100 years of insulin: celebrating the past, present and future of diabetes therapy', Nature Medicine, 27: 1154-64.
Sinha, V. P., S. L. Choi, D. K. Soon, K. F. Mace, K. P. Yeo, S. T. Lim, and D. C. Howey. 2014. 'Single-dose pharmacokinetics and glucodynamics of the novel, long-acting basal insulin LY2605541 in healthy subjects', J Clin Pharmacol, 54: 792-9.
Smith, Nicholas A., John G. Menting, Michael A. Weiss, Michael C. Lawrence, and Brian J. Smith. 2022. 'Single-chain insulin analogs threaded by the insulin receptor αCT domain', Biophysical Journal, 121: 4063-77.
Sonne, O. 1988. 'Receptor-mediated endocytosis and degradation of insulin', Physiological Reviews, 68: 1129-96.
Stark, George R. 1965. 'Reactions of Cyanate with Functional Groups of Proteins. III. Reactions with Amino and Carboxyl Groups*', Biochemistry, 4: 1030-36.
Thalluri, Kishore, John P. Mayer, Joseph R. Chabenne, Vasily Gelfanov, and Richard D. DiMarchi. 2018. 'Synthesis of disulfide-rich heterodimeric peptides through an auxiliary N, N-crosslink', Communications Chemistry, 1: 36.
Tokarz, Victoria L., Patrick E. MacDonald, and Amira Klip. 2018. 'The cell biology of systemic insulin function', Journal of Cell Biology, 217: 2273-89.
Valera Mora, M. Elena, Antonino Scarfone, Menotti Calvani, Aldo V. Greco, and Geltrude Mingrone. 2003. 'Insulin Clearance in Obesity', Journal of the American College of Nutrition, 22: 487-93.
Wang, Jingyao, Mark R. Berglund, Timothy Braden, Matthew C. Embry, Martin D. Johnson, Stephen R. Groskreutz, Fareed Bhasha Sayyed, Sergey Vladimirovich Tsukanov, Timothy D. White, Ankur Jalan, Kevin D. Seibert, and Michael E. Kopach. 2022. 'Mechanistic Study of Diketopiperazine Formation during Solid-Phase Peptide Synthesis of Tirzepatide', ACS Omega, 7: 46809-24.
Wu, Fangzhou, Joseph R. Chabenne, Vasily M. Gelfanov, John P. Mayer, and Richard D. DiMarchi. 2015. 'Three-chain insulin analogs demonstrate the importance of insulin secondary structure to bioactivity', Journal of Peptide Science, 21: 223-30.
Zaykov, Alexander N., John P. Mayer, and Richard D. DiMarchi. 2016. 'Pursuit of a perfect insulin', Nature Reviews Drug Discovery, 15: 425-39.
Zaykov, Alexander N., John P. Mayer, Vasily M. Gelfanov, and Richard D. DiMarchi. 2014. 'Chemical Synthesis of Insulin Analogs through a Novel Precursor', ACS Chemical Biology, 9: 683-91.

No competing interests reported.

Download PDF

Editorial decision: Revision requested
16 Sep, 2024
Reviews received at journal
23 Aug, 2024
Reviews received at journal
23 Aug, 2024
Reviews received at journal
22 Aug, 2024
Reviews received at journal
15 Aug, 2024
Reviewers agreed at journal
15 Aug, 2024
Reviewers agreed at journal
14 Aug, 2024
Reviewers agreed at journal
13 Aug, 2024
Reviewers agreed at journal
13 Aug, 2024
Reviewers invited by journal
13 Aug, 2024
Editor assigned by journal
12 Aug, 2024
Submission checks completed at journal
12 Aug, 2024
First submitted to journal
10 Aug, 2024

You are reading this latest preprint version

Insulin prodrugs with programmed chemical release of an intramolecular cross-link

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

Conclusion

MATERIALS AND METHODS

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1