A skin-permeable polymer for non-invasive transdermal insulin delivery

Abstract

Non-invasive skin permeation is widely used for convenient transdermal delivery of small-molecule therapeutics (less than 500 Da) with appropriate hydrophobicities¹. However, it has long been deemed infeasible for large molecules-particularly polymers, proteins and peptides^2,3-due to the formidable barrier posed by the skin structure. Here we show that the fast skin-permeable polyzwitterion poly[2-(N-oxide-N,N-dimethylamino)ethyl methacrylate] (OP) can efficiently penetrate the stratum corneum, viable epidermis and dermis into circulation. OP is protonated to be cationic and is therefore enriched in the acidic sebum and paracellular stratum corneum lipids containing fatty acids, and subsequently diffuses through the intercorneocyte lipid lamella. Beneath the stratum corneum, at the normal physiological pH, OP becomes a neutral polyzwitterion, 'hopping' on cell membranes, enabling its efficient migration through the epidermis and dermis and ultimately entering dermal lymphatic vessels and systemic circulation. As a result, OP-conjugated insulin efficiently permeates through the skin into the blood circulation; transdermal administration of OP-conjugated insulin at a dose of 116 U kg^-1 into mice with type 1 diabetes quickly lowers their blood glucose levels to the normal range, and a transdermal dose of 29 U kg^-1 normalizes the blood glucose levels of diabetic minipigs. Thus, the skin-permeable polymer may enable non-invasive transdermal delivery of insulin, relieving patients with diabetes from subcutaneous injections and potentially facilitating patient-friendly use of other protein- and peptide-based therapeutics through transdermal delivery.

Fig. 1. Skin permeability of OP.

a, CLSM images of the section slices of the C57BL/6J mouse dorsal skin after 4 h of post-topical application with OP^Cy5 or PEG^Cy5 (Cy5-equivalent dose, 0.2 ml of 10 μg ml⁻¹) through a diffusion cell (1.13 cm²); the nuclei were stained with DAPI (blue). The images are representative of n = 3 independent experiments. b, Cy5 fluorescence intensities plotted from the skin surface to the subcutis along randomly selected lines (yellow arrows in a). c,d, Transmission electron microscopy characterization of cryosections of the SC layer of the dorsal skin after 4 h of topical application with OP–AuNPs (OP-equivalent dose, 0.2 ml of 0.5 mg ml⁻¹; application area, 1.13 cm²). HAADFI–STEM and EDS elemental mapping (c) and EDS analysis (d) are shown. A large view of the SC region is shown in Supplementary Fig. 4b. The images are representative of n = 3 independent experiments. e–g, Cy5 fluorescence intensity in the blood as a function of the time of topical application with OP^Cy5 on the mouse dorsal skin (e), and ex vivo fluorescence imaging (f) and fluorescence intensity quantification (g) of the major organs of the mice after 2 h of topical application of OP^Cy5 (Cy5-equivalent dose, 0.2 ml of 10 μg ml⁻¹; application area, 1.13 cm²). Data are mean ± s.d. n = 3 mice. Scale bars, 50 μm (a) and 50 nm (c). Source Data

Fig. 2. Skin permeability of OP–I.

a, CLSM images of the distribution of OP–I^FITC, PEG–I^FITC or insulin^FITC across the in vitro 3D skin equivalent EpiKutis model at 4 h after treatment (FITC-equivalent dose, 0.2 ml of 10 μg ml⁻¹; application area, 0.081 cm²; width, 1,272 μm; height, 1,272 μm; depth, 170 μm). b, Time-dependent permeation curves of OP–I, PEG–I and insulin across the EpiKutis model (insulin-equivalent dose, 0.2 ml of 0.5 mg ml⁻¹; application area, 0.081 cm²). Data are mean ± s.d. n = 3 independent experiments. c, CLSM images of the mouse dorsal skin after 4 h of topical application of OP–I^FITC (FITC-equivalent dose, 0.2 ml of 10 μg ml⁻¹; application area, 1.13 cm²). E, epidermis; H, hair follicle; S, subcutis; D, dermis. The images are representative of n = 3 independent experiments. d, CLSM images of immunofluorescence staining of LYVE-1^AF488 (lymphatic vessel endothelial hyaluronan receptor-1, green) in the subcutaneous tissue of Sprague–Dawley rats after 4 h topical application of OP–I^Cy5 (red) (Cy5-equivalent dose, 0.2 ml of 10 μg ml⁻¹; application area, 1.13 cm²). The nuclei were labelled with DAPI (blue). The images are representative of n = 3 independent experiments. e, CLSM images of the slices of the minipig abdominal skin after 4 h of topical application of OP–I^Cy5-containing cream (Cy5-equivalent dose, 10 ml of 10 μg ml⁻¹; application area, 100 cm²). The images are representative of n = 3 independent experiments. f, Cy5-fluorescence intensity profile from the skin surface to the subcutis plotted along a randomly selected line (yellow arrow in e). Scale bars, 100 μm (c,d) and 50 μm (e). Source Data

Fig. 3. Hypoglycaemic effect of topical OP–I in STZ-induced diabetic mice and minipigs.

a, Plasma insulin concentrations in diabetic mice after topical application of PBS, insulin, PEG–I or OP–I (insulin-equivalent dose, 116 U kg⁻¹; 0.2 ml of 0.5 mg ml⁻¹, 1.13 cm² dorsal skin). n = 5 mice. Mice injected subcutaneously with insulin (5 U kg⁻¹) were used as a control. b, The plasma insulin concentrations in diabetic mice after three consecutive days of topical application of OP–I as in a. n = 5 mice. c, BGLs of diabetic mice after treatments as in a. n = 8 mice. d, BGLs of the diabetic mice after topical application with lower doses of OP–I (insulin-equivalent dose, 58 or 29 U kg⁻¹; 0.1 ml or 0.05 ml of 0.5 mg ml⁻¹, 1.13 cm² dorsal skin). n = 8 mice. e,f, The plasma insulin concentrations (e; n = 5 mice) and BGLs (f; n = 8 mice) of diabetic mice after topical application of OP–I on the dorsal or abdominal skin as in a. The two experiments were performed using separate groups. g, IPGTTs in diabetic mice. n = 5 mice. Mice received treatments as in a and, 1 h later, were injected intraperitoneally with glucose (1.5 g per kg); their BGLs were then measured. Healthy mice were used as controls. h, The AUC (0–120 min) in the IPGTT experiment in g. n = 5 mice. i, BGLs in the diabetic minipigs after topical application on the abdominal skin with insulin, PEG–I or OP–I dispersed in water-in-oil cream (insulin-equivalent dose, 29 U kg⁻¹, 1 mg ml⁻¹, 40 ml, 400 cm²; low-dose group (L), 7.25 U kg⁻¹, 1 mg ml⁻¹, 10 ml, 100 cm²); n = 3 minipigs. Data are mean ± s.d. For c, d, f, g and i, the shaded areas outline the normal blood glucose range (50–200 mg dl⁻¹). The diagram in i was created using BioRender. Source Data

Fig. 4. Mechanism study of the SC penetration of OP and OP–I.

a, CLSM images of SC samples from the mouse dorsal skin after 4 h of topical application with OP–I^Cy5 (red; Cy5-equivalent dose, 0.2 ml of 10 μg ml⁻¹, 1.13 cm²). The SC samples were obtained by peeling the skin using adhesive tape, and the SC intercellular lipids were stained with NBD-C6-HPC (green). Additional images are provided in Extended Data Fig. 6a. The images are representative of n = 3 independent experiments. b, 3D reconstructed view from sequential z-stack intravital two-photon microscopy imaging of the mouse skin after 4 h of topical application with OP–I^FITC (FITC-equivalent dose, 0.2 ml of 10 μg ml⁻¹, 1.13 cm²). Separate images are provided in Supplementary Fig. 24. The images are representative of n = 3 independent experiments. c, Representative binding modes from MD simulations of insulin, OP and OP–I on SC lipids at pH 5.5. The SC lipid membrane, composed of equimolar ceramide, cholesterol and free fatty acids, was generated using the membrane builder of CHARMM-GUI. See also Supplementary Video 2. d, PMF results perpendicular to the SC surface, showing the binding free energies of insulin, OP and OP–I to SC lipids at pH 5.5. z represents the distance between the centre of mass of insulin, OP or OP–I, and the SC lipid surface. Umbrella sampling distance, 9 nm; window resolution, 0.1 nm; sampling time, 35 ns per window; restraint force constant, 1,000 kJ mol⁻¹ nm⁻². e, MSD results parallel to the SC surface, comparing the diffusivities of insulin, OP or OP–I on SC lipids at pH 5.5. f, Representative interaction modes of OP–I with SC lipids at pH 5.5 and pH 7.0. See also Supplementary Video 4. g, MSD results comparing the diffusivities of OP–I on SC lipids under weak acidic (pH 5.5) and neutral (pH 7.0) conditions. Data are from n = 3 independent experiments. Scale bar, 50 μm (a). Source Data

Fig. 5. The mechanism of OP–I penetration in the viable epidermis.

a,b, Penetration of OP–I^FITC in HaCat spheroids observed by CLSM (a) and line-scan analysis of fluorescence intensity along the yellow arrows in a (b). Spheroids were incubated with each formulation (FITC-equivalent dose, 1 μg ml⁻¹) for 12 h, and the middle layers of the spheroids were imaged. The images are representative of n = 3 independent experiments. c, Fluorescence intensities of NBD-C6-HPC (green) and OP–I^Cy5 (red) along the yellow arrows in CLSM images of HaCat cells after 6, 12 or 24 h of incubation with OP–I^Cy5 (Cy5-equivalent dose, 1 μg ml⁻¹). See also the CLSM images in Extended Data Fig. 8a. The images are representative of n = 3 independent experiments. d, CLSM images of OP–I^Cy5 hopping on HaCat cell membranes. The images show the fixed views from a time-lapse acquisition mode in Supplementary Video 7 (Cy5-equivalent dose, 1 μg ml⁻¹). The images are representative of n = 3 independent experiments. e,f, Cell-contact-dependent transfer of OP–I^Cy5 from OP–I^Cy5-pretreated HaCat cells to untreated HaCat^GFP cells. e, Representative CLSM images of n = 3 independent experiments. f, Flow cytometry analysis of transfer efficiency at different timepoints. n = 3 independent experiments. Cy5-equivalent dose, 1 μg ml⁻¹. g, Cell-contact-dependent transfer of OP–I^Cy5 between HaCat cells on two apposed coverslips. HaCat cells on coverslip 1 were treated with OP–I^Cy5 (Cy5-equivalent dose, 1 μg ml⁻¹) for 12 h, washed and imaged using CLSM. Coverslip 1 was then pressed face-to-face with coverslip 2 seeded with untreated HaCat cells. The paired coverslips were incubated in fresh medium for 0.5 or 1 h at 37 °C, washed and analysed using CLSM imaging (left) and ImageJ fluorescence quantification (right). Data are mean ± s.d. n = 3 independent experiments. Significance was determined using one-way analysis of variance for multiple comparisons. Scale bars, 500 μm (g), 100 μm (a), 50 μm (d (left), e (left)), 25 μm (c) and 20 μm (d (right) and e (right)). Source Data

Extended Data Fig. 1

Schematic of the skin penetration mechanism of OP and its conjugate with insulin (OP-I).

Extended Data Fig. 2. Characterization of insulin conjugates OP-I and PEG-I.

a, RP-HPLC traces at an absorbance of 280 nm. b, MALDI-TOF MS spectra (OP-I was reduced to PDMA-I). c, GPC traces in H₂O. d, CD spectra in H₂O. e, BGLs of diabetic mice after subcutaneous (s.c.) injection of OP-I, PEG-I, or native insulin; insulin-equivalent dose, 5 U kg⁻¹. Data are mean ± s.d.; n = 5 mice. Source Data

Extended Data Fig. 3. SPR characterization of insulin receptor (IR) binding and activation of insulin and OP-I.

a, Schematic illustration of the experimental setup, where anti-His antibody was immobilized on the CM5 chip sensor surface, followed by the addition of Histag-ECD-IR/IGF1R, and then insulin/OP-I was introduced to study their interactions. Created in BioRender.com. b, SPR traces showing the binding of insulin and OP-I to ECD-IR; RU, resonance unit. c, Dissociation constant (K_D), association rate constant (k_on), dissociation rate constant (k_off), and half-life (t_1/2) of insulin or OP-I binding to ECD-IR, calculated from SPR binding curves in Extended Data Fig. 3b. d, SPR traces showing minimal binding of insulin and OP-I to ECD-IGF1R; RU, resonance unit. e, Representative binding modes from all-atom MD simulations of insulin and OP-I on the ECD-IR. f, PMF analysis of the binding affinities with two major ECD-IR binding sites. Umbrella sampling distance: 5 nm; window resolution: 0.1 nm; sampling time: 20 ns per window; restraint force constant: 1000 kJ mol⁻¹ nm⁻². Source Data

Extended Data Fig. 4. Characterization of OP-I permeability in mouse skin.

a, Fluorescence distribution in the skin (left) and its s.c. fat (right, enlarged view) of the C57BL/6J mouse dorsal skin after topical application with OP-I^Cy5 for timed intervals (Cy5-equivalent dose, 0.2 ml of 10 μg ml⁻¹ applied on 1.13 cm²). BODIPY (green) was used to label the s.c. adipose tissue, and DAPI (blue) was used to label the nuclei. b, CLSM images of the slices of the C57BL/6J mouse dorsal skin after topical application with OP-I^Cy5 (Cy5-equivalent dose, 0.2 ml of 10 μg ml⁻¹ applied on 1.13 cm²) for 2 or 4 h. c,d, CLSM images (c) and line-scan analysis of FITC fluorescence intensity from the skin surface to the subcutis along randomly drawn lines (yellow arrows in c) (d) of vertically sectioned slices of treated mouse dorsal skin after 4 h topical administration of OP-I^FITC. The images are representative of n = 3 independent experiments. Source Data

Extended Data Fig. 5. Biological safety evaluation of OP-I for transdermal application.

a-c, Representative images of SEM (a), H&E staining (b), and immunohistological staining with TUNEL assay (green) and Hoechst (blue) (c) of the mouse dorsal skin after 4 h of topical treatment with OP-I, PEG-I, native insulin, or PBS (insulin-equivalent dose, 116 U kg⁻¹; 0.2 ml of 0.5 mg ml⁻¹; application area, 1.13 cm²). d,e, Representative images of H&E staining (d) and immunohistological staining with TUNEL assay (green) and Hoechst (blue) (e) of the treated sites of the minipig abdominal skin after 4 h of topical treatment with OP-I, PEG-I, native insulin, or PBS (insulin-equivalent dose, 40 ml of 1 mg ml⁻¹, 29 U kg⁻¹; application area, 400 cm²). The untreated skin was designated as the Control group. The images are representative of n = 3 independent experiments.

Extended Data Fig. 6. Co-localization of OP-I with SC lipids.

a, CLSM images of the adhesive tape peeled from the mouse dorsal skin after 4 h of topical application with OP-I^Cy5, PEG-I^Cy5, or insulin^Cy5 (Cy5-equivalent dose, 0.2 ml of 10 μg ml⁻¹; application area, 1.13 cm²). The images are representative of n = 3 independent experiments. b, CLSM images of the vertically sectioned slices of minipig abdominal skin after 4 h of topical application with OP-I^Cy5 (Cy5-equivalent dose, 10 ml of 10 μg ml⁻¹; application area, 100 cm²). The images are representative of n = 3 independent experiments. c, Enlarged view of the selected area in (b) outlining a corneocyte. d, Line-scan analysis of NBD-C6-HPC and Cy5 fluorescence along the yellow arrow in (c). Cy5 fluorescence is shown in red, and SC intercellular lipids stained with NBD-C6-HPC are in green. Source Data

Extended Data Fig. 7. Modelling of the interactions of OP-I with SC lipids.

a, PMF analyses along the surface of the model SC lipids in different paths to estimate the energy barriers for diffusion. The energy barriers were 2.2 ± 0.3 kcal mol⁻¹ for insulin, 0.9 ± 0.1 kcal mol⁻¹ for OP, and 1.7 ± 0.2 kcal mol⁻¹ for OP-I. Umbrella sampling distance: 2 nm; window resolution: 0.1 nm; sampling time: 35 ns per window; restraint force constant: 1000 kJ mol⁻¹ nm⁻². b, Representative local binding modes of insulin and OP on the rough model SC lipid surface at pH 5.5. c, Traces of insulin, OP, and OP-I on the model SC lipids at pH 5.5 in three 100-ns independent MD trajectories for each. See also Supplementary Video 3. Data are mean ± sem; n = 3 independent experiments. Source Data

Extended Data Fig. 8. Cell-contact-mediated transfer of OP-I between HaCat cells.

a, CLSM imaging of the OP-I^Cy5 retention in cell membranes of HaCat cells cultured with OP-I^Cy5 for 6, 12, or 24 h (Cy5-equivalent dose, 1 μg ml⁻¹); the cell membrane was stained with NBD-C6-HPC in green. The images are representative of n = 3 independent experiments. b, Enlarged view of the viable epidermis layer in Extended Data Fig. 6b. Minipig abdominal skin was treated with OP-I^Cy5 (Cy5-equivalent dose, 10 ml of 10 μg ml⁻¹; application area, 100 cm²) via topical application for 4 h. Cy5 fluorescence is shown in red, DAPI in blue, and NBD-C6-HPC fluorescence in green. The images are representative of n = 3 independent experiments. c, Line-scan analysis of NBD-C6-HPC and Cy5 fluorescence along the yellow arrow in (b). d, Flow cytometry analysis of the effect of cell density on OP-I^Cy5 transfer efficiency (Cy5-equivalent dose, 1 μg ml⁻¹). The cell densities were 1.25 × 10⁵ per well or 2.5 × 10⁵ per well for both cell types (OP-I^Cy5-pretreated HaCat cells and untreated HaCat^GFP cells), with untreated HaCat cells and HaCat^GFP cells as controls. Data are mean ± s.d.; n = 3 independent experiments. Significance was determined using a two-tailed unpaired Student’s t-test. Source Data