Controlled Interfacial Polymer Self‐Assembly Coordinates Ultrahigh Drug Loading and Zero‐Order Release in Particles Prepared under Continuous Flow

Microparticles are successfully engineered through controlled interfacial self‐assembly of polymers to harmonize ultrahigh drug loading with zero‐order release of protein payloads. To address their poor miscibility with carrier materials, protein molecules are transformed into nanoparticles, whose surfaces are covered with polymer molecules. This polymer layer hinders the transfer of cargo nanoparticles from oil to water, achieving superior encapsulation efficiency (up to 99.9%). To control payload release, the polymer density at the oil–water interface is enhanced, forming a compact shell for microparticles. The resultant microparticles can harvest up to 49.9% mass fraction of proteins with zero‐order release kinetics in vivo, enabling an efficient glycemic control in type 1 diabetes. Moreover, the precise control of engineering process offered through continuous flow results in high batch‐to‐batch reproducibility and, ultimately, excellent scale‐up feasibility.


Introduction
The encapsulation of payloads, such as solar modules, [1] drugs, [2] fragrances, [3] food ingredients, [4] gas bubbles, [5] and living www.advmat.de www.advancedsciencenews.com and improve patient compliance. [9] However, the loading capacity of engineered systems is inherently constrained by the miscibility of active ingredients with carrier materials, particularly for water-soluble proteins. [10] By converting drug molecules into nanoparticles (NPs), whose surfaces could adsorb carrier materials, the compatibility limitation with carrier materials could be mitigated [9b,11] and record of drug mass fraction of 77% could be achieved. [12] However, with such a high mass fraction of therapeutics, an initial large bolus of therapeutics is released immediately-the so called "burst release"-upon contacting with release medium. [13] Such a burst release of therapeutics is not only economically inefficient but also pharmacologically dangerous. [14] For instance, burst release of insulin may be fatal owing to hypoglycaemia. Therefore, a versatile technique that can engineer systems simultaneously featuring ultrahigh mass fraction of therapeutics and controlled release of payloads is highly desirable. [15] The preparation of drug-loaded particles under continuous flow offers significant technical and economic advantages over conventional batch methods. [16] For instance, particle preparation using batch-type reactors constantly faces batch-to-batch variations and scale-up difficulties. [9b,17] These shortcomings of conventional bulk methods can be attributed to the variability of conditions during particles formation, which results in insufficient control of the emulsification and solidification processes. [16b] By contrast, the good control of engineering process together with the online preparation of microparticles offered by the continuous flow process leads to a precise control over the physicochemical properties of microparticles, a high batch-to-batch reproducibility, and finally a good industrial scale-up feasibility.
To this end, in the present study, we controlled the interfacial distribution of polymers under continuous flow to integrate high drug loading and controlled release of payloads into a single microparticle. Specifically, a polymer with charged functional groups was used to facilitate electrostatic attraction toward oppositely charged NPs and achieve universal surface decoration for cargo NPs. This adsorbed polymer layer was expected to increase the affinity of cargo NPs to the oil phase, [18] thereby blocking the phase transfer of proteins from oil to water (Scheme 1a), and enable efficient cargo encapsulation. By improving the adsorption of polymer molecules at the Scheme 1. Microparticles with controlled distribution of components integrating ultrahigh drug loading and zero-order release kinetics. a) Surface adsorption of polymers onto cargo NPs blocked their transfer from oil to water and enabled efficient encapsulation. b) Enhanced oil-water interface adsorption of polymers minimized burst release and enabled controlled release kinetics. c) Programmable element distribution integrated efficient encapsulation and controlled release in a single microparticle. d) Molecular structures of AD and ADS. e) Schematic diagram of the continuous flow device used for preparing microparticles with programmable element distribution. Protein NPs were prepared via nanoprecipitation, and cargo-NPencapsulated microparticles were engineered via droplet microfluidics. www.advmat.de www.advancedsciencenews.com oil-water interface, a thick interfacial polymer layer was constructed to minimize the burst release of encapsulated payloads (Scheme 1b). Benefiting from the programmable distribution of elements, the obtained microparticles were expected to integrate ultrahigh mass fraction of cargoes and controlled release kinetics (Scheme 1c).
Considering the impact of electrostatic forces on surface adsorption, spermine-modified acetylated dextran (ADS), a cationic polymer with abundant amino groups, was selected as the carrier material [12,19] (Scheme 1d). A polymer with a similar backbone but without amino groups, acetylated dextran (AD), served as the control. A microfluidic platform offering optimum control of the preparation process together with a continuous preparation feature, [16,20] was used to prepare cargo NPs and the corresponding microparticles. Protein NPs were prepared via nanoprecipitation using a coflow microcapillary device (Scheme 1e). [21] The drug solution flowed through the space between the inner and outer capillaries, and the drug nonsolvent-containing polymer molecules were pumped into the inner capillary. During the solvent diffusion, the protein molecules self-assembled into NPs. Next, cargo-NP-encapsulated microparticles were prepared using the second coflow microcapillary device. Cargo NPs, together with polymers, were pumped into the inner capillary to serve as the dispersed phase, which was emulsified with a continuous phase containing poly(vinyl alcohol) and MgCl 2 . MgCl 2 enables the immiscibility of the continuous phase (water) with the dispersed phase (oil). [22] The resultant droplets were transformed into solidified microparticles after the depletion of solvents in the oil phase.

Electrostatic-Force-Mediated Surface Adsorption Blocked the Phase Transfer of Cargo NPs
We speculated that the adsorbed ADS would hinder the transfer of protein NPs from oil to water. Owing to its short in vivo halflife and the urgent need for improving patient compliance, insulin was selected as the model drug. To verify our hypothesis, acetone containing insulin NPs was stacked on the top of an aqueous phase containing MgCl 2 ·6H 2 O (Figure 1a). The apparent water solubility of insulin improved (>15 mg mL −1 in 1 h at 25 °C) after NP formation. AD did not affect the phase transfer of insulin; ≈71.0% insulin was detected in water after 2 h of incubation (Figure 1b). Under the action of ADS, ≈0.5% insulin was detected in water phase after 2 h. After 24 h of incubation, insulin concentration in the water phase slightly increased to 2.0 ± 0.2%, indicating that the ADS molecules efficiently blocked the transfer of insulin NPs into the water phase.
Further, the adsorption of polymers onto the surface of insulin NPs was studied using a quartz crystal microbalance. Insulin NPs were immobilized on a quartz disk to form a film via spin coating (Figure 1c). When a polymer solution (10 mg mL −1 , in acetone) flowed over, polymer molecules were adsorbed onto the surface of insulin NPs, leading to a frequency decrease (Figure 1d). After rinsing with acetone, the final frequency decrease caused by the adsorption of ADS (−30.0 ± 0.6 Hz) was significantly (p < 0.001) higher than that caused by the adsorption of AD (−6.4 ± 2.0 Hz). Similarly, the adsorbed mass (∆m) [23] of the polymer solution in the ADS group was significantly (p < 0.001) higher than that in the AD group ( Figure S1, Supporting Information).
As expected, AD only slightly increased the zeta potential of insulin NPs from −4.0 ± 1.6 to −0.7 ± 1.4 mV (Figure 1e). Following surface adsorption of ADS, the zeta potential of insulin NPs increased to 14.3 ± 0.4 mV. Within 24 h after preparation, the size of insulin NPs with AD increased from 459.7 to 4240.0 nm (Figure 1f) and the polydispersity index increased from 0.29 to 0.83 (Figure 1g), suggesting the aggregation of insulin NPs in AD acetone solution. By contrast, following surface adsorption of ADS, the particle size of insulin NPs was maintained at ≈200 nm and the polydispersity index was <0.1 at 24 h after preparation. Therefore, surface adsorption of ADS likely stabilized insulin NPs in the oil phase.
Furthermore, isothermal titration calorimetry was used to elucidate the thermodynamic profile of polymer-insulin NP interaction. The polymer solution was titrated into a suspension of insulin NPs, and heat change associated with the reaction was recorded (Figure 1h). The corresponding dilution heat ( Figure S2, Supporting Information) was subtracted to yield the corrected heat, resulting solely from the polymerinsulin NP interaction. The titration of AD or ADS into insulin NPs was monotonically exothermic ( Figure S3, Supporting Information). Since the values of Gibbs free energy change (∆G) were negative (−30.7 ± 2.4 kJ mol −1 for AD complex and −23.9 ± 0.5 kJ mol −1 for ADS complex), the interaction between polymer molecules and insulin NPs was a spontaneous process ( Figure 1i). Moreover, the adsorption of AD or ADS onto insulin NPs was enthalpy-driven, in which the favorable enthalpy change (∆H < 0) was partially offset by the unfavorable entropy loss (∆S < 0) ( Figure S4, Supporting Information). [24] Next, using multiangle (173°, 13°, and 90°) dynamic light scattering, we measured the number concentration of insulin NPs, which is based on the transformation of intensityweighted size distribution into absolute concentration distribution. [25] Accordingly, the total surface area of insulin NPs was estimated. The number of polymer molecules per 10 nm 2 surface of insulin NPs was calculated by considering the stoichiometric relationship between the polymer molecules and insulin NPs (Figure 1j). The number of ADS molecules (3.6 ± 0.3 per 10 nm 2 ) adsorbed on insulin NPs was significantly (p < 0.001) higher than that of AD molecules (0.7 ± 0.1 per 10 nm 2 ), consistent with the results of quartz crystal microbalance analysis.
We then explored the adsorption driving forces using molecular dynamics (MD) simulations. As simulating the structure of amorphous NPs is difficult, one insulin molecule rather than an insulin NP was used for simulations. The initial configurations of the insulin-polymer complexes were obtained from docking simulations ( Figure S5, Supporting Information). For the complexes containing AD, the contact sites were mainly positively charged residues of insulin, while the insulin surface directly facing ADS was a negatively charged region (Figure 2a,b).
Regarding the interaction between ADS and insulin, the contribution of electrostatic forces, derived from the interaction between the amino groups of ADS and the negatively charged www.advmat.de www.advancedsciencenews.com residues of insulin, was larger than that of van der Waals force. The contribution of electrostatic attraction to the formation of the ADS-insulin complex was −317.1 kJ mol −1 (Figure 2c), which was much higher than that to the formation of the AD-insulin complex (−139.2 kJ mol −1 ). By contrast, the contribution of van der Waals forces to the formation of the AD-insulin complex (−120.4 kJ mol −1 ) was close to that to the formation of the ADSinsulin complex (−153.2 kJ mol −1 ) (Figure 2d). Seven insulin residues contributed to the insulin-AD interaction (Figure 2e). Among these, the positively charged arginine (from 1 to 134) and . c) Polymer adsorption onto insulin NPs as monitored using a quartz crystal microbalance. d) Normalized frequency of the third overtone (∆F 3 /n) as a function of time when polymer solution flowed over the surface of insulin NPs. e-g) Effects of AD and ADS on the zeta potential (e), size stability (f), and polydispersity index (g) of insulin NPs (n = 3). h) Adsorption of polymer molecules onto insulin NPs as monitored using isothermal titration calorimetry. i) Gibbs free energy change deduced from isothermal titration calorimetry enthalpograms (n = 3). j) Number of adsorbed polymer molecules per 10 nm 2 surface of insulin NPs (Student's t-test; n = 3; ADS group vs the corresponding AD group; ***p < 0.001).
www.advmat.de www.advancedsciencenews.com asparagine (from 0 to 51) residues showed the highest number of contact points with AD. Meanwhile, 12 insulin residues contributed to insulin-ADS complexation ( Figure 2e). The number of contact points with ADS for the neutral phenylalanine (up to 79) and negatively charged glutamate (up to 71) residues was much higher than those for the remaining 10 residues.
Overall, both AD and ADS were adsorbed onto the surface of insulin NPs. However, compared with AD molecules, more ADS molecules were adsorbed onto the surface of insulin NPs and formed a thicker polymer layer because of the stronger electrostatic forces. While this ADS layer could efficiently hinder the diffusion of insulin into the water phase, the adsorbed AD molecules on the insulin NPs surface showed little effect on insulin diffusion because of their small amount.

Zero-Order Release Kinetics Enabled by the Enhanced Oil-Water Interfacial ADS Adsorption
Compared with AD molecules, more ADS molecules were adsorbed onto the surface of insulin NPs and formed a thicker polymer layer, which efficiently hindered the transfer of insulin NPs into the water phase. The insulin-encapsulated ADS microparticles, Insulin@ADS, were expected to exhibit ultrahigh insulin encapsulation efficiency and, consequently, ultrahigh insulin loading. The obtained Insulin@ADS were spherical and intact, with an average particle size of ≈25 µm (Figure 3a). By contrast, some insulin-NP-encapsulated AD microparticles (Insulin@AD) showed hollow structures, which can be ascribed to the transfer of insulin NPs during the emulsification and droplet solidification processes ( Figure S6, Supporting Information). The low polydispersity index indicates a narrow size distribution of these two microparticles ( Figure S7, Supporting Information). As expected, the encapsulation efficiency of Insulin@ADS was 94.7 ± 2.2%, which was significantly higher than that of Insulin@AD (14.9 ± 8.7%) (Figure 3b). Consequently, the surface adsorption of ADS successfully realized ultrahigh drug loading at 47.4 ± 1.1%; meanwhile, insulin loading with Insulin@AD was only 7.4 ± 4.4% (Figure 3c).
Surprisingly, Insulin@ADS showed a linear payload release profile, with minimal burst release (Figure 3d). Within the first 1 h, only ≈1% insulin was released from Insulin@ADS.   The insulin release profile at 10 days fitted well to a linear model with a R 2 value of 0.994, indicating zero-order insulin release kinetics. Subsequently, the release rate of insulin from Insulin@ADS decreased gradually. By contrast, Insulin@AD showed an obvious "burst release," with 55.8 ± 3.5% insulin released on the first day ( Figure 3d). Further, the secondary structure of the released insulin was studied using circular dichroism (CD) spectroscopy. The CD spectrum of insulin released from Insulin@ADS was consistent with that of native insulin ( Figure 3e). The molar ellipticity ratio between the negative bands at 208 and 222 nm, representing the overall conformation of insulin molecules, [26] for the released insulin (1.22) was close to that for native insulin (1.20). Therefore, the secondary structure of insulin was well-preserved during the encapsulation and release processes. Consistent with previous reports, [27] both AD and ADS microparticles showed good compatibility with primary dermal fibroblast cells, even at the concentration of 10 mg mL −1 ( Figure S8, Supporting Information), validating their potential to serve as drug delivery systems.
Burst release is mainly attributed to the intensive distribution of therapeutics near the surface of the delivery systems. [28] In delivery systems loaded with an ultrahigh mass fraction of cargos, avoiding the burst release of payloads, especially watersoluble ones, is challenging. We speculate that the controlled   (b) and loading degree (c) of insulin in the obtained microparticles (Student's t-test; n = 3; ADS group vs the corresponding AD group; ***p < 0.001). d) Insulin release profile from Insulin@ADS and Insulin@AD (n = 3). e) Circular dichroism spectra of native insulin and insulin released from Insulin@ADS. f) Dynamic interfacial tension over time for DCM and water without or with MgCl 2 ·6H 2 O (60%, w/w) dissolved in the water phase (n = 3). g) Insulin release profile from insulin-loaded ADS microparticles prepared with and without MgCl 2 ·6H 2 O (60%, w/w) in the water phase (n = 3). h) Snapshots of ADS distribution at the DCM-water interface after simulation for 200 ns. For clarity, the water and DCM molecules are drawn as lines, while ions (Mg 2+ and Cl − ) and ADS molecules are drawn as van der Waals spheres. i) The z-distribution density profiles of different components, including DCM, water, and ADS, without or with MgCl 2 ·6H 2 O (60%, w/w) dissolved in the water phase. j) ADS density variation as a function of distance from water molecules without or with MgCl 2 ·6H 2 O (60%, w/w) dissolved in the water phase. k) SEM images of microparticles prepared with and without MgCl 2 ·6H 2 O. l) Effect of MgCl 2 ·6H 2 O concentration on the particle size of Insulin@ADS (n = 3).
www.advmat.de www.advancedsciencenews.com release of insulin from Insulin@ADS, even with an ultrahigh insulin loading degree (47.4 ± 1.1%), can be attributed to the presence of MgCl 2 . ADS is amphiphilic, with cationic hydrophilic (spermine groups) and lipophilic (acetalated dextran) components. In the presence of salts in the water phase, counterions screened the electrostatic repulsion among the ionic amphiphilic polymers, ADS. [29] MgCl 2 likely increased the density of ADS at the oil-water interface, forming a thicker ADS layer at the surface of the obtained microparticles. This thicker ADS layer prevented the initial burst release and provided efficient control of the release rate of encapsulated cargos.
The adsorption of polymers at the oil-water interface affects interfacial tension. [30] To thoroughly illustrate the effect of MgCl 2 on interfacial tension, water immiscible dichloromethane (DCM) was selected as the oil phase. The effects of MgCl 2 on the interfacial tension between ADS DCM solution (10 mg mL −1 ) and water were evaluated. In the absence of MgCl 2 in water phase, the interfacial tension decreased from 10.5 ± 0.2 to 8.9 ± 0.4 mN m −1 within 30 s and then remained constant at around 9.0 mN m −1 (Figure 3f). Meanwhile, in the presence of MgCl 2 ·6H 2 O (60%, w/w), the interfacial tension gradually decreased from 11.0 ± 0.4 to 6.2 ± 0.4 mN m −1 in 200 s.
To further verify the effects of MgCl 2 on the release profiles of payloads, we prepared insulin-loaded ADS microparticles using an emulsion and solvent diffusion method, with DCM as the oil-phase solvent. In the presence of MgCl 2 , cumulative insulin released from the insulin-loaded ADS microparticles decreased from 15.5 ± 0.6% to 9.8 ± 0.8% in 1 h ( Figure 3g) and further from 50 ± 1.2% to 28.3 ± 4.0% in 24 h. Therefore, MgCl 2 in the water phase indeed enhanced the adsorption of ADS molecules at the oil-water interface, diminishing the initial burst release and efficiently decreasing insulin release rate.
Furthermore, MD simulations were performed to examine the effect of MgCl 2 on ADS adsorption at the DCM-water interface. As shown in Figure 3h, the addition of MgCl 2 to the water phase enhanced interfacial ADS adsorption. In the presence of MgCl 2 in water phase, the interfacial ADS density increased from 94.4 to 109.1 g cm −3 , while the width of ADS peak increased from 25.0 to 35.5 Å ( Figure 3i). As MgCl 2 enhanced the oil-water interfacial adsorption of ADS molecules, the calculated interface thickness, that is, the distance between two points where the density of water and oil reached 90% of their corresponding bulk densities, [31] increased from 20.0 to 22.5 Å. A radial distribution function represents the density variation of a particle as a function of its distance from the reference particles. The radial distribution function between ADS and water molecules decreased after the introduction of MgCl 2 into the water phase (Figure 3j), suggesting that MgCl 2 "pushed" ADS molecules to the oil phase. [29a] In the absence of MgCl 2 , Insulin@ADS showed a porous structure (Figure 3k), which can be attributed to the distribution of ADS molecules closer to the water side. The shortened distance between water-insoluble ADS and water phase may accelerate the local solidification of ADS during the solvent depletion process, resulting in the formation of porous structures. After "pushing" ADS molecules to the oil phase, MgCl 2 may further delay the solidification of oil droplets, thereby decreasing the diameter of microparticles ( Figure 3l) and increasing the density of particle matrix. Even in the presence of 5% MgCl 2 ·6H 2 O, the diameter of Insulin@ADS decreased from 79 to 37 µm, which significantly reduced their porosity. With increase in MgCl 2 ·6H 2 O concentration to 60%, the diameter decreased to 25 µm and the structure of Insulin@ADS became more compact.
Strictly steady release of therapeutics, particularly watersoluble proteins, from the carriers is a great challenge. In the presence of MgCl 2 in the water phase, droplet solidification was delayed, leading to the formation of more compact microparticles. More ADS molecules were adsorbed at the oil-water interface, forming a thicker polymer layer at the surface of microparticles after solidification. This thicker polymer layer and more compact particle matrix facilitated the zero-order release kinetics of insulin from the fabricated Insulin@ADS without burst release.

Versatile Efficient Encapsulation and Excellent Batch-to-Batch Reproducibility
The key driving force of the adsorption process is the electrostatic attraction between the amino groups of ADS and the negatively charged residues of insulin. Therefore, we hypothesized that ADS can be adsorbed onto the negatively charged areas of other cargo NPs, forming microparticles with an ultrahigh mass fraction of proteins. Based on our hypothesis, we selected two additional proteins with abundant acid residues, namely bovine serum albumin (BSA) and β-lactoglobulin (β-LG), to demonstrate the versatility of microparticles with programmable element distribution.
Isothermal titration calorimetry (Figures S9 and S10, Supporting Information) was used to elucidate the stoichiometric relationship between the polymer molecules and protein NPs. Considering the number concentration and average diameter of protein NPs, we calculated the number of polymer molecules adsorbed on the surface of cargo NPs. The number of ADS molecules per 10 nm 2 surface of BSA NPs (5.2 ± 0.2) was significantly (p < 0.001) higher than that of AD molecules (1.6 ± 0.2) (Figure 4a). Regarding β-LG NPs, the estimated number of ADS and AD molecules per 10 nm 2 of the surface of NPs was 2.5 ± 0.2 and 0.8 ± 0.7, respectively. As expected, the density of ADS molecules on the surface of β-LG NPs was significantly (p < 0.01) higher than that of AD molecules.
For both BSA and β-LG NPs, the adsorption of polymer solution ( Figure S11 Supporting Information). In the presence of ADS, the size of BSA NPs only slightly increased from 149.7 to 185.7 nm at 24 h after preparation, while the size of β-LG NPs was maintained at ≈140 nm (Figure 4e). Moreover, in the presence of ADS, the polydispersity index for both BSA and β-LG NPs remained below 0.2 ( Figure S12, Supporting Information).
To further confirm the versatility of our interfacial adsorption strategy, we engineered microparticles loaded with BSA (BSA@ADS) and β-LG NPs (β-LG@ADS). Both microparticles exhibited a spherical morphology (Figure 4f) with homogenous size distribution ( Figure S13, Supporting Information). In addition, both BSA@ADS and β-LG@ADS showed ultrahigh protein encapsulation efficiency (>98%) (Figure 4g). Owing to this ultrahigh encapsulation efficiency, the protein loading degree in the obtained microparticles was >49% (Figure 4h). Both BSA@ADS and β-LG@ADS showed a linear payload release profile, with minimal burst release (Figure 4i), which may be ascribed to the thicker ADS layer and more compact particle matrix. Only 0.2% BSA and 0.4% β-LG were released from the corresponding microparticles in the first hour. The cumulative BSA released increased to ≈97.4% after 22 days of the release test, and ≈95.6% β-LG was released after 18 days. The goodness-of-fit for the linear regression model was 0.998 for both BSA and β-LG release profiles within 22 and 18 days, respectively. Compared with that of insulin, the slower release profiles of BSA and β-LG could be attributed to their relatively larger molecular size.
As the microparticle preparation process operates in a continuous mode, the physicochemical properties of the prepared microparticles are independent of batch size. To demonstrate the precise control capability toward the engineering process, we characterized 10 batches of microparticles prepared by different researchers (Figure 5). Across the 10 batches, the average particle size of Insulin@ADS, BSA@ADS, and β-LG@ADS varied in the range of 24-25, 32-34, and 28-29 µm, respectively ( Figure 5a). All the corresponding coefficients of variance for particle size were <6% (Figure 5b  www.advmat.de www.advancedsciencenews.com <1.5%, respectively. Therefore, the developed platform could resolve the controllability and reproducibility of the engineering process for high-drug-loaded microparticles.
Overall, the developed system is sufficiently versatile to achieve ultrahighly efficient protein encapsulation in the obtained microparticles with a programmable element distribution. ADS molecules with abundant amino groups could be adsorbed onto the surfaces of insulin, BSA, and β-LG NPs. The adsorbed ADS layer successfully improved the dispersity of protein NPs in the oil phase, hindered the leakage of cargo NPs into the water phase, and realized ultrahighly efficient drug encapsulation. After formulating insulin, BSA, and β-LG molecules into NPs, decoration was required only at the surface of the NPs. Comparison with molecular-level modification, these NP surface decorations greatly reduced the amount of polymer molecules, thereby achieving ultrahigh mass fraction of therapeutics in the obtained microparticles.

Insulin@ADS Yields Favorable Zero-Order Adsorption Kinetics in Diabetic Rats
With a high loading degree of therapeutics and controlled drug release profile, fewer excipients are required to deliver a certain amount of therapeutics. Therefore, microparticles with programmable element distribution are expected to efficiently increase treatment efficacy and avoid potential side effects caused by excessive excipients. Basal insulin is of paramount importance for controlling blood glucose levels in patients with diabetes. To mimic basal insulin secretion, an ideal system should provide minimal burst release and prolonged insulin supply. With a constant insulin release rate over 10 days, Insulin@ADS was expected to maintain plasma insulin concentration for a prolonged period. To verify the in vivo performance of the obtained microparticles, we evaluated the effects of Insulin@ADS [50 international unit kg −1 (IU kg −1 )] on the stability of plasma insulin concentration and its therapeutic efficacy in a rat model of type 1 diabetes (Figure 6a). Insulin@ AD (50 IU kg −1 ), insulin solution (10 IU kg −1 ), normal saline, and bare ADS microparticles (controls) were administered.
After subcutaneous injection of insulin solution, the plasma insulin concentration rapidly dropped to <0.2 mUI L −1 within 2 days (Figure 6b). Owing to the rapid release of insulin, the enhanced plasma insulin level (>20 mUI L −1 ) lasted only 2 days with Insulin@AD injection. With subcutaneous injection of Insulin@ADS, the plasma insulin level peaked at 8.0 h after administration, reaching the value of 125.4 ± 18.1 mUI L −1 ( Figure S14  Insulin@ADS successfully maintained plasma insulin concentration >50 mUI L −1 for at least 8 days. Thanks to encapsulation within ADS microparticles, the mean residence time of insulin was significantly prolonged (p < 0.001) from 0.4 ± 0.2 (for insulin solution) to 5.2 ± 0.6 days (Figure 6c). By contrast, AD microparticles improved the mean residence time of insulin to only 1.8 ± 0.5 days. The area under the plasma insulin concentration-time curve was 81.9 ± 15.0, 190.9 ± 25.3, and 698.3 ± 82.0 mUI L −1 days, for insulin solution, Insulin@AD, and Insulin@ADS, respectively (Figure 6d). Therefore, Insulin@ADS efficiently enhanced insulin bioavailability, which was 3.6 times that of Insulin@AD.
To further evaluate the in vivo controlled-release capability of Insulin@ADS, we plotted the area under the plasma insulin concentration-time curve versus time (Figure 6e). The high goodness-of-fit of the linear model (R 2 = 0.989) reflected the in vivo zero-order release of insulin from Insulin@ADS within 6 days. Moreover, in vivo insulin absorption fraction was calculated from the pharmacokinetic profile using the deconvolution method ( Figure S15  . Insulin@ADS microparticles yielded zero-order insulin release kinetics and achieved persistent and efficient blood glucose control. a) Schematic illustration of the in vivo experiment. b) Plasma insulin level after subcutaneous injection of Insulin@ADS, Insulin@AD, and insulin solution in diabetic rats (n = 6). c,d) Pharmacokinetic features for Insulin@ADS, Insulin@AD, and insulin solution, including the mean residence time (c) and plasma insulin area under the curve (d) (one-way ANOVA with post-hoc Bonferroni's test; n = 6; **p < 0.01 and ***p < 0.001). e) Plasma insulin area under the curve as a function of time fitting the zero-order kinetics model (n = 6). f) In vivo insulin absorption showed a strong linear correlation with in vitro insulin release. g,h) Effects of Insulin@ADS, Insulin@AD, and insulin solution on blood glucose level (g) in diabetic rats and glucose area under the curve (h) in 6 days after administration (one-way ANOVA with post-hoc Bonferroni's test; n = 6; ***p < 0.001). i-k) Representative histological images (i) and VTEA scatterplots (j) of adjacent tissues to the injection site in the Insulin@ADS group. CD68 is indicated in red and 4′,6-diamidino-2-phenylindole (DAPI) in cyan. Summary of VTEA (k) compared the percentage of CD68-positive cells (Student's t-test; n = 6; Insulin@ADS group vs the corresponding Insulin@AD group; ***p < 0.001). All the results are expressed as mean ± SD.

www.advmat.de www.advancedsciencenews.com
Adv. Mater. 2023, 35, 2211254 expected, in vivo zero-order insulin absorption was observed for Insulin@ADS over 6 days (R 2 = 0.998). As shown in Figure 6f, Insulin@ADS exhibited a point-to-point correlation between in vitro release and in vivo absorption, indicating level-A in vitroin vivo correlation. [32] This in vitro-in vivo correlation indicates that we could accurately predict the in vivo insulin absorption fraction from the in vitro insulin release profile. Furthermore, we monitored the blood glucose levels of engineered formulations, reflecting their therapeutic effects in rats with type 1 diabetes. Insulin solution decreased the blood glucose level to 125.4 ± 58.3 mg dL −1 at 2 h after administration (Figure 6g). However, within 12 h, the blood glucose level returned to a hyperglycemic state (>350 mg dL −1 ). Normal saline and bare ADS microparticles did not affect the glycemic levels in rats ( Figure S16, Supporting Information). The therapeutic effect of Insulin@AD was maintained for ≈2 days. By contrast, Insulin@ADS gradually reduced the blood glucose level to 155.1 ± 23.9 mg dL −1 at 4 h after administration, and its therapeutic effect was maintained for nearly 8 days. After 10 days, the blood glucose level returned to the initial hyperglycemic state. The blood glucose area under the curve for Insulin@ADS was 0.68 ± 0.06 g dL −1 days (Figure 6h), which was significantly smaller than that for both Insulin@AD (1.64 ± 0.23 g dL −1 days; p < 0.001) and insulin solution (2.66 ± 0.26 g dL −1 days; p < 0.001). Therefore, the therapeutic effect of each formulation was consistent with the corresponding pharmacokinetic profiles. Moreover, Insulin@ADS successfully alleviated weight loss in rats, a symptom of type 1 diabetes ( Figure S17, Supporting Information).
Next, the inflammatory response induced by the administered microparticles was evaluated by monitoring the level of cluster of differentiation 68 (CD68), a universal macrophage marker, [33] in tissues adjacent to the injection site (Figure 6i). Volumetric tissue exploration and analysis (VTEA) was used to quantify the fraction of macrophages (Figure 6j) adjacent to the injection site. [34] As opposed to that in untreated tissues (0.9 ± 0.4%, data not shown), both Insulin@ADS and Insulin@AD stimulated an inflammatory response in tissues adjacent to the injection site (Figure 6k). The peak level of macrophages on day 8 after the injection could be attributed to the degradation of the polymeric matrix, which triggered dextran release and macrophage infiltration. [35] Regardless of the time after administration, the density of macrophages around the injection site was significantly lower for Insulin@ ADS than for Insulin@AD. Moreover, the inflammatory reaction induced by Insulin@ADS was much weaker than that induced by Insulin@AD, which may be attributed to the higher mass of AD administered. The insulin loading degree for Insulin@ADS (47.4 ± 1.1%) was ≈5 times higher than that for Insulin@AD (7.4 ± 4.4%). For a dose of 50 IU kg −1 , the administered mass of AD was 10.3 times higher than that of ADS. Moreover, the duration of the therapeutic effect of Insulin@ADS was 3 times longer than that of Insulin@AD. Compared with the mass of AD (for Insulin@AD) injected in 6 days, the total administered mass of ADS (for Insulin@ADS) was 44.2 times lower. Therefore, the ultrahigh insulin loading degree greatly reduced the total mass of carrier material administered and efficiently alleviated the intensity of carrier-material-induced inflammation reaction.
To evaluate its long-term (35 days) therapeutic efficacy, Insulin@ADS (50 IU kg −1 ) was subcutaneously administered once a week. Throughout the 35 days treatment, Insulin@ADS maintained plasma insulin levels between 66 and 115 mIU L −1 (Figure 7a). Compared with that in the Insulin@ADS group, much greater fluctuations in plasma insulin concentration were observed in the Insulin@AD group. At the same dose of insulin administered, the plasma insulin area under the curve for Insulin@ADS (3270.1 ± 303.1 mIU L −1 days) was 3.2 times higher than that for Insulin@AD (778.5 ± 257.0 mIU L −1 days, Figure 7b). Insulin@ADS strictly controlled the glycemic level within the normal range throughout the 35 days treatment; meanwhile, the rats remained in a hyperglycemic state after Insulin@AD administration (Figure 7c). The glucose area under the curve for the Insulin@ADS group was 3.47 ± 0.34 g dL −1 days, which was significantly (p < 0.001) smaller than that for the Insulin@AD group (12.31 ± 0.96 g dL −1 days, Figure 7d). Hemoglobin A1c (HbA1c) is the gold standard metric for longterm glycemic control. Therefore, we monitored the HbA1c percentage throughout the treatment period. HbA1c levels were ≈5.7% before the treatment (Figure 7e). After the 35 days treatment, the HbA1c level in the Insulin@ADS group (4.2 ± 0.8%) was significantly lower than that in the Insulin@AD group (11.5 ± 1.0%).
Despite the high mass fraction of therapeutics, Insulin@ ADS controlled the release of insulin both in vitro and in vivo. Consequently, Insulin@ADS enabled zero-order release kinetics for at least 8 days after subcutaneous injection. With the help of encapsulation, the mean residence time of insulin was successfully prolonged and insulin bioavailability was effectively improved. A high mass fraction of therapeutics reduced the administered mass of the polymeric matrix, which efficiently limited the immune response induced by the carrier materials. Therefore, Insulin@ADS exhibited better biocompatibility than Insulin@AD. As evidenced in the long-term therapeutic efficacy test, Insulin@ADS stringently controlled the glycemic level within the normal range, enhanced the therapeutic efficacy, decreased the frequency of drug administration, and ultimately improved patient compliance.
In clinical practice, insulin is typically injected subcutaneously as a solution through a syringe, pen, or pump to patients with type 1 diabetes. Long-acting insulin, an analog that increases the half-life of insulin by modifying its molecular structure, has been approved for clinical application. For instance, insulin degludec can maintain blood glucose levels within the normal range for up to 42 h. [36] However, this longacting modification strategy shows structural specificity and is difficult to apply to other proteins. Although no insulin delivery system has been approved for clinical application, various poly(lactic-co-glycolic acid)-based carriers have been reported to deliver insulin in the management of type 1 diabetes. [37] However, the loading degree of insulin in these delivery system does not exceed 5%. [37] Moreover, these systems introduce a large bolus of initial insulin, or burst release, which may lead to hypoglycemia or even death. [37] Overall, low loading degree and burst release may hinder the clinical translation of these insulin delivery systems. In the present study, we successfully prepared microparticles with programmable element distribution for versatile protein delivery. The prepared insulin-loaded www.advmat.de www.advancedsciencenews.com Adv. Mater. 2023, 35, 2211254 microparticles simultaneously featured ultrahigh drug loading degree of 47.4% and zero-order release kinetics of payloads both in vitro and in vivo. Recently, Ali et al. developed a straightforward strategy to fabricate glucose-responsive protein-polymer hybrid hydrogels by cross-linking poly(vinyl alcohol) with various proteins using formylphenylboronic-acid-based crosslinkers. [38] The glucose-responsive insulin release from the hydrogels was precisely engineered based on the molecular structure of formylphenylboronic acid ligands. The proteinpolymer hybrid hydrogels may be a potential smart insulin delivery system for in vivo applications.

Conclusion
We have successfully prepared microparticles with programmable element distribution for versatile delivery of protein therapeutics. The enhanced distribution of ADS molecules at two interfaces, the cargo-NP surface and oil-water interface, was the key point in coordinating ultrahigh drug loading and zeroorder release for the microparticles. Meanwhile, the enhanced adsorption of ADS onto the surface of cargo NPs addressed the poor cargo-polymer miscibility, blocking the phase transfer of protein NPs from oil to water and thus achieving ultrahigh encapsulation efficiency (up to 99.9%). Further, by formulating proteins into NPs, decoration with carrier materials was only required at the surface of the drug NPs. Therefore, the obtained microparticles were loaded with an ultrahigh mass fraction of therapeutics. Moreover, the adsorption of carrier materials at the oil-water interface was enhanced by introducing salts into the water phase. The formation of a compact polymer layer at the surface of the microparticles enabled zero-order release kinetics for protein payloads. Consequently, the obtained microparticles simultaneously featured an ultrahigh drug loading degree (up to 49.9%) and zero-order release kinetics of payloads both in vitro and in vivo. The versatility of microparticles with programmable element distribution was verified by successful engineering microparticles loaded with BSA and β-LG NPs. Even with an ultrahigh insulin loading degree of 47.4 ± 1.1%, the microparticles achieved persistent and efficient blood glucose control for at least 8 days in diabetic rats. Benefiting from the ultrahigh mass fraction of insulin, the administered mass of ADS with Insulin@ADS was 44.2 times smaller than the mass of AD with Insulin@AD. This reduction in the administered mass of ADS greatly constrained the immune response induced by carrier materials. Weekly administration of Insulin@ADS maintained steady plasma drug concentrations throughout the 35 days treatment period. Overall, by integrating efficient protein encapsulation and zero-order release of payloads into a single microparticle, microparticles with programmable element distribution showed great potential for revolutionizing the engineering of long-acting injectable microparticles in the pharmaceutical industry.

Experimental Section
Preparation and Characterization of Protein NPs: The protein NPs were prepared in the first coflow device of a continuous flow device. The nonsolvents for protein molecules containing ADS (2.2 mg mL −1 ) and protein water solution (20.0 mg mL −1 ) were pumped into the inner capillary and the space between inner and outer capillaries, respectively. The nonsolvent and protein solutions flowed in the same direction, and their flow ratio was fixed at 9:1. The nonsolvents for insulin (human recombinant, ≥27.5 units mg −1 ; Merck Limited, China), BSA (96%; Aladdin Industrial Inc., China), and β-LG (≥90%; Merck Limited, China) were acetone, acetonitrile, and tetrahydrofuran, respectively. Particle size was analyzed at an angle of 173° using dynamic light scattering (Zetasizer Ultra, Malvern Instruments Ltd., UK) with a synthetic quartz glass cell (Hellma Analytics, Germany) at 25 °C. The concentration of the obtained NPs was analyzed using a multiangle (173°, 13°, and 90°) light scattering detector. Zeta potential of the protein NPs in the oil phase was measured using Zetasizer Ultra equipped with a universal dip cell kit (ZEN 1002; Malvern Instruments, UK) at 10 mV.
Phase-Transfer Kinetics of Insulin NPs: Acetone solution (0.5 mL) containing insulin NPs (0.1 mg mL −1 ) was stacked on the top of water containing MgCl 2 ·6H 2 O (60%, w/w, 0.5 mL). The acetone and water phases were stirred at 400 rpm. The bottom solution was withdrawn at predetermined intervals. Insulin concentration in the water phase was quantified using high-performance liquid chromatography (see the Supporting Information for details). Thereafter, the mass of insulin transferred into the water phase was calculated. ADS and AD were added to acetone at the concentration of 2 mg mL −1 to verify the effects of polymers on the transfer of insulin NPs from acetone to water.
Surface Adsorption as Monitored using a Quartz Crystal Microbalance: The adsorption of polymers onto the surface of protein NPs was evaluated using the QSense Analyzer (Biolin Scientific AB, Sweden) equipped with the QSense Sensor (QSX 303 SiO 2 ; Biolin Scientific AB, Sweden). The sensor was cleaned through immersion in sodium dodecyl sulfate solution (2%, w/v). Next, the sensor was rinsed with water and dried under nitrogen flux. A quartz disk was spin-coated with the suspension (20 µL) of protein NPs (10 mg mL −1 ) at 3000 rpm for 30 s. The sensors coated with protein NPs were placed in the flow cell and allowed to stabilize in organic solvent flow (0.5 mL min −1 ) at 25 °C to establish the baseline. Subsequently, polymer solution (2 mg mL −1 ) was flown over the surface of the sensor for 20 min at the flow rate of 0.5 mL min −1 . Finally, an organic solvent (0.5 mL min −1 ) was pumped into the flow cell to remove unbonded polymers. The organic solvents used for insulin, BSA, and β-LG were acetone, acetonitrile, and tetrahydrofuran, respectively. Frequency and dissipation were measured simultaneously for all overtones (n = 3, 5, 7, 9, 11, or 13), and the adsorbed mass of the film was calculated using the Sauerbrey equation [23] · where, ∆m is the areal mass difference, ∆f/n is the frequency change, and C is a constant = 17.7 ng cm −2 Hz −1 .

Isothermal Titration Calorimetry:
To evaluate the adsorption of polymers onto protein NPs, thermodynamic properties, including adsorption enthalpy, adsorption entropy, and stoichiometry, were quantified using a Nano-ITC (TA, USA). Acetone, acetonitrile, and tetrahydrofuran served as the organic solvents for insulin, BSA, and β-LG NPs, respectively. Protein NPs suspended in an organic solvent were loaded into the calorimeter cell of the fixed cylindrical shape and 190 µL volume. Then, 50 µL of polymer solution was titrated into the suspension of insulin NPs while stirring at 300 rpm. For reference, the AD or ADS solutions were titrated into the pure organic solvent to determine the dilution heat of the polymers. All experiments were conducted at 25 °C. The integrated heat of the adsorption experiments was subtracted from the integrated reference heat. The obtained normalized heat values were fitted using an independent model to obtain the thermodynamic parameters.
Simulation of the Insulin-Polymer Interactions: MD simulations of the interaction between insulin and polymer were conducted using Amber 18 and AmberTools 18, with the protein ff14SB and general force field. The initial structures of the insulin-polymer complexes were obtained using docking simulations. [39] In this simulation, a simulation cube with a thickness of 20 Å filled with acetone was used. Initially, the energy of the solvent was minimized in 10 000 steps. Subsequently, the energy minimization of the entire system was run for 20 000 steps, and then the system was equilibrated to 25 °C through 10 000 steps. The simulation was performed for 150 ns with a time step of 0.0002 ps and cutoff of 10 Å. The balanced trajectory files of the last 5 ns were extracted to calculate the Gibbs free energy change and the corresponding contributions of electrostatic and van der Waals forces. [40] Fabrication of the Continuous-Flow Device: In the continuous-flow device, two coflow devices were connected with a peristaltic pump. The coflow device was fabricated by assembling two coaxially aligned borosilicate glass capillaries (World Precision Instruments, USA) on a glass slide. One end of the inner capillary (inner diameter [i.d.] ≈ 580 µm; outer diameter [o.d.] ≈ 1000 µm) was tapered using a micropipette puller (P-1000, Sutter Instrument, USA). The diameter of the obtained fine tip was enlarged to 50 and 100 µm for the first and second coflow devices, respectively. The inner capillary was inserted coaxially into a larger cylindrical capillary (i.d. = 1120 µm; o.d. = 1500 µm). Transparent epoxy resin was employed to fix the capillaries on the glass slide and seal them where required. Specifically, at the orifice of the inner capillary of the second coflow device, the outer capillary was constricted using a micropipette puller (P-1000, Sutter Instrument, USA) to form a Venturishaped neck.
Preparation of Protein-Encapsulated Microparticles: As the dispersed phase, the obtained protein NP suspension was pumped into the inner capillary of the second coflow device. Aqueous solution containing MgCl 2 ·6H 2 O (60%, w/w) and poly(vinyl alcohol) (2%, w/v) was flown through the space between the inner and outer capillaries in the same www.advmat.de www.advancedsciencenews.com Adv. Mater. 2023, 35,2211254 direction to serve as the continuous phase. The dispersed phase was emulsified by the continuous phase to form droplets, and the flow velocity was adjusted to achieve the dripping mode. The resulting droplets were collected in an aqueous solution containing MgCl 2 ·6H 2 O (30%, w/w). After diffusion of the oil-phase solvent into water, the twophase system was solidified into microparticles. The microparticles were thoroughly washed with water 6 times to remove salts and poly(vinyl alcohol). Acetone, acetonitrile, and tetrahydrofuran served as the oilphase solvents for insulin, BSA, and β-LG, respectively.
The preparation process for bare ADS microparticles was the same as that for Insulin@ADS, except that the ADS acetone solution (4 mg mL −1 ) served as the oil phase. To verify the effect of salts on the interfacial adsorption of ADS, DCM was used as the oil-phase solvent to prepare insulin-loaded microparticles. Specifically, DCM containing insulin NPs (10 mg mL −1 ) and ADS (10 mg mL −1 ) was pumped into the inner capillary, and an aqueous solution containing poly(vinyl alcohol) (2%, w/v) with or without MgCl 2 ·6H 2 O (60%, w/w) was simultaneously pumped into the space between the inner and outer capillaries. The resulting oil-inwater emulsion was collected in aqueous solution containing poly(vinyl alcohol) (2%, w/v) at 4 °C. The solidified microparticles were washed with water to remove the salt and poly(vinyl alcohol).
Characterization of Microparticles: Morphological characteristics of the fabricated microparticles were observed using the ZEISS Gemini 300 scanning electron microscope (Carl Zeiss AG, Germany) and Zeiss Axio Observer 7 microscope (Carl Zeiss AG, Germany). Their size and polydispersity index were analyzed using ImageJ. Encapsulation efficiency was defined as the weight percentage of encapsulated protein among the total protein added and calculated as [(weight of loaded protein/weight of total protein added) × 100%]. Loading degree was defined as the amount of protein loaded per unit weight of microparticles and calculated as [(weight of loaded protein/weight of protein-loaded microparticles) × 100%]. To calculate the encapsulation efficiency and loading degree, the amount of protein encapsulated in the microparticles and that dissolved in the aqueous phase were measured using high-performance liquid chromatography (for details, see the Supporting Information).
In Vitro Protein Release: To study their controlled drug release capability, the obtained microparticles were suspended in phosphatebuffered saline (pH 7.4; 1 mL) under sink conditions. Samples were incubated at 37 °C while shaking (100 rpm). At predetermined time intervals, the samples were centrifuged (3000 rpm for 5 min) to remove the supernatant, and 1 mL of fresh release medium was added to the tube. Insulin released into the supernatant at each time point was quantified using high-performance liquid chromatography (see the Supporting Information for details). The conformational structural stability of the released insulin was evaluated using a CD spectropolarimeter (Jasco J-810) with a cell length of 0.1 cm at 25 °C. A standard insulin solution was prepared in phosphate buffer (pH 7.4) at the concentration of 0.1 mg mL −1 . The CD spectra of the standard and released insulin were scanned from 190 to 260 nm 3 times at the resolution of 1 nm and scanning speed of 50 nm min −1 . All CD data were presented as the mean residue ellipticity.
DCM-Water Interfacial Tension: The interfacial tension between DCM and water without or with MgCl 2 ·6H 2 O (60%, w/w) dissolved in the water phase was determined using pendent drop tensiometry with the DSA30S Drop Shape Analyzer (Kruss, Germany). Here, a drop of ADS DCM solution (10 mg mL −1 ) was suspended from a needle inside the water without or with 60% MgCl 2 ·6H 2 O w/w. The shadow image of the pendent drop was captured with the camera, and its interfacial tension was calculated using drop-shape analysis.
Simulation of ADS Adsorption at the Oil-Water Interface: MD simulations of ADS adsorption at the interface with ff14SB and general force fields were performed using Amber 18 and AmberTools 18 software. The initial simulation cell was a rectangular basic box with dimensions of L x = L y = 50 Å and L z = 220 Å. 6000 water molecules, 600 DCM molecules, 15 ADS molecules, 443 Mg 2+ ions, and 886 Cl − ions were added into the system. Initially, to remove the molecular overlaps in the initial configurations, the energy of the solvent and whole system was minimized using the steepest descent algorithm for 40 000 and 20 000 steps, respectively. The system was then heated to room temperature for 20 ps, and simulation was performed for 200 ns. Radial distribution functions between ADS and water molecules were analyzed using the CPPTRAJ program.
Type 1 Diabetes Model: All animal experiments were conducted according to the protocols approved by the China Pharmaceutical University Institutional Animal Care and Use Committee (202103003). The pharmacokinetic and hypoglycemic effects of Insulin@ADS were evaluated in a streptozotocin-induced type 1 diabetic rat model. After fasting overnight, male Sprague-Dawley rats (380-400 g) were intraperitoneally injected with streptozotocin (55 mg kg −1 ) in citrate buffer (pH 4.5) to induce type 1 diabetes. Blood glucose levels were monitored using the glucose dehydrogenase method with a glucose meter (Contour Plus, Bayer, Germany) equipped with test strips (Bayer, Germany). Rats with blood glucose levels >350 mg dL −1 were used for in vivo experiments.
Pharmacokinetic and Pharmacodynamic Tests: Hyperglycemic rats were subcutaneously injected with Insulin@AD (50 IU kg −1 ) and Insulin@ ADS (50 IU kg −1 ), which were dispersed in a carboxymethylcellulose aqueous solution (2%, w/v; 1 mL). Free insulin solution (10 IU kg −1 ), bare ADS microparticles, and normal saline were used as controls. Blood glucose levels and body weight were monitored throughout the treatment period. Venous blood was collected in heparin tubes (Vacutainer, Becton Dickinson, USA) in the morning (9:00-11:00 a.m.) or within 8 h after administration. After centrifugation (1000 × g, 10 min), insulin concentration in the upper plasma was measured using an insulin enzyme-linked immunosorbent assay kit (MultiSciences, China), according to the manufacturer's instructions. By plotting the plasma insulin level against time, pharmacokinetic parameters, including the maximum plasma concentration, mean residence time, and area under the curve, were calculated using PkSolver. [41] The in vivo absorption percentage was calculated using the deconvolution method. Blood glucose levels were determined using the glucose dehydrogenase method with a glucometer. HbA1c levels were measured using A1CNow SELF-CHECK (Sinocare Inc., China).
Histological Immunofluorescence Analysis: To evaluate the tissue biocompatibility of Insulin@AD and Insulin@ADS, subcutaneous tissues adjacent to the injection site were excised on days 4, 8, and 12 after administration. The obtained tissues were fixed in paraformaldehyde solution (4%, v/v) for 24 h. After embedding in paraffin wax, 3-5 µm thick sections were obtained. The sections were blocked with BSA (3%, v/v) in phosphate-buffered saline for 30 min and then incubated with the anti-CD68 (rabbit polyclonal antibody, immunoglobulin G; 1:4000; Wuhan Servicebio Technology, China) as the primary antibody for 12 h. Next, the sections were washed and incubated with horseradishperoxidase-conjugated goat anti-rabbit immunoglobulin G (1:500; Wuhan Servicebio Technology, China) as the secondary antibody. Cy3-conjugated tyramide served as the fluorescent dye, resulting in antigen-associated fluorescence signals. The nuclei were stained with 4′,6-diamidino-2phenylindole (DAPI; Wuhan Servicebio Technology, China). The samples were imaged using an inverted fluorescence microscope (Axio Observer 7, Carl Zeiss AG, Germany). Cytometry of subcutaneous tissues adjacent to the injection site was performed using the VTEA plugin of Fiji. [34] Statistical Analysis: The results were expressed as mean ± SD of at least three independent experiments. Analysis was performed using Origin 2020 (Origin Lab Corporation, USA). Statistical analyses were performed as indicated in the figure legends. Significance was set at the following probability levels: *p < 0.05, **p < 0.01, and ***p < 0.001.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author. www.advmat.de www.advancedsciencenews.com Adv. Mater. 2023, 35, 2211254