Glucose-Activated Switch Regulating Insulin Analog Secretion Enables Long-term Precise Glucose Control in Mice With Type 1 Diabetes

Type 1 diabetes (T1D) is a chronic disease characterized by the immune system’s destruction of insulin producing pancreatic β-cells, resulting in insulin deficiency (1,2). Currently, ∼537 million people have diabetes globally. According to the International Diabetes Federation, this number will reach 783 million by 2045, an increase of 46% (3). Generally, T1D is managed via multiple daily injections, continuous subcutaneous insulin infusion, or an insulin pump (4–6). However, because of fluctuations, it is still difficult to accurately maintain balanced blood glucose levels, even with multiple daily insulin injections or continuous glucose monitoring (7).

Synthetic biology has provided new insights into the intelligent therapeutic approaches for T1D (8–13). However, most exogenous signal-induced therapeutic systems have focus on transcriptional regulation in target cells, which usually causes a certain time lag between signal reception and insulin expression/secretion. Therefore, these systems have a risk of inducing postprandial hyperglycemia and subsequent hypoglycemia in the treatment of T1D. Compared with transcriptional regulation, direct regulation of insulin secretion can provide faster blood glucose regulation. Inducible secretion has been previously achieved by the accumulation (14) or conditional storage of insulin in the endoplasmic reticulum (ER) (15). These studies prove that gene therapy is an alternative way to supply exogenous insulin for the long-term precise control of T1D. However, these methods still have some limitations in clinical application, such as viral vector application, additional exogenous signal stimulation or bioelectronic device requirements, implantation and maintenance of engineered exogenous cells, and so on. Additionally, the integration of therapy and monitoring should be reasonably designed with minimal potential biosafety risks.

Intramuscular plasmid-encoded functional protein expression has shown potential for the treatment of some diseases, such as cancers and diabetes; however, the gene delivery efficiency and the expression of plasmid-encoded protein were too low to generate efficient therapeutic results. In previous studies, we established high-efficiency intramuscular gene delivery/expression techniques with muscle specificity. On the basis of the aforementioned techniques, prolonged control of T1D was acquired by single intramuscular injection of the plasmid-encoding mouse single-strand insulin analog (SIA). However, the sustained and uncontrolled expression of the SIA can induce hypoglycemia and is unable to treat clinical T1D cases. In this study, to obtain SIAs with high and improved efficiency to synthesize insulin, we introduced a series of flexible peptides to replace proinsulin C-peptide. The biological function of this system was verified by both in vitro and in vivo experiments. On the basis of intramuscular gene delivery and expression techniques, this study aimed to establish a glucose-activated SIA switch (GAIS) supply, achieving safety and efficiency in T1D treatment.

The SIA coding sequences were synthesized by the Tsingke Company (Chengdu, China) and then cloned into the plasmid (pSC) (insertion of the SV40 enhancer in front of the cytomegalovirus promoter of pcDNA3.1[+]) to produce the pSC-SIA plasmids. For protein structure prediction and visual molecular dynamics, SWISS-MODEL software was used.

Each well of a white Greiner Lumitrac 600 96-well plate (Corning, Inc., Corning, NY) was coated with human insulin receptor (IR) (0.1 μg) overnight at 4°C. Then, 50 μL horseradish peroxidase–labeled insulin (2.9 ng/mL) mixed with increasing concentrations of SIAs or unlabeled insulin was added to each well for 1 h at 37°C followed by washing with PBS with Tween detergent; 100 μL tetramethylbenzidine was then added, and the plate was incubated for 10 min at 37°C. Lastly, the reaction was terminated, and absorbance at 450 nm was detected with a microplate analyzer (BioTek Instruments, Inc., Winooski, VT).

NIH3T3-L1 adipoblasts seeded in 96-well plates were induced into adipocytes as described (16). Glucose uptake in response to insulin and the SIA was measured (17). Insulin or the SIA was added at decreasing concentrations (from 10 to 1 nmol/L) and incubated for 30 min at 37°C, followed by rapid cell wash with ice-cold Krebs-Ringer phosphate buffer, and 100 μL (34 μg/mL) glucose analog 2-NBDG was then added. Cell uptake of 2-NBDG was measured using the Screen Quest Fluorometric Glucose Uptake Assay Kit (AAT Bioquest, Pleasanton, CA).

The coding sequences of the signal peptide (ss), the conditional aggregation domain (CAD), and a furin cleavage sequence (FCS) along with the SIA sequence (ss-CAD-FCS-SIA-B2) was synthesized by the Tsingke Company and then cloned into the pSC vector to construct the GAIS system.

Secreted alkaline phosphatase (SEAP) production was quantified with a chemiluminescence-based assay kit (Abcam).

Six-week-old male mice (C57BL/6J) (purchased from the Experimental Animal Center of Sichuan Province; approval no. KS2019070) were fasted for 16 h and then injected with streptozotocin (STZ) (60 mg/kg in 0.1 mol/L citrate buffer; pH 4.5; Sigma) in the peritoneum daily for 5 days. Mice with persistent hyperglycemia (>16.7 mmol/L) for 2 weeks were considered to have diabetes and used for additional experiments.

All mice with T1D were randomly divided into T1D, pSC-SIA, and GAIS groups, and healthy untreated mice were used as control (wild-type [WT] group). Intramuscular gene delivery was carried out using the L/E/G technique (pluronic L64, low-voltage electropulse, and epigallocatechin gallate [EGCG]). Briefly, 100 μg plasmid was mixed and coincubated with 10:1 EGCG for 30 min at 37°C; an equal volume of 0.2% (w/v) pluronic L64 was then added. Approximately 60 μL plasmid DNA/EGCG/L64 mixture was equally injected into both sides of the tibialis anterior (TA) muscle of each mouse. After 60 min, the electropulse was applied using a clinical SDZ-V nerve and muscle stimulator (18).

Data are presented as mean ± SEM and analyzed by unpaired t test and one- and two-way ANOVAs using GraphPad Prism (version 6.0). P values <0.05 were deemed statistically significant.

This study was performed with the approval of the Medical Ethics Committee of Sichuan University (Chengdu, China). All procedures were strictly conducted by the code of ethics. When planning and conducting experiments, we strictly followed the 3R principle (replacement, reduction, refinement). Animal handling and termination standards were in line with the standards of the Society for Experimental Animal Science (Society for Laboratory Animal Science).

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

In comparison with natural insulin, a single-chain structure may facilitate efficient SIA expression in non–β-cells. In this study, we designed four SIAs containing different flexible peptide linkers to replace the long and rigid C-peptide chain in human proinsulin (Supplementary Fig. 1). Compared with that of proinsulin, the predicted structure of the four SIAs showed greater similarity to insulin (Fig. 1A); moreover, the flexible linkers might have offered stretchable space to the SIAs, thereby facilitating the formation of disulfide bonds. Competitive binding assays were performed to assess the binding affinities of the SIAs to the human IR. Notably, SIA-A1, -A2, and -B2 were as effective as natural insulin in receptor binding, suggesting the SIAs had biological functions similar to those of insulin (Fig. 1B). SIA-A2 was equipotent to native insulin, and SIA-A1, -B1, and -B2 were five- to 10-fold more potent in promoting glucose uptake at 10 nmol/L, despite their equal affinity to IR (Fig. 1C). Additionally, these four SIAs and insulin showed no difference in cell proliferation assays, suggesting that malignant proliferation by SIAs could be ignored (Fig. 1D).

To investigate the therapeutic effects of SIAs on STZ-induced T1D, we divided mice into six groups based on the plasmid they received (100 μg per mouse). Fasting blood glucose levels of mice receiving pSC-SIA-A1, -A2, or -B2 were significantly decreased during 9–33 days, whereas blood SIA expression persisted for 30 days (Fig. 1E and F). However, pSC-SIA-A1– and -B2–treated mice did show hypoglycemia and even death (data not shown), suggesting that a sustained supply of an SIA was not safe enough. On the 14th day after gene therapy, mice were fasted for 16 h, after which an intraperitoneal glucose tolerance test (IPGTT) was performed. Because mice with diabetes did not clear blood glucose, pSC-SIA-A1–,-A2–, and -B2–injected mice with diabetes displayed improved glucose tolerance, especially those treated with pSC-SIA-A1 or -B2 (Fig. 1G and H). From the 15th day, body weights of pSC-SIA-A1– and -A2– and especially -B2–treated mice were heavier than those of pSC-treated mice and mice with diabetes (Fig. 1I). Collectively, these results demonstrated the therapeutic benefits of pSC-SIA, especially pSC-SIA-B2, in reducing blood glucose levels and improving blood glucose metabolism. SIA-B2 was used in the subsequent studies and abbreviated as SIA. However, hypoglycemia caused by persistent strong expression of SIA needs to be further solved.

The plasmid pSC-CAD-FCS-SIA was designed to express fusion preprotein ss-CAD-FCS-SIA (19). The CAD can interact with GRP78 (20,21), an ER protein with ATPase activity (3,22), in a ligand-reversible manner. The fusion protein CAD-FCS-SIAs were transiently retained in the ER via CAD-GRP78 interaction. The affinity between GRP78 and the CAD may be changed under different ATP-to-ADP ratios, resulting in the retention of CAD-FCS-SIAs within the ER during hypoglycemia (low ATP-to-ADP ratio) and dissociation of CAD-FCS-SIAs from GRP78 under hyperglycemic conditions (increased ATP-to-ADP ratio). Subsequently, CAD-FCS-SIAs were transferred to the Golgi apparatus; there, the CAD was removed by the furin protease at the FCS site, generating the SIA with biological activity. Therefore, in the GAIS system, the ER served as the storage depot for CAD-FCS-SIAs, which were then cleaved by the furin protease in the Golgi apparatus and released and secreted into the blood (Fig. 2A).

In the colocalization assays of the GAIS system, fluorescent signals of the SIA (green) and GRP78 (red) colocalized in the ER (purple; labeled by protein disulfide isomerase). Meanwhile, the SIA signals (in the CAD-FCS-SIA fusion protein) were extremely strong in the ER during the hypoglycemic state and were significantly reduced during the hyperglycemic state (Fig. 2B). The results showed that the CAD-FCS-SIA fusion proteins expressed by the GAIS system were temporarily kept in the ER in low glucose conditions but were released and secreted from the ER in high glucose conditions. In comparison, the signal intensity of the pSC-SIA–encoded SIA did not show any change during increased glucose concentrations. In addition to the colocalization analysis, the immunoprecipitation (IP) assay further proved the binding of CAD-FCS-SIA to GRP78 (Fig. 2C). In the IP products precipitated by the anti-GRP78 antibody from the lysates of pLVX-GAIS–infected C2C12 cells, GRP78-CAD-FCS-SIA was detected. This band was also detected in the cell lysates. The results confirmed the binding of CAD-FCS-SIA to GRP78. The hypothesis that the ER acted as a storage depot for proteins was confirmed after C2C12 cells expressing the fusion protein were treated with 2.5 mmol/L glucose for 24 h, and the fusion protein was precipitated by co-IP using antibodies against GRP78. Collectively, these results confirmed that the ER can serve as a temporary repository for the CAD-FCS-SIA fusion protein because of its binding to GRP78 and that it is released from GRP78 upon high glucose stimulation.

In GAIS-transfected C2C12 cells, glucose treatment resulted in a dose-dependent response across the concentration range from 0 to 15 mmol/L (Fig. 2D). The SEAP reporter gene level reached its peak within 150 min after glucose stimulation (Fig. 2E). Because the transcription-based SEAP expression peak usually appears at 7 h (13), the secretory regulation-based GAIS system had a faster response to blood glucose. In addition, the SEAP expression induced by hyperglycemia in vivo also showed a significant increase at 2.5 h in mice with T1D (Fig. 2F).

Neither pSC-CAD-FCS-SIA– nor pSC-SIA–transfected C2C12 cells showed a significant difference in SIA mRNA levels under different glucose concentrations (Fig. 3A) or different exposure times to 15 mmol/L glucose (Fig. 3B). However, the SIA concentration in the cell culture supernatant greatly increased in response to glucose rise and reached its peak at 15 mmol/L (Fig. 3C). The SIA concentration in the supernatant also increased over time and reached its peak from 120 to 150 min (Fig. 3D). These results demonstrated that the GAIS system could activate secretion of the SIA but not gene transcription.

SIA levels can physiologically fluctuate in response to blood glucose levels; that is, the GAIS system should be opened in high glucose concentrations and closed in low glucose concentrations. As expected, in the on-off-on test, the medium SIA concentration showed a high-low-high fluctuation. In the off-on-off test, there was a low-high-low outcome. At the low level, the SIA was almost undetectable (Fig. 3E), demonstrating that the GAIS system could be reversibly and repeatedly activated and suspended during glucose concentration fluctuation.

To investigate the application potential of this rapid-response secretion system in diabetes, the therapeutic effects of the GAIS system in mice with STZ-induced T1D were investigated (Fig. 4A and B). In comparison, the GAIS system significantly reduced fasting blood glucose levels from the 6th day and kept them in the normal range from the 12th to the 60th day. pSC-SIA also significantly lowered fasting blood glucose levels from the 9th day; however, it caused fluctuations in blood glucose levels within the following days (Fig. 4C). We also detected the random blood glucose of mice. Due to variations in feeding patterns among individual mice, nonfasting glucose levels demonstrated significant fluctuations. However, mice treated with GAIS and pSC-SIA exhibited significantly lower nonfasting glucose levels compared with those in the T1D group (data not shown). In this study, both the GAIS and pSC-SIA treatments successfully restored HbA_1c levels to normal, suggesting the successful control of serum glucose in the 2-month treatment period with two injections administered on days 0 and 30 (Fig. 4D). GTTs performed on days 18, 30, 57 (Fig. 4E–J), 39, and 48 (Supplementary Fig. 4) further indicated the partially recovered ability to metabolize postprandial blood glucose after gene therapy. Although both the GAIS and pSC-SIA treatments reduced blood glucose levels in the IPGTT assays, the GAIS system more effectively dealt with postprandial blood glucose on days 18, 39, and 48 compared with WT mice and mice with T1D (Fig. 4F and Supplementary Fig. 4A and C). On the day-18 IPGTT assay, the blood glucose values in the GAIS-treated mice at 90- and 120-min test points were similar to those in WT mice (P > 0.05) (Fig. 4E). Even on the 57th day, GAIS treatment in mice with T1D was able to stimulate the serum SIA by glucose administration (Fig. 4K). The IPGTTs on days 18, 39, and 48 were more effective than those on days 30 and 57 (Fig. 4E–J and Supplementary Fig. 4), which may have been due to the gradual degradation of the plasmid, resulting in a diminished therapeutic effect, and therefore, in order to maintain the long-term therapeutic effect, plasmids should be administered at appropriate intervals. Polydipsia, polyuria, polyphagia, and weight loss are the typical manifestations of T1D (17). Mice with STZ-induced diabetes continued to lose weight, whereas GAIS and pSC-SIA administrations prevented the continuous loss of body weight from the 12th day (Supplementary Fig. 3A). Polydipsia and polyphagia were ameliorated in the treatment groups compared with in the diabetes group, but these results were still inferior to those in the WT group (Supplementary Fig. 3B and C). Collectively, in vivo studies demonstrated that the GAIS system could imitate the natural and physiological response to glucose stimuli and secrete the SIA in time.

The main causes for the progression of diabetic complications are chronic inflammatory processes, oxygen transport system damage and hypoxia, and redox imbalance (23,24). To study the effects of the treatments on blood glucose control and antioxidation in mice, malondialdehyde, superoxide dismutase, reduced glutathione, and total antioxidant capacity in serum, kidney, and heart were monitored (Fig. 5). We observed by malondialdehyde levels that after 60 days of treatment, oxidative stress in serum, kidney, and heart was significantly reduced. Meanwhile, superoxide dismutase and glutathione as well as total antioxidant capacity levels were significantly increased in serum (Fig. 5B), kidney (Fig. 5C), and heart (Fig. 5D) compared with those in mice with T1D. The aforementioned results indicated that the GAIS system could ameliorate the damage caused by oxidative stress in diabetic blood, kidneys, and heart.

On the day 18, the pancreatic insulin signal (Fig. 6A) and C-peptide (Fig. 6B) were detected only in WT mice, not in STZ-treated groups. The results confirmed that in STZ-treated mice, β-cells were destroyed, losing their insulin-producing function. SIA or insulin signals were shown in mouse TA muscles of the GAIS and pSC-SIA groups but not in other organs analyzed (Supplementary Fig. 7) or in the WT or T1D group (Fig. 6A). Furthermore, the serum SIA levels gradually increased in the GAIS- and pSC-SIA–treated mice, reaching peaks on the day 36 and still detectable on the day 48; however, insulin signals and the SIA were undetectable in mice with T1D within 60 days (Fig. 6C). These results demonstrated that the remission of diabetes in mice with T1D should be due to the SIA encoded by the intramuscularly delivered GAIS or pSC-SIA rather than the endogenous insulin from residual pancreatic β-cells.

There was no significant change in total serum immunoglobulin G or M level between WT and treated diabetic groups (Fig. 7A and B), nor was there a difference seen in inflammation-related factors, such as interleukin-6 and interferon-γ, between WT and treated diabetic groups (Fig. 7C and D). These results illustrate that there was no immunological response or inflammatory reaction to the expressed SIA. In addition, the long-term effective control of blood glucose also implies the lack of immunogenicity of the SIA.

The accumulation of unfolded or misfolded proteins in the ER may cause ER stress, leading to damage, autophagy, and even cell death (25); therefore, it is necessary to detect whether temporary storage of SIAs in the ER causes ER stress. ER stress markers ATF4, CHOP, GRP78, XBP1s, and XBP1u were evaluated, and no significant difference was revealed between WT and GAIS-treated mice, indicating no obvious ER stress in the GAIS-transfected skeletal muscle cells after repeated SIA storage and release (Fig. 7E).

Histological analysis of the plasmid-injected TA muscles showed no pathological changes (Fig. 7F), suggesting that neither exogenous SIAs nor muscle cell–expressed SIAs induce a local inflammatory reaction or pathological damage in situ. Moreover, no significant pathological changes were observed in the heart, liver, spleen, lungs, or kidneys of the four mouse groups (Supplementary Fig. 6), which further confirms the biosafety of exogenous SIAs for diabetes treatment.

The biochemical markers of serum albumin, alkaline phosphatase, ALT, AST, blood urea nitrogen, and creatinine were tested to evaluate liver and kidney function. The results revealed normal liver and kidney function in GAIS- and pSC-SIA–treated as well as WT mice (Supplementary Fig. 5A). In addition, there was no significant difference in routine blood tests (Supplementary Fig. 5B). Therefore, the SIA had no hepatorenal toxicity or hematotoxicity. HOMA for insulin resistance was analyzed on day 24 after GAIS treatment. The results indicated that expression of the GAIS system in muscle did not produce significant peripheral insulin resistance (Supplementary Fig. 8).

Regulating therapeutic protein expression can provide a new perspective on control of hormone release, tumor immunotherapy, and homeostasis regulation (26,27). Currently, treatments are focused on regulating transcriptional levels, a relatively slow method for producing therapeutic proteins. To bypass gene transcription, we designed a system to regulate the secretion of proteins. The GAIS system has a faster response than transcriptional switches and has the characteristics of high sensitivity, dose dependence, and reversibility. In the mice with T1D, the treated group showed faster blood glucose clearance. Moreover, it was demonstrated that the GAIS system realized the integration of therapy and monitoring, overcoming poor compliance with insulin injection and thereby achieving long-term blood glucose control. This study also revealed that the GAIS system can achieve rapid coupling of glucose sensing with the release of therapeutic proteins.

T1D is associated with the loss of β-cells expressing insulin; however, as an alternative, other cells can be used as a new source of insulin (15,28–31). The liver is a target organ of insulin and has glucose sensors (32–34), but it lacks prohormone invertase and regulatory secretion pathways. Neuroendocrine cells have a regulatory secretory pathway but lack a glucose-sensing mechanism (35). Intestinal epithelial cells express GK, GLUT2, and prohormone invertase, but considering the rapid renewal and complex operation, the potential of K cells as targeted cells is still limited (36–39). Skeletal muscles are widely distributed, have slow turnover and extensive vascularization, can secrete therapeutic proteins into the systemic circulation to treat multiple human diseases (40–43), and are easy to operate with lower risk. Moreover, as nonislet β-cells, they are not attacked by the immune system (44,45). The precursor of insulin consists of A, B, and C chains, and the C-peptide is cut off by the prohormone invertase in the Golgi to form mature insulin (46). However, prohormone invertase is mainly expressed in β-cells. In this study, the first step was to design and screen appropriate linkers to substitute for C-peptide. Meanwhile, insulin gene therapy requires an effective gene delivery method to the skeletal muscle cells. Both viral and nonviral vectors have been tested for the treatment of diabetes, and viral vectors undoubtedly have superior delivery efficiency (47). Adeno-associated viral vectors have been considered the most promising vectors for gene therapy; however, it has been reported that some of the therapeutic gene fragments they carry can integrate into a dog’s chromosomes, potentially causing cancer (48). Plasmids can overcome this shortcoming but have lower transfection efficiency. To develop an efficient and safe skeletal muscle delivery system, we used the L/E/G technique to transfer the GAIS system to skeletal muscle (18).

Inducible trafficking was previously achieved by oral administration, biotin release ER-resident hook (49), and light-triggered release (50), but these require additional external signal stimulations. Unfortunately, it is difficult to adjust signal strength to control hyperglycemia or hypoglycemia. The release of SIAs in this study was slightly delayed compared with that observed in the physiological state, and the glucose tolerance was lower than that of normal mice. The reason may be that SIAs need to be transported from the ER to the Golgi, where they are treated by furin, and then further transported to the plasma membrane. To further shorten the time from stimulation to release, an SIA can be selectively stored near the cell membrane. In summary, in this investigation, we designed a GAIS system to enable spontaneous blood glucose regulation of SIA release expressed in skeletal muscle cells. Using a reversible combination of CAD motif and GRP78 can effectively and rapidly control blood glucose in mice with diabetes for a long time. Subsequent studies may combine the secretory regulatory GAIS system with current widely used transcriptional switches to form a compatible system and then integrate this system into genomic safe sites using CRISPR-Cas9 to achieve permanent treatment of diabetes. The integration of the GAIS system and compatible transcription system can also achieve a variety of signals and levels to regulate the expression of therapeutic proteins and design complex biological gate circuits. This system can be beneficial for controlling molecular mechanisms, drug expression, and synthetic biology in the future. In addition to being used for diabetes, the GAIS system–mediated smart production factory platform can be used to express and secrete a variety of therapeutic proteins/peptides for the treatment of different diseases, such as muscular dystrophy, severe osteoporosis, and immunotherapy of tumors, thus achieving long-term, accurate, and controlled protein/peptide drug delivery.

Funding. This study was supported by the National Natural Science Foundation of China (31971390), the International Cooperative Project of Sichuan Province on Science and Technology Innovation (2021YFH0142), and the Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine (ZYYCXTD-C-202209).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. L.X. and W.L. performed the experiments, with assistance from Y.Z., L.D., M.L., and H.G. L.X., C.X., and G.W. analyzed the data. L.X. and G.W. designed the project and wrote the paper. G.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.