Multicomponent Plasmid Protects Mice From Spontaneous Autoimmune Diabetes

Although the etiology of type 1 diabetes remains incompletely understood (1), islet antigen–specific T cells, especially CD8⁺ T cells (2), are hypothesized to drive the loss of β-cells that characterizes type 1 diabetes. Antigen-presenting cells pick up β-cell–derived antigens in the pancreas and present to autoreactive T cells in the draining lymph nodes (3). This promotes T cell infiltration of the pancreas and T cell–mediated β-cell death. Most T cell–targeted therapies tested in clinical trials, including anti-CD3 monoclonal antibodies that delete or anergize all T cells (4–6), may at least temporarily cause broad immunosuppression. When given at diagnosis, all treatments evaluated to date have yielded only transient benefits, and most are associated with side effects such as reactivation of latent viruses (7,8), making them unsuitable for chronic use. In contrast, immunotherapy given as early as possible may prevent clinical type 1 diabetes onset in individuals at risk. The promise of this approach was recently corroborated in a clinical proof of concept phase 2 trial with the anti-CD3 antibody teplizumab, which delayed clinical onset by >2 years (9). Arguably, however, the ideal therapy for prevention of type 1 diabetes should target only autoreactive diabetogenic T cells to avoid systemic immune modulation resulting in immune exhaustion or deficiency.

Antigen-specific immunotherapy (ASIT) has long been proposed as key to type 1 diabetes prevention, primarily based on successful studies in animals. ASIT typically works by immunization with islet-associated proteins or peptides to suppress autoreactive T cells through deletion or anergy. Unfortunately, to date, clinical trials have convincingly demonstrated safety but not efficacy (10,11). In fact, ASIT has failed to alter the disease course after hyperglycemia onset in preclinical and clinical studies (12). Likewise, oral administration of peptides or proteins induces minimal, if any, efficacy before or after disease onset (13,14). Basic tolerogenic plasmids encoding only self-antigens are a promising modality for ASIT and have been shown to be safe in recently diagnosed type 1 diabetes, whereas efficacy remains to be demonstrated (15,16). Efficacy may be enhanced by coexpression of or treatment with immune modifiers to generate a micromilieu for autoantigen presentation that favors tolerization (17). For example, controlled secretion of type 1 diabetes autoantigens, GAD65 or proinsulin, in combination with interleukin (IL)-10 by genetically modified lactococcal bacteria has been shown to preserve functional β-cell mass in recent-onset NOD mice but only with coadministration with anti-CD3 therapy (18,19).

We hypothesized that plasmid-mediated coexpression of antigen and cytokines would create a microenvironment where antigen presentation to autoreactive T cells leads to limited activation and, consequently, limited in vivo disease progression. We demonstrate that continuous administration of this plasmid effectively prevents type 1 diabetes development in NOD mice without broad immune suppression. Collectively, these data provide a strong foundation for translation of this novel approach to clinical testing in individuals at risk for developing type 1 diabetes.

Female NOD/ShiLtJ mice 5–8 weeks of age were from The Jackson Laboratory. Blood glucose was monitored weekly via tail vein bleeding with a CONTOUR NEXT glucometer (Bayer). Mice were considered diabetic when blood glucose was ≥250 mg/dL on two consecutive days and were euthanized when blood glucose was ≥600 mg/dL on two consecutive days. For the cyclophosphamide model, 24 seven-week-old NOD mice were prepared for each group; 1 animal was removed from the plasmid group prior to any administrations due to multiple seizures. Cyclophosphamide (Cytoxan) induction was performed with a single injection of 200 mg/kg i.p. in 200 µL PBS. All mice were handled according to Novo Nordisk guidelines, all local Institutional Animal Care and Use Committee protocols, and Ethical Review Council standards at Novo Nordisk Research Center Seattle, Inc., and La Jolla Institute for Immunology.

The pVAX1 vector backbone was used to generate plasmid constructs (Thermo Fisher Scientific). pVAX1 contains a human cytomegalovirus (CMV) immediate-early (IE) enhancer promoter, kanamycin antibiotic resistance gene (NptII), and pBR322 origin of replication. The following gene products were synthesized using Eurofins Genomics or Integrated DNA Technologies: proinsulin2 (PI), preproinsulin2 (PPI), viral 2A peptide sequences, modified CD74 invariant chain, viral internal ribosome entry site (IRES), and mouse transforming growth factor (TGF)-β1, IL-10, and IL-2. PstI and XhoI restriction sites were used for insertion of the multicistronic gene product into pVAX1 using restriction enzymes. Q5 High-Fidelity DNA Polymerase (New England Biolabs [NEB]) was used for cloning gene products and Gibson Assembly Master Mix (NEB) was used to ligate the gene products into the final plasmids according to manufacturer-suggested protocols. Plasmids were sequence verified using BigDye Terminator v3.1 Cycling Sequencing Kit (Thermo Fisher Scientific). EndoFree Giga prep kits were used to purify plasmids following the manufacturer’s protocol with minor modifications for maximum yield (QIAGEN): 1 mL bacteria was incubated for 8 h in LB Broth (Teknova) to seed 500 mL Super Broth (Teknova) in 1-L baffled flasks.

The pVAX1 vector backbone (Thermo Fisher Scientific) was used as a template for the antibiotic resistance–free vector backbone. The plasmid was constructed in three segments by cloning sequence regions using Q5 High-Fidelity DNA Polymerase. The kanamycin antibiotic resistance open reading frame was removed and substituted with a modified Escherichia coli infA gene. Additionally, the pBR322 origin of replication was converted to pUC to increase copy expression. The three segments were ligated with Gibson Assembly Master Mix following the manufacturer’s protocol. The resulting vector was sequence verified, tested for temperature selection, and designated pNN for differentiation from pVAX1. EndoFree Giga prep kits were used to purify plasmids as described above.

High-efficiency 5-α DH5α E. coli cells were used for plasmid growth (NEB). For modification of the cells for antibiotic-free selection, the upstream noncoding region of translation initiation factor 1 (IF-1/infA) was replaced through integration of a temperature-responsive transcriptional regulator segment of prfA from Listeria monocytogenes into the E. coli genome. After the sequence and growth selection were confirmed, infA::prfA E. coli cells were cultured and generated to be electrocompetent. Competent cells were stored at −80°C and used for pNN transformation.

We conducted endotoxin testing by diluting plasmid in Endosafe LAL reagent water and loading onto an LAL test cartridge to be read on the Charles River Laboratories Endosafe nexgen-MCS system. The quality control thresholds for in vivo studies were set at 0.0025 EU/μg for 40 µg dose, 0.005 EU/μg for 20 µg dose, and 0.001 EU/μg for 100 µg dose. Restriction enzyme digestion was performed using XhoI and PstI-HF enzymes and CutSmart Buffer (NEB) according to the manufacturer’s protocol. The expected fragment size was confirmed by running electrophoresis on a 1% agarose gel. Plasmid sequencing was performed with use of a set of 18 primers by Sanger sequencing with services from GENEWIZ. The .ABI files obtained from GENEWIZ were aligned to the plasmid reference sequence with use of Geneious software to confirm the identity of the plasmid.

All equimolar doses were diluted using sterile PBS to a final volume of 50 or 100 µL for injection. Six to 11-week-old mice were treated with 40 µg empty pVAX1 plasmid, 50 µg pVAX1-PI, 67.5 µg pVAX1-PI/IL-10, 100 µg pVAX1-PPI/TGF-β/IL-10/IL-2, 100 µg pNN-PPI/TGF-β/IL-10/IL-2, or PBS as a negative control.

Blood collection was performed via the submandibular vein for interim time points and via cardiac puncture for terminal sample collection. For serum, blood was allowed to clot, centrifuged (4°C, 700g, 5 min), and carefully collected in prelabeled tubes for storage at −80°C until antibody titers were measured. For plasma, blood samples were collected in K₂EDTA BD Microtainer Tubes, inverted 10 times, and kept on a rocking platform until processing. Plasma were separated (4°C, 2,000g, 10 min), carefully collected in prelabeled tubes, and stored at −80°C until multicytokine measurements. Leukocytes were resuspended for subsequent red blood cell lysis and flow cytometry staining immediately.

Cytokine levels were measured with the mouse V-PLEX Proinflammatory Panel 1 assay (Meso Scale Discovery) following the manufacturer’s instructions and evaluated with the MESO QuickPlex SQ 120reader (Meso Scale Discovery).

Immunophenotype via flow cytometry was used to assess frequencies of innate immune cell populations and their expression of IL-10 receptor, also known as CD210, in peripheral blood. Blood was collected from a subset of mice from each group by submandibular venipuncture in K₂EDTA BD Microtainer Tubes, without anesthesia, at 13 weeks of age. A maximum volume of 150 µL was collected. Leukocytes were resuspended for subsequent red blood cell lysis and flow cytometry staining (Supplementary Table 1). Cells were stained first with LIVE/DEAD Aqua dye (Molecular Probes) for 30 min, incubated with Fc block for at least 5 min, and then stained with surface markers including anti-mouse CD45, CD3, CD19, CD11b, CD11c, CD49b, NKp46, Ly6C, Ly6G, F4/80, and CD210 biotin for 30 min. Lastly, cells were stained with streptavidin-APC-Cy7 for CD210 detection for 30 min. PBS plus 2% FBS was used as wash buffer between the staining steps, and FMOs (fluorescence minus one control) were made for each individual marker to ensure gating strategy. Data were collected on a BD LSRFortessa Flow Cytometer and analyzed using FlowJo software (version 10; BD).

Plasma samples were shipped to Barbara Davis Center for Childhood Diabetes at University of Colorado Health Sciences Center, and insulin autoantibody (IAA) titers were evaluated by radioassay as previously described (20).

Mice were pretreated with 100 µg pVAX1-PPI/TGF-β/IL-10/IL-2 once weekly starting at 6 weeks of age for 5 weeks prior to treatment with 50 µg each OVA and aluminum hydroxide (alum) (Sigma-Aldrich) in 100 µL total volume intraperitoneally at 12 weeks of age. Mice continued weekly plasmid treatment for 4 weeks and were bled on day 30 after OVA/alum challenge. Serum was collected for anti-OVA IgG measurement using a mouse anti-OVA IgG antibody assay kit (Chondrex). All samples were stored at −80°C prior to analysis.

With use of GraphPad Prism, version 7 (GraphPad Software, La Jolla, CA), group comparisons were performed with the Log-rank (Mantel-Cox) test for type 1 diabetes incidence curves, Kruskal-Wallis tests with Dunn posttest comparisons for cytokine and flow cytometry data, and Mann-Whitney U test for OVA challenge studies.

All data generated or analyzed during this study are included in the published article (and Supplementary Material). Raw underlying data can be requested from the corresponding author upon reasonable request and for noncommercial purposes.

The full plasmid sequences used in the current study are available from the corresponding author upon reasonable request and for noncommercial purposes.

We initially generated three plasmids to study the impact of antigen and response modifier expression on diabetes prevention: the plasmid DNA backbone without antigen or cytokine (pVAX1-empty), a plasmid encoding mouse proinsulin2 (PI) antigen only (pVAX1-PI), and one plasmid co-encoding mouse PI and the immune-suppressing cytokine IL-10, using an IRES to generate a single bicistronic mRNA (pVAX1-PI/IL-10) (21) (Fig. 1A). IL-10 was chosen for the initial design, as it is a well-characterized anti-inflammatory cytokine involved in counteracting type 1 diabetes pathogenesis (22,23).

In vitro transfection analyses of the plasmids confirmed that PI and IL-10 proteins were produced as expected (data not shown). The three plasmids were administered via intramuscular (i.m.) injection into NOD mice beginning at 9 weeks of age. Mice were treated once a week for 8 weeks (Fig. 1B) and monitored for disease development for 21 weeks. The pVAX1-empty group was not protected from type 1 diabetes development, and only modest protection was observed for the pVAX1-PI group, but the diabetes incidence was not statistically significant different from that of the pVAX1-empty group. However, marked protection from type 1 diabetes was observed for the pVAX1-PI/IL-10 group; the incidence was statistically significantly lower than for the pVAX1-empty group. This suggested that coexpression of a response modifier such as IL-10 could increase antigen-specific plasmid efficacy in vivo.

Additionally, an analogous treatment regimen was tested in the accelerated cyclophosphamide-induced/synchronized NOD diabetes model. In this model, administration of the alkylating agent cyclophosphamide, a chemotherapeutic for T-cell leukemia, was injected intraperitoneally to promote synchronized hyperglycemia onset 2.5 weeks after administration. The plasmid dosing scheme was the same as described for the spontaneous NOD model, except that two doses were administered prior to, and three following, cyclophosphamide injection. The results confirmed the findings in the less aggressive spontaneous-onset NOD model, with significant protection from hyperglycemia in the pVAX1-PI/IL-10 plasmid treatment group only (Fig. 1C).

For further enhancement of the plasmid-mediated immune modulation and efficacy, PI was replaced with preproinsulin2 (PPI) with a Sec61-impairing mutation to prevent cleavage and secretion and for enhancement of potential antigen diversity (24). We also hypothesized that concomitant addition of TGF-β1 and IL-2 to the plasmid as response modifiers could promote the induction and stabilization of T-cell tolerance and enhance the prevention of type 1 diabetes in vivo. For coexpression of TGF-β1, IL-10, and IL-2, along with PPI, in a single mRNA derived from the vector, a viral IRES and 2A peptides were added to the sequence to promote protein separation. We further added the invariant chain (CD74) leader peptide to direct the PPI to the endosome for MHC class I and II processing to improve tolerogenic antigen presentation to both CD4⁺ and CD8⁺ T cells (Fig. 2A and B) (25).

In vitro transfection analyses of the plasmids confirmed that PPI, TGF-β1, IL-10, and IL-2 proteins were produced as expected (data not shown). For determination of whether the modifications further improved type 1 diabetes prevention, NOD mice were treated with pVAX1-PI/IL-10, pVAX1-PPI/TGF-β/IL-10, or pVAX1-PPI/TGF-β/IL-10/IL-2 weekly (i.m.) for 21 weeks starting at 9 weeks of age. A near-complete prevention of type 1 diabetes was observed with the modified plasmids (pVAX1-PPI/TGF-β/IL-10/IL-2 and pVAX1-PPI/TGF-β/IL-10), and this was statistically significantly different in comparison with the untreated group (Fig. 2C). When treatment was administered to mice at 6 weeks of age, pVAX1-PPI/TGF-β/IL-10/IL-2 significantly outperformed pVAX1-PPI/TGF-β/IL-10 (Fig. 2D). We concluded that the most robust effect was obtained when IL-2 was included in the multicytokine plasmid.

Further, 7 weeks post–first injection, higher IL-10 levels were detectable with pVAX1-PPI/TGF-β/IL-10/IL-2 than with pVAX1-PI/IL-10 and pVAX1-PPI/TGF-β/IL-10 (Fig. 3). We confirmed that the IL-10 measured was of endogenous origin using mass spectrometry–based antibody characterization to discriminate between endogenous and the plasmid-derived IL-10 that carries a short 2A-derived tag (data not shown). Peripheral IL-2 levels remained very low and similar across groups. Circulating TNF-α and IFN-γ were elevated in mice treated with pVAX1-PPI/TGF-β/IL-10/IL-2 in comparison with other treatment groups (Fig. 3). While these cytokines are conventionally considered proinflammatory (26), their protective role during the late prevention phases of autoimmune diabetes in mice has long been known (27,28).

For determination of whether plasmid therapy induced acute changes in immune cell phenotypes, 9-week-old NOD mice were treated i.m. once weekly with 25 µg, 50 µg, 100 µg, or 200 µg pVAX1-PPI/TGF-β/IL-10/IL-2; all groups were significantly and comparably protected from type 1 diabetes development versus untreated controls (Supplementary Fig. 1). At 13 weeks of age, flow cytometry was performed on peripheral blood mononuclear cells from 12 randomly selected mice per group. In comparisons of untreated mice versus mice treated with the plasmid, no statistically significant difference was found regarding innate immune cell frequencies (Supplementary Fig. 2 and Supplementary Tables 1 and 2). However, after treatment with pVAX1-PPI/TGF-β/IL-10/IL-2, CD210, the IL-10 receptor, was dose-dependently upregulated on CD11b⁺ myeloid cells (Fig. 4A), myeloid dendritic cells (Fig. 4B), monocytes, Ly6C⁻ monocytes, and Ly6C⁺ classical monocytes (Supplementary Table 1). Other cell types either showed no (e.g., NK cells) or little (eosinophils and neutrophils) difference in CD210 expression across treatment groups. Additionally, while no increase was detected for circulating IL-2 and plasmid-derived mature TGF-β1, there was a dose-dependent increase in IL-10 and TNF-α (up to the 200-µg dose) and in IFN-γ and IL-6 (up to the 100-µg dose) (Fig. 4C and not shown).

Post hoc analyses from the Diabetes Prevention Trial–Type 1 (DPT-1) with use of oral insulin demonstrated a potential clinical benefit in individuals with IAA ≥80 nU/mL (29). Preclinical data in NOD mice have also shown that preexisting IAA could predict the efficacy of oral insulin in curing type 1 diabetes in combination with anti-CD3 (30). Thus, we sought to determine whether the efficacy of the plasmid therapy was dependent on the baseline IAA status of the mice. Mice were bled at 9 weeks of age before receiving their first plasmid dose, and plasma was analyzed for IAA. Though IAA detection at 9 weeks of age in NOD mice is relatively infrequent (20), the number of IAA⁺ mice in our study was 12 of 70 across all treatment groups, comprising 4 of 24 untreated and 8 of 46 plasmid-treated mice. To gain insights into whether IAA would impact responsiveness to plasmid therapy, we pooled all mice from the 25-µg, 50-µg, 100-µg, and 200-µg pVAX1-PPI/TGF-β/IL-10/IL-2 plasmid-treated groups based on their IAA status (8 IAA⁺, 38 IAA⁻) and plotted versus the diabetes outcome at 30 weeks of age. We found that among untreated mice, the proportion of mice developing type 1 diabetes was statistically significantly greater among those that were IAA⁺ (100% by 18 weeks of age) compared with those that were IAA⁻ (30% by 18 weeks of age and 75% at end point). In contrast, the proportion of plasmid-treated mice developing type 1 diabetes was similar in IAA⁻ (23.7%) and IAA⁺ (25%) animals (Fig. 5). The conclusion was similar based on pooled groups of mice treated subcutaneously (s.c.) with plasmid once and thrice weekly (data not shown). These data seem to indicate that the efficacy of plasmid-mediated type 1 diabetes prevention is not altered by presence of insulin autoantibodies at treatment initiation.

For determination of whether antigen was required for maximal prevention, NOD mice were treated weekly, beginning at 11 weeks of age, with i.m. injections of a plasmid encoding only cytokines and monitored for disease development. Mice treated with pVAX1-PPI/TGF-β/IL-10/IL-2 were almost completely protected. In comparison, the diabetes incidence was statistically significantly greater in mice treated with pVAX1-TGF-β/IL-10/IL-2 encoding immune modifiers but no antigen (Fig. 6A). This demonstrated that the in vivo efficacy associated with the multicytokine plasmid is driven, in part, by the coexpression of an islet antigen.

Next, for determination of the effect of plasmid therapy on the recall response to an irrelevant antigen, the antigenic response to OVA in NOD mice was evaluated with regard to whether our plasmid therapy induced broad immune suppression. NOD mice were treated with pVAX1-PPI/TGF-β/IL-10/IL-2 or left untreated for 5 weeks prior to immunization with OVA in alum (Fig. 6B). Serum anti-OVA IgG antibodies were measured 30 days postchallenge. We did not observe any statistically significant differences between plasmid-treated and untreated groups, suggesting that plasmid therapy did not induce general immune suppression in vivo (Fig. 6C). Mice that were OVA challenged and either left untreated or continued being dosed with pVAX1-PPI/TGF-β/IL-10/IL-2 were followed for type 1 diabetes development until 21 weeks of age. Upon OVA challenge, full protection was seen in plasmid-dosed mice compared with untreated counterparts, suggesting that plasmid efficacy is not hindered by concomitant immunization with an irrelevant antigen (not shown).

Though i.m. is a common route of administration for plasmid therapy, the ability to dose through s.c. injection likely has a more favorable path to clinical translation due to the discomfort associated with i.m. injections. Therefore, we next compared s.c. injection to i.m. and evaluated the dosing frequency. The minimal dosing regimen required for maximal efficacy with i.m. administration was once weekly (Fig. 7A). However, our plasmid is adjuvant free, and we were uncertain about whether transfection efficiency of the plasmid would be sufficient to mount an effective tolerogenic response with injection once weekly via s.c. (31). To compensate for potential loss of transfection efficacy using the s.c. injection method, we evaluated both once weekly and thrice weekly administration schedules beginning at 9 weeks of age. In an earlier study, NOD mice injected s.c. once weekly beginning at 6 weeks of age displayed a diabetes incidence that was statistically significantly higher than in the group dosed once per week i.m. (14% vs. 0%, respectively; P < 0.05) (Fig. 2D). Subsequently, we observed that treatment thrice weekly with pVAX1-PPI/TGF-β/IL-10/IL-2 completely prevented type 1 diabetes (0%) (Fig. 7B); in comparison, with once-weekly injections, statistically significantly more mice (∼29%) developed diabetes (P < 0.01).

Human chromosomal integration of the antibiotic resistance genes, gene transfer to other microorganisms, and downstream production complications are concerns in development of plasmid therapy for clinical use (32). Therefore, to facilitate clinical translation of the multicytokine plasmid, we developed a novel selection system using the temperature selection properties of L. monocytogenes. Specifically, prfA, a thermoregulation virulence regulator for L. monocytogenes that selectively promotes the growth of bacteria at temperatures >30°C (33), was inserted upstream of the IF-1 gene in E. coli to control the expression of the essential gene (34,35). Additionally, we substituted the kanamycin antibiotic resistance gene (nptII) in the pVAX1 vector with InfA, creating a new plasmid backbone (Fig. 8A and B). The engineered infA::prfA E. coli produced antibiotic resistance–free plasmid at both 30°C and 37°C. The antibiotic-free tolerogenic plasmid was then designated pNN-PPI/TGF-β/IL-10/IL-2. Treatment of NOD mice with the new plasmid using once weekly and thrice weekly dosing regimens demonstrated efficacy comparable with that of the pVAX1-based plasmid. (Fig. 8C).

Plasmid-based vaccination and immunotherapy to stimulate or silence antigenic responses, respectively, have advanced into clinic trials (16,36). While safety has consistently been excellent, neither potent immunogenicity nor antigenic tolerization has been convincingly demonstrated. We therefore sought to improve upon previous work with tolerizing antigenic plasmid therapy by generating a novel multicomponent plasmid that encoded both anti-inflammatory cytokines and a disease-relevant autoantigen to promote a tolerance-inducing microenvironment around the expressing cells (17,37). Using this plasmid, we demonstrated that such coexpression of immune modulators markedly improves efficacy as measured by the prevention of hyperglycemia in the NOD mouse model.

Our results demonstrate that a single-autoantigen-encoding plasmid was capable of preventing diabetes almost completely in a mouse model where β-cell destruction is predominantly driven by reactivity to insulin (38), as is also likely the case in many humans with type 1 diabetes (39–41). Additionally, the IAA status at baseline did not influence the therapeutic efficacy of the plasmid at end point. This suggests that a clinical trial testing such a plasmid in people at risk for developing type 1 diabetes might not require participant stratification or exclusion based on IAA status. There was no evidence of broad immune suppression, and we detected little to no systemic levels of plasmid-derived cytokines, supporting the hypothesis that plasmid-encoded cytokines affect the immune system locally to suppress the response to PPI rather than promote global suppression. Furthermore, when one of the components was omitted or its expression level altered, efficacy was reduced, indicating that our extensive compound screening has yielded a unique combination of plasmid backbone elements, antigen, and cytokines that, together, results in potent disease suppression.

A strength in our approach was the stringent in vivo screening procedure by well-powered NOD studies, necessitated by our intention to identify a plasmid candidate for clinical use. It is generally acknowledged that the timing of diabetes development or overall incidence and consistency of results in, and between, NOD studies can differ significantly between mice from different vendors and between laboratories (42). To address this, we conducted our prevention studies at multiple sites, using NOD mouse colonies not only from The Jackson Laboratory but also from Taconic Biosciences (data not shown). This provided assurance that the observed efficacy was robust, as results for both plasmids (pVAX1 and pNN) were similar across sites and colonies.

The in vivo mechanism of action of naked plasmid immunization for antigenic tolerance induction remains ill-defined (43). It has been suggested that ∼10–30% of cells located near the injection site express the encoded protein but that only ∼1–2% of cells in the nearby muscle (i.e., quadriceps) are transfected (44). The use of adjuvants and/or injection equipment such as gene guns or electroporation may increase transfection efficiency considerably, but it is uncertain what the ideal transfection efficiency should be to induce an adequate tolerogenic response. A small fraction of transfected cells are likely antigen-presenting cells that migrate to draining lymph nodes (45), a hypothesis that was recently elegantly confirmed with use of fluorescent plasmid tracers upon intradermal administration (46). These cells could produce regulatory T cell–inducing cytokines and present antigen to circulating CD4⁺ or CD8⁺ T cells, potentially tolerizing them to the encoded antigen, preventing autoimmunity. These transfected cells could also explain the increased CD210 expression on circulating myeloid dendritic cells. Increasing the understanding of the mechanism of action is further complicated by use of s.c. over i.m. administration; s.c. injection likely results in transfection of different cell populations or reduced transfection efficiency, explaining the need for a higher injection frequency with s.c. to mimic the efficacy of once weekly i.m. injection. Studies are required to fully understand the transfection process and to discover the cell populations that are modulated post–plasmid administration.

We unexpectedly observed increased plasma or serum IFN-γ and TNF-α after the first i.m. injection (4 and 8 weeks, respectively). While it remains unclear how the plasmid promotes expression of these proinflammatory cytokines, there is conflicting information as to whether these cytokines may have a protective or harmful effect by limiting autoreactivity or stimulating pathogenic CD8⁺ cells (47–50). It is possible that plasmid therapy directly promotes expression of IFN-γ and TNF-α to limit proliferation and activation of pathogenic CD8⁺ cells, but whether this is an underlying mechanism that facilitates tolerance induction remains unclear. Future studies, including the ongoing phase 1 trial with the pNN plasmid (TOPPLE T1D [clinical trial reg. no. NCT04279613, ClinicalTrials.gov]), are needed to clarify the relevance of these unexpected findings via longitudinal measurements of pro- and anti-inflammatory cytokines and chemokines. This would involve investigations to clarify the relevance of the findings to the human setting. In this context, recent evidence from a trial of anti–TNF-α treatment with golimumab, which appeared to improve β-cell function in people with new-onset type 1 diabetes, indicates a pathogenic role of this proinflammatory cytokine in human type 1 diabetes (51). The current study describes the molecular design, superior preclinical efficacy, and translational biomarker profile of an advanced plasmid-based immunotherapy. This therapy was tailored for chronic use in humans for the prevention of type 1 diabetes in individuals at risk, with IAA positivity, by virtue of its s.c. route, antibiotic resistance gene–free, and nonimmunosuppressant properties. The approach has cleared formal toxicology studies and has been produced at quality and scale for the ongoing, multiple ascending doses trial to investigate its safety, tolerability, and pharmacokinetic profile (TOPPLE T1D). In summary, the results demonstrate that this unique plasmid-based combination immunotherapy has potential to provide a path to prevention of type 1 diabetes.

Acknowledgments. The authors thank Frederik Flindt Kreiner, Novo Nordisk A/S, Søborg, Denmark, for medical writing and submission support; David Rodriguez and staff, Novo Nordisk Research Center Seattle, Inc. [NNRCSI], for vivarium management and animal care; Claire Gibson Bamman, Tamar Boursalian, and Jose Luis Vela, NNRCSI, for scientific discussions; and David Purdy and Bong Kim, NNRCSI, for conducting mass spectrometry experiments.

Duality of Interest. The study was funded by Novo Nordisk A/S. All authors are or have been employees of Novo Nordisk A/S. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. P.P.P., J.Ch., J.D.W., K.C., and M.v.H. generated the idea for the study. P.P.P., J.Ch., M.W., J.D.W., and V.K. planned the experiments. J.Ch., M.W., J.G., J.Cr., C.O., N.P, S.L., S.S.R., A.R., A.M., and C.-l.C. conducted the experiments. P.P.P., J.Ch., M.W., J.D.W., N.P., V.K., and S.L. analyzed the collected data. P.P.P., J.D.W., K.C., and M.v.H. wrote and reviewed the manuscript. M.v.H. 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.